PageRenderTime 132ms CodeModel.GetById 4ms app.highlight 111ms RepoModel.GetById 2ms app.codeStats 0ms

/net/socket.c

https://bitbucket.org/codefirex/kernel_lge_kamakogeebang
C | 3384 lines | 2425 code | 495 blank | 464 comment | 381 complexity | 0e0e00142ecff79c83e1293e19535b2f MD5 | raw file

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

   1/*
   2 * NET		An implementation of the SOCKET network access protocol.
   3 *
   4 * Version:	@(#)socket.c	1.1.93	18/02/95
   5 *
   6 * Authors:	Orest Zborowski, <obz@Kodak.COM>
   7 *		Ross Biro
   8 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
   9 *
  10 * Fixes:
  11 *		Anonymous	:	NOTSOCK/BADF cleanup. Error fix in
  12 *					shutdown()
  13 *		Alan Cox	:	verify_area() fixes
  14 *		Alan Cox	:	Removed DDI
  15 *		Jonathan Kamens	:	SOCK_DGRAM reconnect bug
  16 *		Alan Cox	:	Moved a load of checks to the very
  17 *					top level.
  18 *		Alan Cox	:	Move address structures to/from user
  19 *					mode above the protocol layers.
  20 *		Rob Janssen	:	Allow 0 length sends.
  21 *		Alan Cox	:	Asynchronous I/O support (cribbed from the
  22 *					tty drivers).
  23 *		Niibe Yutaka	:	Asynchronous I/O for writes (4.4BSD style)
  24 *		Jeff Uphoff	:	Made max number of sockets command-line
  25 *					configurable.
  26 *		Matti Aarnio	:	Made the number of sockets dynamic,
  27 *					to be allocated when needed, and mr.
  28 *					Uphoff's max is used as max to be
  29 *					allowed to allocate.
  30 *		Linus		:	Argh. removed all the socket allocation
  31 *					altogether: it's in the inode now.
  32 *		Alan Cox	:	Made sock_alloc()/sock_release() public
  33 *					for NetROM and future kernel nfsd type
  34 *					stuff.
  35 *		Alan Cox	:	sendmsg/recvmsg basics.
  36 *		Tom Dyas	:	Export net symbols.
  37 *		Marcin Dalecki	:	Fixed problems with CONFIG_NET="n".
  38 *		Alan Cox	:	Added thread locking to sys_* calls
  39 *					for sockets. May have errors at the
  40 *					moment.
  41 *		Kevin Buhr	:	Fixed the dumb errors in the above.
  42 *		Andi Kleen	:	Some small cleanups, optimizations,
  43 *					and fixed a copy_from_user() bug.
  44 *		Tigran Aivazian	:	sys_send(args) calls sys_sendto(args, NULL, 0)
  45 *		Tigran Aivazian	:	Made listen(2) backlog sanity checks
  46 *					protocol-independent
  47 *
  48 *
  49 *		This program is free software; you can redistribute it and/or
  50 *		modify it under the terms of the GNU General Public License
  51 *		as published by the Free Software Foundation; either version
  52 *		2 of the License, or (at your option) any later version.
  53 *
  54 *
  55 *	This module is effectively the top level interface to the BSD socket
  56 *	paradigm.
  57 *
  58 *	Based upon Swansea University Computer Society NET3.039
  59 */
  60
  61#include <linux/mm.h>
  62#include <linux/socket.h>
  63#include <linux/file.h>
  64#include <linux/net.h>
  65#include <linux/interrupt.h>
  66#include <linux/thread_info.h>
  67#include <linux/rcupdate.h>
  68#include <linux/netdevice.h>
  69#include <linux/proc_fs.h>
  70#include <linux/seq_file.h>
  71#include <linux/mutex.h>
  72#include <linux/wanrouter.h>
  73#include <linux/if_bridge.h>
  74#include <linux/if_frad.h>
  75#include <linux/if_vlan.h>
  76#include <linux/init.h>
  77#include <linux/poll.h>
  78#include <linux/cache.h>
  79#include <linux/module.h>
  80#include <linux/highmem.h>
  81#include <linux/mount.h>
  82#include <linux/security.h>
  83#include <linux/syscalls.h>
  84#include <linux/compat.h>
  85#include <linux/kmod.h>
  86#include <linux/audit.h>
  87#include <linux/wireless.h>
  88#include <linux/nsproxy.h>
  89#include <linux/magic.h>
  90#include <linux/slab.h>
  91
  92#include <asm/uaccess.h>
  93#include <asm/unistd.h>
  94
  95#include <net/compat.h>
  96#include <net/wext.h>
  97#include <net/cls_cgroup.h>
  98
  99#include <net/sock.h>
 100#include <linux/netfilter.h>
 101
 102#include <linux/if_tun.h>
 103#include <linux/ipv6_route.h>
 104#include <linux/route.h>
 105#include <linux/sockios.h>
 106#include <linux/atalk.h>
 107
 108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
 109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 110			 unsigned long nr_segs, loff_t pos);
 111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 112			  unsigned long nr_segs, loff_t pos);
 113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
 114
 115static int sock_close(struct inode *inode, struct file *file);
 116static unsigned int sock_poll(struct file *file,
 117			      struct poll_table_struct *wait);
 118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 119#ifdef CONFIG_COMPAT
 120static long compat_sock_ioctl(struct file *file,
 121			      unsigned int cmd, unsigned long arg);
 122#endif
 123static int sock_fasync(int fd, struct file *filp, int on);
 124static ssize_t sock_sendpage(struct file *file, struct page *page,
 125			     int offset, size_t size, loff_t *ppos, int more);
 126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 127				struct pipe_inode_info *pipe, size_t len,
 128				unsigned int flags);
 129
 130/*
 131 *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
 132 *	in the operation structures but are done directly via the socketcall() multiplexor.
 133 */
 134
 135static const struct file_operations socket_file_ops = {
 136	.owner =	THIS_MODULE,
 137	.llseek =	no_llseek,
 138	.aio_read =	sock_aio_read,
 139	.aio_write =	sock_aio_write,
 140	.poll =		sock_poll,
 141	.unlocked_ioctl = sock_ioctl,
 142#ifdef CONFIG_COMPAT
 143	.compat_ioctl = compat_sock_ioctl,
 144#endif
 145	.mmap =		sock_mmap,
 146	.open =		sock_no_open,	/* special open code to disallow open via /proc */
 147	.release =	sock_close,
 148	.fasync =	sock_fasync,
 149	.sendpage =	sock_sendpage,
 150	.splice_write = generic_splice_sendpage,
 151	.splice_read =	sock_splice_read,
 152};
 153
 154/*
 155 *	The protocol list. Each protocol is registered in here.
 156 */
 157
 158static DEFINE_SPINLOCK(net_family_lock);
 159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
 160
 161/*
 162 *	Statistics counters of the socket lists
 163 */
 164
 165static DEFINE_PER_CPU(int, sockets_in_use);
 166
 167/*
 168 * Support routines.
 169 * Move socket addresses back and forth across the kernel/user
 170 * divide and look after the messy bits.
 171 */
 172
 173/**
 174 *	move_addr_to_kernel	-	copy a socket address into kernel space
 175 *	@uaddr: Address in user space
 176 *	@kaddr: Address in kernel space
 177 *	@ulen: Length in user space
 178 *
 179 *	The address is copied into kernel space. If the provided address is
 180 *	too long an error code of -EINVAL is returned. If the copy gives
 181 *	invalid addresses -EFAULT is returned. On a success 0 is returned.
