linux - net/socket.c

3711 lines · 90.2 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/bpf-cgroup.h>
56	#include <linux/ethtool.h>
57	#include <linux/mm.h>
58	#include <linux/socket.h>
59	#include <linux/file.h>
60	#include <linux/splice.h>
61	#include <linux/net.h>
62	#include <linux/interrupt.h>
63	#include <linux/thread_info.h>
64	#include <linux/rcupdate.h>
65	#include <linux/netdevice.h>
66	#include <linux/proc_fs.h>
67	#include <linux/seq_file.h>
68	#include <linux/mutex.h>
69	#include <linux/if_bridge.h>
70	#include <linux/if_vlan.h>
71	#include <linux/ptp_classify.h>
72	#include <linux/init.h>
73	#include <linux/poll.h>
74	#include <linux/cache.h>
75	#include <linux/module.h>
76	#include <linux/highmem.h>
77	#include <linux/mount.h>
78	#include <linux/pseudo_fs.h>
79	#include <linux/security.h>
80	#include <linux/syscalls.h>
81	#include <linux/compat.h>
82	#include <linux/kmod.h>
83	#include <linux/audit.h>
84	#include <linux/wireless.h>
85	#include <linux/nsproxy.h>
86	#include <linux/magic.h>
87	#include <linux/slab.h>
88	#include <linux/xattr.h>
89	#include <linux/nospec.h>
90	#include <linux/indirect_call_wrapper.h>
91	#include <linux/io_uring/net.h>
92
93	#include <linux/uaccess.h>
94	#include <asm/unistd.h>
95
96	#include <net/compat.h>
97	#include <net/wext.h>
98	#include <net/cls_cgroup.h>
99
100	#include <net/sock.h>
101	#include <linux/netfilter.h>
102
103	#include <linux/if_tun.h>
104	#include <linux/ipv6_route.h>
105	#include <linux/route.h>
106	#include <linux/termios.h>
107	#include <linux/sockios.h>
108	#include <net/busy_poll.h>
109	#include <linux/errqueue.h>
110	#include <linux/ptp_clock_kernel.h>
111	#include <trace/events/sock.h>
112
113	#include "core/dev.h"
114
115	#ifdef CONFIG_NET_RX_BUSY_POLL
116	unsigned int sysctl_net_busy_read __read_mostly;
117	unsigned int sysctl_net_busy_poll __read_mostly;
118	#endif
119
120	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to);
121	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from);
122	static int sock_mmap(struct file file, struct vm_area_struct vma);
123
124	static int sock_close(struct inode inode, struct file file);
125	static __poll_t sock_poll(struct file *file,
126	struct poll_table_struct *wait);
127	static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
128	#ifdef CONFIG_COMPAT
129	static long compat_sock_ioctl(struct file *file,
130	unsigned int cmd, unsigned long arg);
131	#endif
132	static int sock_fasync(int fd, struct file *filp, int on);
133	static ssize_t sock_splice_read(struct file file, loff_t ppos,
134	struct pipe_inode_info *pipe, size_t len,
135	unsigned int flags);
136	static void sock_splice_eof(struct file *file);
137
138	#ifdef CONFIG_PROC_FS
139	static void sock_show_fdinfo(struct seq_file m, struct file f)
140	{
141	struct socket *sock = f->private_data;
142	const struct proto_ops *ops = READ_ONCE(sock->ops);
143
144	if (ops->show_fdinfo)
145	ops->show_fdinfo(m, sock);
146	}
147	#else
148	#define sock_show_fdinfo NULL
149	#endif
150
151	/*
152	* Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
153	* in the operation structures but are done directly via the socketcall() multiplexor.
154	*/
155
156	static const struct file_operations socket_file_ops = {
157	.owner = THIS_MODULE,
158	.read_iter = sock_read_iter,
159	.write_iter = sock_write_iter,
160	.poll = sock_poll,
161	.unlocked_ioctl = sock_ioctl,
162	#ifdef CONFIG_COMPAT
163	.compat_ioctl = compat_sock_ioctl,
164	#endif
165	.uring_cmd = io_uring_cmd_sock,
166	.mmap = sock_mmap,
167	.release = sock_close,
168	.fasync = sock_fasync,
169	.splice_write = splice_to_socket,
170	.splice_read = sock_splice_read,
171	.splice_eof = sock_splice_eof,
172	.show_fdinfo = sock_show_fdinfo,
173	};
174
175	static const char * const pf_family_names[] = {
176	[PF_UNSPEC] = "PF_UNSPEC",
177	[PF_UNIX] = "PF_UNIX/PF_LOCAL",
178	[PF_INET] = "PF_INET",
179	[PF_AX25] = "PF_AX25",
180	[PF_IPX] = "PF_IPX",
181	[PF_APPLETALK] = "PF_APPLETALK",
182	[PF_NETROM] = "PF_NETROM",
183	[PF_BRIDGE] = "PF_BRIDGE",
184	[PF_ATMPVC] = "PF_ATMPVC",
185	[PF_X25] = "PF_X25",
186	[PF_INET6] = "PF_INET6",
187	[PF_ROSE] = "PF_ROSE",
188	[PF_DECnet] = "PF_DECnet",
189	[PF_NETBEUI] = "PF_NETBEUI",
190	[PF_SECURITY] = "PF_SECURITY",
191	[PF_KEY] = "PF_KEY",
192	[PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
193	[PF_PACKET] = "PF_PACKET",
194	[PF_ASH] = "PF_ASH",
195	[PF_ECONET] = "PF_ECONET",
196	[PF_ATMSVC] = "PF_ATMSVC",
197	[PF_RDS] = "PF_RDS",
198	[PF_SNA] = "PF_SNA",
199	[PF_IRDA] = "PF_IRDA",
200	[PF_PPPOX] = "PF_PPPOX",
201	[PF_WANPIPE] = "PF_WANPIPE",
202	[PF_LLC] = "PF_LLC",
203	[PF_IB] = "PF_IB",
204	[PF_MPLS] = "PF_MPLS",
205	[PF_CAN] = "PF_CAN",
206	[PF_TIPC] = "PF_TIPC",
207	[PF_BLUETOOTH] = "PF_BLUETOOTH",
208	[PF_IUCV] = "PF_IUCV",
209	[PF_RXRPC] = "PF_RXRPC",
210	[PF_ISDN] = "PF_ISDN",
211	[PF_PHONET] = "PF_PHONET",
212	[PF_IEEE802154] = "PF_IEEE802154",
213	[PF_CAIF] = "PF_CAIF",
214	[PF_ALG] = "PF_ALG",
215	[PF_NFC] = "PF_NFC",
216	[PF_VSOCK] = "PF_VSOCK",
217	[PF_KCM] = "PF_KCM",
218	[PF_QIPCRTR] = "PF_QIPCRTR",
219	[PF_SMC] = "PF_SMC",
220	[PF_XDP] = "PF_XDP",
221	[PF_MCTP] = "PF_MCTP",
222	};
223
224	/*
225	* The protocol list. Each protocol is registered in here.
226	*/
227
228	static DEFINE_SPINLOCK(net_family_lock);
229	static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
230
231	/*
232	* Support routines.
233	* Move socket addresses back and forth across the kernel/user
234	* divide and look after the messy bits.
235	*/
236
237	/**
238	* move_addr_to_kernel - copy a socket address into kernel space
239	* @uaddr: Address in user space
240	* @kaddr: Address in kernel space
241	* @ulen: Length in user space
242	*
243	* The address is copied into kernel space. If the provided address is
244	* too long an error code of -EINVAL is returned. If the copy gives
245	* invalid addresses -EFAULT is returned. On a success 0 is returned.
246	*/
247
248	int move_addr_to_kernel(void __user uaddr, int ulen, struct sockaddr_storage kaddr)
249	{
250	if (ulen < 0 \|\| ulen > sizeof(struct sockaddr_storage))
251	return -EINVAL;
252	if (ulen == 0)
253	return 0;
254	if (copy_from_user(kaddr, uaddr, ulen))
255	return -EFAULT;
256	return audit_sockaddr(ulen, kaddr);
257	}
258
259	/**
260	* move_addr_to_user - copy an address to user space
261	* @kaddr: kernel space address
262	* @klen: length of address in kernel
263	* @uaddr: user space address
264	* @ulen: pointer to user length field
265	*
266	* The value pointed to by ulen on entry is the buffer length available.
267	* This is overwritten with the buffer space used. -EINVAL is returned
268	* if an overlong buffer is specified or a negative buffer size. -EFAULT
269	* is returned if either the buffer or the length field are not
270	* accessible.
271	* After copying the data up to the limit the user specifies, the true
272	* length of the data is written over the length limit the user
273	* specified. Zero is returned for a success.
274	*/
275
276	static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
277	void __user uaddr, int __user ulen)
278	{
279	int err;
280	int len;
281
282	BUG_ON(klen > sizeof(struct sockaddr_storage));
283	err = get_user(len, ulen);
284	if (err)
285	return err;
286	if (len > klen)
287	len = klen;
288	if (len < 0)
289	return -EINVAL;
290	if (len) {
291	if (audit_sockaddr(klen, kaddr))
292	return -ENOMEM;
293	if (copy_to_user(uaddr, kaddr, len))
294	return -EFAULT;
295	}
296	/*
297	* "fromlen shall refer to the value before truncation.."
298	* 1003.1g
299	*/
300	return __put_user(klen, ulen);
301	}
302
303	static struct kmem_cache *sock_inode_cachep __ro_after_init;
304
305	static struct inode sock_alloc_inode(struct super_block sb)
306	{
307	struct socket_alloc *ei;
308
309	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
310	if (!ei)
311	return NULL;
312	init_waitqueue_head(&ei->socket.wq.wait);
313	ei->socket.wq.fasync_list = NULL;
314	ei->socket.wq.flags = 0;
315
316	ei->socket.state = SS_UNCONNECTED;
317	ei->socket.flags = 0;
318	ei->socket.ops = NULL;
319	ei->socket.sk = NULL;
320	ei->socket.file = NULL;
321
322	return &ei->vfs_inode;
323	}
324
325	static void sock_free_inode(struct inode *inode)
326	{
327	struct socket_alloc *ei;
328
329	ei = container_of(inode, struct socket_alloc, vfs_inode);
330	kmem_cache_free(sock_inode_cachep, ei);
331	}
332
333	static void init_once(void *foo)
334	{
335	struct socket_alloc ei = (struct socket_alloc )foo;
336
337	inode_init_once(&ei->vfs_inode);
338	}
339
340	static void init_inodecache(void)
341	{
342	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
343	sizeof(struct socket_alloc),
344	0,
345	(SLAB_HWCACHE_ALIGN \|
346	SLAB_RECLAIM_ACCOUNT \|
347	SLAB_ACCOUNT),
348	init_once);
349	BUG_ON(sock_inode_cachep == NULL);
350	}
351
352	static const struct super_operations sockfs_ops = {
353	.alloc_inode = sock_alloc_inode,
354	.free_inode = sock_free_inode,
355	.statfs = simple_statfs,
356	};
357
358	/*
359	* sockfs_dname() is called from d_path().
360	*/
361	static char sockfs_dname(struct dentry dentry, char *buffer, int buflen)
362	{
363	return dynamic_dname(buffer, buflen, "socket:[%lu]",
364	d_inode(dentry)->i_ino);
365	}
366
367	static const struct dentry_operations sockfs_dentry_operations = {
368	.d_dname = sockfs_dname,
369	};
370
371	static int sockfs_xattr_get(const struct xattr_handler *handler,
372	struct dentry dentry, struct inode inode,
373	const char suffix, void value, size_t size)
374	{
375	if (value) {
376	if (dentry->d_name.len + 1 > size)
377	return -ERANGE;
378	memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
379	}
380	return dentry->d_name.len + 1;
381	}
382
383	#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
384	#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
385	#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
386
387	static const struct xattr_handler sockfs_xattr_handler = {
388	.name = XATTR_NAME_SOCKPROTONAME,
389	.get = sockfs_xattr_get,
390	};
391
392	static int sockfs_security_xattr_set(const struct xattr_handler *handler,
393	struct mnt_idmap *idmap,
394	struct dentry dentry, struct inode inode,
395	const char suffix, const void value,
396	size_t size, int flags)
397	{
398	/* Handled by LSM. */
399	return -EAGAIN;
400	}
401
402	static const struct xattr_handler sockfs_security_xattr_handler = {
403	.prefix = XATTR_SECURITY_PREFIX,
404	.set = sockfs_security_xattr_set,
405	};
406
407	static const struct xattr_handler * const sockfs_xattr_handlers[] = {
408	&sockfs_xattr_handler,
409	&sockfs_security_xattr_handler,
410	NULL
411	};
412
413	static int sockfs_init_fs_context(struct fs_context *fc)
414	{
415	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
416	if (!ctx)
417	return -ENOMEM;
418	ctx->ops = &sockfs_ops;
419	ctx->dops = &sockfs_dentry_operations;
420	ctx->xattr = sockfs_xattr_handlers;
421	return 0;
422	}
423
424	static struct vfsmount *sock_mnt __read_mostly;
425
426	static struct file_system_type sock_fs_type = {
427	.name = "sockfs",
428	.init_fs_context = sockfs_init_fs_context,
429	.kill_sb = kill_anon_super,
430	};
431
432	/*
433	* Obtains the first available file descriptor and sets it up for use.
434	*
435	* These functions create file structures and maps them to fd space
436	* of the current process. On success it returns file descriptor
437	* and file struct implicitly stored in sock->file.
