PageRenderTime 12ms CodeModel.GetById 7ms app.highlight 119ms RepoModel.GetById 1ms app.codeStats 1ms

/net/socket.c

https://bitbucket.org/zarboz/evilz.213.blackout_edition
C | 3399 lines | 2440 code | 488 blank | 471 comment | 385 complexity | 0cf80b73d579a1e956248d940af34f75 MD5 | raw file

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

   1/*
   2 * NET		An implementation of the SOCKET network access protocol.
   3 *
   4 * Version:	@(#)socket.c	1.1.93	18/02/95
   5 *
   6 * Authors:	Orest Zborowski, <obz@Kodak.COM>
   7 *		Ross Biro
   8 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
   9 *
  10 * Fixes:
  11 *		Anonymous	:	NOTSOCK/BADF cleanup. Error fix in
  12 *					shutdown()
  13 *		Alan Cox	:	verify_area() fixes
  14 *		Alan Cox	:	Removed DDI
  15 *		Jonathan Kamens	:	SOCK_DGRAM reconnect bug
  16 *		Alan Cox	:	Moved a load of checks to the very
  17 *					top level.
  18 *		Alan Cox	:	Move address structures to/from user
  19 *					mode above the protocol layers.
  20 *		Rob Janssen	:	Allow 0 length sends.
  21 *		Alan Cox	:	Asynchronous I/O support (cribbed from the
  22 *					tty drivers).
  23 *		Niibe Yutaka	:	Asynchronous I/O for writes (4.4BSD style)
  24 *		Jeff Uphoff	:	Made max number of sockets command-line
  25 *					configurable.
  26 *		Matti Aarnio	:	Made the number of sockets dynamic,
  27 *					to be allocated when needed, and mr.
  28 *					Uphoff's max is used as max to be
  29 *					allowed to allocate.
  30 *		Linus		:	Argh. removed all the socket allocation
  31 *					altogether: it's in the inode now.
  32 *		Alan Cox	:	Made sock_alloc()/sock_release() public
  33 *					for NetROM and future kernel nfsd type
  34 *					stuff.
  35 *		Alan Cox	:	sendmsg/recvmsg basics.
  36 *		Tom Dyas	:	Export net symbols.
  37 *		Marcin Dalecki	:	Fixed problems with CONFIG_NET="n".
  38 *		Alan Cox	:	Added thread locking to sys_* calls
  39 *					for sockets. May have errors at the
  40 *					moment.
  41 *		Kevin Buhr	:	Fixed the dumb errors in the above.
  42 *		Andi Kleen	:	Some small cleanups, optimizations,
  43 *					and fixed a copy_from_user() bug.
  44 *		Tigran Aivazian	:	sys_send(args) calls sys_sendto(args, NULL, 0)
  45 *		Tigran Aivazian	:	Made listen(2) backlog sanity checks
  46 *					protocol-independent
  47 *
  48 *
  49 *		This program is free software; you can redistribute it and/or
  50 *		modify it under the terms of the GNU General Public License
  51 *		as published by the Free Software Foundation; either version
  52 *		2 of the License, or (at your option) any later version.
  53 *
  54 *
  55 *	This module is effectively the top level interface to the BSD socket
  56 *	paradigm.
  57 *
  58 *	Based upon Swansea University Computer Society NET3.039
  59 */
  60
  61#include <linux/mm.h>
  62#include <linux/socket.h>
  63#include <linux/file.h>
  64#include <linux/net.h>
  65#include <linux/interrupt.h>
  66#include <linux/thread_info.h>
  67#include <linux/rcupdate.h>
  68#include <linux/netdevice.h>
  69#include <linux/proc_fs.h>
  70#include <linux/seq_file.h>
  71#include <linux/mutex.h>
  72#include <linux/wanrouter.h>
  73#include <linux/if_bridge.h>
  74#include <linux/if_frad.h>
  75#include <linux/if_vlan.h>
  76#include <linux/init.h>
  77#include <linux/poll.h>
  78#include <linux/cache.h>
  79#include <linux/module.h>
  80#include <linux/highmem.h>
  81#include <linux/mount.h>
  82#include <linux/security.h>
  83#include <linux/syscalls.h>
  84#include <linux/compat.h>
  85#include <linux/kmod.h>
  86#include <linux/audit.h>
  87#include <linux/wireless.h>
  88#include <linux/nsproxy.h>
  89#include <linux/magic.h>
  90#include <linux/slab.h>
  91
  92#include <asm/uaccess.h>
  93#include <asm/unistd.h>
  94
  95#include <net/compat.h>
  96#include <net/wext.h>
  97#include <net/cls_cgroup.h>
  98
  99#include <net/sock.h>
 100#include <linux/netfilter.h>
 101
 102#include <linux/if_tun.h>
 103#include <linux/ipv6_route.h>
 104#include <linux/route.h>
 105#include <linux/sockios.h>
 106#include <linux/atalk.h>
 107
 108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
 109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 110			 unsigned long nr_segs, loff_t pos);
 111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 112			  unsigned long nr_segs, loff_t pos);
 113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 114
 115static int sock_close(struct inode *inode, struct file *file);
 116static unsigned int sock_poll(struct file *file,
 117			      struct poll_table_struct *wait);
 118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 119#ifdef CONFIG_COMPAT
 120static long compat_sock_ioctl(struct file *file,
 121			      unsigned int cmd, unsigned long arg);
 122#endif
 123static int sock_fasync(int fd, struct file *filp, int on);
 124static ssize_t sock_sendpage(struct file *file, struct page *page,
 125			     int offset, size_t size, loff_t *ppos, int more);
 126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 127				struct pipe_inode_info *pipe, size_t len,
 128				unsigned int flags);
 129
 130/*
 131 *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 132 *	in the operation structures but are done directly via the socketcall() multiplexor.
 133 */
 134
 135static const struct file_operations socket_file_ops = {
 136	.owner =	THIS_MODULE,
 137	.llseek =	no_llseek,
 138	.aio_read =	sock_aio_read,
 139	.aio_write =	sock_aio_write,
 140	.poll =		sock_poll,
 141	.unlocked_ioctl = sock_ioctl,
 142#ifdef CONFIG_COMPAT
 143	.compat_ioctl = compat_sock_ioctl,
 144#endif
 145	.mmap =		sock_mmap,
 146	.open =		sock_no_open,	/* special open code to disallow open via /proc */
 147	.release =	sock_close,
 148	.fasync =	sock_fasync,
 149	.sendpage =	sock_sendpage,
 150	.splice_write = generic_splice_sendpage,
 151	.splice_read =	sock_splice_read,
 152};
 153
 154/*
 155 *	The protocol list. Each protocol is registered in here.
 156 */
 157
 158static DEFINE_SPINLOCK(net_family_lock);
 159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 160
 161/*
 162 *	Statistics counters of the socket lists
 163 */
 164
 165static DEFINE_PER_CPU(int, sockets_in_use);
 166
 167/*
 168 * Support routines.
 169 * Move socket addresses back and forth across the kernel/user
 170 * divide and look after the messy bits.
 171 */
 172
 173/**
 174 *	move_addr_to_kernel	-	copy a socket address into kernel space
 175 *	@uaddr: Address in user space
 176 *	@kaddr: Address in kernel space
 177 *	@ulen: Length in user space
 178 *
 179 *	The address is copied into kernel space. If the provided address is
 180 *	too long an error code of -EINVAL is returned. If the copy gives
 181 *	invalid addresses -EFAULT is returned. On a success 0 is returned.