 182 */
 183
 184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
 185{
 186	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
 187		return -EINVAL;
 188	if (ulen == 0)
 189		return 0;
 190	if (copy_from_user(kaddr, uaddr, ulen))
 191		return -EFAULT;
 192	return audit_sockaddr(ulen, kaddr);
 193}
 194
 195/**
 196 *	move_addr_to_user	-	copy an address to user space
 197 *	@kaddr: kernel space address
 198 *	@klen: length of address in kernel
 199 *	@uaddr: user space address
 200 *	@ulen: pointer to user length field
 201 *
 202 *	The value pointed to by ulen on entry is the buffer length available.
 203 *	This is overwritten with the buffer space used. -EINVAL is returned
 204 *	if an overlong buffer is specified or a negative buffer size. -EFAULT
 205 *	is returned if either the buffer or the length field are not
 206 *	accessible.
 207 *	After copying the data up to the limit the user specifies, the true
 208 *	length of the data is written over the length limit the user
 209 *	specified. Zero is returned for a success.
 210 */
 211
 212static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
 213			     void __user *uaddr, int __user *ulen)
 214{
 215	int err;
 216	int len;
 217
 218	err = get_user(len, ulen);
 219	if (err)
 220		return err;
 221	if (len > klen)
 222		len = klen;
 223	if (len < 0 || len > sizeof(struct sockaddr_storage))
 224		return -EINVAL;
 225	if (len) {
 226		if (audit_sockaddr(klen, kaddr))
 227			return -ENOMEM;
 228		if (copy_to_user(uaddr, kaddr, len))
 229			return -EFAULT;
 230	}
 231	/*
 232	 *      "fromlen shall refer to the value before truncation.."
 233	 *                      1003.1g
 234	 */
 235	return __put_user(klen, ulen);
 236}
 237
 238static struct kmem_cache *sock_inode_cachep __read_mostly;
 239
 240static struct inode *sock_alloc_inode(struct super_block *sb)
 241{
 242	struct socket_alloc *ei;
 243	struct socket_wq *wq;
 244
 245	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
 246	if (!ei)
 247		return NULL;
 248	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
 249	if (!wq) {
 250		kmem_cache_free(sock_inode_cachep, ei);
 251		return NULL;
 252	}
 253	init_waitqueue_head(&wq->wait);
 254	wq->fasync_list = NULL;
 255	RCU_INIT_POINTER(ei->socket.wq, wq);
 256
 257	ei->socket.state = SS_UNCONNECTED;
 258	ei->socket.flags = 0;
 259	ei->socket.ops = NULL;
 260	ei->socket.sk = NULL;
 261	ei->socket.file = NULL;
 262
 263	return &ei->vfs_inode;
 264}
 265
 266static void sock_destroy_inode(struct inode *inode)
 267{
 268	struct socket_alloc *ei;
 269	struct socket_wq *wq;
 270
 271	ei = container_of(inode, struct socket_alloc, vfs_inode);
 272	wq = rcu_dereference_protected(ei->socket.wq, 1);
 273	kfree_rcu(wq, rcu);
 274	kmem_cache_free(sock_inode_cachep, ei);
 275}
 276
 277static void init_once(void *foo)
 278{
 279	struct socket_alloc *ei = (struct socket_alloc *)foo;
 280
 281	inode_init_once(&ei->vfs_inode);
 282}
 283
 284static int init_inodecache(void)
 285{
 286	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
 287					      sizeof(struct socket_alloc),
 288					      0,
 289					      (SLAB_HWCACHE_ALIGN |
 290					       SLAB_RECLAIM_ACCOUNT |
 291					       SLAB_MEM_SPREAD),
 292					      init_once);
 293	if (sock_inode_cachep == NULL)
 294		return -ENOMEM;
 295	return 0;
 296}
 297
 298static const struct super_operations sockfs_ops = {
 299	.alloc_inode	= sock_alloc_inode,
 300	.destroy_inode	= sock_destroy_inode,
 301	.statfs		= simple_statfs,
 302};
 303
 304/*
 305 * sockfs_dname() is called from d_path().
 306 */
 307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
 308{
 309	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
 310				dentry->d_inode->i_ino);
 311}
 312
 313static const struct dentry_operations sockfs_dentry_operations = {
 314	.d_dname  = sockfs_dname,
 315};
 316
 317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
 318			 int flags, const char *dev_name, void *data)
 319{
 320	return mount_pseudo(fs_type, "socket:", &sockfs_ops,
 321		&sockfs_dentry_operations, SOCKFS_MAGIC);
 322}
 323
 324static struct vfsmount *sock_mnt __read_mostly;
 325
 326static struct file_system_type sock_fs_type = {
 327	.name =		"sockfs",
 328	.mount =	sockfs_mount,
 329	.kill_sb =	kill_anon_super,
 330};
 331
 332/*
 333 *	Obtains the first available file descriptor and sets it up for use.
 334 *
 335 *	These functions create file structures and maps them to fd space
 336 *	of the current process. On success it returns file descriptor
 337 *	and file struct implicitly stored in sock->file.
 338 *	Note that another thread may close file descriptor before we return
 339 *	from this function. We use the fact that now we do not refer
 340 *	to socket after mapping. If one day we will need it, this
 341 *	function will increment ref. count on file by 1.
 342 *
 343 *	In any case returned fd MAY BE not valid!
 344 *	This race condition is unavoidable
 345 *	with shared fd spaces, we cannot solve it inside kernel,
 346 *	but we take care of internal coherence yet.
 347 */
 348
 349static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
 350{
 351	struct qstr name = { .name = "" };
 352	struct path path;
 353	struct file *file;
 354	int fd;
 355
 356	fd = get_unused_fd_flags(flags);
 357	if (unlikely(fd < 0))
 358		return fd;
 359
 360	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
 361	if (unlikely(!path.dentry)) {
 362		put_unused_fd(fd);
 363		return -ENOMEM;
 364	}
 365	path.mnt = mntget(sock_mnt);
 366
 367	d_instantiate(path.dentry, SOCK_INODE(sock));
 368	SOCK_INODE(sock)->i_fop = &socket_file_ops;
 369
 370	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
 371		  &socket_file_ops);
 372	if (unlikely(!file)) {
 373		/* drop dentry, keep inode */
 374		ihold(path.dentry->d_inode);
 375		path_put(&path);
 376		put_unused_fd(fd);
 377		return -ENFILE;
 378	}
 379
 380	sock->file = file;
 381	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
 382	file->f_pos = 0;
 383	file->private_data = sock;
 384
 385	*f = file;
 386	return fd;
 387}
 388
 389int sock_map_fd(struct socket *sock, int flags)
 390{
 391	struct file *newfile;
 392	int fd = sock_alloc_file(sock, &newfile, flags);
 393
 394	if (likely(fd >= 0))
 395		fd_install(fd, newfile);
 396
 397	return fd;
 398}
 399EXPORT_SYMBOL(sock_map_fd);
 400
 401static struct socket *sock_from_file(struct file *file, int *err)
 402{
 403	if (file->f_op == &socket_file_ops)
 404		return file->private_data;	/* set in sock_map_fd */
 405
 406	*err = -ENOTSOCK;
 407	return NULL;
 408}
 409
 410/**
 411 *	sockfd_lookup - Go from a file number to its socket slot
 412 *	@fd: file handle
 413 *	@err: pointer to an error code return
 414 *
 415 *	The file handle passed in is locked and the socket it is bound
 416 *	too is returned. If an error occurs the err pointer is overwritten
 417 *	with a negative errno code and NULL is returned. The function checks
 418 *	for both invalid handles and passing a handle which is not a socket.
 419 *
 420 *	On a success the socket object pointer is returned.
