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/ptp_classify.h>
76#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
86#include <linux/audit.h>
87#include <linux/wireless.h>
88#include <linux/nsproxy.h>
89#include <linux/magic.h>
90#include <linux/slab.h>
91#include <linux/xattr.h>
92
93#include <asm/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
97#include <net/wext.h>
98#include <net/cls_cgroup.h>
99
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
106#include <linux/sockios.h>
107#include <linux/atalk.h>
108#include <net/busy_poll.h>
109#include <linux/errqueue.h>
110
111#ifdef CONFIG_NET_RX_BUSY_POLL
112unsigned int sysctl_net_busy_read __read_mostly;
113unsigned int sysctl_net_busy_poll __read_mostly;
114#endif
115
116static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119
120static int sock_close(struct inode *inode, struct file *file);
121static unsigned int sock_poll(struct file *file,
122			      struct poll_table_struct *wait);
123static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124#ifdef CONFIG_COMPAT
125static long compat_sock_ioctl(struct file *file,
126			      unsigned int cmd, unsigned long arg);
127#endif
128static int sock_fasync(int fd, struct file *filp, int on);
129static ssize_t sock_sendpage(struct file *file, struct page *page,
130			     int offset, size_t size, loff_t *ppos, int more);
131static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132				struct pipe_inode_info *pipe, size_t len,
133				unsigned int flags);
134
135/*
136 *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 *	in the operation structures but are done directly via the socketcall() multiplexor.
138 */
139
140static const struct file_operations socket_file_ops = {
141	.owner =	THIS_MODULE,
142	.llseek =	no_llseek,
143	.read_iter =	sock_read_iter,
144	.write_iter =	sock_write_iter,
145	.poll =		sock_poll,
146	.unlocked_ioctl = sock_ioctl,
147#ifdef CONFIG_COMPAT
148	.compat_ioctl = compat_sock_ioctl,
149#endif
150	.mmap =		sock_mmap,
151	.release =	sock_close,
152	.fasync =	sock_fasync,
153	.sendpage =	sock_sendpage,
154	.splice_write = generic_splice_sendpage,
155	.splice_read =	sock_splice_read,
156};
157
158/*
159 *	The protocol list. Each protocol is registered in here.
160 */
161
162static DEFINE_SPINLOCK(net_family_lock);
163static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164
165/*
166 *	Statistics counters of the socket lists
167 */
168
169static DEFINE_PER_CPU(int, sockets_in_use);
170
171/*
172 * Support routines.
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
175 */
176
177/**
178 *	move_addr_to_kernel	-	copy a socket address into kernel space
179 *	@uaddr: Address in user space
180 *	@kaddr: Address in kernel space
181 *	@ulen: Length in user space
182 *
183 *	The address is copied into kernel space. If the provided address is
184 *	too long an error code of -EINVAL is returned. If the copy gives
185 *	invalid addresses -EFAULT is returned. On a success 0 is returned.
186 */
187
188int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
189{
190	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191		return -EINVAL;
192	if (ulen == 0)
193		return 0;
194	if (copy_from_user(kaddr, uaddr, ulen))
195		return -EFAULT;
196	return audit_sockaddr(ulen, kaddr);
197}
198
199/**
200 *	move_addr_to_user	-	copy an address to user space
201 *	@kaddr: kernel space address
202 *	@klen: length of address in kernel
203 *	@uaddr: user space address
204 *	@ulen: pointer to user length field
205 *
206 *	The value pointed to by ulen on entry is the buffer length available.
207 *	This is overwritten with the buffer space used. -EINVAL is returned
208 *	if an overlong buffer is specified or a negative buffer size. -EFAULT
209 *	is returned if either the buffer or the length field are not
210 *	accessible.
211 *	After copying the data up to the limit the user specifies, the true
212 *	length of the data is written over the length limit the user
213 *	specified. Zero is returned for a success.
214 */
215
216static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217			     void __user *uaddr, int __user *ulen)
218{
219	int err;
220	int len;
221
222	BUG_ON(klen > sizeof(struct sockaddr_storage));
223	err = get_user(len, ulen);
224	if (err)
225		return err;
226	if (len > klen)
227		len = klen;
228	if (len < 0)
229		return -EINVAL;
230	if (len) {
231		if (audit_sockaddr(klen, kaddr))
232			return -ENOMEM;
233		if (copy_to_user(uaddr, kaddr, len))
234			return -EFAULT;
235	}
236	/*
237	 *      "fromlen shall refer to the value before truncation.."
238	 *                      1003.1g
239	 */
240	return __put_user(klen, ulen);
241}
242
243static struct kmem_cache *sock_inode_cachep __read_mostly;
244
245static struct inode *sock_alloc_inode(struct super_block *sb)
246{
247	struct socket_alloc *ei;
248	struct socket_wq *wq;
249
250	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
251	if (!ei)
252		return NULL;
253	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
254	if (!wq) {
255		kmem_cache_free(sock_inode_cachep, ei);
256		return NULL;
257	}
258	init_waitqueue_head(&wq->wait);
259	wq->fasync_list = NULL;
260	wq->flags = 0;
261	RCU_INIT_POINTER(ei->socket.wq, wq);
262
263	ei->socket.state = SS_UNCONNECTED;
264	ei->socket.flags = 0;
265	ei->socket.ops = NULL;
266	ei->socket.sk = NULL;
267	ei->socket.file = NULL;
268
269	return &ei->vfs_inode;
270}
271
272static void sock_destroy_inode(struct inode *inode)
273{
274	struct socket_alloc *ei;
275	struct socket_wq *wq;
276
277	ei = container_of(inode, struct socket_alloc, vfs_inode);
278	wq = rcu_dereference_protected(ei->socket.wq, 1);
279	kfree_rcu(wq, rcu);
280	kmem_cache_free(sock_inode_cachep, ei);
281}
282
283static void init_once(void *foo)
284{
285	struct socket_alloc *ei = (struct socket_alloc *)foo;
286
287	inode_init_once(&ei->vfs_inode);
288}
289
290static int init_inodecache(void)
291{
292	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293					      sizeof(struct socket_alloc),
294					      0,
295					      (SLAB_HWCACHE_ALIGN |
296					       SLAB_RECLAIM_ACCOUNT |
297					       SLAB_MEM_SPREAD),
298					      init_once);
299	if (sock_inode_cachep == NULL)
300		return -ENOMEM;
301	return 0;
302}
303
304static const struct super_operations sockfs_ops = {
305	.alloc_inode	= sock_alloc_inode,
306	.destroy_inode	= sock_destroy_inode,
307	.statfs		= simple_statfs,
308};
309
310/*
311 * sockfs_dname() is called from d_path().
312 */
313static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314{
315	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316				d_inode(dentry)->i_ino);
317}
318
319static const struct dentry_operations sockfs_dentry_operations = {
320	.d_dname  = sockfs_dname,
321};
322
323static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324			 int flags, const char *dev_name, void *data)
325{
326	return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327		&sockfs_dentry_operations, SOCKFS_MAGIC);
328}
329
330static struct vfsmount *sock_mnt __read_mostly;
331
332static struct file_system_type sock_fs_type = {
333	.name =		"sockfs",
334	.mount =	sockfs_mount,
335	.kill_sb =	kill_anon_super,
336};
337
338/*
339 *	Obtains the first available file descriptor and sets it up for use.
340 *
341 *	These functions create file structures and maps them to fd space
342 *	of the current process. On success it returns file descriptor
343 *	and file struct implicitly stored in sock->file.
344 *	Note that another thread may close file descriptor before we return
345 *	from this function. We use the fact that now we do not refer
346 *	to socket after mapping. If one day we will need it, this
347 *	function will increment ref. count on file by 1.
348 *
349 *	In any case returned fd MAY BE not valid!
350 *	This race condition is unavoidable
351 *	with shared fd spaces, we cannot solve it inside kernel,
352 *	but we take care of internal coherence yet.
353 */
354
355struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
356{
357	struct qstr name = { .name = "" };
358	struct path path;
359	struct file *file;
360
361	if (dname) {
362		name.name = dname;
363		name.len = strlen(name.name);
364	} else if (sock->sk) {
365		name.name = sock->sk->sk_prot_creator->name;
366		name.len = strlen(name.name);
367	}
368	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
369	if (unlikely(!path.dentry))
370		return ERR_PTR(-ENOMEM);
371	path.mnt = mntget(sock_mnt);
372
373	d_instantiate(path.dentry, SOCK_INODE(sock));
374
375	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
376		  &socket_file_ops);
377	if (IS_ERR(file)) {
378		/* drop dentry, keep inode */
379		ihold(d_inode(path.dentry));
380		path_put(&path);
381		return file;
382	}
383
384	sock->file = file;
385	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
386	file->private_data = sock;
387	return file;
388}
389EXPORT_SYMBOL(sock_alloc_file);
390
391static int sock_map_fd(struct socket *sock, int flags)
392{
393	struct file *newfile;
394	int fd = get_unused_fd_flags(flags);
395	if (unlikely(fd < 0))
396		return fd;
397
398	newfile = sock_alloc_file(sock, flags, NULL);
399	if (likely(!IS_ERR(newfile))) {
400		fd_install(fd, newfile);
401		return fd;
402	}
403
404	put_unused_fd(fd);
405	return PTR_ERR(newfile);
406}
407
408struct socket *sock_from_file(struct file *file, int *err)
409{
410	if (file->f_op == &socket_file_ops)
411		return file->private_data;	/* set in sock_map_fd */
412
413	*err = -ENOTSOCK;
414	return NULL;
415}
416EXPORT_SYMBOL(sock_from_file);
417
418/**
419 *	sockfd_lookup - Go from a file number to its socket slot
420 *	@fd: file handle
421 *	@err: pointer to an error code return
422 *
423 *	The file handle passed in is locked and the socket it is bound
424 *	too is returned. If an error occurs the err pointer is overwritten
425 *	with a negative errno code and NULL is returned. The function checks
426 *	for both invalid handles and passing a handle which is not a socket.
427 *
428 *	On a success the socket object pointer is returned.
429 */
430
431struct socket *sockfd_lookup(int fd, int *err)
432{
433	struct file *file;
434	struct socket *sock;
435
436	file = fget(fd);
437	if (!file) {
438		*err = -EBADF;
439		return NULL;
440	}
441
442	sock = sock_from_file(file, err);
443	if (!sock)
444		fput(file);
445	return sock;
446}
447EXPORT_SYMBOL(sockfd_lookup);
448
449static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
450{
451	struct fd f = fdget(fd);
452	struct socket *sock;
453
454	*err = -EBADF;
455	if (f.file) {
456		sock = sock_from_file(f.file, err);
457		if (likely(sock)) {
458			*fput_needed = f.flags;
459			return sock;
460		}
461		fdput(f);
462	}
463	return NULL;
464}
465
466#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469static ssize_t sockfs_getxattr(struct dentry *dentry,
470			       const char *name, void *value, size_t size)
471{
472	const char *proto_name;
473	size_t proto_size;
474	int error;
475
476	error = -ENODATA;
477	if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478		proto_name = dentry->d_name.name;
479		proto_size = strlen(proto_name);
480
481		if (value) {
482			error = -ERANGE;
483			if (proto_size + 1 > size)
484				goto out;
485
486			strncpy(value, proto_name, proto_size + 1);
487		}
488		error = proto_size + 1;
489	}
490
491out:
492	return error;
493}
494
495static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496				size_t size)
497{
498	ssize_t len;
499	ssize_t used = 0;
500
501	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
502	if (len < 0)
503		return len;
504	used += len;
505	if (buffer) {
506		if (size < used)
507			return -ERANGE;
508		buffer += len;
509	}
510
511	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512	used += len;
513	if (buffer) {
514		if (size < used)
515			return -ERANGE;
516		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517		buffer += len;
518	}
519
520	return used;
521}
522
523static const struct inode_operations sockfs_inode_ops = {
524	.getxattr = sockfs_getxattr,
525	.listxattr = sockfs_listxattr,
526};
527
528/**
529 *	sock_alloc	-	allocate a socket
530 *
531 *	Allocate a new inode and socket object. The two are bound together
532 *	and initialised. The socket is then returned. If we are out of inodes
533 *	NULL is returned.