438	* Note that another thread may close file descriptor before we return
439	* from this function. We use the fact that now we do not refer
440	* to socket after mapping. If one day we will need it, this
441	* function will increment ref. count on file by 1.
442	*
443	* In any case returned fd MAY BE not valid!
444	* This race condition is unavoidable
445	* with shared fd spaces, we cannot solve it inside kernel,
446	* but we take care of internal coherence yet.
447	*/
448
449	/**
450	* sock_alloc_file - Bind a &socket to a &file
451	* @sock: socket
452	* @flags: file status flags
453	* @dname: protocol name
454	*
455	* Returns the &file bound with @sock, implicitly storing it
456	* in sock->file. If dname is %NULL, sets to "".
457	*
458	* On failure @sock is released, and an ERR pointer is returned.
459	*
460	* This function uses GFP_KERNEL internally.
461	*/
462
463	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
464	{
465	struct file *file;
466
467	if (!dname)
468	dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
469
470	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
471	O_RDWR \| (flags & O_NONBLOCK),
472	&socket_file_ops);
473	if (IS_ERR(file)) {
474	sock_release(sock);
475	return file;
476	}
477
478	file->f_mode \|= FMODE_NOWAIT;
479	sock->file = file;
480	file->private_data = sock;
481	stream_open(SOCK_INODE(sock), file);
482	/*
483	* Disable permission and pre-content events, but enable legacy
484	* inotify events for legacy users.
485	*/
486	file_set_fsnotify_mode(file, FMODE_NONOTIFY_PERM);
487	return file;
488	}
489	EXPORT_SYMBOL(sock_alloc_file);
490
491	static int sock_map_fd(struct socket *sock, int flags)
492	{
493	struct file *newfile;
494	int fd = get_unused_fd_flags(flags);
495	if (unlikely(fd < 0)) {
496	sock_release(sock);
497	return fd;
498	}
499
500	newfile = sock_alloc_file(sock, flags, NULL);
501	if (!IS_ERR(newfile)) {
502	fd_install(fd, newfile);
503	return fd;
504	}
505
506	put_unused_fd(fd);
507	return PTR_ERR(newfile);
508	}
509
510	/**
511	* sock_from_file - Return the &socket bounded to @file.
512	* @file: file
513	*
514	* On failure returns %NULL.
515	*/
516
517	struct socket sock_from_file(struct file file)
518	{
519	if (likely(file->f_op == &socket_file_ops))
520	return file->private_data; /* set in sock_alloc_file */
521
522	return NULL;
523	}
524	EXPORT_SYMBOL(sock_from_file);
525
526	/**
527	* sockfd_lookup - Go from a file number to its socket slot
528	* @fd: file handle
529	* @err: pointer to an error code return
530	*
531	* The file handle passed in is locked and the socket it is bound
532	* to is returned. If an error occurs the err pointer is overwritten
533	* with a negative errno code and NULL is returned. The function checks
534	* for both invalid handles and passing a handle which is not a socket.
535	*
536	* On a success the socket object pointer is returned.
537	*/
538
539	struct socket sockfd_lookup(int fd, int err)
540	{
541	struct file *file;
542	struct socket *sock;
543
544	file = fget(fd);
545	if (!file) {
546	*err = -EBADF;
547	return NULL;
548	}
549
550	sock = sock_from_file(file);
551	if (!sock) {
552	*err = -ENOTSOCK;
553	fput(file);
554	}
555	return sock;
556	}
557	EXPORT_SYMBOL(sockfd_lookup);
558
559	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
560	size_t size)
561	{
562	ssize_t len;
563	ssize_t used = 0;
564
565	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
566	if (len < 0)
567	return len;
568	used += len;
569	if (buffer) {
570	if (size < used)
571	return -ERANGE;
572	buffer += len;
573	}
574
575	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
576	used += len;
577	if (buffer) {
578	if (size < used)
579	return -ERANGE;
580	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
581	buffer += len;
582	}
583
584	return used;
585	}
586
587	static int sockfs_setattr(struct mnt_idmap *idmap,
588	struct dentry dentry, struct iattr iattr)
589	{
590	int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
591
592	if (!err && (iattr->ia_valid & ATTR_UID)) {
593	struct socket *sock = SOCKET_I(d_inode(dentry));
594
595	if (sock->sk)
596	sock->sk->sk_uid = iattr->ia_uid;
597	else
598	err = -ENOENT;
599	}
600
601	return err;
602	}
603
604	static const struct inode_operations sockfs_inode_ops = {
605	.listxattr = sockfs_listxattr,
606	.setattr = sockfs_setattr,
607	};
608
609	/**
610	* sock_alloc - allocate a socket
611	*
612	* Allocate a new inode and socket object. The two are bound together
613	* and initialised. The socket is then returned. If we are out of inodes
614	* NULL is returned. This functions uses GFP_KERNEL internally.
615	*/
616
617	struct socket *sock_alloc(void)
618	{
619	struct inode *inode;
620	struct socket *sock;
621
622	inode = new_inode_pseudo(sock_mnt->mnt_sb);
623	if (!inode)
624	return NULL;
625
626	sock = SOCKET_I(inode);
627
628	inode->i_ino = get_next_ino();
629	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
630	inode->i_uid = current_fsuid();
631	inode->i_gid = current_fsgid();
632	inode->i_op = &sockfs_inode_ops;
633
634	return sock;
635	}
636	EXPORT_SYMBOL(sock_alloc);
637
638	static void __sock_release(struct socket sock, struct inode inode)
639	{
640	const struct proto_ops *ops = READ_ONCE(sock->ops);
641
642	if (ops) {
643	struct module *owner = ops->owner;
644
645	if (inode)
646	inode_lock(inode);
647	ops->release(sock);
648	sock->sk = NULL;
649	if (inode)
650	inode_unlock(inode);
651	sock->ops = NULL;
652	module_put(owner);
653	}
654
655	if (sock->wq.fasync_list)
656	pr_err("%s: fasync list not empty!\n", __func__);
657
658	if (!sock->file) {
659	iput(SOCK_INODE(sock));
660	return;
661	}
662	sock->file = NULL;
663	}
664
665	/**
666	* sock_release - close a socket
667	* @sock: socket to close
668	*
669	* The socket is released from the protocol stack if it has a release
670	* callback, and the inode is then released if the socket is bound to
671	* an inode not a file.
672	*/
673	void sock_release(struct socket *sock)
674	{
675	__sock_release(sock, NULL);
676	}
677	EXPORT_SYMBOL(sock_release);
678
679	void __sock_tx_timestamp(__u32 tsflags, __u8 *tx_flags)
680	{
681	u8 flags = *tx_flags;
682
683	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
684	flags \|= SKBTX_HW_TSTAMP;
685
686	/* PTP hardware clocks can provide a free running cycle counter
687	* as a time base for virtual clocks. Tell driver to use the
688	* free running cycle counter for timestamp if socket is bound
689	* to virtual clock.
690	*/
691	if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
692	flags \|= SKBTX_HW_TSTAMP_USE_CYCLES;
693	}
694
695	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
696	flags \|= SKBTX_SW_TSTAMP;
697
698	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
699	flags \|= SKBTX_SCHED_TSTAMP;
700
701	*tx_flags = flags;
702	}
703	EXPORT_SYMBOL(__sock_tx_timestamp);
704
705	INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket , struct msghdr ,
706	size_t));
707	INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket , struct msghdr ,
708	size_t));
709
710	static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
711	int flags)
712	{
713	trace_sock_send_length(sk, ret, 0);
714	}
715
716	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
717	{
718	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
719	inet_sendmsg, sock, msg,
720	msg_data_left(msg));
721	BUG_ON(ret == -EIOCBQUEUED);
722
723	if (trace_sock_send_length_enabled())
724	call_trace_sock_send_length(sock->sk, ret, 0);
725	return ret;
726	}
727
728	static int __sock_sendmsg(struct socket sock, struct msghdr msg)
729	{
730	int err = security_socket_sendmsg(sock, msg,
731	msg_data_left(msg));
732
733	return err ?: sock_sendmsg_nosec(sock, msg);
734	}
735
736	/**
737	* sock_sendmsg - send a message through @sock
738	* @sock: socket
739	* @msg: message to send
740	*
741	* Sends @msg through @sock, passing through LSM.
742	* Returns the number of bytes sent, or an error code.
743	*/
744	int sock_sendmsg(struct socket sock, struct msghdr msg)
745	{
746	struct sockaddr_storage save_addr = (struct sockaddr_storage )msg->msg_name;
747	struct sockaddr_storage address;
748	int save_len = msg->msg_namelen;
749	int ret;
750
751	if (msg->msg_name) {
752	memcpy(&address, msg->msg_name, msg->msg_namelen);
753	msg->msg_name = &address;
754	}
755
756	ret = __sock_sendmsg(sock, msg);
757	msg->msg_name = save_addr;
758	msg->msg_namelen = save_len;
759
760	return ret;
761	}
762	EXPORT_SYMBOL(sock_sendmsg);
763
764	/**
765	* kernel_sendmsg - send a message through @sock (kernel-space)
766	* @sock: socket
767	* @msg: message header
768	* @vec: kernel vec
769	* @num: vec array length
770	* @size: total message data size
771	*
772	* Builds the message data with @vec and sends it through @sock.
773	* Returns the number of bytes sent, or an error code.
774	*/
775
776	int kernel_sendmsg(struct socket sock, struct msghdr msg,
777	struct kvec *vec, size_t num, size_t size)
778	{
779	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
780	return sock_sendmsg(sock, msg);
781	}
782	EXPORT_SYMBOL(kernel_sendmsg);
783
784	static bool skb_is_err_queue(const struct sk_buff *skb)
785	{
786	/* pkt_type of skbs enqueued on the error queue are set to
787	* PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
788	* in recvmsg, since skbs received on a local socket will never
789	* have a pkt_type of PACKET_OUTGOING.
790	*/
791	return skb->pkt_type == PACKET_OUTGOING;
792	}
793
794	/* On transmit, software and hardware timestamps are returned independently.
795	* As the two skb clones share the hardware timestamp, which may be updated
796	* before the software timestamp is received, a hardware TX timestamp may be
797	* returned only if there is no software TX timestamp. Ignore false software
798	* timestamps, which may be made in the __sock_recv_timestamp() call when the
799	* option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
800	* hardware timestamp.
801	*/
802	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
803	{
804	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
805	}
806
807	static ktime_t get_timestamp(struct sock sk, struct sk_buff skb, int *if_index)
808	{
809	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
810	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
811	struct net_device *orig_dev;
812	ktime_t hwtstamp;
813
814	rcu_read_lock();
815	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
816	if (orig_dev) {
817	*if_index = orig_dev->ifindex;
818	hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
819	} else {
820	hwtstamp = shhwtstamps->hwtstamp;
821	}
822	rcu_read_unlock();
823
824	return hwtstamp;
825	}
826
827	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb,
828	int if_index)
829	{
830	struct scm_ts_pktinfo ts_pktinfo;
831	struct net_device *orig_dev;
832
833	if (!skb_mac_header_was_set(skb))
834	return;
835
836	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
837
838	if (!if_index) {
839	rcu_read_lock();
840	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
841	if (orig_dev)
842	if_index = orig_dev->ifindex;
843	rcu_read_unlock();
844	}
845	ts_pktinfo.if_index = if_index;
846
847	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
848	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
849	sizeof(ts_pktinfo), &ts_pktinfo);
850	}
851
852	/*
853	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
854	*/
855	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
856	struct sk_buff *skb)
857	{
858	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
859	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
860	struct scm_timestamping_internal tss;
861	int empty = 1, false_tstamp = 0;
862	struct skb_shared_hwtstamps *shhwtstamps =
863	skb_hwtstamps(skb);
864	int if_index;
865	ktime_t hwtstamp;
866	u32 tsflags;
867
868	/* Race occurred between timestamp enabling and packet
869	receiving. Fill in the current time for now. */
870	if (need_software_tstamp && skb->tstamp == 0) {
871	__net_timestamp(skb);
872	false_tstamp = 1;
873	}
874
875	if (need_software_tstamp) {
876	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
877	if (new_tstamp) {
878	struct __kernel_sock_timeval tv;
879
880	skb_get_new_timestamp(skb, &tv);
881	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
882	sizeof(tv), &tv);
883	} else {
884	struct __kernel_old_timeval tv;
885
886	skb_get_timestamp(skb, &tv);
887	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
888	sizeof(tv), &tv);
889	}
890	} else {
891	if (new_tstamp) {
892	struct __kernel_timespec ts;
893
894	skb_get_new_timestampns(skb, &ts);
895	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
896	sizeof(ts), &ts);
897	} else {
898	struct __kernel_old_timespec ts;
899
900	skb_get_timestampns(skb, &ts);
901	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
902	sizeof(ts), &ts);
903	}
904	}
905	}
906
907	memset(&tss, 0, sizeof(tss));
908	tsflags = READ_ONCE(sk->sk_tsflags);
909	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
910	(tsflags & SOF_TIMESTAMPING_RX_SOFTWARE \|\|
911	skb_is_err_queue(skb) \|\|
912	!(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
913	ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
914	empty = 0;
915	if (shhwtstamps &&
916	(tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
917	(tsflags & SOF_TIMESTAMPING_RX_HARDWARE \|\|
918	skb_is_err_queue(skb) \|\|
919	!(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
920	!skb_is_swtx_tstamp(skb, false_tstamp)) {
921	if_index = 0;
922	if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
923	hwtstamp = get_timestamp(sk, skb, &if_index);
924	else
925	hwtstamp = shhwtstamps->hwtstamp;
926
927	if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
928	hwtstamp = ptp_convert_timestamp(&hwtstamp,
929	READ_ONCE(sk->sk_bind_phc));
930
931	if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
932	empty = 0;
933
934	if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
935	!skb_is_err_queue(skb))
936	put_ts_pktinfo(msg, skb, if_index);
937	}
938	}
939	if (!empty) {
940	if (sock_flag(sk, SOCK_TSTAMP_NEW))
941	put_cmsg_scm_timestamping64(msg, &tss);
942	else
943	put_cmsg_scm_timestamping(msg, &tss);
944
945	if (skb_is_err_queue(skb) && skb->len &&
946	SKB_EXT_ERR(skb)->opt_stats)
947	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
948	skb->len, skb->data);
949	}
950	}
951	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
952
953	#ifdef CONFIG_WIRELESS
954	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
955	struct sk_buff *skb)
956	{
957	int ack;
958
959	if (!sock_flag(sk, SOCK_WIFI_STATUS))
960	return;
961	if (!skb->wifi_acked_valid)
962	return;
963
964	ack = skb->wifi_acked;
965
966	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
967	}
968	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
969	#endif
970
971	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
972	struct sk_buff *skb)
973	{
974	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
975	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
976	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
977	}
978
979	static void sock_recv_mark(struct msghdr msg, struct sock sk,
980	struct sk_buff *skb)
981	{
982	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
983	/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
984	__u32 mark = skb->mark;
985
986	put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
987	}
988	}
989
990	static void sock_recv_priority(struct msghdr msg, struct sock sk,
991	struct sk_buff *skb)
992	{
993	if (sock_flag(sk, SOCK_RCVPRIORITY) && skb) {
994	__u32 priority = skb->priority;
995
996	put_cmsg(msg, SOL_SOCKET, SO_PRIORITY, sizeof(__u32), &priority);
997	}
998	}
999
1000	void __sock_recv_cmsgs(struct msghdr msg, struct sock sk,
1001	struct sk_buff *skb)
1002	{
1003	sock_recv_timestamp(msg, sk, skb);
1004	sock_recv_drops(msg, sk, skb);
1005	sock_recv_mark(msg, sk, skb);
1006	sock_recv_priority(msg, sk, skb);
1007	}
1008	EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1009
1010	INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket , struct msghdr ,
1011	size_t, int));
1012	INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket , struct msghdr ,
1013	size_t, int));
1014
1015	static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1016	{
1017	trace_sock_recv_length(sk, ret, flags);
1018	}
1019
1020	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
1021	int flags)
1022	{
1023	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1024	inet6_recvmsg,
1025	inet_recvmsg, sock, msg,
1026	msg_data_left(msg), flags);
1027	if (trace_sock_recv_length_enabled())
1028	call_trace_sock_recv_length(sock->sk, ret, flags);
1029	return ret;
1030	}
1031
1032	/**
1033	* sock_recvmsg - receive a message from @sock
1034	* @sock: socket
1035	* @msg: message to receive
1036	* @flags: message flags
1037	*
1038	* Receives @msg from @sock, passing through LSM. Returns the total number
1039	* of bytes received, or an error.
