xref: /openbmc/linux/net/socket.c (revision 12eb4683)
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/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/mount.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/compat.h>
84 #include <linux/kmod.h>
85 #include <linux/audit.h>
86 #include <linux/wireless.h>
87 #include <linux/nsproxy.h>
88 #include <linux/magic.h>
89 #include <linux/slab.h>
90 #include <linux/xattr.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 #include <net/busy_poll.h>
108 
109 #ifdef CONFIG_NET_RX_BUSY_POLL
110 unsigned int sysctl_net_busy_read __read_mostly;
111 unsigned int sysctl_net_busy_poll __read_mostly;
112 #endif
113 
114 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
115 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
116 			 unsigned long nr_segs, loff_t pos);
117 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
118 			  unsigned long nr_segs, loff_t pos);
119 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
120 
121 static int sock_close(struct inode *inode, struct file *file);
122 static unsigned int sock_poll(struct file *file,
123 			      struct poll_table_struct *wait);
124 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
125 #ifdef CONFIG_COMPAT
126 static long compat_sock_ioctl(struct file *file,
127 			      unsigned int cmd, unsigned long arg);
128 #endif
129 static int sock_fasync(int fd, struct file *filp, int on);
130 static ssize_t sock_sendpage(struct file *file, struct page *page,
131 			     int offset, size_t size, loff_t *ppos, int more);
132 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
133 				struct pipe_inode_info *pipe, size_t len,
134 				unsigned int flags);
135 
136 /*
137  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138  *	in the operation structures but are done directly via the socketcall() multiplexor.
139  */
140 
141 static const struct file_operations socket_file_ops = {
142 	.owner =	THIS_MODULE,
143 	.llseek =	no_llseek,
144 	.aio_read =	sock_aio_read,
145 	.aio_write =	sock_aio_write,
146 	.poll =		sock_poll,
147 	.unlocked_ioctl = sock_ioctl,
148 #ifdef CONFIG_COMPAT
149 	.compat_ioctl = compat_sock_ioctl,
150 #endif
151 	.mmap =		sock_mmap,
152 	.open =		sock_no_open,	/* special open code to disallow open via /proc */
153 	.release =	sock_close,
154 	.fasync =	sock_fasync,
155 	.sendpage =	sock_sendpage,
156 	.splice_write = generic_splice_sendpage,
157 	.splice_read =	sock_splice_read,
158 };
159 
160 /*
161  *	The protocol list. Each protocol is registered in here.
162  */
163 
164 static DEFINE_SPINLOCK(net_family_lock);
165 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
166 
167 /*
168  *	Statistics counters of the socket lists
169  */
170 
171 static DEFINE_PER_CPU(int, sockets_in_use);
172 
173 /*
174  * Support routines.
175  * Move socket addresses back and forth across the kernel/user
176  * divide and look after the messy bits.
177  */
178 
179 /**
180  *	move_addr_to_kernel	-	copy a socket address into kernel space
181  *	@uaddr: Address in user space
182  *	@kaddr: Address in kernel space
183  *	@ulen: Length in user space
184  *
185  *	The address is copied into kernel space. If the provided address is
186  *	too long an error code of -EINVAL is returned. If the copy gives
187  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
188  */
189 
190 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
191 {
192 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
193 		return -EINVAL;
194 	if (ulen == 0)
195 		return 0;
196 	if (copy_from_user(kaddr, uaddr, ulen))
197 		return -EFAULT;
198 	return audit_sockaddr(ulen, kaddr);
199 }
200 
201 /**
202  *	move_addr_to_user	-	copy an address to user space
203  *	@kaddr: kernel space address
204  *	@klen: length of address in kernel
205  *	@uaddr: user space address
206  *	@ulen: pointer to user length field
207  *
208  *	The value pointed to by ulen on entry is the buffer length available.
209  *	This is overwritten with the buffer space used. -EINVAL is returned
210  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
211  *	is returned if either the buffer or the length field are not
212  *	accessible.
213  *	After copying the data up to the limit the user specifies, the true
214  *	length of the data is written over the length limit the user
215  *	specified. Zero is returned for a success.
216  */
217 
218 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
219 			     void __user *uaddr, int __user *ulen)
220 {
221 	int err;
222 	int len;
223 
224 	BUG_ON(klen > sizeof(struct sockaddr_storage));
225 	err = get_user(len, ulen);
226 	if (err)
227 		return err;
228 	if (len > klen)
229 		len = klen;
230 	if (len < 0)
231 		return -EINVAL;
232 	if (len) {
233 		if (audit_sockaddr(klen, kaddr))
234 			return -ENOMEM;
235 		if (copy_to_user(uaddr, kaddr, len))
236 			return -EFAULT;
237 	}
238 	/*
239 	 *      "fromlen shall refer to the value before truncation.."
240 	 *                      1003.1g
241 	 */
242 	return __put_user(klen, ulen);
243 }
244 
245 static struct kmem_cache *sock_inode_cachep __read_mostly;
246 
247 static struct inode *sock_alloc_inode(struct super_block *sb)
248 {
249 	struct socket_alloc *ei;
250 	struct socket_wq *wq;
251 
252 	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
253 	if (!ei)
254 		return NULL;
255 	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256 	if (!wq) {
257 		kmem_cache_free(sock_inode_cachep, ei);
258 		return NULL;
259 	}
260 	init_waitqueue_head(&wq->wait);
261 	wq->fasync_list = NULL;
262 	RCU_INIT_POINTER(ei->socket.wq, wq);
263 
264 	ei->socket.state = SS_UNCONNECTED;
265 	ei->socket.flags = 0;
266 	ei->socket.ops = NULL;
267 	ei->socket.sk = NULL;
268 	ei->socket.file = NULL;
269 
270 	return &ei->vfs_inode;
271 }
272 
273 static void sock_destroy_inode(struct inode *inode)
274 {
275 	struct socket_alloc *ei;
276 	struct socket_wq *wq;
277 
278 	ei = container_of(inode, struct socket_alloc, vfs_inode);
279 	wq = rcu_dereference_protected(ei->socket.wq, 1);
280 	kfree_rcu(wq, rcu);
281 	kmem_cache_free(sock_inode_cachep, ei);
282 }
283 
284 static void init_once(void *foo)
285 {
286 	struct socket_alloc *ei = (struct socket_alloc *)foo;
287 
288 	inode_init_once(&ei->vfs_inode);
289 }
290 
291 static int init_inodecache(void)
292 {
293 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
294 					      sizeof(struct socket_alloc),
295 					      0,
296 					      (SLAB_HWCACHE_ALIGN |
297 					       SLAB_RECLAIM_ACCOUNT |
298 					       SLAB_MEM_SPREAD),
299 					      init_once);
300 	if (sock_inode_cachep == NULL)
301 		return -ENOMEM;
302 	return 0;
303 }
304 
305 static const struct super_operations sockfs_ops = {
306 	.alloc_inode	= sock_alloc_inode,
307 	.destroy_inode	= sock_destroy_inode,
308 	.statfs		= simple_statfs,
309 };
310 
311 /*
312  * sockfs_dname() is called from d_path().
313  */
314 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315 {
316 	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 				dentry->d_inode->i_ino);
318 }
319 
320 static const struct dentry_operations sockfs_dentry_operations = {
321 	.d_dname  = sockfs_dname,
322 };
323 
324 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 			 int flags, const char *dev_name, void *data)
326 {
327 	return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 		&sockfs_dentry_operations, SOCKFS_MAGIC);
329 }
330 
331 static struct vfsmount *sock_mnt __read_mostly;
332 
333 static struct file_system_type sock_fs_type = {
334 	.name =		"sockfs",
335 	.mount =	sockfs_mount,
336 	.kill_sb =	kill_anon_super,
337 };
338 
339 /*
340  *	Obtains the first available file descriptor and sets it up for use.
341  *
342  *	These functions create file structures and maps them to fd space
343  *	of the current process. On success it returns file descriptor
344  *	and file struct implicitly stored in sock->file.
345  *	Note that another thread may close file descriptor before we return
346  *	from this function. We use the fact that now we do not refer
347  *	to socket after mapping. If one day we will need it, this
348  *	function will increment ref. count on file by 1.
349  *
350  *	In any case returned fd MAY BE not valid!
351  *	This race condition is unavoidable
352  *	with shared fd spaces, we cannot solve it inside kernel,
353  *	but we take care of internal coherence yet.
354  */
355 
356 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
357 {
358 	struct qstr name = { .name = "" };
359 	struct path path;
360 	struct file *file;
361 
362 	if (dname) {
363 		name.name = dname;
364 		name.len = strlen(name.name);
365 	} else if (sock->sk) {
366 		name.name = sock->sk->sk_prot_creator->name;
367 		name.len = strlen(name.name);
368 	}
369 	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
370 	if (unlikely(!path.dentry))
371 		return ERR_PTR(-ENOMEM);
372 	path.mnt = mntget(sock_mnt);
373 
374 	d_instantiate(path.dentry, SOCK_INODE(sock));
375 	SOCK_INODE(sock)->i_fop = &socket_file_ops;
376 
377 	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
378 		  &socket_file_ops);
379 	if (unlikely(IS_ERR(file))) {
380 		/* drop dentry, keep inode */
381 		ihold(path.dentry->d_inode);
382 		path_put(&path);
383 		return file;
384 	}
385 
386 	sock->file = file;
387 	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
388 	file->private_data = sock;
389 	return file;
390 }
391 EXPORT_SYMBOL(sock_alloc_file);
392 
393 static int sock_map_fd(struct socket *sock, int flags)
394 {
395 	struct file *newfile;
396 	int fd = get_unused_fd_flags(flags);
397 	if (unlikely(fd < 0))
398 		return fd;
399 
400 	newfile = sock_alloc_file(sock, flags, NULL);
401 	if (likely(!IS_ERR(newfile))) {
402 		fd_install(fd, newfile);
403 		return fd;
404 	}
405 
406 	put_unused_fd(fd);
407 	return PTR_ERR(newfile);
408 }
409 
410 struct socket *sock_from_file(struct file *file, int *err)
411 {
412 	if (file->f_op == &socket_file_ops)
413 		return file->private_data;	/* set in sock_map_fd */
414 
415 	*err = -ENOTSOCK;
416 	return NULL;
417 }
418 EXPORT_SYMBOL(sock_from_file);
419 
420 /**
421  *	sockfd_lookup - Go from a file number to its socket slot
422  *	@fd: file handle
423  *	@err: pointer to an error code return
424  *
425  *	The file handle passed in is locked and the socket it is bound
426  *	too is returned. If an error occurs the err pointer is overwritten
427  *	with a negative errno code and NULL is returned. The function checks
428  *	for both invalid handles and passing a handle which is not a socket.
429  *
430  *	On a success the socket object pointer is returned.
431  */
432 
433 struct socket *sockfd_lookup(int fd, int *err)
434 {
435 	struct file *file;
436 	struct socket *sock;
437 
438 	file = fget(fd);
439 	if (!file) {
440 		*err = -EBADF;
441 		return NULL;
442 	}
443 
444 	sock = sock_from_file(file, err);
445 	if (!sock)
446 		fput(file);
447 	return sock;
448 }
449 EXPORT_SYMBOL(sockfd_lookup);
450 
451 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452 {
453 	struct file *file;
454 	struct socket *sock;
455 
456 	*err = -EBADF;
457 	file = fget_light(fd, fput_needed);
458 	if (file) {
459 		sock = sock_from_file(file, err);
460 		if (sock)
461 			return sock;
462 		fput_light(file, *fput_needed);
463 	}
464 	return NULL;
465 }
466 
467 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470 static ssize_t sockfs_getxattr(struct dentry *dentry,
471 			       const char *name, void *value, size_t size)
472 {
473 	const char *proto_name;
474 	size_t proto_size;
475 	int error;
476 
477 	error = -ENODATA;
478 	if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 		proto_name = dentry->d_name.name;
480 		proto_size = strlen(proto_name);
481 
482 		if (value) {
483 			error = -ERANGE;
484 			if (proto_size + 1 > size)
485 				goto out;
486 
487 			strncpy(value, proto_name, proto_size + 1);
488 		}
489 		error = proto_size + 1;
490 	}
491 
492 out:
493 	return error;
494 }
495 
496 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
497 				size_t size)
498 {
499 	ssize_t len;
500 	ssize_t used = 0;
501 
502 	len = security_inode_listsecurity(dentry->d_inode, buffer, size);
503 	if (len < 0)
504 		return len;
505 	used += len;
506 	if (buffer) {
507 		if (size < used)
508 			return -ERANGE;
509 		buffer += len;
510 	}
511 
512 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
513 	used += len;
514 	if (buffer) {
515 		if (size < used)
516 			return -ERANGE;
517 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
518 		buffer += len;
519 	}
520 
521 	return used;
522 }
523 
524 static const struct inode_operations sockfs_inode_ops = {
525 	.getxattr = sockfs_getxattr,
526 	.listxattr = sockfs_listxattr,
527 };
528 
529 /**
530  *	sock_alloc	-	allocate a socket
531  *
532  *	Allocate a new inode and socket object. The two are bound together
533  *	and initialised. The socket is then returned. If we are out of inodes
534  *	NULL is returned.