 182 */
 183
 184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
 185{
 186	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 187		return -EINVAL;
 188	if (ulen == 0)
 189		return 0;
 190	if (copy_from_user(kaddr, uaddr, ulen))
 191		return -EFAULT;
 192	return audit_sockaddr(ulen, kaddr);
 193}
 194
 195/**
 196 *	move_addr_to_user	-	copy an address to user space
 197 *	@kaddr: kernel space address
 198 *	@klen: length of address in kernel
 199 *	@uaddr: user space address
 200 *	@ulen: pointer to user length field
 201 *
 202 *	The value pointed to by ulen on entry is the buffer length available.
 203 *	This is overwritten with the buffer space used. -EINVAL is returned
 204 *	if an overlong buffer is specified or a negative buffer size. -EFAULT
 205 *	is returned if either the buffer or the length field are not
 206 *	accessible.
 207 *	After copying the data up to the limit the user specifies, the true
 208 *	length of the data is written over the length limit the user
 209 *	specified. Zero is returned for a success.
 210 */
 211
 212static int move_addr_to_user(struct sockaddr *kaddr, int klen,
 213			     void __user *uaddr, int __user *ulen)
 214{
 215	int err;
 216	int len;
 217
 218	err = get_user(len, ulen);
 219	if (err)
 220		return err;
 221	if (len > klen)
 222		len = klen;
 223	if (len < 0 || len > sizeof(struct sockaddr_storage))
 224		return -EINVAL;
 225	if (len) {
 226		if (audit_sockaddr(klen, kaddr))
 227			return -ENOMEM;
 228		if (copy_to_user(uaddr, kaddr, len))
 229			return -EFAULT;
 230	}
 231	/*
 232	 *      "fromlen shall refer to the value before truncation.."
 233	 *                      1003.1g
 234	 */
 235	return __put_user(klen, ulen);
 236}
 237
 238static struct kmem_cache *sock_inode_cachep __read_mostly;
 239
 240static struct inode *sock_alloc_inode(struct super_block *sb)
 241{
 242	struct socket_alloc *ei;
 243	struct socket_wq *wq;
 244
 245	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 246	if (!ei)
 247		return NULL;
 248	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 249	if (!wq) {
 250		kmem_cache_free(sock_inode_cachep, ei);
 251		return NULL;
 252	}
 253	init_waitqueue_head(&wq->wait);
 254	wq->fasync_list = NULL;
 255	RCU_INIT_POINTER(ei->socket.wq, wq);
 256
 257	ei->socket.state = SS_UNCONNECTED;
 258	ei->socket.flags = 0;
 259	ei->socket.ops = NULL;
 260	ei->socket.sk = NULL;
 261	ei->socket.file = NULL;
 262
 263	return &ei->vfs_inode;
 264}
 265
 266static void sock_destroy_inode(struct inode *inode)
 267{
 268	struct socket_alloc *ei;
 269	struct socket_wq *wq;
 270
 271	ei = container_of(inode, struct socket_alloc, vfs_inode);
 272	wq = rcu_dereference_protected(ei->socket.wq, 1);
 273	kfree_rcu(wq, rcu);
 274	kmem_cache_free(sock_inode_cachep, ei);
 275}
 276
 277static void init_once(void *foo)
 278{
 279	struct socket_alloc *ei = (struct socket_alloc *)foo;
 280
 281	inode_init_once(&ei->vfs_inode);
 282}
 283
 284static int init_inodecache(void)
 285{
 286	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 287					      sizeof(struct socket_alloc),
 288					      0,
 289					      (SLAB_HWCACHE_ALIGN |
 290					       SLAB_RECLAIM_ACCOUNT |
 291					       SLAB_MEM_SPREAD),
 292					      init_once);
 293	if (sock_inode_cachep == NULL)
 294		return -ENOMEM;
 295	return 0;
 296}
 297
 298static const struct super_operations sockfs_ops = {
 299	.alloc_inode	= sock_alloc_inode,
 300	.destroy_inode	= sock_destroy_inode,
 301	.statfs		= simple_statfs,
 302};
 303
 304/*
 305 * sockfs_dname() is called from d_path().
 306 */
 307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 308{
 309	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 310				dentry->d_inode->i_ino);
 311}
 312
 313static const struct dentry_operations sockfs_dentry_operations = {
 314	.d_dname  = sockfs_dname,
 315};
 316
 317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 318			 int flags, const char *dev_name, void *data)
 319{
 320	return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 321		&sockfs_dentry_operations, SOCKFS_MAGIC);
 322}
 323
 324static struct vfsmount *sock_mnt __read_mostly;
 325
 326static struct file_system_type sock_fs_type = {
 327	.name =		"sockfs",
 328	.mount =	sockfs_mount,
 329	.kill_sb =	kill_anon_super,
 330};
 331
 332/*
 333 *	Obtains the first available file descriptor and sets it up for use.
 334 *
 335 *	These functions create file structures and maps them to fd space
 336 *	of the current process. On success it returns file descriptor
 337 *	and file struct implicitly stored in sock->file.
 338 *	Note that another thread may close file descriptor before we return
 339 *	from this function. We use the fact that now we do not refer
 340 *	to socket after mapping. If one day we will need it, this
 341 *	function will increment ref. count on file by 1.
 342 *
 343 *	In any case returned fd MAY BE not valid!
 344 *	This race condition is unavoidable
 345 *	with shared fd spaces, we cannot solve it inside kernel,
 346 *	but we take care of internal coherence yet.
 347 */
 348
 349static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 350{
 351	struct qstr name = { .name = "" };
 352	struct path path;
 353	struct file *file;
 354	int fd;
 355
 356	fd = get_unused_fd_flags(flags);
 357	if (unlikely(fd < 0))
 358		return fd;
 359
 360	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 361	if (unlikely(!path.dentry)) {
 362		put_unused_fd(fd);
 363		return -ENOMEM;
 364	}
 365	path.mnt = mntget(sock_mnt);
 366
 367	d_instantiate(path.dentry, SOCK_INODE(sock));
 368	SOCK_INODE(sock)->i_fop = &socket_file_ops;
 369
 370	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 371		  &socket_file_ops);
 372	if (unlikely(!file)) {
 373		/* drop dentry, keep inode */
 374		ihold(path.dentry->d_inode);
 375		path_put(&path);
 376		put_unused_fd(fd);
 377		return -ENFILE;
 378	}
 379
 380	sock->file = file;
 381	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 382	file->f_pos = 0;
 383	file->private_data = sock;
 384
 385	*f = file;
 386	return fd;
 387}
 388
 389int sock_map_fd(struct socket *sock, int flags)
 390{
 391	struct file *newfile;
 392	int fd = sock_alloc_file(sock, &newfile, flags);
 393
 394	if (likely(fd >= 0))
 395		fd_install(fd, newfile);
 396
 397	return fd;
 398}
 399EXPORT_SYMBOL(sock_map_fd);
 400
 401static struct socket *sock_from_file(struct file *file, int *err)
 402{
 403	if (file->f_op == &socket_file_ops)
 404		return file->private_data;	/* set in sock_map_fd */
 405
 406	*err = -ENOTSOCK;
 407	return NULL;
 408}
 409
 410/**
 411 *	sockfd_lookup - Go from a file number to its socket slot
 412 *	@fd: file handle
 413 *	@err: pointer to an error code return
 414 *
 415 *	The file handle passed in is locked and the socket it is bound
 416 *	too is returned. If an error occurs the err pointer is overwritten
 417 *	with a negative errno code and NULL is returned. The function checks
 418 *	for both invalid handles and passing a handle which is not a socket.
