/net/socket.c

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
 108#ifdef CONFIG_UID_STAT
 109#include <linux/uid_stat.h>
 110#endif
 111
 112static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
 113static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 114			 unsigned long nr_segs, loff_t pos);
 115static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 116			  unsigned long nr_segs, loff_t pos);
 117static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 118
 119static int sock_close(struct inode *inode, struct file *file);
 120static unsigned int sock_poll(struct file *file,
 121			      struct poll_table_struct *wait);
 122static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 123#ifdef CONFIG_COMPAT
 124static long compat_sock_ioctl(struct file *file,
 125			      unsigned int cmd, unsigned long arg);
 126#endif
 127static int sock_fasync(int fd, struct file *filp, int on);
 128static ssize_t sock_sendpage(struct file *file, struct page *page,
 129			     int offset, size_t size, loff_t *ppos, int more);
 130static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 131				struct pipe_inode_info *pipe, size_t len,
 132				unsigned int flags);
 133
 134/*
 135 *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 136 *	in the operation structures but are done directly via the socketcall() multiplexor.
 137 */
 138
 139static const struct file_operations socket_file_ops = {
 140	.owner =	THIS_MODULE,
 141	.llseek =	no_llseek,
 142	.aio_read =	sock_aio_read,
 143	.aio_write =	sock_aio_write,
 144	.poll =		sock_poll,
 145	.unlocked_ioctl = sock_ioctl,
 146#ifdef CONFIG_COMPAT
 147	.compat_ioctl = compat_sock_ioctl,
 148#endif
 149	.mmap =		sock_mmap,
 150	.open =		sock_no_open,	/* special open code to disallow open via /proc */
 151	.release =	sock_close,
 152	.fasync =	sock_fasync,
 153	.sendpage =	sock_sendpage,
 154	.splice_write = generic_splice_sendpage,
 155	.splice_read =	sock_splice_read,
 156};
 157
 158/*
 159 *	The protocol list. Each protocol is registered in here.
 160 */
 161
 162static DEFINE_SPINLOCK(net_family_lock);
 163static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 164
 165/*
 166 *	Statistics counters of the socket lists
 167 */
 168
 169static DEFINE_PER_CPU(int, sockets_in_use);
 170
 171/*
 172 * Support routines.
 173 * Move socket addresses back and forth across the kernel/user
 174 * divide and look after the messy bits.
 175 */
 176
 177/**
 178 *	move_addr_to_kernel	-	copy a socket address into kernel space
 179 *	@uaddr: Address in user space
 180 *	@kaddr: Address in kernel space
 181 *	@ulen: Length in user space
 182 *
 183 *	The address is copied into kernel space. If the provided address is
 184 *	too long an error code of -EINVAL is returned. If the copy gives
 185 *	invalid addresses -EFAULT is returned. On a success 0 is returned.
 186 */
 187
 188int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
 189{
 190	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 191		return -EINVAL;
 192	if (ulen == 0)
 193		return 0;
 194	if (copy_from_user(kaddr, uaddr, ulen))
 195		return -EFAULT;
 196	return audit_sockaddr(ulen, kaddr);
 197}
 198
 199/**
 200 *	move_addr_to_user	-	copy an address to user space
 201 *	@kaddr: kernel space address
 202 *	@klen: length of address in kernel
 203 *	@uaddr: user space address
 204 *	@ulen: pointer to user length field
 205 *
 206 *	The value pointed to by ulen on entry is the buffer length available.
 207 *	This is overwritten with the buffer space used. -EINVAL is returned
 208 *	if an overlong buffer is specified or a negative buffer size. -EFAULT
 209 *	is returned if either the buffer or the length field are not
 210 *	accessible.
 211 *	After copying the data up to the limit the user specifies, the true
 212 *	length of the data is written over the length limit the user
 213 *	specified. Zero is returned for a success.
 214 */
 215
 216static int move_addr_to_user(struct sockaddr *kaddr, int klen,
 217			     void __user *uaddr, int __user *ulen)
 218{
 219	int err;
 220	int len;
 221
 222	err = get_user(len, ulen);
 223	if (err)
 224		return err;
 225	if (len > klen)
 226		len = klen;
 227	if (len < 0 || len > sizeof(struct sockaddr_storage))
 228		return -EINVAL;
 229	if (len) {
 230		if (audit_sockaddr(klen, kaddr))
 231			return -ENOMEM;
 232		if (copy_to_user(uaddr, kaddr, len))
 233			return -EFAULT;
 234	}
 235	/*
 236	 *      "fromlen shall refer to the value before truncation.."
 237	 *                      1003.1g
 238	 */
 239	return __put_user(klen, ulen);
 240}
 241
 242static struct kmem_cache *sock_inode_cachep __read_mostly;
 243
 244static struct inode *sock_alloc_inode(struct super_block *sb)
 245{
 246	struct socket_alloc *ei;
 247	struct socket_wq *wq;
 248
 249	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 250	if (!ei)
 251		return NULL;
 252	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 253	if (!wq) {
 254		kmem_cache_free(sock_inode_cachep, ei);
 255		return NULL;
 256	}
 257	init_waitqueue_head(&wq->wait);
 258	wq->fasync_list = NULL;
 259	RCU_INIT_POINTER(ei->socket.wq, wq);
 260
 261	ei->socket.state = SS_UNCONNECTED;
 262	ei->socket.flags = 0;
 263	ei->socket.ops = NULL;
 264	ei->socket.sk = NULL;
 265	ei->socket.file = NULL;
 266
 267	return &ei->vfs_inode;
 268}
 269
 270static void sock_destroy_inode(struct inode *inode)
 271{
 272	struct socket_alloc *ei;
 273	struct socket_wq *wq;
 274
 275	ei = container_of(inode, struct socket_alloc, vfs_inode);
 276	wq = rcu_dereference_protected(ei->socket.wq, 1);
 277	kfree_rcu(wq, rcu);
 278	kmem_cache_free(sock_inode_cachep, ei);
 279}
 280
 281static void init_once(void *foo)
 282{
 283	struct socket_alloc *ei = (struct socket_alloc *)foo;
 284
 285	inode_init_once(&ei->vfs_inode);
 286}
 287
 288static int init_inodecache(void)
 289{
 290	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 291					      sizeof(struct socket_alloc),
 292					      0,
 293					      (SLAB_HWCACHE_ALIGN |
 294					       SLAB_RECLAIM_ACCOUNT |
 295					       SLAB_MEM_SPREAD),
 296					      init_once);
 297	if (sock_inode_cachep == NULL)
 298		return -ENOMEM;
 299	return 0;
 300}
 301
 302static const struct super_operations sockfs_ops = {
 303	.alloc_inode	= sock_alloc_inode,
 304	.destroy_inode	= sock_destroy_inode,
 305	.statfs		= simple_statfs,
 306};
 307
 308/*
 309 * sockfs_dname() is called from d_path().