535  */
536 
537 static struct socket *sock_alloc(void)
538 {
539 	struct inode *inode;
540 	struct socket *sock;
541 
542 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
543 	if (!inode)
544 		return NULL;
545 
546 	sock = SOCKET_I(inode);
547 
548 	kmemcheck_annotate_bitfield(sock, type);
549 	inode->i_ino = get_next_ino();
550 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
551 	inode->i_uid = current_fsuid();
552 	inode->i_gid = current_fsgid();
553 	inode->i_op = &sockfs_inode_ops;
554 
555 	this_cpu_add(sockets_in_use, 1);
556 	return sock;
557 }
558 
559 /*
560  *	In theory you can't get an open on this inode, but /proc provides
561  *	a back door. Remember to keep it shut otherwise you'll let the
562  *	creepy crawlies in.
563  */
564 
565 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
566 {
567 	return -ENXIO;
568 }
569 
570 const struct file_operations bad_sock_fops = {
571 	.owner = THIS_MODULE,
572 	.open = sock_no_open,
573 	.llseek = noop_llseek,
574 };
575 
576 /**
577  *	sock_release	-	close a socket
578  *	@sock: socket to close
579  *
580  *	The socket is released from the protocol stack if it has a release
581  *	callback, and the inode is then released if the socket is bound to
582  *	an inode not a file.
583  */
584 
585 void sock_release(struct socket *sock)
586 {
587 	if (sock->ops) {
588 		struct module *owner = sock->ops->owner;
589 
590 		sock->ops->release(sock);
591 		sock->ops = NULL;
592 		module_put(owner);
593 	}
594 
595 	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
596 		printk(KERN_ERR "sock_release: fasync list not empty!\n");
597 
598 	if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
599 		return;
600 
601 	this_cpu_sub(sockets_in_use, 1);
602 	if (!sock->file) {
603 		iput(SOCK_INODE(sock));
604 		return;
605 	}
606 	sock->file = NULL;
607 }
608 EXPORT_SYMBOL(sock_release);
609 
610 void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
611 {
612 	*tx_flags = 0;
613 	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
614 		*tx_flags |= SKBTX_HW_TSTAMP;
615 	if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
616 		*tx_flags |= SKBTX_SW_TSTAMP;
617 	if (sock_flag(sk, SOCK_WIFI_STATUS))
618 		*tx_flags |= SKBTX_WIFI_STATUS;
619 }
620 EXPORT_SYMBOL(sock_tx_timestamp);
621 
622 static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
623 				       struct msghdr *msg, size_t size)
624 {
625 	struct sock_iocb *si = kiocb_to_siocb(iocb);
626 
627 	si->sock = sock;
628 	si->scm = NULL;
629 	si->msg = msg;
630 	si->size = size;
631 
632 	return sock->ops->sendmsg(iocb, sock, msg, size);
633 }
634 
635 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
636 				 struct msghdr *msg, size_t size)
637 {
638 	int err = security_socket_sendmsg(sock, msg, size);
639 
640 	return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
641 }
642 
643 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
644 {
645 	struct kiocb iocb;
646 	struct sock_iocb siocb;
647 	int ret;
648 
649 	init_sync_kiocb(&iocb, NULL);
650 	iocb.private = &siocb;
651 	ret = __sock_sendmsg(&iocb, sock, msg, size);
652 	if (-EIOCBQUEUED == ret)
653 		ret = wait_on_sync_kiocb(&iocb);
654 	return ret;
655 }
656 EXPORT_SYMBOL(sock_sendmsg);
657 
658 static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
659 {
660 	struct kiocb iocb;
661 	struct sock_iocb siocb;
662 	int ret;
663 
664 	init_sync_kiocb(&iocb, NULL);
665 	iocb.private = &siocb;
666 	ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
667 	if (-EIOCBQUEUED == ret)
668 		ret = wait_on_sync_kiocb(&iocb);
669 	return ret;
670 }
671 
672 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
673 		   struct kvec *vec, size_t num, size_t size)
674 {
675 	mm_segment_t oldfs = get_fs();
676 	int result;
677 
678 	set_fs(KERNEL_DS);
679 	/*
680 	 * the following is safe, since for compiler definitions of kvec and
681 	 * iovec are identical, yielding the same in-core layout and alignment
682 	 */
683 	msg->msg_iov = (struct iovec *)vec;
684 	msg->msg_iovlen = num;
685 	result = sock_sendmsg(sock, msg, size);
686 	set_fs(oldfs);
687 	return result;
688 }
689 EXPORT_SYMBOL(kernel_sendmsg);
690 
691 /*
692  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
693  */
694 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
695 	struct sk_buff *skb)
696 {
697 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
698 	struct timespec ts[3];
699 	int empty = 1;
700 	struct skb_shared_hwtstamps *shhwtstamps =
701 		skb_hwtstamps(skb);
702 
703 	/* Race occurred between timestamp enabling and packet
704 	   receiving.  Fill in the current time for now. */
705 	if (need_software_tstamp && skb->tstamp.tv64 == 0)
706 		__net_timestamp(skb);
707 
708 	if (need_software_tstamp) {
709 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
710 			struct timeval tv;
711 			skb_get_timestamp(skb, &tv);
712 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
713 				 sizeof(tv), &tv);
714 		} else {
715 			skb_get_timestampns(skb, &ts[0]);
716 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
717 				 sizeof(ts[0]), &ts[0]);
718 		}
719 	}
720 
721 
722 	memset(ts, 0, sizeof(ts));
723 	if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
724 	    ktime_to_timespec_cond(skb->tstamp, ts + 0))
725 		empty = 0;
726 	if (shhwtstamps) {
727 		if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
728 		    ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
729 			empty = 0;
730 		if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
731 		    ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
732 			empty = 0;
733 	}
734 	if (!empty)
735 		put_cmsg(msg, SOL_SOCKET,
736 			 SCM_TIMESTAMPING, sizeof(ts), &ts);
737 }
738 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
739 
740 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
741 	struct sk_buff *skb)
742 {
743 	int ack;
744 
745 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
746 		return;
747 	if (!skb->wifi_acked_valid)
748 		return;
749 
750 	ack = skb->wifi_acked;
751 
752 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
753 }
754 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
755 
756 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
757 				   struct sk_buff *skb)
758 {
759 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
760 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
761 			sizeof(__u32), &skb->dropcount);
762 }
763 
764 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
765 	struct sk_buff *skb)
766 {
767 	sock_recv_timestamp(msg, sk, skb);
768 	sock_recv_drops(msg, sk, skb);
769 }
770 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
771 
772 static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
773 				       struct msghdr *msg, size_t size, int flags)
774 {
775 	struct sock_iocb *si = kiocb_to_siocb(iocb);
776 
777 	si->sock = sock;
778 	si->scm = NULL;
779 	si->msg = msg;
780 	si->size = size;
781 	si->flags = flags;
782 
783 	return sock->ops->recvmsg(iocb, sock, msg, size, flags);
784 }
785 
786 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
787 				 struct msghdr *msg, size_t size, int flags)
788 {
789 	int err = security_socket_recvmsg(sock, msg, size, flags);
790 
791 	return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
792 }
793 
794 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
795 		 size_t size, int flags)
796 {
797 	struct kiocb iocb;
798 	struct sock_iocb siocb;
799 	int ret;
800 
801 	init_sync_kiocb(&iocb, NULL);
802 	iocb.private = &siocb;
803 	ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
804 	if (-EIOCBQUEUED == ret)
805 		ret = wait_on_sync_kiocb(&iocb);
806 	return ret;
807 }
808 EXPORT_SYMBOL(sock_recvmsg);
809 
810 static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
811 			      size_t size, int flags)
812 {
813 	struct kiocb iocb;
814 	struct sock_iocb siocb;
815 	int ret;
816 
817 	init_sync_kiocb(&iocb, NULL);
818 	iocb.private = &siocb;
819 	ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
820 	if (-EIOCBQUEUED == ret)
821 		ret = wait_on_sync_kiocb(&iocb);
822 	return ret;
823 }
824 
825 /**
826  * kernel_recvmsg - Receive a message from a socket (kernel space)
827  * @sock:       The socket to receive the message from
828  * @msg:        Received message
829  * @vec:        Input s/g array for message data
830  * @num:        Size of input s/g array
831  * @size:       Number of bytes to read
832  * @flags:      Message flags (MSG_DONTWAIT, etc...)
833  *
834  * On return the msg structure contains the scatter/gather array passed in the
835  * vec argument. The array is modified so that it consists of the unfilled
836  * portion of the original array.
837  *
838  * The returned value is the total number of bytes received, or an error.
839  */
840 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
841 		   struct kvec *vec, size_t num, size_t size, int flags)
842 {
843 	mm_segment_t oldfs = get_fs();
844 	int result;
845 
846 	set_fs(KERNEL_DS);
847 	/*
848 	 * the following is safe, since for compiler definitions of kvec and
849 	 * iovec are identical, yielding the same in-core layout and alignment
850 	 */
851 	msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
852 	result = sock_recvmsg(sock, msg, size, flags);
853 	set_fs(oldfs);
854 	return result;
855 }
856 EXPORT_SYMBOL(kernel_recvmsg);
857 
858 static ssize_t sock_sendpage(struct file *file, struct page *page,
859 			     int offset, size_t size, loff_t *ppos, int more)
860 {
861 	struct socket *sock;
862 	int flags;
863 
864 	sock = file->private_data;
865 
866 	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
867 	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
868 	flags |= more;
869 
870 	return kernel_sendpage(sock, page, offset, size, flags);
871 }
872 
873 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
874 				struct pipe_inode_info *pipe, size_t len,
875 				unsigned int flags)
876 {
877 	struct socket *sock = file->private_data;
878 
879 	if (unlikely(!sock->ops->splice_read))
880 		return -EINVAL;
881 
882 	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
883 }
884 
885 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
886 					 struct sock_iocb *siocb)
887 {
888 	if (!is_sync_kiocb(iocb))
889 		BUG();
890 
891 	siocb->kiocb = iocb;
892 	iocb->private = siocb;
893 	return siocb;
894 }
895 
896 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
897 		struct file *file, const struct iovec *iov,
898 		unsigned long nr_segs)
899 {
900 	struct socket *sock = file->private_data;
901 	size_t size = 0;
902 	int i;
903 
904 	for (i = 0; i < nr_segs; i++)
905 		size += iov[i].iov_len;
906 
907 	msg->msg_name = NULL;
908 	msg->msg_namelen = 0;
909 	msg->msg_control = NULL;
910 	msg->msg_controllen = 0;
911 	msg->msg_iov = (struct iovec *)iov;
912 	msg->msg_iovlen = nr_segs;
913 	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
914 
915 	return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
916 }
917 
918 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
919 				unsigned long nr_segs, loff_t pos)
920 {
921 	struct sock_iocb siocb, *x;
922 
923 	if (pos != 0)
924 		return -ESPIPE;
925 
926 	if (iocb->ki_nbytes == 0)	/* Match SYS5 behaviour */
927 		return 0;
928 
929 
930 	x = alloc_sock_iocb(iocb, &siocb);
931 	if (!x)
932 		return -ENOMEM;
933 	return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
934 }
935 
936 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
937 			struct file *file, const struct iovec *iov,
938 			unsigned long nr_segs)
939 {
940 	struct socket *sock = file->private_data;
941 	size_t size = 0;
942 	int i;
943 
944 	for (i = 0; i < nr_segs; i++)
945 		size += iov[i].iov_len;
946 
947 	msg->msg_name = NULL;
948 	msg->msg_namelen = 0;
949 	msg->msg_control = NULL;
950 	msg->msg_controllen = 0;
951 	msg->msg_iov = (struct iovec *)iov;
952 	msg->msg_iovlen = nr_segs;
953 	msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
954 	if (sock->type == SOCK_SEQPACKET)
955 		msg->msg_flags |= MSG_EOR;
956 
957 	return __sock_sendmsg(iocb, sock, msg, size);
958 }
959 
960 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
961 			  unsigned long nr_segs, loff_t pos)
962 {
963 	struct sock_iocb siocb, *x;
964 
965 	if (pos != 0)
966 		return -ESPIPE;
967 
968 	x = alloc_sock_iocb(iocb, &siocb);
969 	if (!x)
970 		return -ENOMEM;
971 
972 	return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
973 }
974 
975 /*
976  * Atomic setting of ioctl hooks to avoid race
977  * with module unload.