 419 *
 420 *	On a success the socket object pointer is returned.
 421 */
 422
 423struct socket *sockfd_lookup(int fd, int *err)
 424{
 425	struct file *file;
 426	struct socket *sock;
 427
 428	file = fget(fd);
 429	if (!file) {
 430		*err = -EBADF;
 431		return NULL;
 432	}
 433
 434	sock = sock_from_file(file, err);
 435	if (!sock)
 436		fput(file);
 437	return sock;
 438}
 439EXPORT_SYMBOL(sockfd_lookup);
 440
 441static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 442{
 443	struct file *file;
 444	struct socket *sock;
 445
 446	*err = -EBADF;
 447	file = fget_light(fd, fput_needed);
 448	if (file) {
 449		sock = sock_from_file(file, err);
 450		if (sock)
 451			return sock;
 452		fput_light(file, *fput_needed);
 453	}
 454	return NULL;
 455}
 456
 457/**
 458 *	sock_alloc	-	allocate a socket
 459 *
 460 *	Allocate a new inode and socket object. The two are bound together
 461 *	and initialised. The socket is then returned. If we are out of inodes
 462 *	NULL is returned.
 463 */
 464
 465static struct socket *sock_alloc(void)
 466{
 467	struct inode *inode;
 468	struct socket *sock;
 469
 470	inode = new_inode_pseudo(sock_mnt->mnt_sb);
 471	if (!inode)
 472		return NULL;
 473
 474	sock = SOCKET_I(inode);
 475
 476	kmemcheck_annotate_bitfield(sock, type);
 477	inode->i_ino = get_next_ino();
 478	inode->i_mode = S_IFSOCK | S_IRWXUGO;
 479	inode->i_uid = current_fsuid();
 480	inode->i_gid = current_fsgid();
 481
 482	percpu_add(sockets_in_use, 1);
 483	return sock;
 484}
 485
 486/*
 487 *	In theory you can't get an open on this inode, but /proc provides
 488 *	a back door. Remember to keep it shut otherwise you'll let the
 489 *	creepy crawlies in.
 490 */
 491
 492static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 493{
 494	return -ENXIO;
 495}
 496
 497const struct file_operations bad_sock_fops = {
 498	.owner = THIS_MODULE,
 499	.open = sock_no_open,
 500	.llseek = noop_llseek,
 501};
 502
 503/**
 504 *	sock_release	-	close a socket
 505 *	@sock: socket to close
 506 *
 507 *	The socket is released from the protocol stack if it has a release
 508 *	callback, and the inode is then released if the socket is bound to
 509 *	an inode not a file.
 510 */
 511
 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	sock_update_classid(sock->sk);
 557
 558	si->sock = sock;
 559	si->scm = NULL;
 560	si->msg = msg;
 561	si->size = size;
 562
 563	return sock->ops->sendmsg(iocb, sock, msg, size);
 564}
 565
 566static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 567				 struct msghdr *msg, size_t size)
 568{
 569	int err = security_socket_sendmsg(sock, msg, size);
 570
 571	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 572}
 573
 574int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 575{
 576	struct kiocb iocb;
 577	struct sock_iocb siocb;
 578	int ret;
 579
 580	init_sync_kiocb(&iocb, NULL);
 581	iocb.private = &siocb;
 582	ret = __sock_sendmsg(&iocb, sock, msg, size);
 583	if (-EIOCBQUEUED == ret)
 584		ret = wait_on_sync_kiocb(&iocb);
 585	return ret;
 586}
 587EXPORT_SYMBOL(sock_sendmsg);
 588
 589int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 590{
 591	struct kiocb iocb;
 592	struct sock_iocb siocb;
 593	int ret;
 594
 595	init_sync_kiocb(&iocb, NULL);
 596	iocb.private = &siocb;
 597	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 598	if (-EIOCBQUEUED == ret)
 599		ret = wait_on_sync_kiocb(&iocb);
 600	return ret;
 601}
 602
 603int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 604		   struct kvec *vec, size_t num, size_t size)
 605{
 606	mm_segment_t oldfs = get_fs();
 607	int result;
 608
 609	set_fs(KERNEL_DS);
 610	/*
 611	 * the following is safe, since for compiler definitions of kvec and
 612	 * iovec are identical, yielding the same in-core layout and alignment
 613	 */
 614	msg->msg_iov = (struct iovec *)vec;
 615	msg->msg_iovlen = num;
 616	result = sock_sendmsg(sock, msg, size);
 617	set_fs(oldfs);
 618	return result;
 619}
 620EXPORT_SYMBOL(kernel_sendmsg);
 621
 622static int ktime2ts(ktime_t kt, struct timespec *ts)
 623{
 624	if (kt.tv64) {
 625		*ts = ktime_to_timespec(kt);
 626		return 1;
 627	} else {
 628		return 0;
 629	}
 630}
 631
 632/*
 633 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 634 */
 635void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 636	struct sk_buff *skb)
 637{
 638	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 639	struct timespec ts[3];
 640	int empty = 1;
 641	struct skb_shared_hwtstamps *shhwtstamps =
 642		skb_hwtstamps(skb);
 643
 644	/* Race occurred between timestamp enabling and packet
 645	   receiving.  Fill in the current time for now. */
 646	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 647		__net_timestamp(skb);
 648
 649	if (need_software_tstamp) {
 650		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 651			struct timeval tv;
 652			skb_get_timestamp(skb, &tv);
 653			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 654				 sizeof(tv), &tv);
 655		} else {
 656			skb_get_timestampns(skb, &ts[0]);
 657			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 658				 sizeof(ts[0]), &ts[0]);
 659		}
 660	}
 661
 662
 663	memset(ts, 0, sizeof(ts));
 664	if (skb->tstamp.tv64 &&
 665	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 666		skb_get_timestampns(skb, ts + 0);
 667		empty = 0;
 668	}
 669	if (shhwtstamps) {
 670		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 671		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 672			empty = 0;
 673		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 674		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 675			empty = 0;
 676	}
 677	if (!empty)
 678		put_cmsg(msg, SOL_SOCKET,
 679			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 680}
 681EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 682
 683static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 684				   struct sk_buff *skb)
 685{
 686	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 687		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 688			sizeof(__u32), &skb->dropcount);
 689}
 690
 691void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 692	struct sk_buff *skb)
 693{
 694	sock_recv_timestamp(msg, sk, skb);
 695	sock_recv_drops(msg, sk, skb);
 696}
 697EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 698
 699static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 700				       struct msghdr *msg, size_t size, int flags)
 701{
 702	struct sock_iocb *si = kiocb_to_siocb(iocb);
 703	sock_update_classid(sock->sk);
 704
 705	si->sock = sock;
 706	si->scm = NULL;
 707	si->msg = msg;
 708	si->size = size;
 709	si->flags = flags;
 710	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 711}
 712
 713static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 714				 struct msghdr *msg, size_t size, int flags)
 715{
 716	int err = security_socket_recvmsg(sock, msg, size, flags);
 717
 718	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 719}
 720
 721int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 722		 size_t size, int flags)
 723{
 724	struct kiocb iocb;
 725	struct sock_iocb siocb;
 726	int ret;
 727
 728	init_sync_kiocb(&iocb, NULL);
 729	iocb.private = &siocb;
 730	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 731	if (-EIOCBQUEUED == ret)
 732		ret = wait_on_sync_kiocb(&iocb);
 733	return ret;
 734}
 735EXPORT_SYMBOL(sock_recvmsg);
 736
 737static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 738			      size_t size, int flags)
 739{
 740	struct kiocb iocb;
 741	struct sock_iocb siocb;
 742	int ret;
 743
 744	init_sync_kiocb(&iocb, NULL);
 745	iocb.private = &siocb;
 746	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 747	if (-EIOCBQUEUED == ret)
 748		ret = wait_on_sync_kiocb(&iocb);
 749	return ret;
 750}
 751
 752/**
 753 * kernel_recvmsg - Receive a message from a socket (kernel space)
 754 * @sock:       The socket to receive the message from
 755 * @msg:        Received message
 756 * @vec:        Input s/g array for message data
 757 * @num:        Size of input s/g array
 758 * @size:       Number of bytes to read
 759 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 760 *
 761 * On return the msg structure contains the scatter/gather array passed in the
 762 * vec argument. The array is modified so that it consists of the unfilled
 763 * portion of the original array.
