linux - net/socket.c

3296 lines · 80.8 KiB

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, <[email protected]>
7	* Ross Biro
8	* Fred N. van Kempen, <[email protected]>
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 <linux/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
112	unsigned int sysctl_net_busy_read __read_mostly;
113	unsigned int sysctl_net_busy_poll __read_mostly;
114	#endif
115
116	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to);
117	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from);
118	static int sock_mmap(struct file file, struct vm_area_struct vma);
119
120	static int sock_close(struct inode inode, struct file file);
121	static __poll_t sock_poll(struct file *file,
122	struct poll_table_struct *wait);
123	static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124	#ifdef CONFIG_COMPAT
125	static long compat_sock_ioctl(struct file *file,
126	unsigned int cmd, unsigned long arg);
127	#endif
128	static int sock_fasync(int fd, struct file *filp, int on);
129	static ssize_t sock_sendpage(struct file file, struct page page,
130	int offset, size_t size, loff_t *ppos, int more);
131	static 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
140	static 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
162	static DEFINE_SPINLOCK(net_family_lock);
163	static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164
165	/*
166	* Support routines.
167	* Move socket addresses back and forth across the kernel/user
168	* divide and look after the messy bits.
169	*/
170
171	/**
172	* move_addr_to_kernel - copy a socket address into kernel space
173	* @uaddr: Address in user space
174	* @kaddr: Address in kernel space
175	* @ulen: Length in user space
176	*
177	* The address is copied into kernel space. If the provided address is
178	* too long an error code of -EINVAL is returned. If the copy gives
179	* invalid addresses -EFAULT is returned. On a success 0 is returned.
180	*/
181
182	int move_addr_to_kernel(void __user uaddr, int ulen, struct sockaddr_storage kaddr)
183	{
184	if (ulen < 0 \|\| ulen > sizeof(struct sockaddr_storage))
185	return -EINVAL;
186	if (ulen == 0)
187	return 0;
188	if (copy_from_user(kaddr, uaddr, ulen))
189	return -EFAULT;
190	return audit_sockaddr(ulen, kaddr);
191	}
192
193	/**
194	* move_addr_to_user - copy an address to user space
195	* @kaddr: kernel space address
196	* @klen: length of address in kernel
197	* @uaddr: user space address
198	* @ulen: pointer to user length field
199	*
200	* The value pointed to by ulen on entry is the buffer length available.
201	* This is overwritten with the buffer space used. -EINVAL is returned
202	* if an overlong buffer is specified or a negative buffer size. -EFAULT
203	* is returned if either the buffer or the length field are not
204	* accessible.
205	* After copying the data up to the limit the user specifies, the true
206	* length of the data is written over the length limit the user
207	* specified. Zero is returned for a success.
208	*/
209
210	static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
211	void __user uaddr, int __user ulen)
212	{
213	int err;
214	int len;
215
216	BUG_ON(klen > sizeof(struct sockaddr_storage));
217	err = get_user(len, ulen);
218	if (err)
219	return err;
220	if (len > klen)
221	len = klen;
222	if (len < 0)
223	return -EINVAL;
224	if (len) {
225	if (audit_sockaddr(klen, kaddr))
226	return -ENOMEM;
227	if (copy_to_user(uaddr, kaddr, len))
228	return -EFAULT;
229	}
230	/*
231	* "fromlen shall refer to the value before truncation.."
232	* 1003.1g
233	*/
234	return __put_user(klen, ulen);
235	}
236
237	static struct kmem_cache *sock_inode_cachep __read_mostly;
238
239	static struct inode sock_alloc_inode(struct super_block sb)
240	{
241	struct socket_alloc *ei;
242	struct socket_wq *wq;
243
244	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
245	if (!ei)
246	return NULL;
247	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
248	if (!wq) {
249	kmem_cache_free(sock_inode_cachep, ei);
250	return NULL;
251	}
252	init_waitqueue_head(&wq->wait);
253	wq->fasync_list = NULL;
254	wq->flags = 0;
255	RCU_INIT_POINTER(ei->socket.wq, wq);
256
257	ei->socket.state = SS_UNCONNECTED;
258	ei->socket.flags = 0;
259	ei->socket.ops = NULL;
260	ei->socket.sk = NULL;
261	ei->socket.file = NULL;
262
263	return &ei->vfs_inode;
264	}
265
266	static void sock_destroy_inode(struct inode *inode)
267	{
268	struct socket_alloc *ei;
269	struct socket_wq *wq;
270
271	ei = container_of(inode, struct socket_alloc, vfs_inode);
272	wq = rcu_dereference_protected(ei->socket.wq, 1);
273	kfree_rcu(wq, rcu);
274	kmem_cache_free(sock_inode_cachep, ei);
275	}
276
277	static void init_once(void *foo)
278	{
279	struct socket_alloc ei = (struct socket_alloc )foo;
280
281	inode_init_once(&ei->vfs_inode);
282	}
283
284	static void init_inodecache(void)
285	{
286	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
287	sizeof(struct socket_alloc),
288	0,
289	(SLAB_HWCACHE_ALIGN \|
290	SLAB_RECLAIM_ACCOUNT \|
291	SLAB_MEM_SPREAD \| SLAB_ACCOUNT),
292	init_once);
293	BUG_ON(sock_inode_cachep == NULL);
294	}
295
296	static const struct super_operations sockfs_ops = {
297	.alloc_inode = sock_alloc_inode,
298	.destroy_inode = sock_destroy_inode,
299	.statfs = simple_statfs,
300	};
301
302	/*
303	* sockfs_dname() is called from d_path().
304	*/
305	static char sockfs_dname(struct dentry dentry, char *buffer, int buflen)
306	{
307	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
308	d_inode(dentry)->i_ino);
309	}
310
311	static const struct dentry_operations sockfs_dentry_operations = {
312	.d_dname = sockfs_dname,
313	};
314
315	static int sockfs_xattr_get(const struct xattr_handler *handler,
316	struct dentry dentry, struct inode inode,
317	const char suffix, void value, size_t size)
318	{
319	if (value) {
320	if (dentry->d_name.len + 1 > size)
321	return -ERANGE;
322	memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
323	}
324	return dentry->d_name.len + 1;
325	}
326
327	#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
328	#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
329	#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
330
331	static const struct xattr_handler sockfs_xattr_handler = {
332	.name = XATTR_NAME_SOCKPROTONAME,
333	.get = sockfs_xattr_get,
334	};
335
336	static int sockfs_security_xattr_set(const struct xattr_handler *handler,
337	struct dentry dentry, struct inode inode,
338	const char suffix, const void value,
339	size_t size, int flags)
340	{
341	/* Handled by LSM. */
342	return -EAGAIN;
343	}
344
345	static const struct xattr_handler sockfs_security_xattr_handler = {
346	.prefix = XATTR_SECURITY_PREFIX,
347	.set = sockfs_security_xattr_set,
348	};
349
350	static const struct xattr_handler *sockfs_xattr_handlers[] = {
351	&sockfs_xattr_handler,
352	&sockfs_security_xattr_handler,
353	NULL
354	};
355
356	static struct dentry sockfs_mount(struct file_system_type fs_type,
357	int flags, const char dev_name, void data)
358	{
359	return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
360	sockfs_xattr_handlers,
361	&sockfs_dentry_operations, SOCKFS_MAGIC);
362	}
363
364	static struct vfsmount *sock_mnt __read_mostly;
365
366	static struct file_system_type sock_fs_type = {
367	.name = "sockfs",
368	.mount = sockfs_mount,
369	.kill_sb = kill_anon_super,
370	};
371
372	/*
373	* Obtains the first available file descriptor and sets it up for use.
374	*
375	* These functions create file structures and maps them to fd space
376	* of the current process. On success it returns file descriptor
377	* and file struct implicitly stored in sock->file.
378	* Note that another thread may close file descriptor before we return
379	* from this function. We use the fact that now we do not refer
380	* to socket after mapping. If one day we will need it, this
381	* function will increment ref. count on file by 1.
382	*
383	* In any case returned fd MAY BE not valid!
384	* This race condition is unavoidable
385	* with shared fd spaces, we cannot solve it inside kernel,
386	* but we take care of internal coherence yet.
387	*/
388
389	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
390	{
391	struct qstr name = { .name = "" };
392	struct path path;
393	struct file *file;
394
395	if (dname) {
396	name.name = dname;
397	name.len = strlen(name.name);
398	} else if (sock->sk) {
399	name.name = sock->sk->sk_prot_creator->name;
400	name.len = strlen(name.name);
401	}
402	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
403	if (unlikely(!path.dentry)) {
404	sock_release(sock);
405	return ERR_PTR(-ENOMEM);
406	}
407	path.mnt = mntget(sock_mnt);
408
409	d_instantiate(path.dentry, SOCK_INODE(sock));
410
411	file = alloc_file(&path, FMODE_READ \| FMODE_WRITE,
412	&socket_file_ops);
413	if (IS_ERR(file)) {
414	/* drop dentry, keep inode for a bit */
415	ihold(d_inode(path.dentry));
416	path_put(&path);
417	/* ... and now kill it properly */
418	sock_release(sock);
419	return file;
420	}
421
422	sock->file = file;
423	file->f_flags = O_RDWR \| (flags & O_NONBLOCK);
424	file->private_data = sock;
425	return file;
426	}
427	EXPORT_SYMBOL(sock_alloc_file);
428
429	static int sock_map_fd(struct socket *sock, int flags)
430	{
431	struct file *newfile;
432	int fd = get_unused_fd_flags(flags);
433	if (unlikely(fd < 0)) {
434	sock_release(sock);
435	return fd;
436	}
437
438	newfile = sock_alloc_file(sock, flags, NULL);
439	if (likely(!IS_ERR(newfile))) {
440	fd_install(fd, newfile);
441	return fd;
442	}
443
444	put_unused_fd(fd);
445	return PTR_ERR(newfile);
446	}
447
448	struct socket sock_from_file(struct file file, int *err)
449	{
450	if (file->f_op == &socket_file_ops)
451	return file->private_data; /* set in sock_map_fd */
452
453	*err = -ENOTSOCK;
454	return NULL;
455	}
456	EXPORT_SYMBOL(sock_from_file);
457
458	/**
459	* sockfd_lookup - Go from a file number to its socket slot
460	* @fd: file handle
461	* @err: pointer to an error code return
462	*
463	* The file handle passed in is locked and the socket it is bound
464	* to is returned. If an error occurs the err pointer is overwritten
465	* with a negative errno code and NULL is returned. The function checks
466	* for both invalid handles and passing a handle which is not a socket.
467	*
468	* On a success the socket object pointer is returned.
469	*/
470
471	struct socket sockfd_lookup(int fd, int err)
472	{
473	struct file *file;
474	struct socket *sock;
475
476	file = fget(fd);
477	if (!file) {
478	*err = -EBADF;
479	return NULL;
480	}
481
482	sock = sock_from_file(file, err);
483	if (!sock)
484	fput(file);
485	return sock;
486	}
487	EXPORT_SYMBOL(sockfd_lookup);
488
489	static struct socket sockfd_lookup_light(int fd, int err, int *fput_needed)
490	{
491	struct fd f = fdget(fd);
492	struct socket *sock;
493
494	*err = -EBADF;
495	if (f.file) {
496	sock = sock_from_file(f.file, err);
497	if (likely(sock)) {
498	*fput_needed = f.flags;
499	return sock;
500	}
501	fdput(f);
502	}
503	return NULL;
504	}
505
506	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
507	size_t size)
508	{
509	ssize_t len;
510	ssize_t used = 0;
511
512	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
513	if (len < 0)
514	return len;
515	used += len;
516	if (buffer) {
517	if (size < used)
518	return -ERANGE;
519	buffer += len;
520	}
521
522	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
523	used += len;
524	if (buffer) {
525	if (size < used)
526	return -ERANGE;
527	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
528	buffer += len;
529	}
530
531	return used;
532	}
533
534	static int sockfs_setattr(struct dentry dentry, struct iattr iattr)
535	{
536	int err = simple_setattr(dentry, iattr);
537
538	if (!err && (iattr->ia_valid & ATTR_UID)) {
539	struct socket *sock = SOCKET_I(d_inode(dentry));
540
541	sock->sk->sk_uid = iattr->ia_uid;
542	}
543
544	return err;
545	}
546
547	static const struct inode_operations sockfs_inode_ops = {
548	.listxattr = sockfs_listxattr,
549	.setattr = sockfs_setattr,
550	};
551
552	/**
553	* sock_alloc - allocate a socket
554	*
555	* Allocate a new inode and socket object. The two are bound together
556	* and initialised. The socket is then returned. If we are out of inodes
557	* NULL is returned.
