PageRenderTime 23ms CodeModel.GetById 3ms app.highlight 114ms RepoModel.GetById 1ms app.codeStats 1ms

/net/socket.c

https://bitbucket.org/teamblackout/vivo-blackout-edition
C | 3404 lines | 2442 code | 490 blank | 472 comment | 386 complexity | 0d0a4e9afa5e2a1f522d20c2d495f48e 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(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
 512int add_or_remove_port(struct sock *sk, int add_or_remove);	/* SSD_RIL: Garbage_Filter_TCP */
 513
 514void sock_release(struct socket *sock)
 515{
 516	/* ++SSD_RIL: Garbage_Filter_TCP */
 517	if (sock->sk != NULL)
 518		add_or_remove_port(sock->sk, 0);
 519	/* --SSD_RIL: Garbage_Filter_TCP */
 520
 521	if (sock->ops) {
 522		struct module *owner = sock->ops->owner;
 523
 524		sock->ops->release(sock);
 525		sock->ops = NULL;
 526		module_put(owner);
 527	}
 528
 529	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 530		printk(KERN_ERR "sock_release: fasync list not empty!\n");
 531
 532	percpu_sub(sockets_in_use, 1);
 533	if (!sock->file) {
 534		iput(SOCK_INODE(sock));
 535		return;
 536	}
 537	sock->file = NULL;
 538}
 539EXPORT_SYMBOL(sock_release);
 540
 541int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 542{
 543	*tx_flags = 0;
 544	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 545		*tx_flags |= SKBTX_HW_TSTAMP;
 546	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 547		*tx_flags |= SKBTX_SW_TSTAMP;
 548	return 0;
 549}
 550EXPORT_SYMBOL(sock_tx_timestamp);
 551
 552static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 553				       struct msghdr *msg, size_t size)
 554{
 555	struct sock_iocb *si = kiocb_to_siocb(iocb);
 556
 557	sock_update_classid(sock->sk);
 558
 559	si->sock = sock;
 560	si->scm = NULL;
 561	si->msg = msg;
 562	si->size = size;
 563
 564	return sock->ops->sendmsg(iocb, sock, msg, size);
 565}
 566
 567static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 568				 struct msghdr *msg, size_t size)
 569{
 570	int err = security_socket_sendmsg(sock, msg, size);
 571
 572	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 573}
 574
 575int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 576{
 577	struct kiocb iocb;
 578	struct sock_iocb siocb;
 579	int ret;
 580
 581	init_sync_kiocb(&iocb, NULL);
 582	iocb.private = &siocb;
 583	ret = __sock_sendmsg(&iocb, sock, msg, size);
 584	if (-EIOCBQUEUED == ret)
 585		ret = wait_on_sync_kiocb(&iocb);
 586	return ret;
 587}
 588EXPORT_SYMBOL(sock_sendmsg);
 589
 590int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 591{
 592	struct kiocb iocb;
 593	struct sock_iocb siocb;
 594	int ret;
 595
 596	init_sync_kiocb(&iocb, NULL);
 597	iocb.private = &siocb;
 598	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 599	if (-EIOCBQUEUED == ret)
 600		ret = wait_on_sync_kiocb(&iocb);
 601	return ret;
 602}
 603
 604int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 605		   struct kvec *vec, size_t num, size_t size)
 606{
 607	mm_segment_t oldfs = get_fs();
 608	int result;
 609
 610	set_fs(KERNEL_DS);
 611	/*
 612	 * the following is safe, since for compiler definitions of kvec and
 613	 * iovec are identical, yielding the same in-core layout and alignment
 614	 */
 615	msg->msg_iov = (struct iovec *)vec;
 616	msg->msg_iovlen = num;
 617	result = sock_sendmsg(sock, msg, size);
 618	set_fs(oldfs);
 619	return result;
 620}
 621EXPORT_SYMBOL(kernel_sendmsg);
 622
 623static int ktime2ts(ktime_t kt, struct timespec *ts)
 624{
 625	if (kt.tv64) {
 626		*ts = ktime_to_timespec(kt);
 627		return 1;
 628	} else {
 629		return 0;
 630	}
 631}
 632
 633/*
 634 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 635 */
 636void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 637	struct sk_buff *skb)
 638{
 639	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 640	struct timespec ts[3];
 641	int empty = 1;
 642	struct skb_shared_hwtstamps *shhwtstamps =
 643		skb_hwtstamps(skb);
 644
 645	/* Race occurred between timestamp enabling and packet
 646	   receiving.  Fill in the current time for now. */
 647	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 648		__net_timestamp(skb);
 649
 650	if (need_software_tstamp) {
 651		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 652			struct timeval tv;
 653			skb_get_timestamp(skb, &tv);
 654			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 655				 sizeof(tv), &tv);
 656		} else {
 657			skb_get_timestampns(skb, &ts[0]);
 658			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 659				 sizeof(ts[0]), &ts[0]);
 660		}
 661	}
 662
 663
 664	memset(ts, 0, sizeof(ts));
 665	if (skb->tstamp.tv64 &&
 666	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 667		skb_get_timestampns(skb, ts + 0);
 668		empty = 0;
 669	}
 670	if (shhwtstamps) {
 671		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 672		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 673			empty = 0;
 674		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 675		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 676			empty = 0;
 677	}
 678	if (!empty)
 679		put_cmsg(msg, SOL_SOCKET,
 680			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 681}
 682EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 683
 684static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 685				   struct sk_buff *skb)
 686{
 687	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 688		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 689			sizeof(__u32), &skb->dropcount);
 690}
 691
 692void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 693	struct sk_buff *skb)
 694{
 695	sock_recv_timestamp(msg, sk, skb);
 696	sock_recv_drops(msg, sk, skb);
 697}
 698EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 699
 700static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 701				       struct msghdr *msg, size_t size, int flags)
 702{
 703	struct sock_iocb *si = kiocb_to_siocb(iocb);
 704
 705	sock_update_classid(sock->sk);