 421 */
 422
 423struct socket *sockfd_lookup(int fd, int *err)
 424{
 425	struct file *file;
 426	struct socket *sock;
 427
 428	file = fget(fd);
 429	if (!file) {
 430		*err = -EBADF;
 431		return NULL;
 432	}
 433
 434	sock = sock_from_file(file, err);
 435	if (!sock)
 436		fput(file);
 437	return sock;
 438}
 439EXPORT_SYMBOL(sockfd_lookup);
 440
 441static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
 442{
 443	struct file *file;
 444	struct socket *sock;
 445
 446	*err = -EBADF;
 447	file = fget_light(fd, fput_needed);
 448	if (file) {
 449		sock = sock_from_file(file, err);
 450		if (sock)
 451			return sock;
 452		fput_light(file, *fput_needed);
 453	}
 454	return NULL;
 455}
 456
 457/**
 458 *	sock_alloc	-	allocate a socket
 459 *
 460 *	Allocate a new inode and socket object. The two are bound together
 461 *	and initialised. The socket is then returned. If we are out of inodes
 462 *	NULL is returned.
 463 */
 464
 465static struct socket *sock_alloc(void)
 466{
 467	struct inode *inode;
 468	struct socket *sock;
 469
 470	inode = new_inode_pseudo(sock_mnt->mnt_sb);
 471	if (!inode)
 472		return NULL;
 473
 474	sock = SOCKET_I(inode);
 475
 476	kmemcheck_annotate_bitfield(sock, type);
 477	inode->i_ino = get_next_ino();
 478	inode->i_mode = S_IFSOCK | S_IRWXUGO;
 479	inode->i_uid = current_fsuid();
 480	inode->i_gid = current_fsgid();
 481
 482	percpu_add(sockets_in_use, 1);
 483	return sock;
 484}
 485
 486/*
 487 *	In theory you can't get an open on this inode, but /proc provides
 488 *	a back door. Remember to keep it shut otherwise you'll let the
 489 *	creepy crawlies in.
 490 */
 491
 492static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
 493{
 494	return -ENXIO;
 495}
 496
 497const struct file_operations bad_sock_fops = {
 498	.owner = THIS_MODULE,
 499	.open = sock_no_open,
 500	.llseek = noop_llseek,
 501};
 502
 503/**
 504 *	sock_release	-	close a socket
 505 *	@sock: socket to close
 506 *
 507 *	The socket is released from the protocol stack if it has a release
 508 *	callback, and the inode is then released if the socket is bound to
 509 *	an inode not a file.
 510 */
 511
 512void sock_release(struct socket *sock)
 513{
 514	if (sock->ops) {
 515		struct module *owner = sock->ops->owner;
 516
 517		sock->ops->release(sock);
 518		sock->ops = NULL;
 519		module_put(owner);
 520	}
 521
 522	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
 523		printk(KERN_ERR "sock_release: fasync list not empty!\n");
 524
 525	if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
 526		return;
 527
 528	percpu_sub(sockets_in_use, 1);
 529	if (!sock->file) {
 530		iput(SOCK_INODE(sock));
 531		return;
 532	}
 533	sock->file = NULL;
 534}
 535EXPORT_SYMBOL(sock_release);
 536
 537int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
 538{
 539	*tx_flags = 0;
 540	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
 541		*tx_flags |= SKBTX_HW_TSTAMP;
 542	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
 543		*tx_flags |= SKBTX_SW_TSTAMP;
 544	if (sock_flag(sk, SOCK_WIFI_STATUS))
 545		*tx_flags |= SKBTX_WIFI_STATUS;
 546	return 0;
 547}
 548EXPORT_SYMBOL(sock_tx_timestamp);
 549
 550static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
 551				       struct msghdr *msg, size_t size)
 552{
 553	struct sock_iocb *si = kiocb_to_siocb(iocb);
 554
 555	sock_update_classid(sock->sk);
 556
 557	sock_update_netprioidx(sock->sk);
 558
 559	si->sock = sock;
 560	si->scm = NULL;
 561	si->msg = msg;
 562	si->size = size;
 563
 564	return sock->ops->sendmsg(iocb, sock, msg, size);
 565}
 566
 567static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
 568				 struct msghdr *msg, size_t size)
 569{
 570	int err = security_socket_sendmsg(sock, msg, size);
 571
 572	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
 573}
 574
 575int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
 576{
 577	struct kiocb iocb;
 578	struct sock_iocb siocb;
 579	int ret;
 580
 581	init_sync_kiocb(&iocb, NULL);
 582	iocb.private = &siocb;
 583	ret = __sock_sendmsg(&iocb, sock, msg, size);
 584	if (-EIOCBQUEUED == ret)
 585		ret = wait_on_sync_kiocb(&iocb);
 586	return ret;
 587}
 588EXPORT_SYMBOL(sock_sendmsg);
 589
 590static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
 591{
 592	struct kiocb iocb;
 593	struct sock_iocb siocb;
 594	int ret;
 595
 596	init_sync_kiocb(&iocb, NULL);
 597	iocb.private = &siocb;
 598	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
 599	if (-EIOCBQUEUED == ret)
 600		ret = wait_on_sync_kiocb(&iocb);
 601	return ret;
 602}
 603
 604int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
 605		   struct kvec *vec, size_t num, size_t size)
 606{
 607	mm_segment_t oldfs = get_fs();
 608	int result;
 609
 610	set_fs(KERNEL_DS);
 611	/*
 612	 * the following is safe, since for compiler definitions of kvec and
 613	 * iovec are identical, yielding the same in-core layout and alignment
 614	 */
 615	msg->msg_iov = (struct iovec *)vec;
 616	msg->msg_iovlen = num;
 617	result = sock_sendmsg(sock, msg, size);
 618	set_fs(oldfs);
 619	return result;
 620}
 621EXPORT_SYMBOL(kernel_sendmsg);
 622
 623static int ktime2ts(ktime_t kt, struct timespec *ts)
 624{
 625	if (kt.tv64) {
 626		*ts = ktime_to_timespec(kt);
 627		return 1;
 628	} else {
 629		return 0;
 630	}
 631}
 632
 633/*
 634 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
 635 */
 636void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
 637	struct sk_buff *skb)
 638{
 639	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
 640	struct timespec ts[3];
 641	int empty = 1;
 642	struct skb_shared_hwtstamps *shhwtstamps =
 643		skb_hwtstamps(skb);
 644
 645	/* Race occurred between timestamp enabling and packet
 646	   receiving.  Fill in the current time for now. */
 647	if (need_software_tstamp && skb->tstamp.tv64 == 0)
 648		__net_timestamp(skb);
 649
 650	if (need_software_tstamp) {
 651		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
 652			struct timeval tv;
 653			skb_get_timestamp(skb, &tv);
 654			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
 655				 sizeof(tv), &tv);
 656		} else {
 657			skb_get_timestampns(skb, &ts[0]);
 658			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
 659				 sizeof(ts[0]), &ts[0]);
 660		}
 661	}
 662
 663
 664	memset(ts, 0, sizeof(ts));
 665	if (skb->tstamp.tv64 &&
 666	    sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
 667		skb_get_timestampns(skb, ts + 0);
 668		empty = 0;
 669	}
 670	if (shhwtstamps) {
 671		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
 672		    ktime2ts(shhwtstamps->syststamp, ts + 1))
 673			empty = 0;
 674		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
 675		    ktime2ts(shhwtstamps->hwtstamp, ts + 2))
 676			empty = 0;
 677	}
 678	if (!empty)
 679		put_cmsg(msg, SOL_SOCKET,
 680			 SCM_TIMESTAMPING, sizeof(ts), &ts);
 681}
 682EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
 683
 684void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