 310 */
 311static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 312{
 313	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 314				dentry->d_inode->i_ino);
 315}
 316
 317static const struct dentry_operations sockfs_dentry_operations = {
 318	.d_dname  = sockfs_dname,
 319};
 320
 321static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 322			 int flags, const char *dev_name, void *data)
 323{
 324	return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 325		&sockfs_dentry_operations, SOCKFS_MAGIC);
 326}
 327
 328static struct vfsmount *sock_mnt __read_mostly;
 329
 330static struct file_system_type sock_fs_type = {
 331	.name =		"sockfs",
 332	.mount =	sockfs_mount,
 333	.kill_sb =	kill_anon_super,
 334};
 335
 336/*
 337 *	Obtains the first available file descriptor and sets it up for use.
 338 *
 339 *	These functions create file structures and maps them to fd space
 340 *	of the current process. On success it returns file descriptor
 341 *	and file struct implicitly stored in sock->file.
 342 *	Note that another thread may close file descriptor before we return
 343 *	from this function. We use the fact that now we do not refer
 344 *	to socket after mapping. If one day we will need it, this
 345 *	function will increment ref. count on file by 1.
 346 *
 347 *	In any case returned fd MAY BE not valid!
 348 *	This race condition is unavoidable
 349 *	with shared fd spaces, we cannot solve it inside kernel,
 350 *	but we take care of internal coherence yet.
 351 */
 352
 353static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 354{
 355	struct qstr name = { .name = "" };
 356	struct path path;
 357	struct file *file;
 358	int fd;
 359
 360	fd = get_unused_fd_flags(flags);
 361	if (unlikely(fd < 0))
 362		return fd;
 363
 364	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 365	if (unlikely(!path.dentry)) {
 366		put_unused_fd(fd);
 367		return -ENOMEM;
 368	}
 369	path.mnt = mntget(sock_mnt);
 370
 371	d_instantiate(path.dentry, SOCK_INODE(sock));
 372	SOCK_INODE(sock)->i_fop = &socket_file_ops;
 373
 374	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 375		  &socket_file_ops);
 376	if (unlikely(!file)) {
 377		/* drop dentry, keep inode */
 378		ihold(path.dentry->d_inode);
 379		path_put(&path);
 380		put_unused_fd(fd);
 381		return -ENFILE;
 382	}
 383
 384	sock->file = file;
 385	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 386	file->f_pos = 0;
 387	file->private_data = sock;
 388
 389	*f = file;
 390	return fd;
 391}
 392
 393int sock_map_fd(struct socket *sock, int flags)
 394{
 395	struct file *newfile;
 396	int fd = sock_alloc_file(sock, &newfile, flags);
 397
 398	if (likely(fd >= 0))
 399		fd_install(fd, newfile);
 400
 401	return fd;
 402}
 403EXPORT_SYMBOL(sock_map_fd);
 404
 405static struct socket *sock_from_file(struct file *file, int *err)
 406{
 407	if (file->f_op == &socket_file_ops)
 408		return file->private_data;	/* set in sock_map_fd */
 409
 410	*err = -ENOTSOCK;
 411	return NULL;
 412}
 413
 414/**
 415 *	sockfd_lookup - Go from a file number to its socket slot
 416 *	@fd: file handle
 417 *	@err: pointer to an error code return
 418 *
 419 *	The file handle passed in is locked and the socket it is bound
 420 *	too is returned. If an error occurs the err pointer is overwritten
 421 *	with a negative errno code and NULL is returned. The function checks
 422 *	for both invalid handles and passing a handle which is not a socket.
 423 *
 424 *	On a success the socket object pointer is returned.
 425 */
 426
 427struct socket *sockfd_lookup(int fd, int *err)
 428{
 429	struct file *file;
 430	struct socket *sock;
 431
 432	file = fget(fd);
 433	if (!file) {
 434		*err = -EBADF;
 435		return NULL;
 436	}
 437
 438	sock = sock_from_file(file, err);
 439	if (!sock)
 440		fput(file);
 441	return sock;
 442}
 443EXPORT_SYMBOL(sockfd_lookup);
 444
 445static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 446{
 447	struct file *file;
 448	struct socket *sock;
 449
 450	*err = -EBADF;
 451	file = fget_light(fd, fput_needed);
 452	if (file) {
 453		sock = sock_from_file(file, err);
 454		if (sock)
 455			return sock;
 456		fput_light(file, *fput_needed);
 457	}
 458	return NULL;
 459}
 460
 461/**
 462 *	sock_alloc	-	allocate a socket
 463 *
 464 *	Allocate a new inode and socket object. The two are bound together
 465 *	and initialised. The socket is then returned. If we are out of inodes
 466 *	NULL is returned.
 467 */
 468
 469static struct socket *sock_alloc(void)
 470{
 471	struct inode *inode;
 472	struct socket *sock;
 473
 474	inode = new_inode_pseudo(sock_mnt->mnt_sb);
 475	if (!inode)
 476		return NULL;
 477
 478	sock = SOCKET_I(inode);
 479
 480	kmemcheck_annotate_bitfield(sock, type);
 481	inode->i_ino = get_next_ino();
 482	inode->i_mode = S_IFSOCK | S_IRWXUGO;
 483	inode->i_uid = current_fsuid();
 484	inode->i_gid = current_fsgid();
 485
 486	percpu_add(sockets_in_use, 1);
 487	return sock;
 488}
 489
 490/*
 491 *	In theory you can't get an open on this inode, but /proc provides
 492 *	a back door. Remember to keep it shut otherwise you'll let the
 493 *	creepy crawlies in.
 494 */
 495
 496static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 497{
 498	return -ENXIO;
 499}
 500
 501const struct file_operations bad_sock_fops = {
 502	.owner = THIS_MODULE,
 503	.open = sock_no_open,
 504	.llseek = noop_llseek,
 505};
 506
 507/**
 508 *	sock_release	-	close a socket
 509 *	@sock: socket to close
 510 *
 511 *	The socket is released from the protocol stack if it has a release
 512 *	callback, and the inode is then released if the socket is bound to
 513 *	an inode not a file.