 706
 707	si->sock = sock;
 708	si->scm = NULL;
 709	si->msg = msg;
 710	si->size = size;
 711	si->flags = flags;
 712
 713	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 714}
 715
 716static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 717				 struct msghdr *msg, size_t size, int flags)
 718{
 719	int err = security_socket_recvmsg(sock, msg, size, flags);
 720
 721	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 722}
 723
 724int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 725		 size_t size, int flags)
 726{
 727	struct kiocb iocb;
 728	struct sock_iocb siocb;
 729	int ret;
 730
 731	init_sync_kiocb(&iocb, NULL);
 732	iocb.private = &siocb;
 733	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 734	if (-EIOCBQUEUED == ret)
 735		ret = wait_on_sync_kiocb(&iocb);
 736	return ret;
 737}
 738EXPORT_SYMBOL(sock_recvmsg);
 739
 740static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 741			      size_t size, int flags)
 742{
 743	struct kiocb iocb;
 744	struct sock_iocb siocb;
 745	int ret;
 746
 747	init_sync_kiocb(&iocb, NULL);
 748	iocb.private = &siocb;
 749	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 750	if (-EIOCBQUEUED == ret)
 751		ret = wait_on_sync_kiocb(&iocb);
 752	return ret;
 753}
 754
 755/**
 756 * kernel_recvmsg - Receive a message from a socket (kernel space)
 757 * @sock:       The socket to receive the message from
 758 * @msg:        Received message
 759 * @vec:        Input s/g array for message data
 760 * @num:        Size of input s/g array
 761 * @size:       Number of bytes to read
 762 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 763 *
 764 * On return the msg structure contains the scatter/gather array passed in the
 765 * vec argument. The array is modified so that it consists of the unfilled
 766 * portion of the original array.
 767 *
 768 * The returned value is the total number of bytes received, or an error.
 769 */
 770int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 771		   struct kvec *vec, size_t num, size_t size, int flags)
 772{
 773	mm_segment_t oldfs = get_fs();
 774	int result;
 775
 776	set_fs(KERNEL_DS);
 777	/*
 778	 * the following is safe, since for compiler definitions of kvec and
 779	 * iovec are identical, yielding the same in-core layout and alignment
 780	 */
 781	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 782	result = sock_recvmsg(sock, msg, size, flags);
 783	set_fs(oldfs);
 784	return result;
 785}
 786EXPORT_SYMBOL(kernel_recvmsg);
 787
 788static void sock_aio_dtor(struct kiocb *iocb)
 789{
 790	kfree(iocb->private);
 791}
 792
 793static ssize_t sock_sendpage(struct file *file, struct page *page,
 794			     int offset, size_t size, loff_t *ppos, int more)
 795{
 796	struct socket *sock;
 797	int flags;
 798
 799	sock = file->private_data;
 800
 801	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 802	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 803	flags |= more;
 804
 805	return kernel_sendpage(sock, page, offset, size, flags);
 806}
 807
 808static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 809				struct pipe_inode_info *pipe, size_t len,
 810				unsigned int flags)
 811{
 812	struct socket *sock = file->private_data;
 813
 814	if (unlikely(!sock->ops->splice_read))
 815		return -EINVAL;
 816
 817	sock_update_classid(sock->sk);
 818
 819	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 820}
 821
 822static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 823					 struct sock_iocb *siocb)
 824{
 825	if (!is_sync_kiocb(iocb)) {
 826		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 827		if (!siocb)
 828			return NULL;
 829		iocb->ki_dtor = sock_aio_dtor;
 830	}
 831
 832	siocb->kiocb = iocb;
 833	iocb->private = siocb;
 834	return siocb;
 835}
 836
 837static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 838		struct file *file, const struct iovec *iov,
 839		unsigned long nr_segs)
 840{
 841	struct socket *sock = file->private_data;
 842	size_t size = 0;
 843	int i;
 844
 845	for (i = 0; i < nr_segs; i++)
 846		size += iov[i].iov_len;
 847
 848	msg->msg_name = NULL;
 849	msg->msg_namelen = 0;
 850	msg->msg_control = NULL;
 851	msg->msg_controllen = 0;
 852	msg->msg_iov = (struct iovec *)iov;
 853	msg->msg_iovlen = nr_segs;
 854	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 855
 856	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 857}
 858
 859static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 860				unsigned long nr_segs, loff_t pos)
 861{
 862	struct sock_iocb siocb, *x;
 863
 864	if (pos != 0)
 865		return -ESPIPE;
 866
 867	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 868		return 0;
 869
 870
 871	x = alloc_sock_iocb(iocb, &siocb);
 872	if (!x)
 873		return -ENOMEM;
 874	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 875}
 876
 877static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 878			struct file *file, const struct iovec *iov,
 879			unsigned long nr_segs)
 880{
 881	struct socket *sock = file->private_data;
 882	size_t size = 0;
 883	int i;
 884
 885	for (i = 0; i < nr_segs; i++)
 886		size += iov[i].iov_len;
 887
 888	msg->msg_name = NULL;
 889	msg->msg_namelen = 0;
 890	msg->msg_control = NULL;
 891	msg->msg_controllen = 0;
 892	msg->msg_iov = (struct iovec *)iov;
 893	msg->msg_iovlen = nr_segs;
 894	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 895	if (sock->type == SOCK_SEQPACKET)
 896		msg->msg_flags |= MSG_EOR;
 897
 898	return __sock_sendmsg(iocb, sock, msg, size);
 899}
 900
 901static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 902			  unsigned long nr_segs, loff_t pos)
 903{
 904	struct sock_iocb siocb, *x;
 905
 906	if (pos != 0)
 907		return -ESPIPE;
 908
 909	x = alloc_sock_iocb(iocb, &siocb);
 910	if (!x)
 911		return -ENOMEM;
 912
 913	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 914}
 915
 916/*
 917 * Atomic setting of ioctl hooks to avoid race
 918 * with module unload.