978  */
979 
980 static DEFINE_MUTEX(br_ioctl_mutex);
981 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
982 
983 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
984 {
985 	mutex_lock(&br_ioctl_mutex);
986 	br_ioctl_hook = hook;
987 	mutex_unlock(&br_ioctl_mutex);
988 }
989 EXPORT_SYMBOL(brioctl_set);
990 
991 static DEFINE_MUTEX(vlan_ioctl_mutex);
992 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
993 
994 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
995 {
996 	mutex_lock(&vlan_ioctl_mutex);
997 	vlan_ioctl_hook = hook;
998 	mutex_unlock(&vlan_ioctl_mutex);
999 }
1000 EXPORT_SYMBOL(vlan_ioctl_set);
1001 
1002 static DEFINE_MUTEX(dlci_ioctl_mutex);
1003 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1004 
1005 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1006 {
1007 	mutex_lock(&dlci_ioctl_mutex);
1008 	dlci_ioctl_hook = hook;
1009 	mutex_unlock(&dlci_ioctl_mutex);
1010 }
1011 EXPORT_SYMBOL(dlci_ioctl_set);
1012 
1013 static long sock_do_ioctl(struct net *net, struct socket *sock,
1014 				 unsigned int cmd, unsigned long arg)
1015 {
1016 	int err;
1017 	void __user *argp = (void __user *)arg;
1018 
1019 	err = sock->ops->ioctl(sock, cmd, arg);
1020 
1021 	/*
1022 	 * If this ioctl is unknown try to hand it down
1023 	 * to the NIC driver.
1024 	 */
1025 	if (err == -ENOIOCTLCMD)
1026 		err = dev_ioctl(net, cmd, argp);
1027 
1028 	return err;
1029 }
1030 
1031 /*
1032  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1033  *	what to do with it - that's up to the protocol still.
1034  */
1035 
1036 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1037 {
1038 	struct socket *sock;
1039 	struct sock *sk;
1040 	void __user *argp = (void __user *)arg;
1041 	int pid, err;
1042 	struct net *net;
1043 
1044 	sock = file->private_data;
1045 	sk = sock->sk;
1046 	net = sock_net(sk);
1047 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
1048 		err = dev_ioctl(net, cmd, argp);
1049 	} else
1050 #ifdef CONFIG_WEXT_CORE
1051 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1052 		err = dev_ioctl(net, cmd, argp);
1053 	} else
1054 #endif
1055 		switch (cmd) {
1056 		case FIOSETOWN:
1057 		case SIOCSPGRP:
1058 			err = -EFAULT;
1059 			if (get_user(pid, (int __user *)argp))
1060 				break;
1061 			err = f_setown(sock->file, pid, 1);
1062 			break;
1063 		case FIOGETOWN:
1064 		case SIOCGPGRP:
1065 			err = put_user(f_getown(sock->file),
1066 				       (int __user *)argp);
1067 			break;
1068 		case SIOCGIFBR:
1069 		case SIOCSIFBR:
1070 		case SIOCBRADDBR:
1071 		case SIOCBRDELBR:
1072 			err = -ENOPKG;
1073 			if (!br_ioctl_hook)
1074 				request_module("bridge");
1075 
1076 			mutex_lock(&br_ioctl_mutex);
1077 			if (br_ioctl_hook)
1078 				err = br_ioctl_hook(net, cmd, argp);
1079 			mutex_unlock(&br_ioctl_mutex);
1080 			break;
1081 		case SIOCGIFVLAN:
1082 		case SIOCSIFVLAN:
1083 			err = -ENOPKG;
1084 			if (!vlan_ioctl_hook)
1085 				request_module("8021q");
1086 
1087 			mutex_lock(&vlan_ioctl_mutex);
1088 			if (vlan_ioctl_hook)
1089 				err = vlan_ioctl_hook(net, argp);
1090 			mutex_unlock(&vlan_ioctl_mutex);
1091 			break;
1092 		case SIOCADDDLCI:
1093 		case SIOCDELDLCI:
1094 			err = -ENOPKG;
1095 			if (!dlci_ioctl_hook)
1096 				request_module("dlci");
1097 
1098 			mutex_lock(&dlci_ioctl_mutex);
1099 			if (dlci_ioctl_hook)
1100 				err = dlci_ioctl_hook(cmd, argp);
1101 			mutex_unlock(&dlci_ioctl_mutex);
1102 			break;
1103 		default:
1104 			err = sock_do_ioctl(net, sock, cmd, arg);
1105 			break;
1106 		}
1107 	return err;
1108 }
1109 
1110 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1111 {
1112 	int err;
1113 	struct socket *sock = NULL;
1114 
1115 	err = security_socket_create(family, type, protocol, 1);
1116 	if (err)
1117 		goto out;
1118 
1119 	sock = sock_alloc();
1120 	if (!sock) {
1121 		err = -ENOMEM;
1122 		goto out;
1123 	}
1124 
1125 	sock->type = type;
1126 	err = security_socket_post_create(sock, family, type, protocol, 1);
1127 	if (err)
1128 		goto out_release;
1129 
1130 out:
1131 	*res = sock;
1132 	return err;
1133 out_release:
1134 	sock_release(sock);
1135 	sock = NULL;
1136 	goto out;
1137 }
1138 EXPORT_SYMBOL(sock_create_lite);
1139 
1140 /* No kernel lock held - perfect */
1141 static unsigned int sock_poll(struct file *file, poll_table *wait)
1142 {
1143 	unsigned int busy_flag = 0;
1144 	struct socket *sock;
1145 
1146 	/*
1147 	 *      We can't return errors to poll, so it's either yes or no.
1148 	 */
1149 	sock = file->private_data;
1150 
1151 	if (sk_can_busy_loop(sock->sk)) {
1152 		/* this socket can poll_ll so tell the system call */
1153 		busy_flag = POLL_BUSY_LOOP;
1154 
1155 		/* once, only if requested by syscall */
1156 		if (wait && (wait->_key & POLL_BUSY_LOOP))
1157 			sk_busy_loop(sock->sk, 1);
1158 	}
1159 
1160 	return busy_flag | sock->ops->poll(file, sock, wait);
1161 }
1162 
1163 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1164 {
1165 	struct socket *sock = file->private_data;
1166 
1167 	return sock->ops->mmap(file, sock, vma);
1168 }
1169 
1170 static int sock_close(struct inode *inode, struct file *filp)
1171 {
1172 	sock_release(SOCKET_I(inode));
1173 	return 0;
1174 }
1175 
1176 /*
1177  *	Update the socket async list
1178  *
1179  *	Fasync_list locking strategy.
1180  *
1181  *	1. fasync_list is modified only under process context socket lock
1182  *	   i.e. under semaphore.
1183  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1184  *	   or under socket lock
1185  */
1186 
1187 static int sock_fasync(int fd, struct file *filp, int on)
1188 {
1189 	struct socket *sock = filp->private_data;
1190 	struct sock *sk = sock->sk;
1191 	struct socket_wq *wq;
1192 
1193 	if (sk == NULL)
1194 		return -EINVAL;
1195 
1196 	lock_sock(sk);
1197 	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1198 	fasync_helper(fd, filp, on, &wq->fasync_list);
1199 
1200 	if (!wq->fasync_list)
1201 		sock_reset_flag(sk, SOCK_FASYNC);
1202 	else
1203 		sock_set_flag(sk, SOCK_FASYNC);
1204 
1205 	release_sock(sk);
1206 	return 0;
1207 }
1208 
1209 /* This function may be called only under socket lock or callback_lock or rcu_lock */
1210 
1211 int sock_wake_async(struct socket *sock, int how, int band)
1212 {
1213 	struct socket_wq *wq;
1214 
1215 	if (!sock)
1216 		return -1;
1217 	rcu_read_lock();
1218 	wq = rcu_dereference(sock->wq);
1219 	if (!wq || !wq->fasync_list) {
1220 		rcu_read_unlock();
1221 		return -1;
1222 	}
1223 	switch (how) {
1224 	case SOCK_WAKE_WAITD:
1225 		if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1226 			break;
1227 		goto call_kill;
1228 	case SOCK_WAKE_SPACE:
1229 		if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1230 			break;
1231 		/* fall through */
1232 	case SOCK_WAKE_IO:
1233 call_kill:
1234 		kill_fasync(&wq->fasync_list, SIGIO, band);
1235 		break;
1236 	case SOCK_WAKE_URG:
1237 		kill_fasync(&wq->fasync_list, SIGURG, band);
1238 	}
1239 	rcu_read_unlock();
1240 	return 0;
1241 }
1242 EXPORT_SYMBOL(sock_wake_async);
1243 
1244 int __sock_create(struct net *net, int family, int type, int protocol,
1245 			 struct socket **res, int kern)
1246 {
1247 	int err;
1248 	struct socket *sock;
1249 	const struct net_proto_family *pf;
1250 
1251 	/*
1252 	 *      Check protocol is in range
1253 	 */
1254 	if (family < 0 || family >= NPROTO)
1255 		return -EAFNOSUPPORT;
1256 	if (type < 0 || type >= SOCK_MAX)
1257 		return -EINVAL;
1258 
1259 	/* Compatibility.
1260 
1261 	   This uglymoron is moved from INET layer to here to avoid
1262 	   deadlock in module load.
1263 	 */
1264 	if (family == PF_INET && type == SOCK_PACKET) {
1265 		static int warned;
1266 		if (!warned) {
1267 			warned = 1;
1268 			printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1269 			       current->comm);
1270 		}
1271 		family = PF_PACKET;
1272 	}
1273 
1274 	err = security_socket_create(family, type, protocol, kern);
1275 	if (err)
1276 		return err;
1277 
1278 	/*
1279 	 *	Allocate the socket and allow the family to set things up. if
1280 	 *	the protocol is 0, the family is instructed to select an appropriate
1281 	 *	default.
1282 	 */
1283 	sock = sock_alloc();
1284 	if (!sock) {
1285 		net_warn_ratelimited("socket: no more sockets\n");
1286 		return -ENFILE;	/* Not exactly a match, but its the
1287 				   closest posix thing */
1288 	}
1289 
1290 	sock->type = type;
1291 
1292 #ifdef CONFIG_MODULES
1293 	/* Attempt to load a protocol module if the find failed.
1294 	 *
1295 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1296 	 * requested real, full-featured networking support upon configuration.
1297 	 * Otherwise module support will break!
1298 	 */
1299 	if (rcu_access_pointer(net_families[family]) == NULL)
1300 		request_module("net-pf-%d", family);
1301 #endif
1302 
1303 	rcu_read_lock();
1304 	pf = rcu_dereference(net_families[family]);
1305 	err = -EAFNOSUPPORT;
1306 	if (!pf)
1307 		goto out_release;
1308 
1309 	/*
1310 	 * We will call the ->create function, that possibly is in a loadable
1311 	 * module, so we have to bump that loadable module refcnt first.
1312 	 */
1313 	if (!try_module_get(pf->owner))
1314 		goto out_release;
1315 
1316 	/* Now protected by module ref count */
1317 	rcu_read_unlock();
1318 
1319 	err = pf->create(net, sock, protocol, kern);
1320 	if (err < 0)
1321 		goto out_module_put;
1322 
1323 	/*
1324 	 * Now to bump the refcnt of the [loadable] module that owns this
1325 	 * socket at sock_release time we decrement its refcnt.
1326 	 */
1327 	if (!try_module_get(sock->ops->owner))
1328 		goto out_module_busy;
1329 
1330 	/*
1331 	 * Now that we're done with the ->create function, the [loadable]
1332 	 * module can have its refcnt decremented
1333 	 */
1334 	module_put(pf->owner);
1335 	err = security_socket_post_create(sock, family, type, protocol, kern);
1336 	if (err)
1337 		goto out_sock_release;
1338 	*res = sock;
1339 
1340 	return 0;
1341 
1342 out_module_busy:
1343 	err = -EAFNOSUPPORT;
1344 out_module_put:
1345 	sock->ops = NULL;
1346 	module_put(pf->owner);
1347 out_sock_release:
1348 	sock_release(sock);
1349 	return err;
1350 
1351 out_release:
1352 	rcu_read_unlock();
1353 	goto out_sock_release;
1354 }
1355 EXPORT_SYMBOL(__sock_create);
1356 
1357 int sock_create(int family, int type, int protocol, struct socket **res)
1358 {
1359 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1360 }
1361 EXPORT_SYMBOL(sock_create);
1362 
1363 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1364 {
1365 	return __sock_create(&init_net, family, type, protocol, res, 1);
1366 }
1367 EXPORT_SYMBOL(sock_create_kern);
1368 
1369 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1370 {
1371 	int retval;
1372 	struct socket *sock;
1373 	int flags;
1374 
1375 	/* Check the SOCK_* constants for consistency.  */
1376 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1377 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1378 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1379 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1380 
1381 	flags = type & ~SOCK_TYPE_MASK;
1382 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1383 		return -EINVAL;
1384 	type &= SOCK_TYPE_MASK;
1385 
1386 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1387 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1388 
1389 	retval = sock_create(family, type, protocol, &sock);
1390 	if (retval < 0)
1391 		goto out;
1392 
1393 	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1394 	if (retval < 0)
1395 		goto out_release;
1396 
1397 out:
1398 	/* It may be already another descriptor 8) Not kernel problem. */
1399 	return retval;
1400 
1401 out_release:
1402 	sock_release(sock);
1403 	return retval;
1404 }
1405 
1406 /*
1407  *	Create a pair of connected sockets.