1040	*/
1041	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
1042	{
1043	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1044
1045	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1046	}
1047	EXPORT_SYMBOL(sock_recvmsg);
1048
1049	/**
1050	* kernel_recvmsg - Receive a message from a socket (kernel space)
1051	* @sock: The socket to receive the message from
1052	* @msg: Received message
1053	* @vec: Input s/g array for message data
1054	* @num: Size of input s/g array
1055	* @size: Number of bytes to read
1056	* @flags: Message flags (MSG_DONTWAIT, etc...)
1057	*
1058	* On return the msg structure contains the scatter/gather array passed in the
1059	* vec argument. The array is modified so that it consists of the unfilled
1060	* portion of the original array.
1061	*
1062	* The returned value is the total number of bytes received, or an error.
1063	*/
1064
1065	int kernel_recvmsg(struct socket sock, struct msghdr msg,
1066	struct kvec *vec, size_t num, size_t size, int flags)
1067	{
1068	msg->msg_control_is_user = false;
1069	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1070	return sock_recvmsg(sock, msg, flags);
1071	}
1072	EXPORT_SYMBOL(kernel_recvmsg);
1073
1074	static ssize_t sock_splice_read(struct file file, loff_t ppos,
1075	struct pipe_inode_info *pipe, size_t len,
1076	unsigned int flags)
1077	{
1078	struct socket *sock = file->private_data;
1079	const struct proto_ops *ops;
1080
1081	ops = READ_ONCE(sock->ops);
1082	if (unlikely(!ops->splice_read))
1083	return copy_splice_read(file, ppos, pipe, len, flags);
1084
1085	return ops->splice_read(sock, ppos, pipe, len, flags);
1086	}
1087
1088	static void sock_splice_eof(struct file *file)
1089	{
1090	struct socket *sock = file->private_data;
1091	const struct proto_ops *ops;
1092
1093	ops = READ_ONCE(sock->ops);
1094	if (ops->splice_eof)
1095	ops->splice_eof(sock);
1096	}
1097
1098	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
1099	{
1100	struct file *file = iocb->ki_filp;
1101	struct socket *sock = file->private_data;
1102	struct msghdr msg = {.msg_iter = *to,
1103	.msg_iocb = iocb};
1104	ssize_t res;
1105
1106	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
1107	msg.msg_flags = MSG_DONTWAIT;
1108
1109	if (iocb->ki_pos != 0)
1110	return -ESPIPE;
1111
1112	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1113	return 0;
1114
1115	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1116	*to = msg.msg_iter;
1117	return res;
1118	}
1119
1120	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
1121	{
1122	struct file *file = iocb->ki_filp;
1123	struct socket *sock = file->private_data;
1124	struct msghdr msg = {.msg_iter = *from,
1125	.msg_iocb = iocb};
1126	ssize_t res;
1127
1128	if (iocb->ki_pos != 0)
1129	return -ESPIPE;
1130
1131	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
1132	msg.msg_flags = MSG_DONTWAIT;
1133
1134	if (sock->type == SOCK_SEQPACKET)
1135	msg.msg_flags \|= MSG_EOR;
1136
1137	res = __sock_sendmsg(sock, &msg);
1138	*from = msg.msg_iter;
1139	return res;
1140	}
1141
1142	/*
1143	* Atomic setting of ioctl hooks to avoid race
1144	* with module unload.
1145	*/
1146
1147	static DEFINE_MUTEX(br_ioctl_mutex);
1148	static int (br_ioctl_hook)(struct net net, struct net_bridge *br,
1149	unsigned int cmd, struct ifreq *ifr,
1150	void __user *uarg);
1151
1152	void brioctl_set(int (hook)(struct net net, struct net_bridge *br,
1153	unsigned int cmd, struct ifreq *ifr,
1154	void __user *uarg))
1155	{
1156	mutex_lock(&br_ioctl_mutex);
1157	br_ioctl_hook = hook;
1158	mutex_unlock(&br_ioctl_mutex);
1159	}
1160	EXPORT_SYMBOL(brioctl_set);
1161
1162	int br_ioctl_call(struct net net, struct net_bridge br, unsigned int cmd,
1163	struct ifreq ifr, void __user uarg)
1164	{
1165	int err = -ENOPKG;
1166
1167	if (!br_ioctl_hook)
1168	request_module("bridge");
1169
1170	mutex_lock(&br_ioctl_mutex);
1171	if (br_ioctl_hook)
1172	err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1173	mutex_unlock(&br_ioctl_mutex);
1174
1175	return err;
1176	}
1177
1178	static DEFINE_MUTEX(vlan_ioctl_mutex);
1179	static int (vlan_ioctl_hook) (struct net , void __user *arg);
1180
1181	void vlan_ioctl_set(int (hook) (struct net , void __user *))
1182	{
1183	mutex_lock(&vlan_ioctl_mutex);
1184	vlan_ioctl_hook = hook;
1185	mutex_unlock(&vlan_ioctl_mutex);
1186	}
1187	EXPORT_SYMBOL(vlan_ioctl_set);
1188
1189	static long sock_do_ioctl(struct net net, struct socket sock,
1190	unsigned int cmd, unsigned long arg)
1191	{
1192	const struct proto_ops *ops = READ_ONCE(sock->ops);
1193	struct ifreq ifr;
1194	bool need_copyout;
1195	int err;
1196	void __user argp = (void __user )arg;
1197	void __user *data;
1198
1199	err = ops->ioctl(sock, cmd, arg);
1200
1201	/*
1202	* If this ioctl is unknown try to hand it down
1203	* to the NIC driver.
1204	*/
1205	if (err != -ENOIOCTLCMD)
1206	return err;
1207
1208	if (!is_socket_ioctl_cmd(cmd))
1209	return -ENOTTY;
1210
1211	if (get_user_ifreq(&ifr, &data, argp))
1212	return -EFAULT;
1213	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1214	if (!err && need_copyout)
1215	if (put_user_ifreq(&ifr, argp))
1216	return -EFAULT;
1217
1218	return err;
1219	}
1220
1221	/*
1222	* With an ioctl, arg may well be a user mode pointer, but we don't know
1223	* what to do with it - that's up to the protocol still.
1224	*/
1225
1226	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1227	{
1228	const struct proto_ops *ops;
1229	struct socket *sock;
1230	struct sock *sk;
1231	void __user argp = (void __user )arg;
1232	int pid, err;
1233	struct net *net;
1234
1235	sock = file->private_data;
1236	ops = READ_ONCE(sock->ops);
1237	sk = sock->sk;
1238	net = sock_net(sk);
1239	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1240	struct ifreq ifr;
1241	void __user *data;
1242	bool need_copyout;
1243	if (get_user_ifreq(&ifr, &data, argp))
1244	return -EFAULT;
1245	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1246	if (!err && need_copyout)
1247	if (put_user_ifreq(&ifr, argp))
1248	return -EFAULT;
1249	} else
1250	#ifdef CONFIG_WEXT_CORE
1251	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1252	err = wext_handle_ioctl(net, cmd, argp);
1253	} else
1254	#endif
1255	switch (cmd) {
1256	case FIOSETOWN:
1257	case SIOCSPGRP:
1258	err = -EFAULT;
1259	if (get_user(pid, (int __user *)argp))
1260	break;
1261	err = f_setown(sock->file, pid, 1);
1262	break;
1263	case FIOGETOWN:
1264	case SIOCGPGRP:
1265	err = put_user(f_getown(sock->file),
1266	(int __user *)argp);
1267	break;
1268	case SIOCGIFBR:
1269	case SIOCSIFBR:
1270	case SIOCBRADDBR:
1271	case SIOCBRDELBR:
1272	err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1273	break;
1274	case SIOCGIFVLAN:
1275	case SIOCSIFVLAN:
1276	err = -ENOPKG;
1277	if (!vlan_ioctl_hook)
1278	request_module("8021q");
1279
1280	mutex_lock(&vlan_ioctl_mutex);
1281	if (vlan_ioctl_hook)
1282	err = vlan_ioctl_hook(net, argp);
1283	mutex_unlock(&vlan_ioctl_mutex);
1284	break;
1285	case SIOCGSKNS:
1286	err = -EPERM;
1287	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1288	break;
1289
1290	err = open_related_ns(&net->ns, get_net_ns);
1291	break;
1292	case SIOCGSTAMP_OLD:
1293	case SIOCGSTAMPNS_OLD:
1294	if (!ops->gettstamp) {
1295	err = -ENOIOCTLCMD;
1296	break;
1297	}
1298	err = ops->gettstamp(sock, argp,
1299	cmd == SIOCGSTAMP_OLD,
1300	!IS_ENABLED(CONFIG_64BIT));
1301	break;
1302	case SIOCGSTAMP_NEW:
1303	case SIOCGSTAMPNS_NEW:
1304	if (!ops->gettstamp) {
1305	err = -ENOIOCTLCMD;
1306	break;
1307	}
1308	err = ops->gettstamp(sock, argp,
1309	cmd == SIOCGSTAMP_NEW,
1310	false);
1311	break;
1312
1313	case SIOCGIFCONF:
1314	err = dev_ifconf(net, argp);
1315	break;
1316
1317	default:
1318	err = sock_do_ioctl(net, sock, cmd, arg);
1319	break;
1320	}
1321	return err;
1322	}
1323
1324	/**
1325	* sock_create_lite - creates a socket
1326	* @family: protocol family (AF_INET, ...)
1327	* @type: communication type (SOCK_STREAM, ...)
1328	* @protocol: protocol (0, ...)
1329	* @res: new socket
1330	*
1331	* Creates a new socket and assigns it to @res, passing through LSM.
1332	* The new socket initialization is not complete, see kernel_accept().
1333	* Returns 0 or an error. On failure @res is set to %NULL.
1334	* This function internally uses GFP_KERNEL.