534 */
535
536static struct socket *sock_alloc(void)
537{
538	struct inode *inode;
539	struct socket *sock;
540
541	inode = new_inode_pseudo(sock_mnt->mnt_sb);
542	if (!inode)
543		return NULL;
544
545	sock = SOCKET_I(inode);
546
547	kmemcheck_annotate_bitfield(sock, type);
548	inode->i_ino = get_next_ino();
549	inode->i_mode = S_IFSOCK | S_IRWXUGO;
550	inode->i_uid = current_fsuid();
551	inode->i_gid = current_fsgid();
552	inode->i_op = &sockfs_inode_ops;
553
554	this_cpu_add(sockets_in_use, 1);
555	return sock;
556}
557
558/**
559 *	sock_release	-	close a socket
560 *	@sock: socket to close
561 *
562 *	The socket is released from the protocol stack if it has a release
563 *	callback, and the inode is then released if the socket is bound to
564 *	an inode not a file.
565 */
566
567void sock_release(struct socket *sock)
568{
569	if (sock->ops) {
570		struct module *owner = sock->ops->owner;
571
572		sock->ops->release(sock);
573		sock->ops = NULL;
574		module_put(owner);
575	}
576
577	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
578		pr_err("%s: fasync list not empty!\n", __func__);
579
580	this_cpu_sub(sockets_in_use, 1);
581	if (!sock->file) {
582		iput(SOCK_INODE(sock));
583		return;
584	}
585	sock->file = NULL;
586}
587EXPORT_SYMBOL(sock_release);
588
589void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
590{
591	u8 flags = *tx_flags;
592
593	if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
594		flags |= SKBTX_HW_TSTAMP;
595
596	if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
597		flags |= SKBTX_SW_TSTAMP;
598
599	if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
600		flags |= SKBTX_SCHED_TSTAMP;
601
602	if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
603		flags |= SKBTX_ACK_TSTAMP;
604
605	*tx_flags = flags;
606}
607EXPORT_SYMBOL(__sock_tx_timestamp);
608
609static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
610{
611	int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
612	BUG_ON(ret == -EIOCBQUEUED);
613	return ret;
614}
615
616int sock_sendmsg(struct socket *sock, struct msghdr *msg)
617{
618	int err = security_socket_sendmsg(sock, msg,
619					  msg_data_left(msg));
620
621	return err ?: sock_sendmsg_nosec(sock, msg);
622}
623EXPORT_SYMBOL(sock_sendmsg);
624
625int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
626		   struct kvec *vec, size_t num, size_t size)
627{
628	iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
629	return sock_sendmsg(sock, msg);
630}
631EXPORT_SYMBOL(kernel_sendmsg);
632
633/*
634 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
635 */
636void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
637	struct sk_buff *skb)
638{
639	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
640	struct scm_timestamping tss;
641	int empty = 1;
642	struct skb_shared_hwtstamps *shhwtstamps =
643		skb_hwtstamps(skb);
644
645	/* Race occurred between timestamp enabling and packet
646	   receiving.  Fill in the current time for now. */
647	if (need_software_tstamp && skb->tstamp.tv64 == 0)
648		__net_timestamp(skb);
649
650	if (need_software_tstamp) {
651		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
652			struct timeval tv;
653			skb_get_timestamp(skb, &tv);
654			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
655				 sizeof(tv), &tv);
656		} else {
657			struct timespec ts;
658			skb_get_timestampns(skb, &ts);
659			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
660				 sizeof(ts), &ts);
661		}
662	}
663
664	memset(&tss, 0, sizeof(tss));
665	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
666	    ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
667		empty = 0;
668	if (shhwtstamps &&
669	    (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
670	    ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
671		empty = 0;
672	if (!empty)
673		put_cmsg(msg, SOL_SOCKET,
674			 SCM_TIMESTAMPING, sizeof(tss), &tss);
675}
676EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
677
678void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
679	struct sk_buff *skb)
680{
681	int ack;
682
683	if (!sock_flag(sk, SOCK_WIFI_STATUS))
684		return;
685	if (!skb->wifi_acked_valid)
686		return;
687
688	ack = skb->wifi_acked;
689
690	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
691}
692EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
693
694static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
695				   struct sk_buff *skb)
696{
697	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
698		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
699			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
700}
701
702void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
703	struct sk_buff *skb)
704{
705	sock_recv_timestamp(msg, sk, skb);
706	sock_recv_drops(msg, sk, skb);
707}
708EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
709
710static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
711				     size_t size, int flags)
712{
713	return sock->ops->recvmsg(sock, msg, size, flags);
714}
715
716int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
717		 int flags)
718{
719	int err = security_socket_recvmsg(sock, msg, size, flags);
720
721	return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
722}
723EXPORT_SYMBOL(sock_recvmsg);
724
725/**
726 * kernel_recvmsg - Receive a message from a socket (kernel space)
727 * @sock:       The socket to receive the message from
728 * @msg:        Received message
729 * @vec:        Input s/g array for message data
730 * @num:        Size of input s/g array
731 * @size:       Number of bytes to read
732 * @flags:      Message flags (MSG_DONTWAIT, etc...)
733 *
734 * On return the msg structure contains the scatter/gather array passed in the
735 * vec argument. The array is modified so that it consists of the unfilled
736 * portion of the original array.
737 *
738 * The returned value is the total number of bytes received, or an error.
739 */
740int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
741		   struct kvec *vec, size_t num, size_t size, int flags)
742{
743	mm_segment_t oldfs = get_fs();
744	int result;
745
746	iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
747	set_fs(KERNEL_DS);
748	result = sock_recvmsg(sock, msg, size, flags);
749	set_fs(oldfs);
750	return result;
751}
752EXPORT_SYMBOL(kernel_recvmsg);
753
754static ssize_t sock_sendpage(struct file *file, struct page *page,
755			     int offset, size_t size, loff_t *ppos, int more)
756{
757	struct socket *sock;
758	int flags;
759
760	sock = file->private_data;
761
762	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
763	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
764	flags |= more;
765
766	return kernel_sendpage(sock, page, offset, size, flags);
767}
768
769static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
770				struct pipe_inode_info *pipe, size_t len,
771				unsigned int flags)
772{
773	struct socket *sock = file->private_data;
774
775	if (unlikely(!sock->ops->splice_read))
776		return -EINVAL;
777
778	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
779}
780
781static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
782{
783	struct file *file = iocb->ki_filp;
784	struct socket *sock = file->private_data;
785	struct msghdr msg = {.msg_iter = *to,
786			     .msg_iocb = iocb};
787	ssize_t res;
788
789	if (file->f_flags & O_NONBLOCK)
790		msg.msg_flags = MSG_DONTWAIT;
791
792	if (iocb->ki_pos != 0)
793		return -ESPIPE;
794
795	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
796		return 0;
797
798	res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
799	*to = msg.msg_iter;
800	return res;
801}
802
803static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
804{
805	struct file *file = iocb->ki_filp;
806	struct socket *sock = file->private_data;
807	struct msghdr msg = {.msg_iter = *from,
808			     .msg_iocb = iocb};
809	ssize_t res;
810
811	if (iocb->ki_pos != 0)
812		return -ESPIPE;
813
814	if (file->f_flags & O_NONBLOCK)
815		msg.msg_flags = MSG_DONTWAIT;
816
817	if (sock->type == SOCK_SEQPACKET)
818		msg.msg_flags |= MSG_EOR;
819
820	res = sock_sendmsg(sock, &msg);
821	*from = msg.msg_iter;
822	return res;
823}
824
825/*
826 * Atomic setting of ioctl hooks to avoid race
827 * with module unload.
828 */
829
830static DEFINE_MUTEX(br_ioctl_mutex);
831static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
832
833void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
834{
835	mutex_lock(&br_ioctl_mutex);
836	br_ioctl_hook = hook;
837	mutex_unlock(&br_ioctl_mutex);
838}
839EXPORT_SYMBOL(brioctl_set);
840
841static DEFINE_MUTEX(vlan_ioctl_mutex);
842static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
843
844void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
845{
846	mutex_lock(&vlan_ioctl_mutex);
847	vlan_ioctl_hook = hook;
848	mutex_unlock(&vlan_ioctl_mutex);
849}
850EXPORT_SYMBOL(vlan_ioctl_set);
851
852static DEFINE_MUTEX(dlci_ioctl_mutex);
853static int (*dlci_ioctl_hook) (unsigned int, void __user *);
854
855void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
856{
857	mutex_lock(&dlci_ioctl_mutex);
858	dlci_ioctl_hook = hook;
859	mutex_unlock(&dlci_ioctl_mutex);
860}
861EXPORT_SYMBOL(dlci_ioctl_set);
862
863static long sock_do_ioctl(struct net *net, struct socket *sock,
864				 unsigned int cmd, unsigned long arg)
865{
866	int err;
867	void __user *argp = (void __user *)arg;
868
869	err = sock->ops->ioctl(sock, cmd, arg);
870
871	/*
872	 * If this ioctl is unknown try to hand it down
873	 * to the NIC driver.
874	 */
875	if (err == -ENOIOCTLCMD)
876		err = dev_ioctl(net, cmd, argp);
877
878	return err;
879}
880
881/*
882 *	With an ioctl, arg may well be a user mode pointer, but we don't know
883 *	what to do with it - that's up to the protocol still.