1408  */
1409 
1410 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1411 		int __user *, usockvec)
1412 {
1413 	struct socket *sock1, *sock2;
1414 	int fd1, fd2, err;
1415 	struct file *newfile1, *newfile2;
1416 	int flags;
1417 
1418 	flags = type & ~SOCK_TYPE_MASK;
1419 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1420 		return -EINVAL;
1421 	type &= SOCK_TYPE_MASK;
1422 
1423 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1424 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1425 
1426 	/*
1427 	 * Obtain the first socket and check if the underlying protocol
1428 	 * supports the socketpair call.
1429 	 */
1430 
1431 	err = sock_create(family, type, protocol, &sock1);
1432 	if (err < 0)
1433 		goto out;
1434 
1435 	err = sock_create(family, type, protocol, &sock2);
1436 	if (err < 0)
1437 		goto out_release_1;
1438 
1439 	err = sock1->ops->socketpair(sock1, sock2);
1440 	if (err < 0)
1441 		goto out_release_both;
1442 
1443 	fd1 = get_unused_fd_flags(flags);
1444 	if (unlikely(fd1 < 0)) {
1445 		err = fd1;
1446 		goto out_release_both;
1447 	}
1448 	fd2 = get_unused_fd_flags(flags);
1449 	if (unlikely(fd2 < 0)) {
1450 		err = fd2;
1451 		put_unused_fd(fd1);
1452 		goto out_release_both;
1453 	}
1454 
1455 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1456 	if (unlikely(IS_ERR(newfile1))) {
1457 		err = PTR_ERR(newfile1);
1458 		put_unused_fd(fd1);
1459 		put_unused_fd(fd2);
1460 		goto out_release_both;
1461 	}
1462 
1463 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1464 	if (IS_ERR(newfile2)) {
1465 		err = PTR_ERR(newfile2);
1466 		fput(newfile1);
1467 		put_unused_fd(fd1);
1468 		put_unused_fd(fd2);
1469 		sock_release(sock2);
1470 		goto out;
1471 	}
1472 
1473 	audit_fd_pair(fd1, fd2);
1474 	fd_install(fd1, newfile1);
1475 	fd_install(fd2, newfile2);
1476 	/* fd1 and fd2 may be already another descriptors.
1477 	 * Not kernel problem.
1478 	 */
1479 
1480 	err = put_user(fd1, &usockvec[0]);
1481 	if (!err)
1482 		err = put_user(fd2, &usockvec[1]);
1483 	if (!err)
1484 		return 0;
1485 
1486 	sys_close(fd2);
1487 	sys_close(fd1);
1488 	return err;
1489 
1490 out_release_both:
1491 	sock_release(sock2);
1492 out_release_1:
1493 	sock_release(sock1);
1494 out:
1495 	return err;
1496 }
1497 
1498 /*
1499  *	Bind a name to a socket. Nothing much to do here since it's
1500  *	the protocol's responsibility to handle the local address.
1501  *
1502  *	We move the socket address to kernel space before we call
1503  *	the protocol layer (having also checked the address is ok).
1504  */
1505 
1506 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1507 {
1508 	struct socket *sock;
1509 	struct sockaddr_storage address;
1510 	int err, fput_needed;
1511 
1512 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1513 	if (sock) {
1514 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1515 		if (err >= 0) {
1516 			err = security_socket_bind(sock,
1517 						   (struct sockaddr *)&address,
1518 						   addrlen);
1519 			if (!err)
1520 				err = sock->ops->bind(sock,
1521 						      (struct sockaddr *)
1522 						      &address, addrlen);
1523 		}
1524 		fput_light(sock->file, fput_needed);
1525 	}
1526 	return err;
1527 }
1528 
1529 /*
1530  *	Perform a listen. Basically, we allow the protocol to do anything
1531  *	necessary for a listen, and if that works, we mark the socket as
1532  *	ready for listening.
1533  */
1534 
1535 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1536 {
1537 	struct socket *sock;
1538 	int err, fput_needed;
1539 	int somaxconn;
1540 
1541 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1542 	if (sock) {
1543 		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1544 		if ((unsigned int)backlog > somaxconn)
1545 			backlog = somaxconn;
1546 
1547 		err = security_socket_listen(sock, backlog);
1548 		if (!err)
1549 			err = sock->ops->listen(sock, backlog);
1550 
1551 		fput_light(sock->file, fput_needed);
1552 	}
1553 	return err;
1554 }
1555 
1556 /*
1557  *	For accept, we attempt to create a new socket, set up the link
1558  *	with the client, wake up the client, then return the new
1559  *	connected fd. We collect the address of the connector in kernel
1560  *	space and move it to user at the very end. This is unclean because
1561  *	we open the socket then return an error.
1562  *
1563  *	1003.1g adds the ability to recvmsg() to query connection pending
1564  *	status to recvmsg. We need to add that support in a way thats
1565  *	clean when we restucture accept also.
1566  */
1567 
1568 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1569 		int __user *, upeer_addrlen, int, flags)
1570 {
1571 	struct socket *sock, *newsock;
1572 	struct file *newfile;
1573 	int err, len, newfd, fput_needed;
1574 	struct sockaddr_storage address;
1575 
1576 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1577 		return -EINVAL;
1578 
1579 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1580 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1581 
1582 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1583 	if (!sock)
1584 		goto out;
1585 
1586 	err = -ENFILE;
1587 	newsock = sock_alloc();
1588 	if (!newsock)
1589 		goto out_put;
1590 
1591 	newsock->type = sock->type;
1592 	newsock->ops = sock->ops;
1593 
1594 	/*
1595 	 * We don't need try_module_get here, as the listening socket (sock)
1596 	 * has the protocol module (sock->ops->owner) held.
1597 	 */
1598 	__module_get(newsock->ops->owner);
1599 
1600 	newfd = get_unused_fd_flags(flags);
1601 	if (unlikely(newfd < 0)) {
1602 		err = newfd;
1603 		sock_release(newsock);
1604 		goto out_put;
1605 	}
1606 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1607 	if (unlikely(IS_ERR(newfile))) {
1608 		err = PTR_ERR(newfile);
1609 		put_unused_fd(newfd);
1610 		sock_release(newsock);
1611 		goto out_put;
1612 	}
1613 
1614 	err = security_socket_accept(sock, newsock);
1615 	if (err)
1616 		goto out_fd;
1617 
1618 	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1619 	if (err < 0)
1620 		goto out_fd;
1621 
1622 	if (upeer_sockaddr) {
1623 		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1624 					  &len, 2) < 0) {
1625 			err = -ECONNABORTED;
1626 			goto out_fd;
1627 		}
1628 		err = move_addr_to_user(&address,
1629 					len, upeer_sockaddr, upeer_addrlen);
1630 		if (err < 0)
1631 			goto out_fd;
1632 	}
1633 
1634 	/* File flags are not inherited via accept() unlike another OSes. */
1635 
1636 	fd_install(newfd, newfile);
1637 	err = newfd;
1638 
1639 out_put:
1640 	fput_light(sock->file, fput_needed);
1641 out:
1642 	return err;
1643 out_fd:
1644 	fput(newfile);
1645 	put_unused_fd(newfd);
1646 	goto out_put;
1647 }
1648 
1649 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1650 		int __user *, upeer_addrlen)
1651 {
1652 	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1653 }
1654 
1655 /*
1656  *	Attempt to connect to a socket with the server address.  The address
1657  *	is in user space so we verify it is OK and move it to kernel space.
1658  *
1659  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1660  *	break bindings
1661  *
1662  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1663  *	other SEQPACKET protocols that take time to connect() as it doesn't
1664  *	include the -EINPROGRESS status for such sockets.
1665  */
1666 
1667 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1668 		int, addrlen)
1669 {
1670 	struct socket *sock;
1671 	struct sockaddr_storage address;
1672 	int err, fput_needed;
1673 
1674 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1675 	if (!sock)
1676 		goto out;
1677 	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1678 	if (err < 0)
1679 		goto out_put;
1680 
1681 	err =
1682 	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1683 	if (err)
1684 		goto out_put;
1685 
1686 	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1687 				 sock->file->f_flags);
1688 out_put:
1689 	fput_light(sock->file, fput_needed);
1690 out:
1691 	return err;
1692 }
1693 
1694 /*
1695  *	Get the local address ('name') of a socket object. Move the obtained
1696  *	name to user space.
1697  */
1698 
1699 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1700 		int __user *, usockaddr_len)
1701 {
1702 	struct socket *sock;
1703 	struct sockaddr_storage address;
1704 	int len, err, fput_needed;
1705 
1706 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1707 	if (!sock)
1708 		goto out;
1709 
1710 	err = security_socket_getsockname(sock);
1711 	if (err)
1712 		goto out_put;
1713 
1714 	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1715 	if (err)
1716 		goto out_put;
1717 	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1718 
1719 out_put:
1720 	fput_light(sock->file, fput_needed);
1721 out:
1722 	return err;
1723 }
1724 
1725 /*
1726  *	Get the remote address ('name') of a socket object. Move the obtained
1727  *	name to user space.
1728  */
1729 
1730 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1731 		int __user *, usockaddr_len)
1732 {
1733 	struct socket *sock;
1734 	struct sockaddr_storage address;
1735 	int len, err, fput_needed;
1736 
1737 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1738 	if (sock != NULL) {
1739 		err = security_socket_getpeername(sock);
1740 		if (err) {
1741 			fput_light(sock->file, fput_needed);
1742 			return err;
1743 		}
1744 
1745 		err =
1746 		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1747 				       1);
1748 		if (!err)
1749 			err = move_addr_to_user(&address, len, usockaddr,
1750 						usockaddr_len);
1751 		fput_light(sock->file, fput_needed);
1752 	}
1753 	return err;
1754 }
1755 
1756 /*
1757  *	Send a datagram to a given address. We move the address into kernel
1758  *	space and check the user space data area is readable before invoking
1759  *	the protocol.
1760  */
1761 
1762 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1763 		unsigned int, flags, struct sockaddr __user *, addr,
1764 		int, addr_len)
1765 {
1766 	struct socket *sock;
1767 	struct sockaddr_storage address;
1768 	int err;
1769 	struct msghdr msg;
1770 	struct iovec iov;
1771 	int fput_needed;
1772 
1773 	if (len > INT_MAX)
1774 		len = INT_MAX;
1775 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1776 	if (!sock)
1777 		goto out;
1778 
1779 	iov.iov_base = buff;
1780 	iov.iov_len = len;
1781 	msg.msg_name = NULL;
1782 	msg.msg_iov = &iov;
1783 	msg.msg_iovlen = 1;
1784 	msg.msg_control = NULL;
1785 	msg.msg_controllen = 0;
1786 	msg.msg_namelen = 0;
1787 	if (addr) {
1788 		err = move_addr_to_kernel(addr, addr_len, &address);
1789 		if (err < 0)
1790 			goto out_put;
1791 		msg.msg_name = (struct sockaddr *)&address;
1792 		msg.msg_namelen = addr_len;
1793 	}
1794 	if (sock->file->f_flags & O_NONBLOCK)
1795 		flags |= MSG_DONTWAIT;
1796 	msg.msg_flags = flags;
1797 	err = sock_sendmsg(sock, &msg, len);
1798 
1799 out_put:
1800 	fput_light(sock->file, fput_needed);
1801 out:
1802 	return err;
1803 }
1804 
1805 /*
1806  *	Send a datagram down a socket.
1807  */
1808 
1809 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1810 		unsigned int, flags)
1811 {
1812 	return sys_sendto(fd, buff, len, flags, NULL, 0);
1813 }
1814 
1815 /*
1816  *	Receive a frame from the socket and optionally record the address of the
1817  *	sender. We verify the buffers are writable and if needed move the
1818  *	sender address from kernel to user space.