 919 */
 920
 921static DEFINE_MUTEX(br_ioctl_mutex);
 922static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 923
 924void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 925{
 926	mutex_lock(&br_ioctl_mutex);
 927	br_ioctl_hook = hook;
 928	mutex_unlock(&br_ioctl_mutex);
 929}
 930EXPORT_SYMBOL(brioctl_set);
 931
 932static DEFINE_MUTEX(vlan_ioctl_mutex);
 933static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 934
 935void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 936{
 937	mutex_lock(&vlan_ioctl_mutex);
 938	vlan_ioctl_hook = hook;
 939	mutex_unlock(&vlan_ioctl_mutex);
 940}
 941EXPORT_SYMBOL(vlan_ioctl_set);
 942
 943static DEFINE_MUTEX(dlci_ioctl_mutex);
 944static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 945
 946void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 947{
 948	mutex_lock(&dlci_ioctl_mutex);
 949	dlci_ioctl_hook = hook;
 950	mutex_unlock(&dlci_ioctl_mutex);
 951}
 952EXPORT_SYMBOL(dlci_ioctl_set);
 953
 954static long sock_do_ioctl(struct net *net, struct socket *sock,
 955				 unsigned int cmd, unsigned long arg)
 956{
 957	int err;
 958	void __user *argp = (void __user *)arg;
 959
 960	err = sock->ops->ioctl(sock, cmd, arg);
 961
 962	/*
 963	 * If this ioctl is unknown try to hand it down
 964	 * to the NIC driver.
 965	 */
 966	if (err == -ENOIOCTLCMD)
 967		err = dev_ioctl(net, cmd, argp);
 968
 969	return err;
 970}
 971
 972/*
 973 *	With an ioctl, arg may well be a user mode pointer, but we don't know
 974 *	what to do with it - that's up to the protocol still.
 975 */
 976
 977static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 978{
 979	struct socket *sock;
 980	struct sock *sk;
 981	void __user *argp = (void __user *)arg;
 982	int pid, err;
 983	struct net *net;
 984
 985	sock = file->private_data;
 986	sk = sock->sk;
 987	net = sock_net(sk);
 988	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
 989		err = dev_ioctl(net, cmd, argp);
 990	} else
 991#ifdef CONFIG_WEXT_CORE
 992	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
 993		err = dev_ioctl(net, cmd, argp);
 994	} else
 995#endif
 996		switch (cmd) {
 997		case FIOSETOWN:
 998		case SIOCSPGRP:
 999			err = -EFAULT;
1000			if (get_user(pid, (int __user *)argp))
1001				break;
1002			err = f_setown(sock->file, pid, 1);
1003			break;
1004		case FIOGETOWN:
1005		case SIOCGPGRP:
1006			err = put_user(f_getown(sock->file),
1007				       (int __user *)argp);
1008			break;
1009		case SIOCGIFBR:
1010		case SIOCSIFBR:
1011		case SIOCBRADDBR:
1012		case SIOCBRDELBR:
1013			err = -ENOPKG;
1014			if (!br_ioctl_hook)
1015				request_module("bridge");
1016
1017			mutex_lock(&br_ioctl_mutex);
1018			if (br_ioctl_hook)
1019				err = br_ioctl_hook(net, cmd, argp);
1020			mutex_unlock(&br_ioctl_mutex);
1021			break;
1022		case SIOCGIFVLAN:
1023		case SIOCSIFVLAN:
1024			err = -ENOPKG;
1025			if (!vlan_ioctl_hook)
1026				request_module("8021q");
1027
1028			mutex_lock(&vlan_ioctl_mutex);
1029			if (vlan_ioctl_hook)
1030				err = vlan_ioctl_hook(net, argp);
1031			mutex_unlock(&vlan_ioctl_mutex);
1032			break;
1033		case SIOCADDDLCI:
1034		case SIOCDELDLCI:
1035			err = -ENOPKG;
1036			if (!dlci_ioctl_hook)
1037				request_module("dlci");
1038
1039			mutex_lock(&dlci_ioctl_mutex);
1040			if (dlci_ioctl_hook)
1041				err = dlci_ioctl_hook(cmd, argp);
1042			mutex_unlock(&dlci_ioctl_mutex);
1043			break;
1044		default:
1045			err = sock_do_ioctl(net, sock, cmd, arg);
1046			break;
1047		}
1048	return err;
1049}
1050
1051int sock_create_lite(int family, int type, int protocol, struct socket **res)
1052{
1053	int err;
1054	struct socket *sock = NULL;
1055
1056	err = security_socket_create(family, type, protocol, 1);
1057	if (err)
1058		goto out;
1059
1060	sock = sock_alloc();
1061	if (!sock) {
1062		err = -ENOMEM;
1063		goto out;
1064	}
1065
1066	sock->type = type;
1067	err = security_socket_post_create(sock, family, type, protocol, 1);
1068	if (err)
1069		goto out_release;
1070
1071out:
1072	*res = sock;
1073	return err;
1074out_release:
1075	sock_release(sock);
1076	sock = NULL;
1077	goto out;
1078}
1079EXPORT_SYMBOL(sock_create_lite);
1080
1081/* No kernel lock held - perfect */
1082static unsigned int sock_poll(struct file *file, poll_table *wait)
1083{
1084	struct socket *sock;
1085
1086	/*
1087	 *      We can't return errors to poll, so it's either yes or no.
