linux - net/socket.c

3710 lines · 91.0 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/ethtool.h>
56	#include <linux/mm.h>
57	#include <linux/socket.h>
58	#include <linux/file.h>
59	#include <linux/net.h>
60	#include <linux/interrupt.h>
61	#include <linux/thread_info.h>
62	#include <linux/rcupdate.h>
63	#include <linux/netdevice.h>
64	#include <linux/proc_fs.h>
65	#include <linux/seq_file.h>
66	#include <linux/mutex.h>
67	#include <linux/if_bridge.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 user_namespace *mnt_userns,
338	struct dentry dentry, struct inode inode,
339	const char suffix, const void value,
340	size_t size, int flags)
341	{
342	/* Handled by LSM. */
343	return -EAGAIN;
344	}
345
346	static const struct xattr_handler sockfs_security_xattr_handler = {
347	.prefix = XATTR_SECURITY_PREFIX,
348	.set = sockfs_security_xattr_set,
349	};
350
351	static const struct xattr_handler *sockfs_xattr_handlers[] = {
352	&sockfs_xattr_handler,
353	&sockfs_security_xattr_handler,
354	NULL
355	};
356
357	static int sockfs_init_fs_context(struct fs_context *fc)
358	{
359	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
360	if (!ctx)
361	return -ENOMEM;
362	ctx->ops = &sockfs_ops;
363	ctx->dops = &sockfs_dentry_operations;
364	ctx->xattr = sockfs_xattr_handlers;
365	return 0;
366	}
367
368	static struct vfsmount *sock_mnt __read_mostly;
369
370	static struct file_system_type sock_fs_type = {
371	.name = "sockfs",
372	.init_fs_context = sockfs_init_fs_context,
373	.kill_sb = kill_anon_super,
374	};
375
376	/*
377	* Obtains the first available file descriptor and sets it up for use.
378	*
379	* These functions create file structures and maps them to fd space
380	* of the current process. On success it returns file descriptor
381	* and file struct implicitly stored in sock->file.
382	* Note that another thread may close file descriptor before we return
383	* from this function. We use the fact that now we do not refer
384	* to socket after mapping. If one day we will need it, this
385	* function will increment ref. count on file by 1.
386	*
387	* In any case returned fd MAY BE not valid!
388	* This race condition is unavoidable
389	* with shared fd spaces, we cannot solve it inside kernel,
390	* but we take care of internal coherence yet.
391	*/
392
393	/**
394	* sock_alloc_file - Bind a &socket to a &file
395	* @sock: socket
396	* @flags: file status flags
397	* @dname: protocol name
398	*
399	* Returns the &file bound with @sock, implicitly storing it
400	* in sock->file. If dname is %NULL, sets to "".
401	* On failure the return is a ERR pointer (see linux/err.h).
402	* This function uses GFP_KERNEL internally.
403	*/
404
405	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
406	{
407	struct file *file;
408
409	if (!dname)
410	dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
411
412	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
413	O_RDWR \| (flags & O_NONBLOCK),
414	&socket_file_ops);
415	if (IS_ERR(file)) {
416	sock_release(sock);
417	return file;
418	}
419
420	sock->file = file;
421	file->private_data = sock;
422	stream_open(SOCK_INODE(sock), file);
423	return file;
424	}
425	EXPORT_SYMBOL(sock_alloc_file);
426
427	static int sock_map_fd(struct socket *sock, int flags)
428	{
429	struct file *newfile;
430	int fd = get_unused_fd_flags(flags);
431	if (unlikely(fd < 0)) {
432	sock_release(sock);
433	return fd;
434	}
435
436	newfile = sock_alloc_file(sock, flags, NULL);
437	if (!IS_ERR(newfile)) {
438	fd_install(fd, newfile);
439	return fd;
440	}
441
442	put_unused_fd(fd);
443	return PTR_ERR(newfile);
444	}
445
446	/**
447	* sock_from_file - Return the &socket bounded to @file.
448	* @file: file
449	*
450	* On failure returns %NULL.
451	*/
452
453	struct socket sock_from_file(struct file file)
454	{
455	if (file->f_op == &socket_file_ops)
456	return file->private_data; /* set in sock_map_fd */
457
458	return NULL;
459	}
460	EXPORT_SYMBOL(sock_from_file);
461
462	/**
463	* sockfd_lookup - Go from a file number to its socket slot
464	* @fd: file handle
465	* @err: pointer to an error code return
466	*
467	* The file handle passed in is locked and the socket it is bound
468	* to is returned. If an error occurs the err pointer is overwritten
469	* with a negative errno code and NULL is returned. The function checks
470	* for both invalid handles and passing a handle which is not a socket.
471	*
472	* On a success the socket object pointer is returned.
473	*/
474
475	struct socket sockfd_lookup(int fd, int err)
476	{
477	struct file *file;
478	struct socket *sock;
479
480	file = fget(fd);
481	if (!file) {
482	*err = -EBADF;
483	return NULL;
484	}
485
486	sock = sock_from_file(file);
487	if (!sock) {
488	*err = -ENOTSOCK;
489	fput(file);
490	}
491	return sock;
492	}
493	EXPORT_SYMBOL(sockfd_lookup);
494
495	static struct socket sockfd_lookup_light(int fd, int err, int *fput_needed)
496	{
497	struct fd f = fdget(fd);
498	struct socket *sock;
499
500	*err = -EBADF;
501	if (f.file) {
502	sock = sock_from_file(f.file);
503	if (likely(sock)) {
504	*fput_needed = f.flags & FDPUT_FPUT;
505	return sock;
506	}
507	*err = -ENOTSOCK;
508	fdput(f);
509	}
510	return NULL;
511	}
512
513	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
514	size_t size)
515	{
516	ssize_t len;
517	ssize_t used = 0;
518
519	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
520	if (len < 0)
521	return len;
522	used += len;
523	if (buffer) {
524	if (size < used)
525	return -ERANGE;
526	buffer += len;
527	}
528
529	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
530	used += len;
531	if (buffer) {
532	if (size < used)
533	return -ERANGE;
534	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
535	buffer += len;
536	}
537
538	return used;
539	}
540
541	static int sockfs_setattr(struct user_namespace *mnt_userns,
542	struct dentry dentry, struct iattr iattr)
543	{
544	int err = simple_setattr(&init_user_ns, dentry, iattr);
545
546	if (!err && (iattr->ia_valid & ATTR_UID)) {
547	struct socket *sock = SOCKET_I(d_inode(dentry));
548
549	if (sock->sk)
550	sock->sk->sk_uid = iattr->ia_uid;
551	else
552	err = -ENOENT;
553	}
554
555	return err;
556	}
557
558	static const struct inode_operations sockfs_inode_ops = {
559	.listxattr = sockfs_listxattr,
560	.setattr = sockfs_setattr,
561	};
562
563	/**
564	* sock_alloc - allocate a socket
565	*
566	* Allocate a new inode and socket object. The two are bound together
567	* and initialised. The socket is then returned. If we are out of inodes
568	* NULL is returned. This functions uses GFP_KERNEL internally.
569	*/
570
571	struct socket *sock_alloc(void)
572	{
573	struct inode *inode;
574	struct socket *sock;
575
576	inode = new_inode_pseudo(sock_mnt->mnt_sb);
577	if (!inode)
578	return NULL;
579
580	sock = SOCKET_I(inode);
581
582	inode->i_ino = get_next_ino();
583	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
584	inode->i_uid = current_fsuid();
585	inode->i_gid = current_fsgid();
586	inode->i_op = &sockfs_inode_ops;
587
588	return sock;
589	}
590	EXPORT_SYMBOL(sock_alloc);
591
592	static void __sock_release(struct socket sock, struct inode inode)
593	{
594	if (sock->ops) {
595	struct module *owner = sock->ops->owner;
596
597	if (inode)
598	inode_lock(inode);
599	sock->ops->release(sock);
600	sock->sk = NULL;
601	if (inode)
602	inode_unlock(inode);
603	sock->ops = NULL;
604	module_put(owner);
605	}
606
607	if (sock->wq.fasync_list)
608	pr_err("%s: fasync list not empty!\n", __func__);
609
610	if (!sock->file) {
611	iput(SOCK_INODE(sock));
612	return;
613	}
614	sock->file = NULL;
615	}
616
617	/**
618	* sock_release - close a socket
619	* @sock: socket to close
620	*
621	* The socket is released from the protocol stack if it has a release
622	* callback, and the inode is then released if the socket is bound to
623	* an inode not a file.
624	*/
625	void sock_release(struct socket *sock)
626	{
627	__sock_release(sock, NULL);
628	}
629	EXPORT_SYMBOL(sock_release);
630
631	void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
632	{
633	u8 flags = *tx_flags;
634
635	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
636	flags \|= SKBTX_HW_TSTAMP;
637
638	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
639	flags \|= SKBTX_SW_TSTAMP;
640
641	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
642	flags \|= SKBTX_SCHED_TSTAMP;
643
644	*tx_flags = flags;
645	}
646	EXPORT_SYMBOL(__sock_tx_timestamp);
647
648	INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket , struct msghdr ,
649	size_t));
650	INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket , struct msghdr ,
651	size_t));
652	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
653	{
654	int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
655	inet_sendmsg, sock, msg,
656	msg_data_left(msg));
657	BUG_ON(ret == -EIOCBQUEUED);
658	return ret;
659	}
660
661	/**
662	* sock_sendmsg - send a message through @sock
663	* @sock: socket
664	* @msg: message to send
665	*
666	* Sends @msg through @sock, passing through LSM.
667	* Returns the number of bytes sent, or an error code.
668	*/
669	int sock_sendmsg(struct socket sock, struct msghdr msg)
670	{
671	int err = security_socket_sendmsg(sock, msg,
672	msg_data_left(msg));
673
674	return err ?: sock_sendmsg_nosec(sock, msg);
675	}
676	EXPORT_SYMBOL(sock_sendmsg);
677
678	/**
679	* kernel_sendmsg - send a message through @sock (kernel-space)
680	* @sock: socket
681	* @msg: message header
682	* @vec: kernel vec
683	* @num: vec array length
684	* @size: total message data size
685	*
686	* Builds the message data with @vec and sends it through @sock.
687	* Returns the number of bytes sent, or an error code.
688	*/
689
690	int kernel_sendmsg(struct socket sock, struct msghdr msg,
691	struct kvec *vec, size_t num, size_t size)
692	{
693	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
694	return sock_sendmsg(sock, msg);
695	}
696	EXPORT_SYMBOL(kernel_sendmsg);
697
698	/**
699	* kernel_sendmsg_locked - send a message through @sock (kernel-space)
700	* @sk: sock
701	* @msg: message header
702	* @vec: output s/g array
703	* @num: output s/g array length
704	* @size: total message data size
705	*
706	* Builds the message data with @vec and sends it through @sock.
707	* Returns the number of bytes sent, or an error code.
708	* Caller must hold @sk.
709	*/
710
711	int kernel_sendmsg_locked(struct sock sk, struct msghdr msg,
712	struct kvec *vec, size_t num, size_t size)
713	{
714	struct socket *sock = sk->sk_socket;
715
716	if (!sock->ops->sendmsg_locked)
717	return sock_no_sendmsg_locked(sk, msg, size);
718
719	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
720
721	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
722	}
723	EXPORT_SYMBOL(kernel_sendmsg_locked);
724
725	static bool skb_is_err_queue(const struct sk_buff *skb)
726	{
727	/* pkt_type of skbs enqueued on the error queue are set to
728	* PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
729	* in recvmsg, since skbs received on a local socket will never
730	* have a pkt_type of PACKET_OUTGOING.
731	*/
732	return skb->pkt_type == PACKET_OUTGOING;
733	}
734
735	/* On transmit, software and hardware timestamps are returned independently.
736	* As the two skb clones share the hardware timestamp, which may be updated
737	* before the software timestamp is received, a hardware TX timestamp may be
738	* returned only if there is no software TX timestamp. Ignore false software
739	* timestamps, which may be made in the __sock_recv_timestamp() call when the
740	* option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
741	* hardware timestamp.