884 */
885
886static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
887{
888	struct socket *sock;
889	struct sock *sk;
890	void __user *argp = (void __user *)arg;
891	int pid, err;
892	struct net *net;
893
894	sock = file->private_data;
895	sk = sock->sk;
896	net = sock_net(sk);
897	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
898		err = dev_ioctl(net, cmd, argp);
899	} else
900#ifdef CONFIG_WEXT_CORE
901	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
902		err = dev_ioctl(net, cmd, argp);
903	} else
904#endif
905		switch (cmd) {
906		case FIOSETOWN:
907		case SIOCSPGRP:
908			err = -EFAULT;
909			if (get_user(pid, (int __user *)argp))
910				break;
911			f_setown(sock->file, pid, 1);
912			err = 0;
913			break;
914		case FIOGETOWN:
915		case SIOCGPGRP:
916			err = put_user(f_getown(sock->file),
917				       (int __user *)argp);
918			break;
919		case SIOCGIFBR:
920		case SIOCSIFBR:
921		case SIOCBRADDBR:
922		case SIOCBRDELBR:
923			err = -ENOPKG;
924			if (!br_ioctl_hook)
925				request_module("bridge");
926
927			mutex_lock(&br_ioctl_mutex);
928			if (br_ioctl_hook)
929				err = br_ioctl_hook(net, cmd, argp);
930			mutex_unlock(&br_ioctl_mutex);
931			break;
932		case SIOCGIFVLAN:
933		case SIOCSIFVLAN:
934			err = -ENOPKG;
935			if (!vlan_ioctl_hook)
936				request_module("8021q");
937
938			mutex_lock(&vlan_ioctl_mutex);
939			if (vlan_ioctl_hook)
940				err = vlan_ioctl_hook(net, argp);
941			mutex_unlock(&vlan_ioctl_mutex);
942			break;
943		case SIOCADDDLCI:
944		case SIOCDELDLCI:
945			err = -ENOPKG;
946			if (!dlci_ioctl_hook)
947				request_module("dlci");
948
949			mutex_lock(&dlci_ioctl_mutex);
950			if (dlci_ioctl_hook)
951				err = dlci_ioctl_hook(cmd, argp);
952			mutex_unlock(&dlci_ioctl_mutex);
953			break;
954		default:
955			err = sock_do_ioctl(net, sock, cmd, arg);
956			break;
957		}
958	return err;
959}
960
961int sock_create_lite(int family, int type, int protocol, struct socket **res)
962{
963	int err;
964	struct socket *sock = NULL;
965
966	err = security_socket_create(family, type, protocol, 1);
967	if (err)
968		goto out;
969
970	sock = sock_alloc();
971	if (!sock) {
972		err = -ENOMEM;
973		goto out;
974	}
975
976	sock->type = type;
977	err = security_socket_post_create(sock, family, type, protocol, 1);
978	if (err)
979		goto out_release;
980
981out:
982	*res = sock;
983	return err;
984out_release:
985	sock_release(sock);
986	sock = NULL;
987	goto out;
988}
989EXPORT_SYMBOL(sock_create_lite);
990
991/* No kernel lock held - perfect */
992static unsigned int sock_poll(struct file *file, poll_table *wait)
993{
994	unsigned int busy_flag = 0;
995	struct socket *sock;
996
997	/*
998	 *      We can't return errors to poll, so it's either yes or no.
999	 */
1000	sock = file->private_data;
1001
1002	if (sk_can_busy_loop(sock->sk)) {
1003		/* this socket can poll_ll so tell the system call */
1004		busy_flag = POLL_BUSY_LOOP;
1005
1006		/* once, only if requested by syscall */
1007		if (wait && (wait->_key & POLL_BUSY_LOOP))
1008			sk_busy_loop(sock->sk, 1);
1009	}
1010
1011	return busy_flag | sock->ops->poll(file, sock, wait);
1012}
1013
1014static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1015{
1016	struct socket *sock = file->private_data;
1017
1018	return sock->ops->mmap(file, sock, vma);
1019}
1020
1021static int sock_close(struct inode *inode, struct file *filp)
1022{
1023	sock_release(SOCKET_I(inode));
1024	return 0;
1025}
1026
1027/*
1028 *	Update the socket async list
1029 *
1030 *	Fasync_list locking strategy.
1031 *
1032 *	1. fasync_list is modified only under process context socket lock
1033 *	   i.e. under semaphore.
1034 *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1035 *	   or under socket lock
1036 */
1037
1038static int sock_fasync(int fd, struct file *filp, int on)
1039{
1040	struct socket *sock = filp->private_data;
1041	struct sock *sk = sock->sk;
1042	struct socket_wq *wq;
1043
1044	if (sk == NULL)
1045		return -EINVAL;
1046
1047	lock_sock(sk);
1048	wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1049	fasync_helper(fd, filp, on, &wq->fasync_list);
1050
1051	if (!wq->fasync_list)
1052		sock_reset_flag(sk, SOCK_FASYNC);
1053	else
1054		sock_set_flag(sk, SOCK_FASYNC);
1055
1056	release_sock(sk);
1057	return 0;
1058}
1059
1060/* This function may be called only under rcu_lock */
1061
1062int sock_wake_async(struct socket_wq *wq, int how, int band)
1063{
1064	if (!wq || !wq->fasync_list)
1065		return -1;
1066
1067	switch (how) {
1068	case SOCK_WAKE_WAITD:
1069		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1070			break;
1071		goto call_kill;
1072	case SOCK_WAKE_SPACE:
1073		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1074			break;
1075		/* fall through */
1076	case SOCK_WAKE_IO:
1077call_kill:
1078		kill_fasync(&wq->fasync_list, SIGIO, band);
1079		break;
1080	case SOCK_WAKE_URG:
1081		kill_fasync(&wq->fasync_list, SIGURG, band);
1082	}
1083
1084	return 0;
1085}
1086EXPORT_SYMBOL(sock_wake_async);
1087
1088int __sock_create(struct net *net, int family, int type, int protocol,
1089			 struct socket **res, int kern)
1090{
1091	int err;
1092	struct socket *sock;
1093	const struct net_proto_family *pf;
1094
1095	/*
1096	 *      Check protocol is in range
1097	 */
1098	if (family < 0 || family >= NPROTO)
1099		return -EAFNOSUPPORT;
1100	if (type < 0 || type >= SOCK_MAX)
1101		return -EINVAL;
1102
1103	/* Compatibility.
1104
1105	   This uglymoron is moved from INET layer to here to avoid
1106	   deadlock in module load.
1107	 */
1108	if (family == PF_INET && type == SOCK_PACKET) {
1109		static int warned;
1110		if (!warned) {
1111			warned = 1;
1112			pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1113				current->comm);
1114		}
1115		family = PF_PACKET;
1116	}
1117
1118	err = security_socket_create(family, type, protocol, kern);
1119	if (err)
1120		return err;
1121
1122	/*
1123	 *	Allocate the socket and allow the family to set things up. if
1124	 *	the protocol is 0, the family is instructed to select an appropriate
1125	 *	default.
1126	 */
1127	sock = sock_alloc();
1128	if (!sock) {
1129		net_warn_ratelimited("socket: no more sockets\n");
1130		return -ENFILE;	/* Not exactly a match, but its the
1131				   closest posix thing */
1132	}
1133
1134	sock->type = type;
1135
1136#ifdef CONFIG_MODULES
1137	/* Attempt to load a protocol module if the find failed.
1138	 *
1139	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1140	 * requested real, full-featured networking support upon configuration.
1141	 * Otherwise module support will break!
1142	 */
1143	if (rcu_access_pointer(net_families[family]) == NULL)
1144		request_module("net-pf-%d", family);
1145#endif
1146
1147	rcu_read_lock();
1148	pf = rcu_dereference(net_families[family]);
1149	err = -EAFNOSUPPORT;
1150	if (!pf)
1151		goto out_release;
1152
1153	/*
1154	 * We will call the ->create function, that possibly is in a loadable
1155	 * module, so we have to bump that loadable module refcnt first.
1156	 */
1157	if (!try_module_get(pf->owner))
1158		goto out_release;
1159
1160	/* Now protected by module ref count */
1161	rcu_read_unlock();
1162
1163	err = pf->create(net, sock, protocol, kern);
1164	if (err < 0)
1165		goto out_module_put;
1166
1167	/*
1168	 * Now to bump the refcnt of the [loadable] module that owns this
1169	 * socket at sock_release time we decrement its refcnt.
1170	 */
1171	if (!try_module_get(sock->ops->owner))
1172		goto out_module_busy;
1173
1174	/*
1175	 * Now that we're done with the ->create function, the [loadable]
1176	 * module can have its refcnt decremented
1177	 */
1178	module_put(pf->owner);
1179	err = security_socket_post_create(sock, family, type, protocol, kern);
1180	if (err)
1181		goto out_sock_release;
1182	*res = sock;
1183
1184	return 0;
1185
1186out_module_busy:
1187	err = -EAFNOSUPPORT;
1188out_module_put:
1189	sock->ops = NULL;
1190	module_put(pf->owner);
1191out_sock_release:
1192	sock_release(sock);
1193	return err;
1194
1195out_release:
1196	rcu_read_unlock();
1197	goto out_sock_release;
1198}
1199EXPORT_SYMBOL(__sock_create);
1200
1201int sock_create(int family, int type, int protocol, struct socket **res)
1202{
1203	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1204}
1205EXPORT_SYMBOL(sock_create);
1206
1207int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1208{
1209	return __sock_create(net, family, type, protocol, res, 1);
1210}
1211EXPORT_SYMBOL(sock_create_kern);
1212
1213SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1214{
1215	int retval;
1216	struct socket *sock;
1217	int flags;
1218
1219	/* Check the SOCK_* constants for consistency.  */
1220	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1221	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1222	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1223	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1224
1225	flags = type & ~SOCK_TYPE_MASK;
1226	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1227		return -EINVAL;
1228	type &= SOCK_TYPE_MASK;
1229
1230	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1231		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1232
1233	retval = sock_create(family, type, protocol, &sock);
1234	if (retval < 0)
1235		goto out;
1236
1237	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1238	if (retval < 0)
1239		goto out_release;
1240
1241out:
1242	/* It may be already another descriptor 8) Not kernel problem. */
1243	return retval;
1244
1245out_release:
1246	sock_release(sock);
1247	return retval;
1248}
1249
1250/*
1251 *	Create a pair of connected sockets.
1252 */
1253
1254SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1255		int __user *, usockvec)
1256{
1257	struct socket *sock1, *sock2;
1258	int fd1, fd2, err;
1259	struct file *newfile1, *newfile2;
1260	int flags;
1261
1262	flags = type & ~SOCK_TYPE_MASK;
1263	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1264		return -EINVAL;
1265	type &= SOCK_TYPE_MASK;
1266
1267	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1268		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1269
1270	/*
1271	 * Obtain the first socket and check if the underlying protocol
1272	 * supports the socketpair call.
1273	 */
1274
1275	err = sock_create(family, type, protocol, &sock1);
1276	if (err < 0)
1277		goto out;
1278
1279	err = sock_create(family, type, protocol, &sock2);
1280	if (err < 0)
1281		goto out_release_1;
1282
1283	err = sock1->ops->socketpair(sock1, sock2);
1284	if (err < 0)
1285		goto out_release_both;
1286
1287	fd1 = get_unused_fd_flags(flags);
1288	if (unlikely(fd1 < 0)) {
1289		err = fd1;
1290		goto out_release_both;
1291	}
1292
1293	fd2 = get_unused_fd_flags(flags);
1294	if (unlikely(fd2 < 0)) {
1295		err = fd2;
1296		goto out_put_unused_1;
1297	}
1298
1299	newfile1 = sock_alloc_file(sock1, flags, NULL);
1300	if (IS_ERR(newfile1)) {
1301		err = PTR_ERR(newfile1);
1302		goto out_put_unused_both;
1303	}
1304
1305	newfile2 = sock_alloc_file(sock2, flags, NULL);
1306	if (IS_ERR(newfile2)) {
1307		err = PTR_ERR(newfile2);
1308		goto out_fput_1;
1309	}
1310
1311	err = put_user(fd1, &usockvec[0]);
1312	if (err)
1313		goto out_fput_both;
1314
1315	err = put_user(fd2, &usockvec[1]);
1316	if (err)
1317		goto out_fput_both;
1318
1319	audit_fd_pair(fd1, fd2);
1320
1321	fd_install(fd1, newfile1);
1322	fd_install(fd2, newfile2);
1323	/* fd1 and fd2 may be already another descriptors.