1088	 */
1089	sock = file->private_data;
1090	return sock->ops->poll(file, sock, wait);
1091}
1092
1093static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1094{
1095	struct socket *sock = file->private_data;
1096
1097	return sock->ops->mmap(file, sock, vma);
1098}
1099
1100static int sock_close(struct inode *inode, struct file *filp)
1101{
1102	/*
1103	 *      It was possible the inode is NULL we were
1104	 *      closing an unfinished socket.
1105	 */
1106
1107	if (!inode) {
1108		printk(KERN_DEBUG "sock_close: NULL inode\n");
1109		return 0;
1110	}
1111	sock_release(SOCKET_I(inode));
1112	return 0;
1113}
1114
1115/*
1116 *	Update the socket async list
1117 *
1118 *	Fasync_list locking strategy.
1119 *
1120 *	1. fasync_list is modified only under process context socket lock
1121 *	   i.e. under semaphore.
1122 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1123 *	   or under socket lock
1124 */
1125
1126static int sock_fasync(int fd, struct file *filp, int on)
1127{
1128	struct socket *sock = filp->private_data;
1129	struct sock *sk = sock->sk;
1130	struct socket_wq *wq;
1131
1132	if (sk == NULL)
1133		return -EINVAL;
1134
1135	lock_sock(sk);
1136	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1137	fasync_helper(fd, filp, on, &wq->fasync_list);
1138
1139	if (!wq->fasync_list)
1140		sock_reset_flag(sk, SOCK_FASYNC);
1141	else
1142		sock_set_flag(sk, SOCK_FASYNC);
1143
1144	release_sock(sk);
1145	return 0;
1146}
1147
1148/* This function may be called only under socket lock or callback_lock or rcu_lock */
1149
1150int sock_wake_async(struct socket *sock, int how, int band)
1151{
1152	struct socket_wq *wq;
1153
1154	if (!sock)
1155		return -1;
1156	rcu_read_lock();
1157	wq = rcu_dereference(sock->wq);
1158	if (!wq || !wq->fasync_list) {
1159		rcu_read_unlock();
1160		return -1;
1161	}
1162	switch (how) {
1163	case SOCK_WAKE_WAITD:
1164		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1165			break;
1166		goto call_kill;
1167	case SOCK_WAKE_SPACE:
1168		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1169			break;
1170		/* fall through */
1171	case SOCK_WAKE_IO:
1172call_kill:
1173		kill_fasync(&wq->fasync_list, SIGIO, band);
1174		break;
1175	case SOCK_WAKE_URG:
1176		kill_fasync(&wq->fasync_list, SIGURG, band);
1177	}
1178	rcu_read_unlock();
1179	return 0;
1180}
1181EXPORT_SYMBOL(sock_wake_async);
1182
1183int __sock_create(struct net *net, int family, int type, int protocol,
1184			 struct socket **res, int kern)
1185{
1186	int err;
1187	struct socket *sock;
1188	const struct net_proto_family *pf;
1189
1190	/*
1191	 *      Check protocol is in range
1192	 */
1193	if (family < 0 || family >= NPROTO)
1194		return -EAFNOSUPPORT;
1195	if (type < 0 || type >= SOCK_MAX)
1196		return -EINVAL;
1197
1198	/* Compatibility.
1199
1200	   This uglymoron is moved from INET layer to here to avoid
1201	   deadlock in module load.
1202	 */
1203	if (family == PF_INET && type == SOCK_PACKET) {
1204		static int warned;
1205		if (!warned) {
1206			warned = 1;
1207			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1208			       current->comm);
1209		}
1210		family = PF_PACKET;
1211	}
1212
1213	err = security_socket_create(family, type, protocol, kern);
1214	if (err)
1215		return err;
1216
1217	/*
1218	 *	Allocate the socket and allow the family to set things up. if
1219	 *	the protocol is 0, the family is instructed to select an appropriate
1220	 *	default.
1221	 */
1222	sock = sock_alloc();
1223	if (!sock) {
1224		if (net_ratelimit())
1225			printk(KERN_WARNING "socket: no more sockets\n");
1226		return -ENFILE;	/* Not exactly a match, but its the
1227				   closest posix thing */
1228	}
1229
1230	sock->type = type;
1231
1232#ifdef CONFIG_MODULES
1233	/* Attempt to load a protocol module if the find failed.