 514 */
 515
 516int add_or_remove_port(struct sock *sk, int add_or_remove);	/* SSD_RIL: Garbage_Filter_TCP */
 517
 518void sock_release(struct socket *sock)
 519{
 520	/* ++SSD_RIL: Garbage_Filter_TCP */
 521	if (sock->sk != NULL)
 522		add_or_remove_port(sock->sk, 0);
 523	/* --SSD_RIL: Garbage_Filter_TCP */
 524
 525	if (sock->ops) {
 526		struct module *owner = sock->ops->owner;
 527
 528		sock->ops->release(sock);
 529		sock->ops = NULL;
 530		module_put(owner);
 531	}
 532
 533	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 534		printk(KERN_ERR "sock_release: fasync list not empty!\n");
 535
 536	percpu_sub(sockets_in_use, 1);
 537	if (!sock->file) {
 538		iput(SOCK_INODE(sock));
 539		return;
 540	}
 541	sock->file = NULL;
 542}
 543EXPORT_SYMBOL(sock_release);
 544
 545int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 546{
 547	*tx_flags = 0;
 548	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 549		*tx_flags |= SKBTX_HW_TSTAMP;
 550	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 551		*tx_flags |= SKBTX_SW_TSTAMP;
 552	return 0;
 553}
 554EXPORT_SYMBOL(sock_tx_timestamp);
 555
 556static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 557				       struct msghdr *msg, size_t size)
 558{
 559	struct sock_iocb *si = kiocb_to_siocb(iocb);
 560#ifdef CONFIG_UID_STAT
 561	int err;
 562#endif
 563	sock_update_classid(sock->sk);
 564
 565	si->sock = sock;
 566	si->scm = NULL;
 567	si->msg = msg;
 568	si->size = size;
 569
 570#ifdef CONFIG_UID_STAT
 571	err = sock->ops->sendmsg(iocb, sock, msg, size);
 572	if(err > 0)
 573		uid_stat_tcp_snd(current_uid(), err);
 574	return err;
 575#else
 576	return sock->ops->sendmsg(iocb, sock, msg, size);
 577#endif
 578}
 579
 580static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 581				 struct msghdr *msg, size_t size)
 582{
 583	int err = security_socket_sendmsg(sock, msg, size);
 584
 585	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 586}
 587
 588int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 589{
 590	struct kiocb iocb;
 591	struct sock_iocb siocb;
 592	int ret;
 593
 594	init_sync_kiocb(&iocb, NULL);
 595	iocb.private = &siocb;
 596	ret = __sock_sendmsg(&iocb, sock, msg, size);
 597	if (-EIOCBQUEUED == ret)
 598		ret = wait_on_sync_kiocb(&iocb);
 599	return ret;
 600}
 601EXPORT_SYMBOL(sock_sendmsg);
 602
 603int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 604{
 605	struct kiocb iocb;
 606	struct sock_iocb siocb;
 607	int ret;
 608
 609	init_sync_kiocb(&iocb, NULL);
 610	iocb.private = &siocb;
 611	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 612	if (-EIOCBQUEUED == ret)
 613		ret = wait_on_sync_kiocb(&iocb);
 614	return ret;
 615}
 616
 617int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 618		   struct kvec *vec, size_t num, size_t size)
 619{
 620	mm_segment_t oldfs = get_fs();
 621	int result;
 622
 623	set_fs(KERNEL_DS);
 624	/*
 625	 * the following is safe, since for compiler definitions of kvec and
 626	 * iovec are identical, yielding the same in-core layout and alignment
 627	 */
 628	msg->msg_iov = (struct iovec *)vec;
 629	msg->msg_iovlen = num;
 630	result = sock_sendmsg(sock, msg, size);
 631	set_fs(oldfs);
 632	return result;
 633}
 634EXPORT_SYMBOL(kernel_sendmsg);
 635
 636static int ktime2ts(ktime_t kt, struct timespec *ts)
 637{
 638	if (kt.tv64) {
 639		*ts = ktime_to_timespec(kt);
 640		return 1;
 641	} else {
 642		return 0;
 643	}
 644}
 645
 646/*
 647 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 648 */
 649void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 650	struct sk_buff *skb)
 651{
 652	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 653	struct timespec ts[3];
 654	int empty = 1;
 655	struct skb_shared_hwtstamps *shhwtstamps =
 656		skb_hwtstamps(skb);
 657
 658	/* Race occurred between timestamp enabling and packet
 659	   receiving.  Fill in the current time for now. */
 660	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 661		__net_timestamp(skb);
 662
 663	if (need_software_tstamp) {
 664		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 665			struct timeval tv;
 666			skb_get_timestamp(skb, &tv);
 667			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 668				 sizeof(tv), &tv);
 669		} else {
 670			skb_get_timestampns(skb, &ts[0]);
 671			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 672				 sizeof(ts[0]), &ts[0]);
 673		}
 674	}
 675
 676
 677	memset(ts, 0, sizeof(ts));
 678	if (skb->tstamp.tv64 &&
 679	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 680		skb_get_timestampns(skb, ts + 0);
 681		empty = 0;
 682	}
 683	if (shhwtstamps) {
 684		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 685		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 686			empty = 0;
 687		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 688		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 689			empty = 0;
 690	}
 691	if (!empty)
 692		put_cmsg(msg, SOL_SOCKET,
 693			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 694}
 695EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 696
 697static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 698				   struct sk_buff *skb)
 699{
 700	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 701		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 702			sizeof(__u32), &skb->dropcount);
 703}
 704
 705void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 706	struct sk_buff *skb)
 707{
 708	sock_recv_timestamp(msg, sk, skb);
 709	sock_recv_drops(msg, sk, skb);
 710}
 711EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 712
 713static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 714				       struct msghdr *msg, size_t size, int flags)
 715{
 716	struct sock_iocb *si = kiocb_to_siocb(iocb);
 717#ifdef CONFIG_UID_STAT
 718	int err;
 719#endif
 720	sock_update_classid(sock->sk);
 721
 722	si->sock = sock;
 723	si->scm = NULL;
 724	si->msg = msg;
 725	si->size = size;
 726	si->flags = flags;
 727#ifdef CONFIG_UID_STAT
 728	if(sock->ops != NULL) {
 729		err = sock->ops->recvmsg(iocb, sock, msg, size, flags);
 730	} else {
 731		err = -EFAULT;
 732		printk(KERN_ERR "[NET]__sock_recvmsg_nosec:sock->ops is NULL\n");
 733	}
 734	if(err > 0)
 735		uid_stat_tcp_rcv(current_uid(), err);
 736	return err;
 737#else
 738	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 739#endif
 740}
 741
 742static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 743				 struct msghdr *msg, size_t size, int flags)
 744{
 745	int err = security_socket_recvmsg(sock, msg, size, flags);
 746
 747	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 748}
 749
 750int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 751		 size_t size, int flags)
 752{
 753	struct kiocb iocb;
 754	struct sock_iocb siocb;
 755	int ret;
 756
 757	init_sync_kiocb(&iocb, NULL);
 758	iocb.private = &siocb;
 759	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 760	if (-EIOCBQUEUED == ret)
 761		ret = wait_on_sync_kiocb(&iocb);
 762	return ret;
 763}
 764EXPORT_SYMBOL(sock_recvmsg);
 765
 766static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 767			      size_t size, int flags)
 768{
 769	struct kiocb iocb;
 770	struct sock_iocb siocb;
 771	int ret;
 772
 773	init_sync_kiocb(&iocb, NULL);
 774	iocb.private = &siocb;
 775	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 776	if (-EIOCBQUEUED == ret)
 777		ret = wait_on_sync_kiocb(&iocb);
 778	return ret;
 779}
 780
 781/**
 782 * kernel_recvmsg - Receive a message from a socket (kernel space)
 783 * @sock:       The socket to receive the message from
 784 * @msg:        Received message
 785 * @vec:        Input s/g array for message data
 786 * @num:        Size of input s/g array
 787 * @size:       Number of bytes to read
 788 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 789 *
 790 * On return the msg structure contains the scatter/gather array passed in the
 791 * vec argument. The array is modified so that it consists of the unfilled
 792 * portion of the original array.