1324	 * Not kernel problem.
1325	 */
1326
1327	return 0;
1328
1329out_fput_both:
1330	fput(newfile2);
1331	fput(newfile1);
1332	put_unused_fd(fd2);
1333	put_unused_fd(fd1);
1334	goto out;
1335
1336out_fput_1:
1337	fput(newfile1);
1338	put_unused_fd(fd2);
1339	put_unused_fd(fd1);
1340	sock_release(sock2);
1341	goto out;
1342
1343out_put_unused_both:
1344	put_unused_fd(fd2);
1345out_put_unused_1:
1346	put_unused_fd(fd1);
1347out_release_both:
1348	sock_release(sock2);
1349out_release_1:
1350	sock_release(sock1);
1351out:
1352	return err;
1353}
1354
1355/*
1356 *	Bind a name to a socket. Nothing much to do here since it's
1357 *	the protocol's responsibility to handle the local address.
1358 *
1359 *	We move the socket address to kernel space before we call
1360 *	the protocol layer (having also checked the address is ok).
1361 */
1362
1363SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1364{
1365	struct socket *sock;
1366	struct sockaddr_storage address;
1367	int err, fput_needed;
1368
1369	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1370	if (sock) {
1371		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1372		if (err >= 0) {
1373			err = security_socket_bind(sock,
1374						   (struct sockaddr *)&address,
1375						   addrlen);
1376			if (!err)
1377				err = sock->ops->bind(sock,
1378						      (struct sockaddr *)
1379						      &address, addrlen);
1380		}
1381		fput_light(sock->file, fput_needed);
1382	}
1383	return err;
1384}
1385
1386/*
1387 *	Perform a listen. Basically, we allow the protocol to do anything
1388 *	necessary for a listen, and if that works, we mark the socket as
1389 *	ready for listening.
1390 */
1391
1392SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1393{
1394	struct socket *sock;
1395	int err, fput_needed;
1396	int somaxconn;
1397
1398	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1399	if (sock) {
1400		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1401		if ((unsigned int)backlog > somaxconn)
1402			backlog = somaxconn;
1403
1404		err = security_socket_listen(sock, backlog);
1405		if (!err)
1406			err = sock->ops->listen(sock, backlog);
1407
1408		fput_light(sock->file, fput_needed);
1409	}
1410	return err;
1411}
1412
1413/*
1414 *	For accept, we attempt to create a new socket, set up the link
1415 *	with the client, wake up the client, then return the new
1416 *	connected fd. We collect the address of the connector in kernel
1417 *	space and move it to user at the very end. This is unclean because
1418 *	we open the socket then return an error.
1419 *
1420 *	1003.1g adds the ability to recvmsg() to query connection pending
1421 *	status to recvmsg. We need to add that support in a way thats
1422 *	clean when we restucture accept also.
1423 */
1424
1425SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1426		int __user *, upeer_addrlen, int, flags)
1427{
1428	struct socket *sock, *newsock;
1429	struct file *newfile;
1430	int err, len, newfd, fput_needed;
1431	struct sockaddr_storage address;
1432
1433	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1434		return -EINVAL;
1435
1436	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1437		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1438
1439	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1440	if (!sock)
1441		goto out;
1442
1443	err = -ENFILE;
1444	newsock = sock_alloc();
1445	if (!newsock)
1446		goto out_put;
1447
1448	newsock->type = sock->type;
1449	newsock->ops = sock->ops;
1450
1451	/*
1452	 * We don't need try_module_get here, as the listening socket (sock)
1453	 * has the protocol module (sock->ops->owner) held.
1454	 */
1455	__module_get(newsock->ops->owner);
1456
1457	newfd = get_unused_fd_flags(flags);
1458	if (unlikely(newfd < 0)) {
1459		err = newfd;
1460		sock_release(newsock);
1461		goto out_put;
1462	}
1463	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1464	if (IS_ERR(newfile)) {
1465		err = PTR_ERR(newfile);
1466		put_unused_fd(newfd);
1467		sock_release(newsock);
1468		goto out_put;
1469	}
1470
1471	err = security_socket_accept(sock, newsock);
1472	if (err)
1473		goto out_fd;
1474
1475	err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1476	if (err < 0)
1477		goto out_fd;
1478
1479	if (upeer_sockaddr) {
1480		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1481					  &len, 2) < 0) {
1482			err = -ECONNABORTED;
1483			goto out_fd;
1484		}
1485		err = move_addr_to_user(&address,
1486					len, upeer_sockaddr, upeer_addrlen);
1487		if (err < 0)
1488			goto out_fd;
1489	}
1490
1491	/* File flags are not inherited via accept() unlike another OSes. */
1492
1493	fd_install(newfd, newfile);
1494	err = newfd;
1495
1496out_put:
1497	fput_light(sock->file, fput_needed);
1498out:
1499	return err;
1500out_fd:
1501	fput(newfile);
1502	put_unused_fd(newfd);
1503	goto out_put;
1504}
1505
1506SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1507		int __user *, upeer_addrlen)
1508{
1509	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1510}
1511
1512/*
1513 *	Attempt to connect to a socket with the server address.  The address
1514 *	is in user space so we verify it is OK and move it to kernel space.
1515 *
1516 *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1517 *	break bindings
1518 *
1519 *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1520 *	other SEQPACKET protocols that take time to connect() as it doesn't
1521 *	include the -EINPROGRESS status for such sockets.
1522 */
1523
1524SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1525		int, addrlen)
1526{
1527	struct socket *sock;
1528	struct sockaddr_storage address;
1529	int err, fput_needed;
1530
1531	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1532	if (!sock)
1533		goto out;
1534	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1535	if (err < 0)
1536		goto out_put;
1537
1538	err =
1539	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1540	if (err)
1541		goto out_put;
1542
1543	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1544				 sock->file->f_flags);
1545out_put:
1546	fput_light(sock->file, fput_needed);
1547out:
1548	return err;
1549}
1550
1551/*
1552 *	Get the local address ('name') of a socket object. Move the obtained
1553 *	name to user space.
1554 */
1555
1556SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1557		int __user *, usockaddr_len)
1558{
1559	struct socket *sock;
1560	struct sockaddr_storage address;
1561	int len, err, fput_needed;
1562
1563	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1564	if (!sock)
1565		goto out;
1566
1567	err = security_socket_getsockname(sock);
1568	if (err)
1569		goto out_put;
1570
1571	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1572	if (err)
1573		goto out_put;
1574	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1575
1576out_put:
1577	fput_light(sock->file, fput_needed);
1578out:
1579	return err;
1580}
1581
1582/*
1583 *	Get the remote address ('name') of a socket object. Move the obtained
1584 *	name to user space.
1585 */
1586
1587SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1588		int __user *, usockaddr_len)
1589{
1590	struct socket *sock;
1591	struct sockaddr_storage address;
1592	int len, err, fput_needed;
1593
1594	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1595	if (sock != NULL) {
1596		err = security_socket_getpeername(sock);
1597		if (err) {
1598			fput_light(sock->file, fput_needed);
1599			return err;
1600		}
1601
1602		err =
1603		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1604				       1);
1605		if (!err)
1606			err = move_addr_to_user(&address, len, usockaddr,
1607						usockaddr_len);
1608		fput_light(sock->file, fput_needed);
1609	}
1610	return err;
1611}
1612
1613/*
1614 *	Send a datagram to a given address. We move the address into kernel
1615 *	space and check the user space data area is readable before invoking
1616 *	the protocol.
1617 */
1618
1619SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1620		unsigned int, flags, struct sockaddr __user *, addr,
1621		int, addr_len)
1622{
1623	struct socket *sock;
1624	struct sockaddr_storage address;
1625	int err;
1626	struct msghdr msg;
1627	struct iovec iov;
1628	int fput_needed;
1629
1630	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1631	if (unlikely(err))
1632		return err;
1633	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1634	if (!sock)
1635		goto out;
1636
1637	msg.msg_name = NULL;
1638	msg.msg_control = NULL;
1639	msg.msg_controllen = 0;
1640	msg.msg_namelen = 0;
1641	if (addr) {
1642		err = move_addr_to_kernel(addr, addr_len, &address);
1643		if (err < 0)
1644			goto out_put;
1645		msg.msg_name = (struct sockaddr *)&address;
1646		msg.msg_namelen = addr_len;
1647	}
1648	if (sock->file->f_flags & O_NONBLOCK)
1649		flags |= MSG_DONTWAIT;
1650	msg.msg_flags = flags;
1651	err = sock_sendmsg(sock, &msg);
1652
1653out_put:
1654	fput_light(sock->file, fput_needed);
1655out:
1656	return err;
1657}
1658
1659/*
1660 *	Send a datagram down a socket.
1661 */
1662
1663SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1664		unsigned int, flags)
1665{
1666	return sys_sendto(fd, buff, len, flags, NULL, 0);
1667}
1668
1669/*
1670 *	Receive a frame from the socket and optionally record the address of the
1671 *	sender. We verify the buffers are writable and if needed move the
1672 *	sender address from kernel to user space.
1673 */
1674
1675SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1676		unsigned int, flags, struct sockaddr __user *, addr,
1677		int __user *, addr_len)
1678{
1679	struct socket *sock;
1680	struct iovec iov;
1681	struct msghdr msg;
1682	struct sockaddr_storage address;
1683	int err, err2;
1684	int fput_needed;
1685
1686	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1687	if (unlikely(err))
1688		return err;
1689	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1690	if (!sock)
1691		goto out;
1692
1693	msg.msg_control = NULL;
1694	msg.msg_controllen = 0;
1695	/* Save some cycles and don't copy the address if not needed */
1696	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1697	/* We assume all kernel code knows the size of sockaddr_storage */
1698	msg.msg_namelen = 0;
1699	msg.msg_iocb = NULL;
1700	if (sock->file->f_flags & O_NONBLOCK)
1701		flags |= MSG_DONTWAIT;
1702	err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1703
1704	if (err >= 0 && addr != NULL) {
1705		err2 = move_addr_to_user(&address,
1706					 msg.msg_namelen, addr, addr_len);
1707		if (err2 < 0)
1708			err = err2;
1709	}
1710
1711	fput_light(sock->file, fput_needed);
1712out:
1713	return err;
1714}
1715
1716/*
1717 *	Receive a datagram from a socket.
1718 */
1719
1720SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1721		unsigned int, flags)
1722{
1723	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1724}
1725
1726/*
1727 *	Set a socket option. Because we don't know the option lengths we have
1728 *	to pass the user mode parameter for the protocols to sort out.
1729 */
1730
1731SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1732		char __user *, optval, int, optlen)
1733{
1734	int err, fput_needed;
1735	struct socket *sock;
1736
1737	if (optlen < 0)
1738		return -EINVAL;
1739
1740	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1741	if (sock != NULL) {
1742		err = security_socket_setsockopt(sock, level, optname);
1743		if (err)
1744			goto out_put;
1745
1746		if (level == SOL_SOCKET)
1747			err =
1748			    sock_setsockopt(sock, level, optname, optval,
1749					    optlen);
1750		else
1751			err =
1752			    sock->ops->setsockopt(sock, level, optname, optval,
1753						  optlen);
1754out_put:
1755		fput_light(sock->file, fput_needed);
1756	}
1757	return err;
1758}
1759
1760/*
1761 *	Get a socket option. Because we don't know the option lengths we have
1762 *	to pass a user mode parameter for the protocols to sort out.