 685	struct sk_buff *skb)
 686{
 687	int ack;
 688
 689	if (!sock_flag(sk, SOCK_WIFI_STATUS))
 690		return;
 691	if (!skb->wifi_acked_valid)
 692		return;
 693
 694	ack = skb->wifi_acked;
 695
 696	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
 697}
 698EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
 699
 700static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
 701				   struct sk_buff *skb)
 702{
 703	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
 704		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
 705			sizeof(__u32), &skb->dropcount);
 706}
 707
 708void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
 709	struct sk_buff *skb)
 710{
 711	sock_recv_timestamp(msg, sk, skb);
 712	sock_recv_drops(msg, sk, skb);
 713}
 714EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
 715
 716static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
 717				       struct msghdr *msg, size_t size, int flags)
 718{
 719	struct sock_iocb *si = kiocb_to_siocb(iocb);
 720
 721	sock_update_classid(sock->sk);
 722
 723	si->sock = sock;
 724	si->scm = NULL;
 725	si->msg = msg;
 726	si->size = size;
 727	si->flags = flags;
 728
 729	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
 730}
 731
 732static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
 733				 struct msghdr *msg, size_t size, int flags)
 734{
 735	int err = security_socket_recvmsg(sock, msg, size, flags);
 736
 737	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
 738}
 739
 740int sock_recvmsg(struct socket *sock, struct msghdr *msg,
 741		 size_t size, int flags)
 742{
 743	struct kiocb iocb;
 744	struct sock_iocb siocb;
 745	int ret;
 746
 747	init_sync_kiocb(&iocb, NULL);
 748	iocb.private = &siocb;
 749	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
 750	if (-EIOCBQUEUED == ret)
 751		ret = wait_on_sync_kiocb(&iocb);
 752	return ret;
 753}
 754EXPORT_SYMBOL(sock_recvmsg);
 755
 756static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
 757			      size_t size, int flags)
 758{
 759	struct kiocb iocb;
 760	struct sock_iocb siocb;
 761	int ret;
 762
 763	init_sync_kiocb(&iocb, NULL);
 764	iocb.private = &siocb;
 765	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
 766	if (-EIOCBQUEUED == ret)
 767		ret = wait_on_sync_kiocb(&iocb);
 768	return ret;
 769}
 770
 771/**
 772 * kernel_recvmsg - Receive a message from a socket (kernel space)
 773 * @sock:       The socket to receive the message from
 774 * @msg:        Received message
 775 * @vec:        Input s/g array for message data
 776 * @num:        Size of input s/g array
 777 * @size:       Number of bytes to read
 778 * @flags:      Message flags (MSG_DONTWAIT, etc...)
 779 *
 780 * On return the msg structure contains the scatter/gather array passed in the
 781 * vec argument. The array is modified so that it consists of the unfilled
 782 * portion of the original array.
 783 *
 784 * The returned value is the total number of bytes received, or an error.
 785 */
 786int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
 787		   struct kvec *vec, size_t num, size_t size, int flags)
 788{
 789	mm_segment_t oldfs = get_fs();
 790	int result;
 791
 792	set_fs(KERNEL_DS);
 793	/*
 794	 * the following is safe, since for compiler definitions of kvec and
 795	 * iovec are identical, yielding the same in-core layout and alignment
 796	 */
 797	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
 798	result = sock_recvmsg(sock, msg, size, flags);
 799	set_fs(oldfs);
 800	return result;
 801}
 802EXPORT_SYMBOL(kernel_recvmsg);
 803
 804static void sock_aio_dtor(struct kiocb *iocb)
 805{
 806	kfree(iocb->private);
 807}
 808
 809static ssize_t sock_sendpage(struct file *file, struct page *page,
 810			     int offset, size_t size, loff_t *ppos, int more)
 811{
 812	struct socket *sock;
 813	int flags;
 814
 815	sock = file->private_data;
 816
 817	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 818	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
 819	flags |= more;
 820
 821	return kernel_sendpage(sock, page, offset, size, flags);
 822}
 823
 824static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
 825				struct pipe_inode_info *pipe, size_t len,
 826				unsigned int flags)
 827{
 828	struct socket *sock = file->private_data;
 829
 830	if (unlikely(!sock->ops->splice_read))
 831		return -EINVAL;
 832
 833	sock_update_classid(sock->sk);
 834
 835	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
 836}
 837
 838static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
 839					 struct sock_iocb *siocb)
 840{
 841	if (!is_sync_kiocb(iocb)) {
 842		siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
 843		if (!siocb)
 844			return NULL;
 845		iocb->ki_dtor = sock_aio_dtor;
 846	}
 847
 848	siocb->kiocb = iocb;
 849	iocb->private = siocb;
 850	return siocb;
 851}
 852
 853static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
 854		struct file *file, const struct iovec *iov,
 855		unsigned long nr_segs)
 856{
 857	struct socket *sock = file->private_data;
 858	size_t size = 0;
 859	int i;
 860
 861	for (i = 0; i < nr_segs; i++)
 862		size += iov[i].iov_len;
 863
 864	msg->msg_name = NULL;
 865	msg->msg_namelen = 0;
 866	msg->msg_control = NULL;
 867	msg->msg_controllen = 0;
 868	msg->msg_iov = (struct iovec *)iov;
 869	msg->msg_iovlen = nr_segs;
 870	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 871
 872	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
 873}
 874
 875static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
 876				unsigned long nr_segs, loff_t pos)
 877{
 878	struct sock_iocb siocb, *x;
 879
 880	if (pos != 0)
 881		return -ESPIPE;
 882
 883	if (iocb->ki_left == 0)	/* Match SYS5 behaviour */
 884		return 0;
 885
 886
 887	x = alloc_sock_iocb(iocb, &siocb);
 888	if (!x)
 889		return -ENOMEM;
 890	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 891}
 892
 893static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
 894			struct file *file, const struct iovec *iov,
 895			unsigned long nr_segs)
 896{
 897	struct socket *sock = file->private_data;
 898	size_t size = 0;
 899	int i;
 900
 901	for (i = 0; i < nr_segs; i++)
 902		size += iov[i].iov_len;
 903
 904	msg->msg_name = NULL;
 905	msg->msg_namelen = 0;
 906	msg->msg_control = NULL;
 907	msg->msg_controllen = 0;
 908	msg->msg_iov = (struct iovec *)iov;
 909	msg->msg_iovlen = nr_segs;
 910	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
 911	if (sock->type == SOCK_SEQPACKET)
 912		msg->msg_flags |= MSG_EOR;
 913
 914	return __sock_sendmsg(iocb, sock, msg, size);
 915}
 916
 917static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
 918			  unsigned long nr_segs, loff_t pos)
 919{
 920	struct sock_iocb siocb, *x;
 921
 922	if (pos != 0)
 923		return -ESPIPE;
 924
 925	x = alloc_sock_iocb(iocb, &siocb);
 926	if (!x)
 927		return -ENOMEM;
 928
 929	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
 930}
 931
 932/*
 933 * Atomic setting of ioctl hooks to avoid race
 934 * with module unload.
