linux - net/socket.c

3734 lines · 91.4 KiB

1	// SPDX-License-Identifier: GPL-2.0-or-later
2	/*
3	* NET An implementation of the SOCKET network access protocol.
4	*
5	* Version: @(#)socket.c 1.1.93 18/02/95
6	*
7	* Authors: Orest Zborowski, <[email protected]>
8	* Ross Biro
9	* Fred N. van Kempen, <[email protected]>
10	*
11	* Fixes:
12	* Anonymous : NOTSOCK/BADF cleanup. Error fix in
13	* shutdown()
14	* Alan Cox : verify_area() fixes
15	* Alan Cox : Removed DDI
16	* Jonathan Kamens : SOCK_DGRAM reconnect bug
17	* Alan Cox : Moved a load of checks to the very
18	* top level.
19	* Alan Cox : Move address structures to/from user
20	* mode above the protocol layers.
21	* Rob Janssen : Allow 0 length sends.
22	* Alan Cox : Asynchronous I/O support (cribbed from the
23	* tty drivers).
24	* Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25	* Jeff Uphoff : Made max number of sockets command-line
26	* configurable.
27	* Matti Aarnio : Made the number of sockets dynamic,
28	* to be allocated when needed, and mr.
29	* Uphoff's max is used as max to be
30	* allowed to allocate.
31	* Linus : Argh. removed all the socket allocation
32	* altogether: it's in the inode now.
33	* Alan Cox : Made sock_alloc()/sock_release() public
34	* for NetROM and future kernel nfsd type
35	* stuff.
36	* Alan Cox : sendmsg/recvmsg basics.
37	* Tom Dyas : Export net symbols.
38	* Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39	* Alan Cox : Added thread locking to sys_* calls
40	* for sockets. May have errors at the
41	* moment.
42	* Kevin Buhr : Fixed the dumb errors in the above.
43	* Andi Kleen : Some small cleanups, optimizations,
44	* and fixed a copy_from_user() bug.
45	* Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46	* Tigran Aivazian : Made listen(2) backlog sanity checks
47	* protocol-independent
48	*
49	* This module is effectively the top level interface to the BSD socket
50	* paradigm.
51	*
52	* Based upon Swansea University Computer Society NET3.039
53	*/
54
55	#include <linux/mm.h>
56	#include <linux/socket.h>
57	#include <linux/file.h>
58	#include <linux/net.h>
59	#include <linux/interrupt.h>
60	#include <linux/thread_info.h>
61	#include <linux/rcupdate.h>
62	#include <linux/netdevice.h>
63	#include <linux/proc_fs.h>
64	#include <linux/seq_file.h>
65	#include <linux/mutex.h>
66	#include <linux/if_bridge.h>
67	#include <linux/if_frad.h>
68	#include <linux/if_vlan.h>
69	#include <linux/ptp_classify.h>
70	#include <linux/init.h>
71	#include <linux/poll.h>
72	#include <linux/cache.h>
73	#include <linux/module.h>
74	#include <linux/highmem.h>
75	#include <linux/mount.h>
76	#include <linux/pseudo_fs.h>
77	#include <linux/security.h>
78	#include <linux/syscalls.h>
79	#include <linux/compat.h>
80	#include <linux/kmod.h>
81	#include <linux/audit.h>
82	#include <linux/wireless.h>
83	#include <linux/nsproxy.h>
84	#include <linux/magic.h>
85	#include <linux/slab.h>
86	#include <linux/xattr.h>
87	#include <linux/nospec.h>
88	#include <linux/indirect_call_wrapper.h>
89
90	#include <linux/uaccess.h>
91	#include <asm/unistd.h>
92
93	#include <net/compat.h>
94	#include <net/wext.h>
95	#include <net/cls_cgroup.h>
96
97	#include <net/sock.h>
98	#include <linux/netfilter.h>
99
100	#include <linux/if_tun.h>
101	#include <linux/ipv6_route.h>
102	#include <linux/route.h>
103	#include <linux/termios.h>
104	#include <linux/sockios.h>
105	#include <net/busy_poll.h>
106	#include <linux/errqueue.h>
107
108	#ifdef CONFIG_NET_RX_BUSY_POLL
109	unsigned int sysctl_net_busy_read __read_mostly;
110	unsigned int sysctl_net_busy_poll __read_mostly;
111	#endif
112
113	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to);
114	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from);
115	static int sock_mmap(struct file file, struct vm_area_struct vma);
116
117	static int sock_close(struct inode inode, struct file file);
118	static __poll_t sock_poll(struct file *file,
119	struct poll_table_struct *wait);
120	static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
121	#ifdef CONFIG_COMPAT
122	static long compat_sock_ioctl(struct file *file,
123	unsigned int cmd, unsigned long arg);
124	#endif
125	static int sock_fasync(int fd, struct file *filp, int on);
126	static ssize_t sock_sendpage(struct file file, struct page page,
127	int offset, size_t size, loff_t *ppos, int more);
128	static ssize_t sock_splice_read(struct file file, loff_t ppos,
129	struct pipe_inode_info *pipe, size_t len,
130	unsigned int flags);
131
132	#ifdef CONFIG_PROC_FS
133	static void sock_show_fdinfo(struct seq_file m, struct file f)
134	{
135	struct socket *sock = f->private_data;
136
137	if (sock->ops->show_fdinfo)
138	sock->ops->show_fdinfo(m, sock);
139	}
140	#else
141	#define sock_show_fdinfo NULL
142	#endif
143
144	/*
145	* Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146	* in the operation structures but are done directly via the socketcall() multiplexor.
147	*/
148
149	static const struct file_operations socket_file_ops = {
150	.owner = THIS_MODULE,
151	.llseek = no_llseek,
152	.read_iter = sock_read_iter,
153	.write_iter = sock_write_iter,
154	.poll = sock_poll,
155	.unlocked_ioctl = sock_ioctl,
156	#ifdef CONFIG_COMPAT
157	.compat_ioctl = compat_sock_ioctl,
158	#endif
159	.mmap = sock_mmap,
160	.release = sock_close,
161	.fasync = sock_fasync,
162	.sendpage = sock_sendpage,
163	.splice_write = generic_splice_sendpage,
164	.splice_read = sock_splice_read,
165	.show_fdinfo = sock_show_fdinfo,
166	};
167
168	/*
169	* The protocol list. Each protocol is registered in here.
170	*/
171
172	static DEFINE_SPINLOCK(net_family_lock);
173	static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
174
175	/*
176	* Support routines.
177	* Move socket addresses back and forth across the kernel/user
178	* divide and look after the messy bits.
179	*/
180
181	/**
182	* move_addr_to_kernel - copy a socket address into kernel space
183	* @uaddr: Address in user space
184	* @kaddr: Address in kernel space
185	* @ulen: Length in user space
186	*
187	* The address is copied into kernel space. If the provided address is
188	* too long an error code of -EINVAL is returned. If the copy gives
189	* invalid addresses -EFAULT is returned. On a success 0 is returned.
190	*/
191
192	int move_addr_to_kernel(void __user uaddr, int ulen, struct sockaddr_storage kaddr)
193	{
194	if (ulen < 0 \|\| ulen > sizeof(struct sockaddr_storage))
195	return -EINVAL;
196	if (ulen == 0)
197	return 0;
198	if (copy_from_user(kaddr, uaddr, ulen))
199	return -EFAULT;
200	return audit_sockaddr(ulen, kaddr);
201	}
202
203	/**
204	* move_addr_to_user - copy an address to user space
205	* @kaddr: kernel space address
206	* @klen: length of address in kernel
207	* @uaddr: user space address
208	* @ulen: pointer to user length field
209	*
210	* The value pointed to by ulen on entry is the buffer length available.
211	* This is overwritten with the buffer space used. -EINVAL is returned
212	* if an overlong buffer is specified or a negative buffer size. -EFAULT
213	* is returned if either the buffer or the length field are not
214	* accessible.
215	* After copying the data up to the limit the user specifies, the true
216	* length of the data is written over the length limit the user
217	* specified. Zero is returned for a success.
218	*/
219
220	static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
221	void __user uaddr, int __user ulen)
222	{
223	int err;
224	int len;
225
226	BUG_ON(klen > sizeof(struct sockaddr_storage));
227	err = get_user(len, ulen);
228	if (err)
229	return err;
230	if (len > klen)
231	len = klen;
232	if (len < 0)
233	return -EINVAL;
234	if (len) {
235	if (audit_sockaddr(klen, kaddr))
236	return -ENOMEM;
237	if (copy_to_user(uaddr, kaddr, len))
238	return -EFAULT;
239	}
240	/*
241	* "fromlen shall refer to the value before truncation.."
242	* 1003.1g
243	*/
244	return __put_user(klen, ulen);
245	}
246
247	static struct kmem_cache *sock_inode_cachep __ro_after_init;
248
249	static struct inode sock_alloc_inode(struct super_block sb)
250	{
251	struct socket_alloc *ei;
252
253	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
254	if (!ei)
255	return NULL;
256	init_waitqueue_head(&ei->socket.wq.wait);
257	ei->socket.wq.fasync_list = NULL;
258	ei->socket.wq.flags = 0;
259
260	ei->socket.state = SS_UNCONNECTED;
261	ei->socket.flags = 0;
262	ei->socket.ops = NULL;
263	ei->socket.sk = NULL;
264	ei->socket.file = NULL;
265
266	return &ei->vfs_inode;
267	}
268
269	static void sock_free_inode(struct inode *inode)
270	{
271	struct socket_alloc *ei;
272
273	ei = container_of(inode, struct socket_alloc, vfs_inode);
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	.free_inode = sock_free_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 int sockfs_init_fs_context(struct fs_context *fc)
357	{
358	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
359	if (!ctx)
360	return -ENOMEM;
361	ctx->ops = &sockfs_ops;
362	ctx->dops = &sockfs_dentry_operations;
363	ctx->xattr = sockfs_xattr_handlers;
364	return 0;
365	}
366
367	static struct vfsmount *sock_mnt __read_mostly;
368
369	static struct file_system_type sock_fs_type = {
370	.name = "sockfs",
371	.init_fs_context = sockfs_init_fs_context,
372	.kill_sb = kill_anon_super,
373	};
374
375	/*
376	* Obtains the first available file descriptor and sets it up for use.
377	*
378	* These functions create file structures and maps them to fd space
379	* of the current process. On success it returns file descriptor
380	* and file struct implicitly stored in sock->file.
381	* Note that another thread may close file descriptor before we return
382	* from this function. We use the fact that now we do not refer
383	* to socket after mapping. If one day we will need it, this
384	* function will increment ref. count on file by 1.
385	*
386	* In any case returned fd MAY BE not valid!
387	* This race condition is unavoidable
388	* with shared fd spaces, we cannot solve it inside kernel,
389	* but we take care of internal coherence yet.
390	*/
391
392	/**
393	* sock_alloc_file - Bind a &socket to a &file
394	* @sock: socket
395	* @flags: file status flags
396	* @dname: protocol name
397	*
398	* Returns the &file bound with @sock, implicitly storing it
399	* in sock->file. If dname is %NULL, sets to "".
400	* On failure the return is a ERR pointer (see linux/err.h).
401	* This function uses GFP_KERNEL internally.
402	*/
403
404	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
405	{
406	struct file *file;
407
408	if (!dname)
409	dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
410
411	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
412	O_RDWR \| (flags & O_NONBLOCK),
413	&socket_file_ops);
414	if (IS_ERR(file)) {
415	sock_release(sock);
416	return file;
417	}
418
419	sock->file = file;
420	file->private_data = sock;
421	stream_open(SOCK_INODE(sock), file);
422	return file;
423	}
424	EXPORT_SYMBOL(sock_alloc_file);
425
426	static int sock_map_fd(struct socket *sock, int flags)
427	{
428	struct file *newfile;
429	int fd = get_unused_fd_flags(flags);
430	if (unlikely(fd < 0)) {
431	sock_release(sock);
432	return fd;
433	}
434
435	newfile = sock_alloc_file(sock, flags, NULL);
436	if (!IS_ERR(newfile)) {
437	fd_install(fd, newfile);
438	return fd;
439	}
440
441	put_unused_fd(fd);
442	return PTR_ERR(newfile);
443	}
444
445	/**
446	* sock_from_file - Return the &socket bounded to @file.
447	* @file: file
448	* @err: pointer to an error code return
449	*
450	* On failure returns %NULL and assigns -ENOTSOCK to @err.
451	*/
452
453	struct socket sock_from_file(struct file file, int *err)
454	{
455	if (file->f_op == &socket_file_ops)
456	return file->private_data; /* set in sock_map_fd */
457
458	*err = -ENOTSOCK;
459	return NULL;
460	}
461	EXPORT_SYMBOL(sock_from_file);
462
463	/**
464	* sockfd_lookup - Go from a file number to its socket slot
465	* @fd: file handle
466	* @err: pointer to an error code return
467	*
468	* The file handle passed in is locked and the socket it is bound
469	* to is returned. If an error occurs the err pointer is overwritten
470	* with a negative errno code and NULL is returned. The function checks
471	* for both invalid handles and passing a handle which is not a socket.
472	*
473	* On a success the socket object pointer is returned.
