linux - net/socket.c

3822 lines · 93.6 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/uio.h>
81	#include <linux/syscalls.h>
82	#include <linux/compat.h>
83	#include <linux/kmod.h>
84	#include <linux/audit.h>
85	#include <linux/wireless.h>
86	#include <linux/nsproxy.h>
87	#include <linux/magic.h>
88	#include <linux/slab.h>
89	#include <linux/xattr.h>
90	#include <linux/nospec.h>
91	#include <linux/indirect_call_wrapper.h>
92	#include <linux/io_uring/net.h>
93
94	#include <linux/uaccess.h>
95	#include <asm/unistd.h>
96
97	#include <net/compat.h>
98	#include <net/wext.h>
99	#include <net/cls_cgroup.h>
100
101	#include <net/sock.h>
102	#include <linux/netfilter.h>
103
104	#include <linux/if_tun.h>
105	#include <linux/ipv6_route.h>
106	#include <linux/route.h>
107	#include <linux/termios.h>
108	#include <linux/sockios.h>
109	#include <net/busy_poll.h>
110	#include <linux/errqueue.h>
111	#include <linux/ptp_clock_kernel.h>
112	#include <trace/events/sock.h>
113
114	#include "core/dev.h"
115
116	#ifdef CONFIG_NET_RX_BUSY_POLL
117	unsigned int sysctl_net_busy_read __read_mostly;
118	unsigned int sysctl_net_busy_poll __read_mostly;
119	#endif
120
121	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to);
122	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from);
123	static int sock_mmap(struct file file, struct vm_area_struct vma);
124
125	static int sock_close(struct inode inode, struct file file);
126	static __poll_t sock_poll(struct file *file,
127	struct poll_table_struct *wait);
128	static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
129	#ifdef CONFIG_COMPAT
130	static long compat_sock_ioctl(struct file *file,
131	unsigned int cmd, unsigned long arg);
132	#endif
133	static int sock_fasync(int fd, struct file *filp, int on);
134	static ssize_t sock_splice_read(struct file file, loff_t ppos,
135	struct pipe_inode_info *pipe, size_t len,
136	unsigned int flags);
137	static void sock_splice_eof(struct file *file);
138
139	#ifdef CONFIG_PROC_FS
140	static void sock_show_fdinfo(struct seq_file m, struct file f)
141	{
142	struct socket *sock = f->private_data;
143	const struct proto_ops *ops = READ_ONCE(sock->ops);
144
145	if (ops->show_fdinfo)
146	ops->show_fdinfo(m, sock);
147	}
148	#else
149	#define sock_show_fdinfo NULL
150	#endif
151
152	/*
153	* Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
154	* in the operation structures but are done directly via the socketcall() multiplexor.
155	*/
156
157	static const struct file_operations socket_file_ops = {
158	.owner = THIS_MODULE,
159	.read_iter = sock_read_iter,
160	.write_iter = sock_write_iter,
161	.poll = sock_poll,
162	.unlocked_ioctl = sock_ioctl,
163	#ifdef CONFIG_COMPAT
164	.compat_ioctl = compat_sock_ioctl,
165	#endif
166	.uring_cmd = io_uring_cmd_sock,
167	.mmap = sock_mmap,
168	.release = sock_close,
169	.fasync = sock_fasync,
170	.splice_write = splice_to_socket,
171	.splice_read = sock_splice_read,
172	.splice_eof = sock_splice_eof,
173	.show_fdinfo = sock_show_fdinfo,
174	};
175
176	static const char * const pf_family_names[] = {
177	[PF_UNSPEC] = "PF_UNSPEC",
178	[PF_UNIX] = "PF_UNIX/PF_LOCAL",
179	[PF_INET] = "PF_INET",
180	[PF_AX25] = "PF_AX25",
181	[PF_IPX] = "PF_IPX",
182	[PF_APPLETALK] = "PF_APPLETALK",
183	[PF_NETROM] = "PF_NETROM",
184	[PF_BRIDGE] = "PF_BRIDGE",