 764 *
 765 * The returned value is the total number of bytes received, or an error.
 766 */
 767int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 768		   struct kvec *vec, size_t num, size_t size, int flags)
 769{
 770	mm_segment_t oldfs = get_fs();
 771	int result;
 772
 773	set_fs(KERNEL_DS);
 774	/*
 775	 * the following is safe, since for compiler definitions of kvec and
 776	 * iovec are identical, yielding the same in-core layout and alignment
 777	 */
 778	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 779	result = sock_recvmsg(sock, msg, size, flags);
 780	set_fs(oldfs);
 781	return result;
 782}
 783EXPORT_SYMBOL(kernel_recvmsg);
 784
 785static void sock_aio_dtor(struct kiocb *iocb)
 786{
 787	kfree(iocb->private);
 788}
 789
 790static ssize_t sock_sendpage(struct file *file, struct page *page,
 791			     int offset, size_t size, loff_t *ppos, int more)
 792{
 793	struct socket *sock;
 794	int flags;
 795
 796	sock = file->private_data;
 797
 798	flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
 799	if (more)
 800		flags |= MSG_MORE;
 801
 802	return kernel_sendpage(sock, page, offset, size, flags);
 803}
 804
 805static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 806				struct pipe_inode_info *pipe, size_t len,
 807				unsigned int flags)
 808{
 809	struct socket *sock = file->private_data;
 810
 811	if (unlikely(!sock->ops->splice_read))
 812		return -EINVAL;
 813
 814	sock_update_classid(sock->sk);
 815
 816	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 817}
 818
 819static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 820					 struct sock_iocb *siocb)
 821{
 822	if (!is_sync_kiocb(iocb)) {
 823		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 824		if (!siocb)
 825			return NULL;
 826		iocb->ki_dtor = sock_aio_dtor;
 827	}
 828
 829	siocb->kiocb = iocb;
 830	iocb->private = siocb;
 831	return siocb;
 832}
 833
 834static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 835		struct file *file, const struct iovec *iov,
 836		unsigned long nr_segs)
 837{
 838	struct socket *sock = file->private_data;
 839	size_t size = 0;
 840	int i;
 841
 842	for (i = 0; i < nr_segs; i++)
 843		size += iov[i].iov_len;
 844
 845	msg->msg_name = NULL;
 846	msg->msg_namelen = 0;
 847	msg->msg_control = NULL;
 848	msg->msg_controllen = 0;
 849	msg->msg_iov = (struct iovec *)iov;
 850	msg->msg_iovlen = nr_segs;
 851	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 852
 853	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 854}
 855
 856static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 857				unsigned long nr_segs, loff_t pos)
 858{
 859	struct sock_iocb siocb, *x;
 860
 861	if (pos != 0)
 862		return -ESPIPE;
 863
 864	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 865		return 0;
 866
 867
 868	x = alloc_sock_iocb(iocb, &siocb);
 869	if (!x)
 870		return -ENOMEM;
 871	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 872}
 873
 874static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 875			struct file *file, const struct iovec *iov,
 876			unsigned long nr_segs)
 877{
 878	struct socket *sock = file->private_data;
 879	size_t size = 0;
 880	int i;
 881
 882	for (i = 0; i < nr_segs; i++)
 883		size += iov[i].iov_len;
 884
 885	msg->msg_name = NULL;
 886	msg->msg_namelen = 0;
 887	msg->msg_control = NULL;
 888	msg->msg_controllen = 0;
 889	msg->msg_iov = (struct iovec *)iov;
 890	msg->msg_iovlen = nr_segs;
 891	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 892	if (sock->type == SOCK_SEQPACKET)
 893		msg->msg_flags |= MSG_EOR;
 894
 895	return __sock_sendmsg(iocb, sock, msg, size);
 896}
 897
 898static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 899			  unsigned long nr_segs, loff_t pos)
 900{
 901	struct sock_iocb siocb, *x;
 902
 903	if (pos != 0)
 904		return -ESPIPE;
 905
 906	x = alloc_sock_iocb(iocb, &siocb);
 907	if (!x)
 908		return -ENOMEM;
 909
 910	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 911}
 912
 913/*
 914 * Atomic setting of ioctl hooks to avoid race
 915 * with module unload.
 916 */
 917
 918static DEFINE_MUTEX(br_ioctl_mutex);
 919static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 920
 921void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 922{
 923	mutex_lock(&br_ioctl_mutex);
 924	br_ioctl_hook = hook;
 925	mutex_unlock(&br_ioctl_mutex);
 926}
 927EXPORT_SYMBOL(brioctl_set);
 928
 929static DEFINE_MUTEX(vlan_ioctl_mutex);
 930static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 931
 932void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 933{
 934	mutex_lock(&vlan_ioctl_mutex);
 935	vlan_ioctl_hook = hook;
 936	mutex_unlock(&vlan_ioctl_mutex);
 937}
 938EXPORT_SYMBOL(vlan_ioctl_set);
 939
 940static DEFINE_MUTEX(dlci_ioctl_mutex);
 941static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 942
 943void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 944{
 945	mutex_lock(&dlci_ioctl_mutex);
 946	dlci_ioctl_hook = hook;
 947	mutex_unlock(&dlci_ioctl_mutex);
 948}
 949EXPORT_SYMBOL(dlci_ioctl_set);
 950
 951static long sock_do_ioctl(struct net *net, struct socket *sock,
 952				 unsigned int cmd, unsigned long arg)
 953{
 954	int err;
 955	void __user *argp = (void __user *)arg;
 956
 957	err = sock->ops->ioctl(sock, cmd, arg);
 958
 959	/*
 960	 * If this ioctl is unknown try to hand it down
 961	 * to the NIC driver.
 962	 */
 963	if (err == -ENOIOCTLCMD)
 964		err = dev_ioctl(net, cmd, argp);
 965
 966	return err;
 967}
 968
 969/*
 970 *	With an ioctl, arg may well be a user mode pointer, but we don't know
 971 *	what to do with it - that's up to the protocol still.