 935 */
 936
 937static DEFINE_MUTEX(br_ioctl_mutex);
 938static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
 939
 940void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
 941{
 942	mutex_lock(&br_ioctl_mutex);
 943	br_ioctl_hook = hook;
 944	mutex_unlock(&br_ioctl_mutex);
 945}
 946EXPORT_SYMBOL(brioctl_set);
 947
 948static DEFINE_MUTEX(vlan_ioctl_mutex);
 949static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
 950
 951void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
 952{
 953	mutex_lock(&vlan_ioctl_mutex);
 954	vlan_ioctl_hook = hook;
 955	mutex_unlock(&vlan_ioctl_mutex);
 956}
 957EXPORT_SYMBOL(vlan_ioctl_set);
 958
 959static DEFINE_MUTEX(dlci_ioctl_mutex);
 960static int (*dlci_ioctl_hook) (unsigned int, void __user *);
 961
 962void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
 963{
 964	mutex_lock(&dlci_ioctl_mutex);
 965	dlci_ioctl_hook = hook;
 966	mutex_unlock(&dlci_ioctl_mutex);
 967}
 968EXPORT_SYMBOL(dlci_ioctl_set);
 969
 970static long sock_do_ioctl(struct net *net, struct socket *sock,
 971				 unsigned int cmd, unsigned long arg)
 972{
 973	int err;
 974	void __user *argp = (void __user *)arg;
 975
 976	err = sock->ops->ioctl(sock, cmd, arg);
 977
 978	/*
 979	 * If this ioctl is unknown try to hand it down
 980	 * to the NIC driver.
 981	 */
 982	if (err == -ENOIOCTLCMD)
 983		err = dev_ioctl(net, cmd, argp);
 984
 985	return err;
 986}
 987
 988/*
 989 *	With an ioctl, arg may well be a user mode pointer, but we don't know
 990 *	what to do with it - that's up to the protocol still.
 991 */
 992
 993static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
 994{
 995	struct socket *sock;
 996	struct sock *sk;
 997	void __user *argp = (void __user *)arg;
 998	int pid, err;
 999	struct net *net;
1000
1001	sock = file->private_data;
1002	sk = sock->sk;
1003	net = sock_net(sk);
1004	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1005		err = dev_ioctl(net, cmd, argp);
1006	} else
1007#ifdef CONFIG_WEXT_CORE
1008	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1009		err = dev_ioctl(net, cmd, argp);
1010	} else
1011#endif
1012		switch (cmd) {
1013		case FIOSETOWN:
1014		case SIOCSPGRP:
1015			err = -EFAULT;
1016			if (get_user(pid, (int __user *)argp))
1017				break;
1018			err = f_setown(sock->file, pid, 1);
1019			break;
1020		case FIOGETOWN:
1021		case SIOCGPGRP:
1022			err = put_user(f_getown(sock->file),
1023				       (int __user *)argp);
1024			break;
1025		case SIOCGIFBR:
1026		case SIOCSIFBR:
1027		case SIOCBRADDBR:
1028		case SIOCBRDELBR:
1029			err = -ENOPKG;
1030			if (!br_ioctl_hook)
1031				request_module("bridge");
1032
1033			mutex_lock(&br_ioctl_mutex);
1034			if (br_ioctl_hook)
1035				err = br_ioctl_hook(net, cmd, argp);
1036			mutex_unlock(&br_ioctl_mutex);
1037			break;
1038		case SIOCGIFVLAN:
1039		case SIOCSIFVLAN:
1040			err = -ENOPKG;
1041			if (!vlan_ioctl_hook)
1042				request_module("8021q");
1043
1044			mutex_lock(&vlan_ioctl_mutex);
1045			if (vlan_ioctl_hook)
1046				err = vlan_ioctl_hook(net, argp);
1047			mutex_unlock(&vlan_ioctl_mutex);
1048			break;
1049		case SIOCADDDLCI:
1050		case SIOCDELDLCI:
1051			err = -ENOPKG;
1052			if (!dlci_ioctl_hook)
1053				request_module("dlci");
1054
1055			mutex_lock(&dlci_ioctl_mutex);
1056			if (dlci_ioctl_hook)
1057				err = dlci_ioctl_hook(cmd, argp);
1058			mutex_unlock(&dlci_ioctl_mutex);
1059			break;
1060		default:
1061			err = sock_do_ioctl(net, sock, cmd, arg);
1062			break;
1063		}
1064	return err;
1065}
1066
1067int sock_create_lite(int family, int type, int protocol, struct socket **res)
1068{
1069	int err;
1070	struct socket *sock = NULL;
1071
1072	err = security_socket_create(family, type, protocol, 1);
1073	if (err)
1074		goto out;
1075
1076	sock = sock_alloc();
1077	if (!sock) {
1078		err = -ENOMEM;
1079		goto out;
1080	}
1081
1082	sock->type = type;
1083	err = security_socket_post_create(sock, family, type, protocol, 1);
1084	if (err)
1085		goto out_release;
1086
1087out:
1088	*res = sock;
1089	return err;
1090out_release:
1091	sock_release(sock);
1092	sock = NULL;
1093	goto out;
1094}
1095EXPORT_SYMBOL(sock_create_lite);
1096
1097/* No kernel lock held - perfect */
1098static unsigned int sock_poll(struct file *file, poll_table *wait)
1099{
1100	struct socket *sock;
1101
1102	/*
1103	 *      We can't return errors to poll, so it's either yes or no.
1104	 */
1105	sock = file->private_data;
1106	return sock->ops->poll(file, sock, wait);
1107}
1108
1109static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1110{
1111	struct socket *sock = file->private_data;
1112
1113	return sock->ops->mmap(file, sock, vma);
1114}
1115
1116static int sock_close(struct inode *inode, struct file *filp)
1117{
1118	/*
1119	 *      It was possible the inode is NULL we were
1120	 *      closing an unfinished socket.
1121	 */
1122
1123	if (!inode) {
1124		printk(KERN_DEBUG "sock_close: NULL inode\n");
1125		return 0;
1126	}
1127	sock_release(SOCKET_I(inode));
1128	return 0;
1129}
1130
1131/*
1132 *	Update the socket async list
1133 *
1134 *	Fasync_list locking strategy.
1135 *
1136 *	1. fasync_list is modified only under process context socket lock
1137 *	   i.e. under semaphore.
1138 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1139 *	   or under socket lock
1140 */
1141
1142static int sock_fasync(int fd, struct file *filp, int on)
1143{
1144	struct socket *sock = filp->private_data;
1145	struct sock *sk = sock->sk;
1146	struct socket_wq *wq;
1147
1148	if (sk == NULL)
1149		return -EINVAL;
1150
1151	lock_sock(sk);
1152	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1153	fasync_helper(fd, filp, on, &wq->fasync_list);
1154
1155	if (!wq->fasync_list)
1156		sock_reset_flag(sk, SOCK_FASYNC);
1157	else
1158		sock_set_flag(sk, SOCK_FASYNC);
1159
1160	release_sock(sk);
1161	return 0;
1162}
1163
1164/* This function may be called only under socket lock or callback_lock or rcu_lock */
1165
1166int sock_wake_async(struct socket *sock, int how, int band)
1167{
1168	struct socket_wq *wq;
1169
1170	if (!sock)
1171		return -1;
1172	rcu_read_lock();
1173	wq = rcu_dereference(sock->wq);
1174	if (!wq || !wq->fasync_list) {
1175		rcu_read_unlock();
1176		return -1;
1177	}
1178	switch (how) {
1179	case SOCK_WAKE_WAITD:
1180		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1181			break;
1182		goto call_kill;
1183	case SOCK_WAKE_SPACE:
1184		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1185			break;
1186		/* fall through */
1187	case SOCK_WAKE_IO:
1188call_kill:
1189		kill_fasync(&wq->fasync_list, SIGIO, band);
1190		break;
1191	case SOCK_WAKE_URG:
1192		kill_fasync(&wq->fasync_list, SIGURG, band);
1193	}
1194	rcu_read_unlock();
1195	return 0;
1196}
1197EXPORT_SYMBOL(sock_wake_async);
1198
1199int __sock_create(struct net *net, int family, int type, int protocol,
1200			 struct socket **res, int kern)
1201{
1202	int err;
1203	struct socket *sock;
1204	const struct net_proto_family *pf;
1205
1206	/*
1207	 *      Check protocol is in range
1208	 */
1209	if (family < 0 || family >= NPROTO)
1210		return -EAFNOSUPPORT;
1211	if (type < 0 || type >= SOCK_MAX)
1212		return -EINVAL;
1213
1214	/* Compatibility.