1335	*/
1336
1337	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1338	{
1339	int err;
1340	struct socket *sock = NULL;
1341
1342	err = security_socket_create(family, type, protocol, 1);
1343	if (err)
1344	goto out;
1345
1346	sock = sock_alloc();
1347	if (!sock) {
1348	err = -ENOMEM;
1349	goto out;
1350	}
1351
1352	sock->type = type;
1353	err = security_socket_post_create(sock, family, type, protocol, 1);
1354	if (err)
1355	goto out_release;
1356
1357	out:
1358	*res = sock;
1359	return err;
1360	out_release:
1361	sock_release(sock);
1362	sock = NULL;
1363	goto out;
1364	}
1365	EXPORT_SYMBOL(sock_create_lite);
1366
1367	/* No kernel lock held - perfect */
1368	static __poll_t sock_poll(struct file file, poll_table wait)
1369	{
1370	struct socket *sock = file->private_data;
1371	const struct proto_ops *ops = READ_ONCE(sock->ops);
1372	__poll_t events = poll_requested_events(wait), flag = 0;
1373
1374	if (!ops->poll)
1375	return 0;
1376
1377	if (sk_can_busy_loop(sock->sk)) {
1378	/* poll once if requested by the syscall */
1379	if (events & POLL_BUSY_LOOP)
1380	sk_busy_loop(sock->sk, 1);
1381
1382	/* if this socket can poll_ll, tell the system call */
1383	flag = POLL_BUSY_LOOP;
1384	}
1385
1386	return ops->poll(file, sock, wait) \| flag;
1387	}
1388
1389	static int sock_mmap(struct file file, struct vm_area_struct vma)
1390	{
1391	struct socket *sock = file->private_data;
1392
1393	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1394	}
1395
1396	static int sock_close(struct inode inode, struct file filp)
1397	{
1398	__sock_release(SOCKET_I(inode), inode);
1399	return 0;
1400	}
1401
1402	/*
1403	* Update the socket async list
1404	*
1405	* Fasync_list locking strategy.
1406	*
1407	* 1. fasync_list is modified only under process context socket lock
1408	* i.e. under semaphore.
1409	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1410	* or under socket lock
1411	*/
1412
1413	static int sock_fasync(int fd, struct file *filp, int on)
1414	{
1415	struct socket *sock = filp->private_data;
1416	struct sock *sk = sock->sk;
1417	struct socket_wq *wq = &sock->wq;
1418
1419	if (sk == NULL)
1420	return -EINVAL;
1421
1422	lock_sock(sk);
1423	fasync_helper(fd, filp, on, &wq->fasync_list);
1424
1425	if (!wq->fasync_list)
1426	sock_reset_flag(sk, SOCK_FASYNC);
1427	else
1428	sock_set_flag(sk, SOCK_FASYNC);
1429
1430	release_sock(sk);
1431	return 0;
1432	}
1433
1434	/* This function may be called only under rcu_lock */
1435
1436	int sock_wake_async(struct socket_wq *wq, int how, int band)
1437	{
1438	if (!wq \|\| !wq->fasync_list)
1439	return -1;
1440
1441	switch (how) {
1442	case SOCK_WAKE_WAITD:
1443	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1444	break;
1445	goto call_kill;
1446	case SOCK_WAKE_SPACE:
1447	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1448	break;
1449	fallthrough;
1450	case SOCK_WAKE_IO:
1451	call_kill:
1452	kill_fasync(&wq->fasync_list, SIGIO, band);
1453	break;
1454	case SOCK_WAKE_URG:
1455	kill_fasync(&wq->fasync_list, SIGURG, band);
1456	}
1457
1458	return 0;
1459	}
1460	EXPORT_SYMBOL(sock_wake_async);
1461
1462	/**
1463	* __sock_create - creates a socket
1464	* @net: net namespace
1465	* @family: protocol family (AF_INET, ...)
1466	* @type: communication type (SOCK_STREAM, ...)
1467	* @protocol: protocol (0, ...)
1468	* @res: new socket
1469	* @kern: boolean for kernel space sockets
1470	*
1471	* Creates a new socket and assigns it to @res, passing through LSM.
1472	* Returns 0 or an error. On failure @res is set to %NULL. @kern must
1473	* be set to true if the socket resides in kernel space.
1474	* This function internally uses GFP_KERNEL.
1475	*/
1476
1477	int __sock_create(struct net *net, int family, int type, int protocol,
1478	struct socket **res, int kern)
1479	{
1480	int err;
1481	struct socket *sock;
1482	const struct net_proto_family *pf;
1483
1484	/*
1485	* Check protocol is in range
1486	*/
1487	if (family < 0 \|\| family >= NPROTO)
1488	return -EAFNOSUPPORT;
1489	if (type < 0 \|\| type >= SOCK_MAX)
1490	return -EINVAL;
1491
1492	/* Compatibility.
1493
1494	This uglymoron is moved from INET layer to here to avoid
1495	deadlock in module load.
1496	*/
1497	if (family == PF_INET && type == SOCK_PACKET) {
1498	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1499	current->comm);
1500	family = PF_PACKET;
1501	}
1502
1503	err = security_socket_create(family, type, protocol, kern);
1504	if (err)
1505	return err;
1506
1507	/*
1508	* Allocate the socket and allow the family to set things up. if
1509	* the protocol is 0, the family is instructed to select an appropriate
1510	* default.
1511	*/
1512	sock = sock_alloc();
1513	if (!sock) {
1514	net_warn_ratelimited("socket: no more sockets\n");
1515	return -ENFILE; /* Not exactly a match, but its the
1516	closest posix thing */
1517	}
1518
1519	sock->type = type;
1520
1521	#ifdef CONFIG_MODULES
1522	/* Attempt to load a protocol module if the find failed.
1523	*
1524	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1525	* requested real, full-featured networking support upon configuration.
1526	* Otherwise module support will break!
1527	*/
1528	if (rcu_access_pointer(net_families[family]) == NULL)
1529	request_module("net-pf-%d", family);
1530	#endif
1531
1532	rcu_read_lock();
1533	pf = rcu_dereference(net_families[family]);
1534	err = -EAFNOSUPPORT;
1535	if (!pf)
1536	goto out_release;
1537
1538	/*
1539	* We will call the ->create function, that possibly is in a loadable
1540	* module, so we have to bump that loadable module refcnt first.
1541	*/
1542	if (!try_module_get(pf->owner))
1543	goto out_release;
1544
1545	/* Now protected by module ref count */
1546	rcu_read_unlock();
1547
1548	err = pf->create(net, sock, protocol, kern);
1549	if (err < 0) {
1550	/* ->create should release the allocated sock->sk object on error
1551	* and make sure sock->sk is set to NULL to avoid use-after-free
1552	*/
1553	DEBUG_NET_WARN_ONCE(sock->sk,
1554	"%ps must clear sock->sk on failure, family: %d, type: %d, protocol: %d\n",
1555	pf->create, family, type, protocol);
1556	goto out_module_put;
1557	}
1558
1559	/*
1560	* Now to bump the refcnt of the [loadable] module that owns this
1561	* socket at sock_release time we decrement its refcnt.
1562	*/
1563	if (!try_module_get(sock->ops->owner))
1564	goto out_module_busy;
1565
1566	/*
1567	* Now that we're done with the ->create function, the [loadable]
1568	* module can have its refcnt decremented
1569	*/
1570	module_put(pf->owner);
1571	err = security_socket_post_create(sock, family, type, protocol, kern);
1572	if (err)
1573	goto out_sock_release;
1574	*res = sock;
1575
1576	return 0;
1577
1578	out_module_busy:
1579	err = -EAFNOSUPPORT;
1580	out_module_put:
1581	sock->ops = NULL;
1582	module_put(pf->owner);
1583	out_sock_release:
1584	sock_release(sock);
1585	return err;
1586
1587	out_release:
1588	rcu_read_unlock();
1589	goto out_sock_release;
1590	}
1591	EXPORT_SYMBOL(__sock_create);
1592
1593	/**
1594	* sock_create - creates a socket
1595	* @family: protocol family (AF_INET, ...)
1596	* @type: communication type (SOCK_STREAM, ...)
1597	* @protocol: protocol (0, ...)
1598	* @res: new socket
1599	*
1600	* A wrapper around __sock_create().
1601	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1602	*/
1603
1604	int sock_create(int family, int type, int protocol, struct socket **res)
1605	{
1606	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1607	}
1608	EXPORT_SYMBOL(sock_create);
1609
1610	/**
1611	* sock_create_kern - creates a socket (kernel space)
1612	* @net: net namespace
1613	* @family: protocol family (AF_INET, ...)
1614	* @type: communication type (SOCK_STREAM, ...)
1615	* @protocol: protocol (0, ...)
1616	* @res: new socket
1617	*
1618	* A wrapper around __sock_create().
1619	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1620	*/
1621
1622	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1623	{
1624	return __sock_create(net, family, type, protocol, res, 1);
1625	}
1626	EXPORT_SYMBOL(sock_create_kern);
1627
1628	static struct socket *__sys_socket_create(int family, int type, int protocol)
1629	{
1630	struct socket *sock;
1631	int retval;
1632
1633	/* Check the SOCK_* constants for consistency. */
1634	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1635	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1636	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1637	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1638
1639	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1640	return ERR_PTR(-EINVAL);
1641	type &= SOCK_TYPE_MASK;
1642
1643	retval = sock_create(family, type, protocol, &sock);
1644	if (retval < 0)
1645	return ERR_PTR(retval);
1646
1647	return sock;
1648	}
1649
1650	struct file *__sys_socket_file(int family, int type, int protocol)
1651	{
1652	struct socket *sock;
1653	int flags;
1654
1655	sock = __sys_socket_create(family, type, protocol);
1656	if (IS_ERR(sock))
1657	return ERR_CAST(sock);
1658
1659	flags = type & ~SOCK_TYPE_MASK;
1660	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1661	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1662
1663	return sock_alloc_file(sock, flags, NULL);
1664	}
1665
1666	/* A hook for bpf progs to attach to and update socket protocol.
1667	*
1668	* A static noinline declaration here could cause the compiler to
1669	* optimize away the function. A global noinline declaration will
1670	* keep the definition, but may optimize away the callsite.
1671	* Therefore, __weak is needed to ensure that the call is still
1672	* emitted, by telling the compiler that we don't know what the
1673	* function might eventually be.
1674	*/
1675
1676	__bpf_hook_start();
1677
1678	__weak noinline int update_socket_protocol(int family, int type, int protocol)
1679	{
1680	return protocol;
1681	}
1682
1683	__bpf_hook_end();
1684
1685	int __sys_socket(int family, int type, int protocol)
1686	{
1687	struct socket *sock;
1688	int flags;
1689
1690	sock = __sys_socket_create(family, type,
1691	update_socket_protocol(family, type, protocol));
1692	if (IS_ERR(sock))
1693	return PTR_ERR(sock);
1694
1695	flags = type & ~SOCK_TYPE_MASK;
1696	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1697	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1698
1699	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1700	}
1701
1702	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1703	{
1704	return __sys_socket(family, type, protocol);
1705	}
1706
1707	/*
1708	* Create a pair of connected sockets.
1709	*/
1710
1711	int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1712	{
1713	struct socket sock1, sock2;
1714	int fd1, fd2, err;
1715	struct file newfile1, newfile2;
1716	int flags;
1717
1718	flags = type & ~SOCK_TYPE_MASK;
1719	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1720	return -EINVAL;
1721	type &= SOCK_TYPE_MASK;
1722
1723	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1724	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1725
1726	/*
1727	* reserve descriptors and make sure we won't fail
1728	* to return them to userland.
1729	*/
1730	fd1 = get_unused_fd_flags(flags);
1731	if (unlikely(fd1 < 0))
1732	return fd1;
1733
1734	fd2 = get_unused_fd_flags(flags);
1735	if (unlikely(fd2 < 0)) {
1736	put_unused_fd(fd1);
1737	return fd2;
1738	}
1739
1740	err = put_user(fd1, &usockvec[0]);
1741	if (err)
1742	goto out;
1743
1744	err = put_user(fd2, &usockvec[1]);
1745	if (err)
1746	goto out;
1747
1748	/*
1749	* Obtain the first socket and check if the underlying protocol
1750	* supports the socketpair call.
1751	*/
1752
1753	err = sock_create(family, type, protocol, &sock1);
1754	if (unlikely(err < 0))
1755	goto out;
1756
1757	err = sock_create(family, type, protocol, &sock2);
1758	if (unlikely(err < 0)) {
1759	sock_release(sock1);
1760	goto out;
1761	}
1762
1763	err = security_socket_socketpair(sock1, sock2);
1764	if (unlikely(err)) {
1765	sock_release(sock2);
1766	sock_release(sock1);
1767	goto out;
1768	}
1769
1770	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1771	if (unlikely(err < 0)) {
1772	sock_release(sock2);
1773	sock_release(sock1);
1774	goto out;
1775	}
1776
1777	newfile1 = sock_alloc_file(sock1, flags, NULL);
1778	if (IS_ERR(newfile1)) {
1779	err = PTR_ERR(newfile1);
1780	sock_release(sock2);
1781	goto out;
1782	}
1783
1784	newfile2 = sock_alloc_file(sock2, flags, NULL);
1785	if (IS_ERR(newfile2)) {
1786	err = PTR_ERR(newfile2);
1787	fput(newfile1);
1788	goto out;
1789	}
1790
1791	audit_fd_pair(fd1, fd2);
1792
1793	fd_install(fd1, newfile1);
1794	fd_install(fd2, newfile2);
1795	return 0;
1796
1797	out:
1798	put_unused_fd(fd2);
1799	put_unused_fd(fd1);
1800	return err;
1801	}
1802
1803	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1804	int __user *, usockvec)
1805	{
1806	return __sys_socketpair(family, type, protocol, usockvec);
1807	}
1808
1809	int __sys_bind_socket(struct socket sock, struct sockaddr_storage address,
1810	int addrlen)
1811	{
1812	int err;
1813
1814	err = security_socket_bind(sock, (struct sockaddr *)address,
1815	addrlen);
1816	if (!err)
1817	err = READ_ONCE(sock->ops)->bind(sock,
1818	(struct sockaddr *)address,
1819	addrlen);
1820	return err;
1821	}
1822
1823	/*
1824	* Bind a name to a socket. Nothing much to do here since it's
1825	* the protocol's responsibility to handle the local address.