1763 */
1764
1765SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1766		char __user *, optval, int __user *, optlen)
1767{
1768	int err, fput_needed;
1769	struct socket *sock;
1770
1771	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1772	if (sock != NULL) {
1773		err = security_socket_getsockopt(sock, level, optname);
1774		if (err)
1775			goto out_put;
1776
1777		if (level == SOL_SOCKET)
1778			err =
1779			    sock_getsockopt(sock, level, optname, optval,
1780					    optlen);
1781		else
1782			err =
1783			    sock->ops->getsockopt(sock, level, optname, optval,
1784						  optlen);
1785out_put:
1786		fput_light(sock->file, fput_needed);
1787	}
1788	return err;
1789}
1790
1791/*
1792 *	Shutdown a socket.
1793 */
1794
1795SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1796{
1797	int err, fput_needed;
1798	struct socket *sock;
1799
1800	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1801	if (sock != NULL) {
1802		err = security_socket_shutdown(sock, how);
1803		if (!err)
1804			err = sock->ops->shutdown(sock, how);
1805		fput_light(sock->file, fput_needed);
1806	}
1807	return err;
1808}
1809
1810/* A couple of helpful macros for getting the address of the 32/64 bit
1811 * fields which are the same type (int / unsigned) on our platforms.
1812 */
1813#define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1814#define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1815#define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1816
1817struct used_address {
1818	struct sockaddr_storage name;
1819	unsigned int name_len;
1820};
1821
1822static int copy_msghdr_from_user(struct msghdr *kmsg,
1823				 struct user_msghdr __user *umsg,
1824				 struct sockaddr __user **save_addr,
1825				 struct iovec **iov)
1826{
1827	struct sockaddr __user *uaddr;
1828	struct iovec __user *uiov;
1829	size_t nr_segs;
1830	ssize_t err;
1831
1832	if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1833	    __get_user(uaddr, &umsg->msg_name) ||
1834	    __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1835	    __get_user(uiov, &umsg->msg_iov) ||
1836	    __get_user(nr_segs, &umsg->msg_iovlen) ||
1837	    __get_user(kmsg->msg_control, &umsg->msg_control) ||
1838	    __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1839	    __get_user(kmsg->msg_flags, &umsg->msg_flags))
1840		return -EFAULT;
1841
1842	if (!uaddr)
1843		kmsg->msg_namelen = 0;
1844
1845	if (kmsg->msg_namelen < 0)
1846		return -EINVAL;
1847
1848	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1849		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1850
1851	if (save_addr)
1852		*save_addr = uaddr;
1853
1854	if (uaddr && kmsg->msg_namelen) {
1855		if (!save_addr) {
1856			err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1857						  kmsg->msg_name);
1858			if (err < 0)
1859				return err;
1860		}
1861	} else {
1862		kmsg->msg_name = NULL;
1863		kmsg->msg_namelen = 0;
1864	}
1865
1866	if (nr_segs > UIO_MAXIOV)
1867		return -EMSGSIZE;
1868
1869	kmsg->msg_iocb = NULL;
1870
1871	return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1872			    UIO_FASTIOV, iov, &kmsg->msg_iter);
1873}
1874
1875static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1876			 struct msghdr *msg_sys, unsigned int flags,
1877			 struct used_address *used_address)
1878{
1879	struct compat_msghdr __user *msg_compat =
1880	    (struct compat_msghdr __user *)msg;
1881	struct sockaddr_storage address;
1882	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1883	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1884	    __attribute__ ((aligned(sizeof(__kernel_size_t))));
1885	/* 20 is size of ipv6_pktinfo */
1886	unsigned char *ctl_buf = ctl;
1887	int ctl_len;
1888	ssize_t err;
1889
1890	msg_sys->msg_name = &address;
1891
1892	if (MSG_CMSG_COMPAT & flags)
1893		err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1894	else
1895		err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1896	if (err < 0)
1897		return err;
1898
1899	err = -ENOBUFS;
1900
1901	if (msg_sys->msg_controllen > INT_MAX)
1902		goto out_freeiov;
1903	ctl_len = msg_sys->msg_controllen;
1904	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1905		err =
1906		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1907						     sizeof(ctl));
1908		if (err)
1909			goto out_freeiov;
1910		ctl_buf = msg_sys->msg_control;
1911		ctl_len = msg_sys->msg_controllen;
1912	} else if (ctl_len) {
1913		if (ctl_len > sizeof(ctl)) {
1914			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1915			if (ctl_buf == NULL)
1916				goto out_freeiov;
1917		}
1918		err = -EFAULT;
1919		/*
1920		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1921		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1922		 * checking falls down on this.
1923		 */
1924		if (copy_from_user(ctl_buf,
1925				   (void __user __force *)msg_sys->msg_control,
1926				   ctl_len))
1927			goto out_freectl;
1928		msg_sys->msg_control = ctl_buf;
1929	}
1930	msg_sys->msg_flags = flags;
1931
1932	if (sock->file->f_flags & O_NONBLOCK)
1933		msg_sys->msg_flags |= MSG_DONTWAIT;
1934	/*
1935	 * If this is sendmmsg() and current destination address is same as
1936	 * previously succeeded address, omit asking LSM's decision.
1937	 * used_address->name_len is initialized to UINT_MAX so that the first
1938	 * destination address never matches.
1939	 */
1940	if (used_address && msg_sys->msg_name &&
1941	    used_address->name_len == msg_sys->msg_namelen &&
1942	    !memcmp(&used_address->name, msg_sys->msg_name,
1943		    used_address->name_len)) {
1944		err = sock_sendmsg_nosec(sock, msg_sys);
1945		goto out_freectl;
1946	}
1947	err = sock_sendmsg(sock, msg_sys);
1948	/*
1949	 * If this is sendmmsg() and sending to current destination address was
1950	 * successful, remember it.
1951	 */
1952	if (used_address && err >= 0) {
1953		used_address->name_len = msg_sys->msg_namelen;
1954		if (msg_sys->msg_name)
1955			memcpy(&used_address->name, msg_sys->msg_name,
1956			       used_address->name_len);
1957	}
1958
1959out_freectl:
1960	if (ctl_buf != ctl)
1961		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1962out_freeiov:
1963	kfree(iov);
1964	return err;
1965}
1966
1967/*
1968 *	BSD sendmsg interface
1969 */
1970
1971long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1972{
1973	int fput_needed, err;
1974	struct msghdr msg_sys;
1975	struct socket *sock;
1976
1977	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1978	if (!sock)
1979		goto out;
1980
1981	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
1982
1983	fput_light(sock->file, fput_needed);
1984out:
1985	return err;
1986}
1987
1988SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1989{
1990	if (flags & MSG_CMSG_COMPAT)
1991		return -EINVAL;
1992	return __sys_sendmsg(fd, msg, flags);
1993}
1994
1995/*
1996 *	Linux sendmmsg interface
1997 */
1998
1999int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2000		   unsigned int flags)
2001{
2002	int fput_needed, err, datagrams;
2003	struct socket *sock;
2004	struct mmsghdr __user *entry;
2005	struct compat_mmsghdr __user *compat_entry;
2006	struct msghdr msg_sys;
2007	struct used_address used_address;
2008
2009	if (vlen > UIO_MAXIOV)
2010		vlen = UIO_MAXIOV;
2011
2012	datagrams = 0;
2013
2014	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2015	if (!sock)
2016		return err;
2017
2018	used_address.name_len = UINT_MAX;
2019	entry = mmsg;
2020	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2021	err = 0;
2022
2023	while (datagrams < vlen) {
2024		if (MSG_CMSG_COMPAT & flags) {
2025			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2026					     &msg_sys, flags, &used_address);
2027			if (err < 0)
2028				break;
2029			err = __put_user(err, &compat_entry->msg_len);
2030			++compat_entry;
2031		} else {
2032			err = ___sys_sendmsg(sock,
2033					     (struct user_msghdr __user *)entry,
2034					     &msg_sys, flags, &used_address);
2035			if (err < 0)
2036				break;
2037			err = put_user(err, &entry->msg_len);
2038			++entry;
2039		}
2040
2041		if (err)
2042			break;
2043		++datagrams;
2044	}
2045
2046	fput_light(sock->file, fput_needed);
2047
2048	/* We only return an error if no datagrams were able to be sent */
2049	if (datagrams != 0)
2050		return datagrams;
2051
2052	return err;
2053}
2054
2055SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2056		unsigned int, vlen, unsigned int, flags)
2057{
2058	if (flags & MSG_CMSG_COMPAT)
2059		return -EINVAL;
2060	return __sys_sendmmsg(fd, mmsg, vlen, flags);
2061}
2062
2063static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2064			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2065{
2066	struct compat_msghdr __user *msg_compat =
2067	    (struct compat_msghdr __user *)msg;
2068	struct iovec iovstack[UIO_FASTIOV];
2069	struct iovec *iov = iovstack;
2070	unsigned long cmsg_ptr;
2071	int total_len, len;
2072	ssize_t err;
2073
2074	/* kernel mode address */
2075	struct sockaddr_storage addr;
2076
2077	/* user mode address pointers */
2078	struct sockaddr __user *uaddr;
2079	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2080
2081	msg_sys->msg_name = &addr;
2082
2083	if (MSG_CMSG_COMPAT & flags)
2084		err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2085	else
2086		err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2087	if (err < 0)
2088		return err;
2089	total_len = iov_iter_count(&msg_sys->msg_iter);
2090
2091	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2092	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2093
2094	/* We assume all kernel code knows the size of sockaddr_storage */
2095	msg_sys->msg_namelen = 0;
2096
2097	if (sock->file->f_flags & O_NONBLOCK)
2098		flags |= MSG_DONTWAIT;
2099	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2100							  total_len, flags);
2101	if (err < 0)
2102		goto out_freeiov;
2103	len = err;
2104
2105	if (uaddr != NULL) {
2106		err = move_addr_to_user(&addr,
2107					msg_sys->msg_namelen, uaddr,
2108					uaddr_len);
2109		if (err < 0)
2110			goto out_freeiov;
2111	}
2112	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2113			 COMPAT_FLAGS(msg));
2114	if (err)
2115		goto out_freeiov;
2116	if (MSG_CMSG_COMPAT & flags)
2117		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2118				 &msg_compat->msg_controllen);
2119	else
2120		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2121				 &msg->msg_controllen);
2122	if (err)
2123		goto out_freeiov;
2124	err = len;
2125
2126out_freeiov:
2127	kfree(iov);
2128	return err;
2129}
2130
2131/*
2132 *	BSD recvmsg interface
2133 */
2134
2135long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2136{
2137	int fput_needed, err;
2138	struct msghdr msg_sys;
2139	struct socket *sock;
2140
2141	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2142	if (!sock)
2143		goto out;
2144
2145	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2146
2147	fput_light(sock->file, fput_needed);
2148out:
2149	return err;
2150}
2151
2152SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2153		unsigned int, flags)
2154{
2155	if (flags & MSG_CMSG_COMPAT)
2156		return -EINVAL;
2157	return __sys_recvmsg(fd, msg, flags);
2158}
2159
2160/*
2161 *     Linux recvmmsg interface
2162 */
2163
2164int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2165		   unsigned int flags, struct timespec *timeout)
2166{
2167	int fput_needed, err, datagrams;
2168	struct socket *sock;
2169	struct mmsghdr __user *entry;
2170	struct compat_mmsghdr __user *compat_entry;
2171	struct msghdr msg_sys;
2172	struct timespec end_time;
2173
2174	if (timeout &&
2175	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2176				    timeout->tv_nsec))
2177		return -EINVAL;
2178
2179	datagrams = 0;
2180
2181	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2182	if (!sock)
2183		return err;
2184
2185	err = sock_error(sock->sk);
2186	if (err)
2187		goto out_put;
2188
2189	entry = mmsg;
2190	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2191
2192	while (datagrams < vlen) {
2193		/*
2194		 * No need to ask LSM for more than the first datagram.