 972 */
 973
 974static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 975{
 976	struct socket *sock;
 977	struct sock *sk;
 978	void __user *argp = (void __user *)arg;
 979	int pid, err;
 980	struct net *net;
 981
 982	sock = file->private_data;
 983	sk = sock->sk;
 984	net = sock_net(sk);
 985	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
 986		err = dev_ioctl(net, cmd, argp);
 987	} else
 988#ifdef CONFIG_WEXT_CORE
 989	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
 990		err = dev_ioctl(net, cmd, argp);
 991	} else
 992#endif
 993		switch (cmd) {
 994		case FIOSETOWN:
 995		case SIOCSPGRP:
 996			err = -EFAULT;
 997			if (get_user(pid, (int __user *)argp))
 998				break;
 999			err = f_setown(sock->file, pid, 1);
1000			break;
1001		case FIOGETOWN:
1002		case SIOCGPGRP:
1003			err = put_user(f_getown(sock->file),
1004				       (int __user *)argp);
1005			break;
1006		case SIOCGIFBR:
1007		case SIOCSIFBR:
1008		case SIOCBRADDBR:
1009		case SIOCBRDELBR:
1010			err = -ENOPKG;
1011			if (!br_ioctl_hook)
1012				request_module("bridge");
1013
1014			mutex_lock(&br_ioctl_mutex);
1015			if (br_ioctl_hook)
1016				err = br_ioctl_hook(net, cmd, argp);
1017			mutex_unlock(&br_ioctl_mutex);
1018			break;
1019		case SIOCGIFVLAN:
1020		case SIOCSIFVLAN:
1021			err = -ENOPKG;
1022			if (!vlan_ioctl_hook)
1023				request_module("8021q");
1024
1025			mutex_lock(&vlan_ioctl_mutex);
1026			if (vlan_ioctl_hook)
1027				err = vlan_ioctl_hook(net, argp);
1028			mutex_unlock(&vlan_ioctl_mutex);
1029			break;
1030		case SIOCADDDLCI:
1031		case SIOCDELDLCI:
1032			err = -ENOPKG;
1033			if (!dlci_ioctl_hook)
1034				request_module("dlci");
1035
1036			mutex_lock(&dlci_ioctl_mutex);
1037			if (dlci_ioctl_hook)
1038				err = dlci_ioctl_hook(cmd, argp);
1039			mutex_unlock(&dlci_ioctl_mutex);
1040			break;
1041		default:
1042			err = sock_do_ioctl(net, sock, cmd, arg);
1043			break;
1044		}
1045	return err;
1046}
1047
1048int sock_create_lite(int family, int type, int protocol, struct socket **res)
1049{
1050	int err;
1051	struct socket *sock = NULL;
1052
1053	err = security_socket_create(family, type, protocol, 1);
1054	if (err)
1055		goto out;
1056
1057	sock = sock_alloc();
1058	if (!sock) {
1059		err = -ENOMEM;
1060		goto out;
1061	}
1062
1063	sock->type = type;
1064	err = security_socket_post_create(sock, family, type, protocol, 1);
1065	if (err)
1066		goto out_release;
1067
1068out:
1069	*res = sock;
1070	return err;
1071out_release:
1072	sock_release(sock);
1073	sock = NULL;
1074	goto out;
1075}
1076EXPORT_SYMBOL(sock_create_lite);
1077
1078/* No kernel lock held - perfect */
1079static unsigned int sock_poll(struct file *file, poll_table *wait)
1080{
1081	struct socket *sock;
1082
1083	/*
1084	 *      We can't return errors to poll, so it's either yes or no.
1085	 */
1086	sock = file->private_data;
1087	return sock->ops->poll(file, sock, wait);
1088}
1089
1090static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1091{
1092	struct socket *sock = file->private_data;
1093
1094	return sock->ops->mmap(file, sock, vma);
1095}
1096
1097static int sock_close(struct inode *inode, struct file *filp)
1098{
1099	/*
1100	 *      It was possible the inode is NULL we were
1101	 *      closing an unfinished socket.
1102	 */
1103
1104	if (!inode) {
1105		printk(KERN_DEBUG "sock_close: NULL inode\n");
1106		return 0;
1107	}
1108	sock_release(SOCKET_I(inode));
1109	return 0;
1110}
1111
1112/*
1113 *	Update the socket async list
1114 *
1115 *	Fasync_list locking strategy.
1116 *
1117 *	1. fasync_list is modified only under process context socket lock
1118 *	   i.e. under semaphore.
1119 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1120 *	   or under socket lock
1121 */
1122
1123static int sock_fasync(int fd, struct file *filp, int on)
1124{
1125	struct socket *sock = filp->private_data;
1126	struct sock *sk = sock->sk;
1127	struct socket_wq *wq;
1128
1129	if (sk == NULL)
1130		return -EINVAL;
1131
1132	lock_sock(sk);
1133	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1134	fasync_helper(fd, filp, on, &wq->fasync_list);
1135
1136	if (!wq->fasync_list)
1137		sock_reset_flag(sk, SOCK_FASYNC);
1138	else
1139		sock_set_flag(sk, SOCK_FASYNC);
1140
1141	release_sock(sk);
1142	return 0;
1143}
1144
1145/* This function may be called only under socket lock or callback_lock or rcu_lock */
1146
1147int sock_wake_async(struct socket *sock, int how, int band)
1148{
1149	struct socket_wq *wq;
1150
1151	if (!sock)
1152		return -1;
1153	rcu_read_lock();
1154	wq = rcu_dereference(sock->wq);
1155	if (!wq || !wq->fasync_list) {
1156		rcu_read_unlock();
1157		return -1;
1158	}
1159	switch (how) {
1160	case SOCK_WAKE_WAITD:
1161		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1162			break;
1163		goto call_kill;
1164	case SOCK_WAKE_SPACE:
1165		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1166			break;
1167		/* fall through */
1168	case SOCK_WAKE_IO:
1169call_kill:
1170		kill_fasync(&wq->fasync_list, SIGIO, band);
1171		break;
1172	case SOCK_WAKE_URG:
1173		kill_fasync(&wq->fasync_list, SIGURG, band);
1174	}
1175	rcu_read_unlock();
1176	return 0;
1177}
1178EXPORT_SYMBOL(sock_wake_async);
1179
1180int __sock_create(struct net *net, int family, int type, int protocol,
1181			 struct socket **res, int kern)
1182{
1183	int err;
1184	struct socket *sock;
1185	const struct net_proto_family *pf;
1186
1187	/*
1188	 *      Check protocol is in range
1189	 */
1190	if (family < 0 || family >= NPROTO)
1191		return -EAFNOSUPPORT;
1192	if (type < 0 || type >= SOCK_MAX)
1193		return -EINVAL;
1194
1195	/* Compatibility.
1196
1197	   This uglymoron is moved from INET layer to here to avoid
1198	   deadlock in module load.
1199	 */
1200	if (family == PF_INET && type == SOCK_PACKET) {
1201		static int warned;
1202		if (!warned) {
1203			warned = 1;
1204			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1205			       current->comm);
1206		}
1207		family = PF_PACKET;
1208	}
1209
1210	err = security_socket_create(family, type, protocol, kern);
1211	if (err)
1212		return err;
1213
1214	/*
1215	 *	Allocate the socket and allow the family to set things up. if
1216	 *	the protocol is 0, the family is instructed to select an appropriate
1217	 *	default.