742	*/
743	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
744	{
745	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
746	}
747
748	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb)
749	{
750	struct scm_ts_pktinfo ts_pktinfo;
751	struct net_device *orig_dev;
752
753	if (!skb_mac_header_was_set(skb))
754	return;
755
756	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
757
758	rcu_read_lock();
759	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
760	if (orig_dev)
761	ts_pktinfo.if_index = orig_dev->ifindex;
762	rcu_read_unlock();
763
764	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
765	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
766	sizeof(ts_pktinfo), &ts_pktinfo);
767	}
768
769	/*
770	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
771	*/
772	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
773	struct sk_buff *skb)
774	{
775	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
776	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
777	struct scm_timestamping_internal tss;
778
779	int empty = 1, false_tstamp = 0;
780	struct skb_shared_hwtstamps *shhwtstamps =
781	skb_hwtstamps(skb);
782
783	/* Race occurred between timestamp enabling and packet
784	receiving. Fill in the current time for now. */
785	if (need_software_tstamp && skb->tstamp == 0) {
786	__net_timestamp(skb);
787	false_tstamp = 1;
788	}
789
790	if (need_software_tstamp) {
791	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
792	if (new_tstamp) {
793	struct __kernel_sock_timeval tv;
794
795	skb_get_new_timestamp(skb, &tv);
796	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
797	sizeof(tv), &tv);
798	} else {
799	struct __kernel_old_timeval tv;
800
801	skb_get_timestamp(skb, &tv);
802	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
803	sizeof(tv), &tv);
804	}
805	} else {
806	if (new_tstamp) {
807	struct __kernel_timespec ts;
808
809	skb_get_new_timestampns(skb, &ts);
810	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
811	sizeof(ts), &ts);
812	} else {
813	struct __kernel_old_timespec ts;
814
815	skb_get_timestampns(skb, &ts);
816	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
817	sizeof(ts), &ts);
818	}
819	}
820	}
821
822	memset(&tss, 0, sizeof(tss));
823	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
824	ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
825	empty = 0;
826	if (shhwtstamps &&
827	(sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
828	!skb_is_swtx_tstamp(skb, false_tstamp) &&
829	ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
830	empty = 0;
831	if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
832	!skb_is_err_queue(skb))
833	put_ts_pktinfo(msg, skb);
834	}
835	if (!empty) {
836	if (sock_flag(sk, SOCK_TSTAMP_NEW))
837	put_cmsg_scm_timestamping64(msg, &tss);
838	else
839	put_cmsg_scm_timestamping(msg, &tss);
840
841	if (skb_is_err_queue(skb) && skb->len &&
842	SKB_EXT_ERR(skb)->opt_stats)
843	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
844	skb->len, skb->data);
845	}
846	}
847	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
848
849	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
850	struct sk_buff *skb)
851	{
852	int ack;
853
854	if (!sock_flag(sk, SOCK_WIFI_STATUS))
855	return;
856	if (!skb->wifi_acked_valid)
857	return;
858
859	ack = skb->wifi_acked;
860
861	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
862	}
863	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
864
865	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
866	struct sk_buff *skb)
867	{
868	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
869	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
870	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
871	}
872
873	void __sock_recv_ts_and_drops(struct msghdr msg, struct sock sk,
874	struct sk_buff *skb)
875	{
876	sock_recv_timestamp(msg, sk, skb);
877	sock_recv_drops(msg, sk, skb);
878	}
879	EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
880
881	INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket , struct msghdr ,
882	size_t, int));
883	INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket , struct msghdr ,
884	size_t, int));
885	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
886	int flags)
887	{
888	return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
889	inet_recvmsg, sock, msg, msg_data_left(msg),
890	flags);
891	}
892
893	/**
894	* sock_recvmsg - receive a message from @sock
895	* @sock: socket
896	* @msg: message to receive
897	* @flags: message flags
898	*
899	* Receives @msg from @sock, passing through LSM. Returns the total number
900	* of bytes received, or an error.
901	*/
902	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
903	{
904	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
905
906	return err ?: sock_recvmsg_nosec(sock, msg, flags);
907	}
908	EXPORT_SYMBOL(sock_recvmsg);
909
910	/**
911	* kernel_recvmsg - Receive a message from a socket (kernel space)
912	* @sock: The socket to receive the message from
913	* @msg: Received message
914	* @vec: Input s/g array for message data
915	* @num: Size of input s/g array
916	* @size: Number of bytes to read
917	* @flags: Message flags (MSG_DONTWAIT, etc...)
918	*
919	* On return the msg structure contains the scatter/gather array passed in the
920	* vec argument. The array is modified so that it consists of the unfilled
921	* portion of the original array.
922	*
923	* The returned value is the total number of bytes received, or an error.
924	*/
925
926	int kernel_recvmsg(struct socket sock, struct msghdr msg,
927	struct kvec *vec, size_t num, size_t size, int flags)
928	{
929	msg->msg_control_is_user = false;
930	iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
931	return sock_recvmsg(sock, msg, flags);
932	}
933	EXPORT_SYMBOL(kernel_recvmsg);
934
935	static ssize_t sock_sendpage(struct file file, struct page page,
936	int offset, size_t size, loff_t *ppos, int more)
937	{
938	struct socket *sock;
939	int flags;
940
941	sock = file->private_data;
942
943	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
944	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
945	flags \|= more;
946
947	return kernel_sendpage(sock, page, offset, size, flags);
948	}
949
950	static ssize_t sock_splice_read(struct file file, loff_t ppos,
951	struct pipe_inode_info *pipe, size_t len,
952	unsigned int flags)
953	{
954	struct socket *sock = file->private_data;
955
956	if (unlikely(!sock->ops->splice_read))
957	return generic_file_splice_read(file, ppos, pipe, len, flags);
958
959	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
960	}
961
962	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
963	{
964	struct file *file = iocb->ki_filp;
965	struct socket *sock = file->private_data;
966	struct msghdr msg = {.msg_iter = *to,
967	.msg_iocb = iocb};
968	ssize_t res;
969
970	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
971	msg.msg_flags = MSG_DONTWAIT;
972
973	if (iocb->ki_pos != 0)
974	return -ESPIPE;
975
976	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
977	return 0;
978
979	res = sock_recvmsg(sock, &msg, msg.msg_flags);
980	*to = msg.msg_iter;
981	return res;
982	}
983
984	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
985	{
986	struct file *file = iocb->ki_filp;
987	struct socket *sock = file->private_data;
988	struct msghdr msg = {.msg_iter = *from,
989	.msg_iocb = iocb};
990	ssize_t res;
991
992	if (iocb->ki_pos != 0)
993	return -ESPIPE;
994
995	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
996	msg.msg_flags = MSG_DONTWAIT;
997
998	if (sock->type == SOCK_SEQPACKET)
999	msg.msg_flags \|= MSG_EOR;
1000
1001	res = sock_sendmsg(sock, &msg);
1002	*from = msg.msg_iter;
1003	return res;
1004	}
1005
1006	/*
1007	* Atomic setting of ioctl hooks to avoid race
1008	* with module unload.
1009	*/
1010
1011	static DEFINE_MUTEX(br_ioctl_mutex);
1012	static int (br_ioctl_hook) (struct net , unsigned int cmd, void __user *arg);
1013
1014	void brioctl_set(int (hook) (struct net , unsigned int, void __user *))
1015	{
1016	mutex_lock(&br_ioctl_mutex);
1017	br_ioctl_hook = hook;
1018	mutex_unlock(&br_ioctl_mutex);
1019	}
1020	EXPORT_SYMBOL(brioctl_set);
1021
1022	static DEFINE_MUTEX(vlan_ioctl_mutex);
1023	static int (vlan_ioctl_hook) (struct net , void __user *arg);
1024
1025	void vlan_ioctl_set(int (hook) (struct net , void __user *))
1026	{
1027	mutex_lock(&vlan_ioctl_mutex);
1028	vlan_ioctl_hook = hook;
1029	mutex_unlock(&vlan_ioctl_mutex);
1030	}
1031	EXPORT_SYMBOL(vlan_ioctl_set);
1032
1033	static long sock_do_ioctl(struct net net, struct socket sock,
1034	unsigned int cmd, unsigned long arg)
1035	{
1036	int err;
1037	void __user argp = (void __user )arg;
1038
1039	err = sock->ops->ioctl(sock, cmd, arg);
1040
1041	/*
1042	* If this ioctl is unknown try to hand it down
1043	* to the NIC driver.
1044	*/
1045	if (err != -ENOIOCTLCMD)
1046	return err;
1047
1048	if (cmd == SIOCGIFCONF) {
1049	struct ifconf ifc;
1050	if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1051	return -EFAULT;
1052	rtnl_lock();
1053	err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1054	rtnl_unlock();
1055	if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1056	err = -EFAULT;
1057	} else {
1058	struct ifreq ifr;
1059	bool need_copyout;
1060	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1061	return -EFAULT;
1062	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1063	if (!err && need_copyout)
1064	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1065	return -EFAULT;
1066	}
1067	return err;
1068	}
1069
1070	/*
1071	* With an ioctl, arg may well be a user mode pointer, but we don't know
1072	* what to do with it - that's up to the protocol still.
1073	*/
1074
1075	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1076	{
1077	struct socket *sock;
1078	struct sock *sk;
1079	void __user argp = (void __user )arg;
1080	int pid, err;
1081	struct net *net;
1082
1083	sock = file->private_data;
1084	sk = sock->sk;
1085	net = sock_net(sk);
1086	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1087	struct ifreq ifr;
1088	bool need_copyout;
1089	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1090	return -EFAULT;
1091	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1092	if (!err && need_copyout)
1093	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1094	return -EFAULT;
1095	} else
1096	#ifdef CONFIG_WEXT_CORE
1097	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1098	err = wext_handle_ioctl(net, cmd, argp);
1099	} else
1100	#endif
1101	switch (cmd) {
1102	case FIOSETOWN:
1103	case SIOCSPGRP:
1104	err = -EFAULT;
1105	if (get_user(pid, (int __user *)argp))
1106	break;
1107	err = f_setown(sock->file, pid, 1);
1108	break;
1109	case FIOGETOWN:
1110	case SIOCGPGRP:
1111	err = put_user(f_getown(sock->file),
1112	(int __user *)argp);
1113	break;
1114	case SIOCGIFBR:
1115	case SIOCSIFBR:
1116	case SIOCBRADDBR:
1117	case SIOCBRDELBR:
1118	err = -ENOPKG;
1119	if (!br_ioctl_hook)
1120	request_module("bridge");
1121
1122	mutex_lock(&br_ioctl_mutex);
1123	if (br_ioctl_hook)
1124	err = br_ioctl_hook(net, cmd, argp);
1125	mutex_unlock(&br_ioctl_mutex);
1126	break;
1127	case SIOCGIFVLAN:
1128	case SIOCSIFVLAN:
1129	err = -ENOPKG;
1130	if (!vlan_ioctl_hook)
1131	request_module("8021q");
1132
1133	mutex_lock(&vlan_ioctl_mutex);
1134	if (vlan_ioctl_hook)
1135	err = vlan_ioctl_hook(net, argp);
1136	mutex_unlock(&vlan_ioctl_mutex);
1137	break;
1138	case SIOCGSKNS:
1139	err = -EPERM;
1140	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1141	break;
1142
1143	err = open_related_ns(&net->ns, get_net_ns);
1144	break;
1145	case SIOCGSTAMP_OLD:
1146	case SIOCGSTAMPNS_OLD:
1147	if (!sock->ops->gettstamp) {
1148	err = -ENOIOCTLCMD;
1149	break;
1150	}
1151	err = sock->ops->gettstamp(sock, argp,
1152	cmd == SIOCGSTAMP_OLD,
1153	!IS_ENABLED(CONFIG_64BIT));
1154	break;
1155	case SIOCGSTAMP_NEW:
1156	case SIOCGSTAMPNS_NEW:
1157	if (!sock->ops->gettstamp) {
1158	err = -ENOIOCTLCMD;
1159	break;
1160	}
1161	err = sock->ops->gettstamp(sock, argp,
1162	cmd == SIOCGSTAMP_NEW,
1163	false);
1164	break;
1165	default:
1166	err = sock_do_ioctl(net, sock, cmd, arg);
1167	break;
1168	}
1169	return err;
1170	}
1171
1172	/**
1173	* sock_create_lite - creates a socket
1174	* @family: protocol family (AF_INET, ...)
1175	* @type: communication type (SOCK_STREAM, ...)
1176	* @protocol: protocol (0, ...)
1177	* @res: new socket
1178	*
1179	* Creates a new socket and assigns it to @res, passing through LSM.