 793 *
 794 * The returned value is the total number of bytes received, or an error.
 795 */
 796int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 797		   struct kvec *vec, size_t num, size_t size, int flags)
 798{
 799	mm_segment_t oldfs = get_fs();
 800	int result;
 801
 802	set_fs(KERNEL_DS);
 803	/*
 804	 * the following is safe, since for compiler definitions of kvec and
 805	 * iovec are identical, yielding the same in-core layout and alignment
 806	 */
 807	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 808	result = sock_recvmsg(sock, msg, size, flags);
 809	set_fs(oldfs);
 810	return result;
 811}
 812EXPORT_SYMBOL(kernel_recvmsg);
 813
 814static void sock_aio_dtor(struct kiocb *iocb)
 815{
 816	kfree(iocb->private);
 817}
 818
 819static ssize_t sock_sendpage(struct file *file, struct page *page,
 820			     int offset, size_t size, loff_t *ppos, int more)
 821{
 822	struct socket *sock;
 823	int flags;
 824
 825	sock = file->private_data;
 826
 827	flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
 828	if (more)
 829		flags |= MSG_MORE;
 830
 831	return kernel_sendpage(sock, page, offset, size, flags);
 832}
 833
 834static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 835				struct pipe_inode_info *pipe, size_t len,
 836				unsigned int flags)
 837{
 838	struct socket *sock = file->private_data;
 839
 840	if (unlikely(!sock->ops->splice_read))
 841		return -EINVAL;
 842
 843	sock_update_classid(sock->sk);
 844
 845	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 846}
 847
 848static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 849					 struct sock_iocb *siocb)
 850{
 851	if (!is_sync_kiocb(iocb)) {
 852		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 853		if (!siocb)
 854			return NULL;
 855		iocb->ki_dtor = sock_aio_dtor;
 856	}
 857
 858	siocb->kiocb = iocb;
 859	iocb->private = siocb;
 860	return siocb;
 861}
 862
 863static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 864		struct file *file, const struct iovec *iov,
 865		unsigned long nr_segs)
 866{
 867	struct socket *sock = file->private_data;
 868	size_t size = 0;
 869	int i;
 870
 871	for (i = 0; i < nr_segs; i++)
 872		size += iov[i].iov_len;
 873
 874	msg->msg_name = NULL;
 875	msg->msg_namelen = 0;
 876	msg->msg_control = NULL;
 877	msg->msg_controllen = 0;
 878	msg->msg_iov = (struct iovec *)iov;
 879	msg->msg_iovlen = nr_segs;
 880	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 881
 882	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 883}
 884
 885static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 886				unsigned long nr_segs, loff_t pos)
 887{
 888	struct sock_iocb siocb, *x;
 889
 890	if (pos != 0)
 891		return -ESPIPE;
 892
 893	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 894		return 0;
 895
 896
 897	x = alloc_sock_iocb(iocb, &siocb);
 898	if (!x)
 899		return -ENOMEM;
 900	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 901}
 902
 903static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 904			struct file *file, const struct iovec *iov,
 905			unsigned long nr_segs)
 906{
 907	struct socket *sock = file->private_data;
 908	size_t size = 0;
 909	int i;
 910
 911	for (i = 0; i < nr_segs; i++)
 912		size += iov[i].iov_len;
 913
 914	msg->msg_name = NULL;
 915	msg->msg_namelen = 0;
 916	msg->msg_control = NULL;
 917	msg->msg_controllen = 0;
 918	msg->msg_iov = (struct iovec *)iov;
 919	msg->msg_iovlen = nr_segs;
 920	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 921	if (sock->type == SOCK_SEQPACKET)
 922		msg->msg_flags |= MSG_EOR;
 923
 924	return __sock_sendmsg(iocb, sock, msg, size);
 925}
 926
 927static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 928			  unsigned long nr_segs, loff_t pos)
 929{
 930	struct sock_iocb siocb, *x;
 931
 932	if (pos != 0)
 933		return -ESPIPE;
 934
 935	x = alloc_sock_iocb(iocb, &siocb);
 936	if (!x)
 937		return -ENOMEM;
 938
 939	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 940}
 941
 942/*
 943 * Atomic setting of ioctl hooks to avoid race
 944 * with module unload.
 945 */
 946
 947static DEFINE_MUTEX(br_ioctl_mutex);
 948static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 949
 950void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 951{
 952	mutex_lock(&br_ioctl_mutex);
 953	br_ioctl_hook = hook;
 954	mutex_unlock(&br_ioctl_mutex);
 955}
 956EXPORT_SYMBOL(brioctl_set);
 957
 958static DEFINE_MUTEX(vlan_ioctl_mutex);
 959static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 960
 961void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 962{
 963	mutex_lock(&vlan_ioctl_mutex);
 964	vlan_ioctl_hook = hook;
 965	mutex_unlock(&vlan_ioctl_mutex);
 966}
 967EXPORT_SYMBOL(vlan_ioctl_set);
 968
 969static DEFINE_MUTEX(dlci_ioctl_mutex);
 970static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 971
 972void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 973{
 974	mutex_lock(&dlci_ioctl_mutex);
 975	dlci_ioctl_hook = hook;
 976	mutex_unlock(&dlci_ioctl_mutex);
 977}
 978EXPORT_SYMBOL(dlci_ioctl_set);
 979
 980static long sock_do_ioctl(struct net *net, struct socket *sock,
 981				 unsigned int cmd, unsigned long arg)
 982{
 983	int err;
 984	void __user *argp = (void __user *)arg;
 985
 986	err = sock->ops->ioctl(sock, cmd, arg);
 987
 988	/*
 989	 * If this ioctl is unknown try to hand it down
 990	 * to the NIC driver.
 991	 */
 992	if (err == -ENOIOCTLCMD)
 993		err = dev_ioctl(net, cmd, argp);
 994
 995	return err;
 996}
 997
 998/*
 999 *	With an ioctl, arg may well be a user mode pointer, but we don't know
1000 *	what to do with it - that's up to the protocol still.