2195		 */
2196		if (MSG_CMSG_COMPAT & flags) {
2197			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2198					     &msg_sys, flags & ~MSG_WAITFORONE,
2199					     datagrams);
2200			if (err < 0)
2201				break;
2202			err = __put_user(err, &compat_entry->msg_len);
2203			++compat_entry;
2204		} else {
2205			err = ___sys_recvmsg(sock,
2206					     (struct user_msghdr __user *)entry,
2207					     &msg_sys, flags & ~MSG_WAITFORONE,
2208					     datagrams);
2209			if (err < 0)
2210				break;
2211			err = put_user(err, &entry->msg_len);
2212			++entry;
2213		}
2214
2215		if (err)
2216			break;
2217		++datagrams;
2218
2219		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2220		if (flags & MSG_WAITFORONE)
2221			flags |= MSG_DONTWAIT;
2222
2223		if (timeout) {
2224			ktime_get_ts(timeout);
2225			*timeout = timespec_sub(end_time, *timeout);
2226			if (timeout->tv_sec < 0) {
2227				timeout->tv_sec = timeout->tv_nsec = 0;
2228				break;
2229			}
2230
2231			/* Timeout, return less than vlen datagrams */
2232			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2233				break;
2234		}
2235
2236		/* Out of band data, return right away */
2237		if (msg_sys.msg_flags & MSG_OOB)
2238			break;
2239	}
2240
2241	if (err == 0)
2242		goto out_put;
2243
2244	if (datagrams == 0) {
2245		datagrams = err;
2246		goto out_put;
2247	}
2248
2249	/*
2250	 * We may return less entries than requested (vlen) if the
2251	 * sock is non block and there aren't enough datagrams...
2252	 */
2253	if (err != -EAGAIN) {
2254		/*
2255		 * ... or  if recvmsg returns an error after we
2256		 * received some datagrams, where we record the
2257		 * error to return on the next call or if the
2258		 * app asks about it using getsockopt(SO_ERROR).
2259		 */
2260		sock->sk->sk_err = -err;
2261	}
2262out_put:
2263	fput_light(sock->file, fput_needed);
2264
2265	return datagrams;
2266}
2267
2268SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2269		unsigned int, vlen, unsigned int, flags,
2270		struct timespec __user *, timeout)
2271{
2272	int datagrams;
2273	struct timespec timeout_sys;
2274
2275	if (flags & MSG_CMSG_COMPAT)
2276		return -EINVAL;
2277
2278	if (!timeout)
2279		return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2280
2281	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2282		return -EFAULT;
2283
2284	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2285
2286	if (datagrams > 0 &&
2287	    copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2288		datagrams = -EFAULT;
2289
2290	return datagrams;
2291}
2292
2293#ifdef __ARCH_WANT_SYS_SOCKETCALL
2294/* Argument list sizes for sys_socketcall */
2295#define AL(x) ((x) * sizeof(unsigned long))
2296static const unsigned char nargs[21] = {
2297	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2298	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2299	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2300	AL(4), AL(5), AL(4)
2301};
2302
2303#undef AL
2304
2305/*
2306 *	System call vectors.
2307 *
2308 *	Argument checking cleaned up. Saved 20% in size.
2309 *  This function doesn't need to set the kernel lock because
2310 *  it is set by the callees.
2311 */
2312
2313SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2314{
2315	unsigned long a[AUDITSC_ARGS];
2316	unsigned long a0, a1;
2317	int err;
2318	unsigned int len;
2319
2320	if (call < 1 || call > SYS_SENDMMSG)
2321		return -EINVAL;
2322
2323	len = nargs[call];
2324	if (len > sizeof(a))
2325		return -EINVAL;
2326
2327	/* copy_from_user should be SMP safe. */
2328	if (copy_from_user(a, args, len))
2329		return -EFAULT;
2330
2331	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2332	if (err)
2333		return err;
2334
2335	a0 = a[0];
2336	a1 = a[1];
2337
2338	switch (call) {
2339	case SYS_SOCKET:
2340		err = sys_socket(a0, a1, a[2]);
2341		break;
2342	case SYS_BIND:
2343		err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2344		break;
2345	case SYS_CONNECT:
2346		err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2347		break;
2348	case SYS_LISTEN:
2349		err = sys_listen(a0, a1);
2350		break;
2351	case SYS_ACCEPT:
2352		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2353				  (int __user *)a[2], 0);
2354		break;
2355	case SYS_GETSOCKNAME:
2356		err =
2357		    sys_getsockname(a0, (struct sockaddr __user *)a1,
2358				    (int __user *)a[2]);
2359		break;
2360	case SYS_GETPEERNAME:
2361		err =
2362		    sys_getpeername(a0, (struct sockaddr __user *)a1,
2363				    (int __user *)a[2]);
2364		break;
2365	case SYS_SOCKETPAIR:
2366		err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2367		break;
2368	case SYS_SEND:
2369		err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2370		break;
2371	case SYS_SENDTO:
2372		err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2373				 (struct sockaddr __user *)a[4], a[5]);
2374		break;
2375	case SYS_RECV:
2376		err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2377		break;
2378	case SYS_RECVFROM:
2379		err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2380				   (struct sockaddr __user *)a[4],
2381				   (int __user *)a[5]);
2382		break;
2383	case SYS_SHUTDOWN:
2384		err = sys_shutdown(a0, a1);
2385		break;
2386	case SYS_SETSOCKOPT:
2387		err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2388		break;
2389	case SYS_GETSOCKOPT:
2390		err =
2391		    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2392				   (int __user *)a[4]);
2393		break;
2394	case SYS_SENDMSG:
2395		err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2396		break;
2397	case SYS_SENDMMSG:
2398		err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2399		break;
2400	case SYS_RECVMSG:
2401		err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2402		break;
2403	case SYS_RECVMMSG:
2404		err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2405				   (struct timespec __user *)a[4]);
2406		break;
2407	case SYS_ACCEPT4:
2408		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2409				  (int __user *)a[2], a[3]);
2410		break;
2411	default:
2412		err = -EINVAL;
2413		break;
2414	}
2415	return err;
2416}
2417
2418#endif				/* __ARCH_WANT_SYS_SOCKETCALL */
2419
2420/**
2421 *	sock_register - add a socket protocol handler
2422 *	@ops: description of protocol
2423 *
2424 *	This function is called by a protocol handler that wants to
2425 *	advertise its address family, and have it linked into the
2426 *	socket interface. The value ops->family corresponds to the
2427 *	socket system call protocol family.
2428 */
2429int sock_register(const struct net_proto_family *ops)
2430{
2431	int err;
2432
2433	if (ops->family >= NPROTO) {
2434		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2435		return -ENOBUFS;
2436	}
2437
2438	spin_lock(&net_family_lock);
2439	if (rcu_dereference_protected(net_families[ops->family],
2440				      lockdep_is_held(&net_family_lock)))
2441		err = -EEXIST;
2442	else {
2443		rcu_assign_pointer(net_families[ops->family], ops);
2444		err = 0;
2445	}
2446	spin_unlock(&net_family_lock);
2447
2448	pr_info("NET: Registered protocol family %d\n", ops->family);
2449	return err;
2450}
2451EXPORT_SYMBOL(sock_register);
2452
2453/**
2454 *	sock_unregister - remove a protocol handler
2455 *	@family: protocol family to remove
2456 *
2457 *	This function is called by a protocol handler that wants to
2458 *	remove its address family, and have it unlinked from the
2459 *	new socket creation.
2460 *
2461 *	If protocol handler is a module, then it can use module reference
2462 *	counts to protect against new references. If protocol handler is not
2463 *	a module then it needs to provide its own protection in
2464 *	the ops->create routine.
2465 */
2466void sock_unregister(int family)
2467{
2468	BUG_ON(family < 0 || family >= NPROTO);
2469
2470	spin_lock(&net_family_lock);
2471	RCU_INIT_POINTER(net_families[family], NULL);
2472	spin_unlock(&net_family_lock);
2473
2474	synchronize_rcu();
2475
2476	pr_info("NET: Unregistered protocol family %d\n", family);
2477}
2478EXPORT_SYMBOL(sock_unregister);
2479
2480static int __init sock_init(void)
2481{
2482	int err;
2483	/*
2484	 *      Initialize the network sysctl infrastructure.
2485	 */
2486	err = net_sysctl_init();
2487	if (err)
2488		goto out;
2489
2490	/*
2491	 *      Initialize skbuff SLAB cache
2492	 */
2493	skb_init();
2494
2495	/*
2496	 *      Initialize the protocols module.
2497	 */
2498
2499	init_inodecache();
2500
2501	err = register_filesystem(&sock_fs_type);
2502	if (err)
2503		goto out_fs;
2504	sock_mnt = kern_mount(&sock_fs_type);
2505	if (IS_ERR(sock_mnt)) {
2506		err = PTR_ERR(sock_mnt);
2507		goto out_mount;
2508	}
2509
2510	/* The real protocol initialization is performed in later initcalls.