1218	 */
1219	sock = sock_alloc();
1220	if (!sock) {
1221		if (net_ratelimit())
1222			printk(KERN_WARNING "socket: no more sockets\n");
1223		return -ENFILE;	/* Not exactly a match, but its the
1224				   closest posix thing */
1225	}
1226
1227	sock->type = type;
1228
1229#ifdef CONFIG_MODULES
1230	/* Attempt to load a protocol module if the find failed.
1231	 *
1232	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1233	 * requested real, full-featured networking support upon configuration.
1234	 * Otherwise module support will break!
1235	 */
1236	if (rcu_access_pointer(net_families[family]) == NULL)
1237		request_module("net-pf-%d", family);
1238#endif
1239
1240	rcu_read_lock();
1241	pf = rcu_dereference(net_families[family]);
1242	err = -EAFNOSUPPORT;
1243	if (!pf)
1244		goto out_release;
1245
1246	/*
1247	 * We will call the ->create function, that possibly is in a loadable
1248	 * module, so we have to bump that loadable module refcnt first.
1249	 */
1250	if (!try_module_get(pf->owner))
1251		goto out_release;
1252
1253	/* Now protected by module ref count */
1254	rcu_read_unlock();
1255
1256	err = pf->create(net, sock, protocol, kern);
1257	if (err < 0)
1258		goto out_module_put;
1259
1260	/*
1261	 * Now to bump the refcnt of the [loadable] module that owns this
1262	 * socket at sock_release time we decrement its refcnt.
1263	 */
1264	if (!try_module_get(sock->ops->owner))
1265		goto out_module_busy;
1266
1267	/*
1268	 * Now that we're done with the ->create function, the [loadable]
1269	 * module can have its refcnt decremented
1270	 */
1271	module_put(pf->owner);
1272	err = security_socket_post_create(sock, family, type, protocol, kern);
1273	if (err)
1274		goto out_sock_release;
1275	*res = sock;
1276
1277	/* ++SSD_RIL: Garbage_Filter_UDP */
1278	#ifdef CONFIG_ARCH_MSM8960
1279	if (sock->sk->sk_protocol == IPPROTO_UDP)
1280		add_or_remove_port(sock->sk, 1);
1281	#endif
1282	/* --SSD_RIL: Garbage_Filter_UDP */
1283
1284	return 0;
1285
1286out_module_busy:
1287	err = -EAFNOSUPPORT;
1288out_module_put:
1289	sock->ops = NULL;
1290	module_put(pf->owner);
1291out_sock_release:
1292	sock_release(sock);
1293	return err;
1294
1295out_release:
1296	rcu_read_unlock();
1297	goto out_sock_release;
1298}
1299EXPORT_SYMBOL(__sock_create);
1300
1301int sock_create(int family, int type, int protocol, struct socket **res)
1302{
1303	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1304}
1305EXPORT_SYMBOL(sock_create);
1306
1307int sock_create_kern(int family, int type, int protocol, struct socket **res)
1308{
1309	return __sock_create(&init_net, family, type, protocol, res, 1);
1310}
1311EXPORT_SYMBOL(sock_create_kern);
1312
1313SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1314{
1315	int retval;
1316	struct socket *sock;
1317	int flags;
1318
1319	/* Check the SOCK_* constants for consistency.  */
1320	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1321	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1322	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1323	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1324
1325	flags = type & ~SOCK_TYPE_MASK;
1326	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1327		return -EINVAL;
1328	type &= SOCK_TYPE_MASK;
1329
1330	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1331		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1332
1333	retval = sock_create(family, type, protocol, &sock);
1334	if (retval < 0)
1335		goto out;
1336
1337	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1338	if (retval < 0)
1339		goto out_release;
1340
1341out:
1342	/* It may be already another descriptor 8) Not kernel problem. */
1343	return retval;
1344
1345out_release:
1346	sock_release(sock);
1347	return retval;
1348}
1349
1350/*
1351 *	Create a pair of connected sockets.
1352 */
1353
1354SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1355		int __user *, usockvec)
1356{
1357	struct socket *sock1, *sock2;
1358	int fd1, fd2, err;
1359	struct file *newfile1, *newfile2;
1360	int flags;
1361
1362	flags = type & ~SOCK_TYPE_MASK;
1363	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1364		return -EINVAL;
1365	type &= SOCK_TYPE_MASK;
1366
1367	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1368		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1369
1370	/*
1371	 * Obtain the first socket and check if the underlying protocol
1372	 * supports the socketpair call.
1373	 */
1374
1375	err = sock_create(family, type, protocol, &sock1);
1376	if (err < 0)
1377		goto out;
1378
1379	err = sock_create(family, type, protocol, &sock2);
1380	if (err < 0)
1381		goto out_release_1;
1382
1383	err = sock1->ops->socketpair(sock1, sock2);
1384	if (err < 0)
1385		goto out_release_both;
1386
1387	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1388	if (unlikely(fd1 < 0)) {
1389		err = fd1;
1390		goto out_release_both;
1391	}
1392
1393	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1394	if (unlikely(fd2 < 0)) {
1395		err = fd2;
1396		fput(newfile1);
1397		put_unused_fd(fd1);
1398		sock_release(sock2);
1399		goto out;
1400	}
1401
1402	audit_fd_pair(fd1, fd2);
1403	fd_install(fd1, newfile1);
1404	fd_install(fd2, newfile2);
1405	/* fd1 and fd2 may be already another descriptors.
1406	 * Not kernel problem.
1407	 */
1408
1409	err = put_user(fd1, &usockvec[0]);
1410	if (!err)
1411		err = put_user(fd2, &usockvec[1]);
1412	if (!err)
1413		return 0;
1414
1415	sys_close(fd2);
1416	sys_close(fd1);
1417	return err;
1418
1419out_release_both:
1420	sock_release(sock2);
1421out_release_1:
1422	sock_release(sock1);
1423out:
1424	return err;
1425}
1426
1427/*
1428 *	Bind a name to a socket. Nothing much to do here since it's
1429 *	the protocol's responsibility to handle the local address.
1430 *
1431 *	We move the socket address to kernel space before we call
1432 *	the protocol layer (having also checked the address is ok).
1433 */
1434
1435SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1436{
1437	struct socket *sock;
1438	struct sockaddr_storage address;
1439	int err, fput_needed;
1440
1441	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1442	if (sock) {
1443		err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1444		if (err >= 0) {
1445			err = security_socket_bind(sock,
1446						   (struct sockaddr *)&address,
1447						   addrlen);
1448			if (!err)
1449				err = sock->ops->bind(sock,
1450						      (struct sockaddr *)
1451						      &address, addrlen);
1452		}
1453		fput_light(sock->file, fput_needed);
1454	}
1455	return err;
1456}
1457
1458/*
1459 *	Perform a listen. Basically, we allow the protocol to do anything
1460 *	necessary for a listen, and if that works, we mark the socket as
1461 *	ready for listening.