1001 */
1002
1003static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1004{
1005	struct socket *sock;
1006	struct sock *sk;
1007	void __user *argp = (void __user *)arg;
1008	int pid, err;
1009	struct net *net;
1010
1011	sock = file->private_data;
1012	sk = sock->sk;
1013	net = sock_net(sk);
1014	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1015		err = dev_ioctl(net, cmd, argp);
1016	} else
1017#ifdef CONFIG_WEXT_CORE
1018	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1019		err = dev_ioctl(net, cmd, argp);
1020	} else
1021#endif
1022		switch (cmd) {
1023		case FIOSETOWN:
1024		case SIOCSPGRP:
1025			err = -EFAULT;
1026			if (get_user(pid, (int __user *)argp))
1027				break;
1028			err = f_setown(sock->file, pid, 1);
1029			break;
1030		case FIOGETOWN:
1031		case SIOCGPGRP:
1032			err = put_user(f_getown(sock->file),
1033				       (int __user *)argp);
1034			break;
1035		case SIOCGIFBR:
1036		case SIOCSIFBR:
1037		case SIOCBRADDBR:
1038		case SIOCBRDELBR:
1039			err = -ENOPKG;
1040			if (!br_ioctl_hook)
1041				request_module("bridge");
1042
1043			mutex_lock(&br_ioctl_mutex);
1044			if (br_ioctl_hook)
1045				err = br_ioctl_hook(net, cmd, argp);
1046			mutex_unlock(&br_ioctl_mutex);
1047			break;
1048		case SIOCGIFVLAN:
1049		case SIOCSIFVLAN:
1050			err = -ENOPKG;
1051			if (!vlan_ioctl_hook)
1052				request_module("8021q");
1053
1054			mutex_lock(&vlan_ioctl_mutex);
1055			if (vlan_ioctl_hook)
1056				err = vlan_ioctl_hook(net, argp);
1057			mutex_unlock(&vlan_ioctl_mutex);
1058			break;
1059		case SIOCADDDLCI:
1060		case SIOCDELDLCI:
1061			err = -ENOPKG;
1062			if (!dlci_ioctl_hook)
1063				request_module("dlci");
1064
1065			mutex_lock(&dlci_ioctl_mutex);
1066			if (dlci_ioctl_hook)
1067				err = dlci_ioctl_hook(cmd, argp);
1068			mutex_unlock(&dlci_ioctl_mutex);
1069			break;
1070		default:
1071			err = sock_do_ioctl(net, sock, cmd, arg);
1072			break;
1073		}
1074	return err;
1075}
1076
1077int sock_create_lite(int family, int type, int protocol, struct socket **res)
1078{
1079	int err;
1080	struct socket *sock = NULL;
1081
1082	err = security_socket_create(family, type, protocol, 1);
1083	if (err)
1084		goto out;
1085
1086	sock = sock_alloc();
1087	if (!sock) {
1088		err = -ENOMEM;
1089		goto out;
1090	}
1091
1092	sock->type = type;
1093	err = security_socket_post_create(sock, family, type, protocol, 1);
1094	if (err)
1095		goto out_release;
1096
1097out:
1098	*res = sock;
1099	return err;
1100out_release:
1101	sock_release(sock);
1102	sock = NULL;
1103	goto out;
1104}
1105EXPORT_SYMBOL(sock_create_lite);
1106
1107/* No kernel lock held - perfect */
1108static unsigned int sock_poll(struct file *file, poll_table *wait)
1109{
1110	struct socket *sock;
1111
1112	/*
1113	 *      We can't return errors to poll, so it's either yes or no.
1114	 */
1115	sock = file->private_data;
1116	if(sock->ops == NULL) {
1117		printk(KERN_ERR "[NET]sock_poll: sock->ops is NULL\n");
1118		return -EFAULT;
1119	} else {
1120		return sock->ops->poll(file, sock, wait);
1121	}
1122}
1123
1124static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1125{
1126	struct socket *sock = file->private_data;
1127
1128	return sock->ops->mmap(file, sock, vma);
1129}
1130
1131static int sock_close(struct inode *inode, struct file *filp)
1132{
1133	/*
1134	 *      It was possible the inode is NULL we were
1135	 *      closing an unfinished socket.
1136	 */
1137
1138	if (!inode) {
1139		printk(KERN_DEBUG "sock_close: NULL inode\n");
1140		return 0;
1141	}
1142	sock_release(SOCKET_I(inode));
1143	return 0;
1144}
1145
1146/*
1147 *	Update the socket async list
1148 *
1149 *	Fasync_list locking strategy.
1150 *
1151 *	1. fasync_list is modified only under process context socket lock
1152 *	   i.e. under semaphore.
1153 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1154 *	   or under socket lock
1155 */
1156
1157static int sock_fasync(int fd, struct file *filp, int on)
1158{
1159	struct socket *sock = filp->private_data;
1160	struct sock *sk = sock->sk;
1161	struct socket_wq *wq;
1162
1163	if (sk == NULL)
1164		return -EINVAL;
1165
1166	lock_sock(sk);
1167	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1168	fasync_helper(fd, filp, on, &wq->fasync_list);
1169
1170	if (!wq->fasync_list)
1171		sock_reset_flag(sk, SOCK_FASYNC);
1172	else
1173		sock_set_flag(sk, SOCK_FASYNC);
1174
1175	release_sock(sk);
1176	return 0;
1177}
1178
1179/* This function may be called only under socket lock or callback_lock or rcu_lock */
1180
1181int sock_wake_async(struct socket *sock, int how, int band)
1182{
1183	struct socket_wq *wq;
1184
1185	if (!sock)
1186		return -1;
1187	rcu_read_lock();
1188	wq = rcu_dereference(sock->wq);
1189	if (!wq || !wq->fasync_list) {
1190		rcu_read_unlock();
1191		return -1;
1192	}
1193	switch (how) {
1194	case SOCK_WAKE_WAITD:
1195		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1196			break;
1197		goto call_kill;
1198	case SOCK_WAKE_SPACE:
1199		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1200			break;
1201		/* fall through */
1202	case SOCK_WAKE_IO:
1203call_kill:
1204		kill_fasync(&wq->fasync_list, SIGIO, band);
1205		break;
1206	case SOCK_WAKE_URG:
1207		kill_fasync(&wq->fasync_list, SIGURG, band);
1208	}
1209	rcu_read_unlock();
1210	return 0;
1211}
1212EXPORT_SYMBOL(sock_wake_async);
1213
1214int __sock_create(struct net *net, int family, int type, int protocol,
1215			 struct socket **res, int kern)
1216{
1217	int err;
1218	struct socket *sock;
1219	const struct net_proto_family *pf;
1220
1221	/*
1222	 *      Check protocol is in range
1223	 */
1224	if (family < 0 || family >= NPROTO)
1225		return -EAFNOSUPPORT;
1226	if (type < 0 || type >= SOCK_MAX)
1227		return -EINVAL;
1228
1229	/* Compatibility.
1230
1231	   This uglymoron is moved from INET layer to here to avoid
1232	   deadlock in module load.
1233	 */
1234	if (family == PF_INET && type == SOCK_PACKET) {
1235		static int warned;
1236		if (!warned) {
1237			warned = 1;
1238			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1239			       current->comm);
1240		}
1241		family = PF_PACKET;
1242	}
1243
1244	err = security_socket_create(family, type, protocol, kern);
1245	if (err)
1246		return err;
1247
1248	/*
1249	 *	Allocate the socket and allow the family to set things up. if
1250	 *	the protocol is 0, the family is instructed to select an appropriate
1251	 *	default.