1234	 *
1235	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1236	 * requested real, full-featured networking support upon configuration.
1237	 * Otherwise module support will break!
1238	 */
1239	if (rcu_access_pointer(net_families[family]) == NULL)
1240		request_module("net-pf-%d", family);
1241#endif
1242
1243	rcu_read_lock();
1244	pf = rcu_dereference(net_families[family]);
1245	err = -EAFNOSUPPORT;
1246	if (!pf)
1247		goto out_release;
1248
1249	/*
1250	 * We will call the ->create function, that possibly is in a loadable
1251	 * module, so we have to bump that loadable module refcnt first.
1252	 */
1253	if (!try_module_get(pf->owner))
1254		goto out_release;
1255
1256	/* Now protected by module ref count */
1257	rcu_read_unlock();
1258
1259	err = pf->create(net, sock, protocol, kern);
1260	if (err < 0)
1261		goto out_module_put;
1262
1263	/*
1264	 * Now to bump the refcnt of the [loadable] module that owns this
1265	 * socket at sock_release time we decrement its refcnt.
1266	 */
1267	if (!try_module_get(sock->ops->owner))
1268		goto out_module_busy;
1269
1270	/*
1271	 * Now that we're done with the ->create function, the [loadable]
1272	 * module can have its refcnt decremented
1273	 */
1274	module_put(pf->owner);
1275	err = security_socket_post_create(sock, family, type, protocol, kern);
1276	if (err)
1277		goto out_sock_release;
1278	*res = sock;
1279
1280	return 0;
1281
1282out_module_busy:
1283	err = -EAFNOSUPPORT;
1284out_module_put:
1285	sock->ops = NULL;
1286	module_put(pf->owner);
1287out_sock_release:
1288	sock_release(sock);
1289	return err;
1290
1291out_release:
1292	rcu_read_unlock();
1293	goto out_sock_release;
1294}
1295EXPORT_SYMBOL(__sock_create);
1296
1297int sock_create(int family, int type, int protocol, struct socket **res)
1298{
1299	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1300}
1301EXPORT_SYMBOL(sock_create);
1302
1303int sock_create_kern(int family, int type, int protocol, struct socket **res)
1304{
1305	return __sock_create(&init_net, family, type, protocol, res, 1);
1306}
1307EXPORT_SYMBOL(sock_create_kern);
1308
1309SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1310{
1311	int retval;
1312	struct socket *sock;
1313	int flags;
1314
1315	/* Check the SOCK_* constants for consistency.  */
1316	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1317	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1318	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1319	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1320
1321	flags = type & ~SOCK_TYPE_MASK;
1322	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1323		return -EINVAL;
1324	type &= SOCK_TYPE_MASK;
1325
1326	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1327		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1328
1329	retval = sock_create(family, type, protocol, &sock);
1330	if (retval < 0)
1331		goto out;
1332
1333	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1334	if (retval < 0)
1335		goto out_release;
1336
1337out:
1338	/* It may be already another descriptor 8) Not kernel problem. */
1339	return retval;
1340
1341out_release:
1342	sock_release(sock);
1343	return retval;
1344}
1345
1346/*
1347 *	Create a pair of connected sockets.
1348 */
1349
1350SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1351		int __user *, usockvec)
1352{
1353	struct socket *sock1, *sock2;
1354	int fd1, fd2, err;
1355	struct file *newfile1, *newfile2;
1356	int flags;
1357
1358	flags = type & ~SOCK_TYPE_MASK;
1359	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1360		return -EINVAL;
1361	type &= SOCK_TYPE_MASK;
1362
1363	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1364		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1365
1366	/*
1367	 * Obtain the first socket and check if the underlying protocol
1368	 * supports the socketpair call.
1369	 */
1370
1371	err = sock_create(family, type, protocol, &sock1);
1372	if (err < 0)
1373		goto out;
1374
1375	err = sock_create(family, type, protocol, &sock2);
1376	if (err < 0)
1377		goto out_release_1;
1378
1379	err = sock1->ops->socketpair(sock1, sock2);
1380	if (err < 0)
1381		goto out_release_both;
1382
1383	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1384	if (unlikely(fd1 < 0)) {
1385		err = fd1;
1386		goto out_release_both;
1387	}
1388
1389	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1390	if (unlikely(fd2 < 0)) {
1391		err = fd2;
1392		fput(newfile1);
1393		put_unused_fd(fd1);
1394		sock_release(sock2);
1395		goto out;
1396	}
1397
1398	audit_fd_pair(fd1, fd2);
1399	fd_install(fd1, newfile1);
1400	fd_install(fd2, newfile2);
1401	/* fd1 and fd2 may be already another descriptors.
1402	 * Not kernel problem.
1403	 */
1404
1405	err = put_user(fd1, &usockvec[0]);
1406	if (!err)
1407		err = put_user(fd2, &usockvec[1]);
1408	if (!err)
1409		return 0;
1410
1411	sys_close(fd2);
1412	sys_close(fd1);
1413	return err;
1414
1415out_release_both:
1416	sock_release(sock2);
1417out_release_1:
1418	sock_release(sock1);
1419out:
1420	return err;
1421}
1422
1423/*
1424 *	Bind a name to a socket. Nothing much to do here since it's
1425 *	the protocol's responsibility to handle the local address.