185	[PF_ATMPVC] = "PF_ATMPVC",
186	[PF_X25] = "PF_X25",
187	[PF_INET6] = "PF_INET6",
188	[PF_ROSE] = "PF_ROSE",
189	[PF_DECnet] = "PF_DECnet",
190	[PF_NETBEUI] = "PF_NETBEUI",
191	[PF_SECURITY] = "PF_SECURITY",
192	[PF_KEY] = "PF_KEY",
193	[PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
194	[PF_PACKET] = "PF_PACKET",
195	[PF_ASH] = "PF_ASH",
196	[PF_ECONET] = "PF_ECONET",
197	[PF_ATMSVC] = "PF_ATMSVC",
198	[PF_RDS] = "PF_RDS",
199	[PF_SNA] = "PF_SNA",
200	[PF_IRDA] = "PF_IRDA",
201	[PF_PPPOX] = "PF_PPPOX",
202	[PF_WANPIPE] = "PF_WANPIPE",
203	[PF_LLC] = "PF_LLC",
204	[PF_IB] = "PF_IB",
205	[PF_MPLS] = "PF_MPLS",
206	[PF_CAN] = "PF_CAN",
207	[PF_TIPC] = "PF_TIPC",
208	[PF_BLUETOOTH] = "PF_BLUETOOTH",
209	[PF_IUCV] = "PF_IUCV",
210	[PF_RXRPC] = "PF_RXRPC",
211	[PF_ISDN] = "PF_ISDN",
212	[PF_PHONET] = "PF_PHONET",
213	[PF_IEEE802154] = "PF_IEEE802154",
214	[PF_CAIF] = "PF_CAIF",
215	[PF_ALG] = "PF_ALG",
216	[PF_NFC] = "PF_NFC",
217	[PF_VSOCK] = "PF_VSOCK",
218	[PF_KCM] = "PF_KCM",
219	[PF_QIPCRTR] = "PF_QIPCRTR",
220	[PF_SMC] = "PF_SMC",
221	[PF_XDP] = "PF_XDP",
222	[PF_MCTP] = "PF_MCTP",
223	};
224
225	/*
226	* The protocol list. Each protocol is registered in here.
227	*/
228
229	static DEFINE_SPINLOCK(net_family_lock);
230	static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
231
232	/*
233	* Support routines.
234	* Move socket addresses back and forth across the kernel/user
235	* divide and look after the messy bits.
236	*/
237
238	/**
239	* move_addr_to_kernel - copy a socket address into kernel space
240	* @uaddr: Address in user space
241	* @kaddr: Address in kernel space
242	* @ulen: Length in user space
243	*
244	* The address is copied into kernel space. If the provided address is
245	* too long an error code of -EINVAL is returned. If the copy gives
246	* invalid addresses -EFAULT is returned. On a success 0 is returned.
247	*/
248
249	int move_addr_to_kernel(void __user uaddr, int ulen, struct sockaddr_storage kaddr)
250	{
251	if (ulen < 0 \|\| ulen > sizeof(struct sockaddr_storage))
252	return -EINVAL;
253	if (ulen == 0)
254	return 0;
255	if (copy_from_user(kaddr, uaddr, ulen))
256	return -EFAULT;
257	return audit_sockaddr(ulen, kaddr);
258	}
259
260	/**
261	* move_addr_to_user - copy an address to user space
262	* @kaddr: kernel space address
263	* @klen: length of address in kernel
264	* @uaddr: user space address
265	* @ulen: pointer to user length field
266	*
267	* The value pointed to by ulen on entry is the buffer length available.
268	* This is overwritten with the buffer space used. -EINVAL is returned
269	* if an overlong buffer is specified or a negative buffer size. -EFAULT
270	* is returned if either the buffer or the length field are not
271	* accessible.
272	* After copying the data up to the limit the user specifies, the true
273	* length of the data is written over the length limit the user
274	* specified. Zero is returned for a success.
275	*/
276
277	static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
278	void __user uaddr, int __user ulen)
279	{
280	int len;
281
282	BUG_ON(klen > sizeof(struct sockaddr_storage));
283
284	scoped_user_rw_access_size(ulen, 4, efault_end) {
285	unsafe_get_user(len, ulen, efault_end);
286
287	if (len > klen)
288	len = klen;
289	/*
290	* "fromlen shall refer to the value before truncation.."