1252	 */
1253	sock = sock_alloc();
1254	if (!sock) {
1255		if (net_ratelimit())
1256			printk(KERN_WARNING "socket: no more sockets\n");
1257		return -ENFILE;	/* Not exactly a match, but its the
1258				   closest posix thing */
1259	}
1260
1261	sock->type = type;
1262
1263#ifdef CONFIG_MODULES
1264	/* Attempt to load a protocol module if the find failed.
1265	 *
1266	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1267	 * requested real, full-featured networking support upon configuration.
1268	 * Otherwise module support will break!
1269	 */
1270	if (rcu_access_pointer(net_families[family]) == NULL)
1271		request_module("net-pf-%d", family);
1272#endif
1273
1274	rcu_read_lock();
1275	pf = rcu_dereference(net_families[family]);
1276	err = -EAFNOSUPPORT;
1277	if (!pf)
1278		goto out_release;
1279
1280	/*
1281	 * We will call the ->create function, that possibly is in a loadable
1282	 * module, so we have to bump that loadable module refcnt first.
1283	 */
1284	if (!try_module_get(pf->owner))
1285		goto out_release;
1286
1287	/* Now protected by module ref count */
1288	rcu_read_unlock();
1289
1290	err = pf->create(net, sock, protocol, kern);
1291	if (err < 0)
1292		goto out_module_put;
1293
1294	/*
1295	 * Now to bump the refcnt of the [loadable] module that owns this
1296	 * socket at sock_release time we decrement its refcnt.
1297	 */
1298	if (!try_module_get(sock->ops->owner))
1299		goto out_module_busy;
1300
1301	/*
1302	 * Now that we're done with the ->create function, the [loadable]
1303	 * module can have its refcnt decremented
1304	 */
1305	module_put(pf->owner);
1306	err = security_socket_post_create(sock, family, type, protocol, kern);
1307	if (err)
1308		goto out_sock_release;
1309	*res = sock;
1310
1311	/* ++SSD_RIL: Garbage_Filter_UDP */
1312	#ifdef CONFIG_ARCH_MSM8960
1313	if (sock->sk->sk_protocol == IPPROTO_UDP)
1314		add_or_remove_port(sock->sk, 1);
1315	#endif
1316	/* --SSD_RIL: Garbage_Filter_UDP */
1317
1318	return 0;
1319
1320out_module_busy:
1321	err = -EAFNOSUPPORT;
1322out_module_put:
1323	sock->ops = NULL;
1324	module_put(pf->owner);
1325out_sock_release:
1326	sock_release(sock);
1327	return err;
1328
1329out_release:
1330	rcu_read_unlock();
1331	goto out_sock_release;
1332}
1333EXPORT_SYMBOL(__sock_create);
1334
1335int sock_create(int family, int type, int protocol, struct socket **res)
1336{
1337	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1338}
1339EXPORT_SYMBOL(sock_create);
1340
1341int sock_create_kern(int family, int type, int protocol, struct socket **res)
1342{
1343	return __sock_create(&init_net, family, type, protocol, res, 1);
1344}
1345EXPORT_SYMBOL(sock_create_kern);
1346
1347SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1348{
1349	int retval;
1350	struct socket *sock;
1351	int flags;
1352
1353	/* Check the SOCK_* constants for consistency.  */
1354	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1355	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1356	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1357	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1358
1359	flags = type & ~SOCK_TYPE_MASK;
1360	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1361		return -EINVAL;
1362	type &= SOCK_TYPE_MASK;
1363
1364	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1365		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1366
1367	retval = sock_create(family, type, protocol, &sock);
1368	if (retval < 0)
1369		goto out;
1370
1371	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1372	if (retval < 0)
1373		goto out_release;
1374
1375out:
1376	/* It may be already another descriptor 8) Not kernel problem. */
1377	return retval;
1378
1379out_release:
1380	sock_release(sock);
1381	return retval;
1382}
1383
1384/*
1385 *	Create a pair of connected sockets.
1386 */
1387
1388SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1389		int __user *, usockvec)
1390{
1391	struct socket *sock1, *sock2;
1392	int fd1, fd2, err;
1393	struct file *newfile1, *newfile2;
1394	int flags;
1395
1396	flags = type & ~SOCK_TYPE_MASK;
1397	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1398		return -EINVAL;
1399	type &= SOCK_TYPE_MASK;
1400
1401	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1402		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1403
1404	/*
1405	 * Obtain the first socket and check if the underlying protocol
1406	 * supports the socketpair call.
1407	 */
1408
1409	err = sock_create(family, type, protocol, &sock1);
1410	if (err < 0)
1411		goto out;
1412
1413	err = sock_create(family, type, protocol, &sock2);
1414	if (err < 0)
1415		goto out_release_1;
1416
1417	err = sock1->ops->socketpair(sock1, sock2);
1418	if (err < 0)
1419		goto out_release_both;
1420
1421	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1422	if (unlikely(fd1 < 0)) {
1423		err = fd1;
1424		goto out_release_both;
1425	}
1426
1427	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1428	if (unlikely(fd2 < 0)) {
1429		err = fd2;
1430		fput(newfile1);
1431		put_unused_fd(fd1);
1432		sock_release(sock2);
1433		goto out;
1434	}
1435
1436	audit_fd_pair(fd1, fd2);
1437	fd_install(fd1, newfile1);
1438	fd_install(fd2, newfile2);
1439	/* fd1 and fd2 may be already another descriptors.
1440	 * Not kernel problem.
1441	 */
1442
1443	err = put_user(fd1, &usockvec[0]);
1444	if (!err)
1445		err = put_user(fd2, &usockvec[1]);
1446	if (!err)
1447		return 0;
1448
1449	sys_close(fd2);
1450	sys_close(fd1);
1451	return err;
1452
1453out_release_both:
1454	sock_release(sock2);
1455out_release_1:
1456	sock_release(sock1);
1457out:
1458	return err;
1459}
1460
1461/*
1462 *	Bind a name to a socket. Nothing much to do here since it's
1463 *	the protocol's responsibility to handle the local address.
1464 *
1465 *	We move the socket address to kernel space before we call
1466 *	the protocol layer (having also checked the address is ok).
1467 */
1468
1469SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1470{
1471	struct socket *sock;
1472	struct sockaddr_storage address;
1473	int err, fput_needed;
1474
1475	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1476	if (sock) {
1477		err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1478		if (err >= 0) {
1479			err = security_socket_bind(sock,
1480						   (struct sockaddr *)&address,
1481						   addrlen);
1482			if (!err)
1483				err = sock->ops->bind(sock,
1484						      (struct sockaddr *)
1485						      &address, addrlen);
1486		}
1487		fput_light(sock->file, fput_needed);
1488	}
1489	return err;
1490}
1491
1492/*
1493 *	Perform a listen. Basically, we allow the protocol to do anything
1494 *	necessary for a listen, and if that works, we mark the socket as
1495 *	ready for listening.