1180	* The new socket initialization is not complete, see kernel_accept().
1181	* Returns 0 or an error. On failure @res is set to %NULL.
1182	* This function internally uses GFP_KERNEL.
1183	*/
1184
1185	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1186	{
1187	int err;
1188	struct socket *sock = NULL;
1189
1190	err = security_socket_create(family, type, protocol, 1);
1191	if (err)
1192	goto out;
1193
1194	sock = sock_alloc();
1195	if (!sock) {
1196	err = -ENOMEM;
1197	goto out;
1198	}
1199
1200	sock->type = type;
1201	err = security_socket_post_create(sock, family, type, protocol, 1);
1202	if (err)
1203	goto out_release;
1204
1205	out:
1206	*res = sock;
1207	return err;
1208	out_release:
1209	sock_release(sock);
1210	sock = NULL;
1211	goto out;
1212	}
1213	EXPORT_SYMBOL(sock_create_lite);
1214
1215	/* No kernel lock held - perfect */
1216	static __poll_t sock_poll(struct file file, poll_table wait)
1217	{
1218	struct socket *sock = file->private_data;
1219	__poll_t events = poll_requested_events(wait), flag = 0;
1220
1221	if (!sock->ops->poll)
1222	return 0;
1223
1224	if (sk_can_busy_loop(sock->sk)) {
1225	/* poll once if requested by the syscall */
1226	if (events & POLL_BUSY_LOOP)
1227	sk_busy_loop(sock->sk, 1);
1228
1229	/* if this socket can poll_ll, tell the system call */
1230	flag = POLL_BUSY_LOOP;
1231	}
1232
1233	return sock->ops->poll(file, sock, wait) \| flag;
1234	}
1235
1236	static int sock_mmap(struct file file, struct vm_area_struct vma)
1237	{
1238	struct socket *sock = file->private_data;
1239
1240	return sock->ops->mmap(file, sock, vma);
1241	}
1242
1243	static int sock_close(struct inode inode, struct file filp)
1244	{
1245	__sock_release(SOCKET_I(inode), inode);
1246	return 0;
1247	}
1248
1249	/*
1250	* Update the socket async list
1251	*
1252	* Fasync_list locking strategy.
1253	*
1254	* 1. fasync_list is modified only under process context socket lock
1255	* i.e. under semaphore.
1256	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1257	* or under socket lock
1258	*/
1259
1260	static int sock_fasync(int fd, struct file *filp, int on)
1261	{
1262	struct socket *sock = filp->private_data;
1263	struct sock *sk = sock->sk;
1264	struct socket_wq *wq = &sock->wq;
1265
1266	if (sk == NULL)
1267	return -EINVAL;
1268
1269	lock_sock(sk);
1270	fasync_helper(fd, filp, on, &wq->fasync_list);
1271
1272	if (!wq->fasync_list)
1273	sock_reset_flag(sk, SOCK_FASYNC);
1274	else
1275	sock_set_flag(sk, SOCK_FASYNC);
1276
1277	release_sock(sk);
1278	return 0;
1279	}
1280
1281	/* This function may be called only under rcu_lock */
1282
1283	int sock_wake_async(struct socket_wq *wq, int how, int band)
1284	{
1285	if (!wq \|\| !wq->fasync_list)
1286	return -1;
1287
1288	switch (how) {
1289	case SOCK_WAKE_WAITD:
1290	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1291	break;
1292	goto call_kill;
1293	case SOCK_WAKE_SPACE:
1294	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1295	break;
1296	fallthrough;
1297	case SOCK_WAKE_IO:
1298	call_kill:
1299	kill_fasync(&wq->fasync_list, SIGIO, band);
1300	break;
1301	case SOCK_WAKE_URG:
1302	kill_fasync(&wq->fasync_list, SIGURG, band);
1303	}
1304
1305	return 0;
1306	}
1307	EXPORT_SYMBOL(sock_wake_async);
1308
1309	/**
1310	* __sock_create - creates a socket
1311	* @net: net namespace
1312	* @family: protocol family (AF_INET, ...)
1313	* @type: communication type (SOCK_STREAM, ...)
1314	* @protocol: protocol (0, ...)
1315	* @res: new socket
1316	* @kern: boolean for kernel space sockets
1317	*
1318	* Creates a new socket and assigns it to @res, passing through LSM.
1319	* Returns 0 or an error. On failure @res is set to %NULL. @kern must
1320	* be set to true if the socket resides in kernel space.
1321	* This function internally uses GFP_KERNEL.
1322	*/
1323
1324	int __sock_create(struct net *net, int family, int type, int protocol,
1325	struct socket **res, int kern)
1326	{
1327	int err;
1328	struct socket *sock;
1329	const struct net_proto_family *pf;
1330
1331	/*
1332	* Check protocol is in range
1333	*/
1334	if (family < 0 \|\| family >= NPROTO)
1335	return -EAFNOSUPPORT;
1336	if (type < 0 \|\| type >= SOCK_MAX)
1337	return -EINVAL;
1338
1339	/* Compatibility.
1340
1341	This uglymoron is moved from INET layer to here to avoid
1342	deadlock in module load.
1343	*/
1344	if (family == PF_INET && type == SOCK_PACKET) {
1345	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1346	current->comm);
1347	family = PF_PACKET;
1348	}
1349
1350	err = security_socket_create(family, type, protocol, kern);
1351	if (err)
1352	return err;
1353
1354	/*
1355	* Allocate the socket and allow the family to set things up. if
1356	* the protocol is 0, the family is instructed to select an appropriate
1357	* default.
1358	*/
1359	sock = sock_alloc();
1360	if (!sock) {
1361	net_warn_ratelimited("socket: no more sockets\n");
1362	return -ENFILE; /* Not exactly a match, but its the
1363	closest posix thing */
1364	}
1365
1366	sock->type = type;
1367
1368	#ifdef CONFIG_MODULES
1369	/* Attempt to load a protocol module if the find failed.
1370	*
1371	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1372	* requested real, full-featured networking support upon configuration.
1373	* Otherwise module support will break!
1374	*/
1375	if (rcu_access_pointer(net_families[family]) == NULL)
1376	request_module("net-pf-%d", family);
1377	#endif
1378
1379	rcu_read_lock();
1380	pf = rcu_dereference(net_families[family]);
1381	err = -EAFNOSUPPORT;
1382	if (!pf)
1383	goto out_release;
1384
1385	/*
1386	* We will call the ->create function, that possibly is in a loadable
1387	* module, so we have to bump that loadable module refcnt first.
1388	*/
1389	if (!try_module_get(pf->owner))
1390	goto out_release;
1391
1392	/* Now protected by module ref count */
1393	rcu_read_unlock();
1394
1395	err = pf->create(net, sock, protocol, kern);
1396	if (err < 0)
1397	goto out_module_put;
1398
1399	/*
1400	* Now to bump the refcnt of the [loadable] module that owns this
1401	* socket at sock_release time we decrement its refcnt.
1402	*/
1403	if (!try_module_get(sock->ops->owner))
1404	goto out_module_busy;
1405
1406	/*
1407	* Now that we're done with the ->create function, the [loadable]
1408	* module can have its refcnt decremented
1409	*/
1410	module_put(pf->owner);
1411	err = security_socket_post_create(sock, family, type, protocol, kern);
1412	if (err)
1413	goto out_sock_release;
1414	*res = sock;
1415
1416	return 0;
1417
1418	out_module_busy:
1419	err = -EAFNOSUPPORT;
1420	out_module_put:
1421	sock->ops = NULL;
1422	module_put(pf->owner);
1423	out_sock_release:
1424	sock_release(sock);
1425	return err;
1426
1427	out_release:
1428	rcu_read_unlock();
1429	goto out_sock_release;
1430	}
1431	EXPORT_SYMBOL(__sock_create);
1432
1433	/**
1434	* sock_create - creates a socket
1435	* @family: protocol family (AF_INET, ...)
1436	* @type: communication type (SOCK_STREAM, ...)
1437	* @protocol: protocol (0, ...)
1438	* @res: new socket
1439	*
1440	* A wrapper around __sock_create().
1441	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1442	*/
1443
1444	int sock_create(int family, int type, int protocol, struct socket **res)
1445	{
1446	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1447	}
1448	EXPORT_SYMBOL(sock_create);
1449
1450	/**
1451	* sock_create_kern - creates a socket (kernel space)
1452	* @net: net namespace
1453	* @family: protocol family (AF_INET, ...)
1454	* @type: communication type (SOCK_STREAM, ...)
1455	* @protocol: protocol (0, ...)
1456	* @res: new socket
1457	*
1458	* A wrapper around __sock_create().
1459	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1460	*/
1461
1462	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1463	{
1464	return __sock_create(net, family, type, protocol, res, 1);
1465	}
1466	EXPORT_SYMBOL(sock_create_kern);
1467
1468	int __sys_socket(int family, int type, int protocol)
1469	{
1470	int retval;
1471	struct socket *sock;
1472	int flags;
1473
1474	/* Check the SOCK_* constants for consistency. */
1475	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1476	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1477	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1478	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1479
1480	flags = type & ~SOCK_TYPE_MASK;
1481	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1482	return -EINVAL;
1483	type &= SOCK_TYPE_MASK;
1484
1485	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1486	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1487
1488	retval = sock_create(family, type, protocol, &sock);
1489	if (retval < 0)
1490	return retval;
1491
1492	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1493	}
1494
1495	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1496	{
1497	return __sys_socket(family, type, protocol);
1498	}
1499
1500	/*
1501	* Create a pair of connected sockets.
1502	*/
1503
1504	int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1505	{
1506	struct socket sock1, sock2;
1507	int fd1, fd2, err;
1508	struct file newfile1, newfile2;
1509	int flags;
1510
1511	flags = type & ~SOCK_TYPE_MASK;
1512	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1513	return -EINVAL;
1514	type &= SOCK_TYPE_MASK;
1515
1516	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1517	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1518
1519	/*
1520	* reserve descriptors and make sure we won't fail
1521	* to return them to userland.
1522	*/
1523	fd1 = get_unused_fd_flags(flags);
1524	if (unlikely(fd1 < 0))
1525	return fd1;
1526
1527	fd2 = get_unused_fd_flags(flags);
1528	if (unlikely(fd2 < 0)) {
1529	put_unused_fd(fd1);
1530	return fd2;
1531	}
1532
1533	err = put_user(fd1, &usockvec[0]);
1534	if (err)
1535	goto out;
1536
1537	err = put_user(fd2, &usockvec[1]);
1538	if (err)
1539	goto out;
1540
1541	/*
1542	* Obtain the first socket and check if the underlying protocol
1543	* supports the socketpair call.
1544	*/
1545
1546	err = sock_create(family, type, protocol, &sock1);
1547	if (unlikely(err < 0))
1548	goto out;
1549
1550	err = sock_create(family, type, protocol, &sock2);
1551	if (unlikely(err < 0)) {
1552	sock_release(sock1);
1553	goto out;
1554	}
1555
1556	err = security_socket_socketpair(sock1, sock2);
1557	if (unlikely(err)) {
1558	sock_release(sock2);
1559	sock_release(sock1);
1560	goto out;
1561	}
1562
1563	err = sock1->ops->socketpair(sock1, sock2);
1564	if (unlikely(err < 0)) {
1565	sock_release(sock2);
1566	sock_release(sock1);
1567	goto out;
1568	}
1569
1570	newfile1 = sock_alloc_file(sock1, flags, NULL);
1571	if (IS_ERR(newfile1)) {
1572	err = PTR_ERR(newfile1);
1573	sock_release(sock2);
1574	goto out;
1575	}
1576
1577	newfile2 = sock_alloc_file(sock2, flags, NULL);
1578	if (IS_ERR(newfile2)) {
1579	err = PTR_ERR(newfile2);
1580	fput(newfile1);
1581	goto out;
1582	}
1583
1584	audit_fd_pair(fd1, fd2);
1585
1586	fd_install(fd1, newfile1);
1587	fd_install(fd2, newfile2);
1588	return 0;
1589
1590	out:
1591	put_unused_fd(fd2);
1592	put_unused_fd(fd1);
1593	return err;
1594	}
1595
1596	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1597	int __user *, usockvec)
1598	{
1599	return __sys_socketpair(family, type, protocol, usockvec);
1600	}
1601
1602	/*
1603	* Bind a name to a socket. Nothing much to do here since it's
1604	* the protocol's responsibility to handle the local address.