1462 */
1463
1464SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1465{
1466	struct socket *sock;
1467	int err, fput_needed;
1468	int somaxconn;
1469
1470	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1471	if (sock) {
1472		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1473		if ((unsigned)backlog > somaxconn)
1474			backlog = somaxconn;
1475
1476		err = security_socket_listen(sock, backlog);
1477		if (!err)
1478			err = sock->ops->listen(sock, backlog);
1479
1480		fput_light(sock->file, fput_needed);
1481		/* ++SSD_RIL: Garbage_Filter_TCP */
1482		if (sock->sk != NULL)
1483			add_or_remove_port(sock->sk, 1);
1484		/* --SSD_RIL: Garbage_Filter_TCP */
1485	}
1486	return err;
1487}
1488
1489/*
1490 *	For accept, we attempt to create a new socket, set up the link
1491 *	with the client, wake up the client, then return the new
1492 *	connected fd. We collect the address of the connector in kernel
1493 *	space and move it to user at the very end. This is unclean because
1494 *	we open the socket then return an error.
1495 *
1496 *	1003.1g adds the ability to recvmsg() to query connection pending
1497 *	status to recvmsg. We need to add that support in a way thats
1498 *	clean when we restucture accept also.
1499 */
1500
1501SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1502		int __user *, upeer_addrlen, int, flags)
1503{
1504	struct socket *sock, *newsock;
1505	struct file *newfile;
1506	int err, len, newfd, fput_needed;
1507	struct sockaddr_storage address;
1508
1509	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1510		return -EINVAL;
1511
1512	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1513		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1514
1515	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1516	if (!sock)
1517		goto out;
1518
1519	err = -ENFILE;
1520	newsock = sock_alloc();
1521	if (!newsock)
1522		goto out_put;
1523
1524	newsock->type = sock->type;
1525	newsock->ops = sock->ops;
1526
1527	/*
1528	 * We don't need try_module_get here, as the listening socket (sock)
1529	 * has the protocol module (sock->ops->owner) held.
1530	 */
1531	__module_get(newsock->ops->owner);
1532
1533	newfd = sock_alloc_file(newsock, &newfile, flags);
1534	if (unlikely(newfd < 0)) {
1535		err = newfd;
1536		sock_release(newsock);
1537		goto out_put;
1538	}
1539
1540	err = security_socket_accept(sock, newsock);
1541	if (err)
1542		goto out_fd;
1543
1544	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1545	if (err < 0)
1546		goto out_fd;
1547
1548	if (upeer_sockaddr) {
1549		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1550					  &len, 2) < 0) {
1551			err = -ECONNABORTED;
1552			goto out_fd;
1553		}
1554		err = move_addr_to_user((struct sockaddr *)&address,
1555					len, upeer_sockaddr, upeer_addrlen);
1556		if (err < 0)
1557			goto out_fd;
1558	}
1559
1560	/* File flags are not inherited via accept() unlike another OSes. */
1561
1562	fd_install(newfd, newfile);
1563	err = newfd;
1564
1565out_put:
1566	fput_light(sock->file, fput_needed);
1567out:
1568	return err;
1569out_fd:
1570	fput(newfile);
1571	put_unused_fd(newfd);
1572	goto out_put;
1573}
1574
1575SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1576		int __user *, upeer_addrlen)
1577{
1578	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1579}
1580
1581/*
1582 *	Attempt to connect to a socket with the server address.  The address
1583 *	is in user space so we verify it is OK and move it to kernel space.
1584 *
1585 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1586 *	break bindings
1587 *
1588 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1589 *	other SEQPACKET protocols that take time to connect() as it doesn't
1590 *	include the -EINPROGRESS status for such sockets.
1591 */
1592
1593SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1594		int, addrlen)
1595{
1596	struct socket *sock;
1597	struct sockaddr_storage address;
1598	int err, fput_needed;
1599
1600	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1601	if (!sock)
1602		goto out;
1603	err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1604	if (err < 0)
1605		goto out_put;
1606
1607	err =
1608	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1609	if (err)
1610		goto out_put;
1611
1612	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1613				 sock->file->f_flags);
1614out_put:
1615	fput_light(sock->file, fput_needed);
1616out:
1617	return err;
1618}
1619
1620/*
1621 *	Get the local address ('name') of a socket object. Move the obtained
1622 *	name to user space.
1623 */
1624
1625SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1626		int __user *, usockaddr_len)
1627{
1628	struct socket *sock;
1629	struct sockaddr_storage address;
1630	int len, err, fput_needed;
1631
1632	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1633	if (!sock)
1634		goto out;
1635
1636	err = security_socket_getsockname(sock);
1637	if (err)
1638		goto out_put;
1639
1640	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1641	if (err)
1642		goto out_put;
1643	err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1644
1645out_put:
1646	fput_light(sock->file, fput_needed);
1647out:
1648	return err;
1649}
1650
1651/*
1652 *	Get the remote address ('name') of a socket object. Move the obtained
1653 *	name to user space.
1654 */
1655
1656SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1657		int __user *, usockaddr_len)
1658{
1659	struct socket *sock;
1660	struct sockaddr_storage address;
1661	int len, err, fput_needed;
1662
1663	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1664	if (sock != NULL) {
1665		err = security_socket_getpeername(sock);
1666		if (err) {
1667			fput_light(sock->file, fput_needed);
1668			return err;
1669		}
1670
1671		err =
1672		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1673				       1);
1674		if (!err)
1675			err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1676						usockaddr_len);
1677		fput_light(sock->file, fput_needed);
1678	}
1679	return err;
1680}
1681
1682/*
1683 *	Send a datagram to a given address. We move the address into kernel
1684 *	space and check the user space data area is readable before invoking
1685 *	the protocol.
1686 */
1687
1688SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1689		unsigned, flags, struct sockaddr __user *, addr,
1690		int, addr_len)
1691{
1692	struct socket *sock;
1693	struct sockaddr_storage address;
1694	int err;
1695	struct msghdr msg;
1696	struct iovec iov;
1697	int fput_needed;
1698
1699	if (len > INT_MAX)
1700		len = INT_MAX;
1701	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1702	if (!sock)
1703		goto out;
1704
1705	iov.iov_base = buff;
1706	iov.iov_len = len;
1707	msg.msg_name = NULL;
1708	msg.msg_iov = &iov;
1709	msg.msg_iovlen = 1;
1710	msg.msg_control = NULL;
1711	msg.msg_controllen = 0;
1712	msg.msg_namelen = 0;
1713	if (addr) {
1714		err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1715		if (err < 0)
1716			goto out_put;
1717		msg.msg_name = (struct sockaddr *)&address;
1718		msg.msg_namelen = addr_len;
1719	}
1720	if (sock->file->f_flags & O_NONBLOCK)
1721		flags |= MSG_DONTWAIT;
1722	msg.msg_flags = flags;
1723	err = sock_sendmsg(sock, &msg, len);
1724
1725out_put:
1726	fput_light(sock->file, fput_needed);
1727out:
1728	return err;
1729}
1730
1731/*
1732 *	Send a datagram down a socket.
1733 */
1734
1735SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1736		unsigned, flags)
1737{
1738	return sys_sendto(fd, buff, len, flags, NULL, 0);
1739}
1740
1741/*
1742 *	Receive a frame from the socket and optionally record the address of the
1743 *	sender. We verify the buffers are writable and if needed move the
1744 *	sender address from kernel to user space.