558	*/
559
560	struct socket *sock_alloc(void)
561	{
562	struct inode *inode;
563	struct socket *sock;
564
565	inode = new_inode_pseudo(sock_mnt->mnt_sb);
566	if (!inode)
567	return NULL;
568
569	sock = SOCKET_I(inode);
570
571	inode->i_ino = get_next_ino();
572	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
573	inode->i_uid = current_fsuid();
574	inode->i_gid = current_fsgid();
575	inode->i_op = &sockfs_inode_ops;
576
577	return sock;
578	}
579	EXPORT_SYMBOL(sock_alloc);
580
581	/**
582	* sock_release - close a socket
583	* @sock: socket to close
584	*
585	* The socket is released from the protocol stack if it has a release
586	* callback, and the inode is then released if the socket is bound to
587	* an inode not a file.
588	*/
589
590	void sock_release(struct socket *sock)
591	{
592	if (sock->ops) {
593	struct module *owner = sock->ops->owner;
594
595	sock->ops->release(sock);
596	sock->ops = NULL;
597	module_put(owner);
598	}
599
600	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
601	pr_err("%s: fasync list not empty!\n", __func__);
602
603	if (!sock->file) {
604	iput(SOCK_INODE(sock));
605	return;
606	}
607	sock->file = NULL;
608	}
609	EXPORT_SYMBOL(sock_release);
610
611	void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
612	{
613	u8 flags = *tx_flags;
614
615	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
616	flags \|= SKBTX_HW_TSTAMP;
617
618	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
619	flags \|= SKBTX_SW_TSTAMP;
620
621	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
622	flags \|= SKBTX_SCHED_TSTAMP;
623
624	*tx_flags = flags;
625	}
626	EXPORT_SYMBOL(__sock_tx_timestamp);
627
628	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
629	{
630	int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
631	BUG_ON(ret == -EIOCBQUEUED);
632	return ret;
633	}
634
635	int sock_sendmsg(struct socket sock, struct msghdr msg)
636	{
637	int err = security_socket_sendmsg(sock, msg,
638	msg_data_left(msg));
639
640	return err ?: sock_sendmsg_nosec(sock, msg);
641	}
642	EXPORT_SYMBOL(sock_sendmsg);
643
644	int kernel_sendmsg(struct socket sock, struct msghdr msg,
645	struct kvec *vec, size_t num, size_t size)
646	{
647	iov_iter_kvec(&msg->msg_iter, WRITE \| ITER_KVEC, vec, num, size);
648	return sock_sendmsg(sock, msg);
649	}
650	EXPORT_SYMBOL(kernel_sendmsg);
651
652	int kernel_sendmsg_locked(struct sock sk, struct msghdr msg,
653	struct kvec *vec, size_t num, size_t size)
654	{
655	struct socket *sock = sk->sk_socket;
656
657	if (!sock->ops->sendmsg_locked)
658	return sock_no_sendmsg_locked(sk, msg, size);
659
660	iov_iter_kvec(&msg->msg_iter, WRITE \| ITER_KVEC, vec, num, size);
661
662	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
663	}
664	EXPORT_SYMBOL(kernel_sendmsg_locked);
665
666	static bool skb_is_err_queue(const struct sk_buff *skb)
667	{
668	/* pkt_type of skbs enqueued on the error queue are set to
669	* PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
670	* in recvmsg, since skbs received on a local socket will never
671	* have a pkt_type of PACKET_OUTGOING.
672	*/
673	return skb->pkt_type == PACKET_OUTGOING;
674	}
675
676	/* On transmit, software and hardware timestamps are returned independently.
677	* As the two skb clones share the hardware timestamp, which may be updated
678	* before the software timestamp is received, a hardware TX timestamp may be
679	* returned only if there is no software TX timestamp. Ignore false software
680	* timestamps, which may be made in the __sock_recv_timestamp() call when the
681	* option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
682	* hardware timestamp.
683	*/
684	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
685	{
686	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
687	}
688
689	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb)
690	{
691	struct scm_ts_pktinfo ts_pktinfo;
692	struct net_device *orig_dev;
693
694	if (!skb_mac_header_was_set(skb))
695	return;
696
697	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
698
699	rcu_read_lock();
700	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
701	if (orig_dev)
702	ts_pktinfo.if_index = orig_dev->ifindex;
703	rcu_read_unlock();
704
705	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
706	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
707	sizeof(ts_pktinfo), &ts_pktinfo);
708	}
709
710	/*
711	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
712	*/
713	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
714	struct sk_buff *skb)
715	{
716	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
717	struct scm_timestamping tss;
718	int empty = 1, false_tstamp = 0;
719	struct skb_shared_hwtstamps *shhwtstamps =
720	skb_hwtstamps(skb);
721
722	/* Race occurred between timestamp enabling and packet
723	receiving. Fill in the current time for now. */
724	if (need_software_tstamp && skb->tstamp == 0) {
725	__net_timestamp(skb);
726	false_tstamp = 1;
727	}
728
729	if (need_software_tstamp) {
730	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
731	struct timeval tv;
732	skb_get_timestamp(skb, &tv);
733	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
734	sizeof(tv), &tv);
735	} else {
736	struct timespec ts;
737	skb_get_timestampns(skb, &ts);
738	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
739	sizeof(ts), &ts);
740	}
741	}
742
743	memset(&tss, 0, sizeof(tss));
744	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
745	ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
746	empty = 0;
747	if (shhwtstamps &&
748	(sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
749	!skb_is_swtx_tstamp(skb, false_tstamp) &&
750	ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
751	empty = 0;
752	if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
753	!skb_is_err_queue(skb))
754	put_ts_pktinfo(msg, skb);
755	}
756	if (!empty) {
757	put_cmsg(msg, SOL_SOCKET,
758	SCM_TIMESTAMPING, sizeof(tss), &tss);
759
760	if (skb_is_err_queue(skb) && skb->len &&
761	SKB_EXT_ERR(skb)->opt_stats)
762	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
763	skb->len, skb->data);
764	}
765	}
766	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
767
768	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
769	struct sk_buff *skb)
770	{
771	int ack;
772
773	if (!sock_flag(sk, SOCK_WIFI_STATUS))
774	return;
775	if (!skb->wifi_acked_valid)
776	return;
777
778	ack = skb->wifi_acked;
779
780	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
781	}
782	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
783
784	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
785	struct sk_buff *skb)
786	{
787	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
788	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
789	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
790	}
791
792	void __sock_recv_ts_and_drops(struct msghdr msg, struct sock sk,
793	struct sk_buff *skb)
794	{
795	sock_recv_timestamp(msg, sk, skb);
796	sock_recv_drops(msg, sk, skb);
797	}
798	EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
799
800	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
801	int flags)
802	{
803	return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
804	}
805
806	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
807	{
808	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
809
810	return err ?: sock_recvmsg_nosec(sock, msg, flags);
811	}
812	EXPORT_SYMBOL(sock_recvmsg);
813
814	/**
815	* kernel_recvmsg - Receive a message from a socket (kernel space)
816	* @sock: The socket to receive the message from
817	* @msg: Received message
818	* @vec: Input s/g array for message data
819	* @num: Size of input s/g array
820	* @size: Number of bytes to read
821	* @flags: Message flags (MSG_DONTWAIT, etc...)
822	*
823	* On return the msg structure contains the scatter/gather array passed in the
824	* vec argument. The array is modified so that it consists of the unfilled
825	* portion of the original array.
826	*
827	* The returned value is the total number of bytes received, or an error.
828	*/
829	int kernel_recvmsg(struct socket sock, struct msghdr msg,
830	struct kvec *vec, size_t num, size_t size, int flags)
831	{
832	mm_segment_t oldfs = get_fs();
833	int result;
834
835	iov_iter_kvec(&msg->msg_iter, READ \| ITER_KVEC, vec, num, size);
836	set_fs(KERNEL_DS);
837	result = sock_recvmsg(sock, msg, flags);
838	set_fs(oldfs);
839	return result;
840	}
841	EXPORT_SYMBOL(kernel_recvmsg);
842
843	static ssize_t sock_sendpage(struct file file, struct page page,
844	int offset, size_t size, loff_t *ppos, int more)
845	{
846	struct socket *sock;
847	int flags;
848
849	sock = file->private_data;
850
851	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
852	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
853	flags \|= more;
854
855	return kernel_sendpage(sock, page, offset, size, flags);
856	}
857
858	static ssize_t sock_splice_read(struct file file, loff_t ppos,
859	struct pipe_inode_info *pipe, size_t len,
860	unsigned int flags)
861	{
862	struct socket *sock = file->private_data;
863
864	if (unlikely(!sock->ops->splice_read))
865	return -EINVAL;
866
867	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
868	}
869
870	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
871	{
872	struct file *file = iocb->ki_filp;
873	struct socket *sock = file->private_data;
874	struct msghdr msg = {.msg_iter = *to,
875	.msg_iocb = iocb};
876	ssize_t res;
877
878	if (file->f_flags & O_NONBLOCK)
879	msg.msg_flags = MSG_DONTWAIT;
880
881	if (iocb->ki_pos != 0)
882	return -ESPIPE;
883
884	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
885	return 0;
886
887	res = sock_recvmsg(sock, &msg, msg.msg_flags);
888	*to = msg.msg_iter;
889	return res;
890	}
891
892	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
893	{
894	struct file *file = iocb->ki_filp;
895	struct socket *sock = file->private_data;
896	struct msghdr msg = {.msg_iter = *from,
897	.msg_iocb = iocb};
898	ssize_t res;
899
900	if (iocb->ki_pos != 0)
901	return -ESPIPE;
902
903	if (file->f_flags & O_NONBLOCK)
904	msg.msg_flags = MSG_DONTWAIT;
905
906	if (sock->type == SOCK_SEQPACKET)
907	msg.msg_flags \|= MSG_EOR;
908
909	res = sock_sendmsg(sock, &msg);
910	*from = msg.msg_iter;
911	return res;
912	}
913
914	/*
915	* Atomic setting of ioctl hooks to avoid race
916	* with module unload.