1605	*
1606	* We move the socket address to kernel space before we call
1607	* the protocol layer (having also checked the address is ok).
1608	*/
1609
1610	int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1611	{
1612	struct socket *sock;
1613	struct sockaddr_storage address;
1614	int err, fput_needed;
1615
1616	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1617	if (sock) {
1618	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1619	if (!err) {
1620	err = security_socket_bind(sock,
1621	(struct sockaddr *)&address,
1622	addrlen);
1623	if (!err)
1624	err = sock->ops->bind(sock,
1625	(struct sockaddr *)
1626	&address, addrlen);
1627	}
1628	fput_light(sock->file, fput_needed);
1629	}
1630	return err;
1631	}
1632
1633	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1634	{
1635	return __sys_bind(fd, umyaddr, addrlen);
1636	}
1637
1638	/*
1639	* Perform a listen. Basically, we allow the protocol to do anything
1640	* necessary for a listen, and if that works, we mark the socket as
1641	* ready for listening.
1642	*/
1643
1644	int __sys_listen(int fd, int backlog)
1645	{
1646	struct socket *sock;
1647	int err, fput_needed;
1648	int somaxconn;
1649
1650	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1651	if (sock) {
1652	somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1653	if ((unsigned int)backlog > somaxconn)
1654	backlog = somaxconn;
1655
1656	err = security_socket_listen(sock, backlog);
1657	if (!err)
1658	err = sock->ops->listen(sock, backlog);
1659
1660	fput_light(sock->file, fput_needed);
1661	}
1662	return err;
1663	}
1664
1665	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1666	{
1667	return __sys_listen(fd, backlog);
1668	}
1669
1670	int __sys_accept4_file(struct file *file, unsigned file_flags,
1671	struct sockaddr __user *upeer_sockaddr,
1672	int __user *upeer_addrlen, int flags,
1673	unsigned long nofile)
1674	{
1675	struct socket sock, newsock;
1676	struct file *newfile;
1677	int err, len, newfd;
1678	struct sockaddr_storage address;
1679
1680	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1681	return -EINVAL;
1682
1683	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1684	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1685
1686	sock = sock_from_file(file);
1687	if (!sock) {
1688	err = -ENOTSOCK;
1689	goto out;
1690	}
1691
1692	err = -ENFILE;
1693	newsock = sock_alloc();
1694	if (!newsock)
1695	goto out;
1696
1697	newsock->type = sock->type;
1698	newsock->ops = sock->ops;
1699
1700	/*
1701	* We don't need try_module_get here, as the listening socket (sock)
1702	* has the protocol module (sock->ops->owner) held.
1703	*/
1704	__module_get(newsock->ops->owner);
1705
1706	newfd = __get_unused_fd_flags(flags, nofile);
1707	if (unlikely(newfd < 0)) {
1708	err = newfd;
1709	sock_release(newsock);
1710	goto out;
1711	}
1712	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1713	if (IS_ERR(newfile)) {
1714	err = PTR_ERR(newfile);
1715	put_unused_fd(newfd);
1716	goto out;
1717	}
1718
1719	err = security_socket_accept(sock, newsock);
1720	if (err)
1721	goto out_fd;
1722
1723	err = sock->ops->accept(sock, newsock, sock->file->f_flags \| file_flags,
1724	false);
1725	if (err < 0)
1726	goto out_fd;
1727
1728	if (upeer_sockaddr) {
1729	len = newsock->ops->getname(newsock,
1730	(struct sockaddr *)&address, 2);
1731	if (len < 0) {
1732	err = -ECONNABORTED;
1733	goto out_fd;
1734	}
1735	err = move_addr_to_user(&address,
1736	len, upeer_sockaddr, upeer_addrlen);
1737	if (err < 0)
1738	goto out_fd;
1739	}
1740
1741	/* File flags are not inherited via accept() unlike another OSes. */
1742
1743	fd_install(newfd, newfile);
1744	err = newfd;
1745	out:
1746	return err;
1747	out_fd:
1748	fput(newfile);
1749	put_unused_fd(newfd);
1750	goto out;
1751
1752	}
1753
1754	/*
1755	* For accept, we attempt to create a new socket, set up the link
1756	* with the client, wake up the client, then return the new
1757	* connected fd. We collect the address of the connector in kernel
1758	* space and move it to user at the very end. This is unclean because
1759	* we open the socket then return an error.
1760	*
1761	* 1003.1g adds the ability to recvmsg() to query connection pending
1762	* status to recvmsg. We need to add that support in a way thats
1763	* clean when we restructure accept also.
1764	*/
1765
1766	int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1767	int __user *upeer_addrlen, int flags)
1768	{
1769	int ret = -EBADF;
1770	struct fd f;
1771
1772	f = fdget(fd);
1773	if (f.file) {
1774	ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1775	upeer_addrlen, flags,
1776	rlimit(RLIMIT_NOFILE));
1777	fdput(f);
1778	}
1779
1780	return ret;
1781	}
1782
1783	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1784	int __user *, upeer_addrlen, int, flags)
1785	{
1786	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1787	}
1788
1789	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1790	int __user *, upeer_addrlen)
1791	{
1792	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1793	}
1794
1795	/*
1796	* Attempt to connect to a socket with the server address. The address
1797	* is in user space so we verify it is OK and move it to kernel space.
1798	*
1799	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1800	* break bindings
1801	*
1802	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1803	* other SEQPACKET protocols that take time to connect() as it doesn't
1804	* include the -EINPROGRESS status for such sockets.
1805	*/
1806
1807	int __sys_connect_file(struct file file, struct sockaddr_storage address,
1808	int addrlen, int file_flags)
1809	{
1810	struct socket *sock;
1811	int err;
1812
1813	sock = sock_from_file(file);
1814	if (!sock) {
1815	err = -ENOTSOCK;
1816	goto out;
1817	}
1818
1819	err =
1820	security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1821	if (err)
1822	goto out;
1823
1824	err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1825	sock->file->f_flags \| file_flags);
1826	out:
1827	return err;
1828	}
1829
1830	int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1831	{
1832	int ret = -EBADF;
1833	struct fd f;
1834
1835	f = fdget(fd);
1836	if (f.file) {
1837	struct sockaddr_storage address;
1838
1839	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1840	if (!ret)
1841	ret = __sys_connect_file(f.file, &address, addrlen, 0);
1842	fdput(f);
1843	}
1844
1845	return ret;
1846	}
1847
1848	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1849	int, addrlen)
1850	{
1851	return __sys_connect(fd, uservaddr, addrlen);
1852	}
1853
1854	/*
1855	* Get the local address ('name') of a socket object. Move the obtained
1856	* name to user space.
1857	*/
1858
1859	int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1860	int __user *usockaddr_len)
1861	{
1862	struct socket *sock;
1863	struct sockaddr_storage address;
1864	int err, fput_needed;
1865
1866	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1867	if (!sock)
1868	goto out;
1869
1870	err = security_socket_getsockname(sock);
1871	if (err)
1872	goto out_put;
1873
1874	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1875	if (err < 0)
1876	goto out_put;
1877	/* "err" is actually length in this case */
1878	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1879
1880	out_put:
1881	fput_light(sock->file, fput_needed);
1882	out:
1883	return err;
1884	}
1885
1886	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1887	int __user *, usockaddr_len)
1888	{
1889	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1890	}
1891
1892	/*
1893	* Get the remote address ('name') of a socket object. Move the obtained
1894	* name to user space.
1895	*/
1896
1897	int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1898	int __user *usockaddr_len)
1899	{
1900	struct socket *sock;
1901	struct sockaddr_storage address;
1902	int err, fput_needed;
1903
1904	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1905	if (sock != NULL) {
1906	err = security_socket_getpeername(sock);
1907	if (err) {
1908	fput_light(sock->file, fput_needed);
1909	return err;
1910	}
1911
1912	err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1913	if (err >= 0)
1914	/* "err" is actually length in this case */
1915	err = move_addr_to_user(&address, err, usockaddr,
1916	usockaddr_len);
1917	fput_light(sock->file, fput_needed);
1918	}
1919	return err;
1920	}
1921
1922	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1923	int __user *, usockaddr_len)
1924	{
1925	return __sys_getpeername(fd, usockaddr, usockaddr_len);
1926	}
1927
1928	/*
1929	* Send a datagram to a given address. We move the address into kernel
1930	* space and check the user space data area is readable before invoking
1931	* the protocol.
1932	*/
1933	int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1934	struct sockaddr __user *addr, int addr_len)
1935	{
1936	struct socket *sock;
1937	struct sockaddr_storage address;
1938	int err;
1939	struct msghdr msg;
1940	struct iovec iov;
1941	int fput_needed;
1942
1943	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1944	if (unlikely(err))
1945	return err;
1946	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1947	if (!sock)
1948	goto out;
1949
1950	msg.msg_name = NULL;
1951	msg.msg_control = NULL;
1952	msg.msg_controllen = 0;
1953	msg.msg_namelen = 0;
1954	if (addr) {
1955	err = move_addr_to_kernel(addr, addr_len, &address);
1956	if (err < 0)
1957	goto out_put;
1958	msg.msg_name = (struct sockaddr *)&address;
1959	msg.msg_namelen = addr_len;
1960	}
1961	if (sock->file->f_flags & O_NONBLOCK)
1962	flags \|= MSG_DONTWAIT;
1963	msg.msg_flags = flags;
1964	err = sock_sendmsg(sock, &msg);
1965
1966	out_put:
1967	fput_light(sock->file, fput_needed);
1968	out:
1969	return err;
1970	}
1971
1972	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1973	unsigned int, flags, struct sockaddr __user *, addr,
1974	int, addr_len)
1975	{
1976	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1977	}
1978
1979	/*
1980	* Send a datagram down a socket.
1981	*/
1982
1983	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1984	unsigned int, flags)
1985	{
1986	return __sys_sendto(fd, buff, len, flags, NULL, 0);
1987	}
1988
1989	/*
1990	* Receive a frame from the socket and optionally record the address of the
1991	* sender. We verify the buffers are writable and if needed move the
1992	* sender address from kernel to user space.
1993	*/
1994	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1995	struct sockaddr __user addr, int __user addr_len)
1996	{
1997	struct socket *sock;
1998	struct iovec iov;
1999	struct msghdr msg;
2000	struct sockaddr_storage address;
2001	int err, err2;
2002	int fput_needed;
2003
2004	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2005	if (unlikely(err))
2006	return err;
2007	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2008	if (!sock)
2009	goto out;
2010
2011	msg.msg_control = NULL;
2012	msg.msg_controllen = 0;
2013	/* Save some cycles and don't copy the address if not needed */
2014	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2015	/* We assume all kernel code knows the size of sockaddr_storage */
2016	msg.msg_namelen = 0;
2017	msg.msg_iocb = NULL;
2018	msg.msg_flags = 0;
2019	if (sock->file->f_flags & O_NONBLOCK)
2020	flags \|= MSG_DONTWAIT;
2021	err = sock_recvmsg(sock, &msg, flags);
2022
2023	if (err >= 0 && addr != NULL) {
2024	err2 = move_addr_to_user(&address,
2025	msg.msg_namelen, addr, addr_len);
2026	if (err2 < 0)
2027	err = err2;
2028	}
2029
2030	fput_light(sock->file, fput_needed);
2031	out:
2032	return err;
2033	}
2034
2035	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2036	unsigned int, flags, struct sockaddr __user *, addr,
2037	int __user *, addr_len)
2038	{
2039	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2040	}
2041
2042	/*
2043	* Receive a datagram from a socket.
2044	*/
2045
2046	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2047	unsigned int, flags)
2048	{
2049	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2050	}
2051
2052	static bool sock_use_custom_sol_socket(const struct socket *sock)
2053	{
2054	const struct sock *sk = sock->sk;
2055
2056	/* Use sock->ops->setsockopt() for MPTCP */
2057	return IS_ENABLED(CONFIG_MPTCP) &&
2058	sk->sk_protocol == IPPROTO_MPTCP &&
2059	sk->sk_type == SOCK_STREAM &&
2060	(sk->sk_family == AF_INET \|\| sk->sk_family == AF_INET6);
2061	}
2062
2063	/*
2064	* Set a socket option. Because we don't know the option lengths we have
2065	* to pass the user mode parameter for the protocols to sort out.