474	*/
475
476	struct socket sockfd_lookup(int fd, int err)
477	{
478	struct file *file;
479	struct socket *sock;
480
481	file = fget(fd);
482	if (!file) {
483	*err = -EBADF;
484	return NULL;
485	}
486
487	sock = sock_from_file(file, err);
488	if (!sock)
489	fput(file);
490	return sock;
491	}
492	EXPORT_SYMBOL(sockfd_lookup);
493
494	static struct socket sockfd_lookup_light(int fd, int err, int *fput_needed)
495	{
496	struct fd f = fdget(fd);
497	struct socket *sock;
498
499	*err = -EBADF;
500	if (f.file) {
501	sock = sock_from_file(f.file, err);
502	if (likely(sock)) {
503	*fput_needed = f.flags;
504	return sock;
505	}
506	fdput(f);
507	}
508	return NULL;
509	}
510
511	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
512	size_t size)
513	{
514	ssize_t len;
515	ssize_t used = 0;
516
517	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
518	if (len < 0)
519	return len;
520	used += len;
521	if (buffer) {
522	if (size < used)
523	return -ERANGE;
524	buffer += len;
525	}
526
527	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
528	used += len;
529	if (buffer) {
530	if (size < used)
531	return -ERANGE;
532	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
533	buffer += len;
534	}
535
536	return used;
537	}
538
539	static int sockfs_setattr(struct dentry dentry, struct iattr iattr)
540	{
541	int err = simple_setattr(dentry, iattr);
542
543	if (!err && (iattr->ia_valid & ATTR_UID)) {
544	struct socket *sock = SOCKET_I(d_inode(dentry));
545
546	if (sock->sk)
547	sock->sk->sk_uid = iattr->ia_uid;
548	else
549	err = -ENOENT;
550	}
551
552	return err;
553	}
554
555	static const struct inode_operations sockfs_inode_ops = {
556	.listxattr = sockfs_listxattr,
557	.setattr = sockfs_setattr,
558	};
559
560	/**
561	* sock_alloc - allocate a socket
562	*
563	* Allocate a new inode and socket object. The two are bound together
564	* and initialised. The socket is then returned. If we are out of inodes
565	* NULL is returned. This functions uses GFP_KERNEL internally.
566	*/
567
568	struct socket *sock_alloc(void)
569	{
570	struct inode *inode;
571	struct socket *sock;
572
573	inode = new_inode_pseudo(sock_mnt->mnt_sb);
574	if (!inode)
575	return NULL;
576
577	sock = SOCKET_I(inode);
578
579	inode->i_ino = get_next_ino();
580	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
581	inode->i_uid = current_fsuid();
582	inode->i_gid = current_fsgid();
583	inode->i_op = &sockfs_inode_ops;
584
585	return sock;
586	}
587	EXPORT_SYMBOL(sock_alloc);
588
589	/**
590	* sock_release - close a socket
591	* @sock: socket to close
592	*
593	* The socket is released from the protocol stack if it has a release
594	* callback, and the inode is then released if the socket is bound to
595	* an inode not a file.
596	*/
597
598	static void __sock_release(struct socket sock, struct inode inode)
599	{
600	if (sock->ops) {
601	struct module *owner = sock->ops->owner;
602
603	if (inode)
604	inode_lock(inode);
605	sock->ops->release(sock);
606	sock->sk = NULL;
607	if (inode)
608	inode_unlock(inode);
609	sock->ops = NULL;
610	module_put(owner);
611	}
612
613	if (sock->wq.fasync_list)
614	pr_err("%s: fasync list not empty!\n", __func__);
615
616	if (!sock->file) {
617	iput(SOCK_INODE(sock));
618	return;
619	}
620	sock->file = NULL;
621	}
622
623	void sock_release(struct socket *sock)
624	{
625	__sock_release(sock, NULL);
626	}
627	EXPORT_SYMBOL(sock_release);
628
629	void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
630	{
631	u8 flags = *tx_flags;
632
633	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
634	flags \|= SKBTX_HW_TSTAMP;
635
636	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
637	flags \|= SKBTX_SW_TSTAMP;
638
639	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
640	flags \|= SKBTX_SCHED_TSTAMP;
641
642	*tx_flags = flags;
643	}
644	EXPORT_SYMBOL(__sock_tx_timestamp);
645
646	INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket , struct msghdr ,
647	size_t));
648	INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket , struct msghdr ,
649	size_t));
650	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
651	{
652	int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
653	inet_sendmsg, sock, msg,
654	msg_data_left(msg));
655	BUG_ON(ret == -EIOCBQUEUED);
656	return ret;
657	}
658
659	/**
660	* sock_sendmsg - send a message through @sock
661	* @sock: socket
662	* @msg: message to send
663	*
664	* Sends @msg through @sock, passing through LSM.
665	* Returns the number of bytes sent, or an error code.
666	*/
667	int sock_sendmsg(struct socket sock, struct msghdr msg)
668	{
669	int err = security_socket_sendmsg(sock, msg,
670	msg_data_left(msg));
671
672	return err ?: sock_sendmsg_nosec(sock, msg);
673	}
674	EXPORT_SYMBOL(sock_sendmsg);
675
676	/**
677	* kernel_sendmsg - send a message through @sock (kernel-space)
678	* @sock: socket
679	* @msg: message header
680	* @vec: kernel vec
681	* @num: vec array length
682	* @size: total message data size
683	*
684	* Builds the message data with @vec and sends it through @sock.
685	* Returns the number of bytes sent, or an error code.
686	*/
687
688	int kernel_sendmsg(struct socket sock, struct msghdr msg,
689	struct kvec *vec, size_t num, size_t size)
690	{
691	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
692	return sock_sendmsg(sock, msg);
693	}
694	EXPORT_SYMBOL(kernel_sendmsg);
695
696	/**
697	* kernel_sendmsg_locked - send a message through @sock (kernel-space)
698	* @sk: sock
699	* @msg: message header
700	* @vec: output s/g array
701	* @num: output s/g array length
702	* @size: total message data size
703	*
704	* Builds the message data with @vec and sends it through @sock.
705	* Returns the number of bytes sent, or an error code.
706	* Caller must hold @sk.
707	*/
708
709	int kernel_sendmsg_locked(struct sock sk, struct msghdr msg,
710	struct kvec *vec, size_t num, size_t size)
711	{
712	struct socket *sock = sk->sk_socket;
713
714	if (!sock->ops->sendmsg_locked)
715	return sock_no_sendmsg_locked(sk, msg, size);
716
717	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
718
719	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
720	}
721	EXPORT_SYMBOL(kernel_sendmsg_locked);
722
723	static bool skb_is_err_queue(const struct sk_buff *skb)
724	{
725	/* pkt_type of skbs enqueued on the error queue are set to
726	* PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
727	* in recvmsg, since skbs received on a local socket will never
728	* have a pkt_type of PACKET_OUTGOING.
729	*/
730	return skb->pkt_type == PACKET_OUTGOING;
731	}
732
733	/* On transmit, software and hardware timestamps are returned independently.
734	* As the two skb clones share the hardware timestamp, which may be updated
735	* before the software timestamp is received, a hardware TX timestamp may be
736	* returned only if there is no software TX timestamp. Ignore false software
737	* timestamps, which may be made in the __sock_recv_timestamp() call when the
738	* option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
739	* hardware timestamp.
740	*/
741	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
742	{
743	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
744	}
745
746	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb)
747	{
748	struct scm_ts_pktinfo ts_pktinfo;
749	struct net_device *orig_dev;
750
751	if (!skb_mac_header_was_set(skb))
752	return;
753
754	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
755
756	rcu_read_lock();
757	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
758	if (orig_dev)
759	ts_pktinfo.if_index = orig_dev->ifindex;
760	rcu_read_unlock();
761
762	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
763	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
764	sizeof(ts_pktinfo), &ts_pktinfo);
765	}
766
767	/*
768	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
769	*/
770	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
771	struct sk_buff *skb)
772	{
773	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
774	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
775	struct scm_timestamping_internal tss;
776
777	int empty = 1, false_tstamp = 0;
778	struct skb_shared_hwtstamps *shhwtstamps =
779	skb_hwtstamps(skb);
780
781	/* Race occurred between timestamp enabling and packet
782	receiving. Fill in the current time for now. */
783	if (need_software_tstamp && skb->tstamp == 0) {
784	__net_timestamp(skb);
785	false_tstamp = 1;
786	}
787
788	if (need_software_tstamp) {
789	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
790	if (new_tstamp) {
791	struct __kernel_sock_timeval tv;
792
793	skb_get_new_timestamp(skb, &tv);
794	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
795	sizeof(tv), &tv);
796	} else {
797	struct __kernel_old_timeval tv;
798
799	skb_get_timestamp(skb, &tv);
800	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
801	sizeof(tv), &tv);
802	}
803	} else {
804	if (new_tstamp) {
805	struct __kernel_timespec ts;
806
807	skb_get_new_timestampns(skb, &ts);
808	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
809	sizeof(ts), &ts);
810	} else {
811	struct __kernel_old_timespec ts;
812
813	skb_get_timestampns(skb, &ts);
814	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
815	sizeof(ts), &ts);
816	}
817	}
818	}
819
820	memset(&tss, 0, sizeof(tss));
821	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
822	ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
823	empty = 0;
824	if (shhwtstamps &&
825	(sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
826	!skb_is_swtx_tstamp(skb, false_tstamp) &&
827	ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
828	empty = 0;
829	if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
830	!skb_is_err_queue(skb))
831	put_ts_pktinfo(msg, skb);
832	}
833	if (!empty) {
834	if (sock_flag(sk, SOCK_TSTAMP_NEW))
835	put_cmsg_scm_timestamping64(msg, &tss);
836	else
837	put_cmsg_scm_timestamping(msg, &tss);
838
839	if (skb_is_err_queue(skb) && skb->len &&
840	SKB_EXT_ERR(skb)->opt_stats)
841	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
842	skb->len, skb->data);
843	}
844	}
845	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
846
847	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
848	struct sk_buff *skb)
849	{
850	int ack;
851
852	if (!sock_flag(sk, SOCK_WIFI_STATUS))
853	return;
854	if (!skb->wifi_acked_valid)
855	return;
856
857	ack = skb->wifi_acked;
858
859	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
860	}
861	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
862
863	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
864	struct sk_buff *skb)
865	{
866	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
867	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
868	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
869	}
870
871	void __sock_recv_ts_and_drops(struct msghdr msg, struct sock sk,
872	struct sk_buff *skb)
873	{
874	sock_recv_timestamp(msg, sk, skb);
875	sock_recv_drops(msg, sk, skb);
876	}
877	EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
878
879	INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket , struct msghdr ,
880	size_t, int));
881	INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket , struct msghdr ,
882	size_t, int));
883	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
884	int flags)
885	{
886	return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
887	inet_recvmsg, sock, msg, msg_data_left(msg),
888	flags);
889	}
890
891	/**
892	* sock_recvmsg - receive a message from @sock
893	* @sock: socket
894	* @msg: message to receive
895	* @flags: message flags
896	*
897	* Receives @msg from @sock, passing through LSM. Returns the total number
898	* of bytes received, or an error.
899	*/
900	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
901	{
902	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
903
904	return err ?: sock_recvmsg_nosec(sock, msg, flags);
905	}
906	EXPORT_SYMBOL(sock_recvmsg);
907
908	/**
909	* kernel_recvmsg - Receive a message from a socket (kernel space)
910	* @sock: The socket to receive the message from
911	* @msg: Received message
912	* @vec: Input s/g array for message data
913	* @num: Size of input s/g array
914	* @size: Number of bytes to read
915	* @flags: Message flags (MSG_DONTWAIT, etc...)
916	*
917	* On return the msg structure contains the scatter/gather array passed in the
918	* vec argument. The array is modified so that it consists of the unfilled
919	* portion of the original array.
920	*
921	* The returned value is the total number of bytes received, or an error.
922	*/
923
924	int kernel_recvmsg(struct socket sock, struct msghdr msg,
925	struct kvec *vec, size_t num, size_t size, int flags)
926	{
927	msg->msg_control_is_user = false;
928	iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
929	return sock_recvmsg(sock, msg, flags);
930	}
931	EXPORT_SYMBOL(kernel_recvmsg);
932
933	static ssize_t sock_sendpage(struct file file, struct page page,
934	int offset, size_t size, loff_t *ppos, int more)
935	{
936	struct socket *sock;
937	int flags;
938
939	sock = file->private_data;
940
941	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
942	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
943	flags \|= more;
944
945	return kernel_sendpage(sock, page, offset, size, flags);
946	}
947
948	static ssize_t sock_splice_read(struct file file, loff_t ppos,
949	struct pipe_inode_info *pipe, size_t len,
950	unsigned int flags)
951	{
952	struct socket *sock = file->private_data;
953
954	if (unlikely(!sock->ops->splice_read))
955	return generic_file_splice_read(file, ppos, pipe, len, flags);
956
957	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
958	}
959
960	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
961	{
962	struct file *file = iocb->ki_filp;
963	struct socket *sock = file->private_data;
964	struct msghdr msg = {.msg_iter = *to,
965	.msg_iocb = iocb};
966	ssize_t res;
967
968	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
969	msg.msg_flags = MSG_DONTWAIT;
970
971	if (iocb->ki_pos != 0)
972	return -ESPIPE;
973
974	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
975	return 0;
976
977	res = sock_recvmsg(sock, &msg, msg.msg_flags);
978	*to = msg.msg_iter;
979	return res;
980	}
981
982	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
983	{
984	struct file *file = iocb->ki_filp;
985	struct socket *sock = file->private_data;
986	struct msghdr msg = {.msg_iter = *from,
987	.msg_iocb = iocb};
988	ssize_t res;
989
990	if (iocb->ki_pos != 0)
991	return -ESPIPE;
992
993	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
994	msg.msg_flags = MSG_DONTWAIT;
995
996	if (sock->type == SOCK_SEQPACKET)
997	msg.msg_flags \|= MSG_EOR;
998
999	res = sock_sendmsg(sock, &msg);
1000	*from = msg.msg_iter;
1001	return res;
1002	}
1003
1004	/*
1005	* Atomic setting of ioctl hooks to avoid race
1006	* with module unload.