291	* 1003.1g
292	*/
293	if (len >= 0)
294	unsafe_put_user(klen, ulen, efault_end);
295	}
296
297	if (len) {
298	if (len < 0)
299	return -EINVAL;
300	if (audit_sockaddr(klen, kaddr))
301	return -ENOMEM;
302	if (copy_to_user(uaddr, kaddr, len))
303	return -EFAULT;
304	}
305	return 0;
306
307	efault_end:
308	return -EFAULT;
309	}
310
311	static struct kmem_cache *sock_inode_cachep __ro_after_init;
312
313	static struct simple_xattr_cache sockfs_xa_cache;
314
315	struct sockfs_inode {
316	struct list_head xattrs;
317	struct simple_xattr_limits xattr_limits;
318	struct socket_alloc;
319	};
320
321	static struct sockfs_inode SOCKFS_I(struct inode inode)
322	{
323	return container_of(inode, struct sockfs_inode, vfs_inode);
324	}
325
326	static struct inode sock_alloc_inode(struct super_block sb)
327	{
328	struct sockfs_inode *si;
329
330	si = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
331	if (!si)
332	return NULL;
333	INIT_LIST_HEAD_RCU(&si->xattrs);
334	simple_xattr_limits_init(&si->xattr_limits);
335
336	init_waitqueue_head(&si->socket.wq.wait);
337	si->socket.wq.fasync_list = NULL;
338	si->socket.wq.flags = 0;
339
340	si->socket.state = SS_UNCONNECTED;
341	si->socket.flags = 0;
342	si->socket.ops = NULL;
343	si->socket.sk = NULL;
344	si->socket.file = NULL;
345
346	return &si->vfs_inode;
347	}
348
349	static void sock_evict_inode(struct inode *inode)
350	{
351	struct sockfs_inode *si = SOCKFS_I(inode);
352
353	simple_xattrs_free(&sockfs_xa_cache, &si->xattrs, NULL);
354	clear_inode(inode);
355	}
356
357	static void sock_free_inode(struct inode *inode)
358	{
359	struct sockfs_inode *si = SOCKFS_I(inode);
360
361	kmem_cache_free(sock_inode_cachep, si);
362	}
363
364	static void init_once(void *foo)
365	{
366	struct sockfs_inode si = (struct sockfs_inode )foo;
367
368	inode_init_once(&si->vfs_inode);
369	}
370
371	static void init_inodecache(void)
372	{
373	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
374	sizeof(struct sockfs_inode),
375	0,
376	(SLAB_HWCACHE_ALIGN \|
377	SLAB_RECLAIM_ACCOUNT \|
378	SLAB_ACCOUNT),
379	init_once);
380	BUG_ON(sock_inode_cachep == NULL);
381	}
382
383	static const struct super_operations sockfs_ops = {
384	.alloc_inode = sock_alloc_inode,
385	.free_inode = sock_free_inode,
386	.evict_inode = sock_evict_inode,
387	.statfs = simple_statfs,
388	};
389
390	/*
391	* sockfs_dname() is called from d_path().