917	*/
918
919	static DEFINE_MUTEX(br_ioctl_mutex);
920	static int (br_ioctl_hook) (struct net , unsigned int cmd, void __user *arg);
921
922	void brioctl_set(int (hook) (struct net , unsigned int, void __user *))
923	{
924	mutex_lock(&br_ioctl_mutex);
925	br_ioctl_hook = hook;
926	mutex_unlock(&br_ioctl_mutex);
927	}
928	EXPORT_SYMBOL(brioctl_set);
929
930	static DEFINE_MUTEX(vlan_ioctl_mutex);
931	static int (vlan_ioctl_hook) (struct net , void __user *arg);
932
933	void vlan_ioctl_set(int (hook) (struct net , void __user *))
934	{
935	mutex_lock(&vlan_ioctl_mutex);
936	vlan_ioctl_hook = hook;
937	mutex_unlock(&vlan_ioctl_mutex);
938	}
939	EXPORT_SYMBOL(vlan_ioctl_set);
940
941	static DEFINE_MUTEX(dlci_ioctl_mutex);
942	static int (dlci_ioctl_hook) (unsigned int, void __user );
943
944	void dlci_ioctl_set(int (hook) (unsigned int, void __user ))
945	{
946	mutex_lock(&dlci_ioctl_mutex);
947	dlci_ioctl_hook = hook;
948	mutex_unlock(&dlci_ioctl_mutex);
949	}
950	EXPORT_SYMBOL(dlci_ioctl_set);
951
952	static long sock_do_ioctl(struct net net, struct socket sock,
953	unsigned int cmd, unsigned long arg)
954	{
955	int err;
956	void __user argp = (void __user )arg;
957
958	err = sock->ops->ioctl(sock, cmd, arg);
959
960	/*
961	* If this ioctl is unknown try to hand it down
962	* to the NIC driver.
963	*/
964	if (err != -ENOIOCTLCMD)
965	return err;
966
967	if (cmd == SIOCGIFCONF) {
968	struct ifconf ifc;
969	if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
970	return -EFAULT;
971	rtnl_lock();
972	err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
973	rtnl_unlock();
974	if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
975	err = -EFAULT;
976	} else {
977	struct ifreq ifr;
978	bool need_copyout;
979	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
980	return -EFAULT;
981	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
982	if (!err && need_copyout)
983	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
984	return -EFAULT;
985	}
986	return err;
987	}
988
989	/*
990	* With an ioctl, arg may well be a user mode pointer, but we don't know
991	* what to do with it - that's up to the protocol still.
992	*/
993
994	static struct ns_common get_net_ns(struct ns_common ns)
995	{
996	return &get_net(container_of(ns, struct net, ns))->ns;
997	}
998
999	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1000	{
1001	struct socket *sock;
1002	struct sock *sk;
1003	void __user argp = (void __user )arg;
1004	int pid, err;
1005	struct net *net;
1006
1007	sock = file->private_data;
1008	sk = sock->sk;
1009	net = sock_net(sk);
1010	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1011	struct ifreq ifr;
1012	bool need_copyout;
1013	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1014	return -EFAULT;
1015	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1016	if (!err && need_copyout)
1017	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1018	return -EFAULT;
1019	} else
1020	#ifdef CONFIG_WEXT_CORE
1021	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1022	err = wext_handle_ioctl(net, cmd, argp);
1023	} else
1024	#endif
1025	switch (cmd) {
1026	case FIOSETOWN:
1027	case SIOCSPGRP:
1028	err = -EFAULT;
1029	if (get_user(pid, (int __user *)argp))
1030	break;
1031	err = f_setown(sock->file, pid, 1);
1032	break;
1033	case FIOGETOWN:
1034	case SIOCGPGRP:
1035	err = put_user(f_getown(sock->file),
1036	(int __user *)argp);
1037	break;
1038	case SIOCGIFBR:
1039	case SIOCSIFBR:
1040	case SIOCBRADDBR:
1041	case SIOCBRDELBR:
1042	err = -ENOPKG;
1043	if (!br_ioctl_hook)
1044	request_module("bridge");
1045
1046	mutex_lock(&br_ioctl_mutex);
1047	if (br_ioctl_hook)
1048	err = br_ioctl_hook(net, cmd, argp);
1049	mutex_unlock(&br_ioctl_mutex);
1050	break;
1051	case SIOCGIFVLAN:
1052	case SIOCSIFVLAN:
1053	err = -ENOPKG;
1054	if (!vlan_ioctl_hook)
1055	request_module("8021q");
1056
1057	mutex_lock(&vlan_ioctl_mutex);
1058	if (vlan_ioctl_hook)
1059	err = vlan_ioctl_hook(net, argp);
1060	mutex_unlock(&vlan_ioctl_mutex);
1061	break;
1062	case SIOCADDDLCI:
1063	case SIOCDELDLCI:
1064	err = -ENOPKG;
1065	if (!dlci_ioctl_hook)
1066	request_module("dlci");
1067
1068	mutex_lock(&dlci_ioctl_mutex);
1069	if (dlci_ioctl_hook)
1070	err = dlci_ioctl_hook(cmd, argp);
1071	mutex_unlock(&dlci_ioctl_mutex);
1072	break;
1073	case SIOCGSKNS:
1074	err = -EPERM;
1075	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1076	break;
1077
1078	err = open_related_ns(&net->ns, get_net_ns);
1079	break;
1080	default:
1081	err = sock_do_ioctl(net, sock, cmd, arg);
1082	break;
1083	}
1084	return err;
1085	}
1086
1087	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1088	{
1089	int err;
1090	struct socket *sock = NULL;
1091
1092	err = security_socket_create(family, type, protocol, 1);
1093	if (err)
1094	goto out;
1095
1096	sock = sock_alloc();
1097	if (!sock) {
1098	err = -ENOMEM;
1099	goto out;
1100	}
1101
1102	sock->type = type;
1103	err = security_socket_post_create(sock, family, type, protocol, 1);
1104	if (err)
1105	goto out_release;
1106
1107	out:
1108	*res = sock;
1109	return err;
1110	out_release:
1111	sock_release(sock);
1112	sock = NULL;
1113	goto out;
1114	}
1115	EXPORT_SYMBOL(sock_create_lite);
1116
1117	/* No kernel lock held - perfect */
1118	static __poll_t sock_poll(struct file file, poll_table wait)
1119	{
1120	__poll_t busy_flag = 0;
1121	struct socket *sock;
1122
1123	/*
1124	* We can't return errors to poll, so it's either yes or no.
1125	*/
1126	sock = file->private_data;
1127
1128	if (sk_can_busy_loop(sock->sk)) {
1129	/* this socket can poll_ll so tell the system call */
1130	busy_flag = POLL_BUSY_LOOP;
1131
1132	/* once, only if requested by syscall */
1133	if (wait && (wait->_key & POLL_BUSY_LOOP))
1134	sk_busy_loop(sock->sk, 1);
1135	}
1136
1137	return busy_flag \| sock->ops->poll(file, sock, wait);
1138	}
1139
1140	static int sock_mmap(struct file file, struct vm_area_struct vma)
1141	{
1142	struct socket *sock = file->private_data;
1143
1144	return sock->ops->mmap(file, sock, vma);
1145	}
1146
1147	static int sock_close(struct inode inode, struct file filp)
1148	{
1149	sock_release(SOCKET_I(inode));
1150	return 0;
1151	}
1152
1153	/*
1154	* Update the socket async list
1155	*
1156	* Fasync_list locking strategy.
1157	*
1158	* 1. fasync_list is modified only under process context socket lock
1159	* i.e. under semaphore.
1160	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1161	* or under socket lock
1162	*/
1163
1164	static int sock_fasync(int fd, struct file *filp, int on)
1165	{
1166	struct socket *sock = filp->private_data;
1167	struct sock *sk = sock->sk;
1168	struct socket_wq *wq;
1169
1170	if (sk == NULL)
1171	return -EINVAL;
1172
1173	lock_sock(sk);
1174	wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1175	fasync_helper(fd, filp, on, &wq->fasync_list);
1176
1177	if (!wq->fasync_list)
1178	sock_reset_flag(sk, SOCK_FASYNC);
1179	else
1180	sock_set_flag(sk, SOCK_FASYNC);
1181
1182	release_sock(sk);
1183	return 0;
1184	}
1185
1186	/* This function may be called only under rcu_lock */
1187
1188	int sock_wake_async(struct socket_wq *wq, int how, int band)
1189	{
1190	if (!wq \|\| !wq->fasync_list)
1191	return -1;
1192
1193	switch (how) {
1194	case SOCK_WAKE_WAITD:
1195	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1196	break;
1197	goto call_kill;
1198	case SOCK_WAKE_SPACE:
1199	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1200	break;
1201	/* fall through */
1202	case SOCK_WAKE_IO:
1203	call_kill:
1204	kill_fasync(&wq->fasync_list, SIGIO, band);
1205	break;
1206	case SOCK_WAKE_URG:
1207	kill_fasync(&wq->fasync_list, SIGURG, band);
1208	}
1209
1210	return 0;
1211	}
1212	EXPORT_SYMBOL(sock_wake_async);
1213
1214	int __sock_create(struct net *net, int family, int type, int protocol,
1215	struct socket **res, int kern)
1216	{
1217	int err;
1218	struct socket *sock;
1219	const struct net_proto_family *pf;
1220
1221	/*
1222	* Check protocol is in range
1223	*/
1224	if (family < 0 \|\| family >= NPROTO)
1225	return -EAFNOSUPPORT;
1226	if (type < 0 \|\| type >= SOCK_MAX)
1227	return -EINVAL;
1228
1229	/* Compatibility.
1230
1231	This uglymoron is moved from INET layer to here to avoid
1232	deadlock in module load.
1233	*/
1234	if (family == PF_INET && type == SOCK_PACKET) {
1235	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1236	current->comm);
1237	family = PF_PACKET;
1238	}
1239
1240	err = security_socket_create(family, type, protocol, kern);
1241	if (err)
1242	return err;
1243
1244	/*
1245	* Allocate the socket and allow the family to set things up. if
1246	* the protocol is 0, the family is instructed to select an appropriate
1247	* default.
1248	*/
1249	sock = sock_alloc();
1250	if (!sock) {
1251	net_warn_ratelimited("socket: no more sockets\n");
1252	return -ENFILE; /* Not exactly a match, but its the
1253	closest posix thing */
1254	}
1255
1256	sock->type = type;
1257
1258	#ifdef CONFIG_MODULES
1259	/* Attempt to load a protocol module if the find failed.
1260	*
1261	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1262	* requested real, full-featured networking support upon configuration.
1263	* Otherwise module support will break!
1264	*/
1265	if (rcu_access_pointer(net_families[family]) == NULL)
1266	request_module("net-pf-%d", family);
1267	#endif
1268
1269	rcu_read_lock();
1270	pf = rcu_dereference(net_families[family]);
1271	err = -EAFNOSUPPORT;
1272	if (!pf)
1273	goto out_release;
1274
1275	/*
1276	* We will call the ->create function, that possibly is in a loadable
1277	* module, so we have to bump that loadable module refcnt first.
1278	*/
1279	if (!try_module_get(pf->owner))
1280	goto out_release;
1281
1282	/* Now protected by module ref count */
1283	rcu_read_unlock();
1284
1285	err = pf->create(net, sock, protocol, kern);
1286	if (err < 0)
1287	goto out_module_put;
1288
1289	/*
1290	* Now to bump the refcnt of the [loadable] module that owns this
1291	* socket at sock_release time we decrement its refcnt.