1007	*/
1008
1009	static DEFINE_MUTEX(br_ioctl_mutex);
1010	static int (br_ioctl_hook) (struct net , unsigned int cmd, void __user *arg);
1011
1012	void brioctl_set(int (hook) (struct net , unsigned int, void __user *))
1013	{
1014	mutex_lock(&br_ioctl_mutex);
1015	br_ioctl_hook = hook;
1016	mutex_unlock(&br_ioctl_mutex);
1017	}
1018	EXPORT_SYMBOL(brioctl_set);
1019
1020	static DEFINE_MUTEX(vlan_ioctl_mutex);
1021	static int (vlan_ioctl_hook) (struct net , void __user *arg);
1022
1023	void vlan_ioctl_set(int (hook) (struct net , void __user *))
1024	{
1025	mutex_lock(&vlan_ioctl_mutex);
1026	vlan_ioctl_hook = hook;
1027	mutex_unlock(&vlan_ioctl_mutex);
1028	}
1029	EXPORT_SYMBOL(vlan_ioctl_set);
1030
1031	static DEFINE_MUTEX(dlci_ioctl_mutex);
1032	static int (dlci_ioctl_hook) (unsigned int, void __user );
1033
1034	void dlci_ioctl_set(int (hook) (unsigned int, void __user ))
1035	{
1036	mutex_lock(&dlci_ioctl_mutex);
1037	dlci_ioctl_hook = hook;
1038	mutex_unlock(&dlci_ioctl_mutex);
1039	}
1040	EXPORT_SYMBOL(dlci_ioctl_set);
1041
1042	static long sock_do_ioctl(struct net net, struct socket sock,
1043	unsigned int cmd, unsigned long arg)
1044	{
1045	int err;
1046	void __user argp = (void __user )arg;
1047
1048	err = sock->ops->ioctl(sock, cmd, arg);
1049
1050	/*
1051	* If this ioctl is unknown try to hand it down
1052	* to the NIC driver.
1053	*/
1054	if (err != -ENOIOCTLCMD)
1055	return err;
1056
1057	if (cmd == SIOCGIFCONF) {
1058	struct ifconf ifc;
1059	if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1060	return -EFAULT;
1061	rtnl_lock();
1062	err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1063	rtnl_unlock();
1064	if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1065	err = -EFAULT;
1066	} else {
1067	struct ifreq ifr;
1068	bool need_copyout;
1069	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1070	return -EFAULT;
1071	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1072	if (!err && need_copyout)
1073	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1074	return -EFAULT;
1075	}
1076	return err;
1077	}
1078
1079	/*
1080	* With an ioctl, arg may well be a user mode pointer, but we don't know
1081	* what to do with it - that's up to the protocol still.
1082	*/
1083
1084	/**
1085	* get_net_ns - increment the refcount of the network namespace
1086	* @ns: common namespace (net)
1087	*
1088	* Returns the net's common namespace.
1089	*/
1090
1091	struct ns_common get_net_ns(struct ns_common ns)
1092	{
1093	return &get_net(container_of(ns, struct net, ns))->ns;
1094	}
1095	EXPORT_SYMBOL_GPL(get_net_ns);
1096
1097	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1098	{
1099	struct socket *sock;
1100	struct sock *sk;
1101	void __user argp = (void __user )arg;
1102	int pid, err;
1103	struct net *net;
1104
1105	sock = file->private_data;
1106	sk = sock->sk;
1107	net = sock_net(sk);
1108	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1109	struct ifreq ifr;
1110	bool need_copyout;
1111	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1112	return -EFAULT;
1113	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1114	if (!err && need_copyout)
1115	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1116	return -EFAULT;
1117	} else
1118	#ifdef CONFIG_WEXT_CORE
1119	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1120	err = wext_handle_ioctl(net, cmd, argp);
1121	} else
1122	#endif
1123	switch (cmd) {
1124	case FIOSETOWN:
1125	case SIOCSPGRP:
1126	err = -EFAULT;
1127	if (get_user(pid, (int __user *)argp))
1128	break;
1129	err = f_setown(sock->file, pid, 1);
1130	break;
1131	case FIOGETOWN:
1132	case SIOCGPGRP:
1133	err = put_user(f_getown(sock->file),
1134	(int __user *)argp);
1135	break;
1136	case SIOCGIFBR:
1137	case SIOCSIFBR:
1138	case SIOCBRADDBR:
1139	case SIOCBRDELBR:
1140	err = -ENOPKG;
1141	if (!br_ioctl_hook)
1142	request_module("bridge");
1143
1144	mutex_lock(&br_ioctl_mutex);
1145	if (br_ioctl_hook)
1146	err = br_ioctl_hook(net, cmd, argp);
1147	mutex_unlock(&br_ioctl_mutex);
1148	break;
1149	case SIOCGIFVLAN:
1150	case SIOCSIFVLAN:
1151	err = -ENOPKG;
1152	if (!vlan_ioctl_hook)
1153	request_module("8021q");
1154
1155	mutex_lock(&vlan_ioctl_mutex);
1156	if (vlan_ioctl_hook)
1157	err = vlan_ioctl_hook(net, argp);
1158	mutex_unlock(&vlan_ioctl_mutex);
1159	break;
1160	case SIOCADDDLCI:
1161	case SIOCDELDLCI:
1162	err = -ENOPKG;
1163	if (!dlci_ioctl_hook)
1164	request_module("dlci");
1165
1166	mutex_lock(&dlci_ioctl_mutex);
1167	if (dlci_ioctl_hook)
1168	err = dlci_ioctl_hook(cmd, argp);
1169	mutex_unlock(&dlci_ioctl_mutex);
1170	break;
1171	case SIOCGSKNS:
1172	err = -EPERM;
1173	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1174	break;
1175
1176	err = open_related_ns(&net->ns, get_net_ns);
1177	break;
1178	case SIOCGSTAMP_OLD:
1179	case SIOCGSTAMPNS_OLD:
1180	if (!sock->ops->gettstamp) {
1181	err = -ENOIOCTLCMD;
1182	break;
1183	}
1184	err = sock->ops->gettstamp(sock, argp,
1185	cmd == SIOCGSTAMP_OLD,
1186	!IS_ENABLED(CONFIG_64BIT));
1187	break;
1188	case SIOCGSTAMP_NEW:
1189	case SIOCGSTAMPNS_NEW:
1190	if (!sock->ops->gettstamp) {
1191	err = -ENOIOCTLCMD;
1192	break;
1193	}
1194	err = sock->ops->gettstamp(sock, argp,
1195	cmd == SIOCGSTAMP_NEW,
1196	false);
1197	break;
1198	default:
1199	err = sock_do_ioctl(net, sock, cmd, arg);
1200	break;
1201	}
1202	return err;
1203	}
1204
1205	/**
1206	* sock_create_lite - creates a socket
1207	* @family: protocol family (AF_INET, ...)
1208	* @type: communication type (SOCK_STREAM, ...)
1209	* @protocol: protocol (0, ...)
1210	* @res: new socket
1211	*
1212	* Creates a new socket and assigns it to @res, passing through LSM.
1213	* The new socket initialization is not complete, see kernel_accept().
1214	* Returns 0 or an error. On failure @res is set to %NULL.
1215	* This function internally uses GFP_KERNEL.
1216	*/
1217
1218	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1219	{
1220	int err;
1221	struct socket *sock = NULL;
1222
1223	err = security_socket_create(family, type, protocol, 1);
1224	if (err)
1225	goto out;
1226
1227	sock = sock_alloc();
1228	if (!sock) {
1229	err = -ENOMEM;
1230	goto out;
1231	}
1232
1233	sock->type = type;
1234	err = security_socket_post_create(sock, family, type, protocol, 1);
1235	if (err)
1236	goto out_release;
1237
1238	out:
1239	*res = sock;
1240	return err;
1241	out_release:
1242	sock_release(sock);
1243	sock = NULL;
1244	goto out;
1245	}
1246	EXPORT_SYMBOL(sock_create_lite);
1247
1248	/* No kernel lock held - perfect */
1249	static __poll_t sock_poll(struct file file, poll_table wait)
1250	{
1251	struct socket *sock = file->private_data;
1252	__poll_t events = poll_requested_events(wait), flag = 0;
1253
1254	if (!sock->ops->poll)
1255	return 0;
1256
1257	if (sk_can_busy_loop(sock->sk)) {
1258	/* poll once if requested by the syscall */
1259	if (events & POLL_BUSY_LOOP)
1260	sk_busy_loop(sock->sk, 1);
1261
1262	/* if this socket can poll_ll, tell the system call */
1263	flag = POLL_BUSY_LOOP;
1264	}
1265
1266	return sock->ops->poll(file, sock, wait) \| flag;
1267	}
1268
1269	static int sock_mmap(struct file file, struct vm_area_struct vma)
1270	{
1271	struct socket *sock = file->private_data;
1272
1273	return sock->ops->mmap(file, sock, vma);
1274	}
1275
1276	static int sock_close(struct inode inode, struct file filp)
1277	{
1278	__sock_release(SOCKET_I(inode), inode);
1279	return 0;
1280	}
1281
1282	/*
1283	* Update the socket async list
1284	*
1285	* Fasync_list locking strategy.
1286	*
1287	* 1. fasync_list is modified only under process context socket lock
1288	* i.e. under semaphore.
1289	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1290	* or under socket lock
1291	*/
1292
1293	static int sock_fasync(int fd, struct file *filp, int on)
1294	{
1295	struct socket *sock = filp->private_data;
1296	struct sock *sk = sock->sk;
1297	struct socket_wq *wq = &sock->wq;
1298
1299	if (sk == NULL)
1300	return -EINVAL;
1301
1302	lock_sock(sk);
1303	fasync_helper(fd, filp, on, &wq->fasync_list);
1304
1305	if (!wq->fasync_list)
1306	sock_reset_flag(sk, SOCK_FASYNC);
1307	else
1308	sock_set_flag(sk, SOCK_FASYNC);
1309
1310	release_sock(sk);
1311	return 0;
1312	}
1313
1314	/* This function may be called only under rcu_lock */
1315
1316	int sock_wake_async(struct socket_wq *wq, int how, int band)
1317	{
1318	if (!wq \|\| !wq->fasync_list)
1319	return -1;
1320
1321	switch (how) {
1322	case SOCK_WAKE_WAITD:
1323	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1324	break;
1325	goto call_kill;
1326	case SOCK_WAKE_SPACE:
1327	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1328	break;
1329	/* fall through */
1330	case SOCK_WAKE_IO:
1331	call_kill:
1332	kill_fasync(&wq->fasync_list, SIGIO, band);
1333	break;
1334	case SOCK_WAKE_URG:
1335	kill_fasync(&wq->fasync_list, SIGURG, band);
1336	}
1337
1338	return 0;
1339	}
1340	EXPORT_SYMBOL(sock_wake_async);
1341
1342	/**
1343	* __sock_create - creates a socket
1344	* @net: net namespace
1345	* @family: protocol family (AF_INET, ...)
1346	* @type: communication type (SOCK_STREAM, ...)
1347	* @protocol: protocol (0, ...)
1348	* @res: new socket
1349	* @kern: boolean for kernel space sockets
1350	*
1351	* Creates a new socket and assigns it to @res, passing through LSM.
1352	* Returns 0 or an error. On failure @res is set to %NULL. @kern must
1353	* be set to true if the socket resides in kernel space.
1354	* This function internally uses GFP_KERNEL.
1355	*/
1356
1357	int __sock_create(struct net *net, int family, int type, int protocol,
1358	struct socket **res, int kern)
1359	{
1360	int err;
1361	struct socket *sock;
1362	const struct net_proto_family *pf;
1363
1364	/*
1365	* Check protocol is in range
1366	*/
1367	if (family < 0 \|\| family >= NPROTO)
1368	return -EAFNOSUPPORT;
1369	if (type < 0 \|\| type >= SOCK_MAX)
1370	return -EINVAL;
1371
1372	/* Compatibility.
1373
1374	This uglymoron is moved from INET layer to here to avoid
1375	deadlock in module load.
1376	*/
1377	if (family == PF_INET && type == SOCK_PACKET) {
1378	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1379	current->comm);
1380	family = PF_PACKET;
1381	}
1382
1383	err = security_socket_create(family, type, protocol, kern);
1384	if (err)
1385	return err;
1386
1387	/*
1388	* Allocate the socket and allow the family to set things up. if
1389	* the protocol is 0, the family is instructed to select an appropriate
1390	* default.
1391	*/
1392	sock = sock_alloc();
1393	if (!sock) {
1394	net_warn_ratelimited("socket: no more sockets\n");
1395	return -ENFILE; /* Not exactly a match, but its the
1396	closest posix thing */
1397	}
1398
1399	sock->type = type;
1400
1401	#ifdef CONFIG_MODULES
1402	/* Attempt to load a protocol module if the find failed.
1403	*
1404	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1405	* requested real, full-featured networking support upon configuration.
1406	* Otherwise module support will break!