1745 */
1746
1747SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1748		unsigned, flags, struct sockaddr __user *, addr,
1749		int __user *, addr_len)
1750{
1751	struct socket *sock;
1752	struct iovec iov;
1753	struct msghdr msg;
1754	struct sockaddr_storage address;
1755	int err, err2;
1756	int fput_needed;
1757
1758	if (size > INT_MAX)
1759		size = INT_MAX;
1760	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1761	if (!sock)
1762		goto out;
1763
1764	msg.msg_control = NULL;
1765	msg.msg_controllen = 0;
1766	msg.msg_iovlen = 1;
1767	msg.msg_iov = &iov;
1768	iov.iov_len = size;
1769	iov.iov_base = ubuf;
1770	msg.msg_name = (struct sockaddr *)&address;
1771	msg.msg_namelen = sizeof(address);
1772	if (sock->file->f_flags & O_NONBLOCK)
1773		flags |= MSG_DONTWAIT;
1774	err = sock_recvmsg(sock, &msg, size, flags);
1775
1776	if (err >= 0 && addr != NULL) {
1777		err2 = move_addr_to_user((struct sockaddr *)&address,
1778					 msg.msg_namelen, addr, addr_len);
1779		if (err2 < 0)
1780			err = err2;
1781	}
1782
1783	fput_light(sock->file, fput_needed);
1784out:
1785	return err;
1786}
1787
1788/*
1789 *	Receive a datagram from a socket.
1790 */
1791
1792asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1793			 unsigned flags)
1794{
1795	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1796}
1797
1798/*
1799 *	Set a socket option. Because we don't know the option lengths we have
1800 *	to pass the user mode parameter for the protocols to sort out.
1801 */
1802
1803SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1804		char __user *, optval, int, optlen)
1805{
1806	int err, fput_needed;
1807	struct socket *sock;
1808
1809	if (optlen < 0)
1810		return -EINVAL;
1811
1812	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1813	if (sock != NULL) {
1814		err = security_socket_setsockopt(sock, level, optname);
1815		if (err)
1816			goto out_put;
1817
1818		if (level == SOL_SOCKET)
1819			err =
1820			    sock_setsockopt(sock, level, optname, optval,
1821					    optlen);
1822		else
1823			err =
1824			    sock->ops->setsockopt(sock, level, optname, optval,
1825						  optlen);
1826out_put:
1827		fput_light(sock->file, fput_needed);
1828	}
1829	return err;
1830}
1831
1832/*
1833 *	Get a socket option. Because we don't know the option lengths we have
1834 *	to pass a user mode parameter for the protocols to sort out.
1835 */
1836
1837SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1838		char __user *, optval, int __user *, optlen)
1839{
1840	int err, fput_needed;
1841	struct socket *sock;
1842
1843	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1844	if (sock != NULL) {
1845		err = security_socket_getsockopt(sock, level, optname);
1846		if (err)
1847			goto out_put;
1848
1849		if (level == SOL_SOCKET)
1850			err =
1851			    sock_getsockopt(sock, level, optname, optval,
1852					    optlen);
1853		else
1854			err =
1855			    sock->ops->getsockopt(sock, level, optname, optval,
1856						  optlen);
1857out_put:
1858		fput_light(sock->file, fput_needed);
1859	}
1860	return err;
1861}
1862
1863/*
1864 *	Shutdown a socket.
1865 */
1866
1867SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1868{
1869	int err, fput_needed;
1870	struct socket *sock;
1871
1872	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1873	if (sock != NULL) {
1874		err = security_socket_shutdown(sock, how);
1875		if (!err)
1876			err = sock->ops->shutdown(sock, how);
1877		fput_light(sock->file, fput_needed);
1878	}
1879	return err;
1880}
1881
1882/* A couple of helpful macros for getting the address of the 32/64 bit
1883 * fields which are the same type (int / unsigned) on our platforms.
1884 */
1885#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1886#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1887#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1888
1889struct used_address {
1890	struct sockaddr_storage name;
1891	unsigned int name_len;
1892};
1893
1894static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1895			 struct msghdr *msg_sys, unsigned flags,
1896			 struct used_address *used_address)
1897{
1898	struct compat_msghdr __user *msg_compat =
1899	    (struct compat_msghdr __user *)msg;
1900	struct sockaddr_storage address;
1901	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1902	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1903	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1904	/* 20 is size of ipv6_pktinfo */
1905	unsigned char *ctl_buf = ctl;
1906	int err, ctl_len, iov_size, total_len;
1907
1908	err = -EFAULT;
1909	if (MSG_CMSG_COMPAT & flags) {
1910		if (get_compat_msghdr(msg_sys, msg_compat))
1911			return -EFAULT;
1912	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1913		return -EFAULT;
1914
1915	/* do not move before msg_sys is valid */
1916	err = -EMSGSIZE;
1917	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1918		goto out;
1919
1920	/* Check whether to allocate the iovec area */
1921	err = -ENOMEM;
1922	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1923	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1924		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1925		if (!iov)
1926			goto out;
1927	}
1928
1929	/* This will also move the address data into kernel space */
1930	if (MSG_CMSG_COMPAT & flags) {
1931		err = verify_compat_iovec(msg_sys, iov,
1932					  (struct sockaddr *)&address,
1933					  VERIFY_READ);
1934	} else
1935		err = verify_iovec(msg_sys, iov,
1936				   (struct sockaddr *)&address,
1937				   VERIFY_READ);
1938	if (err < 0)
1939		goto out_freeiov;
1940	total_len = err;
1941
1942	err = -ENOBUFS;
1943
1944	if (msg_sys->msg_controllen > INT_MAX)
1945		goto out_freeiov;
1946	ctl_len = msg_sys->msg_controllen;
1947	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1948		err =
1949		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1950						     sizeof(ctl));
1951		if (err)
1952			goto out_freeiov;
1953		ctl_buf = msg_sys->msg_control;
1954		ctl_len = msg_sys->msg_controllen;
1955	} else if (ctl_len) {
1956		if (ctl_len > sizeof(ctl)) {
1957			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1958			if (ctl_buf == NULL)
1959				goto out_freeiov;
1960		}
1961		err = -EFAULT;
1962		/*
1963		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1964		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1965		 * checking falls down on this.
1966		 */
1967		if (copy_from_user(ctl_buf,
1968				   (void __user __force *)msg_sys->msg_control,
1969				   ctl_len))
1970			goto out_freectl;
1971		msg_sys->msg_control = ctl_buf;
1972	}
1973	msg_sys->msg_flags = flags;
1974
1975	if (sock->file->f_flags & O_NONBLOCK)
1976		msg_sys->msg_flags |= MSG_DONTWAIT;
1977	/*
1978	 * If this is sendmmsg() and current destination address is same as
1979	 * previously succeeded address, omit asking LSM's decision.
1980	 * used_address->name_len is initialized to UINT_MAX so that the first
1981	 * destination address never matches.
1982	 */
1983	if (used_address && msg_sys->msg_name &&
1984	    used_address->name_len == msg_sys->msg_namelen &&
1985	    !memcmp(&used_address->name, msg_sys->msg_name,
1986		    used_address->name_len)) {
1987		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1988		goto out_freectl;
1989	}
1990	err = sock_sendmsg(sock, msg_sys, total_len);
1991	/*
1992	 * If this is sendmmsg() and sending to current destination address was
1993	 * successful, remember it.
1994	 */
1995	if (used_address && err >= 0) {
1996		used_address->name_len = msg_sys->msg_namelen;

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