1426 *
1427 *	We move the socket address to kernel space before we call
1428 *	the protocol layer (having also checked the address is ok).
1429 */
1430
1431SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1432{
1433	struct socket *sock;
1434	struct sockaddr_storage address;
1435	int err, fput_needed;
1436
1437	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1438	if (sock) {
1439		err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1440		if (err >= 0) {
1441			err = security_socket_bind(sock,
1442						   (struct sockaddr *)&address,
1443						   addrlen);
1444			if (!err)
1445				err = sock->ops->bind(sock,
1446						      (struct sockaddr *)
1447						      &address, addrlen);
1448		}
1449		fput_light(sock->file, fput_needed);
1450		/* ++SSD_RIL: Garbage_Filter_UDP */
1451		#ifdef CONFIG_ARCH_MSM8960
1452		if (sock->sk != NULL) {
1453			if (sock->sk->sk_protocol == IPPROTO_UDP)
1454				add_or_remove_port(sock->sk, 1);
1455		}
1456		#endif
1457		/* --SSD_RIL: Garbage_Filter_UDP */
1458	}
1459	return err;
1460}
1461
1462/*
1463 *	Perform a listen. Basically, we allow the protocol to do anything
1464 *	necessary for a listen, and if that works, we mark the socket as
1465 *	ready for listening.
1466 */
1467
1468SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1469{
1470	struct socket *sock;
1471	int err, fput_needed;
1472	int somaxconn;
1473
1474	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1475	if (sock) {
1476		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1477		if ((unsigned)backlog > somaxconn)
1478			backlog = somaxconn;
1479
1480		err = security_socket_listen(sock, backlog);
1481		if (!err)
1482			err = sock->ops->listen(sock, backlog);
1483
1484		fput_light(sock->file, fput_needed);
1485		/* ++SSD_RIL: Garbage_Filter_TCP */
1486		if (sock->sk != NULL)
1487			add_or_remove_port(sock->sk, 1);
1488		/* --SSD_RIL: Garbage_Filter_TCP */
1489	}
1490	return err;
1491}
1492
1493/*
1494 *	For accept, we attempt to create a new socket, set up the link
1495 *	with the client, wake up the client, then return the new
1496 *	connected fd. We collect the address of the connector in kernel
1497 *	space and move it to user at the very end. This is unclean because
1498 *	we open the socket then return an error.
1499 *
1500 *	1003.1g adds the ability to recvmsg() to query connection pending
1501 *	status to recvmsg. We need to add that support in a way thats
1502 *	clean when we restucture accept also.
1503 */
1504
1505SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1506		int __user *, upeer_addrlen, int, flags)
1507{
1508	struct socket *sock, *newsock;
1509	struct file *newfile;
1510	int err, len, newfd, fput_needed;
1511	struct sockaddr_storage address;
1512
1513	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1514		return -EINVAL;
1515
1516	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1517		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1518
1519	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1520	if (!sock)
1521		goto out;
1522
1523	err = -ENFILE;
1524	newsock = sock_alloc();
1525	if (!newsock)
1526		goto out_put;
1527
1528	newsock->type = sock->type;
1529	newsock->ops = sock->ops;
1530
1531	/*
1532	 * We don't need try_module_get here, as the listening socket (sock)
1533	 * has the protocol module (sock->ops->owner) held.
1534	 */
1535	__module_get(newsock->ops->owner);
1536
1537	newfd = sock_alloc_file(newsock, &newfile, flags);
1538	if (unlikely(newfd < 0)) {
1539		err = newfd;
1540		sock_release(newsock);
1541		goto out_put;
1542	}
1543
1544	err = security_socket_accept(sock, newsock);
1545	if (err)
1546		goto out_fd;
1547
1548	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1549	if (err < 0)
1550		goto out_fd;
1551
1552	if (upeer_sockaddr) {
1553		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1554					  &len, 2) < 0) {
1555			err = -ECONNABORTED;
1556			goto out_fd;
1557		}
1558		err = move_addr_to_user((struct sockaddr *)&address,
1559					len, upeer_sockaddr, upeer_addrlen);
1560		if (err < 0)
1561			goto out_fd;
1562	}
1563
1564	/* File flags are not inherited via accept() unlike another OSes. */
1565
1566	fd_install(newfd, newfile);
1567	err = newfd;
1568
1569out_put:
1570	fput_light(sock->file, fput_needed);
1571out:
1572	return err;
1573out_fd:
1574	fput(newfile);
1575	put_unused_fd(newfd);
1576	goto out_put;
1577}
1578
1579SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1580		int __user *, upeer_addrlen)
1581{
1582	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1583}
1584
1585/*
1586 *	Attempt to connect to a socket with the server address.  The address
1587 *	is in user space so we verify it is OK and move it to kernel space.
1588 *
1589 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1590 *	break bindings
1591 *
1592 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1593 *	other SEQPACKET protocols that take time to connect() as it doesn't
1594 *	include the -EINPROGRESS status for such sockets.
1595 */
1596
1597SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1598		int, addrlen)
1599{
1600	struct socket *sock;
1601	struct sockaddr_storage address;
1602	int err, fput_needed;
1603
1604	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1605	if (!sock)
1606		goto out;
1607	err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1608	if (err < 0)
1609		goto out_put;
1610
1611	err =
1612	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1613	if (err)
1614		goto out_put;
1615
1616	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1617				 sock->file->f_flags);
1618out_put:
1619	fput_light(sock->file, fput_needed);
1620out:
1621	return err;
1622}
1623
1624/*
1625 *	Get the local address ('name') of a socket object. Move the obtained
1626 *	name to user space.