1292	*/
1293	if (!try_module_get(sock->ops->owner))
1294	goto out_module_busy;
1295
1296	/*
1297	* Now that we're done with the ->create function, the [loadable]
1298	* module can have its refcnt decremented
1299	*/
1300	module_put(pf->owner);
1301	err = security_socket_post_create(sock, family, type, protocol, kern);
1302	if (err)
1303	goto out_sock_release;
1304	*res = sock;
1305
1306	return 0;
1307
1308	out_module_busy:
1309	err = -EAFNOSUPPORT;
1310	out_module_put:
1311	sock->ops = NULL;
1312	module_put(pf->owner);
1313	out_sock_release:
1314	sock_release(sock);
1315	return err;
1316
1317	out_release:
1318	rcu_read_unlock();
1319	goto out_sock_release;
1320	}
1321	EXPORT_SYMBOL(__sock_create);
1322
1323	int sock_create(int family, int type, int protocol, struct socket **res)
1324	{
1325	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1326	}
1327	EXPORT_SYMBOL(sock_create);
1328
1329	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1330	{
1331	return __sock_create(net, family, type, protocol, res, 1);
1332	}
1333	EXPORT_SYMBOL(sock_create_kern);
1334
1335	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1336	{
1337	int retval;
1338	struct socket *sock;
1339	int flags;
1340
1341	/* Check the SOCK_* constants for consistency. */
1342	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1343	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1344	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1345	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1346
1347	flags = type & ~SOCK_TYPE_MASK;
1348	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1349	return -EINVAL;
1350	type &= SOCK_TYPE_MASK;
1351
1352	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1353	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1354
1355	retval = sock_create(family, type, protocol, &sock);
1356	if (retval < 0)
1357	return retval;
1358
1359	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1360	}
1361
1362	/*
1363	* Create a pair of connected sockets.
1364	*/
1365
1366	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1367	int __user *, usockvec)
1368	{
1369	struct socket sock1, sock2;
1370	int fd1, fd2, err;
1371	struct file newfile1, newfile2;
1372	int flags;
1373
1374	flags = type & ~SOCK_TYPE_MASK;
1375	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1376	return -EINVAL;
1377	type &= SOCK_TYPE_MASK;
1378
1379	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1380	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1381
1382	/*
1383	* reserve descriptors and make sure we won't fail
1384	* to return them to userland.
1385	*/
1386	fd1 = get_unused_fd_flags(flags);
1387	if (unlikely(fd1 < 0))
1388	return fd1;
1389
1390	fd2 = get_unused_fd_flags(flags);
1391	if (unlikely(fd2 < 0)) {
1392	put_unused_fd(fd1);
1393	return fd2;
1394	}
1395
1396	err = put_user(fd1, &usockvec[0]);
1397	if (err)
1398	goto out;
1399
1400	err = put_user(fd2, &usockvec[1]);
1401	if (err)
1402	goto out;
1403
1404	/*
1405	* Obtain the first socket and check if the underlying protocol
1406	* supports the socketpair call.
1407	*/
1408
1409	err = sock_create(family, type, protocol, &sock1);
1410	if (unlikely(err < 0))
1411	goto out;
1412
1413	err = sock_create(family, type, protocol, &sock2);
1414	if (unlikely(err < 0)) {
1415	sock_release(sock1);
1416	goto out;
1417	}
1418
1419	err = sock1->ops->socketpair(sock1, sock2);
1420	if (unlikely(err < 0)) {
1421	sock_release(sock2);
1422	sock_release(sock1);
1423	goto out;
1424	}
1425
1426	newfile1 = sock_alloc_file(sock1, flags, NULL);
1427	if (IS_ERR(newfile1)) {
1428	err = PTR_ERR(newfile1);
1429	sock_release(sock2);
1430	goto out;
1431	}
1432
1433	newfile2 = sock_alloc_file(sock2, flags, NULL);
1434	if (IS_ERR(newfile2)) {
1435	err = PTR_ERR(newfile2);
1436	fput(newfile1);
1437	goto out;
1438	}
1439
1440	audit_fd_pair(fd1, fd2);
1441
1442	fd_install(fd1, newfile1);
1443	fd_install(fd2, newfile2);
1444	return 0;
1445
1446	out:
1447	put_unused_fd(fd2);
1448	put_unused_fd(fd1);
1449	return err;
1450	}
1451
1452	/*
1453	* Bind a name to a socket. Nothing much to do here since it's
1454	* the protocol's responsibility to handle the local address.
1455	*
1456	* We move the socket address to kernel space before we call
1457	* the protocol layer (having also checked the address is ok).
1458	*/
1459
1460	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1461	{
1462	struct socket *sock;
1463	struct sockaddr_storage address;
1464	int err, fput_needed;
1465
1466	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1467	if (sock) {
1468	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1469	if (err >= 0) {
1470	err = security_socket_bind(sock,
1471	(struct sockaddr *)&address,
1472	addrlen);
1473	if (!err)
1474	err = sock->ops->bind(sock,
1475	(struct sockaddr *)
1476	&address, addrlen);
1477	}
1478	fput_light(sock->file, fput_needed);
1479	}
1480	return err;
1481	}
1482
1483	/*
1484	* Perform a listen. Basically, we allow the protocol to do anything
1485	* necessary for a listen, and if that works, we mark the socket as
1486	* ready for listening.
1487	*/
1488
1489	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1490	{
1491	struct socket *sock;
1492	int err, fput_needed;
1493	int somaxconn;
1494
1495	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1496	if (sock) {
1497	somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1498	if ((unsigned int)backlog > somaxconn)
1499	backlog = somaxconn;
1500
1501	err = security_socket_listen(sock, backlog);
1502	if (!err)
1503	err = sock->ops->listen(sock, backlog);
1504
1505	fput_light(sock->file, fput_needed);
1506	}
1507	return err;
1508	}
1509
1510	/*
1511	* For accept, we attempt to create a new socket, set up the link
1512	* with the client, wake up the client, then return the new
1513	* connected fd. We collect the address of the connector in kernel
1514	* space and move it to user at the very end. This is unclean because
1515	* we open the socket then return an error.
1516	*
1517	* 1003.1g adds the ability to recvmsg() to query connection pending
1518	* status to recvmsg. We need to add that support in a way thats
1519	* clean when we restucture accept also.
1520	*/
1521
1522	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1523	int __user *, upeer_addrlen, int, flags)
1524	{
1525	struct socket sock, newsock;
1526	struct file *newfile;
1527	int err, len, newfd, fput_needed;
1528	struct sockaddr_storage address;
1529
1530	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1531	return -EINVAL;
1532
1533	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1534	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1535
1536	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1537	if (!sock)
1538	goto out;
1539
1540	err = -ENFILE;
1541	newsock = sock_alloc();
1542	if (!newsock)
1543	goto out_put;
1544
1545	newsock->type = sock->type;
1546	newsock->ops = sock->ops;
1547
1548	/*
1549	* We don't need try_module_get here, as the listening socket (sock)
1550	* has the protocol module (sock->ops->owner) held.
1551	*/
1552	__module_get(newsock->ops->owner);
1553
1554	newfd = get_unused_fd_flags(flags);
1555	if (unlikely(newfd < 0)) {
1556	err = newfd;
1557	sock_release(newsock);
1558	goto out_put;
1559	}
1560	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1561	if (IS_ERR(newfile)) {
1562	err = PTR_ERR(newfile);
1563	put_unused_fd(newfd);
1564	goto out_put;
1565	}
1566
1567	err = security_socket_accept(sock, newsock);
1568	if (err)
1569	goto out_fd;
1570
1571	err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1572	if (err < 0)
1573	goto out_fd;
1574
1575	if (upeer_sockaddr) {
1576	if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1577	&len, 2) < 0) {
1578	err = -ECONNABORTED;
1579	goto out_fd;
1580	}
1581	err = move_addr_to_user(&address,
1582	len, upeer_sockaddr, upeer_addrlen);
1583	if (err < 0)
1584	goto out_fd;
1585	}
1586
1587	/* File flags are not inherited via accept() unlike another OSes. */
1588
1589	fd_install(newfd, newfile);
1590	err = newfd;
1591
1592	out_put:
1593	fput_light(sock->file, fput_needed);
1594	out:
1595	return err;
1596	out_fd:
1597	fput(newfile);
1598	put_unused_fd(newfd);
1599	goto out_put;
1600	}
1601
1602	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1603	int __user *, upeer_addrlen)
1604	{
1605	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1606	}
1607
1608	/*
1609	* Attempt to connect to a socket with the server address. The address
1610	* is in user space so we verify it is OK and move it to kernel space.
1611	*
1612	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1613	* break bindings
1614	*
1615	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1616	* other SEQPACKET protocols that take time to connect() as it doesn't
1617	* include the -EINPROGRESS status for such sockets.
1618	*/
1619
1620	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1621	int, addrlen)
1622	{
1623	struct socket *sock;
1624	struct sockaddr_storage address;
1625	int err, fput_needed;
1626
1627	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1628	if (!sock)
1629	goto out;
1630	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1631	if (err < 0)
1632	goto out_put;
1633
1634	err =
1635	security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1636	if (err)
1637	goto out_put;
1638
1639	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1640	sock->file->f_flags);
1641	out_put:
1642	fput_light(sock->file, fput_needed);
1643	out:
1644	return err;
1645	}
1646
1647	/*
1648	* Get the local address ('name') of a socket object. Move the obtained
1649	* name to user space.
1650	*/
1651
1652	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1653	int __user *, usockaddr_len)
1654	{
1655	struct socket *sock;
1656	struct sockaddr_storage address;
1657	int len, err, fput_needed;
1658
1659	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1660	if (!sock)
1661	goto out;
1662
1663	err = security_socket_getsockname(sock);
1664	if (err)
1665	goto out_put;
1666
1667	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1668	if (err)
1669	goto out_put;
1670	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1671
1672	out_put:
1673	fput_light(sock->file, fput_needed);
1674	out:
1675	return err;
1676	}
1677
1678	/*
1679	* Get the remote address ('name') of a socket object. Move the obtained
1680	* name to user space.
1681	*/
1682
1683	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1684	int __user *, usockaddr_len)
1685	{
1686	struct socket *sock;
1687	struct sockaddr_storage address;
1688	int len, err, fput_needed;
1689
1690	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1691	if (sock != NULL) {
1692	err = security_socket_getpeername(sock);
1693	if (err) {
1694	fput_light(sock->file, fput_needed);
1695	return err;
1696	}
1697
1698	err =
1699	sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1700	1);
1701	if (!err)
1702	err = move_addr_to_user(&address, len, usockaddr,
1703	usockaddr_len);
1704	fput_light(sock->file, fput_needed);
1705	}
1706	return err;
1707	}
1708
1709	/*
1710	* Send a datagram to a given address. We move the address into kernel
1711	* space and check the user space data area is readable before invoking
1712	* the protocol.
1713	*/
1714
1715	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1716	unsigned int, flags, struct sockaddr __user *, addr,
1717	int, addr_len)
1718	{
1719	struct socket *sock;
1720	struct sockaddr_storage address;
1721	int err;
1722	struct msghdr msg;
1723	struct iovec iov;
1724	int fput_needed;
1725
1726	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1727	if (unlikely(err))
1728	return err;
1729	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1730	if (!sock)
1731	goto out;
1732
1733	msg.msg_name = NULL;
1734	msg.msg_control = NULL;
1735	msg.msg_controllen = 0;
1736	msg.msg_namelen = 0;
1737	if (addr) {
1738	err = move_addr_to_kernel(addr, addr_len, &address);
1739	if (err < 0)
1740	goto out_put;
1741	msg.msg_name = (struct sockaddr *)&address;
1742	msg.msg_namelen = addr_len;
1743	}
1744	if (sock->file->f_flags & O_NONBLOCK)
1745	flags \|= MSG_DONTWAIT;
1746	msg.msg_flags = flags;
1747	err = sock_sendmsg(sock, &msg);
1748
1749	out_put:
1750	fput_light(sock->file, fput_needed);
1751	out:
1752	return err;
1753	}
1754
1755	/*
1756	* Send a datagram down a socket.
1757	*/
1758
1759	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1760	unsigned int, flags)
1761	{
1762	return sys_sendto(fd, buff, len, flags, NULL, 0);
1763	}
1764
1765	/*
1766	* Receive a frame from the socket and optionally record the address of the
1767	* sender. We verify the buffers are writable and if needed move the
1768	* sender address from kernel to user space.