1496 */
1497
1498SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1499{
1500	struct socket *sock;
1501	int err, fput_needed;
1502	int somaxconn;
1503
1504	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1505	if (sock) {
1506		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1507		if ((unsigned)backlog > somaxconn)
1508			backlog = somaxconn;
1509
1510		err = security_socket_listen(sock, backlog);
1511		if (!err)
1512			err = sock->ops->listen(sock, backlog);
1513
1514		fput_light(sock->file, fput_needed);
1515		/* ++SSD_RIL: Garbage_Filter_TCP */
1516		if (sock->sk != NULL)
1517			add_or_remove_port(sock->sk, 1);
1518		/* --SSD_RIL: Garbage_Filter_TCP */
1519	}
1520	return err;
1521}
1522
1523/*
1524 *	For accept, we attempt to create a new socket, set up the link
1525 *	with the client, wake up the client, then return the new
1526 *	connected fd. We collect the address of the connector in kernel
1527 *	space and move it to user at the very end. This is unclean because
1528 *	we open the socket then return an error.
1529 *
1530 *	1003.1g adds the ability to recvmsg() to query connection pending
1531 *	status to recvmsg. We need to add that support in a way thats
1532 *	clean when we restucture accept also.
1533 */
1534
1535SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1536		int __user *, upeer_addrlen, int, flags)
1537{
1538	struct socket *sock, *newsock;
1539	struct file *newfile;
1540	int err, len, newfd, fput_needed;
1541	struct sockaddr_storage address;
1542
1543	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1544		return -EINVAL;
1545
1546	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1547		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1548
1549	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1550	if (!sock)
1551		goto out;
1552
1553	err = -ENFILE;
1554	newsock = sock_alloc();
1555	if (!newsock)
1556		goto out_put;
1557
1558	newsock->type = sock->type;
1559	newsock->ops = sock->ops;
1560
1561	/*
1562	 * We don't need try_module_get here, as the listening socket (sock)
1563	 * has the protocol module (sock->ops->owner) held.
1564	 */
1565	__module_get(newsock->ops->owner);
1566
1567	newfd = sock_alloc_file(newsock, &newfile, flags);
1568	if (unlikely(newfd < 0)) {
1569		err = newfd;
1570		sock_release(newsock);
1571		goto out_put;
1572	}
1573
1574	err = security_socket_accept(sock, newsock);
1575	if (err)
1576		goto out_fd;
1577
1578	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1579	if (err < 0)
1580		goto out_fd;
1581
1582	if (upeer_sockaddr) {
1583		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1584					  &len, 2) < 0) {
1585			err = -ECONNABORTED;
1586			goto out_fd;
1587		}
1588		err = move_addr_to_user((struct sockaddr *)&address,
1589					len, upeer_sockaddr, upeer_addrlen);
1590		if (err < 0)
1591			goto out_fd;
1592	}
1593
1594	/* File flags are not inherited via accept() unlike another OSes. */
1595
1596	fd_install(newfd, newfile);
1597	err = newfd;
1598
1599out_put:
1600	fput_light(sock->file, fput_needed);
1601out:
1602	return err;
1603out_fd:
1604	fput(newfile);
1605	put_unused_fd(newfd);
1606	goto out_put;
1607}
1608
1609SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1610		int __user *, upeer_addrlen)
1611{
1612	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1613}
1614
1615/*
1616 *	Attempt to connect to a socket with the server address.  The address
1617 *	is in user space so we verify it is OK and move it to kernel space.
1618 *
1619 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1620 *	break bindings
1621 *
1622 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1623 *	other SEQPACKET protocols that take time to connect() as it doesn't
1624 *	include the -EINPROGRESS status for such sockets.
1625 */
1626
1627SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1628		int, addrlen)
1629{
1630	struct socket *sock;
1631	struct sockaddr_storage address;
1632	int err, fput_needed;
1633
1634	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1635	if (!sock)
1636		goto out;
1637	err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1638	if (err < 0)
1639		goto out_put;
1640
1641	err =
1642	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1643	if (err)
1644		goto out_put;
1645
1646	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1647				 sock->file->f_flags);
1648out_put:
1649	fput_light(sock->file, fput_needed);
1650out:
1651	return err;
1652}
1653
1654/*
1655 *	Get the local address ('name') of a socket object. Move the obtained
1656 *	name to user space.
1657 */
1658
1659SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1660		int __user *, usockaddr_len)
1661{
1662	struct socket *sock;
1663	struct sockaddr_storage address;
1664	int len, err, fput_needed;
1665
1666	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1667	if (!sock)
1668		goto out;
1669
1670	err = security_socket_getsockname(sock);
1671	if (err)
1672		goto out_put;
1673
1674	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1675	if (err)
1676		goto out_put;
1677	err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1678
1679out_put:
1680	fput_light(sock->file, fput_needed);
1681out:
1682	return err;
1683}
1684
1685/*
1686 *	Get the remote address ('name') of a socket object. Move the obtained
1687 *	name to user space.
1688 */
1689
1690SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1691		int __user *, usockaddr_len)
1692{
1693	struct socket *sock;
1694	struct sockaddr_storage address;
1695	int len, err, fput_needed;
1696
1697	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1698	if (sock != NULL) {
1699		err = security_socket_getpeername(sock);
1700		if (err) {
1701			fput_light(sock->file, fput_needed);
1702			return err;
1703		}
1704
1705		err =
1706		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1707				       1);
1708		if (!err)
1709			err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1710						usockaddr_len);
1711		fput_light(sock->file, fput_needed);
1712	}
1713	return err;
1714}
1715
1716/*
1717 *	Send a datagram to a given address. We move the address into kernel
1718 *	space and check the user space data area is readable before invoking
1719 *	the protocol.
1720 */
1721
1722SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1723		unsigned, flags, struct sockaddr __user *, addr,
1724		int, addr_len)
1725{
1726	struct socket *sock;
1727	struct sockaddr_storage address;
1728	int err;
1729	struct msghdr msg;
1730	struct iovec iov;
1731	int fput_needed;
1732
1733	if (len > INT_MAX)
1734		len = INT_MAX;
1735	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1736	if (!sock)
1737		goto out;
1738
1739	iov.iov_base = buff;
1740	iov.iov_len = len;
1741	msg.msg_name = NULL;
1742	msg.msg_iov = &iov;
1743	msg.msg_iovlen = 1;
1744	msg.msg_control = NULL;
1745	msg.msg_controllen = 0;
1746	msg.msg_namelen = 0;
1747	if (addr) {
1748		err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1749		if (err < 0)
1750			goto out_put;
1751		msg.msg_name = (struct sockaddr *)&address;
1752		msg.msg_namelen = addr_len;
1753	}
1754	if (sock->file->f_flags & O_NONBLOCK)
1755		flags |= MSG_DONTWAIT;
1756	msg.msg_flags = flags;
1757	err = sock_sendmsg(sock, &msg, len);
1758
1759out_put:
1760	fput_light(sock->file, fput_needed);
1761out:
1762	return err;
1763}
1764
1765/*
1766 *	Send a datagram down a socket.