2066	*/
2067	int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2068	int optlen)
2069	{
2070	sockptr_t optval = USER_SOCKPTR(user_optval);
2071	char *kernel_optval = NULL;
2072	int err, fput_needed;
2073	struct socket *sock;
2074
2075	if (optlen < 0)
2076	return -EINVAL;
2077
2078	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2079	if (!sock)
2080	return err;
2081
2082	err = security_socket_setsockopt(sock, level, optname);
2083	if (err)
2084	goto out_put;
2085
2086	if (!in_compat_syscall())
2087	err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2088	user_optval, &optlen,
2089	&kernel_optval);
2090	if (err < 0)
2091	goto out_put;
2092	if (err > 0) {
2093	err = 0;
2094	goto out_put;
2095	}
2096
2097	if (kernel_optval)
2098	optval = KERNEL_SOCKPTR(kernel_optval);
2099	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2100	err = sock_setsockopt(sock, level, optname, optval, optlen);
2101	else if (unlikely(!sock->ops->setsockopt))
2102	err = -EOPNOTSUPP;
2103	else
2104	err = sock->ops->setsockopt(sock, level, optname, optval,
2105	optlen);
2106	kfree(kernel_optval);
2107	out_put:
2108	fput_light(sock->file, fput_needed);
2109	return err;
2110	}
2111
2112	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2113	char __user *, optval, int, optlen)
2114	{
2115	return __sys_setsockopt(fd, level, optname, optval, optlen);
2116	}
2117
2118	INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2119	int optname));
2120
2121	/*
2122	* Get a socket option. Because we don't know the option lengths we have
2123	* to pass a user mode parameter for the protocols to sort out.
2124	*/
2125	int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2126	int __user *optlen)
2127	{
2128	int err, fput_needed;
2129	struct socket *sock;
2130	int max_optlen;
2131
2132	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2133	if (!sock)
2134	return err;
2135
2136	err = security_socket_getsockopt(sock, level, optname);
2137	if (err)
2138	goto out_put;
2139
2140	if (!in_compat_syscall())
2141	max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2142
2143	if (level == SOL_SOCKET)
2144	err = sock_getsockopt(sock, level, optname, optval, optlen);
2145	else if (unlikely(!sock->ops->getsockopt))
2146	err = -EOPNOTSUPP;
2147	else
2148	err = sock->ops->getsockopt(sock, level, optname, optval,
2149	optlen);
2150
2151	if (!in_compat_syscall())
2152	err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2153	optval, optlen, max_optlen,
2154	err);
2155	out_put:
2156	fput_light(sock->file, fput_needed);
2157	return err;
2158	}
2159
2160	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2161	char __user , optval, int __user , optlen)
2162	{
2163	return __sys_getsockopt(fd, level, optname, optval, optlen);
2164	}
2165
2166	/*
2167	* Shutdown a socket.
2168	*/
2169
2170	int __sys_shutdown_sock(struct socket *sock, int how)
2171	{
2172	int err;
2173
2174	err = security_socket_shutdown(sock, how);
2175	if (!err)
2176	err = sock->ops->shutdown(sock, how);
2177
2178	return err;
2179	}
2180
2181	int __sys_shutdown(int fd, int how)
2182	{
2183	int err, fput_needed;
2184	struct socket *sock;
2185
2186	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2187	if (sock != NULL) {
2188	err = __sys_shutdown_sock(sock, how);
2189	fput_light(sock->file, fput_needed);
2190	}
2191	return err;
2192	}
2193
2194	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2195	{
2196	return __sys_shutdown(fd, how);
2197	}
2198
2199	/* A couple of helpful macros for getting the address of the 32/64 bit
2200	* fields which are the same type (int / unsigned) on our platforms.
2201	*/
2202	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2203	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2204	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2205
2206	struct used_address {
2207	struct sockaddr_storage name;
2208	unsigned int name_len;
2209	};
2210
2211	int __copy_msghdr_from_user(struct msghdr *kmsg,
2212	struct user_msghdr __user *umsg,
2213	struct sockaddr __user **save_addr,
2214	struct iovec __user *uiov, size_t nsegs)
2215	{
2216	struct user_msghdr msg;
2217	ssize_t err;
2218
2219	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2220	return -EFAULT;
2221
2222	kmsg->msg_control_is_user = true;
2223	kmsg->msg_control_user = msg.msg_control;
2224	kmsg->msg_controllen = msg.msg_controllen;
2225	kmsg->msg_flags = msg.msg_flags;
2226
2227	kmsg->msg_namelen = msg.msg_namelen;
2228	if (!msg.msg_name)
2229	kmsg->msg_namelen = 0;
2230
2231	if (kmsg->msg_namelen < 0)
2232	return -EINVAL;
2233
2234	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2235	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2236
2237	if (save_addr)
2238	*save_addr = msg.msg_name;
2239
2240	if (msg.msg_name && kmsg->msg_namelen) {
2241	if (!save_addr) {
2242	err = move_addr_to_kernel(msg.msg_name,
2243	kmsg->msg_namelen,
2244	kmsg->msg_name);
2245	if (err < 0)
2246	return err;
2247	}
2248	} else {
2249	kmsg->msg_name = NULL;
2250	kmsg->msg_namelen = 0;
2251	}
2252
2253	if (msg.msg_iovlen > UIO_MAXIOV)
2254	return -EMSGSIZE;
2255
2256	kmsg->msg_iocb = NULL;
2257	*uiov = msg.msg_iov;
2258	*nsegs = msg.msg_iovlen;
2259	return 0;
2260	}
2261
2262	static int copy_msghdr_from_user(struct msghdr *kmsg,
2263	struct user_msghdr __user *umsg,
2264	struct sockaddr __user **save_addr,
2265	struct iovec **iov)
2266	{
2267	struct user_msghdr msg;
2268	ssize_t err;
2269
2270	err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2271	&msg.msg_iovlen);
2272	if (err)
2273	return err;
2274
2275	err = import_iovec(save_addr ? READ : WRITE,
2276	msg.msg_iov, msg.msg_iovlen,
2277	UIO_FASTIOV, iov, &kmsg->msg_iter);
2278	return err < 0 ? err : 0;
2279	}
2280
2281	static int ____sys_sendmsg(struct socket sock, struct msghdr msg_sys,
2282	unsigned int flags, struct used_address *used_address,
2283	unsigned int allowed_msghdr_flags)
2284	{
2285	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2286	__aligned(sizeof(__kernel_size_t));
2287	/* 20 is size of ipv6_pktinfo */
2288	unsigned char *ctl_buf = ctl;
2289	int ctl_len;
2290	ssize_t err;
2291
2292	err = -ENOBUFS;
2293
2294	if (msg_sys->msg_controllen > INT_MAX)
2295	goto out;
2296	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2297	ctl_len = msg_sys->msg_controllen;
2298	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2299	err =
2300	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2301	sizeof(ctl));
2302	if (err)
2303	goto out;
2304	ctl_buf = msg_sys->msg_control;
2305	ctl_len = msg_sys->msg_controllen;
2306	} else if (ctl_len) {
2307	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2308	CMSG_ALIGN(sizeof(struct cmsghdr)));
2309	if (ctl_len > sizeof(ctl)) {
2310	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2311	if (ctl_buf == NULL)
2312	goto out;
2313	}
2314	err = -EFAULT;
2315	if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2316	goto out_freectl;
2317	msg_sys->msg_control = ctl_buf;
2318	msg_sys->msg_control_is_user = false;
2319	}
2320	msg_sys->msg_flags = flags;
2321
2322	if (sock->file->f_flags & O_NONBLOCK)
2323	msg_sys->msg_flags \|= MSG_DONTWAIT;
2324	/*
2325	* If this is sendmmsg() and current destination address is same as
2326	* previously succeeded address, omit asking LSM's decision.
2327	* used_address->name_len is initialized to UINT_MAX so that the first
2328	* destination address never matches.
2329	*/
2330	if (used_address && msg_sys->msg_name &&
2331	used_address->name_len == msg_sys->msg_namelen &&
2332	!memcmp(&used_address->name, msg_sys->msg_name,
2333	used_address->name_len)) {
2334	err = sock_sendmsg_nosec(sock, msg_sys);
2335	goto out_freectl;
2336	}
2337	err = sock_sendmsg(sock, msg_sys);
2338	/*
2339	* If this is sendmmsg() and sending to current destination address was
2340	* successful, remember it.
2341	*/
2342	if (used_address && err >= 0) {
2343	used_address->name_len = msg_sys->msg_namelen;
2344	if (msg_sys->msg_name)
2345	memcpy(&used_address->name, msg_sys->msg_name,
2346	used_address->name_len);
2347	}
2348
2349	out_freectl:
2350	if (ctl_buf != ctl)
2351	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2352	out:
2353	return err;
2354	}
2355
2356	int sendmsg_copy_msghdr(struct msghdr *msg,
2357	struct user_msghdr __user *umsg, unsigned flags,
2358	struct iovec **iov)
2359	{
2360	int err;
2361
2362	if (flags & MSG_CMSG_COMPAT) {
2363	struct compat_msghdr __user *msg_compat;
2364
2365	msg_compat = (struct compat_msghdr __user *) umsg;
2366	err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2367	} else {
2368	err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2369	}
2370	if (err < 0)
2371	return err;
2372
2373	return 0;
2374	}
2375
2376	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
2377	struct msghdr *msg_sys, unsigned int flags,
2378	struct used_address *used_address,
2379	unsigned int allowed_msghdr_flags)
2380	{
2381	struct sockaddr_storage address;
2382	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2383	ssize_t err;
2384
2385	msg_sys->msg_name = &address;
2386
2387	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2388	if (err < 0)
2389	return err;
2390
2391	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2392	allowed_msghdr_flags);
2393	kfree(iov);
2394	return err;
2395	}
2396
2397	/*
2398	* BSD sendmsg interface
2399	*/
2400	long __sys_sendmsg_sock(struct socket sock, struct msghdr msg,
2401	unsigned int flags)
2402	{
2403	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2404	}
2405
2406	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2407	bool forbid_cmsg_compat)
2408	{
2409	int fput_needed, err;
2410	struct msghdr msg_sys;
2411	struct socket *sock;
2412
2413	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2414	return -EINVAL;
2415
2416	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2417	if (!sock)
2418	goto out;
2419
2420	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2421
2422	fput_light(sock->file, fput_needed);
2423	out:
2424	return err;
2425	}
2426
2427	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2428	{
2429	return __sys_sendmsg(fd, msg, flags, true);
2430	}
2431
2432	/*
2433	* Linux sendmmsg interface
2434	*/
2435
2436	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2437	unsigned int flags, bool forbid_cmsg_compat)
2438	{
2439	int fput_needed, err, datagrams;
2440	struct socket *sock;
2441	struct mmsghdr __user *entry;
2442	struct compat_mmsghdr __user *compat_entry;
2443	struct msghdr msg_sys;
2444	struct used_address used_address;
2445	unsigned int oflags = flags;
2446
2447	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2448	return -EINVAL;
2449
2450	if (vlen > UIO_MAXIOV)
2451	vlen = UIO_MAXIOV;
2452
2453	datagrams = 0;
2454
2455	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2456	if (!sock)
2457	return err;
2458
2459	used_address.name_len = UINT_MAX;
2460	entry = mmsg;
2461	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2462	err = 0;
2463	flags \|= MSG_BATCH;
2464
2465	while (datagrams < vlen) {
2466	if (datagrams == vlen - 1)
2467	flags = oflags;
2468
2469	if (MSG_CMSG_COMPAT & flags) {
2470	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2471	&msg_sys, flags, &used_address, MSG_EOR);
2472	if (err < 0)
2473	break;
2474	err = __put_user(err, &compat_entry->msg_len);
2475	++compat_entry;
2476	} else {
2477	err = ___sys_sendmsg(sock,
2478	(struct user_msghdr __user *)entry,
2479	&msg_sys, flags, &used_address, MSG_EOR);
2480	if (err < 0)
2481	break;
2482	err = put_user(err, &entry->msg_len);
2483	++entry;
2484	}
2485
2486	if (err)
2487	break;
2488	++datagrams;
2489	if (msg_data_left(&msg_sys))
2490	break;
2491	cond_resched();
2492	}
2493
2494	fput_light(sock->file, fput_needed);
2495
2496	/* We only return an error if no datagrams were able to be sent */
2497	if (datagrams != 0)
2498	return datagrams;
2499
2500	return err;
2501	}
2502
2503	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2504	unsigned int, vlen, unsigned int, flags)
2505	{
2506	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2507	}
2508