1769	*/
1770
1771	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1772	unsigned int, flags, struct sockaddr __user *, addr,
1773	int __user *, addr_len)
1774	{
1775	struct socket *sock;
1776	struct iovec iov;
1777	struct msghdr msg;
1778	struct sockaddr_storage address;
1779	int err, err2;
1780	int fput_needed;
1781
1782	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1783	if (unlikely(err))
1784	return err;
1785	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1786	if (!sock)
1787	goto out;
1788
1789	msg.msg_control = NULL;
1790	msg.msg_controllen = 0;
1791	/* Save some cycles and don't copy the address if not needed */
1792	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1793	/* We assume all kernel code knows the size of sockaddr_storage */
1794	msg.msg_namelen = 0;
1795	msg.msg_iocb = NULL;
1796	msg.msg_flags = 0;
1797	if (sock->file->f_flags & O_NONBLOCK)
1798	flags \|= MSG_DONTWAIT;
1799	err = sock_recvmsg(sock, &msg, flags);
1800
1801	if (err >= 0 && addr != NULL) {
1802	err2 = move_addr_to_user(&address,
1803	msg.msg_namelen, addr, addr_len);
1804	if (err2 < 0)
1805	err = err2;
1806	}
1807
1808	fput_light(sock->file, fput_needed);
1809	out:
1810	return err;
1811	}
1812
1813	/*
1814	* Receive a datagram from a socket.
1815	*/
1816
1817	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1818	unsigned int, flags)
1819	{
1820	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1821	}
1822
1823	/*
1824	* Set a socket option. Because we don't know the option lengths we have
1825	* to pass the user mode parameter for the protocols to sort out.
1826	*/
1827
1828	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1829	char __user *, optval, int, optlen)
1830	{
1831	int err, fput_needed;
1832	struct socket *sock;
1833
1834	if (optlen < 0)
1835	return -EINVAL;
1836
1837	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1838	if (sock != NULL) {
1839	err = security_socket_setsockopt(sock, level, optname);
1840	if (err)
1841	goto out_put;
1842
1843	if (level == SOL_SOCKET)
1844	err =
1845	sock_setsockopt(sock, level, optname, optval,
1846	optlen);
1847	else
1848	err =
1849	sock->ops->setsockopt(sock, level, optname, optval,
1850	optlen);
1851	out_put:
1852	fput_light(sock->file, fput_needed);
1853	}
1854	return err;
1855	}
1856
1857	/*
1858	* Get a socket option. Because we don't know the option lengths we have
1859	* to pass a user mode parameter for the protocols to sort out.
1860	*/
1861
1862	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1863	char __user , optval, int __user , optlen)
1864	{
1865	int err, fput_needed;
1866	struct socket *sock;
1867
1868	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1869	if (sock != NULL) {
1870	err = security_socket_getsockopt(sock, level, optname);
1871	if (err)
1872	goto out_put;
1873
1874	if (level == SOL_SOCKET)
1875	err =
1876	sock_getsockopt(sock, level, optname, optval,
1877	optlen);
1878	else
1879	err =
1880	sock->ops->getsockopt(sock, level, optname, optval,
1881	optlen);
1882	out_put:
1883	fput_light(sock->file, fput_needed);
1884	}
1885	return err;
1886	}
1887
1888	/*
1889	* Shutdown a socket.
1890	*/
1891
1892	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1893	{
1894	int err, fput_needed;
1895	struct socket *sock;
1896
1897	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1898	if (sock != NULL) {
1899	err = security_socket_shutdown(sock, how);
1900	if (!err)
1901	err = sock->ops->shutdown(sock, how);
1902	fput_light(sock->file, fput_needed);
1903	}
1904	return err;
1905	}
1906
1907	/* A couple of helpful macros for getting the address of the 32/64 bit
1908	* fields which are the same type (int / unsigned) on our platforms.
1909	*/
1910	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1911	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1912	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1913
1914	struct used_address {
1915	struct sockaddr_storage name;
1916	unsigned int name_len;
1917	};
1918
1919	static int copy_msghdr_from_user(struct msghdr *kmsg,
1920	struct user_msghdr __user *umsg,
1921	struct sockaddr __user **save_addr,
1922	struct iovec **iov)
1923	{
1924	struct user_msghdr msg;
1925	ssize_t err;
1926
1927	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1928	return -EFAULT;
1929
1930	kmsg->msg_control = (void __force *)msg.msg_control;
1931	kmsg->msg_controllen = msg.msg_controllen;
1932	kmsg->msg_flags = msg.msg_flags;
1933
1934	kmsg->msg_namelen = msg.msg_namelen;
1935	if (!msg.msg_name)
1936	kmsg->msg_namelen = 0;
1937
1938	if (kmsg->msg_namelen < 0)
1939	return -EINVAL;
1940
1941	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1942	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1943
1944	if (save_addr)
1945	*save_addr = msg.msg_name;
1946
1947	if (msg.msg_name && kmsg->msg_namelen) {
1948	if (!save_addr) {
1949	err = move_addr_to_kernel(msg.msg_name,
1950	kmsg->msg_namelen,
1951	kmsg->msg_name);
1952	if (err < 0)
1953	return err;
1954	}
1955	} else {
1956	kmsg->msg_name = NULL;
1957	kmsg->msg_namelen = 0;
1958	}
1959
1960	if (msg.msg_iovlen > UIO_MAXIOV)
1961	return -EMSGSIZE;
1962
1963	kmsg->msg_iocb = NULL;
1964
1965	return import_iovec(save_addr ? READ : WRITE,
1966	msg.msg_iov, msg.msg_iovlen,
1967	UIO_FASTIOV, iov, &kmsg->msg_iter);
1968	}
1969
1970	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
1971	struct msghdr *msg_sys, unsigned int flags,
1972	struct used_address *used_address,
1973	unsigned int allowed_msghdr_flags)
1974	{
1975	struct compat_msghdr __user *msg_compat =
1976	(struct compat_msghdr __user *)msg;
1977	struct sockaddr_storage address;
1978	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1979	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1980	__aligned(sizeof(__kernel_size_t));
1981	/* 20 is size of ipv6_pktinfo */
1982	unsigned char *ctl_buf = ctl;
1983	int ctl_len;
1984	ssize_t err;
1985
1986	msg_sys->msg_name = &address;
1987
1988	if (MSG_CMSG_COMPAT & flags)
1989	err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1990	else
1991	err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1992	if (err < 0)
1993	return err;
1994
1995	err = -ENOBUFS;
1996
1997	if (msg_sys->msg_controllen > INT_MAX)
1998	goto out_freeiov;
1999	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2000	ctl_len = msg_sys->msg_controllen;
2001	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2002	err =
2003	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2004	sizeof(ctl));
2005	if (err)
2006	goto out_freeiov;
2007	ctl_buf = msg_sys->msg_control;
2008	ctl_len = msg_sys->msg_controllen;
2009	} else if (ctl_len) {
2010	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2011	CMSG_ALIGN(sizeof(struct cmsghdr)));
2012	if (ctl_len > sizeof(ctl)) {
2013	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2014	if (ctl_buf == NULL)
2015	goto out_freeiov;
2016	}
2017	err = -EFAULT;
2018	/*
2019	* Careful! Before this, msg_sys->msg_control contains a user pointer.
2020	* Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2021	* checking falls down on this.
2022	*/
2023	if (copy_from_user(ctl_buf,
2024	(void __user __force *)msg_sys->msg_control,
2025	ctl_len))
2026	goto out_freectl;
2027	msg_sys->msg_control = ctl_buf;
2028	}
2029	msg_sys->msg_flags = flags;
2030
2031	if (sock->file->f_flags & O_NONBLOCK)
2032	msg_sys->msg_flags \|= MSG_DONTWAIT;
2033	/*
2034	* If this is sendmmsg() and current destination address is same as
2035	* previously succeeded address, omit asking LSM's decision.
2036	* used_address->name_len is initialized to UINT_MAX so that the first
2037	* destination address never matches.
2038	*/
2039	if (used_address && msg_sys->msg_name &&
2040	used_address->name_len == msg_sys->msg_namelen &&
2041	!memcmp(&used_address->name, msg_sys->msg_name,
2042	used_address->name_len)) {
2043	err = sock_sendmsg_nosec(sock, msg_sys);
2044	goto out_freectl;
2045	}
2046	err = sock_sendmsg(sock, msg_sys);
2047	/*
2048	* If this is sendmmsg() and sending to current destination address was
2049	* successful, remember it.