1407	*/
1408	if (rcu_access_pointer(net_families[family]) == NULL)
1409	request_module("net-pf-%d", family);
1410	#endif
1411
1412	rcu_read_lock();
1413	pf = rcu_dereference(net_families[family]);
1414	err = -EAFNOSUPPORT;
1415	if (!pf)
1416	goto out_release;
1417
1418	/*
1419	* We will call the ->create function, that possibly is in a loadable
1420	* module, so we have to bump that loadable module refcnt first.
1421	*/
1422	if (!try_module_get(pf->owner))
1423	goto out_release;
1424
1425	/* Now protected by module ref count */
1426	rcu_read_unlock();
1427
1428	err = pf->create(net, sock, protocol, kern);
1429	if (err < 0)
1430	goto out_module_put;
1431
1432	/*
1433	* Now to bump the refcnt of the [loadable] module that owns this
1434	* socket at sock_release time we decrement its refcnt.
1435	*/
1436	if (!try_module_get(sock->ops->owner))
1437	goto out_module_busy;
1438
1439	/*
1440	* Now that we're done with the ->create function, the [loadable]
1441	* module can have its refcnt decremented
1442	*/
1443	module_put(pf->owner);
1444	err = security_socket_post_create(sock, family, type, protocol, kern);
1445	if (err)
1446	goto out_sock_release;
1447	*res = sock;
1448
1449	return 0;
1450
1451	out_module_busy:
1452	err = -EAFNOSUPPORT;
1453	out_module_put:
1454	sock->ops = NULL;
1455	module_put(pf->owner);
1456	out_sock_release:
1457	sock_release(sock);
1458	return err;
1459
1460	out_release:
1461	rcu_read_unlock();
1462	goto out_sock_release;
1463	}
1464	EXPORT_SYMBOL(__sock_create);
1465
1466	/**
1467	* sock_create - creates a socket
1468	* @family: protocol family (AF_INET, ...)
1469	* @type: communication type (SOCK_STREAM, ...)
1470	* @protocol: protocol (0, ...)
1471	* @res: new socket
1472	*
1473	* A wrapper around __sock_create().
1474	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1475	*/
1476
1477	int sock_create(int family, int type, int protocol, struct socket **res)
1478	{
1479	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1480	}
1481	EXPORT_SYMBOL(sock_create);
1482
1483	/**
1484	* sock_create_kern - creates a socket (kernel space)
1485	* @net: net namespace
1486	* @family: protocol family (AF_INET, ...)
1487	* @type: communication type (SOCK_STREAM, ...)
1488	* @protocol: protocol (0, ...)
1489	* @res: new socket
1490	*
1491	* A wrapper around __sock_create().
1492	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1493	*/
1494
1495	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1496	{
1497	return __sock_create(net, family, type, protocol, res, 1);
1498	}
1499	EXPORT_SYMBOL(sock_create_kern);
1500
1501	int __sys_socket(int family, int type, int protocol)
1502	{
1503	int retval;
1504	struct socket *sock;
1505	int flags;
1506
1507	/* Check the SOCK_* constants for consistency. */
1508	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1509	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1510	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1511	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1512
1513	flags = type & ~SOCK_TYPE_MASK;
1514	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1515	return -EINVAL;
1516	type &= SOCK_TYPE_MASK;
1517
1518	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1519	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1520
1521	retval = sock_create(family, type, protocol, &sock);
1522	if (retval < 0)
1523	return retval;
1524
1525	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1526	}
1527
1528	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1529	{
1530	return __sys_socket(family, type, protocol);
1531	}
1532
1533	/*
1534	* Create a pair of connected sockets.
1535	*/
1536
1537	int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1538	{
1539	struct socket sock1, sock2;
1540	int fd1, fd2, err;
1541	struct file newfile1, newfile2;
1542	int flags;
1543
1544	flags = type & ~SOCK_TYPE_MASK;
1545	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1546	return -EINVAL;
1547	type &= SOCK_TYPE_MASK;
1548
1549	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1550	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1551
1552	/*
1553	* reserve descriptors and make sure we won't fail
1554	* to return them to userland.
1555	*/
1556	fd1 = get_unused_fd_flags(flags);
1557	if (unlikely(fd1 < 0))
1558	return fd1;
1559
1560	fd2 = get_unused_fd_flags(flags);
1561	if (unlikely(fd2 < 0)) {
1562	put_unused_fd(fd1);
1563	return fd2;
1564	}
1565
1566	err = put_user(fd1, &usockvec[0]);
1567	if (err)
1568	goto out;
1569
1570	err = put_user(fd2, &usockvec[1]);
1571	if (err)
1572	goto out;
1573
1574	/*
1575	* Obtain the first socket and check if the underlying protocol
1576	* supports the socketpair call.
1577	*/
1578
1579	err = sock_create(family, type, protocol, &sock1);
1580	if (unlikely(err < 0))
1581	goto out;
1582
1583	err = sock_create(family, type, protocol, &sock2);
1584	if (unlikely(err < 0)) {
1585	sock_release(sock1);
1586	goto out;
1587	}
1588
1589	err = security_socket_socketpair(sock1, sock2);
1590	if (unlikely(err)) {
1591	sock_release(sock2);
1592	sock_release(sock1);
1593	goto out;
1594	}
1595
1596	err = sock1->ops->socketpair(sock1, sock2);
1597	if (unlikely(err < 0)) {
1598	sock_release(sock2);
1599	sock_release(sock1);
1600	goto out;
1601	}
1602
1603	newfile1 = sock_alloc_file(sock1, flags, NULL);
1604	if (IS_ERR(newfile1)) {
1605	err = PTR_ERR(newfile1);
1606	sock_release(sock2);
1607	goto out;
1608	}
1609
1610	newfile2 = sock_alloc_file(sock2, flags, NULL);
1611	if (IS_ERR(newfile2)) {
1612	err = PTR_ERR(newfile2);
1613	fput(newfile1);
1614	goto out;
1615	}
1616
1617	audit_fd_pair(fd1, fd2);
1618
1619	fd_install(fd1, newfile1);
1620	fd_install(fd2, newfile2);
1621	return 0;
1622
1623	out:
1624	put_unused_fd(fd2);
1625	put_unused_fd(fd1);
1626	return err;
1627	}
1628
1629	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1630	int __user *, usockvec)
1631	{
1632	return __sys_socketpair(family, type, protocol, usockvec);
1633	}
1634
1635	/*
1636	* Bind a name to a socket. Nothing much to do here since it's
1637	* the protocol's responsibility to handle the local address.
1638	*
1639	* We move the socket address to kernel space before we call
1640	* the protocol layer (having also checked the address is ok).
1641	*/
1642
1643	int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1644	{
1645	struct socket *sock;
1646	struct sockaddr_storage address;
1647	int err, fput_needed;
1648
1649	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1650	if (sock) {
1651	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1652	if (!err) {
1653	err = security_socket_bind(sock,
1654	(struct sockaddr *)&address,
1655	addrlen);
1656	if (!err)
1657	err = sock->ops->bind(sock,
1658	(struct sockaddr *)
1659	&address, addrlen);
1660	}
1661	fput_light(sock->file, fput_needed);
1662	}
1663	return err;
1664	}
1665
1666	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1667	{
1668	return __sys_bind(fd, umyaddr, addrlen);
1669	}
1670
1671	/*
1672	* Perform a listen. Basically, we allow the protocol to do anything
1673	* necessary for a listen, and if that works, we mark the socket as
1674	* ready for listening.
1675	*/
1676
1677	int __sys_listen(int fd, int backlog)
1678	{
1679	struct socket *sock;
1680	int err, fput_needed;
1681	int somaxconn;
1682
1683	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1684	if (sock) {
1685	somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1686	if ((unsigned int)backlog > somaxconn)
1687	backlog = somaxconn;
1688
1689	err = security_socket_listen(sock, backlog);
1690	if (!err)
1691	err = sock->ops->listen(sock, backlog);
1692
1693	fput_light(sock->file, fput_needed);
1694	}
1695	return err;
1696	}
1697
1698	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1699	{
1700	return __sys_listen(fd, backlog);
1701	}
1702
1703	int __sys_accept4_file(struct file *file, unsigned file_flags,
1704	struct sockaddr __user *upeer_sockaddr,
1705	int __user *upeer_addrlen, int flags,
1706	unsigned long nofile)
1707	{
1708	struct socket sock, newsock;
1709	struct file *newfile;
1710	int err, len, newfd;
1711	struct sockaddr_storage address;
1712
1713	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1714	return -EINVAL;
1715
1716	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1717	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1718
1719	sock = sock_from_file(file, &err);
1720	if (!sock)
1721	goto out;
1722
1723	err = -ENFILE;
1724	newsock = sock_alloc();
1725	if (!newsock)
1726	goto out;
1727
1728	newsock->type = sock->type;
1729	newsock->ops = sock->ops;
1730
1731	/*
1732	* We don't need try_module_get here, as the listening socket (sock)
1733	* has the protocol module (sock->ops->owner) held.
1734	*/
1735	__module_get(newsock->ops->owner);
1736
1737	newfd = __get_unused_fd_flags(flags, nofile);
1738	if (unlikely(newfd < 0)) {
1739	err = newfd;
1740	sock_release(newsock);
1741	goto out;
1742	}
1743	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1744	if (IS_ERR(newfile)) {
1745	err = PTR_ERR(newfile);
1746	put_unused_fd(newfd);
1747	goto out;
1748	}
1749
1750	err = security_socket_accept(sock, newsock);
1751	if (err)
1752	goto out_fd;
1753
1754	err = sock->ops->accept(sock, newsock, sock->file->f_flags \| file_flags,
1755	false);
1756	if (err < 0)
1757	goto out_fd;
1758
1759	if (upeer_sockaddr) {
1760	len = newsock->ops->getname(newsock,
1761	(struct sockaddr *)&address, 2);
1762	if (len < 0) {
1763	err = -ECONNABORTED;
1764	goto out_fd;
1765	}
1766	err = move_addr_to_user(&address,
1767	len, upeer_sockaddr, upeer_addrlen);
1768	if (err < 0)
1769	goto out_fd;
1770	}
1771
1772	/* File flags are not inherited via accept() unlike another OSes. */
1773
1774	fd_install(newfd, newfile);
1775	err = newfd;
1776	out:
1777	return err;
1778	out_fd:
1779	fput(newfile);
1780	put_unused_fd(newfd);
1781	goto out;
1782
1783	}
1784
1785	/*
1786	* For accept, we attempt to create a new socket, set up the link
1787	* with the client, wake up the client, then return the new
1788	* connected fd. We collect the address of the connector in kernel
1789	* space and move it to user at the very end. This is unclean because
1790	* we open the socket then return an error.
1791	*
1792	* 1003.1g adds the ability to recvmsg() to query connection pending
1793	* status to recvmsg. We need to add that support in a way thats
1794	* clean when we restructure accept also.
1795	*/
1796
1797	int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1798	int __user *upeer_addrlen, int flags)
1799	{
1800	int ret = -EBADF;
1801	struct fd f;
1802
1803	f = fdget(fd);
1804	if (f.file) {
1805	ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1806	upeer_addrlen, flags,
1807	rlimit(RLIMIT_NOFILE));
1808	if (f.flags)
1809	fput(f.file);
1810	}
1811
1812	return ret;
1813	}
1814
1815	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1816	int __user *, upeer_addrlen, int, flags)
1817	{
1818	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1819	}
1820
1821	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1822	int __user *, upeer_addrlen)
1823	{
1824	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1825	}
1826
1827	/*
1828	* Attempt to connect to a socket with the server address. The address
1829	* is in user space so we verify it is OK and move it to kernel space.
1830	*
1831	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1832	* break bindings
1833	*
1834	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1835	* other SEQPACKET protocols that take time to connect() as it doesn't
1836	* include the -EINPROGRESS status for such sockets.
1837	*/
1838
1839	int __sys_connect_file(struct file file, struct sockaddr_storage address,
1840	int addrlen, int file_flags)
1841	{
1842	struct socket *sock;
1843	int err;
1844
1845	sock = sock_from_file(file, &err);
1846	if (!sock)
1847	goto out;
1848
1849	err =
1850	security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1851	if (err)
1852	goto out;
1853
1854	err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1855	sock->file->f_flags \| file_flags);
1856	out:
1857	return err;
1858	}
1859
1860	int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1861	{
1862	int ret = -EBADF;
1863	struct fd f;
1864
1865	f = fdget(fd);
1866	if (f.file) {
1867	struct sockaddr_storage address;
1868
1869	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1870	if (!ret)
1871	ret = __sys_connect_file(f.file, &address, addrlen, 0);
1872	if (f.flags)
1873	fput(f.file);
1874	}
1875
1876	return ret;
1877	}
1878
1879	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1880	int, addrlen)
1881	{
1882	return __sys_connect(fd, uservaddr, addrlen);
1883	}
1884
1885	/*
1886	* Get the local address ('name') of a socket object. Move the obtained
1887	* name to user space.
1888	*/
1889
1890	int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1891	int __user *usockaddr_len)
1892	{
1893	struct socket *sock;
1894	struct sockaddr_storage address;
1895	int err, fput_needed;
1896
1897	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1898	if (!sock)
1899	goto out;
1900
1901	err = security_socket_getsockname(sock);
1902	if (err)
1903	goto out_put;
1904
1905	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1906	if (err < 0)
1907	goto out_put;
1908	/* "err" is actually length in this case */
1909	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1910
1911	out_put:
1912	fput_light(sock->file, fput_needed);
1913	out:
1914	return err;
1915	}
1916
1917	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1918	int __user *, usockaddr_len)
1919	{
1920	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1921	}
1922
1923	/*
1924	* Get the remote address ('name') of a socket object. Move the obtained
1925	* name to user space.