2511	 */
2512
2513#ifdef CONFIG_NETFILTER
2514	err = netfilter_init();
2515	if (err)
2516		goto out;
2517#endif
2518
2519	ptp_classifier_init();
2520
2521out:
2522	return err;
2523
2524out_mount:
2525	unregister_filesystem(&sock_fs_type);
2526out_fs:
2527	goto out;
2528}
2529
2530core_initcall(sock_init);	/* early initcall */
2531
2532#ifdef CONFIG_PROC_FS
2533void socket_seq_show(struct seq_file *seq)
2534{
2535	int cpu;
2536	int counter = 0;
2537
2538	for_each_possible_cpu(cpu)
2539	    counter += per_cpu(sockets_in_use, cpu);
2540
2541	/* It can be negative, by the way. 8) */
2542	if (counter < 0)
2543		counter = 0;
2544
2545	seq_printf(seq, "sockets: used %d\n", counter);
2546}
2547#endif				/* CONFIG_PROC_FS */
2548
2549#ifdef CONFIG_COMPAT
2550static int do_siocgstamp(struct net *net, struct socket *sock,
2551			 unsigned int cmd, void __user *up)
2552{
2553	mm_segment_t old_fs = get_fs();
2554	struct timeval ktv;
2555	int err;
2556
2557	set_fs(KERNEL_DS);
2558	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2559	set_fs(old_fs);
2560	if (!err)
2561		err = compat_put_timeval(&ktv, up);
2562
2563	return err;
2564}
2565
2566static int do_siocgstampns(struct net *net, struct socket *sock,
2567			   unsigned int cmd, void __user *up)
2568{
2569	mm_segment_t old_fs = get_fs();
2570	struct timespec kts;
2571	int err;
2572
2573	set_fs(KERNEL_DS);
2574	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2575	set_fs(old_fs);
2576	if (!err)
2577		err = compat_put_timespec(&kts, up);
2578
2579	return err;
2580}
2581
2582static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2583{
2584	struct ifreq __user *uifr;
2585	int err;
2586
2587	uifr = compat_alloc_user_space(sizeof(struct ifreq));
2588	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2589		return -EFAULT;
2590
2591	err = dev_ioctl(net, SIOCGIFNAME, uifr);
2592	if (err)
2593		return err;
2594
2595	if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2596		return -EFAULT;
2597
2598	return 0;
2599}
2600
2601static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2602{
2603	struct compat_ifconf ifc32;
2604	struct ifconf ifc;
2605	struct ifconf __user *uifc;
2606	struct compat_ifreq __user *ifr32;
2607	struct ifreq __user *ifr;
2608	unsigned int i, j;
2609	int err;
2610
2611	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2612		return -EFAULT;
2613
2614	memset(&ifc, 0, sizeof(ifc));
2615	if (ifc32.ifcbuf == 0) {
2616		ifc32.ifc_len = 0;
2617		ifc.ifc_len = 0;
2618		ifc.ifc_req = NULL;
2619		uifc = compat_alloc_user_space(sizeof(struct ifconf));
2620	} else {
2621		size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2622			sizeof(struct ifreq);
2623		uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2624		ifc.ifc_len = len;
2625		ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2626		ifr32 = compat_ptr(ifc32.ifcbuf);
2627		for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2628			if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2629				return -EFAULT;
2630			ifr++;
2631			ifr32++;
2632		}
2633	}
2634	if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2635		return -EFAULT;
2636
2637	err = dev_ioctl(net, SIOCGIFCONF, uifc);
2638	if (err)
2639		return err;
2640
2641	if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2642		return -EFAULT;
2643
2644	ifr = ifc.ifc_req;
2645	ifr32 = compat_ptr(ifc32.ifcbuf);
2646	for (i = 0, j = 0;
2647	     i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2648	     i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2649		if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2650			return -EFAULT;
2651		ifr32++;
2652		ifr++;
2653	}
2654
2655	if (ifc32.ifcbuf == 0) {
2656		/* Translate from 64-bit structure multiple to
2657		 * a 32-bit one.
2658		 */
2659		i = ifc.ifc_len;
2660		i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2661		ifc32.ifc_len = i;
2662	} else {
2663		ifc32.ifc_len = i;
2664	}
2665	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2666		return -EFAULT;
2667
2668	return 0;
2669}
2670
2671static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2672{
2673	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2674	bool convert_in = false, convert_out = false;
2675	size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2676	struct ethtool_rxnfc __user *rxnfc;
2677	struct ifreq __user *ifr;
2678	u32 rule_cnt = 0, actual_rule_cnt;
2679	u32 ethcmd;
2680	u32 data;
2681	int ret;
2682
2683	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2684		return -EFAULT;
2685
2686	compat_rxnfc = compat_ptr(data);
2687
2688	if (get_user(ethcmd, &compat_rxnfc->cmd))
2689		return -EFAULT;
2690
2691	/* Most ethtool structures are defined without padding.
2692	 * Unfortunately struct ethtool_rxnfc is an exception.
2693	 */
2694	switch (ethcmd) {
2695	default:
2696		break;
2697	case ETHTOOL_GRXCLSRLALL:
2698		/* Buffer size is variable */
2699		if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2700			return -EFAULT;
2701		if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2702			return -ENOMEM;
2703		buf_size += rule_cnt * sizeof(u32);
2704		/* fall through */
2705	case ETHTOOL_GRXRINGS:
2706	case ETHTOOL_GRXCLSRLCNT:
2707	case ETHTOOL_GRXCLSRULE:
2708	case ETHTOOL_SRXCLSRLINS:
2709		convert_out = true;
2710		/* fall through */
2711	case ETHTOOL_SRXCLSRLDEL:
2712		buf_size += sizeof(struct ethtool_rxnfc);
2713		convert_in = true;
2714		break;
2715	}
2716
2717	ifr = compat_alloc_user_space(buf_size);
2718	rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2719
2720	if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2721		return -EFAULT;
2722
2723	if (put_user(convert_in ? rxnfc : compat_ptr(data),
2724		     &ifr->ifr_ifru.ifru_data))
2725		return -EFAULT;
2726
2727	if (convert_in) {
2728		/* We expect there to be holes between fs.m_ext and
2729		 * fs.ring_cookie and at the end of fs, but nowhere else.
2730		 */
2731		BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2732			     sizeof(compat_rxnfc->fs.m_ext) !=
2733			     offsetof(struct ethtool_rxnfc, fs.m_ext) +
2734			     sizeof(rxnfc->fs.m_ext));
2735		BUILD_BUG_ON(
2736			offsetof(struct compat_ethtool_rxnfc, fs.location) -
2737			offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2738			offsetof(struct ethtool_rxnfc, fs.location) -
2739			offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2740
2741		if (copy_in_user(rxnfc, compat_rxnfc,
2742				 (void __user *)(&rxnfc->fs.m_ext + 1) -
2743				 (void __user *)rxnfc) ||
2744		    copy_in_user(&rxnfc->fs.ring_cookie,
2745				 &compat_rxnfc->fs.ring_cookie,
2746				 (void __user *)(&rxnfc->fs.location + 1) -
2747				 (void __user *)&rxnfc->fs.ring_cookie) ||
2748		    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2749				 sizeof(rxnfc->rule_cnt)))
2750			return -EFAULT;
2751	}
2752
2753	ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2754	if (ret)
2755		return ret;
2756
2757	if (convert_out) {
2758		if (copy_in_user(compat_rxnfc, rxnfc,
2759				 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2760				 (const void __user *)rxnfc) ||
2761		    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2762				 &rxnfc->fs.ring_cookie,
2763				 (const void __user *)(&rxnfc->fs.location + 1) -
2764				 (const void __user *)&rxnfc->fs.ring_cookie) ||
2765		    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2766				 sizeof(rxnfc->rule_cnt)))
2767			return -EFAULT;
2768
2769		if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2770			/* As an optimisation, we only copy the actual
2771			 * number of rules that the underlying
2772			 * function returned.  Since Mallory might
2773			 * change the rule count in user memory, we
2774			 * check that it is less than the rule count
2775			 * originally given (as the user buffer size),
2776			 * which has been range-checked.
2777			 */
2778			if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2779				return -EFAULT;
2780			if (actual_rule_cnt < rule_cnt)
2781				rule_cnt = actual_rule_cnt;
2782			if (copy_in_user(&compat_rxnfc->rule_locs[0],
2783					 &rxnfc->rule_locs[0],
2784					 rule_cnt * sizeof(u32)))
2785				return -EFAULT;
2786		}
2787	}
2788
2789	return 0;
2790}
2791
2792static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2793{
2794	void __user *uptr;
2795	compat_uptr_t uptr32;
2796	struct ifreq __user *uifr;
2797
2798	uifr = compat_alloc_user_space(sizeof(*uifr));
2799	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2800		return -EFAULT;
2801
2802	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2803		return -EFAULT;
2804
2805	uptr = compat_ptr(uptr32);
2806
2807	if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2808		return -EFAULT;
2809
2810	return dev_ioctl(net, SIOCWANDEV, uifr);
2811}
2812
2813static int bond_ioctl(struct net *net, unsigned int cmd,
2814			 struct compat_ifreq __user *ifr32)
2815{
2816	struct ifreq kifr;
2817	mm_segment_t old_fs;
2818	int err;
2819
2820	switch (cmd) {
2821	case SIOCBONDENSLAVE:
2822	case SIOCBONDRELEASE:
2823	case SIOCBONDSETHWADDR:
2824	case SIOCBONDCHANGEACTIVE:
2825		if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2826			return -EFAULT;
2827
2828		old_fs = get_fs();
2829		set_fs(KERNEL_DS);
2830		err = dev_ioctl(net, cmd,
2831				(struct ifreq __user __force *) &kifr);
2832		set_fs(old_fs);
2833
2834		return err;
2835	default:
2836		return -ENOIOCTLCMD;
2837	}
2838}
2839
2840/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2841static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2842				 struct compat_ifreq __user *u_ifreq32)
2843{
2844	struct ifreq __user *u_ifreq64;
2845	char tmp_buf[IFNAMSIZ];
2846	void __user *data64;
2847	u32 data32;
2848
2849	if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2850			   IFNAMSIZ))
2851		return -EFAULT;
2852	if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2853		return -EFAULT;
2854	data64 = compat_ptr(data32);
2855
2856	u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2857
2858	if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2859			 IFNAMSIZ))
2860		return -EFAULT;
2861	if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2862		return -EFAULT;
2863
2864	return dev_ioctl(net, cmd, u_ifreq64);
2865}
2866
2867static int dev_ifsioc(struct net *net, struct socket *sock,
2868			 unsigned int cmd, struct compat_ifreq __user *uifr32)
2869{
2870	struct ifreq __user *uifr;
2871	int err;
2872
2873	uifr = compat_alloc_user_space(sizeof(*uifr));
2874	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2875		return -EFAULT;
2876
2877	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2878
2879	if (!err) {
2880		switch (cmd) {
2881		case SIOCGIFFLAGS:
2882		case SIOCGIFMETRIC:
2883		case SIOCGIFMTU:
2884		case SIOCGIFMEM:
2885		case SIOCGIFHWADDR:
2886		case SIOCGIFINDEX:
2887		case SIOCGIFADDR:
2888		case SIOCGIFBRDADDR:
2889		case SIOCGIFDSTADDR:
2890		case SIOCGIFNETMASK:
2891		case SIOCGIFPFLAGS:
2892		case SIOCGIFTXQLEN:
2893		case SIOCGMIIPHY:
2894		case SIOCGMIIREG:
2895			if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2896				err = -EFAULT;
2897			break;
2898		}
2899	}
2900	return err;
2901}
2902
2903static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2904			struct compat_ifreq __user *uifr32)
2905{
2906	struct ifreq ifr;
2907	struct compat_ifmap __user *uifmap32;