1826	*
1827	* We move the socket address to kernel space before we call
1828	* the protocol layer (having also checked the address is ok).
1829	*/
1830
1831	int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1832	{
1833	struct socket *sock;
1834	struct sockaddr_storage address;
1835	CLASS(fd, f)(fd);
1836	int err;
1837
1838	if (fd_empty(f))
1839	return -EBADF;
1840	sock = sock_from_file(fd_file(f));
1841	if (unlikely(!sock))
1842	return -ENOTSOCK;
1843
1844	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1845	if (unlikely(err))
1846	return err;
1847
1848	return __sys_bind_socket(sock, &address, addrlen);
1849	}
1850
1851	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1852	{
1853	return __sys_bind(fd, umyaddr, addrlen);
1854	}
1855
1856	/*
1857	* Perform a listen. Basically, we allow the protocol to do anything
1858	* necessary for a listen, and if that works, we mark the socket as
1859	* ready for listening.
1860	*/
1861	int __sys_listen_socket(struct socket *sock, int backlog)
1862	{
1863	int somaxconn, err;
1864
1865	somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1866	if ((unsigned int)backlog > somaxconn)
1867	backlog = somaxconn;
1868
1869	err = security_socket_listen(sock, backlog);
1870	if (!err)
1871	err = READ_ONCE(sock->ops)->listen(sock, backlog);
1872	return err;
1873	}
1874
1875	int __sys_listen(int fd, int backlog)
1876	{
1877	CLASS(fd, f)(fd);
1878	struct socket *sock;
1879
1880	if (fd_empty(f))
1881	return -EBADF;
1882	sock = sock_from_file(fd_file(f));
1883	if (unlikely(!sock))
1884	return -ENOTSOCK;
1885
1886	return __sys_listen_socket(sock, backlog);
1887	}
1888
1889	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1890	{
1891	return __sys_listen(fd, backlog);
1892	}
1893
1894	struct file do_accept(struct file file, struct proto_accept_arg *arg,
1895	struct sockaddr __user *upeer_sockaddr,
1896	int __user *upeer_addrlen, int flags)
1897	{
1898	struct socket sock, newsock;
1899	struct file *newfile;
1900	int err, len;
1901	struct sockaddr_storage address;
1902	const struct proto_ops *ops;
1903
1904	sock = sock_from_file(file);
1905	if (!sock)
1906	return ERR_PTR(-ENOTSOCK);
1907
1908	newsock = sock_alloc();
1909	if (!newsock)
1910	return ERR_PTR(-ENFILE);
1911	ops = READ_ONCE(sock->ops);
1912
1913	newsock->type = sock->type;
1914	newsock->ops = ops;
1915
1916	/*
1917	* We don't need try_module_get here, as the listening socket (sock)
1918	* has the protocol module (sock->ops->owner) held.
1919	*/
1920	__module_get(ops->owner);
1921
1922	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1923	if (IS_ERR(newfile))
1924	return newfile;
1925
1926	err = security_socket_accept(sock, newsock);
1927	if (err)
1928	goto out_fd;
1929
1930	arg->flags \|= sock->file->f_flags;
1931	err = ops->accept(sock, newsock, arg);
1932	if (err < 0)
1933	goto out_fd;
1934
1935	if (upeer_sockaddr) {
1936	len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1937	if (len < 0) {
1938	err = -ECONNABORTED;
1939	goto out_fd;
1940	}
1941	err = move_addr_to_user(&address,
1942	len, upeer_sockaddr, upeer_addrlen);
1943	if (err < 0)
1944	goto out_fd;
1945	}
1946
1947	/* File flags are not inherited via accept() unlike another OSes. */
1948	return newfile;
1949	out_fd:
1950	fput(newfile);
1951	return ERR_PTR(err);
1952	}
1953
1954	static int __sys_accept4_file(struct file file, struct sockaddr __user upeer_sockaddr,
1955	int __user *upeer_addrlen, int flags)
1956	{
1957	struct proto_accept_arg arg = { };
1958	struct file *newfile;
1959	int newfd;
1960
1961	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1962	return -EINVAL;
1963
1964	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1965	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1966
1967	newfd = get_unused_fd_flags(flags);
1968	if (unlikely(newfd < 0))
1969	return newfd;
1970
1971	newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1972	flags);
1973	if (IS_ERR(newfile)) {
1974	put_unused_fd(newfd);
1975	return PTR_ERR(newfile);
1976	}
1977	fd_install(newfd, newfile);
1978	return newfd;
1979	}
1980
1981	/*
1982	* For accept, we attempt to create a new socket, set up the link
1983	* with the client, wake up the client, then return the new
1984	* connected fd. We collect the address of the connector in kernel
1985	* space and move it to user at the very end. This is unclean because
1986	* we open the socket then return an error.
1987	*
1988	* 1003.1g adds the ability to recvmsg() to query connection pending
1989	* status to recvmsg. We need to add that support in a way thats
1990	* clean when we restructure accept also.
1991	*/
1992
1993	int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1994	int __user *upeer_addrlen, int flags)
1995	{
1996	CLASS(fd, f)(fd);
1997
1998	if (fd_empty(f))
1999	return -EBADF;
2000	return __sys_accept4_file(fd_file(f), upeer_sockaddr,
2001	upeer_addrlen, flags);
2002	}
2003
2004	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2005	int __user *, upeer_addrlen, int, flags)
2006	{
2007	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2008	}
2009
2010	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2011	int __user *, upeer_addrlen)
2012	{
2013	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2014	}
2015
2016	/*
2017	* Attempt to connect to a socket with the server address. The address
2018	* is in user space so we verify it is OK and move it to kernel space.
2019	*
2020	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2021	* break bindings
2022	*
2023	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2024	* other SEQPACKET protocols that take time to connect() as it doesn't
2025	* include the -EINPROGRESS status for such sockets.
2026	*/
2027
2028	int __sys_connect_file(struct file file, struct sockaddr_storage address,
2029	int addrlen, int file_flags)
2030	{
2031	struct socket *sock;
2032	int err;
2033
2034	sock = sock_from_file(file);
2035	if (!sock) {
2036	err = -ENOTSOCK;
2037	goto out;
2038	}
2039
2040	err =
2041	security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2042	if (err)
2043	goto out;
2044
2045	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2046	addrlen, sock->file->f_flags \| file_flags);
2047	out:
2048	return err;
2049	}
2050
2051	int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2052	{
2053	struct sockaddr_storage address;
2054	CLASS(fd, f)(fd);
2055	int ret;
2056
2057	if (fd_empty(f))
2058	return -EBADF;
2059
2060	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2061	if (ret)
2062	return ret;
2063
2064	return __sys_connect_file(fd_file(f), &address, addrlen, 0);
2065	}
2066
2067	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2068	int, addrlen)
2069	{
2070	return __sys_connect(fd, uservaddr, addrlen);
2071	}
2072
2073	/*
2074	* Get the local address ('name') of a socket object. Move the obtained
2075	* name to user space.
2076	*/
2077
2078	int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2079	int __user *usockaddr_len)
2080	{
2081	struct socket *sock;
2082	struct sockaddr_storage address;
2083	CLASS(fd, f)(fd);
2084	int err;
2085
2086	if (fd_empty(f))
2087	return -EBADF;
2088	sock = sock_from_file(fd_file(f));
2089	if (unlikely(!sock))
2090	return -ENOTSOCK;
2091
2092	err = security_socket_getsockname(sock);
2093	if (err)
2094	return err;
2095
2096	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2097	if (err < 0)
2098	return err;
2099
2100	/* "err" is actually length in this case */
2101	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2102	}
2103
2104	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2105	int __user *, usockaddr_len)
2106	{
2107	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2108	}
2109
2110	/*
2111	* Get the remote address ('name') of a socket object. Move the obtained
2112	* name to user space.
2113	*/
2114
2115	int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2116	int __user *usockaddr_len)
2117	{
2118	struct socket *sock;
2119	struct sockaddr_storage address;
2120	CLASS(fd, f)(fd);
2121	int err;
2122
2123	if (fd_empty(f))
2124	return -EBADF;
2125	sock = sock_from_file(fd_file(f));
2126	if (unlikely(!sock))
2127	return -ENOTSOCK;
2128
2129	err = security_socket_getpeername(sock);
2130	if (err)
2131	return err;
2132
2133	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 1);
2134	if (err < 0)
2135	return err;
2136
2137	/* "err" is actually length in this case */
2138	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2139	}
2140
2141	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2142	int __user *, usockaddr_len)
2143	{
2144	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2145	}
2146
2147	/*
2148	* Send a datagram to a given address. We move the address into kernel
2149	* space and check the user space data area is readable before invoking
2150	* the protocol.
2151	*/
2152	int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2153	struct sockaddr __user *addr, int addr_len)
2154	{
2155	struct socket *sock;
2156	struct sockaddr_storage address;
2157	int err;
2158	struct msghdr msg;
2159
2160	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2161	if (unlikely(err))
2162	return err;
2163
2164	CLASS(fd, f)(fd);
2165	if (fd_empty(f))
2166	return -EBADF;
2167	sock = sock_from_file(fd_file(f));
2168	if (unlikely(!sock))
2169	return -ENOTSOCK;
2170
2171	msg.msg_name = NULL;
2172	msg.msg_control = NULL;
2173	msg.msg_controllen = 0;
2174	msg.msg_namelen = 0;
2175	msg.msg_ubuf = NULL;
2176	if (addr) {
2177	err = move_addr_to_kernel(addr, addr_len, &address);
2178	if (err < 0)
2179	return err;
2180	msg.msg_name = (struct sockaddr *)&address;
2181	msg.msg_namelen = addr_len;
2182	}
2183	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2184	if (sock->file->f_flags & O_NONBLOCK)
2185	flags \|= MSG_DONTWAIT;
2186	msg.msg_flags = flags;
2187	return __sock_sendmsg(sock, &msg);
2188	}
2189
2190	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2191	unsigned int, flags, struct sockaddr __user *, addr,
2192	int, addr_len)
2193	{
2194	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2195	}
2196
2197	/*
2198	* Send a datagram down a socket.
2199	*/
2200
2201	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2202	unsigned int, flags)
2203	{
2204	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2205	}
2206
2207	/*
2208	* Receive a frame from the socket and optionally record the address of the
2209	* sender. We verify the buffers are writable and if needed move the
2210	* sender address from kernel to user space.
2211	*/
2212	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2213	struct sockaddr __user addr, int __user addr_len)
2214	{
2215	struct sockaddr_storage address;
2216	struct msghdr msg = {
2217	/* Save some cycles and don't copy the address if not needed */
2218	.msg_name = addr ? (struct sockaddr *)&address : NULL,
2219	};
2220	struct socket *sock;
2221	int err, err2;
2222
2223	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2224	if (unlikely(err))
2225	return err;
2226
2227	CLASS(fd, f)(fd);
2228
2229	if (fd_empty(f))
2230	return -EBADF;
2231	sock = sock_from_file(fd_file(f));
2232	if (unlikely(!sock))
2233	return -ENOTSOCK;
2234
2235	if (sock->file->f_flags & O_NONBLOCK)
2236	flags \|= MSG_DONTWAIT;
2237	err = sock_recvmsg(sock, &msg, flags);
2238
2239	if (err >= 0 && addr != NULL) {
2240	err2 = move_addr_to_user(&address,
2241	msg.msg_namelen, addr, addr_len);
2242	if (err2 < 0)
2243	err = err2;
2244	}
2245	return err;
2246	}
2247
2248	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2249	unsigned int, flags, struct sockaddr __user *, addr,
2250	int __user *, addr_len)
2251	{
2252	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2253	}
2254
2255	/*
2256	* Receive a datagram from a socket.
2257	*/
2258
2259	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2260	unsigned int, flags)
2261	{
2262	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2263	}
2264
2265	static bool sock_use_custom_sol_socket(const struct socket *sock)
2266	{
2267	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2268	}
2269
2270	int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2271	int optname, sockptr_t optval, int optlen)
2272	{
2273	const struct proto_ops *ops;
2274	char *kernel_optval = NULL;
2275	int err;
2276
2277	if (optlen < 0)
2278	return -EINVAL;
2279
2280	err = security_socket_setsockopt(sock, level, optname);
2281	if (err)
2282	goto out_put;
2283
2284	if (!compat)
2285	err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2286	optval, &optlen,
2287	&kernel_optval);
2288	if (err < 0)
2289	goto out_put;
2290	if (err > 0) {
2291	err = 0;
2292	goto out_put;
2293	}
2294
2295	if (kernel_optval)
2296	optval = KERNEL_SOCKPTR(kernel_optval);
2297	ops = READ_ONCE(sock->ops);
2298	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2299	err = sock_setsockopt(sock, level, optname, optval, optlen);
2300	else if (unlikely(!ops->setsockopt))
2301	err = -EOPNOTSUPP;
2302	else
2303	err = ops->setsockopt(sock, level, optname, optval,
2304	optlen);
2305	kfree(kernel_optval);
2306	out_put:
2307	return err;
2308	}
2309	EXPORT_SYMBOL(do_sock_setsockopt);
2310
2311	/* Set a socket option. Because we don't know the option lengths we have
2312	* to pass the user mode parameter for the protocols to sort out.