1215
1216	   This uglymoron is moved from INET layer to here to avoid
1217	   deadlock in module load.
1218	 */
1219	if (family == PF_INET && type == SOCK_PACKET) {
1220		static int warned;
1221		if (!warned) {
1222			warned = 1;
1223			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1224			       current->comm);
1225		}
1226		family = PF_PACKET;
1227	}
1228
1229	err = security_socket_create(family, type, protocol, kern);
1230	if (err)
1231		return err;
1232
1233	/*
1234	 *	Allocate the socket and allow the family to set things up. if
1235	 *	the protocol is 0, the family is instructed to select an appropriate
1236	 *	default.
1237	 */
1238	sock = sock_alloc();
1239	if (!sock) {
1240		if (net_ratelimit())
1241			printk(KERN_WARNING "socket: no more sockets\n");
1242		return -ENFILE;	/* Not exactly a match, but its the
1243				   closest posix thing */
1244	}
1245
1246	sock->type = type;
1247
1248#ifdef CONFIG_MODULES
1249	/* Attempt to load a protocol module if the find failed.
1250	 *
1251	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1252	 * requested real, full-featured networking support upon configuration.
1253	 * Otherwise module support will break!
1254	 */
1255	if (rcu_access_pointer(net_families[family]) == NULL)
1256		request_module("net-pf-%d", family);
1257#endif
1258
1259	rcu_read_lock();
1260	pf = rcu_dereference(net_families[family]);
1261	err = -EAFNOSUPPORT;
1262	if (!pf)
1263		goto out_release;
1264
1265	/*
1266	 * We will call the ->create function, that possibly is in a loadable
1267	 * module, so we have to bump that loadable module refcnt first.
1268	 */
1269	if (!try_module_get(pf->owner))
1270		goto out_release;
1271
1272	/* Now protected by module ref count */
1273	rcu_read_unlock();
1274
1275	err = pf->create(net, sock, protocol, kern);
1276	if (err < 0)
1277		goto out_module_put;
1278
1279	/*
1280	 * Now to bump the refcnt of the [loadable] module that owns this
1281	 * socket at sock_release time we decrement its refcnt.
1282	 */
1283	if (!try_module_get(sock->ops->owner))
1284		goto out_module_busy;
1285
1286	/*
1287	 * Now that we're done with the ->create function, the [loadable]
1288	 * module can have its refcnt decremented
1289	 */
1290	module_put(pf->owner);
1291	err = security_socket_post_create(sock, family, type, protocol, kern);
1292	if (err)
1293		goto out_sock_release;
1294	*res = sock;
1295
1296	return 0;
1297
1298out_module_busy:
1299	err = -EAFNOSUPPORT;
1300out_module_put:
1301	sock->ops = NULL;
1302	module_put(pf->owner);
1303out_sock_release:
1304	sock_release(sock);
1305	return err;
1306
1307out_release:
1308	rcu_read_unlock();
1309	goto out_sock_release;
1310}
1311EXPORT_SYMBOL(__sock_create);
1312
1313int sock_create(int family, int type, int protocol, struct socket **res)
1314{
1315	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1316}
1317EXPORT_SYMBOL(sock_create);
1318
1319int sock_create_kern(int family, int type, int protocol, struct socket **res)
1320{
1321	return __sock_create(&init_net, family, type, protocol, res, 1);
1322}
1323EXPORT_SYMBOL(sock_create_kern);
1324
1325SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1326{
1327	int retval;
1328	struct socket *sock;
1329	int flags;
1330
1331	/* Check the SOCK_* constants for consistency.  */
1332	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1333	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1334	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1335	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1336
1337	flags = type & ~SOCK_TYPE_MASK;
1338	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1339		return -EINVAL;
1340	type &= SOCK_TYPE_MASK;
1341
1342	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1343		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1344
1345	retval = sock_create(family, type, protocol, &sock);
1346	if (retval < 0)
1347		goto out;
1348
1349	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1350	if (retval < 0)
1351		goto out_release;
1352
1353out:
1354	/* It may be already another descriptor 8) Not kernel problem. */
1355	return retval;
1356
1357out_release:
1358	sock_release(sock);
1359	return retval;
1360}
1361
1362/*
1363 *	Create a pair of connected sockets.
1364 */
1365
1366SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1367		int __user *, usockvec)
1368{
1369	struct socket *sock1, *sock2;
1370	int fd1, fd2, err;
1371	struct file *newfile1, *newfile2;
1372	int flags;
1373
1374	flags = type & ~SOCK_TYPE_MASK;
1375	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1376		return -EINVAL;
1377	type &= SOCK_TYPE_MASK;
1378
1379	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1380		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1381
1382	/*
1383	 * Obtain the first socket and check if the underlying protocol
1384	 * supports the socketpair call.
1385	 */
1386
1387	err = sock_create(family, type, protocol, &sock1);
1388	if (err < 0)
1389		goto out;
1390
1391	err = sock_create(family, type, protocol, &sock2);
1392	if (err < 0)
1393		goto out_release_1;
1394
1395	err = sock1->ops->socketpair(sock1, sock2);
1396	if (err < 0)
1397		goto out_release_both;
1398
1399	fd1 = sock_alloc_file(sock1, &newfile1, flags);
1400	if (unlikely(fd1 < 0)) {
1401		err = fd1;
1402		goto out_release_both;
1403	}
1404
1405	fd2 = sock_alloc_file(sock2, &newfile2, flags);
1406	if (unlikely(fd2 < 0)) {
1407		err = fd2;
1408		fput(newfile1);
1409		put_unused_fd(fd1);
1410		sock_release(sock2);
1411		goto out;
1412	}
1413
1414	audit_fd_pair(fd1, fd2);
1415	fd_install(fd1, newfile1);
1416	fd_install(fd2, newfile2);
1417	/* fd1 and fd2 may be already another descriptors.
1418	 * Not kernel problem.
1419	 */
1420
1421	err = put_user(fd1, &usockvec[0]);
1422	if (!err)
1423		err = put_user(fd2, &usockvec[1]);
1424	if (!err)
1425		return 0;
1426
1427	sys_close(fd2);
1428	sys_close(fd1);
1429	return err;
1430
1431out_release_both:
1432	sock_release(sock2);
1433out_release_1:
1434	sock_release(sock1);
1435out:
1436	return err;
1437}
1438
1439/*
1440 *	Bind a name to a socket. Nothing much to do here since it's
1441 *	the protocol's responsibility to handle the local address.
1442 *
1443 *	We move the socket address to kernel space before we call
1444 *	the protocol layer (having also checked the address is ok).
1445 */
1446
1447SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1448{
1449	struct socket *sock;
1450	struct sockaddr_storage address;
1451	int err, fput_needed;
1452
1453	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1454	if (sock) {
1455		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1456		if (err >= 0) {
1457			err = security_socket_bind(sock,
1458						   (struct sockaddr *)&address,
1459						   addrlen);
1460			if (!err)
1461				err = sock->ops->bind(sock,
1462						      (struct sockaddr *)
1463						      &address, addrlen);
1464		}
1465		fput_light(sock->file, fput_needed);
1466	}
1467	return err;
1468}
1469
1470/*
1471 *	Perform a listen. Basically, we allow the protocol to do anything
1472 *	necessary for a listen, and if that works, we mark the socket as
1473 *	ready for listening.