2509	int recvmsg_copy_msghdr(struct msghdr *msg,
2510	struct user_msghdr __user *umsg, unsigned flags,
2511	struct sockaddr __user **uaddr,
2512	struct iovec **iov)
2513	{
2514	ssize_t err;
2515
2516	if (MSG_CMSG_COMPAT & flags) {
2517	struct compat_msghdr __user *msg_compat;
2518
2519	msg_compat = (struct compat_msghdr __user *) umsg;
2520	err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2521	} else {
2522	err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2523	}
2524	if (err < 0)
2525	return err;
2526
2527	return 0;
2528	}
2529
2530	static int ____sys_recvmsg(struct socket sock, struct msghdr msg_sys,
2531	struct user_msghdr __user *msg,
2532	struct sockaddr __user *uaddr,
2533	unsigned int flags, int nosec)
2534	{
2535	struct compat_msghdr __user *msg_compat =
2536	(struct compat_msghdr __user *) msg;
2537	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2538	struct sockaddr_storage addr;
2539	unsigned long cmsg_ptr;
2540	int len;
2541	ssize_t err;
2542
2543	msg_sys->msg_name = &addr;
2544	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2545	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2546
2547	/* We assume all kernel code knows the size of sockaddr_storage */
2548	msg_sys->msg_namelen = 0;
2549
2550	if (sock->file->f_flags & O_NONBLOCK)
2551	flags \|= MSG_DONTWAIT;
2552
2553	if (unlikely(nosec))
2554	err = sock_recvmsg_nosec(sock, msg_sys, flags);
2555	else
2556	err = sock_recvmsg(sock, msg_sys, flags);
2557
2558	if (err < 0)
2559	goto out;
2560	len = err;
2561
2562	if (uaddr != NULL) {
2563	err = move_addr_to_user(&addr,
2564	msg_sys->msg_namelen, uaddr,
2565	uaddr_len);
2566	if (err < 0)
2567	goto out;
2568	}
2569	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2570	COMPAT_FLAGS(msg));
2571	if (err)
2572	goto out;
2573	if (MSG_CMSG_COMPAT & flags)
2574	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2575	&msg_compat->msg_controllen);
2576	else
2577	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2578	&msg->msg_controllen);
2579	if (err)
2580	goto out;
2581	err = len;
2582	out:
2583	return err;
2584	}
2585
2586	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2587	struct msghdr *msg_sys, unsigned int flags, int nosec)
2588	{
2589	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2590	/* user mode address pointers */
2591	struct sockaddr __user *uaddr;
2592	ssize_t err;
2593
2594	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2595	if (err < 0)
2596	return err;
2597
2598	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2599	kfree(iov);
2600	return err;
2601	}
2602
2603	/*
2604	* BSD recvmsg interface
2605	*/
2606
2607	long __sys_recvmsg_sock(struct socket sock, struct msghdr msg,
2608	struct user_msghdr __user *umsg,
2609	struct sockaddr __user *uaddr, unsigned int flags)
2610	{
2611	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2612	}
2613
2614	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2615	bool forbid_cmsg_compat)
2616	{
2617	int fput_needed, err;
2618	struct msghdr msg_sys;
2619	struct socket *sock;
2620
2621	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2622	return -EINVAL;
2623
2624	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2625	if (!sock)
2626	goto out;
2627
2628	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2629
2630	fput_light(sock->file, fput_needed);
2631	out:
2632	return err;
2633	}
2634
2635	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2636	unsigned int, flags)
2637	{
2638	return __sys_recvmsg(fd, msg, flags, true);
2639	}
2640
2641	/*
2642	* Linux recvmmsg interface
2643	*/
2644
2645	static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2646	unsigned int vlen, unsigned int flags,
2647	struct timespec64 *timeout)
2648	{
2649	int fput_needed, err, datagrams;
2650	struct socket *sock;
2651	struct mmsghdr __user *entry;
2652	struct compat_mmsghdr __user *compat_entry;
2653	struct msghdr msg_sys;
2654	struct timespec64 end_time;
2655	struct timespec64 timeout64;
2656
2657	if (timeout &&
2658	poll_select_set_timeout(&end_time, timeout->tv_sec,
2659	timeout->tv_nsec))
2660	return -EINVAL;
2661
2662	datagrams = 0;
2663
2664	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2665	if (!sock)
2666	return err;
2667
2668	if (likely(!(flags & MSG_ERRQUEUE))) {
2669	err = sock_error(sock->sk);
2670	if (err) {
2671	datagrams = err;
2672	goto out_put;
2673	}
2674	}
2675
2676	entry = mmsg;
2677	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2678
2679	while (datagrams < vlen) {
2680	/*
2681	* No need to ask LSM for more than the first datagram.
2682	*/
2683	if (MSG_CMSG_COMPAT & flags) {
2684	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2685	&msg_sys, flags & ~MSG_WAITFORONE,
2686	datagrams);
2687	if (err < 0)
2688	break;
2689	err = __put_user(err, &compat_entry->msg_len);
2690	++compat_entry;
2691	} else {
2692	err = ___sys_recvmsg(sock,
2693	(struct user_msghdr __user *)entry,
2694	&msg_sys, flags & ~MSG_WAITFORONE,
2695	datagrams);
2696	if (err < 0)
2697	break;
2698	err = put_user(err, &entry->msg_len);
2699	++entry;
2700	}
2701
2702	if (err)
2703	break;
2704	++datagrams;
2705
2706	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2707	if (flags & MSG_WAITFORONE)
2708	flags \|= MSG_DONTWAIT;
2709
2710	if (timeout) {
2711	ktime_get_ts64(&timeout64);
2712	*timeout = timespec64_sub(end_time, timeout64);
2713	if (timeout->tv_sec < 0) {
2714	timeout->tv_sec = timeout->tv_nsec = 0;
2715	break;
2716	}
2717
2718	/* Timeout, return less than vlen datagrams */
2719	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2720	break;
2721	}
2722
2723	/* Out of band data, return right away */
2724	if (msg_sys.msg_flags & MSG_OOB)
2725	break;
2726	cond_resched();
2727	}
2728
2729	if (err == 0)
2730	goto out_put;
2731
2732	if (datagrams == 0) {
2733	datagrams = err;
2734	goto out_put;
2735	}
2736
2737	/*
2738	* We may return less entries than requested (vlen) if the
2739	* sock is non block and there aren't enough datagrams...
2740	*/
2741	if (err != -EAGAIN) {
2742	/*
2743	* ... or if recvmsg returns an error after we
2744	* received some datagrams, where we record the
2745	* error to return on the next call or if the
2746	* app asks about it using getsockopt(SO_ERROR).
2747	*/
2748	sock->sk->sk_err = -err;
2749	}
2750	out_put:
2751	fput_light(sock->file, fput_needed);
2752
2753	return datagrams;
2754	}
2755
2756	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2757	unsigned int vlen, unsigned int flags,
2758	struct __kernel_timespec __user *timeout,
2759	struct old_timespec32 __user *timeout32)
2760	{
2761	int datagrams;
2762	struct timespec64 timeout_sys;
2763
2764	if (timeout && get_timespec64(&timeout_sys, timeout))
2765	return -EFAULT;
2766
2767	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2768	return -EFAULT;
2769
2770	if (!timeout && !timeout32)
2771	return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2772
2773	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2774
2775	if (datagrams <= 0)
2776	return datagrams;
2777
2778	if (timeout && put_timespec64(&timeout_sys, timeout))
2779	datagrams = -EFAULT;
2780
2781	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2782	datagrams = -EFAULT;
2783
2784	return datagrams;
2785	}
2786
2787	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2788	unsigned int, vlen, unsigned int, flags,
2789	struct __kernel_timespec __user *, timeout)
2790	{
2791	if (flags & MSG_CMSG_COMPAT)
2792	return -EINVAL;
2793
2794	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2795	}
2796
2797	#ifdef CONFIG_COMPAT_32BIT_TIME
2798	SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2799	unsigned int, vlen, unsigned int, flags,
2800	struct old_timespec32 __user *, timeout)
2801	{
2802	if (flags & MSG_CMSG_COMPAT)
2803	return -EINVAL;
2804
2805	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2806	}
2807	#endif
2808
2809	#ifdef __ARCH_WANT_SYS_SOCKETCALL
2810	/* Argument list sizes for sys_socketcall */
2811	#define AL(x) ((x) * sizeof(unsigned long))
2812	static const unsigned char nargs[21] = {
2813	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2814	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2815	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2816	AL(4), AL(5), AL(4)
2817	};
2818
2819	#undef AL
2820
2821	/*
2822	* System call vectors.
2823	*
2824	* Argument checking cleaned up. Saved 20% in size.
2825	* This function doesn't need to set the kernel lock because
2826	* it is set by the callees.
2827	*/
2828
2829	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2830	{
2831	unsigned long a[AUDITSC_ARGS];
2832	unsigned long a0, a1;
2833	int err;
2834	unsigned int len;
2835
2836	if (call < 1 \|\| call > SYS_SENDMMSG)
2837	return -EINVAL;
2838	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2839
2840	len = nargs[call];
2841	if (len > sizeof(a))
2842	return -EINVAL;
2843
2844	/* copy_from_user should be SMP safe. */
2845	if (copy_from_user(a, args, len))
2846	return -EFAULT;
2847
2848	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2849	if (err)
2850	return err;
2851
2852	a0 = a[0];
2853	a1 = a[1];
2854
2855	switch (call) {
2856	case SYS_SOCKET:
2857	err = __sys_socket(a0, a1, a[2]);
2858	break;
2859	case SYS_BIND:
2860	err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2861	break;
2862	case SYS_CONNECT:
2863	err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2864	break;
2865	case SYS_LISTEN:
2866	err = __sys_listen(a0, a1);
2867	break;
2868	case SYS_ACCEPT:
2869	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2870	(int __user *)a[2], 0);
2871	break;
2872	case SYS_GETSOCKNAME:
2873	err =
2874	__sys_getsockname(a0, (struct sockaddr __user *)a1,
2875	(int __user *)a[2]);
2876	break;
2877	case SYS_GETPEERNAME:
2878	err =
2879	__sys_getpeername(a0, (struct sockaddr __user *)a1,
2880	(int __user *)a[2]);
2881	break;
2882	case SYS_SOCKETPAIR:
2883	err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2884	break;
2885	case SYS_SEND:
2886	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2887	NULL, 0);
2888	break;
2889	case SYS_SENDTO:
2890	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2891	(struct sockaddr __user *)a[4], a[5]);
2892	break;
2893	case SYS_RECV:
2894	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2895	NULL, NULL);
2896	break;
2897	case SYS_RECVFROM:
2898	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2899	(struct sockaddr __user *)a[4],
2900	(int __user *)a[5]);
2901	break;
2902	case SYS_SHUTDOWN:
2903	err = __sys_shutdown(a0, a1);
2904	break;
2905	case SYS_SETSOCKOPT:
2906	err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2907	a[4]);
2908	break;
2909	case SYS_GETSOCKOPT:
2910	err =
2911	__sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2912	(int __user *)a[4]);
2913	break;
2914	case SYS_SENDMSG:
2915	err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2916	a[2], true);
2917	break;
2918	case SYS_SENDMMSG:
2919	err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2920	a[3], true);
2921	break;
2922	case SYS_RECVMSG:
2923	err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2924	a[2], true);
2925	break;
2926	case SYS_RECVMMSG:
2927	if (IS_ENABLED(CONFIG_64BIT))
2928	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2929	a[2], a[3],
2930	(struct __kernel_timespec __user *)a[4],
2931	NULL);
2932	else
2933	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2934	a[2], a[3], NULL,
2935	(struct old_timespec32 __user *)a[4]);
2936	break;
2937	case SYS_ACCEPT4:
2938	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2939	(int __user *)a[2], a[3]);
2940	break;
2941	default:
2942	err = -EINVAL;
2943	break;
2944	}
2945	return err;
2946	}
2947
2948	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2949
2950	/**
2951	* sock_register - add a socket protocol handler
2952	* @ops: description of protocol
2953	*
2954	* This function is called by a protocol handler that wants to
2955	* advertise its address family, and have it linked into the
2956	* socket interface. The value ops->family corresponds to the
2957	* socket system call protocol family.