2313	*/
2314	int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2315	int optlen)
2316	{
2317	sockptr_t optval = USER_SOCKPTR(user_optval);
2318	bool compat = in_compat_syscall();
2319	struct socket *sock;
2320	CLASS(fd, f)(fd);
2321
2322	if (fd_empty(f))
2323	return -EBADF;
2324	sock = sock_from_file(fd_file(f));
2325	if (unlikely(!sock))
2326	return -ENOTSOCK;
2327
2328	return do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2329	}
2330
2331	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2332	char __user *, optval, int, optlen)
2333	{
2334	return __sys_setsockopt(fd, level, optname, optval, optlen);
2335	}
2336
2337	INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2338	int optname));
2339
2340	int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2341	int optname, sockptr_t optval, sockptr_t optlen)
2342	{
2343	int max_optlen __maybe_unused = 0;
2344	const struct proto_ops *ops;
2345	int err;
2346
2347	err = security_socket_getsockopt(sock, level, optname);
2348	if (err)
2349	return err;
2350
2351	if (!compat)
2352	copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2353
2354	ops = READ_ONCE(sock->ops);
2355	if (level == SOL_SOCKET) {
2356	err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2357	} else if (unlikely(!ops->getsockopt)) {
2358	err = -EOPNOTSUPP;
2359	} else {
2360	if (WARN_ONCE(optval.is_kernel \|\| optlen.is_kernel,
2361	"Invalid argument type"))
2362	return -EOPNOTSUPP;
2363
2364	err = ops->getsockopt(sock, level, optname, optval.user,
2365	optlen.user);
2366	}
2367
2368	if (!compat)
2369	err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2370	optval, optlen, max_optlen,
2371	err);
2372
2373	return err;
2374	}
2375	EXPORT_SYMBOL(do_sock_getsockopt);
2376
2377	/*
2378	* Get a socket option. Because we don't know the option lengths we have
2379	* to pass a user mode parameter for the protocols to sort out.
2380	*/
2381	int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2382	int __user *optlen)
2383	{
2384	struct socket *sock;
2385	CLASS(fd, f)(fd);
2386
2387	if (fd_empty(f))
2388	return -EBADF;
2389	sock = sock_from_file(fd_file(f));
2390	if (unlikely(!sock))
2391	return -ENOTSOCK;
2392
2393	return do_sock_getsockopt(sock, in_compat_syscall(), level, optname,
2394	USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2395	}
2396
2397	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2398	char __user , optval, int __user , optlen)
2399	{
2400	return __sys_getsockopt(fd, level, optname, optval, optlen);
2401	}
2402
2403	/*
2404	* Shutdown a socket.
2405	*/
2406
2407	int __sys_shutdown_sock(struct socket *sock, int how)
2408	{
2409	int err;
2410
2411	err = security_socket_shutdown(sock, how);
2412	if (!err)
2413	err = READ_ONCE(sock->ops)->shutdown(sock, how);
2414
2415	return err;
2416	}
2417
2418	int __sys_shutdown(int fd, int how)
2419	{
2420	struct socket *sock;
2421	CLASS(fd, f)(fd);
2422
2423	if (fd_empty(f))
2424	return -EBADF;
2425	sock = sock_from_file(fd_file(f));
2426	if (unlikely(!sock))
2427	return -ENOTSOCK;
2428
2429	return __sys_shutdown_sock(sock, how);
2430	}
2431
2432	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2433	{
2434	return __sys_shutdown(fd, how);
2435	}
2436
2437	/* A couple of helpful macros for getting the address of the 32/64 bit
2438	* fields which are the same type (int / unsigned) on our platforms.
2439	*/
2440	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2441	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2442	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2443
2444	struct used_address {
2445	struct sockaddr_storage name;
2446	unsigned int name_len;
2447	};
2448
2449	int __copy_msghdr(struct msghdr *kmsg,
2450	struct user_msghdr *msg,
2451	struct sockaddr __user **save_addr)
2452	{
2453	ssize_t err;
2454
2455	kmsg->msg_control_is_user = true;
2456	kmsg->msg_get_inq = 0;
2457	kmsg->msg_control_user = msg->msg_control;
2458	kmsg->msg_controllen = msg->msg_controllen;
2459	kmsg->msg_flags = msg->msg_flags;
2460
2461	kmsg->msg_namelen = msg->msg_namelen;
2462	if (!msg->msg_name)
2463	kmsg->msg_namelen = 0;
2464
2465	if (kmsg->msg_namelen < 0)
2466	return -EINVAL;
2467
2468	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2469	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2470
2471	if (save_addr)
2472	*save_addr = msg->msg_name;
2473
2474	if (msg->msg_name && kmsg->msg_namelen) {
2475	if (!save_addr) {
2476	err = move_addr_to_kernel(msg->msg_name,
2477	kmsg->msg_namelen,
2478	kmsg->msg_name);
2479	if (err < 0)
2480	return err;
2481	}
2482	} else {
2483	kmsg->msg_name = NULL;
2484	kmsg->msg_namelen = 0;
2485	}
2486
2487	if (msg->msg_iovlen > UIO_MAXIOV)
2488	return -EMSGSIZE;
2489
2490	kmsg->msg_iocb = NULL;
2491	kmsg->msg_ubuf = NULL;
2492	return 0;
2493	}
2494
2495	static int copy_msghdr_from_user(struct msghdr *kmsg,
2496	struct user_msghdr __user *umsg,
2497	struct sockaddr __user **save_addr,
2498	struct iovec **iov)
2499	{
2500	struct user_msghdr msg;
2501	ssize_t err;
2502
2503	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2504	return -EFAULT;
2505
2506	err = __copy_msghdr(kmsg, &msg, save_addr);
2507	if (err)
2508	return err;
2509
2510	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2511	msg.msg_iov, msg.msg_iovlen,
2512	UIO_FASTIOV, iov, &kmsg->msg_iter);
2513	return err < 0 ? err : 0;
2514	}
2515
2516	static int ____sys_sendmsg(struct socket sock, struct msghdr msg_sys,
2517	unsigned int flags, struct used_address *used_address,
2518	unsigned int allowed_msghdr_flags)
2519	{
2520	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2521	__aligned(sizeof(__kernel_size_t));
2522	/* 20 is size of ipv6_pktinfo */
2523	unsigned char *ctl_buf = ctl;
2524	int ctl_len;
2525	ssize_t err;
2526
2527	err = -ENOBUFS;
2528
2529	if (msg_sys->msg_controllen > INT_MAX)
2530	goto out;
2531	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2532	ctl_len = msg_sys->msg_controllen;
2533	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2534	err =
2535	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2536	sizeof(ctl));
2537	if (err)
2538	goto out;
2539	ctl_buf = msg_sys->msg_control;
2540	ctl_len = msg_sys->msg_controllen;
2541	} else if (ctl_len) {
2542	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2543	CMSG_ALIGN(sizeof(struct cmsghdr)));
2544	if (ctl_len > sizeof(ctl)) {
2545	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2546	if (ctl_buf == NULL)
2547	goto out;
2548	}
2549	err = -EFAULT;
2550	if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2551	goto out_freectl;
2552	msg_sys->msg_control = ctl_buf;
2553	msg_sys->msg_control_is_user = false;
2554	}
2555	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2556	msg_sys->msg_flags = flags;
2557
2558	if (sock->file->f_flags & O_NONBLOCK)
2559	msg_sys->msg_flags \|= MSG_DONTWAIT;
2560	/*
2561	* If this is sendmmsg() and current destination address is same as
2562	* previously succeeded address, omit asking LSM's decision.
2563	* used_address->name_len is initialized to UINT_MAX so that the first
2564	* destination address never matches.
2565	*/
2566	if (used_address && msg_sys->msg_name &&
2567	used_address->name_len == msg_sys->msg_namelen &&
2568	!memcmp(&used_address->name, msg_sys->msg_name,
2569	used_address->name_len)) {
2570	err = sock_sendmsg_nosec(sock, msg_sys);
2571	goto out_freectl;
2572	}
2573	err = __sock_sendmsg(sock, msg_sys);
2574	/*
2575	* If this is sendmmsg() and sending to current destination address was
2576	* successful, remember it.
2577	*/
2578	if (used_address && err >= 0) {
2579	used_address->name_len = msg_sys->msg_namelen;
2580	if (msg_sys->msg_name)
2581	memcpy(&used_address->name, msg_sys->msg_name,
2582	used_address->name_len);
2583	}
2584
2585	out_freectl:
2586	if (ctl_buf != ctl)
2587	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2588	out:
2589	return err;
2590	}
2591
2592	static int sendmsg_copy_msghdr(struct msghdr *msg,
2593	struct user_msghdr __user *umsg, unsigned flags,
2594	struct iovec **iov)
2595	{
2596	int err;
2597
2598	if (flags & MSG_CMSG_COMPAT) {
2599	struct compat_msghdr __user *msg_compat;
2600
2601	msg_compat = (struct compat_msghdr __user *) umsg;
2602	err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2603	} else {
2604	err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2605	}
2606	if (err < 0)
2607	return err;
2608
2609	return 0;
2610	}
2611
2612	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
2613	struct msghdr *msg_sys, unsigned int flags,
2614	struct used_address *used_address,
2615	unsigned int allowed_msghdr_flags)
2616	{
2617	struct sockaddr_storage address;
2618	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2619	ssize_t err;
2620
2621	msg_sys->msg_name = &address;
2622
2623	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2624	if (err < 0)
2625	return err;
2626
2627	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2628	allowed_msghdr_flags);
2629	kfree(iov);
2630	return err;
2631	}
2632
2633	/*
2634	* BSD sendmsg interface
2635	*/
2636	long __sys_sendmsg_sock(struct socket sock, struct msghdr msg,
2637	unsigned int flags)
2638	{
2639	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2640	}
2641
2642	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2643	bool forbid_cmsg_compat)
2644	{
2645	struct msghdr msg_sys;
2646	struct socket *sock;
2647
2648	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2649	return -EINVAL;
2650
2651	CLASS(fd, f)(fd);
2652
2653	if (fd_empty(f))
2654	return -EBADF;
2655	sock = sock_from_file(fd_file(f));
2656	if (unlikely(!sock))
2657	return -ENOTSOCK;
2658
2659	return ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2660	}
2661
2662	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2663	{
2664	return __sys_sendmsg(fd, msg, flags, true);
2665	}
2666
2667	/*
2668	* Linux sendmmsg interface
2669	*/
2670
2671	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2672	unsigned int flags, bool forbid_cmsg_compat)
2673	{
2674	int err, datagrams;
2675	struct socket *sock;
2676	struct mmsghdr __user *entry;
2677	struct compat_mmsghdr __user *compat_entry;
2678	struct msghdr msg_sys;
2679	struct used_address used_address;
2680	unsigned int oflags = flags;
2681
2682	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2683	return -EINVAL;
2684
2685	if (vlen > UIO_MAXIOV)
2686	vlen = UIO_MAXIOV;
2687
2688	datagrams = 0;
2689
2690	CLASS(fd, f)(fd);
2691
2692	if (fd_empty(f))
2693	return -EBADF;
2694	sock = sock_from_file(fd_file(f));
2695	if (unlikely(!sock))
2696	return -ENOTSOCK;
2697
2698	used_address.name_len = UINT_MAX;
2699	entry = mmsg;
2700	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2701	err = 0;
2702	flags \|= MSG_BATCH;
2703
2704	while (datagrams < vlen) {
2705	if (datagrams == vlen - 1)
2706	flags = oflags;
2707
2708	if (MSG_CMSG_COMPAT & flags) {
2709	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2710	&msg_sys, flags, &used_address, MSG_EOR);
2711	if (err < 0)
2712	break;
2713	err = __put_user(err, &compat_entry->msg_len);
2714	++compat_entry;
2715	} else {
2716	err = ___sys_sendmsg(sock,
2717	(struct user_msghdr __user *)entry,
2718	&msg_sys, flags, &used_address, MSG_EOR);
2719	if (err < 0)
2720	break;
2721	err = put_user(err, &entry->msg_len);
2722	++entry;
2723	}
2724
2725	if (err)
2726	break;
2727	++datagrams;
2728	if (msg_data_left(&msg_sys))
2729	break;
2730	cond_resched();
2731	}
2732
2733	/* We only return an error if no datagrams were able to be sent */
2734	if (datagrams != 0)
2735	return datagrams;
2736
2737	return err;
2738	}
2739
2740	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2741	unsigned int, vlen, unsigned int, flags)
2742	{
2743	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2744	}
2745
2746	static int recvmsg_copy_msghdr(struct msghdr *msg,
2747	struct user_msghdr __user *umsg, unsigned flags,
2748	struct sockaddr __user **uaddr,
2749	struct iovec **iov)
2750	{
2751	ssize_t err;
2752
2753	if (MSG_CMSG_COMPAT & flags) {
2754	struct compat_msghdr __user *msg_compat;
2755
2756	msg_compat = (struct compat_msghdr __user *) umsg;
2757	err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2758	} else {
2759	err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2760	}
2761	if (err < 0)
2762	return err;
2763
2764	return 0;
2765	}
2766
2767	static int ____sys_recvmsg(struct socket sock, struct msghdr msg_sys,
2768	struct user_msghdr __user *msg,
2769	struct sockaddr __user *uaddr,
2770	unsigned int flags, int nosec)
2771	{
2772	struct compat_msghdr __user *msg_compat =
2773	(struct compat_msghdr __user *) msg;
2774	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2775	struct sockaddr_storage addr;
2776	unsigned long cmsg_ptr;
2777	int len;
2778	ssize_t err;
2779
2780	msg_sys->msg_name = &addr;
2781	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2782	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2783
2784	/* We assume all kernel code knows the size of sockaddr_storage */
2785	msg_sys->msg_namelen = 0;
2786
2787	if (sock->file->f_flags & O_NONBLOCK)
2788	flags \|= MSG_DONTWAIT;
2789
2790	if (unlikely(nosec))
2791	err = sock_recvmsg_nosec(sock, msg_sys, flags);
2792	else
2793	err = sock_recvmsg(sock, msg_sys, flags);
2794
2795	if (err < 0)
2796	goto out;
2797	len = err;
2798
2799	if (uaddr != NULL) {
2800	err = move_addr_to_user(&addr,
2801	msg_sys->msg_namelen, uaddr,
2802	uaddr_len);
2803	if (err < 0)
2804	goto out;
2805	}
2806	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2807	COMPAT_FLAGS(msg));
2808	if (err)
2809	goto out;
2810	if (MSG_CMSG_COMPAT & flags)
2811	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2812	&msg_compat->msg_controllen);
2813	else
2814	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2815	&msg->msg_controllen);
2816	if (err)
2817	goto out;
2818	err = len;
2819	out:
2820	return err;
2821	}
2822
2823	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2824	struct msghdr *msg_sys, unsigned int flags, int nosec)
2825	{
2826	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2827	/* user mode address pointers */
2828	struct sockaddr __user *uaddr;
2829	ssize_t err;
2830
2831	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2832	if (err < 0)
2833	return err;
2834
2835	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2836	kfree(iov);
2837	return err;
2838	}
2839
2840	/*
2841	* BSD recvmsg interface
2842	*/
2843
2844	long __sys_recvmsg_sock(struct socket sock, struct msghdr msg,
2845	struct user_msghdr __user *umsg,
2846	struct sockaddr __user *uaddr, unsigned int flags)
2847	{
2848	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2849	}
2850
2851	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2852	bool forbid_cmsg_compat)
2853	{
2854	struct msghdr msg_sys;
2855	struct socket *sock;
2856
2857	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2858	return -EINVAL;
2859
2860	CLASS(fd, f)(fd);
2861
2862	if (fd_empty(f))
2863	return -EBADF;
2864	sock = sock_from_file(fd_file(f));
2865	if (unlikely(!sock))
2866	return -ENOTSOCK;
2867
2868	return ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2869	}
2870
2871	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2872	unsigned int, flags)
2873	{
2874	return __sys_recvmsg(fd, msg, flags, true);
2875	}
2876
2877	/*
2878	* Linux recvmmsg interface
2879	*/
2880
2881	static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2882	unsigned int vlen, unsigned int flags,
2883	struct timespec64 *timeout)
2884	{
2885	int err = 0, datagrams;
2886	struct socket *sock;
2887	struct mmsghdr __user *entry;
2888	struct compat_mmsghdr __user *compat_entry;
2889	struct msghdr msg_sys;
2890	struct timespec64 end_time;
2891	struct timespec64 timeout64;
2892
2893	if (timeout &&
2894	poll_select_set_timeout(&end_time, timeout->tv_sec,
2895	timeout->tv_nsec))
2896	return -EINVAL;
2897
2898	datagrams = 0;
2899
2900	CLASS(fd, f)(fd);
2901
2902	if (fd_empty(f))
2903	return -EBADF;
2904	sock = sock_from_file(fd_file(f));
2905	if (unlikely(!sock))
2906	return -ENOTSOCK;
2907
2908	if (likely(!(flags & MSG_ERRQUEUE))) {
2909	err = sock_error(sock->sk);
2910	if (err)
2911	return err;
2912	}
2913
2914	entry = mmsg;
2915	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2916
2917	while (datagrams < vlen) {
2918	/*
2919	* No need to ask LSM for more than the first datagram.