2050	*/
2051	if (used_address && err >= 0) {
2052	used_address->name_len = msg_sys->msg_namelen;
2053	if (msg_sys->msg_name)
2054	memcpy(&used_address->name, msg_sys->msg_name,
2055	used_address->name_len);
2056	}
2057
2058	out_freectl:
2059	if (ctl_buf != ctl)
2060	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2061	out_freeiov:
2062	kfree(iov);
2063	return err;
2064	}
2065
2066	/*
2067	* BSD sendmsg interface
2068	*/
2069
2070	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2071	{
2072	int fput_needed, err;
2073	struct msghdr msg_sys;
2074	struct socket *sock;
2075
2076	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2077	if (!sock)
2078	goto out;
2079
2080	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2081
2082	fput_light(sock->file, fput_needed);
2083	out:
2084	return err;
2085	}
2086
2087	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2088	{
2089	if (flags & MSG_CMSG_COMPAT)
2090	return -EINVAL;
2091	return __sys_sendmsg(fd, msg, flags);
2092	}
2093
2094	/*
2095	* Linux sendmmsg interface
2096	*/
2097
2098	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2099	unsigned int flags)
2100	{
2101	int fput_needed, err, datagrams;
2102	struct socket *sock;
2103	struct mmsghdr __user *entry;
2104	struct compat_mmsghdr __user *compat_entry;
2105	struct msghdr msg_sys;
2106	struct used_address used_address;
2107	unsigned int oflags = flags;
2108
2109	if (vlen > UIO_MAXIOV)
2110	vlen = UIO_MAXIOV;
2111
2112	datagrams = 0;
2113
2114	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2115	if (!sock)
2116	return err;
2117
2118	used_address.name_len = UINT_MAX;
2119	entry = mmsg;
2120	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2121	err = 0;
2122	flags \|= MSG_BATCH;
2123
2124	while (datagrams < vlen) {
2125	if (datagrams == vlen - 1)
2126	flags = oflags;
2127
2128	if (MSG_CMSG_COMPAT & flags) {
2129	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2130	&msg_sys, flags, &used_address, MSG_EOR);
2131	if (err < 0)
2132	break;
2133	err = __put_user(err, &compat_entry->msg_len);
2134	++compat_entry;
2135	} else {
2136	err = ___sys_sendmsg(sock,
2137	(struct user_msghdr __user *)entry,
2138	&msg_sys, flags, &used_address, MSG_EOR);
2139	if (err < 0)
2140	break;
2141	err = put_user(err, &entry->msg_len);
2142	++entry;
2143	}
2144
2145	if (err)
2146	break;
2147	++datagrams;
2148	if (msg_data_left(&msg_sys))
2149	break;
2150	cond_resched();
2151	}
2152
2153	fput_light(sock->file, fput_needed);
2154
2155	/* We only return an error if no datagrams were able to be sent */
2156	if (datagrams != 0)
2157	return datagrams;
2158
2159	return err;
2160	}
2161
2162	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2163	unsigned int, vlen, unsigned int, flags)
2164	{
2165	if (flags & MSG_CMSG_COMPAT)
2166	return -EINVAL;
2167	return __sys_sendmmsg(fd, mmsg, vlen, flags);
2168	}
2169
2170	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2171	struct msghdr *msg_sys, unsigned int flags, int nosec)
2172	{
2173	struct compat_msghdr __user *msg_compat =
2174	(struct compat_msghdr __user *)msg;
2175	struct iovec iovstack[UIO_FASTIOV];
2176	struct iovec *iov = iovstack;
2177	unsigned long cmsg_ptr;
2178	int len;
2179	ssize_t err;
2180
2181	/* kernel mode address */
2182	struct sockaddr_storage addr;
2183
2184	/* user mode address pointers */
2185	struct sockaddr __user *uaddr;
2186	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2187
2188	msg_sys->msg_name = &addr;
2189
2190	if (MSG_CMSG_COMPAT & flags)
2191	err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2192	else
2193	err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2194	if (err < 0)
2195	return err;
2196
2197	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2198	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2199
2200	/* We assume all kernel code knows the size of sockaddr_storage */
2201	msg_sys->msg_namelen = 0;
2202
2203	if (sock->file->f_flags & O_NONBLOCK)
2204	flags \|= MSG_DONTWAIT;
2205	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2206	if (err < 0)
2207	goto out_freeiov;
2208	len = err;
2209
2210	if (uaddr != NULL) {
2211	err = move_addr_to_user(&addr,
2212	msg_sys->msg_namelen, uaddr,
2213	uaddr_len);
2214	if (err < 0)
2215	goto out_freeiov;
2216	}
2217	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2218	COMPAT_FLAGS(msg));
2219	if (err)
2220	goto out_freeiov;
2221	if (MSG_CMSG_COMPAT & flags)
2222	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2223	&msg_compat->msg_controllen);
2224	else
2225	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2226	&msg->msg_controllen);
2227	if (err)
2228	goto out_freeiov;
2229	err = len;
2230
2231	out_freeiov:
2232	kfree(iov);
2233	return err;
2234	}
2235
2236	/*
2237	* BSD recvmsg interface
2238	*/
2239
2240	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2241	{
2242	int fput_needed, err;
2243	struct msghdr msg_sys;
2244	struct socket *sock;
2245
2246	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2247	if (!sock)
2248	goto out;
2249
2250	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2251
2252	fput_light(sock->file, fput_needed);
2253	out:
2254	return err;
2255	}
2256
2257	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2258	unsigned int, flags)
2259	{
2260	if (flags & MSG_CMSG_COMPAT)
2261	return -EINVAL;
2262	return __sys_recvmsg(fd, msg, flags);
2263	}
2264
2265	/*
2266	* Linux recvmmsg interface
2267	*/
2268
2269	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2270	unsigned int flags, struct timespec *timeout)
2271	{
2272	int fput_needed, err, datagrams;
2273	struct socket *sock;
2274	struct mmsghdr __user *entry;
2275	struct compat_mmsghdr __user *compat_entry;
2276	struct msghdr msg_sys;
2277	struct timespec64 end_time;
2278	struct timespec64 timeout64;
2279
2280	if (timeout &&
2281	poll_select_set_timeout(&end_time, timeout->tv_sec,
2282	timeout->tv_nsec))
2283	return -EINVAL;
2284
2285	datagrams = 0;
2286
2287	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2288	if (!sock)
2289	return err;
2290
2291	err = sock_error(sock->sk);
2292	if (err) {
2293	datagrams = err;
2294	goto out_put;
2295	}
2296
2297	entry = mmsg;
2298	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2299
2300	while (datagrams < vlen) {
2301	/*
2302	* No need to ask LSM for more than the first datagram.
2303	*/
2304	if (MSG_CMSG_COMPAT & flags) {
2305	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2306	&msg_sys, flags & ~MSG_WAITFORONE,
2307	datagrams);
2308	if (err < 0)
2309	break;
2310	err = __put_user(err, &compat_entry->msg_len);
2311	++compat_entry;
2312	} else {
2313	err = ___sys_recvmsg(sock,
2314	(struct user_msghdr __user *)entry,
2315	&msg_sys, flags & ~MSG_WAITFORONE,
2316	datagrams);
2317	if (err < 0)
2318	break;
2319	err = put_user(err, &entry->msg_len);
2320	++entry;
2321	}
2322
2323	if (err)
2324	break;
2325	++datagrams;
2326
2327	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2328	if (flags & MSG_WAITFORONE)
2329	flags \|= MSG_DONTWAIT;
2330
2331	if (timeout) {
2332	ktime_get_ts64(&timeout64);
2333	*timeout = timespec64_to_timespec(
2334	timespec64_sub(end_time, timeout64));
2335	if (timeout->tv_sec < 0) {
2336	timeout->tv_sec = timeout->tv_nsec = 0;
2337	break;
2338	}
2339
2340	/* Timeout, return less than vlen datagrams */
2341	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2342	break;
2343	}
2344
2345	/* Out of band data, return right away */
2346	if (msg_sys.msg_flags & MSG_OOB)
2347	break;
2348	cond_resched();
2349	}
2350
2351	if (err == 0)
2352	goto out_put;
2353
2354	if (datagrams == 0) {
2355	datagrams = err;
2356	goto out_put;
2357	}
2358
2359	/*
2360	* We may return less entries than requested (vlen) if the
2361	* sock is non block and there aren't enough datagrams...
2362	*/
2363	if (err != -EAGAIN) {
2364	/*
2365	* ... or if recvmsg returns an error after we
2366	* received some datagrams, where we record the
2367	* error to return on the next call or if the
2368	* app asks about it using getsockopt(SO_ERROR).
2369	*/
2370	sock->sk->sk_err = -err;
2371	}
2372	out_put:
2373	fput_light(sock->file, fput_needed);
2374
2375	return datagrams;
2376	}
2377
2378	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2379	unsigned int, vlen, unsigned int, flags,
2380	struct timespec __user *, timeout)
2381	{
2382	int datagrams;
2383	struct timespec timeout_sys;
2384
2385	if (flags & MSG_CMSG_COMPAT)
2386	return -EINVAL;
2387
2388	if (!timeout)
2389	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2390
2391	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2392	return -EFAULT;
2393
2394	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2395
2396	if (datagrams > 0 &&
2397	copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2398	datagrams = -EFAULT;
2399
2400	return datagrams;
2401	}
2402
2403	#ifdef __ARCH_WANT_SYS_SOCKETCALL
2404	/* Argument list sizes for sys_socketcall */
2405	#define AL(x) ((x) * sizeof(unsigned long))
2406	static const unsigned char nargs[21] = {
2407	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2408	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2409	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2410	AL(4), AL(5), AL(4)
2411	};
2412
2413	#undef AL
2414
2415	/*
2416	* System call vectors.
2417	*
2418	* Argument checking cleaned up. Saved 20% in size.
2419	* This function doesn't need to set the kernel lock because
2420	* it is set by the callees.
2421	*/
2422
2423	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2424	{
2425	unsigned long a[AUDITSC_ARGS];
2426	unsigned long a0, a1;
2427	int err;
2428	unsigned int len;
2429
2430	if (call < 1 \|\| call > SYS_SENDMMSG)
2431	return -EINVAL;
2432
2433	len = nargs[call];
2434	if (len > sizeof(a))
2435	return -EINVAL;
2436
2437	/* copy_from_user should be SMP safe. */
2438	if (copy_from_user(a, args, len))
2439	return -EFAULT;
2440
2441	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2442	if (err)
2443	return err;
2444
2445	a0 = a[0];
2446	a1 = a[1];
2447
2448	switch (call) {
2449	case SYS_SOCKET:
2450	err = sys_socket(a0, a1, a[2]);
2451	break;
2452	case SYS_BIND:
2453	err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2454	break;
2455	case SYS_CONNECT:
2456	err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2457	break;
2458	case SYS_LISTEN:
2459	err = sys_listen(a0, a1);
2460	break;
2461	case SYS_ACCEPT:
2462	err = sys_accept4(a0, (struct sockaddr __user *)a1,
2463	(int __user *)a[2], 0);
2464	break;
2465	case SYS_GETSOCKNAME:
2466	err =
2467	sys_getsockname(a0, (struct sockaddr __user *)a1,
2468	(int __user *)a[2]);
2469	break;
2470	case SYS_GETPEERNAME:
2471	err =
2472	sys_getpeername(a0, (struct sockaddr __user *)a1,
2473	(int __user *)a[2]);
2474	break;
2475	case SYS_SOCKETPAIR:
2476	err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2477	break;
2478	case SYS_SEND:
2479	err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2480	break;
2481	case SYS_SENDTO:
2482	err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2483	(struct sockaddr __user *)a[4], a[5]);
2484	break;
2485	case SYS_RECV:
2486	err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2487	break;
2488	case SYS_RECVFROM:
2489	err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2490	(struct sockaddr __user *)a[4],
2491	(int __user *)a[5]);
2492	break;
2493	case SYS_SHUTDOWN:
2494	err = sys_shutdown(a0, a1);
2495	break;
2496	case SYS_SETSOCKOPT:
2497	err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2498	break;
2499	case SYS_GETSOCKOPT:
2500	err =
2501	sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2502	(int __user *)a[4]);
2503	break;
2504	case SYS_SENDMSG:
2505	err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2506	break;
2507	case SYS_SENDMMSG:
2508	err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2509	break;
2510	case SYS_RECVMSG:
2511	err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2512	break;
2513	case SYS_RECVMMSG:
2514	err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2515	(struct timespec __user *)a[4]);
2516	break;
2517	case SYS_ACCEPT4:
2518	err = sys_accept4(a0, (struct sockaddr __user *)a1,
2519	(int __user *)a[2], a[3]);
2520	break;
2521	default:
2522	err = -EINVAL;
2523	break;
2524	}
2525	return err;
2526	}
2527
2528	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2529
2530	/**
2531	* sock_register - add a socket protocol handler
2532	* @ops: description of protocol
2533	*
2534	* This function is called by a protocol handler that wants to
2535	* advertise its address family, and have it linked into the
2536	* socket interface. The value ops->family corresponds to the
2537	* socket system call protocol family.
2538	*/
2539	int sock_register(const struct net_proto_family *ops)
2540	{
2541	int err;
2542
2543	if (ops->family >= NPROTO) {
2544	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2545	return -ENOBUFS;
2546	}
2547
2548	spin_lock(&net_family_lock);
2549	if (rcu_dereference_protected(net_families[ops->family],
2550	lockdep_is_held(&net_family_lock)))
2551	err = -EEXIST;
2552	else {
2553	rcu_assign_pointer(net_families[ops->family], ops);
2554	err = 0;
2555	}
2556	spin_unlock(&net_family_lock);
2557
2558	pr_info("NET: Registered protocol family %d\n", ops->family);
2559	return err;
2560	}
2561	EXPORT_SYMBOL(sock_register);
2562
2563	/**
2564	* sock_unregister - remove a protocol handler
2565	* @family: protocol family to remove
2566	*
2567	* This function is called by a protocol handler that wants to
2568	* remove its address family, and have it unlinked from the
2569	* new socket creation.
2570	*
2571	* If protocol handler is a module, then it can use module reference
2572	* counts to protect against new references. If protocol handler is not
2573	* a module then it needs to provide its own protection in
2574	* the ops->create routine.
2575	*/
2576	void sock_unregister(int family)
2577	{
2578	BUG_ON(family < 0 \|\| family >= NPROTO);
2579
2580	spin_lock(&net_family_lock);
2581	RCU_INIT_POINTER(net_families[family], NULL);
2582	spin_unlock(&net_family_lock);
2583
2584	synchronize_rcu();
2585
2586	pr_info("NET: Unregistered protocol family %d\n", family);
2587	}
2588	EXPORT_SYMBOL(sock_unregister);
2589
2590	static int __init sock_init(void)
2591	{
2592	int err;
2593	/*
2594	* Initialize the network sysctl infrastructure.