1926	*/
1927
1928	int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1929	int __user *usockaddr_len)
1930	{
1931	struct socket *sock;
1932	struct sockaddr_storage address;
1933	int err, fput_needed;
1934
1935	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1936	if (sock != NULL) {
1937	err = security_socket_getpeername(sock);
1938	if (err) {
1939	fput_light(sock->file, fput_needed);
1940	return err;
1941	}
1942
1943	err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1944	if (err >= 0)
1945	/* "err" is actually length in this case */
1946	err = move_addr_to_user(&address, err, usockaddr,
1947	usockaddr_len);
1948	fput_light(sock->file, fput_needed);
1949	}
1950	return err;
1951	}
1952
1953	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1954	int __user *, usockaddr_len)
1955	{
1956	return __sys_getpeername(fd, usockaddr, usockaddr_len);
1957	}
1958
1959	/*
1960	* Send a datagram to a given address. We move the address into kernel
1961	* space and check the user space data area is readable before invoking
1962	* the protocol.
1963	*/
1964	int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1965	struct sockaddr __user *addr, int addr_len)
1966	{
1967	struct socket *sock;
1968	struct sockaddr_storage address;
1969	int err;
1970	struct msghdr msg;
1971	struct iovec iov;
1972	int fput_needed;
1973
1974	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1975	if (unlikely(err))
1976	return err;
1977	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1978	if (!sock)
1979	goto out;
1980
1981	msg.msg_name = NULL;
1982	msg.msg_control = NULL;
1983	msg.msg_controllen = 0;
1984	msg.msg_namelen = 0;
1985	if (addr) {
1986	err = move_addr_to_kernel(addr, addr_len, &address);
1987	if (err < 0)
1988	goto out_put;
1989	msg.msg_name = (struct sockaddr *)&address;
1990	msg.msg_namelen = addr_len;
1991	}
1992	if (sock->file->f_flags & O_NONBLOCK)
1993	flags \|= MSG_DONTWAIT;
1994	msg.msg_flags = flags;
1995	err = sock_sendmsg(sock, &msg);
1996
1997	out_put:
1998	fput_light(sock->file, fput_needed);
1999	out:
2000	return err;
2001	}
2002
2003	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2004	unsigned int, flags, struct sockaddr __user *, addr,
2005	int, addr_len)
2006	{
2007	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2008	}
2009
2010	/*
2011	* Send a datagram down a socket.
2012	*/
2013
2014	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2015	unsigned int, flags)
2016	{
2017	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2018	}
2019
2020	/*
2021	* Receive a frame from the socket and optionally record the address of the
2022	* sender. We verify the buffers are writable and if needed move the
2023	* sender address from kernel to user space.
2024	*/
2025	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2026	struct sockaddr __user addr, int __user addr_len)
2027	{
2028	struct socket *sock;
2029	struct iovec iov;
2030	struct msghdr msg;
2031	struct sockaddr_storage address;
2032	int err, err2;
2033	int fput_needed;
2034
2035	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2036	if (unlikely(err))
2037	return err;
2038	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2039	if (!sock)
2040	goto out;
2041
2042	msg.msg_control = NULL;
2043	msg.msg_controllen = 0;
2044	/* Save some cycles and don't copy the address if not needed */
2045	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2046	/* We assume all kernel code knows the size of sockaddr_storage */
2047	msg.msg_namelen = 0;
2048	msg.msg_iocb = NULL;
2049	msg.msg_flags = 0;
2050	if (sock->file->f_flags & O_NONBLOCK)
2051	flags \|= MSG_DONTWAIT;
2052	err = sock_recvmsg(sock, &msg, flags);
2053
2054	if (err >= 0 && addr != NULL) {
2055	err2 = move_addr_to_user(&address,
2056	msg.msg_namelen, addr, addr_len);
2057	if (err2 < 0)
2058	err = err2;
2059	}
2060
2061	fput_light(sock->file, fput_needed);
2062	out:
2063	return err;
2064	}
2065
2066	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2067	unsigned int, flags, struct sockaddr __user *, addr,
2068	int __user *, addr_len)
2069	{
2070	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2071	}
2072
2073	/*
2074	* Receive a datagram from a socket.
2075	*/
2076
2077	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2078	unsigned int, flags)
2079	{
2080	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2081	}
2082
2083	/*
2084	* Set a socket option. Because we don't know the option lengths we have
2085	* to pass the user mode parameter for the protocols to sort out.
2086	*/
2087
2088	static int __sys_setsockopt(int fd, int level, int optname,
2089	char __user *optval, int optlen)
2090	{
2091	mm_segment_t oldfs = get_fs();
2092	char *kernel_optval = NULL;
2093	int err, fput_needed;
2094	struct socket *sock;
2095
2096	if (optlen < 0)
2097	return -EINVAL;
2098
2099	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2100	if (sock != NULL) {
2101	err = security_socket_setsockopt(sock, level, optname);
2102	if (err)
2103	goto out_put;
2104
2105	err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level,
2106	&optname, optval, &optlen,
2107	&kernel_optval);
2108
2109	if (err < 0) {
2110	goto out_put;
2111	} else if (err > 0) {
2112	err = 0;
2113	goto out_put;
2114	}
2115
2116	if (kernel_optval) {
2117	set_fs(KERNEL_DS);
2118	optval = (char __user __force *)kernel_optval;
2119	}
2120
2121	if (level == SOL_SOCKET)
2122	err =
2123	sock_setsockopt(sock, level, optname, optval,
2124	optlen);
2125	else
2126	err =
2127	sock->ops->setsockopt(sock, level, optname, optval,
2128	optlen);
2129
2130	if (kernel_optval) {
2131	set_fs(oldfs);
2132	kfree(kernel_optval);
2133	}
2134	out_put:
2135	fput_light(sock->file, fput_needed);
2136	}
2137	return err;
2138	}
2139
2140	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2141	char __user *, optval, int, optlen)
2142	{
2143	return __sys_setsockopt(fd, level, optname, optval, optlen);
2144	}
2145
2146	/*
2147	* Get a socket option. Because we don't know the option lengths we have
2148	* to pass a user mode parameter for the protocols to sort out.
2149	*/
2150
2151	static int __sys_getsockopt(int fd, int level, int optname,
2152	char __user optval, int __user optlen)
2153	{
2154	int err, fput_needed;
2155	struct socket *sock;
2156	int max_optlen;
2157
2158	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2159	if (sock != NULL) {
2160	err = security_socket_getsockopt(sock, level, optname);
2161	if (err)
2162	goto out_put;
2163
2164	max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2165
2166	if (level == SOL_SOCKET)
2167	err =
2168	sock_getsockopt(sock, level, optname, optval,
2169	optlen);
2170	else
2171	err =
2172	sock->ops->getsockopt(sock, level, optname, optval,
2173	optlen);
2174
2175	err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2176	optval, optlen,
2177	max_optlen, err);
2178	out_put:
2179	fput_light(sock->file, fput_needed);
2180	}
2181	return err;
2182	}
2183
2184	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2185	char __user , optval, int __user , optlen)
2186	{
2187	return __sys_getsockopt(fd, level, optname, optval, optlen);
2188	}
2189
2190	/*
2191	* Shutdown a socket.
2192	*/
2193
2194	int __sys_shutdown(int fd, int how)
2195	{
2196	int err, fput_needed;
2197	struct socket *sock;
2198
2199	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2200	if (sock != NULL) {
2201	err = security_socket_shutdown(sock, how);
2202	if (!err)
2203	err = sock->ops->shutdown(sock, how);
2204	fput_light(sock->file, fput_needed);
2205	}
2206	return err;
2207	}
2208
2209	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2210	{
2211	return __sys_shutdown(fd, how);
2212	}
2213
2214	/* A couple of helpful macros for getting the address of the 32/64 bit
2215	* fields which are the same type (int / unsigned) on our platforms.
2216	*/
2217	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2218	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2219	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2220
2221	struct used_address {
2222	struct sockaddr_storage name;
2223	unsigned int name_len;
2224	};
2225
2226	int __copy_msghdr_from_user(struct msghdr *kmsg,
2227	struct user_msghdr __user *umsg,
2228	struct sockaddr __user **save_addr,
2229	struct iovec __user *uiov, size_t nsegs)
2230	{
2231	struct user_msghdr msg;
2232	ssize_t err;
2233
2234	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2235	return -EFAULT;
2236
2237	kmsg->msg_control_is_user = true;
2238	kmsg->msg_control_user = msg.msg_control;
2239	kmsg->msg_controllen = msg.msg_controllen;
2240	kmsg->msg_flags = msg.msg_flags;
2241
2242	kmsg->msg_namelen = msg.msg_namelen;
2243	if (!msg.msg_name)
2244	kmsg->msg_namelen = 0;
2245
2246	if (kmsg->msg_namelen < 0)
2247	return -EINVAL;
2248
2249	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2250	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2251
2252	if (save_addr)
2253	*save_addr = msg.msg_name;
2254
2255	if (msg.msg_name && kmsg->msg_namelen) {
2256	if (!save_addr) {
2257	err = move_addr_to_kernel(msg.msg_name,
2258	kmsg->msg_namelen,
2259	kmsg->msg_name);
2260	if (err < 0)
2261	return err;
2262	}
2263	} else {
2264	kmsg->msg_name = NULL;
2265	kmsg->msg_namelen = 0;
2266	}
2267
2268	if (msg.msg_iovlen > UIO_MAXIOV)
2269	return -EMSGSIZE;
2270
2271	kmsg->msg_iocb = NULL;
2272	*uiov = msg.msg_iov;
2273	*nsegs = msg.msg_iovlen;
2274	return 0;
2275	}
2276
2277	static int copy_msghdr_from_user(struct msghdr *kmsg,
2278	struct user_msghdr __user *umsg,
2279	struct sockaddr __user **save_addr,
2280	struct iovec **iov)
2281	{
2282	struct user_msghdr msg;
2283	ssize_t err;
2284
2285	err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2286	&msg.msg_iovlen);
2287	if (err)
2288	return err;
2289
2290	err = import_iovec(save_addr ? READ : WRITE,
2291	msg.msg_iov, msg.msg_iovlen,
2292	UIO_FASTIOV, iov, &kmsg->msg_iter);
2293	return err < 0 ? err : 0;
2294	}
2295
2296	static int ____sys_sendmsg(struct socket sock, struct msghdr msg_sys,
2297	unsigned int flags, struct used_address *used_address,
2298	unsigned int allowed_msghdr_flags)
2299	{
2300	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2301	__aligned(sizeof(__kernel_size_t));
2302	/* 20 is size of ipv6_pktinfo */
2303	unsigned char *ctl_buf = ctl;
2304	int ctl_len;
2305	ssize_t err;
2306
2307	err = -ENOBUFS;
2308
2309	if (msg_sys->msg_controllen > INT_MAX)
2310	goto out;
2311	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2312	ctl_len = msg_sys->msg_controllen;
2313	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2314	err =
2315	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2316	sizeof(ctl));
2317	if (err)
2318	goto out;
2319	ctl_buf = msg_sys->msg_control;
2320	ctl_len = msg_sys->msg_controllen;
2321	} else if (ctl_len) {
2322	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2323	CMSG_ALIGN(sizeof(struct cmsghdr)));
2324	if (ctl_len > sizeof(ctl)) {
2325	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2326	if (ctl_buf == NULL)
2327	goto out;
2328	}
2329	err = -EFAULT;
2330	if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2331	goto out_freectl;
2332	msg_sys->msg_control = ctl_buf;
2333	msg_sys->msg_control_is_user = false;
2334	}
2335	msg_sys->msg_flags = flags;
2336
2337	if (sock->file->f_flags & O_NONBLOCK)
2338	msg_sys->msg_flags \|= MSG_DONTWAIT;
2339	/*
2340	* If this is sendmmsg() and current destination address is same as
2341	* previously succeeded address, omit asking LSM's decision.
2342	* used_address->name_len is initialized to UINT_MAX so that the first
2343	* destination address never matches.
2344	*/
2345	if (used_address && msg_sys->msg_name &&
2346	used_address->name_len == msg_sys->msg_namelen &&
2347	!memcmp(&used_address->name, msg_sys->msg_name,
2348	used_address->name_len)) {
2349	err = sock_sendmsg_nosec(sock, msg_sys);
2350	goto out_freectl;
2351	}
2352	err = sock_sendmsg(sock, msg_sys);
2353	/*
2354	* If this is sendmmsg() and sending to current destination address was
2355	* successful, remember it.