1819  */
1820 
1821 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1822 		unsigned int, flags, struct sockaddr __user *, addr,
1823 		int __user *, addr_len)
1824 {
1825 	struct socket *sock;
1826 	struct iovec iov;
1827 	struct msghdr msg;
1828 	struct sockaddr_storage address;
1829 	int err, err2;
1830 	int fput_needed;
1831 
1832 	if (size > INT_MAX)
1833 		size = INT_MAX;
1834 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1835 	if (!sock)
1836 		goto out;
1837 
1838 	msg.msg_control = NULL;
1839 	msg.msg_controllen = 0;
1840 	msg.msg_iovlen = 1;
1841 	msg.msg_iov = &iov;
1842 	iov.iov_len = size;
1843 	iov.iov_base = ubuf;
1844 	/* Save some cycles and don't copy the address if not needed */
1845 	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1846 	/* We assume all kernel code knows the size of sockaddr_storage */
1847 	msg.msg_namelen = 0;
1848 	if (sock->file->f_flags & O_NONBLOCK)
1849 		flags |= MSG_DONTWAIT;
1850 	err = sock_recvmsg(sock, &msg, size, flags);
1851 
1852 	if (err >= 0 && addr != NULL) {
1853 		err2 = move_addr_to_user(&address,
1854 					 msg.msg_namelen, addr, addr_len);
1855 		if (err2 < 0)
1856 			err = err2;
1857 	}
1858 
1859 	fput_light(sock->file, fput_needed);
1860 out:
1861 	return err;
1862 }
1863 
1864 /*
1865  *	Receive a datagram from a socket.
1866  */
1867 
1868 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1869 			 unsigned int flags)
1870 {
1871 	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1872 }
1873 
1874 /*
1875  *	Set a socket option. Because we don't know the option lengths we have
1876  *	to pass the user mode parameter for the protocols to sort out.
1877  */
1878 
1879 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1880 		char __user *, optval, int, optlen)
1881 {
1882 	int err, fput_needed;
1883 	struct socket *sock;
1884 
1885 	if (optlen < 0)
1886 		return -EINVAL;
1887 
1888 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1889 	if (sock != NULL) {
1890 		err = security_socket_setsockopt(sock, level, optname);
1891 		if (err)
1892 			goto out_put;
1893 
1894 		if (level == SOL_SOCKET)
1895 			err =
1896 			    sock_setsockopt(sock, level, optname, optval,
1897 					    optlen);
1898 		else
1899 			err =
1900 			    sock->ops->setsockopt(sock, level, optname, optval,
1901 						  optlen);
1902 out_put:
1903 		fput_light(sock->file, fput_needed);
1904 	}
1905 	return err;
1906 }
1907 
1908 /*
1909  *	Get a socket option. Because we don't know the option lengths we have
1910  *	to pass a user mode parameter for the protocols to sort out.
1911  */
1912 
1913 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1914 		char __user *, optval, int __user *, optlen)
1915 {
1916 	int err, fput_needed;
1917 	struct socket *sock;
1918 
1919 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1920 	if (sock != NULL) {
1921 		err = security_socket_getsockopt(sock, level, optname);
1922 		if (err)
1923 			goto out_put;
1924 
1925 		if (level == SOL_SOCKET)
1926 			err =
1927 			    sock_getsockopt(sock, level, optname, optval,
1928 					    optlen);
1929 		else
1930 			err =
1931 			    sock->ops->getsockopt(sock, level, optname, optval,
1932 						  optlen);
1933 out_put:
1934 		fput_light(sock->file, fput_needed);
1935 	}
1936 	return err;
1937 }
1938 
1939 /*
1940  *	Shutdown a socket.
1941  */
1942 
1943 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1944 {
1945 	int err, fput_needed;
1946 	struct socket *sock;
1947 
1948 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1949 	if (sock != NULL) {
1950 		err = security_socket_shutdown(sock, how);
1951 		if (!err)
1952 			err = sock->ops->shutdown(sock, how);
1953 		fput_light(sock->file, fput_needed);
1954 	}
1955 	return err;
1956 }
1957 
1958 /* A couple of helpful macros for getting the address of the 32/64 bit
1959  * fields which are the same type (int / unsigned) on our platforms.
1960  */
1961 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1962 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1963 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1964 
1965 struct used_address {
1966 	struct sockaddr_storage name;
1967 	unsigned int name_len;
1968 };
1969 
1970 static int copy_msghdr_from_user(struct msghdr *kmsg,
1971 				 struct msghdr __user *umsg)
1972 {
1973 	if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1974 		return -EFAULT;
1975 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1976 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1977 	return 0;
1978 }
1979 
1980 static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
1981 			 struct msghdr *msg_sys, unsigned int flags,
1982 			 struct used_address *used_address)
1983 {
1984 	struct compat_msghdr __user *msg_compat =
1985 	    (struct compat_msghdr __user *)msg;
1986 	struct sockaddr_storage address;
1987 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1988 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1989 	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1990 	/* 20 is size of ipv6_pktinfo */
1991 	unsigned char *ctl_buf = ctl;
1992 	int err, ctl_len, total_len;
1993 
1994 	err = -EFAULT;
1995 	if (MSG_CMSG_COMPAT & flags) {
1996 		if (get_compat_msghdr(msg_sys, msg_compat))
1997 			return -EFAULT;
1998 	} else {
1999 		err = copy_msghdr_from_user(msg_sys, msg);
2000 		if (err)
2001 			return err;
2002 	}
2003 
2004 	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2005 		err = -EMSGSIZE;
2006 		if (msg_sys->msg_iovlen > UIO_MAXIOV)
2007 			goto out;
2008 		err = -ENOMEM;
2009 		iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2010 			      GFP_KERNEL);
2011 		if (!iov)
2012 			goto out;
2013 	}
2014 
2015 	/* This will also move the address data into kernel space */
2016 	if (MSG_CMSG_COMPAT & flags) {
2017 		err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
2018 	} else
2019 		err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
2020 	if (err < 0)
2021 		goto out_freeiov;
2022 	total_len = err;
2023 
2024 	err = -ENOBUFS;
2025 
2026 	if (msg_sys->msg_controllen > INT_MAX)
2027 		goto out_freeiov;
2028 	ctl_len = msg_sys->msg_controllen;
2029 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2030 		err =
2031 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2032 						     sizeof(ctl));
2033 		if (err)
2034 			goto out_freeiov;
2035 		ctl_buf = msg_sys->msg_control;
2036 		ctl_len = msg_sys->msg_controllen;
2037 	} else if (ctl_len) {
2038 		if (ctl_len > sizeof(ctl)) {
2039 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2040 			if (ctl_buf == NULL)
2041 				goto out_freeiov;
2042 		}
2043 		err = -EFAULT;
2044 		/*
2045 		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2046 		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2047 		 * checking falls down on this.
2048 		 */
2049 		if (copy_from_user(ctl_buf,
2050 				   (void __user __force *)msg_sys->msg_control,
2051 				   ctl_len))
2052 			goto out_freectl;
2053 		msg_sys->msg_control = ctl_buf;
2054 	}
2055 	msg_sys->msg_flags = flags;
2056 
2057 	if (sock->file->f_flags & O_NONBLOCK)
2058 		msg_sys->msg_flags |= MSG_DONTWAIT;
2059 	/*
2060 	 * If this is sendmmsg() and current destination address is same as
2061 	 * previously succeeded address, omit asking LSM's decision.
2062 	 * used_address->name_len is initialized to UINT_MAX so that the first
2063 	 * destination address never matches.
2064 	 */
2065 	if (used_address && msg_sys->msg_name &&
2066 	    used_address->name_len == msg_sys->msg_namelen &&
2067 	    !memcmp(&used_address->name, msg_sys->msg_name,
2068 		    used_address->name_len)) {
2069 		err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2070 		goto out_freectl;
2071 	}
2072 	err = sock_sendmsg(sock, msg_sys, total_len);
2073 	/*
2074 	 * If this is sendmmsg() and sending to current destination address was
2075 	 * successful, remember it.
2076 	 */
2077 	if (used_address && err >= 0) {
2078 		used_address->name_len = msg_sys->msg_namelen;
2079 		if (msg_sys->msg_name)
2080 			memcpy(&used_address->name, msg_sys->msg_name,
2081 			       used_address->name_len);
2082 	}
2083 
2084 out_freectl:
2085 	if (ctl_buf != ctl)
2086 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2087 out_freeiov:
2088 	if (iov != iovstack)
2089 		kfree(iov);
2090 out:
2091 	return err;
2092 }
2093 
2094 /*
2095  *	BSD sendmsg interface
2096  */
2097 
2098 long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
2099 {
2100 	int fput_needed, err;
2101 	struct msghdr msg_sys;
2102 	struct socket *sock;
2103 
2104 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2105 	if (!sock)
2106 		goto out;
2107 
2108 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
2109 
2110 	fput_light(sock->file, fput_needed);
2111 out:
2112 	return err;
2113 }
2114 
2115 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2116 {
2117 	if (flags & MSG_CMSG_COMPAT)
2118 		return -EINVAL;
2119 	return __sys_sendmsg(fd, msg, flags);
2120 }
2121 
2122 /*
2123  *	Linux sendmmsg interface
2124  */
2125 
2126 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2127 		   unsigned int flags)
2128 {
2129 	int fput_needed, err, datagrams;
2130 	struct socket *sock;
2131 	struct mmsghdr __user *entry;
2132 	struct compat_mmsghdr __user *compat_entry;
2133 	struct msghdr msg_sys;
2134 	struct used_address used_address;
2135 
2136 	if (vlen > UIO_MAXIOV)
2137 		vlen = UIO_MAXIOV;
2138 
2139 	datagrams = 0;
2140 
2141 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2142 	if (!sock)
2143 		return err;
2144 
2145 	used_address.name_len = UINT_MAX;
2146 	entry = mmsg;
2147 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2148 	err = 0;
2149 
2150 	while (datagrams < vlen) {
2151 		if (MSG_CMSG_COMPAT & flags) {
2152 			err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2153 					     &msg_sys, flags, &used_address);
2154 			if (err < 0)
2155 				break;
2156 			err = __put_user(err, &compat_entry->msg_len);
2157 			++compat_entry;
2158 		} else {
2159 			err = ___sys_sendmsg(sock,
2160 					     (struct msghdr __user *)entry,
2161 					     &msg_sys, flags, &used_address);
2162 			if (err < 0)
2163 				break;
2164 			err = put_user(err, &entry->msg_len);
2165 			++entry;
2166 		}
2167 
2168 		if (err)
2169 			break;
2170 		++datagrams;
2171 	}
2172 
2173 	fput_light(sock->file, fput_needed);
2174 
2175 	/* We only return an error if no datagrams were able to be sent */
2176 	if (datagrams != 0)
2177 		return datagrams;
2178 
2179 	return err;
2180 }
2181 
2182 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2183 		unsigned int, vlen, unsigned int, flags)
2184 {
2185 	if (flags & MSG_CMSG_COMPAT)
2186 		return -EINVAL;
2187 	return __sys_sendmmsg(fd, mmsg, vlen, flags);
2188 }
2189 
2190 static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2191 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2192 {
2193 	struct compat_msghdr __user *msg_compat =
2194 	    (struct compat_msghdr __user *)msg;
2195 	struct iovec iovstack[UIO_FASTIOV];
2196 	struct iovec *iov = iovstack;
2197 	unsigned long cmsg_ptr;
2198 	int err, total_len, len;
2199 
2200 	/* kernel mode address */
2201 	struct sockaddr_storage addr;
2202 
2203 	/* user mode address pointers */
2204 	struct sockaddr __user *uaddr;
2205 	int __user *uaddr_len;
2206 
2207 	if (MSG_CMSG_COMPAT & flags) {
2208 		if (get_compat_msghdr(msg_sys, msg_compat))
2209 			return -EFAULT;
2210 	} else {
2211 		err = copy_msghdr_from_user(msg_sys, msg);
2212 		if (err)
2213 			return err;
2214 	}
2215 
2216 	if (msg_sys->msg_iovlen > UIO_FASTIOV) {
2217 		err = -EMSGSIZE;
2218 		if (msg_sys->msg_iovlen > UIO_MAXIOV)
2219 			goto out;
2220 		err = -ENOMEM;
2221 		iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2222 			      GFP_KERNEL);
2223 		if (!iov)
2224 			goto out;
2225 	}
2226 
2227 	/* Save the user-mode address (verify_iovec will change the
2228 	 * kernel msghdr to use the kernel address space)
2229 	 */
2230 	uaddr = (__force void __user *)msg_sys->msg_name;
2231 	uaddr_len = COMPAT_NAMELEN(msg);
2232 	if (MSG_CMSG_COMPAT & flags)
2233 		err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2234 	else
2235 		err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
2236 	if (err < 0)
2237 		goto out_freeiov;
2238 	total_len = err;
2239 
2240 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2241 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2242 
2243 	/* We assume all kernel code knows the size of sockaddr_storage */
2244 	msg_sys->msg_namelen = 0;
2245 
2246 	if (sock->file->f_flags & O_NONBLOCK)
2247 		flags |= MSG_DONTWAIT;
2248 	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2249 							  total_len, flags);
2250 	if (err < 0)
2251 		goto out_freeiov;
2252 	len = err;
2253 
2254 	if (uaddr != NULL) {
2255 		err = move_addr_to_user(&addr,
2256 					msg_sys->msg_namelen, uaddr,
2257 					uaddr_len);
2258 		if (err < 0)
2259 			goto out_freeiov;
2260 	}
2261 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2262 			 COMPAT_FLAGS(msg));
2263 	if (err)
2264 		goto out_freeiov;
2265 	if (MSG_CMSG_COMPAT & flags)
2266 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2267 				 &msg_compat->msg_controllen);
2268 	else
2269 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2270 				 &msg->msg_controllen);
2271 	if (err)
2272 		goto out_freeiov;
2273 	err = len;
2274 
2275 out_freeiov:
2276 	if (iov != iovstack)
2277 		kfree(iov);
2278 out:
2279 	return err;
2280 }
2281 
2282 /*
2283  *	BSD recvmsg interface
2284  */
2285 
2286 long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
2287 {
2288 	int fput_needed, err;
2289 	struct msghdr msg_sys;
2290 	struct socket *sock;
2291 
2292 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2293 	if (!sock)
2294 		goto out;
2295 
2296 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2297 
2298 	fput_light(sock->file, fput_needed);
2299 out:
2300 	return err;
2301 }
2302 
2303 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2304 		unsigned int, flags)
2305 {
2306 	if (flags & MSG_CMSG_COMPAT)
2307 		return -EINVAL;
2308 	return __sys_recvmsg(fd, msg, flags);
2309 }
2310 
2311 /*
2312  *     Linux recvmmsg interface
2313  */
2314 
2315 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2316 		   unsigned int flags, struct timespec *timeout)
2317 {
2318 	int fput_needed, err, datagrams;
2319 	struct socket *sock;
2320 	struct mmsghdr __user *entry;
2321 	struct compat_mmsghdr __user *compat_entry;
2322 	struct msghdr msg_sys;
2323 	struct timespec end_time;
2324 
2325 	if (timeout &&
2326 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2327 				    timeout->tv_nsec))
2328 		return -EINVAL;
2329 
2330 	datagrams = 0;
2331 
2332 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2333 	if (!sock)
2334 		return err;
2335 
2336 	err = sock_error(sock->sk);
2337 	if (err)
2338 		goto out_put;
2339 
2340 	entry = mmsg;
2341 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2342 
2343 	while (datagrams < vlen) {
2344 		/*
2345 		 * No need to ask LSM for more than the first datagram.