1474 */
1475
1476SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1477{
1478	struct socket *sock;
1479	int err, fput_needed;
1480	int somaxconn;
1481
1482	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1483	if (sock) {
1484		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1485		if ((unsigned)backlog > somaxconn)
1486			backlog = somaxconn;
1487
1488		err = security_socket_listen(sock, backlog);
1489		if (!err)
1490			err = sock->ops->listen(sock, backlog);
1491
1492		fput_light(sock->file, fput_needed);
1493	}
1494	return err;
1495}
1496
1497/*
1498 *	For accept, we attempt to create a new socket, set up the link
1499 *	with the client, wake up the client, then return the new
1500 *	connected fd. We collect the address of the connector in kernel
1501 *	space and move it to user at the very end. This is unclean because
1502 *	we open the socket then return an error.
1503 *
1504 *	1003.1g adds the ability to recvmsg() to query connection pending
1505 *	status to recvmsg. We need to add that support in a way thats
1506 *	clean when we restucture accept also.
1507 */
1508
1509SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1510		int __user *, upeer_addrlen, int, flags)
1511{
1512	struct socket *sock, *newsock;
1513	struct file *newfile;
1514	int err, len, newfd, fput_needed;
1515	struct sockaddr_storage address;
1516
1517	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1518		return -EINVAL;
1519
1520	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1521		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1522
1523	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1524	if (!sock)
1525		goto out;
1526
1527	err = -ENFILE;
1528	newsock = sock_alloc();
1529	if (!newsock)
1530		goto out_put;
1531
1532	newsock->type = sock->type;
1533	newsock->ops = sock->ops;
1534
1535	/*
1536	 * We don't need try_module_get here, as the listening socket (sock)
1537	 * has the protocol module (sock->ops->owner) held.
1538	 */
1539	__module_get(newsock->ops->owner);
1540
1541	newfd = sock_alloc_file(newsock, &newfile, flags);
1542	if (unlikely(newfd < 0)) {
1543		err = newfd;
1544		sock_release(newsock);
1545		goto out_put;
1546	}
1547
1548	err = security_socket_accept(sock, newsock);
1549	if (err)
1550		goto out_fd;
1551
1552	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1553	if (err < 0)
1554		goto out_fd;
1555
1556	if (upeer_sockaddr) {
1557		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1558					  &len, 2) < 0) {
1559			err = -ECONNABORTED;
1560			goto out_fd;
1561		}
1562		err = move_addr_to_user(&address,
1563					len, upeer_sockaddr, upeer_addrlen);
1564		if (err < 0)
1565			goto out_fd;
1566	}
1567
1568	/* File flags are not inherited via accept() unlike another OSes. */
1569
1570	fd_install(newfd, newfile);
1571	err = newfd;
1572
1573out_put:
1574	fput_light(sock->file, fput_needed);
1575out:
1576	return err;
1577out_fd:
1578	fput(newfile);
1579	put_unused_fd(newfd);
1580	goto out_put;
1581}
1582
1583SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1584		int __user *, upeer_addrlen)
1585{
1586	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1587}
1588
1589/*
1590 *	Attempt to connect to a socket with the server address.  The address
1591 *	is in user space so we verify it is OK and move it to kernel space.
1592 *
1593 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1594 *	break bindings
1595 *
1596 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1597 *	other SEQPACKET protocols that take time to connect() as it doesn't
1598 *	include the -EINPROGRESS status for such sockets.
1599 */
1600
1601SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1602		int, addrlen)
1603{
1604	struct socket *sock;
1605	struct sockaddr_storage address;
1606	int err, fput_needed;
1607
1608	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1609	if (!sock)
1610		goto out;
1611	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1612	if (err < 0)
1613		goto out_put;
1614
1615	err =
1616	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1617	if (err)
1618		goto out_put;
1619
1620	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1621				 sock->file->f_flags);
1622out_put:
1623	fput_light(sock->file, fput_needed);
1624out:
1625	return err;
1626}
1627
1628/*
1629 *	Get the local address ('name') of a socket object. Move the obtained
1630 *	name to user space.
1631 */
1632
1633SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1634		int __user *, usockaddr_len)
1635{
1636	struct socket *sock;
1637	struct sockaddr_storage address;
1638	int len, err, fput_needed;
1639
1640	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1641	if (!sock)
1642		goto out;
1643
1644	err = security_socket_getsockname(sock);
1645	if (err)
1646		goto out_put;
1647
1648	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1649	if (err)
1650		goto out_put;
1651	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1652
1653out_put:
1654	fput_light(sock->file, fput_needed);
1655out:
1656	return err;
1657}
1658
1659/*
1660 *	Get the remote address ('name') of a socket object. Move the obtained
1661 *	name to user space.
1662 */
1663
1664SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1665		int __user *, usockaddr_len)
1666{
1667	struct socket *sock;
1668	struct sockaddr_storage address;
1669	int len, err, fput_needed;
1670
1671	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1672	if (sock != NULL) {
1673		err = security_socket_getpeername(sock);
1674		if (err) {
1675			fput_light(sock->file, fput_needed);
1676			return err;
1677		}
1678
1679		err =
1680		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1681				       1);
1682		if (!err)
1683			err = move_addr_to_user(&address, len, usockaddr,
1684						usockaddr_len);
1685		fput_light(sock->file, fput_needed);
1686	}
1687	return err;
1688}
1689
1690/*
1691 *	Send a datagram to a given address. We move the address into kernel
1692 *	space and check the user space data area is readable before invoking
1693 *	the protocol.
1694 */
1695
1696SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1697		unsigned, flags, struct sockaddr __user *, addr,
1698		int, addr_len)
1699{
1700	struct socket *sock;
1701	struct sockaddr_storage address;
1702	int err;
1703	struct msghdr msg;
1704	struct iovec iov;
1705	int fput_needed;
1706
1707	if (len > INT_MAX)
1708		len = INT_MAX;
1709	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1710	if (!sock)
1711		goto out;
1712
1713	iov.iov_base = buff;
1714	iov.iov_len = len;
1715	msg.msg_name = NULL;
1716	msg.msg_iov = &iov;
1717	msg.msg_iovlen = 1;
1718	msg.msg_control = NULL;
1719	msg.msg_controllen = 0;
1720	msg.msg_namelen = 0;
1721	if (addr) {
1722		err = move_addr_to_kernel(addr, addr_len, &address);
1723		if (err < 0)
1724			goto out_put;
1725		msg.msg_name = (struct sockaddr *)&address;
1726		msg.msg_namelen = addr_len;
1727	}
1728	if (sock->file->f_flags & O_NONBLOCK)
1729		flags |= MSG_DONTWAIT;
1730	msg.msg_flags = flags;
1731	err = sock_sendmsg(sock, &msg, len);
1732
1733out_put:
1734	fput_light(sock->file, fput_needed);
1735out:
1736	return err;
1737}
1738
1739/*
1740 *	Send a datagram down a socket.
1741 */
1742
1743SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1744		unsigned, flags)
1745{
1746	return sys_sendto(fd, buff, len, flags, NULL, 0);
1747}
1748
1749/*
1750 *	Receive a frame from the socket and optionally record the address of the
1751 *	sender. We verify the buffers are writable and if needed move the
1752 *	sender address from kernel to user space.