2356	*/
2357	if (used_address && err >= 0) {
2358	used_address->name_len = msg_sys->msg_namelen;
2359	if (msg_sys->msg_name)
2360	memcpy(&used_address->name, msg_sys->msg_name,
2361	used_address->name_len);
2362	}
2363
2364	out_freectl:
2365	if (ctl_buf != ctl)
2366	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2367	out:
2368	return err;
2369	}
2370
2371	int sendmsg_copy_msghdr(struct msghdr *msg,
2372	struct user_msghdr __user *umsg, unsigned flags,
2373	struct iovec **iov)
2374	{
2375	int err;
2376
2377	if (flags & MSG_CMSG_COMPAT) {
2378	struct compat_msghdr __user *msg_compat;
2379
2380	msg_compat = (struct compat_msghdr __user *) umsg;
2381	err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2382	} else {
2383	err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2384	}
2385	if (err < 0)
2386	return err;
2387
2388	return 0;
2389	}
2390
2391	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
2392	struct msghdr *msg_sys, unsigned int flags,
2393	struct used_address *used_address,
2394	unsigned int allowed_msghdr_flags)
2395	{
2396	struct sockaddr_storage address;
2397	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2398	ssize_t err;
2399
2400	msg_sys->msg_name = &address;
2401
2402	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2403	if (err < 0)
2404	return err;
2405
2406	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2407	allowed_msghdr_flags);
2408	kfree(iov);
2409	return err;
2410	}
2411
2412	/*
2413	* BSD sendmsg interface
2414	*/
2415	long __sys_sendmsg_sock(struct socket sock, struct msghdr msg,
2416	unsigned int flags)
2417	{
2418	/* disallow ancillary data requests from this path */
2419	if (msg->msg_control \|\| msg->msg_controllen)
2420	return -EINVAL;
2421
2422	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2423	}
2424
2425	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2426	bool forbid_cmsg_compat)
2427	{
2428	int fput_needed, err;
2429	struct msghdr msg_sys;
2430	struct socket *sock;
2431
2432	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2433	return -EINVAL;
2434
2435	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2436	if (!sock)
2437	goto out;
2438
2439	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2440
2441	fput_light(sock->file, fput_needed);
2442	out:
2443	return err;
2444	}
2445
2446	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2447	{
2448	return __sys_sendmsg(fd, msg, flags, true);
2449	}
2450
2451	/*
2452	* Linux sendmmsg interface
2453	*/
2454
2455	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2456	unsigned int flags, bool forbid_cmsg_compat)
2457	{
2458	int fput_needed, err, datagrams;
2459	struct socket *sock;
2460	struct mmsghdr __user *entry;
2461	struct compat_mmsghdr __user *compat_entry;
2462	struct msghdr msg_sys;
2463	struct used_address used_address;
2464	unsigned int oflags = flags;
2465
2466	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2467	return -EINVAL;
2468
2469	if (vlen > UIO_MAXIOV)
2470	vlen = UIO_MAXIOV;
2471
2472	datagrams = 0;
2473
2474	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2475	if (!sock)
2476	return err;
2477
2478	used_address.name_len = UINT_MAX;
2479	entry = mmsg;
2480	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2481	err = 0;
2482	flags \|= MSG_BATCH;
2483
2484	while (datagrams < vlen) {
2485	if (datagrams == vlen - 1)
2486	flags = oflags;
2487
2488	if (MSG_CMSG_COMPAT & flags) {
2489	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2490	&msg_sys, flags, &used_address, MSG_EOR);
2491	if (err < 0)
2492	break;
2493	err = __put_user(err, &compat_entry->msg_len);
2494	++compat_entry;
2495	} else {
2496	err = ___sys_sendmsg(sock,
2497	(struct user_msghdr __user *)entry,
2498	&msg_sys, flags, &used_address, MSG_EOR);
2499	if (err < 0)
2500	break;
2501	err = put_user(err, &entry->msg_len);
2502	++entry;
2503	}
2504
2505	if (err)
2506	break;
2507	++datagrams;
2508	if (msg_data_left(&msg_sys))
2509	break;
2510	cond_resched();
2511	}
2512
2513	fput_light(sock->file, fput_needed);
2514
2515	/* We only return an error if no datagrams were able to be sent */
2516	if (datagrams != 0)
2517	return datagrams;
2518
2519	return err;
2520	}
2521
2522	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2523	unsigned int, vlen, unsigned int, flags)
2524	{
2525	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2526	}
2527
2528	int recvmsg_copy_msghdr(struct msghdr *msg,
2529	struct user_msghdr __user *umsg, unsigned flags,
2530	struct sockaddr __user **uaddr,
2531	struct iovec **iov)
2532	{
2533	ssize_t err;
2534
2535	if (MSG_CMSG_COMPAT & flags) {
2536	struct compat_msghdr __user *msg_compat;
2537
2538	msg_compat = (struct compat_msghdr __user *) umsg;
2539	err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2540	} else {
2541	err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2542	}
2543	if (err < 0)
2544	return err;
2545
2546	return 0;
2547	}
2548
2549	static int ____sys_recvmsg(struct socket sock, struct msghdr msg_sys,
2550	struct user_msghdr __user *msg,
2551	struct sockaddr __user *uaddr,
2552	unsigned int flags, int nosec)
2553	{
2554	struct compat_msghdr __user *msg_compat =
2555	(struct compat_msghdr __user *) msg;
2556	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2557	struct sockaddr_storage addr;
2558	unsigned long cmsg_ptr;
2559	int len;
2560	ssize_t err;
2561
2562	msg_sys->msg_name = &addr;
2563	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2564	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2565
2566	/* We assume all kernel code knows the size of sockaddr_storage */
2567	msg_sys->msg_namelen = 0;
2568
2569	if (sock->file->f_flags & O_NONBLOCK)
2570	flags \|= MSG_DONTWAIT;
2571
2572	if (unlikely(nosec))
2573	err = sock_recvmsg_nosec(sock, msg_sys, flags);
2574	else
2575	err = sock_recvmsg(sock, msg_sys, flags);
2576
2577	if (err < 0)
2578	goto out;
2579	len = err;
2580
2581	if (uaddr != NULL) {
2582	err = move_addr_to_user(&addr,
2583	msg_sys->msg_namelen, uaddr,
2584	uaddr_len);
2585	if (err < 0)
2586	goto out;
2587	}
2588	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2589	COMPAT_FLAGS(msg));
2590	if (err)
2591	goto out;
2592	if (MSG_CMSG_COMPAT & flags)
2593	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2594	&msg_compat->msg_controllen);
2595	else
2596	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2597	&msg->msg_controllen);
2598	if (err)
2599	goto out;
2600	err = len;
2601	out:
2602	return err;
2603	}
2604
2605	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2606	struct msghdr *msg_sys, unsigned int flags, int nosec)
2607	{
2608	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2609	/* user mode address pointers */
2610	struct sockaddr __user *uaddr;
2611	ssize_t err;
2612
2613	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2614	if (err < 0)
2615	return err;
2616
2617	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2618	kfree(iov);
2619	return err;
2620	}
2621
2622	/*
2623	* BSD recvmsg interface
2624	*/
2625
2626	long __sys_recvmsg_sock(struct socket sock, struct msghdr msg,
2627	struct user_msghdr __user *umsg,
2628	struct sockaddr __user *uaddr, unsigned int flags)
2629	{
2630	/* disallow ancillary data requests from this path */
2631	if (msg->msg_control \|\| msg->msg_controllen)
2632	return -EINVAL;
2633
2634	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2635	}
2636
2637	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2638	bool forbid_cmsg_compat)
2639	{
2640	int fput_needed, err;
2641	struct msghdr msg_sys;
2642	struct socket *sock;
2643
2644	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2645	return -EINVAL;
2646
2647	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2648	if (!sock)
2649	goto out;
2650
2651	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2652
2653	fput_light(sock->file, fput_needed);
2654	out:
2655	return err;
2656	}
2657
2658	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2659	unsigned int, flags)
2660	{
2661	return __sys_recvmsg(fd, msg, flags, true);
2662	}
2663
2664	/*
2665	* Linux recvmmsg interface
2666	*/
2667
2668	static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2669	unsigned int vlen, unsigned int flags,
2670	struct timespec64 *timeout)
2671	{
2672	int fput_needed, err, datagrams;
2673	struct socket *sock;
2674	struct mmsghdr __user *entry;
2675	struct compat_mmsghdr __user *compat_entry;
2676	struct msghdr msg_sys;
2677	struct timespec64 end_time;
2678	struct timespec64 timeout64;
2679
2680	if (timeout &&
2681	poll_select_set_timeout(&end_time, timeout->tv_sec,
2682	timeout->tv_nsec))
2683	return -EINVAL;
2684
2685	datagrams = 0;
2686
2687	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2688	if (!sock)
2689	return err;
2690
2691	if (likely(!(flags & MSG_ERRQUEUE))) {
2692	err = sock_error(sock->sk);
2693	if (err) {
2694	datagrams = err;
2695	goto out_put;
2696	}
2697	}
2698
2699	entry = mmsg;
2700	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2701
2702	while (datagrams < vlen) {
2703	/*
2704	* No need to ask LSM for more than the first datagram.
2705	*/
2706	if (MSG_CMSG_COMPAT & flags) {
2707	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2708	&msg_sys, flags & ~MSG_WAITFORONE,
2709	datagrams);
2710	if (err < 0)
2711	break;
2712	err = __put_user(err, &compat_entry->msg_len);
2713	++compat_entry;
2714	} else {
2715	err = ___sys_recvmsg(sock,
2716	(struct user_msghdr __user *)entry,
2717	&msg_sys, flags & ~MSG_WAITFORONE,
2718	datagrams);
2719	if (err < 0)
2720	break;
2721	err = put_user(err, &entry->msg_len);
2722	++entry;
2723	}
2724
2725	if (err)
2726	break;
2727	++datagrams;
2728
2729	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2730	if (flags & MSG_WAITFORONE)
2731	flags \|= MSG_DONTWAIT;
2732
2733	if (timeout) {
2734	ktime_get_ts64(&timeout64);
2735	*timeout = timespec64_sub(end_time, timeout64);
2736	if (timeout->tv_sec < 0) {
2737	timeout->tv_sec = timeout->tv_nsec = 0;
2738	break;
2739	}
2740
2741	/* Timeout, return less than vlen datagrams */
2742	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2743	break;
2744	}
2745
2746	/* Out of band data, return right away */
2747	if (msg_sys.msg_flags & MSG_OOB)
2748	break;
2749	cond_resched();
2750	}
2751
2752	if (err == 0)
2753	goto out_put;
2754
2755	if (datagrams == 0) {
2756	datagrams = err;
2757	goto out_put;
2758	}
2759
2760	/*
2761	* We may return less entries than requested (vlen) if the
2762	* sock is non block and there aren't enough datagrams...
2763	*/
2764	if (err != -EAGAIN) {
2765	/*
2766	* ... or if recvmsg returns an error after we
2767	* received some datagrams, where we record the
2768	* error to return on the next call or if the
2769	* app asks about it using getsockopt(SO_ERROR).
2770	*/
2771	sock->sk->sk_err = -err;
2772	}
2773	out_put:
2774	fput_light(sock->file, fput_needed);
2775
2776	return datagrams;
2777	}
2778
2779	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2780	unsigned int vlen, unsigned int flags,
2781	struct __kernel_timespec __user *timeout,
2782	struct old_timespec32 __user *timeout32)
2783	{
2784	int datagrams;
2785	struct timespec64 timeout_sys;
2786
2787	if (timeout && get_timespec64(&timeout_sys, timeout))
2788	return -EFAULT;
2789
2790	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2791	return -EFAULT;
2792
2793	if (!timeout && !timeout32)
2794	return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2795
2796	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2797
2798	if (datagrams <= 0)
2799	return datagrams;
2800
2801	if (timeout && put_timespec64(&timeout_sys, timeout))
2802	datagrams = -EFAULT;
2803
2804	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2805	datagrams = -EFAULT;
2806
2807	return datagrams;
2808	}
2809
2810	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2811	unsigned int, vlen, unsigned int, flags,
2812	struct __kernel_timespec __user *, timeout)
2813	{
2814	if (flags & MSG_CMSG_COMPAT)
2815	return -EINVAL;
2816
2817	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2818	}
2819
2820	#ifdef CONFIG_COMPAT_32BIT_TIME
2821	SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2822	unsigned int, vlen, unsigned int, flags,
2823	struct old_timespec32 __user *, timeout)
2824	{
2825	if (flags & MSG_CMSG_COMPAT)
2826	return -EINVAL;
2827
2828	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2829	}
2830	#endif
2831
2832	#ifdef __ARCH_WANT_SYS_SOCKETCALL
2833	/* Argument list sizes for sys_socketcall */
2834	#define AL(x) ((x) * sizeof(unsigned long))
2835	static const unsigned char nargs[21] = {
2836	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2837	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2838	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2839	AL(4), AL(5), AL(4)
2840	};
2841
2842	#undef AL
2843
2844	/*
2845	* System call vectors.
2846	*
2847	* Argument checking cleaned up. Saved 20% in size.
2848	* This function doesn't need to set the kernel lock because
2849	* it is set by the callees.