2920	*/
2921	if (MSG_CMSG_COMPAT & flags) {
2922	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2923	&msg_sys, flags & ~MSG_WAITFORONE,
2924	datagrams);
2925	if (err < 0)
2926	break;
2927	err = __put_user(err, &compat_entry->msg_len);
2928	++compat_entry;
2929	} else {
2930	err = ___sys_recvmsg(sock,
2931	(struct user_msghdr __user *)entry,
2932	&msg_sys, flags & ~MSG_WAITFORONE,
2933	datagrams);
2934	if (err < 0)
2935	break;
2936	err = put_user(err, &entry->msg_len);
2937	++entry;
2938	}
2939
2940	if (err)
2941	break;
2942	++datagrams;
2943
2944	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2945	if (flags & MSG_WAITFORONE)
2946	flags \|= MSG_DONTWAIT;
2947
2948	if (timeout) {
2949	ktime_get_ts64(&timeout64);
2950	*timeout = timespec64_sub(end_time, timeout64);
2951	if (timeout->tv_sec < 0) {
2952	timeout->tv_sec = timeout->tv_nsec = 0;
2953	break;
2954	}
2955
2956	/* Timeout, return less than vlen datagrams */
2957	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2958	break;
2959	}
2960
2961	/* Out of band data, return right away */
2962	if (msg_sys.msg_flags & MSG_OOB)
2963	break;
2964	cond_resched();
2965	}
2966
2967	if (err == 0)
2968	return datagrams;
2969
2970	if (datagrams == 0)
2971	return err;
2972
2973	/*
2974	* We may return less entries than requested (vlen) if the
2975	* sock is non block and there aren't enough datagrams...
2976	*/
2977	if (err != -EAGAIN) {
2978	/*
2979	* ... or if recvmsg returns an error after we
2980	* received some datagrams, where we record the
2981	* error to return on the next call or if the
2982	* app asks about it using getsockopt(SO_ERROR).
2983	*/
2984	WRITE_ONCE(sock->sk->sk_err, -err);
2985	}
2986	return datagrams;
2987	}
2988
2989	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2990	unsigned int vlen, unsigned int flags,
2991	struct __kernel_timespec __user *timeout,
2992	struct old_timespec32 __user *timeout32)
2993	{
2994	int datagrams;
2995	struct timespec64 timeout_sys;
2996
2997	if (timeout && get_timespec64(&timeout_sys, timeout))
2998	return -EFAULT;
2999
3000	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3001	return -EFAULT;
3002
3003	if (!timeout && !timeout32)
3004	return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3005
3006	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3007
3008	if (datagrams <= 0)
3009	return datagrams;
3010
3011	if (timeout && put_timespec64(&timeout_sys, timeout))
3012	datagrams = -EFAULT;
3013
3014	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3015	datagrams = -EFAULT;
3016
3017	return datagrams;
3018	}
3019
3020	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3021	unsigned int, vlen, unsigned int, flags,
3022	struct __kernel_timespec __user *, timeout)
3023	{
3024	if (flags & MSG_CMSG_COMPAT)
3025	return -EINVAL;
3026
3027	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3028	}
3029
3030	#ifdef CONFIG_COMPAT_32BIT_TIME
3031	SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3032	unsigned int, vlen, unsigned int, flags,
3033	struct old_timespec32 __user *, timeout)
3034	{
3035	if (flags & MSG_CMSG_COMPAT)
3036	return -EINVAL;
3037
3038	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3039	}
3040	#endif
3041
3042	#ifdef __ARCH_WANT_SYS_SOCKETCALL
3043	/* Argument list sizes for sys_socketcall */
3044	#define AL(x) ((x) * sizeof(unsigned long))
3045	static const unsigned char nargs[21] = {
3046	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3047	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3048	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3049	AL(4), AL(5), AL(4)
3050	};
3051
3052	#undef AL
3053
3054	/*
3055	* System call vectors.
3056	*
3057	* Argument checking cleaned up. Saved 20% in size.
3058	* This function doesn't need to set the kernel lock because
3059	* it is set by the callees.
3060	*/
3061
3062	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3063	{
3064	unsigned long a[AUDITSC_ARGS];
3065	unsigned long a0, a1;
3066	int err;
3067	unsigned int len;
3068
3069	if (call < 1 \|\| call > SYS_SENDMMSG)
3070	return -EINVAL;
3071	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3072
3073	len = nargs[call];
3074	if (len > sizeof(a))
3075	return -EINVAL;
3076
3077	/* copy_from_user should be SMP safe. */
3078	if (copy_from_user(a, args, len))
3079	return -EFAULT;
3080
3081	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3082	if (err)
3083	return err;
3084
3085	a0 = a[0];
3086	a1 = a[1];
3087
3088	switch (call) {
3089	case SYS_SOCKET:
3090	err = __sys_socket(a0, a1, a[2]);
3091	break;
3092	case SYS_BIND:
3093	err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3094	break;
3095	case SYS_CONNECT:
3096	err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3097	break;
3098	case SYS_LISTEN:
3099	err = __sys_listen(a0, a1);
3100	break;
3101	case SYS_ACCEPT:
3102	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3103	(int __user *)a[2], 0);
3104	break;
3105	case SYS_GETSOCKNAME:
3106	err =
3107	__sys_getsockname(a0, (struct sockaddr __user *)a1,
3108	(int __user *)a[2]);
3109	break;
3110	case SYS_GETPEERNAME:
3111	err =
3112	__sys_getpeername(a0, (struct sockaddr __user *)a1,
3113	(int __user *)a[2]);
3114	break;
3115	case SYS_SOCKETPAIR:
3116	err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3117	break;
3118	case SYS_SEND:
3119	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3120	NULL, 0);
3121	break;
3122	case SYS_SENDTO:
3123	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3124	(struct sockaddr __user *)a[4], a[5]);
3125	break;
3126	case SYS_RECV:
3127	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3128	NULL, NULL);
3129	break;
3130	case SYS_RECVFROM:
3131	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3132	(struct sockaddr __user *)a[4],
3133	(int __user *)a[5]);
3134	break;
3135	case SYS_SHUTDOWN:
3136	err = __sys_shutdown(a0, a1);
3137	break;
3138	case SYS_SETSOCKOPT:
3139	err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3140	a[4]);
3141	break;
3142	case SYS_GETSOCKOPT:
3143	err =
3144	__sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3145	(int __user *)a[4]);
3146	break;
3147	case SYS_SENDMSG:
3148	err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3149	a[2], true);
3150	break;
3151	case SYS_SENDMMSG:
3152	err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3153	a[3], true);
3154	break;
3155	case SYS_RECVMSG:
3156	err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3157	a[2], true);
3158	break;
3159	case SYS_RECVMMSG:
3160	if (IS_ENABLED(CONFIG_64BIT))
3161	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3162	a[2], a[3],
3163	(struct __kernel_timespec __user *)a[4],
3164	NULL);
3165	else
3166	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3167	a[2], a[3], NULL,
3168	(struct old_timespec32 __user *)a[4]);
3169	break;
3170	case SYS_ACCEPT4:
3171	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3172	(int __user *)a[2], a[3]);
3173	break;
3174	default:
3175	err = -EINVAL;
3176	break;
3177	}
3178	return err;
3179	}
3180
3181	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
3182
3183	/**
3184	* sock_register - add a socket protocol handler
3185	* @ops: description of protocol
3186	*
3187	* This function is called by a protocol handler that wants to
3188	* advertise its address family, and have it linked into the
3189	* socket interface. The value ops->family corresponds to the
3190	* socket system call protocol family.
3191	*/
3192	int sock_register(const struct net_proto_family *ops)
3193	{
3194	int err;
3195
3196	if (ops->family >= NPROTO) {
3197	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3198	return -ENOBUFS;
3199	}
3200
3201	spin_lock(&net_family_lock);
3202	if (rcu_dereference_protected(net_families[ops->family],
3203	lockdep_is_held(&net_family_lock)))
3204	err = -EEXIST;
3205	else {
3206	rcu_assign_pointer(net_families[ops->family], ops);
3207	err = 0;
3208	}
3209	spin_unlock(&net_family_lock);
3210
3211	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3212	return err;
3213	}
3214	EXPORT_SYMBOL(sock_register);
3215
3216	/**
3217	* sock_unregister - remove a protocol handler
3218	* @family: protocol family to remove
3219	*
3220	* This function is called by a protocol handler that wants to
3221	* remove its address family, and have it unlinked from the
3222	* new socket creation.
3223	*
3224	* If protocol handler is a module, then it can use module reference
3225	* counts to protect against new references. If protocol handler is not
3226	* a module then it needs to provide its own protection in
3227	* the ops->create routine.
3228	*/
3229	void sock_unregister(int family)
3230	{
3231	BUG_ON(family < 0 \|\| family >= NPROTO);
3232
3233	spin_lock(&net_family_lock);
3234	RCU_INIT_POINTER(net_families[family], NULL);
3235	spin_unlock(&net_family_lock);
3236
3237	synchronize_rcu();
3238
3239	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3240	}
3241	EXPORT_SYMBOL(sock_unregister);
3242
3243	bool sock_is_registered(int family)
3244	{
3245	return family < NPROTO && rcu_access_pointer(net_families[family]);
3246	}
3247
3248	static int __init sock_init(void)
3249	{
3250	int err;
3251	/*
3252	* Initialize the network sysctl infrastructure.
3253	*/
3254	err = net_sysctl_init();
3255	if (err)
3256	goto out;
3257
3258	/*
3259	* Initialize skbuff SLAB cache
3260	*/
3261	skb_init();
3262
3263	/*
3264	* Initialize the protocols module.
3265	*/
3266
3267	init_inodecache();
3268
3269	err = register_filesystem(&sock_fs_type);
3270	if (err)
3271	goto out;
3272	sock_mnt = kern_mount(&sock_fs_type);
3273	if (IS_ERR(sock_mnt)) {
3274	err = PTR_ERR(sock_mnt);
3275	goto out_mount;
3276	}
3277
3278	/* The real protocol initialization is performed in later initcalls.