2346 		 */
2347 		if (MSG_CMSG_COMPAT & flags) {
2348 			err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2349 					     &msg_sys, flags & ~MSG_WAITFORONE,
2350 					     datagrams);
2351 			if (err < 0)
2352 				break;
2353 			err = __put_user(err, &compat_entry->msg_len);
2354 			++compat_entry;
2355 		} else {
2356 			err = ___sys_recvmsg(sock,
2357 					     (struct msghdr __user *)entry,
2358 					     &msg_sys, flags & ~MSG_WAITFORONE,
2359 					     datagrams);
2360 			if (err < 0)
2361 				break;
2362 			err = put_user(err, &entry->msg_len);
2363 			++entry;
2364 		}
2365 
2366 		if (err)
2367 			break;
2368 		++datagrams;
2369 
2370 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2371 		if (flags & MSG_WAITFORONE)
2372 			flags |= MSG_DONTWAIT;
2373 
2374 		if (timeout) {
2375 			ktime_get_ts(timeout);
2376 			*timeout = timespec_sub(end_time, *timeout);
2377 			if (timeout->tv_sec < 0) {
2378 				timeout->tv_sec = timeout->tv_nsec = 0;
2379 				break;
2380 			}
2381 
2382 			/* Timeout, return less than vlen datagrams */
2383 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2384 				break;
2385 		}
2386 
2387 		/* Out of band data, return right away */
2388 		if (msg_sys.msg_flags & MSG_OOB)
2389 			break;
2390 	}
2391 
2392 out_put:
2393 	fput_light(sock->file, fput_needed);
2394 
2395 	if (err == 0)
2396 		return datagrams;
2397 
2398 	if (datagrams != 0) {
2399 		/*
2400 		 * We may return less entries than requested (vlen) if the
2401 		 * sock is non block and there aren't enough datagrams...
2402 		 */
2403 		if (err != -EAGAIN) {
2404 			/*
2405 			 * ... or  if recvmsg returns an error after we
2406 			 * received some datagrams, where we record the
2407 			 * error to return on the next call or if the
2408 			 * app asks about it using getsockopt(SO_ERROR).
2409 			 */
2410 			sock->sk->sk_err = -err;
2411 		}
2412 
2413 		return datagrams;
2414 	}
2415 
2416 	return err;
2417 }
2418 
2419 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2420 		unsigned int, vlen, unsigned int, flags,
2421 		struct timespec __user *, timeout)
2422 {
2423 	int datagrams;
2424 	struct timespec timeout_sys;
2425 
2426 	if (flags & MSG_CMSG_COMPAT)
2427 		return -EINVAL;
2428 
2429 	if (!timeout)
2430 		return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2431 
2432 	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2433 		return -EFAULT;
2434 
2435 	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2436 
2437 	if (datagrams > 0 &&
2438 	    copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2439 		datagrams = -EFAULT;
2440 
2441 	return datagrams;
2442 }
2443 
2444 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2445 /* Argument list sizes for sys_socketcall */
2446 #define AL(x) ((x) * sizeof(unsigned long))
2447 static const unsigned char nargs[21] = {
2448 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2449 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2450 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2451 	AL(4), AL(5), AL(4)
2452 };
2453 
2454 #undef AL
2455 
2456 /*
2457  *	System call vectors.
2458  *
2459  *	Argument checking cleaned up. Saved 20% in size.
2460  *  This function doesn't need to set the kernel lock because
2461  *  it is set by the callees.
2462  */
2463 
2464 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2465 {
2466 	unsigned long a[AUDITSC_ARGS];
2467 	unsigned long a0, a1;
2468 	int err;
2469 	unsigned int len;
2470 
2471 	if (call < 1 || call > SYS_SENDMMSG)
2472 		return -EINVAL;
2473 
2474 	len = nargs[call];
2475 	if (len > sizeof(a))
2476 		return -EINVAL;
2477 
2478 	/* copy_from_user should be SMP safe. */
2479 	if (copy_from_user(a, args, len))
2480 		return -EFAULT;
2481 
2482 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2483 	if (err)
2484 		return err;
2485 
2486 	a0 = a[0];
2487 	a1 = a[1];
2488 
2489 	switch (call) {
2490 	case SYS_SOCKET:
2491 		err = sys_socket(a0, a1, a[2]);
2492 		break;
2493 	case SYS_BIND:
2494 		err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2495 		break;
2496 	case SYS_CONNECT:
2497 		err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2498 		break;
2499 	case SYS_LISTEN:
2500 		err = sys_listen(a0, a1);
2501 		break;
2502 	case SYS_ACCEPT:
2503 		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2504 				  (int __user *)a[2], 0);
2505 		break;
2506 	case SYS_GETSOCKNAME:
2507 		err =
2508 		    sys_getsockname(a0, (struct sockaddr __user *)a1,
2509 				    (int __user *)a[2]);
2510 		break;
2511 	case SYS_GETPEERNAME:
2512 		err =
2513 		    sys_getpeername(a0, (struct sockaddr __user *)a1,
2514 				    (int __user *)a[2]);
2515 		break;
2516 	case SYS_SOCKETPAIR:
2517 		err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2518 		break;
2519 	case SYS_SEND:
2520 		err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2521 		break;
2522 	case SYS_SENDTO:
2523 		err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2524 				 (struct sockaddr __user *)a[4], a[5]);
2525 		break;
2526 	case SYS_RECV:
2527 		err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2528 		break;
2529 	case SYS_RECVFROM:
2530 		err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2531 				   (struct sockaddr __user *)a[4],
2532 				   (int __user *)a[5]);
2533 		break;
2534 	case SYS_SHUTDOWN:
2535 		err = sys_shutdown(a0, a1);
2536 		break;
2537 	case SYS_SETSOCKOPT:
2538 		err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2539 		break;
2540 	case SYS_GETSOCKOPT:
2541 		err =
2542 		    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2543 				   (int __user *)a[4]);
2544 		break;
2545 	case SYS_SENDMSG:
2546 		err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2547 		break;
2548 	case SYS_SENDMMSG:
2549 		err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2550 		break;
2551 	case SYS_RECVMSG:
2552 		err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2553 		break;
2554 	case SYS_RECVMMSG:
2555 		err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2556 				   (struct timespec __user *)a[4]);
2557 		break;
2558 	case SYS_ACCEPT4:
2559 		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2560 				  (int __user *)a[2], a[3]);
2561 		break;
2562 	default:
2563 		err = -EINVAL;
2564 		break;
2565 	}
2566 	return err;
2567 }
2568 
2569 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
2570 
2571 /**
2572  *	sock_register - add a socket protocol handler
2573  *	@ops: description of protocol
2574  *
2575  *	This function is called by a protocol handler that wants to
2576  *	advertise its address family, and have it linked into the
2577  *	socket interface. The value ops->family coresponds to the
2578  *	socket system call protocol family.
2579  */
2580 int sock_register(const struct net_proto_family *ops)
2581 {
2582 	int err;
2583 
2584 	if (ops->family >= NPROTO) {
2585 		printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2586 		       NPROTO);
2587 		return -ENOBUFS;
2588 	}
2589 
2590 	spin_lock(&net_family_lock);
2591 	if (rcu_dereference_protected(net_families[ops->family],
2592 				      lockdep_is_held(&net_family_lock)))
2593 		err = -EEXIST;
2594 	else {
2595 		rcu_assign_pointer(net_families[ops->family], ops);
2596 		err = 0;
2597 	}
2598 	spin_unlock(&net_family_lock);
2599 
2600 	printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2601 	return err;
2602 }
2603 EXPORT_SYMBOL(sock_register);
2604 
2605 /**
2606  *	sock_unregister - remove a protocol handler
2607  *	@family: protocol family to remove
2608  *
2609  *	This function is called by a protocol handler that wants to
2610  *	remove its address family, and have it unlinked from the
2611  *	new socket creation.
2612  *
2613  *	If protocol handler is a module, then it can use module reference
2614  *	counts to protect against new references. If protocol handler is not
2615  *	a module then it needs to provide its own protection in
2616  *	the ops->create routine.
2617  */
2618 void sock_unregister(int family)
2619 {
2620 	BUG_ON(family < 0 || family >= NPROTO);
2621 
2622 	spin_lock(&net_family_lock);
2623 	RCU_INIT_POINTER(net_families[family], NULL);
2624 	spin_unlock(&net_family_lock);
2625 
2626 	synchronize_rcu();
2627 
2628 	printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2629 }
2630 EXPORT_SYMBOL(sock_unregister);
2631 
2632 static int __init sock_init(void)
2633 {
2634 	int err;
2635 	/*
2636 	 *      Initialize the network sysctl infrastructure.
2637 	 */
2638 	err = net_sysctl_init();
2639 	if (err)
2640 		goto out;
2641 
2642 	/*
2643 	 *      Initialize skbuff SLAB cache
2644 	 */
2645 	skb_init();
2646 
2647 	/*
2648 	 *      Initialize the protocols module.
2649 	 */
2650 
2651 	init_inodecache();
2652 
2653 	err = register_filesystem(&sock_fs_type);
2654 	if (err)
2655 		goto out_fs;
2656 	sock_mnt = kern_mount(&sock_fs_type);
2657 	if (IS_ERR(sock_mnt)) {
2658 		err = PTR_ERR(sock_mnt);
2659 		goto out_mount;
2660 	}
2661 
2662 	/* The real protocol initialization is performed in later initcalls.