2908	mm_segment_t old_fs;
2909	int err;
2910
2911	uifmap32 = &uifr32->ifr_ifru.ifru_map;
2912	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2913	err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2914	err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2915	err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2916	err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2917	err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2918	err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2919	if (err)
2920		return -EFAULT;
2921
2922	old_fs = get_fs();
2923	set_fs(KERNEL_DS);
2924	err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
2925	set_fs(old_fs);
2926
2927	if (cmd == SIOCGIFMAP && !err) {
2928		err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2929		err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2930		err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2931		err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2932		err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2933		err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2934		err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2935		if (err)
2936			err = -EFAULT;
2937	}
2938	return err;
2939}
2940
2941struct rtentry32 {
2942	u32		rt_pad1;
2943	struct sockaddr rt_dst;         /* target address               */
2944	struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
2945	struct sockaddr rt_genmask;     /* target network mask (IP)     */
2946	unsigned short	rt_flags;
2947	short		rt_pad2;
2948	u32		rt_pad3;
2949	unsigned char	rt_tos;
2950	unsigned char	rt_class;
2951	short		rt_pad4;
2952	short		rt_metric;      /* +1 for binary compatibility! */
2953	/* char * */ u32 rt_dev;        /* forcing the device at add    */
2954	u32		rt_mtu;         /* per route MTU/Window         */
2955	u32		rt_window;      /* Window clamping              */
2956	unsigned short  rt_irtt;        /* Initial RTT                  */
2957};
2958
2959struct in6_rtmsg32 {
2960	struct in6_addr		rtmsg_dst;
2961	struct in6_addr		rtmsg_src;
2962	struct in6_addr		rtmsg_gateway;
2963	u32			rtmsg_type;
2964	u16			rtmsg_dst_len;
2965	u16			rtmsg_src_len;
2966	u32			rtmsg_metric;
2967	u32			rtmsg_info;
2968	u32			rtmsg_flags;
2969	s32			rtmsg_ifindex;
2970};
2971
2972static int routing_ioctl(struct net *net, struct socket *sock,
2973			 unsigned int cmd, void __user *argp)
2974{
2975	int ret;
2976	void *r = NULL;
2977	struct in6_rtmsg r6;
2978	struct rtentry r4;
2979	char devname[16];
2980	u32 rtdev;
2981	mm_segment_t old_fs = get_fs();
2982
2983	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2984		struct in6_rtmsg32 __user *ur6 = argp;
2985		ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2986			3 * sizeof(struct in6_addr));
2987		ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2988		ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2989		ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2990		ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2991		ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2992		ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2993		ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2994
2995		r = (void *) &r6;
2996	} else { /* ipv4 */
2997		struct rtentry32 __user *ur4 = argp;
2998		ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
2999					3 * sizeof(struct sockaddr));
3000		ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3001		ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3002		ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3003		ret |= get_user(r4.rt_window, &(ur4->rt_window));
3004		ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3005		ret |= get_user(rtdev, &(ur4->rt_dev));
3006		if (rtdev) {
3007			ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3008			r4.rt_dev = (char __user __force *)devname;
3009			devname[15] = 0;
3010		} else
3011			r4.rt_dev = NULL;
3012
3013		r = (void *) &r4;
3014	}
3015
3016	if (ret) {
3017		ret = -EFAULT;
3018		goto out;
3019	}
3020
3021	set_fs(KERNEL_DS);
3022	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3023	set_fs(old_fs);
3024
3025out:
3026	return ret;
3027}
3028
3029/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3030 * for some operations; this forces use of the newer bridge-utils that
3031 * use compatible ioctls
3032 */
3033static int old_bridge_ioctl(compat_ulong_t __user *argp)
3034{
3035	compat_ulong_t tmp;
3036
3037	if (get_user(tmp, argp))
3038		return -EFAULT;
3039	if (tmp == BRCTL_GET_VERSION)
3040		return BRCTL_VERSION + 1;
3041	return -EINVAL;
3042}
3043
3044static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3045			 unsigned int cmd, unsigned long arg)
3046{
3047	void __user *argp = compat_ptr(arg);
3048	struct sock *sk = sock->sk;
3049	struct net *net = sock_net(sk);
3050
3051	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3052		return compat_ifr_data_ioctl(net, cmd, argp);
3053
3054	switch (cmd) {
3055	case SIOCSIFBR:
3056	case SIOCGIFBR:
3057		return old_bridge_ioctl(argp);
3058	case SIOCGIFNAME:
3059		return dev_ifname32(net, argp);
3060	case SIOCGIFCONF:
3061		return dev_ifconf(net, argp);
3062	case SIOCETHTOOL:
3063		return ethtool_ioctl(net, argp);
3064	case SIOCWANDEV:
3065		return compat_siocwandev(net, argp);
3066	case SIOCGIFMAP:
3067	case SIOCSIFMAP:
3068		return compat_sioc_ifmap(net, cmd, argp);
3069	case SIOCBONDENSLAVE:
3070	case SIOCBONDRELEASE:
3071	case SIOCBONDSETHWADDR:
3072	case SIOCBONDCHANGEACTIVE:
3073		return bond_ioctl(net, cmd, argp);
3074	case SIOCADDRT:
3075	case SIOCDELRT:
3076		return routing_ioctl(net, sock, cmd, argp);
3077	case SIOCGSTAMP:
3078		return do_siocgstamp(net, sock, cmd, argp);
3079	case SIOCGSTAMPNS:
3080		return do_siocgstampns(net, sock, cmd, argp);
3081	case SIOCBONDSLAVEINFOQUERY:
3082	case SIOCBONDINFOQUERY:
3083	case SIOCSHWTSTAMP:
3084	case SIOCGHWTSTAMP:
3085		return compat_ifr_data_ioctl(net, cmd, argp);
3086
3087	case FIOSETOWN:
3088	case SIOCSPGRP:
3089	case FIOGETOWN:
3090	case SIOCGPGRP:
3091	case SIOCBRADDBR:
3092	case SIOCBRDELBR:
3093	case SIOCGIFVLAN:
3094	case SIOCSIFVLAN:
3095	case SIOCADDDLCI:
3096	case SIOCDELDLCI:
3097		return sock_ioctl(file, cmd, arg);
3098
3099	case SIOCGIFFLAGS:
3100	case SIOCSIFFLAGS:
3101	case SIOCGIFMETRIC:
3102	case SIOCSIFMETRIC:
3103	case SIOCGIFMTU:
3104	case SIOCSIFMTU:
3105	case SIOCGIFMEM:
3106	case SIOCSIFMEM:
3107	case SIOCGIFHWADDR:
3108	case SIOCSIFHWADDR:
3109	case SIOCADDMULTI:
3110	case SIOCDELMULTI:
3111	case SIOCGIFINDEX:
3112	case SIOCGIFADDR:
3113	case SIOCSIFADDR:
3114	case SIOCSIFHWBROADCAST:
3115	case SIOCDIFADDR:
3116	case SIOCGIFBRDADDR:
3117	case SIOCSIFBRDADDR:
3118	case SIOCGIFDSTADDR:
3119	case SIOCSIFDSTADDR:
3120	case SIOCGIFNETMASK:
3121	case SIOCSIFNETMASK:
3122	case SIOCSIFPFLAGS:
3123	case SIOCGIFPFLAGS:
3124	case SIOCGIFTXQLEN:
3125	case SIOCSIFTXQLEN:
3126	case SIOCBRADDIF:
3127	case SIOCBRDELIF:
3128	case SIOCSIFNAME:
3129	case SIOCGMIIPHY:
3130	case SIOCGMIIREG:
3131	case SIOCSMIIREG:
3132		return dev_ifsioc(net, sock, cmd, argp);
3133
3134	case SIOCSARP:
3135	case SIOCGARP:
3136	case SIOCDARP:
3137	case SIOCATMARK:
3138		return sock_do_ioctl(net, sock, cmd, arg);
3139	}
3140
3141	return -ENOIOCTLCMD;
3142}
3143
3144static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3145			      unsigned long arg)
3146{
3147	struct socket *sock = file->private_data;
3148	int ret = -ENOIOCTLCMD;
3149	struct sock *sk;
3150	struct net *net;
3151
3152	sk = sock->sk;
3153	net = sock_net(sk);
3154
3155	if (sock->ops->compat_ioctl)
3156		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3157
3158	if (ret == -ENOIOCTLCMD &&
3159	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3160		ret = compat_wext_handle_ioctl(net, cmd, arg);
3161
3162	if (ret == -ENOIOCTLCMD)
3163		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3164
3165	return ret;
3166}
3167#endif
3168
3169int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3170{
3171	return sock->ops->bind(sock, addr, addrlen);
3172}
3173EXPORT_SYMBOL(kernel_bind);
3174
3175int kernel_listen(struct socket *sock, int backlog)
3176{
3177	return sock->ops->listen(sock, backlog);
3178}
3179EXPORT_SYMBOL(kernel_listen);
3180
3181int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3182{
3183	struct sock *sk = sock->sk;
3184	int err;
3185
3186	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3187			       newsock);
3188	if (err < 0)
3189		goto done;
3190
3191	err = sock->ops->accept(sock, *newsock, flags);
3192	if (err < 0) {
3193		sock_release(*newsock);
3194		*newsock = NULL;
3195		goto done;
3196	}
3197
3198	(*newsock)->ops = sock->ops;
3199	__module_get((*newsock)->ops->owner);
3200
3201done:
3202	return err;
3203}
3204EXPORT_SYMBOL(kernel_accept);
3205
3206int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3207		   int flags)
3208{
3209	return sock->ops->connect(sock, addr, addrlen, flags);
3210}
3211EXPORT_SYMBOL(kernel_connect);
3212
3213int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3214			 int *addrlen)
3215{
3216	return sock->ops->getname(sock, addr, addrlen, 0);
3217}
3218EXPORT_SYMBOL(kernel_getsockname);
3219
3220int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3221			 int *addrlen)
3222{
3223	return sock->ops->getname(sock, addr, addrlen, 1);
3224}
3225EXPORT_SYMBOL(kernel_getpeername);
3226
3227int kernel_getsockopt(struct socket *sock, int level, int optname,
3228			char *optval, int *optlen)
3229{
3230	mm_segment_t oldfs = get_fs();
3231	char __user *uoptval;
3232	int __user *uoptlen;
3233	int err;
3234
3235	uoptval = (char __user __force *) optval;
3236	uoptlen = (int __user __force *) optlen;
3237
3238	set_fs(KERNEL_DS);
3239	if (level == SOL_SOCKET)
3240		err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3241	else
3242		err = sock->ops->getsockopt(sock, level, optname, uoptval,
3243					    uoptlen);
3244	set_fs(oldfs);
3245	return err;
3246}
3247EXPORT_SYMBOL(kernel_getsockopt);
3248
3249int kernel_setsockopt(struct socket *sock, int level, int optname,
3250			char *optval, unsigned int optlen)
3251{
3252	mm_segment_t oldfs = get_fs();
3253	char __user *uoptval;
3254	int err;
3255
3256	uoptval = (char __user __force *) optval;
3257
3258	set_fs(KERNEL_DS);
3259	if (level == SOL_SOCKET)
3260		err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3261	else
3262		err = sock->ops->setsockopt(sock, level, optname, uoptval,
3263					    optlen);
3264	set_fs(oldfs);
3265	return err;
3266}
3267EXPORT_SYMBOL(kernel_setsockopt);
3268
3269int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3270		    size_t size, int flags)
3271{
3272	if (sock->ops->sendpage)
3273		return sock->ops->sendpage(sock, page, offset, size, flags);
3274
3275	return sock_no_sendpage(sock, page, offset, size, flags);
3276}
3277EXPORT_SYMBOL(kernel_sendpage);
3278
3279int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3280{
3281	mm_segment_t oldfs = get_fs();
3282	int err;
3283
3284	set_fs(KERNEL_DS);
3285	err = sock->ops->ioctl(sock, cmd, arg);
3286	set_fs(oldfs);
3287
3288	return err;
3289}
3290EXPORT_SYMBOL(kernel_sock_ioctl);
3291
3292int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3293{
3294	return sock->ops->shutdown(sock, how);
3295}
3296EXPORT_SYMBOL(kernel_sock_shutdown);
3297