2958	*/
2959	int sock_register(const struct net_proto_family *ops)
2960	{
2961	int err;
2962
2963	if (ops->family >= NPROTO) {
2964	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2965	return -ENOBUFS;
2966	}
2967
2968	spin_lock(&net_family_lock);
2969	if (rcu_dereference_protected(net_families[ops->family],
2970	lockdep_is_held(&net_family_lock)))
2971	err = -EEXIST;
2972	else {
2973	rcu_assign_pointer(net_families[ops->family], ops);
2974	err = 0;
2975	}
2976	spin_unlock(&net_family_lock);
2977
2978	pr_info("NET: Registered protocol family %d\n", ops->family);
2979	return err;
2980	}
2981	EXPORT_SYMBOL(sock_register);
2982
2983	/**
2984	* sock_unregister - remove a protocol handler
2985	* @family: protocol family to remove
2986	*
2987	* This function is called by a protocol handler that wants to
2988	* remove its address family, and have it unlinked from the
2989	* new socket creation.
2990	*
2991	* If protocol handler is a module, then it can use module reference
2992	* counts to protect against new references. If protocol handler is not
2993	* a module then it needs to provide its own protection in
2994	* the ops->create routine.
2995	*/
2996	void sock_unregister(int family)
2997	{
2998	BUG_ON(family < 0 \|\| family >= NPROTO);
2999
3000	spin_lock(&net_family_lock);
3001	RCU_INIT_POINTER(net_families[family], NULL);
3002	spin_unlock(&net_family_lock);
3003
3004	synchronize_rcu();
3005
3006	pr_info("NET: Unregistered protocol family %d\n", family);
3007	}
3008	EXPORT_SYMBOL(sock_unregister);
3009
3010	bool sock_is_registered(int family)
3011	{
3012	return family < NPROTO && rcu_access_pointer(net_families[family]);
3013	}
3014
3015	static int __init sock_init(void)
3016	{
3017	int err;
3018	/*
3019	* Initialize the network sysctl infrastructure.
3020	*/
3021	err = net_sysctl_init();
3022	if (err)
3023	goto out;
3024
3025	/*
3026	* Initialize skbuff SLAB cache
3027	*/
3028	skb_init();
3029
3030	/*
3031	* Initialize the protocols module.
3032	*/
3033
3034	init_inodecache();
3035
3036	err = register_filesystem(&sock_fs_type);
3037	if (err)
3038	goto out;
3039	sock_mnt = kern_mount(&sock_fs_type);
3040	if (IS_ERR(sock_mnt)) {
3041	err = PTR_ERR(sock_mnt);
3042	goto out_mount;
3043	}
3044
3045	/* The real protocol initialization is performed in later initcalls.
3046	*/
3047
3048	#ifdef CONFIG_NETFILTER
3049	err = netfilter_init();
3050	if (err)
3051	goto out;
3052	#endif
3053
3054	ptp_classifier_init();
3055
3056	out:
3057	return err;
3058
3059	out_mount:
3060	unregister_filesystem(&sock_fs_type);
3061	goto out;
3062	}
3063
3064	core_initcall(sock_init); /* early initcall */
3065
3066	#ifdef CONFIG_PROC_FS
3067	void socket_seq_show(struct seq_file *seq)
3068	{
3069	seq_printf(seq, "sockets: used %d\n",
3070	sock_inuse_get(seq->private));
3071	}
3072	#endif /* CONFIG_PROC_FS */
3073
3074	#ifdef CONFIG_COMPAT
3075	static int compat_dev_ifconf(struct net net, struct compat_ifconf __user uifc32)
3076	{
3077	struct compat_ifconf ifc32;
3078	struct ifconf ifc;
3079	int err;
3080
3081	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
3082	return -EFAULT;
3083
3084	ifc.ifc_len = ifc32.ifc_len;
3085	ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
3086
3087	rtnl_lock();
3088	err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3089	rtnl_unlock();
3090	if (err)
3091	return err;
3092
3093	ifc32.ifc_len = ifc.ifc_len;
3094	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
3095	return -EFAULT;
3096
3097	return 0;
3098	}
3099
3100	static int ethtool_ioctl(struct net net, struct compat_ifreq __user ifr32)
3101	{
3102	struct compat_ethtool_rxnfc __user *compat_rxnfc;
3103	bool convert_in = false, convert_out = false;
3104	size_t buf_size = 0;
3105	struct ethtool_rxnfc __user *rxnfc = NULL;
3106	struct ifreq ifr;
3107	u32 rule_cnt = 0, actual_rule_cnt;
3108	u32 ethcmd;
3109	u32 data;
3110	int ret;
3111
3112	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3113	return -EFAULT;
3114
3115	compat_rxnfc = compat_ptr(data);
3116
3117	if (get_user(ethcmd, &compat_rxnfc->cmd))
3118	return -EFAULT;
3119
3120	/* Most ethtool structures are defined without padding.
3121	* Unfortunately struct ethtool_rxnfc is an exception.
3122	*/
3123	switch (ethcmd) {
3124	default:
3125	break;
3126	case ETHTOOL_GRXCLSRLALL:
3127	/* Buffer size is variable */
3128	if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3129	return -EFAULT;
3130	if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3131	return -ENOMEM;
3132	buf_size += rule_cnt * sizeof(u32);
3133	fallthrough;
3134	case ETHTOOL_GRXRINGS:
3135	case ETHTOOL_GRXCLSRLCNT:
3136	case ETHTOOL_GRXCLSRULE:
3137	case ETHTOOL_SRXCLSRLINS:
3138	convert_out = true;
3139	fallthrough;
3140	case ETHTOOL_SRXCLSRLDEL:
3141	buf_size += sizeof(struct ethtool_rxnfc);
3142	convert_in = true;
3143	rxnfc = compat_alloc_user_space(buf_size);
3144	break;
3145	}
3146
3147	if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
3148	return -EFAULT;
3149
3150	ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
3151
3152	if (convert_in) {
3153	/* We expect there to be holes between fs.m_ext and
3154	* fs.ring_cookie and at the end of fs, but nowhere else.
3155	*/
3156	BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3157	sizeof(compat_rxnfc->fs.m_ext) !=
3158	offsetof(struct ethtool_rxnfc, fs.m_ext) +
3159	sizeof(rxnfc->fs.m_ext));
3160	BUILD_BUG_ON(
3161	offsetof(struct compat_ethtool_rxnfc, fs.location) -
3162	offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3163	offsetof(struct ethtool_rxnfc, fs.location) -
3164	offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3165
3166	if (copy_in_user(rxnfc, compat_rxnfc,
3167	(void __user *)(&rxnfc->fs.m_ext + 1) -
3168	(void __user *)rxnfc) \|\|
3169	copy_in_user(&rxnfc->fs.ring_cookie,
3170	&compat_rxnfc->fs.ring_cookie,
3171	(void __user *)(&rxnfc->fs.location + 1) -
3172	(void __user *)&rxnfc->fs.ring_cookie))
3173	return -EFAULT;
3174	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3175	if (put_user(rule_cnt, &rxnfc->rule_cnt))
3176	return -EFAULT;
3177	} else if (copy_in_user(&rxnfc->rule_cnt,
3178	&compat_rxnfc->rule_cnt,
3179	sizeof(rxnfc->rule_cnt)))
3180	return -EFAULT;
3181	}
3182
3183	ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
3184	if (ret)
3185	return ret;
3186
3187	if (convert_out) {
3188	if (copy_in_user(compat_rxnfc, rxnfc,
3189	(const void __user *)(&rxnfc->fs.m_ext + 1) -
3190	(const void __user *)rxnfc) \|\|
3191	copy_in_user(&compat_rxnfc->fs.ring_cookie,
3192	&rxnfc->fs.ring_cookie,
3193	(const void __user *)(&rxnfc->fs.location + 1) -
3194	(const void __user *)&rxnfc->fs.ring_cookie) \|\|
3195	copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3196	sizeof(rxnfc->rule_cnt)))
3197	return -EFAULT;
3198
3199	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3200	/* As an optimisation, we only copy the actual
3201	* number of rules that the underlying
3202	* function returned. Since Mallory might
3203	* change the rule count in user memory, we
3204	* check that it is less than the rule count
3205	* originally given (as the user buffer size),
3206	* which has been range-checked.
3207	*/
3208	if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3209	return -EFAULT;
3210	if (actual_rule_cnt < rule_cnt)
3211	rule_cnt = actual_rule_cnt;
3212	if (copy_in_user(&compat_rxnfc->rule_locs[0],
3213	&rxnfc->rule_locs[0],
3214	rule_cnt * sizeof(u32)))
3215	return -EFAULT;
3216	}
3217	}
3218
3219	return 0;
3220	}
3221
3222	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
3223	{
3224	compat_uptr_t uptr32;
3225	struct ifreq ifr;
3226	void __user *saved;
3227	int err;
3228
3229	if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
3230	return -EFAULT;
3231
3232	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3233	return -EFAULT;
3234
3235	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3236	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3237
3238	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3239	if (!err) {
3240	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3241	if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3242	err = -EFAULT;
3243	}
3244	return err;
3245	}
3246
3247	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3248	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3249	struct compat_ifreq __user *u_ifreq32)
3250	{
3251	struct ifreq ifreq;
3252	u32 data32;
3253
3254	if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
3255	return -EFAULT;
3256	if (get_user(data32, &u_ifreq32->ifr_data))
3257	return -EFAULT;
3258	ifreq.ifr_data = compat_ptr(data32);
3259
3260	return dev_ioctl(net, cmd, &ifreq, NULL);
3261	}
3262
3263	static int compat_ifreq_ioctl(struct net net, struct socket sock,
3264	unsigned int cmd,
3265	struct compat_ifreq __user *uifr32)
3266	{
3267	struct ifreq __user *uifr;
3268	int err;
3269
3270	/* Handle the fact that while struct ifreq has the same layout on
3271	* 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3272	* which are handled elsewhere, it still has different size due to
3273	* ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3274	* resulting in struct ifreq being 32 and 40 bytes respectively).
3275	* As a result, if the struct happens to be at the end of a page and
3276	* the next page isn't readable/writable, we get a fault. To prevent
3277	* that, copy back and forth to the full size.