2850	*/
2851
2852	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2853	{
2854	unsigned long a[AUDITSC_ARGS];
2855	unsigned long a0, a1;
2856	int err;
2857	unsigned int len;
2858
2859	if (call < 1 \|\| call > SYS_SENDMMSG)
2860	return -EINVAL;
2861	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2862
2863	len = nargs[call];
2864	if (len > sizeof(a))
2865	return -EINVAL;
2866
2867	/* copy_from_user should be SMP safe. */
2868	if (copy_from_user(a, args, len))
2869	return -EFAULT;
2870
2871	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2872	if (err)
2873	return err;
2874
2875	a0 = a[0];
2876	a1 = a[1];
2877
2878	switch (call) {
2879	case SYS_SOCKET:
2880	err = __sys_socket(a0, a1, a[2]);
2881	break;
2882	case SYS_BIND:
2883	err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2884	break;
2885	case SYS_CONNECT:
2886	err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2887	break;
2888	case SYS_LISTEN:
2889	err = __sys_listen(a0, a1);
2890	break;
2891	case SYS_ACCEPT:
2892	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2893	(int __user *)a[2], 0);
2894	break;
2895	case SYS_GETSOCKNAME:
2896	err =
2897	__sys_getsockname(a0, (struct sockaddr __user *)a1,
2898	(int __user *)a[2]);
2899	break;
2900	case SYS_GETPEERNAME:
2901	err =
2902	__sys_getpeername(a0, (struct sockaddr __user *)a1,
2903	(int __user *)a[2]);
2904	break;
2905	case SYS_SOCKETPAIR:
2906	err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2907	break;
2908	case SYS_SEND:
2909	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2910	NULL, 0);
2911	break;
2912	case SYS_SENDTO:
2913	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2914	(struct sockaddr __user *)a[4], a[5]);
2915	break;
2916	case SYS_RECV:
2917	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2918	NULL, NULL);
2919	break;
2920	case SYS_RECVFROM:
2921	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2922	(struct sockaddr __user *)a[4],
2923	(int __user *)a[5]);
2924	break;
2925	case SYS_SHUTDOWN:
2926	err = __sys_shutdown(a0, a1);
2927	break;
2928	case SYS_SETSOCKOPT:
2929	err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2930	a[4]);
2931	break;
2932	case SYS_GETSOCKOPT:
2933	err =
2934	__sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2935	(int __user *)a[4]);
2936	break;
2937	case SYS_SENDMSG:
2938	err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2939	a[2], true);
2940	break;
2941	case SYS_SENDMMSG:
2942	err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2943	a[3], true);
2944	break;
2945	case SYS_RECVMSG:
2946	err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2947	a[2], true);
2948	break;
2949	case SYS_RECVMMSG:
2950	if (IS_ENABLED(CONFIG_64BIT))
2951	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2952	a[2], a[3],
2953	(struct __kernel_timespec __user *)a[4],
2954	NULL);
2955	else
2956	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2957	a[2], a[3], NULL,
2958	(struct old_timespec32 __user *)a[4]);
2959	break;
2960	case SYS_ACCEPT4:
2961	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2962	(int __user *)a[2], a[3]);
2963	break;
2964	default:
2965	err = -EINVAL;
2966	break;
2967	}
2968	return err;
2969	}
2970
2971	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2972
2973	/**
2974	* sock_register - add a socket protocol handler
2975	* @ops: description of protocol
2976	*
2977	* This function is called by a protocol handler that wants to
2978	* advertise its address family, and have it linked into the
2979	* socket interface. The value ops->family corresponds to the
2980	* socket system call protocol family.
2981	*/
2982	int sock_register(const struct net_proto_family *ops)
2983	{
2984	int err;
2985
2986	if (ops->family >= NPROTO) {
2987	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2988	return -ENOBUFS;
2989	}
2990
2991	spin_lock(&net_family_lock);
2992	if (rcu_dereference_protected(net_families[ops->family],
2993	lockdep_is_held(&net_family_lock)))
2994	err = -EEXIST;
2995	else {
2996	rcu_assign_pointer(net_families[ops->family], ops);
2997	err = 0;
2998	}
2999	spin_unlock(&net_family_lock);
3000
3001	pr_info("NET: Registered protocol family %d\n", ops->family);
3002	return err;
3003	}
3004	EXPORT_SYMBOL(sock_register);
3005
3006	/**
3007	* sock_unregister - remove a protocol handler
3008	* @family: protocol family to remove
3009	*
3010	* This function is called by a protocol handler that wants to
3011	* remove its address family, and have it unlinked from the
3012	* new socket creation.
3013	*
3014	* If protocol handler is a module, then it can use module reference
3015	* counts to protect against new references. If protocol handler is not
3016	* a module then it needs to provide its own protection in
3017	* the ops->create routine.
3018	*/
3019	void sock_unregister(int family)
3020	{
3021	BUG_ON(family < 0 \|\| family >= NPROTO);
3022
3023	spin_lock(&net_family_lock);
3024	RCU_INIT_POINTER(net_families[family], NULL);
3025	spin_unlock(&net_family_lock);
3026
3027	synchronize_rcu();
3028
3029	pr_info("NET: Unregistered protocol family %d\n", family);
3030	}
3031	EXPORT_SYMBOL(sock_unregister);
3032
3033	bool sock_is_registered(int family)
3034	{
3035	return family < NPROTO && rcu_access_pointer(net_families[family]);
3036	}
3037
3038	static int __init sock_init(void)
3039	{
3040	int err;
3041	/*
3042	* Initialize the network sysctl infrastructure.
3043	*/
3044	err = net_sysctl_init();
3045	if (err)
3046	goto out;
3047
3048	/*
3049	* Initialize skbuff SLAB cache
3050	*/
3051	skb_init();
3052
3053	/*
3054	* Initialize the protocols module.
3055	*/
3056
3057	init_inodecache();
3058
3059	err = register_filesystem(&sock_fs_type);
3060	if (err)
3061	goto out_fs;
3062	sock_mnt = kern_mount(&sock_fs_type);
3063	if (IS_ERR(sock_mnt)) {
3064	err = PTR_ERR(sock_mnt);
3065	goto out_mount;
3066	}
3067
3068	/* The real protocol initialization is performed in later initcalls.
3069	*/
3070
3071	#ifdef CONFIG_NETFILTER
3072	err = netfilter_init();
3073	if (err)
3074	goto out;
3075	#endif
3076
3077	ptp_classifier_init();
3078
3079	out:
3080	return err;
3081
3082	out_mount:
3083	unregister_filesystem(&sock_fs_type);
3084	out_fs:
3085	goto out;
3086	}
3087
3088	core_initcall(sock_init); /* early initcall */
3089
3090	#ifdef CONFIG_PROC_FS
3091	void socket_seq_show(struct seq_file *seq)
3092	{
3093	seq_printf(seq, "sockets: used %d\n",
3094	sock_inuse_get(seq->private));
3095	}
3096	#endif /* CONFIG_PROC_FS */
3097
3098	#ifdef CONFIG_COMPAT
3099	static int compat_dev_ifconf(struct net net, struct compat_ifconf __user uifc32)
3100	{
3101	struct compat_ifconf ifc32;
3102	struct ifconf ifc;
3103	int err;
3104
3105	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3106	return -EFAULT;
3107
3108	ifc.ifc_len = ifc32.ifc_len;
3109	ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3110
3111	rtnl_lock();
3112	err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3113	rtnl_unlock();
3114	if (err)
3115	return err;
3116
3117	ifc32.ifc_len = ifc.ifc_len;
3118	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3119	return -EFAULT;
3120
3121	return 0;
3122	}
3123
3124	static int ethtool_ioctl(struct net net, struct compat_ifreq __user ifr32)
3125	{
3126	struct compat_ethtool_rxnfc __user *compat_rxnfc;
3127	bool convert_in = false, convert_out = false;
3128	size_t buf_size = 0;
3129	struct ethtool_rxnfc __user *rxnfc = NULL;
3130	struct ifreq ifr;
3131	u32 rule_cnt = 0, actual_rule_cnt;
3132	u32 ethcmd;
3133	u32 data;
3134	int ret;
3135
3136	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3137	return -EFAULT;
3138
3139	compat_rxnfc = compat_ptr(data);
3140
3141	if (get_user(ethcmd, &compat_rxnfc->cmd))
3142	return -EFAULT;
3143
3144	/* Most ethtool structures are defined without padding.
3145	* Unfortunately struct ethtool_rxnfc is an exception.
3146	*/
3147	switch (ethcmd) {
3148	default:
3149	break;
3150	case ETHTOOL_GRXCLSRLALL:
3151	/* Buffer size is variable */
3152	if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3153	return -EFAULT;
3154	if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3155	return -ENOMEM;
3156	buf_size += rule_cnt * sizeof(u32);
3157	/* fall through */
3158	case ETHTOOL_GRXRINGS:
3159	case ETHTOOL_GRXCLSRLCNT:
3160	case ETHTOOL_GRXCLSRULE:
3161	case ETHTOOL_SRXCLSRLINS:
3162	convert_out = true;
3163	/* fall through */
3164	case ETHTOOL_SRXCLSRLDEL:
3165	buf_size += sizeof(struct ethtool_rxnfc);
3166	convert_in = true;
3167	rxnfc = compat_alloc_user_space(buf_size);
3168	break;
3169	}
3170
3171	if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3172	return -EFAULT;
3173
3174	ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3175
3176	if (convert_in) {
3177	/* We expect there to be holes between fs.m_ext and
3178	* fs.ring_cookie and at the end of fs, but nowhere else.
3179	*/
3180	BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3181	sizeof(compat_rxnfc->fs.m_ext) !=
3182	offsetof(struct ethtool_rxnfc, fs.m_ext) +
3183	sizeof(rxnfc->fs.m_ext));
3184	BUILD_BUG_ON(
3185	offsetof(struct compat_ethtool_rxnfc, fs.location) -
3186	offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3187	offsetof(struct ethtool_rxnfc, fs.location) -
3188	offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3189
3190	if (copy_in_user(rxnfc, compat_rxnfc,
3191	(void __user *)(&rxnfc->fs.m_ext + 1) -
3192	(void __user *)rxnfc) \|\|
3193	copy_in_user(&rxnfc->fs.ring_cookie,
3194	&compat_rxnfc->fs.ring_cookie,
3195	(void __user *)(&rxnfc->fs.location + 1) -
3196	(void __user *)&rxnfc->fs.ring_cookie))
3197	return -EFAULT;
3198	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3199	if (put_user(rule_cnt, &rxnfc->rule_cnt))
3200	return -EFAULT;
3201	} else if (copy_in_user(&rxnfc->rule_cnt,
3202	&compat_rxnfc->rule_cnt,
3203	sizeof(rxnfc->rule_cnt)))
3204	return -EFAULT;
3205	}
3206
3207	ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3208	if (ret)
3209	return ret;
3210
3211	if (convert_out) {
3212	if (copy_in_user(compat_rxnfc, rxnfc,
3213	(const void __user *)(&rxnfc->fs.m_ext + 1) -
3214	(const void __user *)rxnfc) \|\|
3215	copy_in_user(&compat_rxnfc->fs.ring_cookie,
3216	&rxnfc->fs.ring_cookie,
3217	(const void __user *)(&rxnfc->fs.location + 1) -
3218	(const void __user *)&rxnfc->fs.ring_cookie) \|\|
3219	copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3220	sizeof(rxnfc->rule_cnt)))
3221	return -EFAULT;
3222
3223	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3224	/* As an optimisation, we only copy the actual
3225	* number of rules that the underlying
3226	* function returned. Since Mallory might
3227	* change the rule count in user memory, we
3228	* check that it is less than the rule count
3229	* originally given (as the user buffer size),
3230	* which has been range-checked.
3231	*/
3232	if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3233	return -EFAULT;
3234	if (actual_rule_cnt < rule_cnt)
3235	rule_cnt = actual_rule_cnt;
3236	if (copy_in_user(&compat_rxnfc->rule_locs[0],
3237	&rxnfc->rule_locs[0],
3238	rule_cnt * sizeof(u32)))
3239	return -EFAULT;
3240	}
3241	}
3242
3243	return 0;
3244	}
3245
3246	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
3247	{
3248	compat_uptr_t uptr32;
3249	struct ifreq ifr;
3250	void __user *saved;
3251	int err;
3252
3253	if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3254	return -EFAULT;
3255
3256	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3257	return -EFAULT;
3258
3259	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3260	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3261
3262	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3263	if (!err) {
3264	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3265	if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3266	err = -EFAULT;
3267	}
3268	return err;
3269	}
3270
3271	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3272	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3273	struct compat_ifreq __user *u_ifreq32)
3274	{
3275	struct ifreq ifreq;
3276	u32 data32;
3277
3278	if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3279	return -EFAULT;
3280	if (get_user(data32, &u_ifreq32->ifr_data))
3281	return -EFAULT;
3282	ifreq.ifr_data = compat_ptr(data32);
3283
3284	return dev_ioctl(net, cmd, &ifreq, NULL);
3285	}
3286
3287	static int compat_ifreq_ioctl(struct net net, struct socket sock,
3288	unsigned int cmd,
3289	struct compat_ifreq __user *uifr32)
3290	{
3291	struct ifreq __user *uifr;
3292	int err;
3293
3294	/* Handle the fact that while struct ifreq has the same layout on
3295	* 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3296	* which are handled elsewhere, it still has different size due to
3297	* ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3298	* resulting in struct ifreq being 32 and 40 bytes respectively).
3299	* As a result, if the struct happens to be at the end of a page and
3300	* the next page isn't readable/writable, we get a fault. To prevent
3301	* that, copy back and forth to the full size.