1767 */
1768
1769SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1770		unsigned, flags)
1771{
1772	return sys_sendto(fd, buff, len, flags, NULL, 0);
1773}
1774
1775/*
1776 *	Receive a frame from the socket and optionally record the address of the
1777 *	sender. We verify the buffers are writable and if needed move the
1778 *	sender address from kernel to user space.
1779 */
1780
1781SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1782		unsigned, flags, struct sockaddr __user *, addr,
1783		int __user *, addr_len)
1784{
1785	struct socket *sock;
1786	struct iovec iov;
1787	struct msghdr msg;
1788	struct sockaddr_storage address;
1789	int err, err2;
1790	int fput_needed;
1791
1792	if (size > INT_MAX)
1793		size = INT_MAX;
1794	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1795	if (!sock)
1796		goto out;
1797
1798	msg.msg_control = NULL;
1799	msg.msg_controllen = 0;
1800	msg.msg_iovlen = 1;
1801	msg.msg_iov = &iov;
1802	iov.iov_len = size;
1803	iov.iov_base = ubuf;
1804	msg.msg_name = (struct sockaddr *)&address;
1805	msg.msg_namelen = sizeof(address);
1806	if (sock->file->f_flags & O_NONBLOCK)
1807		flags |= MSG_DONTWAIT;
1808	err = sock_recvmsg(sock, &msg, size, flags);
1809
1810	if (err >= 0 && addr != NULL) {
1811		err2 = move_addr_to_user((struct sockaddr *)&address,
1812					 msg.msg_namelen, addr, addr_len);
1813		if (err2 < 0)
1814			err = err2;
1815	}
1816
1817	fput_light(sock->file, fput_needed);
1818out:
1819	return err;
1820}
1821
1822/*
1823 *	Receive a datagram from a socket.
1824 */
1825
1826asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1827			 unsigned flags)
1828{
1829	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1830}
1831
1832/*
1833 *	Set a socket option. Because we don't know the option lengths we have
1834 *	to pass the user mode parameter for the protocols to sort out.
1835 */
1836
1837SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1838		char __user *, optval, int, optlen)
1839{
1840	int err, fput_needed;
1841	struct socket *sock;
1842
1843	if (optlen < 0)
1844		return -EINVAL;
1845
1846	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1847	if (sock != NULL) {
1848		err = security_socket_setsockopt(sock, level, optname);
1849		if (err)
1850			goto out_put;
1851
1852		if (level == SOL_SOCKET)
1853			err =
1854			    sock_setsockopt(sock, level, optname, optval,
1855					    optlen);
1856		else
1857			err =
1858			    sock->ops->setsockopt(sock, level, optname, optval,
1859						  optlen);
1860out_put:
1861		fput_light(sock->file, fput_needed);
1862	}
1863	return err;
1864}
1865
1866/*
1867 *	Get a socket option. Because we don't know the option lengths we have
1868 *	to pass a user mode parameter for the protocols to sort out.
1869 */
1870
1871SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1872		char __user *, optval, int __user *, optlen)
1873{
1874	int err, fput_needed;
1875	struct socket *sock;
1876
1877	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1878	if (sock != NULL) {
1879		err = security_socket_getsockopt(sock, level, optname);
1880		if (err)
1881			goto out_put;
1882
1883		if (level == SOL_SOCKET)
1884			err =
1885			    sock_getsockopt(sock, level, optname, optval,
1886					    optlen);
1887		else
1888			err =
1889			    sock->ops->getsockopt(sock, level, optname, optval,
1890						  optlen);
1891out_put:
1892		fput_light(sock->file, fput_needed);
1893	}
1894	return err;
1895}
1896
1897/*
1898 *	Shutdown a socket.
1899 */
1900
1901SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1902{
1903	int err, fput_needed;
1904	struct socket *sock;
1905
1906	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1907	if (sock != NULL) {
1908		err = security_socket_shutdown(sock, how);
1909		if (!err)
1910			err = sock->ops->shutdown(sock, how);
1911		fput_light(sock->file, fput_needed);
1912	}
1913	return err;
1914}
1915
1916/* A couple of helpful macros for getting the address of the 32/64 bit
1917 * fields which are the same type (int / unsigned) on our platforms.
1918 */
1919#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1920#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1921#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1922
1923struct used_address {
1924	struct sockaddr_storage name;
1925	unsigned int name_len;
1926};
1927
1928static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1929			 struct msghdr *msg_sys, unsigned flags,
1930			 struct used_address *used_address)
1931{
1932	struct compat_msghdr __user *msg_compat =
1933	    (struct compat_msghdr __user *)msg;
1934	struct sockaddr_storage address;
1935	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1936	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1937	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1938	/* 20 is size of ipv6_pktinfo */
1939	unsigned char *ctl_buf = ctl;
1940	int err, ctl_len, iov_size, total_len;
1941
1942	err = -EFAULT;
1943	if (MSG_CMSG_COMPAT & flags) {
1944		if (get_compat_msghdr(msg_sys, msg_compat))
1945			return -EFAULT;
1946	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1947		return -EFAULT;
1948
1949	/* do not move before msg_sys is valid */
1950	err = -EMSGSIZE;
1951	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1952		goto out;
1953
1954	/* Check whether to allocate the iovec area */
1955	err = -ENOMEM;
1956	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1957	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1958		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1959		if (!iov)
1960			goto out;
1961	}
1962
1963	/* This will also move the address data into kernel space */
1964	if (MSG_CMSG_COMPAT & flags) {
1965		err = verify_compat_iovec(msg_sys, iov,
1966					  (struct sockaddr *)&address,
1967					  VERIFY_READ);
1968	} else
1969		err = verify_iovec(msg_sys, iov,
1970				   (struct sockaddr *)&address,
1971				   VERIFY_READ);
1972	if (err < 0)
1973		goto out_freeiov;
1974	total_len = err;
1975
1976	err = -ENOBUFS;
1977
1978	if (msg_sys->msg_controllen > INT_MAX)
1979		goto out_freeiov;
1980	ctl_len = msg_sys->msg_controllen;
1981	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1982		err =
1983		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1984						     sizeof(ctl));
1985		if (err)
1986			goto out_freeiov;
1987		ctl_buf = msg_sys->msg_control;
1988		ctl_len = msg_sys->msg_controllen;
1989	} else if (ctl_len) {
1990		if (ctl_len > sizeof(ctl)) {
1991			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1992			if (ctl_buf == NULL)
1993				goto out_freeiov;
1994		}
1995		err = -EFAULT;
1996		/*
1997		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1998		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1999		 * checking falls down on this.
2000		 */
2001		if (copy_from_user(ctl_buf,
2002				   (void __user __force *)msg_sys->msg_control,
2003				   ctl_len))
2004			goto out_freectl;
2005		msg_sys->msg_control = ctl_buf;
2006	}
2007	msg_sys->msg_flags = flags;
2008
2009	if (sock->file->f_flags & O_NONBLOCK)
2010		msg_sys->msg_flags |= MSG_DONTWAIT;
2011	/*
2012	 * If this is sendmmsg() and current destination address…