1627 */
1628
1629SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1630		int __user *, usockaddr_len)
1631{
1632	struct socket *sock;
1633	struct sockaddr_storage address;
1634	int len, err, fput_needed;
1635
1636	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1637	if (!sock)
1638		goto out;
1639
1640	err = security_socket_getsockname(sock);
1641	if (err)
1642		goto out_put;
1643
1644	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1645	if (err)
1646		goto out_put;
1647	err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1648
1649out_put:
1650	fput_light(sock->file, fput_needed);
1651out:
1652	return err;
1653}
1654
1655/*
1656 *	Get the remote address ('name') of a socket object. Move the obtained
1657 *	name to user space.
1658 */
1659
1660SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1661		int __user *, usockaddr_len)
1662{
1663	struct socket *sock;
1664	struct sockaddr_storage address;
1665	int len, err, fput_needed;
1666
1667	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1668	if (sock != NULL) {
1669		err = security_socket_getpeername(sock);
1670		if (err) {
1671			fput_light(sock->file, fput_needed);
1672			return err;
1673		}
1674
1675		err =
1676		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1677				       1);
1678		if (!err)
1679			err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1680						usockaddr_len);
1681		fput_light(sock->file, fput_needed);
1682	}
1683	return err;
1684}
1685
1686/*
1687 *	Send a datagram to a given address. We move the address into kernel
1688 *	space and check the user space data area is readable before invoking
1689 *	the protocol.
1690 */
1691
1692SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1693		unsigned, flags, struct sockaddr __user *, addr,
1694		int, addr_len)
1695{
1696	struct socket *sock;
1697	struct sockaddr_storage address;
1698	int err;
1699	struct msghdr msg;
1700	struct iovec iov;
1701	int fput_needed;
1702
1703	if (len > INT_MAX)
1704		len = INT_MAX;
1705	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1706	if (!sock)
1707		goto out;
1708
1709	iov.iov_base = buff;
1710	iov.iov_len = len;
1711	msg.msg_name = NULL;
1712	msg.msg_iov = &iov;
1713	msg.msg_iovlen = 1;
1714	msg.msg_control = NULL;
1715	msg.msg_controllen = 0;
1716	msg.msg_namelen = 0;
1717	if (addr) {
1718		err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1719		if (err < 0)
1720			goto out_put;
1721		msg.msg_name = (struct sockaddr *)&address;
1722		msg.msg_namelen = addr_len;
1723	}
1724	if (sock->file->f_flags & O_NONBLOCK)
1725		flags |= MSG_DONTWAIT;
1726	msg.msg_flags = flags;
1727	err = sock_sendmsg(sock, &msg, len);
1728
1729out_put:
1730	fput_light(sock->file, fput_needed);
1731out:
1732	return err;
1733}
1734
1735/*
1736 *	Send a datagram down a socket.
1737 */
1738
1739SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1740		unsigned, flags)
1741{
1742	return sys_sendto(fd, buff, len, flags, NULL, 0);
1743}
1744
1745/*
1746 *	Receive a frame from the socket and optionally record the address of the
1747 *	sender. We verify the buffers are writable and if needed move the
1748 *	sender address from kernel to user space.
1749 */
1750
1751SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1752		unsigned, flags, struct sockaddr __user *, addr,
1753		int __user *, addr_len)
1754{
1755	struct socket *sock;
1756	struct iovec iov;
1757	struct msghdr msg;
1758	struct sockaddr_storage address;
1759	int err, err2;
1760	int fput_needed;
1761
1762	if (size > INT_MAX)
1763		size = INT_MAX;
1764	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1765	if (!sock)
1766		goto out;
1767
1768	msg.msg_control = NULL;
1769	msg.msg_controllen = 0;
1770	msg.msg_iovlen = 1;
1771	msg.msg_iov = &iov;
1772	iov.iov_len = size;
1773	iov.iov_base = ubuf;
1774	msg.msg_name = (struct sockaddr *)&address;
1775	msg.msg_namelen = sizeof(address);
1776	if (sock->file->f_flags & O_NONBLOCK)
1777		flags |= MSG_DONTWAIT;
1778	err = sock_recvmsg(sock, &msg, size, flags);
1779
1780	if (err >= 0 && addr != NULL) {
1781		err2 = move_addr_to_user((struct sockaddr *)&address,
1782					 msg.msg_namelen, addr, addr_len);
1783		if (err2 < 0)
1784			err = err2;
1785	}
1786
1787	fput_light(sock->file, fput_needed);
1788out:
1789	return err;
1790}
1791
1792/*
1793 *	Receive a datagram from a socket.
1794 */
1795
1796asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1797			 unsigned flags)
1798{
1799	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1800}
1801
1802/*
1803 *	Set a socket option. Because we don't know the option lengths we have
1804 *	to pass the user mode parameter for the protocols to sort out.