3302	*/
3303
3304	uifr = compat_alloc_user_space(sizeof(*uifr));
3305	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3306	return -EFAULT;
3307
3308	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3309
3310	if (!err) {
3311	switch (cmd) {
3312	case SIOCGIFFLAGS:
3313	case SIOCGIFMETRIC:
3314	case SIOCGIFMTU:
3315	case SIOCGIFMEM:
3316	case SIOCGIFHWADDR:
3317	case SIOCGIFINDEX:
3318	case SIOCGIFADDR:
3319	case SIOCGIFBRDADDR:
3320	case SIOCGIFDSTADDR:
3321	case SIOCGIFNETMASK:
3322	case SIOCGIFPFLAGS:
3323	case SIOCGIFTXQLEN:
3324	case SIOCGMIIPHY:
3325	case SIOCGMIIREG:
3326	case SIOCGIFNAME:
3327	if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3328	err = -EFAULT;
3329	break;
3330	}
3331	}
3332	return err;
3333	}
3334
3335	static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3336	struct compat_ifreq __user *uifr32)
3337	{
3338	struct ifreq ifr;
3339	struct compat_ifmap __user *uifmap32;
3340	int err;
3341
3342	uifmap32 = &uifr32->ifr_ifru.ifru_map;
3343	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3344	err \|= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3345	err \|= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3346	err \|= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3347	err \|= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3348	err \|= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3349	err \|= get_user(ifr.ifr_map.port, &uifmap32->port);
3350	if (err)
3351	return -EFAULT;
3352
3353	err = dev_ioctl(net, cmd, &ifr, NULL);
3354
3355	if (cmd == SIOCGIFMAP && !err) {
3356	err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3357	err \|= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3358	err \|= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3359	err \|= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3360	err \|= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3361	err \|= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3362	err \|= put_user(ifr.ifr_map.port, &uifmap32->port);
3363	if (err)
3364	err = -EFAULT;
3365	}
3366	return err;
3367	}
3368
3369	/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3370	* for some operations; this forces use of the newer bridge-utils that
3371	* use compatible ioctls
3372	*/
3373	static int old_bridge_ioctl(compat_ulong_t __user *argp)
3374	{
3375	compat_ulong_t tmp;
3376
3377	if (get_user(tmp, argp))
3378	return -EFAULT;
3379	if (tmp == BRCTL_GET_VERSION)
3380	return BRCTL_VERSION + 1;
3381	return -EINVAL;
3382	}
3383
3384	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3385	unsigned int cmd, unsigned long arg)
3386	{
3387	void __user *argp = compat_ptr(arg);
3388	struct sock *sk = sock->sk;
3389	struct net *net = sock_net(sk);
3390
3391	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3392	return compat_ifr_data_ioctl(net, cmd, argp);
3393
3394	switch (cmd) {
3395	case SIOCSIFBR:
3396	case SIOCGIFBR:
3397	return old_bridge_ioctl(argp);
3398	case SIOCGIFCONF:
3399	return compat_dev_ifconf(net, argp);
3400	case SIOCETHTOOL:
3401	return ethtool_ioctl(net, argp);
3402	case SIOCWANDEV:
3403	return compat_siocwandev(net, argp);
3404	case SIOCGIFMAP:
3405	case SIOCSIFMAP:
3406	return compat_sioc_ifmap(net, cmd, argp);
3407	case SIOCGSTAMP_OLD:
3408	case SIOCGSTAMPNS_OLD:
3409	if (!sock->ops->gettstamp)
3410	return -ENOIOCTLCMD;
3411	return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3412	!COMPAT_USE_64BIT_TIME);
3413
3414	case SIOCBONDSLAVEINFOQUERY:
3415	case SIOCBONDINFOQUERY:
3416	case SIOCSHWTSTAMP:
3417	case SIOCGHWTSTAMP:
3418	return compat_ifr_data_ioctl(net, cmd, argp);
3419
3420	case FIOSETOWN:
3421	case SIOCSPGRP:
3422	case FIOGETOWN:
3423	case SIOCGPGRP:
3424	case SIOCBRADDBR:
3425	case SIOCBRDELBR:
3426	case SIOCGIFVLAN:
3427	case SIOCSIFVLAN:
3428	case SIOCADDDLCI:
3429	case SIOCDELDLCI:
3430	case SIOCGSKNS:
3431	case SIOCGSTAMP_NEW:
3432	case SIOCGSTAMPNS_NEW:
3433	return sock_ioctl(file, cmd, arg);
3434
3435	case SIOCGIFFLAGS:
3436	case SIOCSIFFLAGS:
3437	case SIOCGIFMETRIC:
3438	case SIOCSIFMETRIC:
3439	case SIOCGIFMTU:
3440	case SIOCSIFMTU:
3441	case SIOCGIFMEM:
3442	case SIOCSIFMEM:
3443	case SIOCGIFHWADDR:
3444	case SIOCSIFHWADDR:
3445	case SIOCADDMULTI:
3446	case SIOCDELMULTI:
3447	case SIOCGIFINDEX:
3448	case SIOCGIFADDR:
3449	case SIOCSIFADDR:
3450	case SIOCSIFHWBROADCAST:
3451	case SIOCDIFADDR:
3452	case SIOCGIFBRDADDR:
3453	case SIOCSIFBRDADDR:
3454	case SIOCGIFDSTADDR:
3455	case SIOCSIFDSTADDR:
3456	case SIOCGIFNETMASK:
3457	case SIOCSIFNETMASK:
3458	case SIOCSIFPFLAGS:
3459	case SIOCGIFPFLAGS:
3460	case SIOCGIFTXQLEN:
3461	case SIOCSIFTXQLEN:
3462	case SIOCBRADDIF:
3463	case SIOCBRDELIF:
3464	case SIOCGIFNAME:
3465	case SIOCSIFNAME:
3466	case SIOCGMIIPHY:
3467	case SIOCGMIIREG:
3468	case SIOCSMIIREG:
3469	case SIOCBONDENSLAVE:
3470	case SIOCBONDRELEASE:
3471	case SIOCBONDSETHWADDR:
3472	case SIOCBONDCHANGEACTIVE:
3473	return compat_ifreq_ioctl(net, sock, cmd, argp);
3474
3475	case SIOCSARP:
3476	case SIOCGARP:
3477	case SIOCDARP:
3478	case SIOCOUTQ:
3479	case SIOCOUTQNSD:
3480	case SIOCATMARK:
3481	return sock_do_ioctl(net, sock, cmd, arg);
3482	}
3483
3484	return -ENOIOCTLCMD;
3485	}
3486
3487	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3488	unsigned long arg)
3489	{
3490	struct socket *sock = file->private_data;
3491	int ret = -ENOIOCTLCMD;
3492	struct sock *sk;
3493	struct net *net;
3494
3495	sk = sock->sk;
3496	net = sock_net(sk);
3497
3498	if (sock->ops->compat_ioctl)
3499	ret = sock->ops->compat_ioctl(sock, cmd, arg);
3500
3501	if (ret == -ENOIOCTLCMD &&
3502	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3503	ret = compat_wext_handle_ioctl(net, cmd, arg);
3504
3505	if (ret == -ENOIOCTLCMD)
3506	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3507
3508	return ret;
3509	}
3510	#endif
3511
3512	/**
3513	* kernel_bind - bind an address to a socket (kernel space)
3514	* @sock: socket
3515	* @addr: address
3516	* @addrlen: length of address
3517	*
3518	* Returns 0 or an error.
3519	*/
3520
3521	int kernel_bind(struct socket sock, struct sockaddr addr, int addrlen)
3522	{
3523	return sock->ops->bind(sock, addr, addrlen);
3524	}
3525	EXPORT_SYMBOL(kernel_bind);
3526
3527	/**
3528	* kernel_listen - move socket to listening state (kernel space)
3529	* @sock: socket
3530	* @backlog: pending connections queue size
3531	*
3532	* Returns 0 or an error.
3533	*/
3534
3535	int kernel_listen(struct socket *sock, int backlog)
3536	{
3537	return sock->ops->listen(sock, backlog);
3538	}
3539	EXPORT_SYMBOL(kernel_listen);
3540
3541	/**
3542	* kernel_accept - accept a connection (kernel space)
3543	* @sock: listening socket
3544	* @newsock: new connected socket
3545	* @flags: flags
3546	*
3547	* @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3548	* If it fails, @newsock is guaranteed to be %NULL.
3549	* Returns 0 or an error.
3550	*/
3551
3552	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3553	{
3554	struct sock *sk = sock->sk;
3555	int err;
3556
3557	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3558	newsock);
3559	if (err < 0)
3560	goto done;
3561
3562	err = sock->ops->accept(sock, *newsock, flags, true);
3563	if (err < 0) {
3564	sock_release(*newsock);
3565	*newsock = NULL;
3566	goto done;
3567	}
3568
3569	(*newsock)->ops = sock->ops;
3570	__module_get((*newsock)->ops->owner);
3571
3572	done:
3573	return err;
3574	}
3575	EXPORT_SYMBOL(kernel_accept);
3576
3577	/**
3578	* kernel_connect - connect a socket (kernel space)
3579	* @sock: socket
3580	* @addr: address
3581	* @addrlen: address length
3582	* @flags: flags (O_NONBLOCK, ...)
3583	*
3584	* For datagram sockets, @addr is the addres to which datagrams are sent
3585	* by default, and the only address from which datagrams are received.
3586	* For stream sockets, attempts to connect to @addr.
3587	* Returns 0 or an error code.
3588	*/
3589
3590	int kernel_connect(struct socket sock, struct sockaddr addr, int addrlen,
3591	int flags)
3592	{
3593	return sock->ops->connect(sock, addr, addrlen, flags);
3594	}
3595	EXPORT_SYMBOL(kernel_connect);
3596
3597	/**
3598	* kernel_getsockname - get the address which the socket is bound (kernel space)
3599	* @sock: socket
3600	* @addr: address holder
3601	*
3602	* Fills the @addr pointer with the address which the socket is bound.
3603	* Returns 0 or an error code.
3604	*/
3605
3606	int kernel_getsockname(struct socket sock, struct sockaddr addr)
3607	{
3608	return sock->ops->getname(sock, addr, 0);
3609	}
3610	EXPORT_SYMBOL(kernel_getsockname);
3611
3612	/**
3613	* kernel_peername - get the address which the socket is connected (kernel space)
3614	* @sock: socket
3615	* @addr: address holder
3616	*
3617	* Fills the @addr pointer with the address which the socket is connected.
3618	* Returns 0 or an error code.
3619	*/
3620
3621	int kernel_getpeername(struct socket sock, struct sockaddr addr)
3622	{
3623	return sock->ops->getname(sock, addr, 1);
3624	}
3625	EXPORT_SYMBOL(kernel_getpeername);
3626
3627	/**
3628	* kernel_sendpage - send a &page through a socket (kernel space)
3629	* @sock: socket
3630	* @page: page
3631	* @offset: page offset
3632	* @size: total size in bytes
3633	* @flags: flags (MSG_DONTWAIT, ...)
3634	*
3635	* Returns the total amount sent in bytes or an error.
3636	*/
3637
3638	int kernel_sendpage(struct socket sock, struct page page, int offset,
3639	size_t size, int flags)
3640	{
3641	if (sock->ops->sendpage)
3642	return sock->ops->sendpage(sock, page, offset, size, flags);
3643
3644	return sock_no_sendpage(sock, page, offset, size, flags);
3645	}
3646	EXPORT_SYMBOL(kernel_sendpage);
3647
3648	/**
3649	* kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3650	* @sk: sock
3651	* @page: page
3652	* @offset: page offset
3653	* @size: total size in bytes
3654	* @flags: flags (MSG_DONTWAIT, ...)
3655	*
3656	* Returns the total amount sent in bytes or an error.
3657	* Caller must hold @sk.
3658	*/
3659
3660	int kernel_sendpage_locked(struct sock sk, struct page page, int offset,
3661	size_t size, int flags)
3662	{
3663	struct socket *sock = sk->sk_socket;
3664
3665	if (sock->ops->sendpage_locked)
3666	return sock->ops->sendpage_locked(sk, page, offset, size,
3667	flags);
3668
3669	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3670	}
3671	EXPORT_SYMBOL(kernel_sendpage_locked);
3672
3673	/**
3674	* kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3675	* @sock: socket
3676	* @how: connection part
3677	*
3678	* Returns 0 or an error.
3679	*/
3680
3681	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3682	{
3683	return sock->ops->shutdown(sock, how);
3684	}
3685	EXPORT_SYMBOL(kernel_sock_shutdown);
3686
3687	/**
3688	* kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3689	* @sk: socket
3690	*
3691	* This routine returns the IP overhead imposed by a socket i.e.
3692	* the length of the underlying IP header, depending on whether
3693	* this is an IPv4 or IPv6 socket and the length from IP options turned
3694	* on at the socket. Assumes that the caller has a lock on the socket.
3695	*/
3696
3697	u32 kernel_sock_ip_overhead(struct sock *sk)
3698	{
3699	struct inet_sock *inet;
3700	struct ip_options_rcu *opt;
3701	u32 overhead = 0;
3702	#if IS_ENABLED(CONFIG_IPV6)
3703	struct ipv6_pinfo *np;
3704	struct ipv6_txoptions *optv6 = NULL;
3705	#endif /* IS_ENABLED(CONFIG_IPV6) */
3706
3707	if (!sk)
3708	return overhead;
3709
3710	switch (sk->sk_family) {
3711	case AF_INET:
3712	inet = inet_sk(sk);
3713	overhead += sizeof(struct iphdr);
3714	opt = rcu_dereference_protected(inet->inet_opt,
3715	sock_owned_by_user(sk));
3716	if (opt)
3717	overhead += opt->opt.optlen;
3718	return overhead;
3719	#if IS_ENABLED(CONFIG_IPV6)
3720	case AF_INET6:
3721	np = inet6_sk(sk);
3722	overhead += sizeof(struct ipv6hdr);
3723	if (np)
3724	optv6 = rcu_dereference_protected(np->opt,
3725	sock_owned_by_user(sk));
3726	if (optv6)
3727	overhead += (optv6->opt_flen + optv6->opt_nflen);
3728	return overhead;
3729	#endif /* IS_ENABLED(CONFIG_IPV6) */
3730	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3731	return overhead;
3732	}
3733	}
3734	EXPORT_SYMBOL(kernel_sock_ip_overhead);