1753 */
1754
1755SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1756		unsigned, flags, struct sockaddr __user *, addr,
1757		int __user *, addr_len)
1758{
1759	struct socket *sock;
1760	struct iovec iov;
1761	struct msghdr msg;
1762	struct sockaddr_storage address;
1763	int err, err2;
1764	int fput_needed;
1765
1766	if (size > INT_MAX)
1767		size = INT_MAX;
1768	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769	if (!sock)
1770		goto out;
1771
1772	msg.msg_control = NULL;
1773	msg.msg_controllen = 0;
1774	msg.msg_iovlen = 1;
1775	msg.msg_iov = &iov;
1776	iov.iov_len = size;
1777	iov.iov_base = ubuf;
1778	msg.msg_name = (struct sockaddr *)&address;
1779	msg.msg_namelen = sizeof(address);
1780	if (sock->file->f_flags & O_NONBLOCK)
1781		flags |= MSG_DONTWAIT;
1782	err = sock_recvmsg(sock, &msg, size, flags);
1783
1784	if (err >= 0 && addr != NULL) {
1785		err2 = move_addr_to_user(&address,
1786					 msg.msg_namelen, addr, addr_len);
1787		if (err2 < 0)
1788			err = err2;
1789	}
1790
1791	fput_light(sock->file, fput_needed);
1792out:
1793	return err;
1794}
1795
1796/*
1797 *	Receive a datagram from a socket.
1798 */
1799
1800asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1801			 unsigned flags)
1802{
1803	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1804}
1805
1806/*
1807 *	Set a socket option. Because we don't know the option lengths we have
1808 *	to pass the user mode parameter for the protocols to sort out.
1809 */
1810
1811SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1812		char __user *, optval, int, optlen)
1813{
1814	int err, fput_needed;
1815	struct socket *sock;
1816
1817	if (optlen < 0)
1818		return -EINVAL;
1819
1820	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1821	if (sock != NULL) {
1822		err = security_socket_setsockopt(sock, level, optname);
1823		if (err)
1824			goto out_put;
1825
1826		if (level == SOL_SOCKET)
1827			err =
1828			    sock_setsockopt(sock, level, optname, optval,
1829					    optlen);
1830		else
1831			err =
1832			    sock->ops->setsockopt(sock, level, optname, optval,
1833						  optlen);
1834out_put:
1835		fput_light(sock->file, fput_needed);
1836	}
1837	return err;
1838}
1839
1840/*
1841 *	Get a socket option. Because we don't know the option lengths we have
1842 *	to pass a user mode parameter for the protocols to sort out.
1843 */
1844
1845SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1846		char __user *, optval, int __user *, optlen)
1847{
1848	int err, fput_needed;
1849	struct socket *sock;
1850
1851	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1852	if (sock != NULL) {
1853		err = security_socket_getsockopt(sock, level, optname);
1854		if (err)
1855			goto out_put;
1856
1857		if (level == SOL_SOCKET)
1858			err =
1859			    sock_getsockopt(sock, level, optname, optval,
1860					    optlen);
1861		else
1862			err =
1863			    sock->ops->getsockopt(sock, level, optname, optval,
1864						  optlen);
1865out_put:
1866		fput_light(sock->file, fput_needed);
1867	}
1868	return err;
1869}
1870
1871/*
1872 *	Shutdown a socket.
1873 */
1874
1875SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1876{
1877	int err, fput_needed;
1878	struct socket *sock;
1879
1880	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1881	if (sock != NULL) {
1882		err = security_socket_shutdown(sock, how);
1883		if (!err)
1884			err = sock->ops->shutdown(sock, how);
1885		fput_light(sock->file, fput_needed);
1886	}
1887	return err;
1888}
1889
1890/* A couple of helpful macros for getting the address of the 32/64 bit
1891 * fields which are the same type (int / unsigned) on our platforms.
1892 */
1893#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1894#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1895#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1896
1897struct used_address {
1898	struct sockaddr_storage name;
1899	unsigned int name_len;
1900};
1901
1902static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1903			 struct msghdr *msg_sys, unsigned flags,
1904			 struct used_address *used_address)
1905{
1906	struct compat_msghdr __user *msg_compat =
1907	    (struct compat_msghdr __user *)msg;
1908	struct sockaddr_storage address;
1909	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1910	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1911	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1912	/* 20 is size of ipv6_pktinfo */
1913	unsigned char *ctl_buf = ctl;
1914	int err, ctl_len, iov_size, total_len;
1915
1916	err = -EFAULT;
1917	if (MSG_CMSG_COMPAT & flags) {
1918		if (get_compat_msghdr(msg_sys, msg_compat))
1919			return -EFAULT;
1920	} else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
1921		return -EFAULT;
1922
1923	/* do not move before msg_sys is valid */
1924	err = -EMSGSIZE;
1925	if (msg_sys->msg_iovlen > UIO_MAXIOV)
1926		goto out;
1927
1928	/* Check whether to allocate the iovec area */
1929	err = -ENOMEM;
1930	iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1931	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1932		iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1933		if (!iov)
1934			goto out;
1935	}
1936
1937	/* This will also move the address data into kernel space */
1938	if (MSG_CMSG_COMPAT & flags) {
1939		err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
1940	} else
1941		err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
1942	if (err < 0)
1943		goto out_freeiov;
1944	total_len = err;
1945
1946	err = -ENOBUFS;
1947
1948	if (msg_sys->msg_controllen > INT_MAX)
1949		goto out_freeiov;
1950	ctl_len = msg_sys->msg_controllen;
1951	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1952		err =
1953		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1954						     sizeof(ctl));
1955		if (err)
1956			goto out_freeiov;
1957		ctl_buf = msg_sys->msg_control;
1958		ctl_len = msg_sys->msg_controllen;
1959	} else if (ctl_len) {
1960		if (ctl_len > sizeof(ctl)) {
1961			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1962			if (ctl_buf == NULL)
1963				goto out_freeiov;
1964		}
1965		err = -EFAULT;
1966		/*
1967		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1968		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1969		 * checking falls down on this.
1970		 */
1971		if (copy_from_user(ctl_buf,
1972				   (void __user __force *)msg_sys->msg_control,
1973				   ctl_len))
1974			goto out_freectl;
1975		msg_sys->msg_control = ctl_buf;
1976	}
1977	msg_sys->msg_flags = flags;
1978
1979	if (sock->file->f_flags & O_NONBLOCK)
1980		msg_sys->msg_flags |= MSG_DONTWAIT;
1981	/*
1982	 * If this is sendmmsg() and current destination address is same as
1983	 * previously succeeded address, omit asking LSM's decision.
1984	 * used_address->name_len is initialized to UINT_MAX so that the first
1985	 * destination address never matches.
1986	 */
1987	if (used_address && msg_sys->msg_name &&
1988	    used_address->name_len == msg_sys->msg_namelen &&
1989	    !memcmp(&used_address->name, msg_sys->msg_name,
1990		    used_address->name_len)) {
1991		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1992		goto out_freectl;
1993	}
1994	err = sock_sendmsg(sock, msg_sys, total_len);
1995	/*
1996	 * If this is sendmmsg() and sending to current destination address was
1997	 * successful, remember it.
1998	 */
1999	if (used_address && err >= 0) {
2000		used_address->name_len = msg_sys->msg_namelen;
2001		if (msg_sys->msg_name)
2002			memcpy(&used_address->name, msg_sys->msg_name,
2003			       used_address->name_len);
2004	}
2005
2006out_freectl:
2007	if (ctl_buf != ctl)
2008		sock_kfree_s(sock->sk, ct

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