2595	*/
2596	err = net_sysctl_init();
2597	if (err)
2598	goto out;
2599
2600	/*
2601	* Initialize skbuff SLAB cache
2602	*/
2603	skb_init();
2604
2605	/*
2606	* Initialize the protocols module.
2607	*/
2608
2609	init_inodecache();
2610
2611	err = register_filesystem(&sock_fs_type);
2612	if (err)
2613	goto out_fs;
2614	sock_mnt = kern_mount(&sock_fs_type);
2615	if (IS_ERR(sock_mnt)) {
2616	err = PTR_ERR(sock_mnt);
2617	goto out_mount;
2618	}
2619
2620	/* The real protocol initialization is performed in later initcalls.
2621	*/
2622
2623	#ifdef CONFIG_NETFILTER
2624	err = netfilter_init();
2625	if (err)
2626	goto out;
2627	#endif
2628
2629	ptp_classifier_init();
2630
2631	out:
2632	return err;
2633
2634	out_mount:
2635	unregister_filesystem(&sock_fs_type);
2636	out_fs:
2637	goto out;
2638	}
2639
2640	core_initcall(sock_init); /* early initcall */
2641
2642	#ifdef CONFIG_PROC_FS
2643	void socket_seq_show(struct seq_file *seq)
2644	{
2645	seq_printf(seq, "sockets: used %d\n",
2646	sock_inuse_get(seq->private));
2647	}
2648	#endif /* CONFIG_PROC_FS */
2649
2650	#ifdef CONFIG_COMPAT
2651	static int do_siocgstamp(struct net net, struct socket sock,
2652	unsigned int cmd, void __user *up)
2653	{
2654	mm_segment_t old_fs = get_fs();
2655	struct timeval ktv;
2656	int err;
2657
2658	set_fs(KERNEL_DS);
2659	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2660	set_fs(old_fs);
2661	if (!err)
2662	err = compat_put_timeval(&ktv, up);
2663
2664	return err;
2665	}
2666
2667	static int do_siocgstampns(struct net net, struct socket sock,
2668	unsigned int cmd, void __user *up)
2669	{
2670	mm_segment_t old_fs = get_fs();
2671	struct timespec kts;
2672	int err;
2673
2674	set_fs(KERNEL_DS);
2675	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2676	set_fs(old_fs);
2677	if (!err)
2678	err = compat_put_timespec(&kts, up);
2679
2680	return err;
2681	}
2682
2683	static int compat_dev_ifconf(struct net net, struct compat_ifconf __user uifc32)
2684	{
2685	struct compat_ifconf ifc32;
2686	struct ifconf ifc;
2687	int err;
2688
2689	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2690	return -EFAULT;
2691
2692	ifc.ifc_len = ifc32.ifc_len;
2693	ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2694
2695	rtnl_lock();
2696	err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2697	rtnl_unlock();
2698	if (err)
2699	return err;
2700
2701	ifc32.ifc_len = ifc.ifc_len;
2702	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2703	return -EFAULT;
2704
2705	return 0;
2706	}
2707
2708	static int ethtool_ioctl(struct net net, struct compat_ifreq __user ifr32)
2709	{
2710	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2711	bool convert_in = false, convert_out = false;
2712	size_t buf_size = 0;
2713	struct ethtool_rxnfc __user *rxnfc = NULL;
2714	struct ifreq ifr;
2715	u32 rule_cnt = 0, actual_rule_cnt;
2716	u32 ethcmd;
2717	u32 data;
2718	int ret;
2719
2720	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2721	return -EFAULT;
2722
2723	compat_rxnfc = compat_ptr(data);
2724
2725	if (get_user(ethcmd, &compat_rxnfc->cmd))
2726	return -EFAULT;
2727
2728	/* Most ethtool structures are defined without padding.
2729	* Unfortunately struct ethtool_rxnfc is an exception.
2730	*/
2731	switch (ethcmd) {
2732	default:
2733	break;
2734	case ETHTOOL_GRXCLSRLALL:
2735	/* Buffer size is variable */
2736	if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2737	return -EFAULT;
2738	if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2739	return -ENOMEM;
2740	buf_size += rule_cnt * sizeof(u32);
2741	/* fall through */
2742	case ETHTOOL_GRXRINGS:
2743	case ETHTOOL_GRXCLSRLCNT:
2744	case ETHTOOL_GRXCLSRULE:
2745	case ETHTOOL_SRXCLSRLINS:
2746	convert_out = true;
2747	/* fall through */
2748	case ETHTOOL_SRXCLSRLDEL:
2749	buf_size += sizeof(struct ethtool_rxnfc);
2750	convert_in = true;
2751	rxnfc = compat_alloc_user_space(buf_size);
2752	break;
2753	}
2754
2755	if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2756	return -EFAULT;
2757
2758	ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2759
2760	if (convert_in) {
2761	/* We expect there to be holes between fs.m_ext and
2762	* fs.ring_cookie and at the end of fs, but nowhere else.
2763	*/
2764	BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2765	sizeof(compat_rxnfc->fs.m_ext) !=
2766	offsetof(struct ethtool_rxnfc, fs.m_ext) +
2767	sizeof(rxnfc->fs.m_ext));
2768	BUILD_BUG_ON(
2769	offsetof(struct compat_ethtool_rxnfc, fs.location) -
2770	offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2771	offsetof(struct ethtool_rxnfc, fs.location) -
2772	offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2773
2774	if (copy_in_user(rxnfc, compat_rxnfc,
2775	(void __user *)(&rxnfc->fs.m_ext + 1) -
2776	(void __user *)rxnfc) \|\|
2777	copy_in_user(&rxnfc->fs.ring_cookie,
2778	&compat_rxnfc->fs.ring_cookie,
2779	(void __user *)(&rxnfc->fs.location + 1) -
2780	(void __user *)&rxnfc->fs.ring_cookie) \|\|
2781	copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2782	sizeof(rxnfc->rule_cnt)))
2783	return -EFAULT;
2784	}
2785
2786	ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2787	if (ret)
2788	return ret;
2789
2790	if (convert_out) {
2791	if (copy_in_user(compat_rxnfc, rxnfc,
2792	(const void __user *)(&rxnfc->fs.m_ext + 1) -
2793	(const void __user *)rxnfc) \|\|
2794	copy_in_user(&compat_rxnfc->fs.ring_cookie,
2795	&rxnfc->fs.ring_cookie,
2796	(const void __user *)(&rxnfc->fs.location + 1) -
2797	(const void __user *)&rxnfc->fs.ring_cookie) \|\|
2798	copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2799	sizeof(rxnfc->rule_cnt)))
2800	return -EFAULT;
2801
2802	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2803	/* As an optimisation, we only copy the actual
2804	* number of rules that the underlying
2805	* function returned. Since Mallory might
2806	* change the rule count in user memory, we
2807	* check that it is less than the rule count
2808	* originally given (as the user buffer size),
2809	* which has been range-checked.
2810	*/
2811	if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2812	return -EFAULT;
2813	if (actual_rule_cnt < rule_cnt)
2814	rule_cnt = actual_rule_cnt;
2815	if (copy_in_user(&compat_rxnfc->rule_locs[0],
2816	&rxnfc->rule_locs[0],
2817	rule_cnt * sizeof(u32)))
2818	return -EFAULT;
2819	}
2820	}
2821
2822	return 0;
2823	}
2824
2825	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
2826	{
2827	compat_uptr_t uptr32;
2828	struct ifreq ifr;
2829	void __user *saved;
2830	int err;
2831
2832	if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2833	return -EFAULT;
2834
2835	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2836	return -EFAULT;
2837
2838	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2839	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2840
2841	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2842	if (!err) {
2843	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2844	if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2845	err = -EFAULT;
2846	}
2847	return err;
2848	}
2849
2850	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2851	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2852	struct compat_ifreq __user *u_ifreq32)
2853	{
2854	struct ifreq ifreq;
2855	u32 data32;
2856
2857	if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2858	return -EFAULT;
2859	if (get_user(data32, &u_ifreq32->ifr_data))
2860	return -EFAULT;
2861	ifreq.ifr_data = compat_ptr(data32);
2862
2863	return dev_ioctl(net, cmd, &ifreq, NULL);
2864	}
2865
2866	static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2867	struct compat_ifreq __user *uifr32)
2868	{
2869	struct ifreq ifr;
2870	struct compat_ifmap __user *uifmap32;
2871	int err;
2872
2873	uifmap32 = &uifr32->ifr_ifru.ifru_map;
2874	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2875	err \|= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2876	err \|= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2877	err \|= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2878	err \|= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2879	err \|= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2880	err \|= get_user(ifr.ifr_map.port, &uifmap32->port);
2881	if (err)
2882	return -EFAULT;
2883
2884	err = dev_ioctl(net, cmd, &ifr, NULL);
2885
2886	if (cmd == SIOCGIFMAP && !err) {
2887	err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2888	err \|= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2889	err \|= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2890	err \|= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2891	err \|= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2892	err \|= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2893	err \|= put_user(ifr.ifr_map.port, &uifmap32->port);
2894	if (err)
2895	err = -EFAULT;
2896	}
2897	return err;
2898	}
2899
2900	struct rtentry32 {
2901	u32 rt_pad1;
2902	struct sockaddr rt_dst; /* target address */
2903	struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2904	struct sockaddr rt_genmask; /* target network mask (IP) */
2905	unsigned short rt_flags;
2906	short rt_pad2;
2907	u32 rt_pad3;
2908	unsigned char rt_tos;
2909	unsigned char rt_class;
2910	short rt_pad4;
2911	short rt_metric; /* +1 for binary compatibility! */
2912	/* char * / u32 rt_dev; / forcing the device at add */
2913	u32 rt_mtu; /* per route MTU/Window */
2914	u32 rt_window; /* Window clamping */
2915	unsigned short rt_irtt; /* Initial RTT */
2916	};
2917
2918	struct in6_rtmsg32 {
2919	struct in6_addr rtmsg_dst;
2920	struct in6_addr rtmsg_src;
2921	struct in6_addr rtmsg_gateway;
2922	u32 rtmsg_type;
2923	u16 rtmsg_dst_len;
2924	u16 rtmsg_src_len;
2925	u32 rtmsg_metric;
2926	u32 rtmsg_info;
2927	u32 rtmsg_flags;
2928	s32 rtmsg_ifindex;
2929	};
2930
2931	static int routing_ioctl(struct net net, struct socket sock,
2932	unsigned int cmd, void __user *argp)
2933	{
2934	int ret;
2935	void *r = NULL;
2936	struct in6_rtmsg r6;
2937	struct rtentry r4;
2938	char devname[16];
2939	u32 rtdev;
2940	mm_segment_t old_fs = get_fs();
2941