2663 	 */
2664 
2665 #ifdef CONFIG_NETFILTER
2666 	err = netfilter_init();
2667 	if (err)
2668 		goto out;
2669 #endif
2670 
2671 #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2672 	skb_timestamping_init();
2673 #endif
2674 
2675 out:
2676 	return err;
2677 
2678 out_mount:
2679 	unregister_filesystem(&sock_fs_type);
2680 out_fs:
2681 	goto out;
2682 }
2683 
2684 core_initcall(sock_init);	/* early initcall */
2685 
2686 #ifdef CONFIG_PROC_FS
2687 void socket_seq_show(struct seq_file *seq)
2688 {
2689 	int cpu;
2690 	int counter = 0;
2691 
2692 	for_each_possible_cpu(cpu)
2693 	    counter += per_cpu(sockets_in_use, cpu);
2694 
2695 	/* It can be negative, by the way. 8) */
2696 	if (counter < 0)
2697 		counter = 0;
2698 
2699 	seq_printf(seq, "sockets: used %d\n", counter);
2700 }
2701 #endif				/* CONFIG_PROC_FS */
2702 
2703 #ifdef CONFIG_COMPAT
2704 static int do_siocgstamp(struct net *net, struct socket *sock,
2705 			 unsigned int cmd, void __user *up)
2706 {
2707 	mm_segment_t old_fs = get_fs();
2708 	struct timeval ktv;
2709 	int err;
2710 
2711 	set_fs(KERNEL_DS);
2712 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2713 	set_fs(old_fs);
2714 	if (!err)
2715 		err = compat_put_timeval(&ktv, up);
2716 
2717 	return err;
2718 }
2719 
2720 static int do_siocgstampns(struct net *net, struct socket *sock,
2721 			   unsigned int cmd, void __user *up)
2722 {
2723 	mm_segment_t old_fs = get_fs();
2724 	struct timespec kts;
2725 	int err;
2726 
2727 	set_fs(KERNEL_DS);
2728 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2729 	set_fs(old_fs);
2730 	if (!err)
2731 		err = compat_put_timespec(&kts, up);
2732 
2733 	return err;
2734 }
2735 
2736 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2737 {
2738 	struct ifreq __user *uifr;
2739 	int err;
2740 
2741 	uifr = compat_alloc_user_space(sizeof(struct ifreq));
2742 	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2743 		return -EFAULT;
2744 
2745 	err = dev_ioctl(net, SIOCGIFNAME, uifr);
2746 	if (err)
2747 		return err;
2748 
2749 	if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2750 		return -EFAULT;
2751 
2752 	return 0;
2753 }
2754 
2755 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2756 {
2757 	struct compat_ifconf ifc32;
2758 	struct ifconf ifc;
2759 	struct ifconf __user *uifc;
2760 	struct compat_ifreq __user *ifr32;
2761 	struct ifreq __user *ifr;
2762 	unsigned int i, j;
2763 	int err;
2764 
2765 	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2766 		return -EFAULT;
2767 
2768 	memset(&ifc, 0, sizeof(ifc));
2769 	if (ifc32.ifcbuf == 0) {
2770 		ifc32.ifc_len = 0;
2771 		ifc.ifc_len = 0;
2772 		ifc.ifc_req = NULL;
2773 		uifc = compat_alloc_user_space(sizeof(struct ifconf));
2774 	} else {
2775 		size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2776 			sizeof(struct ifreq);
2777 		uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2778 		ifc.ifc_len = len;
2779 		ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2780 		ifr32 = compat_ptr(ifc32.ifcbuf);
2781 		for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2782 			if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2783 				return -EFAULT;
2784 			ifr++;
2785 			ifr32++;
2786 		}
2787 	}
2788 	if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2789 		return -EFAULT;
2790 
2791 	err = dev_ioctl(net, SIOCGIFCONF, uifc);
2792 	if (err)
2793 		return err;
2794 
2795 	if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2796 		return -EFAULT;
2797 
2798 	ifr = ifc.ifc_req;
2799 	ifr32 = compat_ptr(ifc32.ifcbuf);
2800 	for (i = 0, j = 0;
2801 	     i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2802 	     i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2803 		if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2804 			return -EFAULT;
2805 		ifr32++;
2806 		ifr++;
2807 	}
2808 
2809 	if (ifc32.ifcbuf == 0) {
2810 		/* Translate from 64-bit structure multiple to
2811 		 * a 32-bit one.
2812 		 */
2813 		i = ifc.ifc_len;
2814 		i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2815 		ifc32.ifc_len = i;
2816 	} else {
2817 		ifc32.ifc_len = i;
2818 	}
2819 	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2820 		return -EFAULT;
2821 
2822 	return 0;
2823 }
2824 
2825 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2826 {
2827 	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2828 	bool convert_in = false, convert_out = false;
2829 	size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2830 	struct ethtool_rxnfc __user *rxnfc;
2831 	struct ifreq __user *ifr;
2832 	u32 rule_cnt = 0, actual_rule_cnt;
2833 	u32 ethcmd;
2834 	u32 data;
2835 	int ret;
2836 
2837 	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2838 		return -EFAULT;
2839 
2840 	compat_rxnfc = compat_ptr(data);
2841 
2842 	if (get_user(ethcmd, &compat_rxnfc->cmd))
2843 		return -EFAULT;
2844 
2845 	/* Most ethtool structures are defined without padding.
2846 	 * Unfortunately struct ethtool_rxnfc is an exception.
2847 	 */
2848 	switch (ethcmd) {
2849 	default:
2850 		break;
2851 	case ETHTOOL_GRXCLSRLALL:
2852 		/* Buffer size is variable */
2853 		if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2854 			return -EFAULT;
2855 		if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2856 			return -ENOMEM;
2857 		buf_size += rule_cnt * sizeof(u32);
2858 		/* fall through */
2859 	case ETHTOOL_GRXRINGS:
2860 	case ETHTOOL_GRXCLSRLCNT:
2861 	case ETHTOOL_GRXCLSRULE:
2862 	case ETHTOOL_SRXCLSRLINS:
2863 		convert_out = true;
2864 		/* fall through */
2865 	case ETHTOOL_SRXCLSRLDEL:
2866 		buf_size += sizeof(struct ethtool_rxnfc);
2867 		convert_in = true;
2868 		break;
2869 	}
2870 
2871 	ifr = compat_alloc_user_space(buf_size);
2872 	rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2873 
2874 	if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2875 		return -EFAULT;
2876 
2877 	if (put_user(convert_in ? rxnfc : compat_ptr(data),
2878 		     &ifr->ifr_ifru.ifru_data))
2879 		return -EFAULT;
2880 
2881 	if (convert_in) {
2882 		/* We expect there to be holes between fs.m_ext and
2883 		 * fs.ring_cookie and at the end of fs, but nowhere else.
2884 		 */
2885 		BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2886 			     sizeof(compat_rxnfc->fs.m_ext) !=
2887 			     offsetof(struct ethtool_rxnfc, fs.m_ext) +
2888 			     sizeof(rxnfc->fs.m_ext));
2889 		BUILD_BUG_ON(
2890 			offsetof(struct compat_ethtool_rxnfc, fs.location) -
2891 			offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2892 			offsetof(struct ethtool_rxnfc, fs.location) -
2893 			offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2894 
2895 		if (copy_in_user(rxnfc, compat_rxnfc,
2896 				 (void __user *)(&rxnfc->fs.m_ext + 1) -
2897 				 (void __user *)rxnfc) ||
2898 		    copy_in_user(&rxnfc->fs.ring_cookie,
2899 				 &compat_rxnfc->fs.ring_cookie,
2900 				 (void __user *)(&rxnfc->fs.location + 1) -
2901 				 (void __user *)&rxnfc->fs.ring_cookie) ||
2902 		    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2903 				 sizeof(rxnfc->rule_cnt)))
2904 			return -EFAULT;
2905 	}
2906 
2907 	ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2908 	if (ret)
2909 		return ret;
2910 
2911 	if (convert_out) {
2912 		if (copy_in_user(compat_rxnfc, rxnfc,
2913 				 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2914 				 (const void __user *)rxnfc) ||
2915 		    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2916 				 &rxnfc->fs.ring_cookie,
2917 				 (const void __user *)(&rxnfc->fs.location + 1) -
2918 				 (const void __user *)&rxnfc->fs.ring_cookie) ||
2919 		    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2920 				 sizeof(rxnfc->rule_cnt)))
2921 			return -EFAULT;
2922 
2923 		if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2924 			/* As an optimisation, we only copy the actual
2925 			 * number of rules that the underlying
2926 			 * function returned.  Since Mallory might
2927 			 * change the rule count in user memory, we
2928 			 * check that it is less than the rule count
2929 			 * originally given (as the user buffer size),
2930 			 * which has been range-checked.
2931 			 */
2932 			if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2933 				return -EFAULT;
2934 			if (actual_rule_cnt < rule_cnt)
2935 				rule_cnt = actual_rule_cnt;
2936 			if (copy_in_user(&compat_rxnfc->rule_locs[0],
2937 					 &rxnfc->rule_locs[0],
2938 					 rule_cnt * sizeof(u32)))
2939 				return -EFAULT;
2940 		}
2941 	}
2942 
2943 	return 0;
2944 }
2945 
2946 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2947 {
2948 	void __user *uptr;
2949 	compat_uptr_t uptr32;
2950 	struct ifreq __user *uifr;
2951 
2952 	uifr = compat_alloc_user_space(sizeof(*uifr));
2953 	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2954 		return -EFAULT;
2955 
2956 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2957 		return -EFAULT;
2958 
2959 	uptr = compat_ptr(uptr32);
2960 
2961 	if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2962 		return -EFAULT;
2963 
2964 	return dev_ioctl(net, SIOCWANDEV, uifr);
2965 }
2966 
2967 static int bond_ioctl(struct net *net, unsigned int cmd,
2968 			 struct compat_ifreq __user *ifr32)
2969 {
2970 	struct ifreq kifr;
2971 	struct ifreq __user *uifr;
2972 	mm_segment_t old_fs;
2973 	int err;
2974 	u32 data;
2975 	void __user *datap;
2976 
2977 	switch (cmd) {
2978 	case SIOCBONDENSLAVE:
2979 	case SIOCBONDRELEASE:
2980 	case SIOCBONDSETHWADDR:
2981 	case SIOCBONDCHANGEACTIVE:
2982 		if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2983 			return -EFAULT;
2984 
2985 		old_fs = get_fs();
2986 		set_fs(KERNEL_DS);
2987 		err = dev_ioctl(net, cmd,
2988 				(struct ifreq __user __force *) &kifr);
2989 		set_fs(old_fs);
2990 
2991 		return err;
2992 	case SIOCBONDSLAVEINFOQUERY:
2993 	case SIOCBONDINFOQUERY:
2994 		uifr = compat_alloc_user_space(sizeof(*uifr));
2995 		if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2996 			return -EFAULT;
2997 
2998 		if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2999 			return -EFAULT;
3000 
3001 		datap = compat_ptr(data);
3002 		if (put_user(datap, &uifr->ifr_ifru.ifru_data))
3003 			return -EFAULT;
3004 
3005 		return dev_ioctl(net, cmd, uifr);
3006 	default:
3007 		return -ENOIOCTLCMD;
3008 	}
3009 }
3010 
3011 static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
3012 				 struct compat_ifreq __user *u_ifreq32)
3013 {
3014 	struct ifreq __user *u_ifreq64;
3015 	char tmp_buf[IFNAMSIZ];
3016 	void __user *data64;
3017 	u32 data32;
3018 
3019 	if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3020 			   IFNAMSIZ))
3021 		return -EFAULT;
3022 	if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3023 		return -EFAULT;
3024 	data64 = compat_ptr(data32);
3025 
3026 	u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3027 
3028 	/* Don't check these user accesses, just let that get trapped
3029 	 * in the ioctl handler instead.