3279	*/
3280
3281	#ifdef CONFIG_NETFILTER
3282	err = netfilter_init();
3283	if (err)
3284	goto out;
3285	#endif
3286
3287	ptp_classifier_init();
3288
3289	out:
3290	return err;
3291
3292	out_mount:
3293	unregister_filesystem(&sock_fs_type);
3294	goto out;
3295	}
3296
3297	core_initcall(sock_init); /* early initcall */
3298
3299	#ifdef CONFIG_PROC_FS
3300	void socket_seq_show(struct seq_file *seq)
3301	{
3302	seq_printf(seq, "sockets: used %d\n",
3303	sock_inuse_get(seq->private));
3304	}
3305	#endif /* CONFIG_PROC_FS */
3306
3307	/* Handle the fact that while struct ifreq has the same layout on
3308	* 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3309	* which are handled elsewhere, it still has different size due to
3310	* ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3311	* resulting in struct ifreq being 32 and 40 bytes respectively).
3312	* As a result, if the struct happens to be at the end of a page and
3313	* the next page isn't readable/writable, we get a fault. To prevent
3314	* that, copy back and forth to the full size.
3315	*/
3316	int get_user_ifreq(struct ifreq ifr, void __user ifrdata, void __user arg)
3317	{
3318	if (in_compat_syscall()) {
3319	struct compat_ifreq ifr32 = (struct compat_ifreq )ifr;
3320
3321	memset(ifr, 0, sizeof(*ifr));
3322	if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3323	return -EFAULT;
3324
3325	if (ifrdata)
3326	*ifrdata = compat_ptr(ifr32->ifr_data);
3327
3328	return 0;
3329	}
3330
3331	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3332	return -EFAULT;
3333
3334	if (ifrdata)
3335	*ifrdata = ifr->ifr_data;
3336
3337	return 0;
3338	}
3339	EXPORT_SYMBOL(get_user_ifreq);
3340
3341	int put_user_ifreq(struct ifreq ifr, void __user arg)
3342	{
3343	size_t size = sizeof(*ifr);
3344
3345	if (in_compat_syscall())
3346	size = sizeof(struct compat_ifreq);
3347
3348	if (copy_to_user(arg, ifr, size))
3349	return -EFAULT;
3350
3351	return 0;
3352	}
3353	EXPORT_SYMBOL(put_user_ifreq);
3354
3355	#ifdef CONFIG_COMPAT
3356	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
3357	{
3358	compat_uptr_t uptr32;
3359	struct ifreq ifr;
3360	void __user *saved;
3361	int err;
3362
3363	if (get_user_ifreq(&ifr, NULL, uifr32))
3364	return -EFAULT;
3365
3366	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3367	return -EFAULT;
3368
3369	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3370	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3371
3372	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3373	if (!err) {
3374	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3375	if (put_user_ifreq(&ifr, uifr32))
3376	err = -EFAULT;
3377	}
3378	return err;
3379	}
3380
3381	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3382	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3383	struct compat_ifreq __user *u_ifreq32)
3384	{
3385	struct ifreq ifreq;
3386	void __user *data;
3387
3388	if (!is_socket_ioctl_cmd(cmd))
3389	return -ENOTTY;
3390	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3391	return -EFAULT;
3392	ifreq.ifr_data = data;
3393
3394	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3395	}
3396
3397	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3398	unsigned int cmd, unsigned long arg)
3399	{
3400	void __user *argp = compat_ptr(arg);
3401	struct sock *sk = sock->sk;
3402	struct net *net = sock_net(sk);
3403	const struct proto_ops *ops;
3404
3405	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3406	return sock_ioctl(file, cmd, (unsigned long)argp);
3407
3408	switch (cmd) {
3409	case SIOCWANDEV:
3410	return compat_siocwandev(net, argp);
3411	case SIOCGSTAMP_OLD:
3412	case SIOCGSTAMPNS_OLD:
3413	ops = READ_ONCE(sock->ops);
3414	if (!ops->gettstamp)
3415	return -ENOIOCTLCMD;
3416	return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3417	!COMPAT_USE_64BIT_TIME);
3418
3419	case SIOCETHTOOL:
3420	case SIOCBONDSLAVEINFOQUERY:
3421	case SIOCBONDINFOQUERY:
3422	case SIOCSHWTSTAMP:
3423	case SIOCGHWTSTAMP:
3424	return compat_ifr_data_ioctl(net, cmd, argp);
3425
3426	case FIOSETOWN:
3427	case SIOCSPGRP:
3428	case FIOGETOWN:
3429	case SIOCGPGRP:
3430	case SIOCBRADDBR:
3431	case SIOCBRDELBR:
3432	case SIOCGIFVLAN:
3433	case SIOCSIFVLAN:
3434	case SIOCGSKNS:
3435	case SIOCGSTAMP_NEW:
3436	case SIOCGSTAMPNS_NEW:
3437	case SIOCGIFCONF:
3438	case SIOCSIFBR:
3439	case SIOCGIFBR:
3440	return sock_ioctl(file, cmd, arg);
3441
3442	case SIOCGIFFLAGS:
3443	case SIOCSIFFLAGS:
3444	case SIOCGIFMAP:
3445	case SIOCSIFMAP:
3446	case SIOCGIFMETRIC:
3447	case SIOCSIFMETRIC:
3448	case SIOCGIFMTU:
3449	case SIOCSIFMTU:
3450	case SIOCGIFMEM:
3451	case SIOCSIFMEM:
3452	case SIOCGIFHWADDR:
3453	case SIOCSIFHWADDR:
3454	case SIOCADDMULTI:
3455	case SIOCDELMULTI:
3456	case SIOCGIFINDEX:
3457	case SIOCGIFADDR:
3458	case SIOCSIFADDR:
3459	case SIOCSIFHWBROADCAST:
3460	case SIOCDIFADDR:
3461	case SIOCGIFBRDADDR:
3462	case SIOCSIFBRDADDR:
3463	case SIOCGIFDSTADDR:
3464	case SIOCSIFDSTADDR:
3465	case SIOCGIFNETMASK:
3466	case SIOCSIFNETMASK:
3467	case SIOCSIFPFLAGS:
3468	case SIOCGIFPFLAGS:
3469	case SIOCGIFTXQLEN:
3470	case SIOCSIFTXQLEN:
3471	case SIOCBRADDIF:
3472	case SIOCBRDELIF:
3473	case SIOCGIFNAME:
3474	case SIOCSIFNAME:
3475	case SIOCGMIIPHY:
3476	case SIOCGMIIREG:
3477	case SIOCSMIIREG:
3478	case SIOCBONDENSLAVE:
3479	case SIOCBONDRELEASE:
3480	case SIOCBONDSETHWADDR:
3481	case SIOCBONDCHANGEACTIVE:
3482	case SIOCSARP:
3483	case SIOCGARP:
3484	case SIOCDARP:
3485	case SIOCOUTQ:
3486	case SIOCOUTQNSD:
3487	case SIOCATMARK:
3488	return sock_do_ioctl(net, sock, cmd, arg);
3489	}
3490
3491	return -ENOIOCTLCMD;
3492	}
3493
3494	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3495	unsigned long arg)
3496	{
3497	struct socket *sock = file->private_data;
3498	const struct proto_ops *ops = READ_ONCE(sock->ops);
3499	int ret = -ENOIOCTLCMD;
3500	struct sock *sk;
3501	struct net *net;
3502
3503	sk = sock->sk;
3504	net = sock_net(sk);
3505
3506	if (ops->compat_ioctl)
3507	ret = ops->compat_ioctl(sock, cmd, arg);
3508
3509	if (ret == -ENOIOCTLCMD &&
3510	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3511	ret = compat_wext_handle_ioctl(net, cmd, arg);
3512
3513	if (ret == -ENOIOCTLCMD)
3514	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3515
3516	return ret;
3517	}
3518	#endif
3519
3520	/**
3521	* kernel_bind - bind an address to a socket (kernel space)
3522	* @sock: socket
3523	* @addr: address
3524	* @addrlen: length of address
3525	*
3526	* Returns 0 or an error.
3527	*/
3528
3529	int kernel_bind(struct socket sock, struct sockaddr addr, int addrlen)
3530	{
3531	struct sockaddr_storage address;
3532
3533	memcpy(&address, addr, addrlen);
3534
3535	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3536	addrlen);
3537	}
3538	EXPORT_SYMBOL(kernel_bind);
3539
3540	/**
3541	* kernel_listen - move socket to listening state (kernel space)
3542	* @sock: socket
3543	* @backlog: pending connections queue size
3544	*
3545	* Returns 0 or an error.
3546	*/
3547
3548	int kernel_listen(struct socket *sock, int backlog)
3549	{
3550	return READ_ONCE(sock->ops)->listen(sock, backlog);
3551	}
3552	EXPORT_SYMBOL(kernel_listen);
3553
3554	/**
3555	* kernel_accept - accept a connection (kernel space)
3556	* @sock: listening socket
3557	* @newsock: new connected socket
3558	* @flags: flags
3559	*
3560	* @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3561	* If it fails, @newsock is guaranteed to be %NULL.
3562	* Returns 0 or an error.
3563	*/
3564
3565	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3566	{
3567	struct sock *sk = sock->sk;
3568	const struct proto_ops *ops = READ_ONCE(sock->ops);
3569	struct proto_accept_arg arg = {
3570	.flags = flags,
3571	.kern = true,
3572	};
3573	int err;
3574
3575	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3576	newsock);
3577	if (err < 0)
3578	goto done;
3579
3580	err = ops->accept(sock, *newsock, &arg);
3581	if (err < 0) {
3582	sock_release(*newsock);
3583	*newsock = NULL;
3584	goto done;
3585	}
3586
3587	(*newsock)->ops = ops;
3588	__module_get(ops->owner);
3589
3590	done:
3591	return err;
3592	}
3593	EXPORT_SYMBOL(kernel_accept);
3594
3595	/**
3596	* kernel_connect - connect a socket (kernel space)
3597	* @sock: socket
3598	* @addr: address
3599	* @addrlen: address length
3600	* @flags: flags (O_NONBLOCK, ...)
3601	*
3602	* For datagram sockets, @addr is the address to which datagrams are sent
3603	* by default, and the only address from which datagrams are received.
3604	* For stream sockets, attempts to connect to @addr.
3605	* Returns 0 or an error code.
3606	*/
3607
3608	int kernel_connect(struct socket sock, struct sockaddr addr, int addrlen,
3609	int flags)
3610	{
3611	struct sockaddr_storage address;
3612
3613	memcpy(&address, addr, addrlen);
3614
3615	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3616	addrlen, flags);
3617	}
3618	EXPORT_SYMBOL(kernel_connect);
3619
3620	/**
3621	* kernel_getsockname - get the address which the socket is bound (kernel space)
3622	* @sock: socket
3623	* @addr: address holder
3624	*
3625	* Fills the @addr pointer with the address which the socket is bound.
3626	* Returns the length of the address in bytes or an error code.
3627	*/
3628
3629	int kernel_getsockname(struct socket sock, struct sockaddr addr)
3630	{
3631	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3632	}
3633	EXPORT_SYMBOL(kernel_getsockname);
3634
3635	/**
3636	* kernel_getpeername - get the address which the socket is connected (kernel space)
3637	* @sock: socket
3638	* @addr: address holder
3639	*
3640	* Fills the @addr pointer with the address which the socket is connected.
3641	* Returns the length of the address in bytes or an error code.
3642	*/
3643
3644	int kernel_getpeername(struct socket sock, struct sockaddr addr)
3645	{
3646	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3647	}
3648	EXPORT_SYMBOL(kernel_getpeername);
3649
3650	/**
3651	* kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3652	* @sock: socket
3653	* @how: connection part
3654	*
3655	* Returns 0 or an error.
3656	*/
3657
3658	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3659	{
3660	return READ_ONCE(sock->ops)->shutdown(sock, how);
3661	}
3662	EXPORT_SYMBOL(kernel_sock_shutdown);
3663
3664	/**
3665	* kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3666	* @sk: socket
3667	*
3668	* This routine returns the IP overhead imposed by a socket i.e.
3669	* the length of the underlying IP header, depending on whether
3670	* this is an IPv4 or IPv6 socket and the length from IP options turned
3671	* on at the socket. Assumes that the caller has a lock on the socket.
3672	*/
3673
3674	u32 kernel_sock_ip_overhead(struct sock *sk)
3675	{
3676	struct inet_sock *inet;
3677	struct ip_options_rcu *opt;
3678	u32 overhead = 0;
3679	#if IS_ENABLED(CONFIG_IPV6)
3680	struct ipv6_pinfo *np;
3681	struct ipv6_txoptions *optv6 = NULL;
3682	#endif /* IS_ENABLED(CONFIG_IPV6) */
3683
3684	if (!sk)
3685	return overhead;
3686
3687	switch (sk->sk_family) {
3688	case AF_INET:
3689	inet = inet_sk(sk);
3690	overhead += sizeof(struct iphdr);
3691	opt = rcu_dereference_protected(inet->inet_opt,
3692	sock_owned_by_user(sk));
3693	if (opt)
3694	overhead += opt->opt.optlen;
3695	return overhead;
3696	#if IS_ENABLED(CONFIG_IPV6)
3697	case AF_INET6:
3698	np = inet6_sk(sk);
3699	overhead += sizeof(struct ipv6hdr);
3700	if (np)
3701	optv6 = rcu_dereference_protected(np->opt,
3702	sock_owned_by_user(sk));
3703	if (optv6)
3704	overhead += (optv6->opt_flen + optv6->opt_nflen);
3705	return overhead;
3706	#endif /* IS_ENABLED(CONFIG_IPV6) */
3707	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3708	return overhead;
3709	}
3710	}
3711	EXPORT_SYMBOL(kernel_sock_ip_overhead);