1805 */
1806
1807SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1808		char __user *, optval, int, optlen)
1809{
1810	int err, fput_needed;
1811	struct socket *sock;
1812
1813	if (optlen < 0)
1814		return -EINVAL;
1815
1816	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1817	if (sock != NULL) {
1818		err = security_socket_setsockopt(sock, level, optname);
1819		if (err)
1820			goto out_put;
1821
1822		if (level == SOL_SOCKET)
1823			err =
1824			    sock_setsockopt(sock, level, optname, optval,
1825					    optlen);
1826		else
1827			err =
1828			    sock->ops->setsockopt(sock, level, optname, optval,
1829						  optlen);
1830out_put:
1831		fput_light(sock->file, fput_needed);
1832	}
1833	return err;
1834}
1835
1836/*
1837 *	Get a socket option. Because we don't know the option lengths we have
1838 *	to pass a user mode parameter for the protocols to sort out.
1839 */
1840
1841SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1842		char __user *, optval, int __user *, optlen)
1843{
1844	int err, fput_needed;
1845	struct socket *sock;
1846
1847	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1848	if (sock != NULL) {
1849		err = security_socket_getsockopt(sock, level, optname);
1850		if (err)
1851			goto out_put;
1852
1853		if (level == SOL_SOCKET)
1854			err =
1855			    sock_getsockopt(sock, level, optname, optval,
1856					    optlen);
1857		else
1858			err =
1859			    sock->ops->getsockopt(sock, level, optname, optval,
1860						  optlen);
1861out_put:
1862		fput_light(sock->file, fput_needed);
1863	}
1864	return err;
1865}
1866
1867/*
1868 *	Shutdown a socket.
1869 */
1870
1871SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1872{
1873	int err, fput_needed;
1874	struct socket *sock;
1875
1876	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1877	if (sock != NULL) {
1878		err = security_socket_shutdown(sock, how);
1879		if (!err)
1880			err = sock->ops->shutdown(sock, how);
1881		fput_light(sock->file, fput_needed);
1882	}
1883	return err;
1884}
1885
1886/* A couple of helpful macros for getting the address of the 32/64 bit
1887 * fields which are the same type (int / unsigned) on our platforms.
1888 */
1889#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1890#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1891#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1892
1893struct used_address {
1894	struct sockaddr_storage name;
1895	unsigned int name_len;
1896};
1897
1898static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1899			 struct msghdr *msg_sys, unsigned flags,
1900			 struct used_address *used_address)
1901{
1902	struct compat_msghdr __user *msg_compat =
1903	    (struct compat_msghdr __user *)msg;
1904	struct sockaddr_storage address;
1905	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1906	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1907	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1908	/* 20 is size of ipv6_pktinfo */
1909	unsigned char *ctl_buf = ctl;
1910	int err, ctl_len, iov_size, total_len;
1911
1912	err = -EFAULT;
1913	if (MSG_CMSG_COMPAT & flags) {
1914		if (get_compat_msghdr(msg_sys, msg_compat))
1915			return -EFAULT;
1916	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1917		return -EFAULT;
1918
1919	/* do not move before msg_sys is valid */
1920	err = -EMSGSIZE;
1921	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1922		goto out;
1923
1924	/* Check whether to allocate the iovec area */
1925	err = -ENOMEM;
1926	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1927	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1928		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1929		if (!iov)
1930			goto out;
1931	}
1932
1933	/* This will also move the address data into kernel space */
1934	if (MSG_CMSG_COMPAT & flags) {
1935		err = verify_compat_iovec(msg_sys, iov,
1936					  (struct sockaddr *)&address,
1937					  VERIFY_READ);
1938	} else
1939		err = verify_iovec(msg_sys, iov,
1940				   (struct sockaddr *)&address,
1941				   VERIFY_READ);
1942	if (err < 0)
1943		goto out_freeiov;
1944	total_len = err;
1945
1946	err = -ENOBUFS;
1947
1948	if (msg_sys->msg_controllen > INT_MAX)
1949		goto out_freeiov;
1950	ctl_len = msg_sys->msg_controllen;
1951	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1952		err =
1953		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1954						     sizeof(ctl));
1955		if (err)
1956			goto out_freeiov;
1957		ctl_buf = msg_sys->msg_control;
1958		ctl_len = msg_sys->msg_controllen;
1959	} else if (ctl_len) {
1960		if (ctl_len > sizeof(ctl)) {
1961			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1962			if (ctl_buf == NULL)
1963				goto out_freeiov;
1964		}
1965		err = -EFAULT;
1966		/*
1967		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1968		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1969		 * checking falls down on this.
1970		 */
1971		if (copy_from_user(ctl_buf,
1972				   (void __user __force *)msg_sys->msg_control,
1973				   ctl_len))
1974			goto out_freectl;
1975		msg_sys->msg_control = ctl_buf;
1976	}
1977	msg_sys->msg_flags = flags;
1978
1979	if (sock->file->f_flags & O_NONBLOCK)
1980		msg_sys->msg_flags |= MSG_DONTWAIT;
1981	/*
1982	 * If this is sendmmsg() and current destination address is same as
1983	 * previously succeeded address, omit asking LSM's decision.
1984	 * used_address->name_len is initialized to UINT_MAX so that the first
1985	 * destination address never matches.
1986	 */
1987	if (used_address && msg_sys->msg_name &&
1988	    used_address->name_len == msg_sys->msg_namelen &&
1989	    !memcmp(&used_address->name, msg_sys->msg_name,
1990		    used_address->name_len)) {
1991		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1992		goto out_freectl;
1993	}
1994	err = sock_sendmsg(sock, msg_sys, total_len);
1995	/*
1996	 * If this is sendmmsg() and sending to current destination address was
1997	 * successful, remember it.
1998	 */

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