3278	*/
3279
3280	uifr = compat_alloc_user_space(sizeof(*uifr));
3281	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3282	return -EFAULT;
3283
3284	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3285
3286	if (!err) {
3287	switch (cmd) {
3288	case SIOCGIFFLAGS:
3289	case SIOCGIFMETRIC:
3290	case SIOCGIFMTU:
3291	case SIOCGIFMEM:
3292	case SIOCGIFHWADDR:
3293	case SIOCGIFINDEX:
3294	case SIOCGIFADDR:
3295	case SIOCGIFBRDADDR:
3296	case SIOCGIFDSTADDR:
3297	case SIOCGIFNETMASK:
3298	case SIOCGIFPFLAGS:
3299	case SIOCGIFTXQLEN:
3300	case SIOCGMIIPHY:
3301	case SIOCGMIIREG:
3302	case SIOCGIFNAME:
3303	if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3304	err = -EFAULT;
3305	break;
3306	}
3307	}
3308	return err;
3309	}
3310
3311	static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3312	struct compat_ifreq __user *uifr32)
3313	{
3314	struct ifreq ifr;
3315	struct compat_ifmap __user *uifmap32;
3316	int err;
3317
3318	uifmap32 = &uifr32->ifr_ifru.ifru_map;
3319	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3320	err \|= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3321	err \|= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3322	err \|= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3323	err \|= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3324	err \|= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3325	err \|= get_user(ifr.ifr_map.port, &uifmap32->port);
3326	if (err)
3327	return -EFAULT;
3328
3329	err = dev_ioctl(net, cmd, &ifr, NULL);
3330
3331	if (cmd == SIOCGIFMAP && !err) {
3332	err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3333	err \|= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3334	err \|= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3335	err \|= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3336	err \|= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3337	err \|= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3338	err \|= put_user(ifr.ifr_map.port, &uifmap32->port);
3339	if (err)
3340	err = -EFAULT;
3341	}
3342	return err;
3343	}
3344
3345	/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3346	* for some operations; this forces use of the newer bridge-utils that
3347	* use compatible ioctls
3348	*/
3349	static int old_bridge_ioctl(compat_ulong_t __user *argp)
3350	{
3351	compat_ulong_t tmp;
3352
3353	if (get_user(tmp, argp))
3354	return -EFAULT;
3355	if (tmp == BRCTL_GET_VERSION)
3356	return BRCTL_VERSION + 1;
3357	return -EINVAL;
3358	}
3359
3360	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3361	unsigned int cmd, unsigned long arg)
3362	{
3363	void __user *argp = compat_ptr(arg);
3364	struct sock *sk = sock->sk;
3365	struct net *net = sock_net(sk);
3366
3367	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3368	return compat_ifr_data_ioctl(net, cmd, argp);
3369
3370	switch (cmd) {
3371	case SIOCSIFBR:
3372	case SIOCGIFBR:
3373	return old_bridge_ioctl(argp);
3374	case SIOCGIFCONF:
3375	return compat_dev_ifconf(net, argp);
3376	case SIOCETHTOOL:
3377	return ethtool_ioctl(net, argp);
3378	case SIOCWANDEV:
3379	return compat_siocwandev(net, argp);
3380	case SIOCGIFMAP:
3381	case SIOCSIFMAP:
3382	return compat_sioc_ifmap(net, cmd, argp);
3383	case SIOCGSTAMP_OLD:
3384	case SIOCGSTAMPNS_OLD:
3385	if (!sock->ops->gettstamp)
3386	return -ENOIOCTLCMD;
3387	return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3388	!COMPAT_USE_64BIT_TIME);
3389
3390	case SIOCBONDSLAVEINFOQUERY:
3391	case SIOCBONDINFOQUERY:
3392	case SIOCSHWTSTAMP:
3393	case SIOCGHWTSTAMP:
3394	return compat_ifr_data_ioctl(net, cmd, argp);
3395
3396	case FIOSETOWN:
3397	case SIOCSPGRP:
3398	case FIOGETOWN:
3399	case SIOCGPGRP:
3400	case SIOCBRADDBR:
3401	case SIOCBRDELBR:
3402	case SIOCGIFVLAN:
3403	case SIOCSIFVLAN:
3404	case SIOCGSKNS:
3405	case SIOCGSTAMP_NEW:
3406	case SIOCGSTAMPNS_NEW:
3407	return sock_ioctl(file, cmd, arg);
3408
3409	case SIOCGIFFLAGS:
3410	case SIOCSIFFLAGS:
3411	case SIOCGIFMETRIC:
3412	case SIOCSIFMETRIC:
3413	case SIOCGIFMTU:
3414	case SIOCSIFMTU:
3415	case SIOCGIFMEM:
3416	case SIOCSIFMEM:
3417	case SIOCGIFHWADDR:
3418	case SIOCSIFHWADDR:
3419	case SIOCADDMULTI:
3420	case SIOCDELMULTI:
3421	case SIOCGIFINDEX:
3422	case SIOCGIFADDR:
3423	case SIOCSIFADDR:
3424	case SIOCSIFHWBROADCAST:
3425	case SIOCDIFADDR:
3426	case SIOCGIFBRDADDR:
3427	case SIOCSIFBRDADDR:
3428	case SIOCGIFDSTADDR:
3429	case SIOCSIFDSTADDR:
3430	case SIOCGIFNETMASK:
3431	case SIOCSIFNETMASK:
3432	case SIOCSIFPFLAGS:
3433	case SIOCGIFPFLAGS:
3434	case SIOCGIFTXQLEN:
3435	case SIOCSIFTXQLEN:
3436	case SIOCBRADDIF:
3437	case SIOCBRDELIF:
3438	case SIOCGIFNAME:
3439	case SIOCSIFNAME:
3440	case SIOCGMIIPHY:
3441	case SIOCGMIIREG:
3442	case SIOCSMIIREG:
3443	case SIOCBONDENSLAVE:
3444	case SIOCBONDRELEASE:
3445	case SIOCBONDSETHWADDR:
3446	case SIOCBONDCHANGEACTIVE:
3447	return compat_ifreq_ioctl(net, sock, cmd, argp);
3448
3449	case SIOCSARP:
3450	case SIOCGARP:
3451	case SIOCDARP:
3452	case SIOCOUTQ:
3453	case SIOCOUTQNSD:
3454	case SIOCATMARK:
3455	return sock_do_ioctl(net, sock, cmd, arg);
3456	}
3457
3458	return -ENOIOCTLCMD;
3459	}
3460
3461	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3462	unsigned long arg)
3463	{
3464	struct socket *sock = file->private_data;
3465	int ret = -ENOIOCTLCMD;
3466	struct sock *sk;
3467	struct net *net;
3468
3469	sk = sock->sk;
3470	net = sock_net(sk);
3471
3472	if (sock->ops->compat_ioctl)
3473	ret = sock->ops->compat_ioctl(sock, cmd, arg);
3474
3475	if (ret == -ENOIOCTLCMD &&
3476	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3477	ret = compat_wext_handle_ioctl(net, cmd, arg);
3478
3479	if (ret == -ENOIOCTLCMD)
3480	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3481
3482	return ret;
3483	}
3484	#endif
3485
3486	/**
3487	* kernel_bind - bind an address to a socket (kernel space)
3488	* @sock: socket
3489	* @addr: address
3490	* @addrlen: length of address
3491	*
3492	* Returns 0 or an error.
3493	*/
3494
3495	int kernel_bind(struct socket sock, struct sockaddr addr, int addrlen)
3496	{
3497	return sock->ops->bind(sock, addr, addrlen);
3498	}
3499	EXPORT_SYMBOL(kernel_bind);
3500
3501	/**
3502	* kernel_listen - move socket to listening state (kernel space)
3503	* @sock: socket
3504	* @backlog: pending connections queue size
3505	*
3506	* Returns 0 or an error.
3507	*/
3508
3509	int kernel_listen(struct socket *sock, int backlog)
3510	{
3511	return sock->ops->listen(sock, backlog);
3512	}
3513	EXPORT_SYMBOL(kernel_listen);
3514
3515	/**
3516	* kernel_accept - accept a connection (kernel space)
3517	* @sock: listening socket
3518	* @newsock: new connected socket
3519	* @flags: flags
3520	*
3521	* @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3522	* If it fails, @newsock is guaranteed to be %NULL.
3523	* Returns 0 or an error.
3524	*/
3525
3526	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3527	{
3528	struct sock *sk = sock->sk;
3529	int err;
3530
3531	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3532	newsock);
3533	if (err < 0)
3534	goto done;
3535
3536	err = sock->ops->accept(sock, *newsock, flags, true);
3537	if (err < 0) {
3538	sock_release(*newsock);
3539	*newsock = NULL;
3540	goto done;
3541	}
3542
3543	(*newsock)->ops = sock->ops;
3544	__module_get((*newsock)->ops->owner);
3545
3546	done:
3547	return err;
3548	}
3549	EXPORT_SYMBOL(kernel_accept);
3550
3551	/**
3552	* kernel_connect - connect a socket (kernel space)
3553	* @sock: socket
3554	* @addr: address
3555	* @addrlen: address length
3556	* @flags: flags (O_NONBLOCK, ...)
3557	*
3558	* For datagram sockets, @addr is the address to which datagrams are sent
3559	* by default, and the only address from which datagrams are received.
3560	* For stream sockets, attempts to connect to @addr.
3561	* Returns 0 or an error code.
3562	*/
3563
3564	int kernel_connect(struct socket sock, struct sockaddr addr, int addrlen,
3565	int flags)
3566	{
3567	return sock->ops->connect(sock, addr, addrlen, flags);
3568	}
3569	EXPORT_SYMBOL(kernel_connect);
3570
3571	/**
3572	* kernel_getsockname - get the address which the socket is bound (kernel space)
3573	* @sock: socket
3574	* @addr: address holder
3575	*
3576	* Fills the @addr pointer with the address which the socket is bound.
3577	* Returns 0 or an error code.
3578	*/
3579
3580	int kernel_getsockname(struct socket sock, struct sockaddr addr)
3581	{
3582	return sock->ops->getname(sock, addr, 0);
3583	}
3584	EXPORT_SYMBOL(kernel_getsockname);
3585
3586	/**
3587	* kernel_getpeername - get the address which the socket is connected (kernel space)
3588	* @sock: socket
3589	* @addr: address holder
3590	*
3591	* Fills the @addr pointer with the address which the socket is connected.
3592	* Returns 0 or an error code.
3593	*/
3594
3595	int kernel_getpeername(struct socket sock, struct sockaddr addr)
3596	{
3597	return sock->ops->getname(sock, addr, 1);
3598	}
3599	EXPORT_SYMBOL(kernel_getpeername);
3600
3601	/**
3602	* kernel_sendpage - send a &page through a socket (kernel space)
3603	* @sock: socket
3604	* @page: page
3605	* @offset: page offset
3606	* @size: total size in bytes
3607	* @flags: flags (MSG_DONTWAIT, ...)
3608	*
3609	* Returns the total amount sent in bytes or an error.
3610	*/
3611
3612	int kernel_sendpage(struct socket sock, struct page page, int offset,
3613	size_t size, int flags)
3614	{
3615	if (sock->ops->sendpage) {
3616	/* Warn in case the improper page to zero-copy send */
3617	WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3618	return sock->ops->sendpage(sock, page, offset, size, flags);
3619	}
3620	return sock_no_sendpage(sock, page, offset, size, flags);
3621	}
3622	EXPORT_SYMBOL(kernel_sendpage);
3623
3624	/**
3625	* kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3626	* @sk: sock
3627	* @page: page
3628	* @offset: page offset
3629	* @size: total size in bytes
3630	* @flags: flags (MSG_DONTWAIT, ...)
3631	*
3632	* Returns the total amount sent in bytes or an error.
3633	* Caller must hold @sk.
3634	*/
3635
3636	int kernel_sendpage_locked(struct sock sk, struct page page, int offset,
3637	size_t size, int flags)
3638	{
3639	struct socket *sock = sk->sk_socket;
3640
3641	if (sock->ops->sendpage_locked)
3642	return sock->ops->sendpage_locked(sk, page, offset, size,
3643	flags);
3644
3645	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3646	}
3647	EXPORT_SYMBOL(kernel_sendpage_locked);
3648
3649	/**
3650	* kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3651	* @sock: socket
3652	* @how: connection part
3653	*
3654	* Returns 0 or an error.
3655	*/
3656
3657	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3658	{
3659	return sock->ops->shutdown(sock, how);
3660	}
3661	EXPORT_SYMBOL(kernel_sock_shutdown);
3662
3663	/**
3664	* kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3665	* @sk: socket
3666	*
3667	* This routine returns the IP overhead imposed by a socket i.e.
3668	* the length of the underlying IP header, depending on whether
3669	* this is an IPv4 or IPv6 socket and the length from IP options turned
3670	* on at the socket. Assumes that the caller has a lock on the socket.
3671	*/
3672
3673	u32 kernel_sock_ip_overhead(struct sock *sk)
3674	{
3675	struct inet_sock *inet;
3676	struct ip_options_rcu *opt;
3677	u32 overhead = 0;
3678	#if IS_ENABLED(CONFIG_IPV6)
3679	struct ipv6_pinfo *np;
3680	struct ipv6_txoptions *optv6 = NULL;
3681	#endif /* IS_ENABLED(CONFIG_IPV6) */
3682
3683	if (!sk)
3684	return overhead;
3685
3686	switch (sk->sk_family) {
3687	case AF_INET:
3688	inet = inet_sk(sk);
3689	overhead += sizeof(struct iphdr);
3690	opt = rcu_dereference_protected(inet->inet_opt,
3691	sock_owned_by_user(sk));
3692	if (opt)
3693	overhead += opt->opt.optlen;
3694	return overhead;
3695	#if IS_ENABLED(CONFIG_IPV6)
3696	case AF_INET6:
3697	np = inet6_sk(sk);
3698	overhead += sizeof(struct ipv6hdr);
3699	if (np)
3700	optv6 = rcu_dereference_protected(np->opt,
3701	sock_owned_by_user(sk));
3702	if (optv6)
3703	overhead += (optv6->opt_flen + optv6->opt_nflen);
3704	return overhead;
3705	#endif /* IS_ENABLED(CONFIG_IPV6) */
3706	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3707	return overhead;
3708	}
3709	}
3710	EXPORT_SYMBOL(kernel_sock_ip_overhead);