2942	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2943	struct in6_rtmsg32 __user *ur6 = argp;
2944	ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2945	3 * sizeof(struct in6_addr));
2946	ret \|= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2947	ret \|= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2948	ret \|= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2949	ret \|= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2950	ret \|= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2951	ret \|= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2952	ret \|= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2953
2954	r = (void *) &r6;
2955	} else { /* ipv4 */
2956	struct rtentry32 __user *ur4 = argp;
2957	ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
2958	3 * sizeof(struct sockaddr));
2959	ret \|= get_user(r4.rt_flags, &(ur4->rt_flags));
2960	ret \|= get_user(r4.rt_metric, &(ur4->rt_metric));
2961	ret \|= get_user(r4.rt_mtu, &(ur4->rt_mtu));
2962	ret \|= get_user(r4.rt_window, &(ur4->rt_window));
2963	ret \|= get_user(r4.rt_irtt, &(ur4->rt_irtt));
2964	ret \|= get_user(rtdev, &(ur4->rt_dev));
2965	if (rtdev) {
2966	ret \|= copy_from_user(devname, compat_ptr(rtdev), 15);
2967	r4.rt_dev = (char __user __force *)devname;
2968	devname[15] = 0;
2969	} else
2970	r4.rt_dev = NULL;
2971
2972	r = (void *) &r4;
2973	}
2974
2975	if (ret) {
2976	ret = -EFAULT;
2977	goto out;
2978	}
2979
2980	set_fs(KERNEL_DS);
2981	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
2982	set_fs(old_fs);
2983
2984	out:
2985	return ret;
2986	}
2987
2988	/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2989	* for some operations; this forces use of the newer bridge-utils that
2990	* use compatible ioctls
2991	*/
2992	static int old_bridge_ioctl(compat_ulong_t __user *argp)
2993	{
2994	compat_ulong_t tmp;
2995
2996	if (get_user(tmp, argp))
2997	return -EFAULT;
2998	if (tmp == BRCTL_GET_VERSION)
2999	return BRCTL_VERSION + 1;
3000	return -EINVAL;
3001	}
3002
3003	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3004	unsigned int cmd, unsigned long arg)
3005	{
3006	void __user *argp = compat_ptr(arg);
3007	struct sock *sk = sock->sk;
3008	struct net *net = sock_net(sk);
3009
3010	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3011	return compat_ifr_data_ioctl(net, cmd, argp);
3012
3013	switch (cmd) {
3014	case SIOCSIFBR:
3015	case SIOCGIFBR:
3016	return old_bridge_ioctl(argp);
3017	case SIOCGIFCONF:
3018	return compat_dev_ifconf(net, argp);
3019	case SIOCETHTOOL:
3020	return ethtool_ioctl(net, argp);
3021	case SIOCWANDEV:
3022	return compat_siocwandev(net, argp);
3023	case SIOCGIFMAP:
3024	case SIOCSIFMAP:
3025	return compat_sioc_ifmap(net, cmd, argp);
3026	case SIOCADDRT:
3027	case SIOCDELRT:
3028	return routing_ioctl(net, sock, cmd, argp);
3029	case SIOCGSTAMP:
3030	return do_siocgstamp(net, sock, cmd, argp);
3031	case SIOCGSTAMPNS:
3032	return do_siocgstampns(net, sock, cmd, argp);
3033	case SIOCBONDSLAVEINFOQUERY:
3034	case SIOCBONDINFOQUERY:
3035	case SIOCSHWTSTAMP:
3036	case SIOCGHWTSTAMP:
3037	return compat_ifr_data_ioctl(net, cmd, argp);
3038
3039	case FIOSETOWN:
3040	case SIOCSPGRP:
3041	case FIOGETOWN:
3042	case SIOCGPGRP:
3043	case SIOCBRADDBR:
3044	case SIOCBRDELBR:
3045	case SIOCGIFVLAN:
3046	case SIOCSIFVLAN:
3047	case SIOCADDDLCI:
3048	case SIOCDELDLCI:
3049	case SIOCGSKNS:
3050	return sock_ioctl(file, cmd, arg);
3051
3052	case SIOCGIFFLAGS:
3053	case SIOCSIFFLAGS:
3054	case SIOCGIFMETRIC:
3055	case SIOCSIFMETRIC:
3056	case SIOCGIFMTU:
3057	case SIOCSIFMTU:
3058	case SIOCGIFMEM:
3059	case SIOCSIFMEM:
3060	case SIOCGIFHWADDR:
3061	case SIOCSIFHWADDR:
3062	case SIOCADDMULTI:
3063	case SIOCDELMULTI:
3064	case SIOCGIFINDEX:
3065	case SIOCGIFADDR:
3066	case SIOCSIFADDR:
3067	case SIOCSIFHWBROADCAST:
3068	case SIOCDIFADDR:
3069	case SIOCGIFBRDADDR:
3070	case SIOCSIFBRDADDR:
3071	case SIOCGIFDSTADDR:
3072	case SIOCSIFDSTADDR:
3073	case SIOCGIFNETMASK:
3074	case SIOCSIFNETMASK:
3075	case SIOCSIFPFLAGS:
3076	case SIOCGIFPFLAGS:
3077	case SIOCGIFTXQLEN:
3078	case SIOCSIFTXQLEN:
3079	case SIOCBRADDIF:
3080	case SIOCBRDELIF:
3081	case SIOCSIFNAME:
3082	case SIOCGMIIPHY:
3083	case SIOCGMIIREG:
3084	case SIOCSMIIREG:
3085	case SIOCSARP:
3086	case SIOCGARP:
3087	case SIOCDARP:
3088	case SIOCATMARK:
3089	case SIOCBONDENSLAVE:
3090	case SIOCBONDRELEASE:
3091	case SIOCBONDSETHWADDR:
3092	case SIOCBONDCHANGEACTIVE:
3093	case SIOCGIFNAME:
3094	return sock_do_ioctl(net, sock, cmd, arg);
3095	}
3096
3097	return -ENOIOCTLCMD;
3098	}
3099
3100	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3101	unsigned long arg)
3102	{
3103	struct socket *sock = file->private_data;
3104	int ret = -ENOIOCTLCMD;
3105	struct sock *sk;
3106	struct net *net;
3107
3108	sk = sock->sk;
3109	net = sock_net(sk);
3110
3111	if (sock->ops->compat_ioctl)
3112	ret = sock->ops->compat_ioctl(sock, cmd, arg);
3113
3114	if (ret == -ENOIOCTLCMD &&
3115	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3116	ret = compat_wext_handle_ioctl(net, cmd, arg);
3117
3118	if (ret == -ENOIOCTLCMD)
3119	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3120
3121	return ret;
3122	}
3123	#endif
3124
3125	int kernel_bind(struct socket sock, struct sockaddr addr, int addrlen)
3126	{
3127	return sock->ops->bind(sock, addr, addrlen);
3128	}
3129	EXPORT_SYMBOL(kernel_bind);
3130
3131	int kernel_listen(struct socket *sock, int backlog)
3132	{
3133	return sock->ops->listen(sock, backlog);
3134	}
3135	EXPORT_SYMBOL(kernel_listen);
3136
3137	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3138	{
3139	struct sock *sk = sock->sk;
3140	int err;
3141
3142	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3143	newsock);
3144	if (err < 0)
3145	goto done;
3146
3147	err = sock->ops->accept(sock, *newsock, flags, true);
3148	if (err < 0) {
3149	sock_release(*newsock);
3150	*newsock = NULL;
3151	goto done;
3152	}
3153
3154	(*newsock)->ops = sock->ops;
3155	__module_get((*newsock)->ops->owner);
3156
3157	done:
3158	return err;
3159	}
3160	EXPORT_SYMBOL(kernel_accept);
3161
3162	int kernel_connect(struct socket sock, struct sockaddr addr, int addrlen,
3163	int flags)
3164	{
3165	return sock->ops->connect(sock, addr, addrlen, flags);
3166	}
3167	EXPORT_SYMBOL(kernel_connect);
3168
3169	int kernel_getsockname(struct socket sock, struct sockaddr addr,
3170	int *addrlen)
3171	{
3172	return sock->ops->getname(sock, addr, addrlen, 0);
3173	}
3174	EXPORT_SYMBOL(kernel_getsockname);
3175
3176	int kernel_getpeername(struct socket sock, struct sockaddr addr,
3177	int *addrlen)
3178	{
3179	return sock->ops->getname(sock, addr, addrlen, 1);
3180	}
3181	EXPORT_SYMBOL(kernel_getpeername);
3182
3183	int kernel_getsockopt(struct socket *sock, int level, int optname,
3184	char optval, int optlen)
3185	{
3186	mm_segment_t oldfs = get_fs();
3187	char __user *uoptval;
3188	int __user *uoptlen;
3189	int err;
3190
3191	uoptval = (char __user __force *) optval;
3192	uoptlen = (int __user __force *) optlen;
3193
3194	set_fs(KERNEL_DS);
3195	if (level == SOL_SOCKET)
3196	err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3197	else
3198	err = sock->ops->getsockopt(sock, level, optname, uoptval,
3199	uoptlen);
3200	set_fs(oldfs);
3201	return err;
3202	}
3203	EXPORT_SYMBOL(kernel_getsockopt);
3204
3205	int kernel_setsockopt(struct socket *sock, int level, int optname,
3206	char *optval, unsigned int optlen)
3207	{
3208	mm_segment_t oldfs = get_fs();
3209	char __user *uoptval;
3210	int err;
3211
3212	uoptval = (char __user __force *) optval;
3213
3214	set_fs(KERNEL_DS);
3215	if (level == SOL_SOCKET)
3216	err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3217	else
3218	err = sock->ops->setsockopt(sock, level, optname, uoptval,
3219	optlen);
3220	set_fs(oldfs);
3221	return err;
3222	}
3223	EXPORT_SYMBOL(kernel_setsockopt);
3224
3225	int kernel_sendpage(struct socket sock, struct page page, int offset,
3226	size_t size, int flags)
3227	{
3228	if (sock->ops->sendpage)
3229	return sock->ops->sendpage(sock, page, offset, size, flags);
3230
3231	return sock_no_sendpage(sock, page, offset, size, flags);
3232	}
3233	EXPORT_SYMBOL(kernel_sendpage);
3234
3235	int kernel_sendpage_locked(struct sock sk, struct page page, int offset,
3236	size_t size, int flags)
3237	{
3238	struct socket *sock = sk->sk_socket;
3239
3240	if (sock->ops->sendpage_locked)
3241	return sock->ops->sendpage_locked(sk, page, offset, size,
3242	flags);
3243
3244	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3245	}
3246	EXPORT_SYMBOL(kernel_sendpage_locked);
3247
3248	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3249	{
3250	return sock->ops->shutdown(sock, how);
3251	}
3252	EXPORT_SYMBOL(kernel_sock_shutdown);
3253
3254	/* This routine returns the IP overhead imposed by a socket i.e.
3255	* the length of the underlying IP header, depending on whether
3256	* this is an IPv4 or IPv6 socket and the length from IP options turned
3257	* on at the socket. Assumes that the caller has a lock on the socket.
3258	*/
3259	u32 kernel_sock_ip_overhead(struct sock *sk)
3260	{
3261	struct inet_sock *inet;
3262	struct ip_options_rcu *opt;
3263	u32 overhead = 0;
3264	#if IS_ENABLED(CONFIG_IPV6)
3265	struct ipv6_pinfo *np;
3266	struct ipv6_txoptions *optv6 = NULL;
3267	#endif /* IS_ENABLED(CONFIG_IPV6) */
3268
3269	if (!sk)
3270	return overhead;
3271
3272	switch (sk->sk_family) {
3273	case AF_INET:
3274	inet = inet_sk(sk);
3275	overhead += sizeof(struct iphdr);
3276	opt = rcu_dereference_protected(inet->inet_opt,
3277	sock_owned_by_user(sk));
3278	if (opt)
3279	overhead += opt->opt.optlen;
3280	return overhead;
3281	#if IS_ENABLED(CONFIG_IPV6)
3282	case AF_INET6:
3283	np = inet6_sk(sk);
3284	overhead += sizeof(struct ipv6hdr);
3285	if (np)
3286	optv6 = rcu_dereference_protected(np->opt,
3287	sock_owned_by_user(sk));
3288	if (optv6)
3289	overhead += (optv6->opt_flen + optv6->opt_nflen);
3290	return overhead;
3291	#endif /* IS_ENABLED(CONFIG_IPV6) */
3292	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3293	return overhead;
3294	}
3295	}
3296	EXPORT_SYMBOL(kernel_sock_ip_overhead);