3030 	 */
3031 	if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3032 			 IFNAMSIZ))
3033 		return -EFAULT;
3034 	if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3035 		return -EFAULT;
3036 
3037 	return dev_ioctl(net, cmd, u_ifreq64);
3038 }
3039 
3040 static int dev_ifsioc(struct net *net, struct socket *sock,
3041 			 unsigned int cmd, struct compat_ifreq __user *uifr32)
3042 {
3043 	struct ifreq __user *uifr;
3044 	int err;
3045 
3046 	uifr = compat_alloc_user_space(sizeof(*uifr));
3047 	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3048 		return -EFAULT;
3049 
3050 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3051 
3052 	if (!err) {
3053 		switch (cmd) {
3054 		case SIOCGIFFLAGS:
3055 		case SIOCGIFMETRIC:
3056 		case SIOCGIFMTU:
3057 		case SIOCGIFMEM:
3058 		case SIOCGIFHWADDR:
3059 		case SIOCGIFINDEX:
3060 		case SIOCGIFADDR:
3061 		case SIOCGIFBRDADDR:
3062 		case SIOCGIFDSTADDR:
3063 		case SIOCGIFNETMASK:
3064 		case SIOCGIFPFLAGS:
3065 		case SIOCGIFTXQLEN:
3066 		case SIOCGMIIPHY:
3067 		case SIOCGMIIREG:
3068 			if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3069 				err = -EFAULT;
3070 			break;
3071 		}
3072 	}
3073 	return err;
3074 }
3075 
3076 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3077 			struct compat_ifreq __user *uifr32)
3078 {
3079 	struct ifreq ifr;
3080 	struct compat_ifmap __user *uifmap32;
3081 	mm_segment_t old_fs;
3082 	int err;
3083 
3084 	uifmap32 = &uifr32->ifr_ifru.ifru_map;
3085 	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3086 	err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3087 	err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3088 	err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3089 	err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3090 	err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3091 	err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3092 	if (err)
3093 		return -EFAULT;
3094 
3095 	old_fs = get_fs();
3096 	set_fs(KERNEL_DS);
3097 	err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3098 	set_fs(old_fs);
3099 
3100 	if (cmd == SIOCGIFMAP && !err) {
3101 		err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3102 		err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3103 		err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3104 		err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3105 		err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3106 		err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3107 		err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3108 		if (err)
3109 			err = -EFAULT;
3110 	}
3111 	return err;
3112 }
3113 
3114 static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3115 {
3116 	void __user *uptr;
3117 	compat_uptr_t uptr32;
3118 	struct ifreq __user *uifr;
3119 
3120 	uifr = compat_alloc_user_space(sizeof(*uifr));
3121 	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3122 		return -EFAULT;
3123 
3124 	if (get_user(uptr32, &uifr32->ifr_data))
3125 		return -EFAULT;
3126 
3127 	uptr = compat_ptr(uptr32);
3128 
3129 	if (put_user(uptr, &uifr->ifr_data))
3130 		return -EFAULT;
3131 
3132 	return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3133 }
3134 
3135 struct rtentry32 {
3136 	u32		rt_pad1;
3137 	struct sockaddr rt_dst;         /* target address               */
3138 	struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3139 	struct sockaddr rt_genmask;     /* target network mask (IP)     */
3140 	unsigned short	rt_flags;
3141 	short		rt_pad2;
3142 	u32		rt_pad3;
3143 	unsigned char	rt_tos;
3144 	unsigned char	rt_class;
3145 	short		rt_pad4;
3146 	short		rt_metric;      /* +1 for binary compatibility! */
3147 	/* char * */ u32 rt_dev;        /* forcing the device at add    */
3148 	u32		rt_mtu;         /* per route MTU/Window         */
3149 	u32		rt_window;      /* Window clamping              */
3150 	unsigned short  rt_irtt;        /* Initial RTT                  */
3151 };
3152 
3153 struct in6_rtmsg32 {
3154 	struct in6_addr		rtmsg_dst;
3155 	struct in6_addr		rtmsg_src;
3156 	struct in6_addr		rtmsg_gateway;
3157 	u32			rtmsg_type;
3158 	u16			rtmsg_dst_len;
3159 	u16			rtmsg_src_len;
3160 	u32			rtmsg_metric;
3161 	u32			rtmsg_info;
3162 	u32			rtmsg_flags;
3163 	s32			rtmsg_ifindex;
3164 };
3165 
3166 static int routing_ioctl(struct net *net, struct socket *sock,
3167 			 unsigned int cmd, void __user *argp)
3168 {
3169 	int ret;
3170 	void *r = NULL;
3171 	struct in6_rtmsg r6;
3172 	struct rtentry r4;
3173 	char devname[16];
3174 	u32 rtdev;
3175 	mm_segment_t old_fs = get_fs();
3176 
3177 	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3178 		struct in6_rtmsg32 __user *ur6 = argp;
3179 		ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3180 			3 * sizeof(struct in6_addr));
3181 		ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3182 		ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3183 		ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3184 		ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3185 		ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3186 		ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3187 		ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3188 
3189 		r = (void *) &r6;
3190 	} else { /* ipv4 */
3191 		struct rtentry32 __user *ur4 = argp;
3192 		ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3193 					3 * sizeof(struct sockaddr));
3194 		ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3195 		ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3196 		ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3197 		ret |= get_user(r4.rt_window, &(ur4->rt_window));
3198 		ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3199 		ret |= get_user(rtdev, &(ur4->rt_dev));
3200 		if (rtdev) {
3201 			ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3202 			r4.rt_dev = (char __user __force *)devname;
3203 			devname[15] = 0;
3204 		} else
3205 			r4.rt_dev = NULL;
3206 
3207 		r = (void *) &r4;
3208 	}
3209 
3210 	if (ret) {
3211 		ret = -EFAULT;
3212 		goto out;
3213 	}
3214 
3215 	set_fs(KERNEL_DS);
3216 	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3217 	set_fs(old_fs);
3218 
3219 out:
3220 	return ret;
3221 }
3222 
3223 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3224  * for some operations; this forces use of the newer bridge-utils that
3225  * use compatible ioctls
3226  */
3227 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3228 {
3229 	compat_ulong_t tmp;
3230 
3231 	if (get_user(tmp, argp))
3232 		return -EFAULT;
3233 	if (tmp == BRCTL_GET_VERSION)
3234 		return BRCTL_VERSION + 1;
3235 	return -EINVAL;
3236 }
3237 
3238 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3239 			 unsigned int cmd, unsigned long arg)
3240 {
3241 	void __user *argp = compat_ptr(arg);
3242 	struct sock *sk = sock->sk;
3243 	struct net *net = sock_net(sk);
3244 
3245 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3246 		return siocdevprivate_ioctl(net, cmd, argp);
3247 
3248 	switch (cmd) {
3249 	case SIOCSIFBR:
3250 	case SIOCGIFBR:
3251 		return old_bridge_ioctl(argp);
3252 	case SIOCGIFNAME:
3253 		return dev_ifname32(net, argp);
3254 	case SIOCGIFCONF:
3255 		return dev_ifconf(net, argp);
3256 	case SIOCETHTOOL:
3257 		return ethtool_ioctl(net, argp);
3258 	case SIOCWANDEV:
3259 		return compat_siocwandev(net, argp);
3260 	case SIOCGIFMAP:
3261 	case SIOCSIFMAP:
3262 		return compat_sioc_ifmap(net, cmd, argp);
3263 	case SIOCBONDENSLAVE:
3264 	case SIOCBONDRELEASE:
3265 	case SIOCBONDSETHWADDR:
3266 	case SIOCBONDSLAVEINFOQUERY:
3267 	case SIOCBONDINFOQUERY:
3268 	case SIOCBONDCHANGEACTIVE:
3269 		return bond_ioctl(net, cmd, argp);
3270 	case SIOCADDRT:
3271 	case SIOCDELRT:
3272 		return routing_ioctl(net, sock, cmd, argp);
3273 	case SIOCGSTAMP:
3274 		return do_siocgstamp(net, sock, cmd, argp);
3275 	case SIOCGSTAMPNS:
3276 		return do_siocgstampns(net, sock, cmd, argp);
3277 	case SIOCSHWTSTAMP:
3278 		return compat_siocshwtstamp(net, argp);
3279 
3280 	case FIOSETOWN:
3281 	case SIOCSPGRP:
3282 	case FIOGETOWN:
3283 	case SIOCGPGRP:
3284 	case SIOCBRADDBR:
3285 	case SIOCBRDELBR:
3286 	case SIOCGIFVLAN:
3287 	case SIOCSIFVLAN:
3288 	case SIOCADDDLCI:
3289 	case SIOCDELDLCI:
3290 		return sock_ioctl(file, cmd, arg);
3291 
3292 	case SIOCGIFFLAGS:
3293 	case SIOCSIFFLAGS:
3294 	case SIOCGIFMETRIC:
3295 	case SIOCSIFMETRIC:
3296 	case SIOCGIFMTU:
3297 	case SIOCSIFMTU:
3298 	case SIOCGIFMEM:
3299 	case SIOCSIFMEM:
3300 	case SIOCGIFHWADDR:
3301 	case SIOCSIFHWADDR:
3302 	case SIOCADDMULTI:
3303 	case SIOCDELMULTI:
3304 	case SIOCGIFINDEX:
3305 	case SIOCGIFADDR:
3306 	case SIOCSIFADDR:
3307 	case SIOCSIFHWBROADCAST:
3308 	case SIOCDIFADDR:
3309 	case SIOCGIFBRDADDR:
3310 	case SIOCSIFBRDADDR:
3311 	case SIOCGIFDSTADDR:
3312 	case SIOCSIFDSTADDR:
3313 	case SIOCGIFNETMASK:
3314 	case SIOCSIFNETMASK:
3315 	case SIOCSIFPFLAGS:
3316 	case SIOCGIFPFLAGS:
3317 	case SIOCGIFTXQLEN:
3318 	case SIOCSIFTXQLEN:
3319 	case SIOCBRADDIF:
3320 	case SIOCBRDELIF:
3321 	case SIOCSIFNAME:
3322 	case SIOCGMIIPHY:
3323 	case SIOCGMIIREG:
3324 	case SIOCSMIIREG:
3325 		return dev_ifsioc(net, sock, cmd, argp);
3326 
3327 	case SIOCSARP:
3328 	case SIOCGARP:
3329 	case SIOCDARP:
3330 	case SIOCATMARK:
3331 		return sock_do_ioctl(net, sock, cmd, arg);
3332 	}
3333 
3334 	return -ENOIOCTLCMD;
3335 }
3336 
3337 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3338 			      unsigned long arg)
3339 {
3340 	struct socket *sock = file->private_data;
3341 	int ret = -ENOIOCTLCMD;
3342 	struct sock *sk;
3343 	struct net *net;
3344 
3345 	sk = sock->sk;
3346 	net = sock_net(sk);
3347 
3348 	if (sock->ops->compat_ioctl)
3349 		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3350 
3351 	if (ret == -ENOIOCTLCMD &&
3352 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3353 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3354 
3355 	if (ret == -ENOIOCTLCMD)
3356 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3357 
3358 	return ret;
3359 }
3360 #endif
3361 
3362 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3363 {
3364 	return sock->ops->bind(sock, addr, addrlen);
3365 }
3366 EXPORT_SYMBOL(kernel_bind);
3367 
3368 int kernel_listen(struct socket *sock, int backlog)
3369 {
3370 	return sock->ops->listen(sock, backlog);
3371 }
3372 EXPORT_SYMBOL(kernel_listen);
3373 
3374 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3375 {
3376 	struct sock *sk = sock->sk;
3377 	int err;
3378 
3379 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3380 			       newsock);
3381 	if (err < 0)
3382 		goto done;
3383 
3384 	err = sock->ops->accept(sock, *newsock, flags);
3385 	if (err < 0) {
3386 		sock_release(*newsock);
3387 		*newsock = NULL;
3388 		goto done;
3389 	}
3390 
3391 	(*newsock)->ops = sock->ops;
3392 	__module_get((*newsock)->ops->owner);
3393 
3394 done:
3395 	return err;
3396 }
3397 EXPORT_SYMBOL(kernel_accept);
3398 
3399 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3400 		   int flags)
3401 {
3402 	return sock->ops->connect(sock, addr, addrlen, flags);
3403 }
3404 EXPORT_SYMBOL(kernel_connect);
3405 
3406 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3407 			 int *addrlen)
3408 {
3409 	return sock->ops->getname(sock, addr, addrlen, 0);
3410 }
3411 EXPORT_SYMBOL(kernel_getsockname);
3412 
3413 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3414 			 int *addrlen)
3415 {
3416 	return sock->ops->getname(sock, addr, addrlen, 1);
3417 }
3418 EXPORT_SYMBOL(kernel_getpeername);
3419 
3420 int kernel_getsockopt(struct socket *sock, int level, int optname,
3421 			char *optval, int *optlen)
3422 {
3423 	mm_segment_t oldfs = get_fs();
3424 	char __user *uoptval;
3425 	int __user *uoptlen;
3426 	int err;
3427 
3428 	uoptval = (char __user __force *) optval;
3429 	uoptlen = (int __user __force *) optlen;
3430 
3431 	set_fs(KERNEL_DS);
3432 	if (level == SOL_SOCKET)
3433 		err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3434 	else
3435 		err = sock->ops->getsockopt(sock, level, optname, uoptval,
3436 					    uoptlen);
3437 	set_fs(oldfs);
3438 	return err;
3439 }
3440 EXPORT_SYMBOL(kernel_getsockopt);
3441 
3442 int kernel_setsockopt(struct socket *sock, int level, int optname,
3443 			char *optval, unsigned int optlen)
3444 {
3445 	mm_segment_t oldfs = get_fs();
3446 	char __user *uoptval;
3447 	int err;
3448 
3449 	uoptval = (char __user __force *) optval;
3450 
3451 	set_fs(KERNEL_DS);
3452 	if (level == SOL_SOCKET)
3453 		err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3454 	else
3455 		err = sock->ops->setsockopt(sock, level, optname, uoptval,
3456 					    optlen);
3457 	set_fs(oldfs);
3458 	return err;
3459 }
3460 EXPORT_SYMBOL(kernel_setsockopt);
3461 
3462 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3463 		    size_t size, int flags)
3464 {
3465 	if (sock->ops->sendpage)
3466 		return sock->ops->sendpage(sock, page, offset, size, flags);
3467 
3468 	return sock_no_sendpage(sock, page, offset, size, flags);
3469 }
3470 EXPORT_SYMBOL(kernel_sendpage);
3471 
3472 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3473 {
3474 	mm_segment_t oldfs = get_fs();
3475 	int err;
3476 
3477 	set_fs(KERNEL_DS);
3478 	err = sock->ops->ioctl(sock, cmd, arg);
3479 	set_fs(oldfs);
3480 
3481 	return err;
3482 }
3483 EXPORT_SYMBOL(kernel_sock_ioctl);
3484 
3485 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3486 {
3487 	return sock->ops->shutdown(sock, how);
3488 }
3489 EXPORT_SYMBOL(kernel_sock_shutdown);
3490