392	*/
393	static char sockfs_dname(struct dentry dentry, char *buffer, int buflen)
394	{
395	return dynamic_dname(buffer, buflen, "socket:[%llu]",
396	d_inode(dentry)->i_ino);
397	}
398
399	static const struct dentry_operations sockfs_dentry_operations = {
400	.d_dname = sockfs_dname,
401	};
402
403	static int sockfs_xattr_get(const struct xattr_handler *handler,
404	struct dentry dentry, struct inode inode,
405	const char suffix, void value, size_t size)
406	{
407	if (value) {
408	if (dentry->d_name.len + 1 > size)
409	return -ERANGE;
410	memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
411	}
412	return dentry->d_name.len + 1;
413	}
414
415	#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
416	#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
417	#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
418
419	static const struct xattr_handler sockfs_xattr_handler = {
420	.name = XATTR_NAME_SOCKPROTONAME,
421	.get = sockfs_xattr_get,
422	};
423
424	static int sockfs_security_xattr_set(const struct xattr_handler *handler,
425	struct mnt_idmap *idmap,
426	struct dentry dentry, struct inode inode,
427	const char suffix, const void value,
428	size_t size, int flags)
429	{
430	/* Handled by LSM. */
431	return -EAGAIN;
432	}
433
434	static const struct xattr_handler sockfs_security_xattr_handler = {
435	.prefix = XATTR_SECURITY_PREFIX,
436	.set = sockfs_security_xattr_set,
437	};
438
439	static int sockfs_user_xattr_get(const struct xattr_handler *handler,
440	struct dentry dentry, struct inode inode,
441	const char suffix, void value, size_t size)
442	{
443	const char *name = xattr_full_name(handler, suffix);
444	struct sockfs_inode *si = SOCKFS_I(inode);
445
446	return simple_xattr_get(&sockfs_xa_cache, &si->xattrs, name, value, size);
447	}
448
449	static int sockfs_user_xattr_set(const struct xattr_handler *handler,
450	struct mnt_idmap *idmap,
451	struct dentry dentry, struct inode inode,
452	const char suffix, const void value,
453	size_t size, int flags)
454	{
455	const char *name = xattr_full_name(handler, suffix);
456	struct sockfs_inode *si = SOCKFS_I(inode);
457
458	return simple_xattr_set_limited(&sockfs_xa_cache, &si->xattrs, &si->xattr_limits,
459	name, value, size, flags);
460	}
461
462	static const struct xattr_handler sockfs_user_xattr_handler = {
463	.prefix = XATTR_USER_PREFIX,
464	.get = sockfs_user_xattr_get,
465	.set = sockfs_user_xattr_set,
466	};
467
468	static const struct xattr_handler * const sockfs_xattr_handlers[] = {
469	&sockfs_xattr_handler,
470	&sockfs_security_xattr_handler,
471	&sockfs_user_xattr_handler,
472	NULL
473	};
474
475	static int sockfs_init_fs_context(struct fs_context *fc)
476	{
477	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
478	if (!ctx)
479	return -ENOMEM;
480	ctx->ops = &sockfs_ops;
481	ctx->dops = &sockfs_dentry_operations;
482	ctx->xattr = sockfs_xattr_handlers;
483	return 0;
484	}
485
486	static struct vfsmount *sock_mnt __read_mostly;
487
488	static struct file_system_type sock_fs_type = {
489	.name = "sockfs",
490	.init_fs_context = sockfs_init_fs_context,
491	.kill_sb = kill_anon_super,
492	};
493
494	/*
495	* Obtains the first available file descriptor and sets it up for use.
496	*
497	* These functions create file structures and maps them to fd space
498	* of the current process. On success it returns file descriptor
499	* and file struct implicitly stored in sock->file.
500	* Note that another thread may close file descriptor before we return
501	* from this function. We use the fact that now we do not refer
502	* to socket after mapping. If one day we will need it, this
503	* function will increment ref. count on file by 1.
504	*
505	* In any case returned fd MAY BE not valid!
506	* This race condition is unavoidable
507	* with shared fd spaces, we cannot solve it inside kernel,
508	* but we take care of internal coherence yet.
509	*/
510
511	/**
512	* sock_alloc_file - Bind a &socket to a &file
513	* @sock: socket
514	* @flags: file status flags
515	* @dname: protocol name
516	*
517	* Returns the &file bound with @sock, implicitly storing it
518	* in sock->file. If dname is %NULL, sets to "".
519	*
520	* On failure @sock is released, and an ERR pointer is returned.
521	*
522	* This function uses GFP_KERNEL internally.
523	*/
524
525	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
526	{
527	struct file *file;
528
529	if (!dname)
530	dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
531
532	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
533	O_RDWR \| (flags & O_NONBLOCK),
534	&socket_file_ops);
535	if (IS_ERR(file)) {
536	sock_release(sock);
537	return file;
538	}
539
540	file->f_mode \|= FMODE_NOWAIT;
541	sock->file = file;
542	file->private_data = sock;
543	stream_open(SOCK_INODE(sock), file);
544	/*
545	* Disable permission and pre-content events, but enable legacy
546	* inotify events for legacy users.
547	*/
548	file_set_fsnotify_mode(file, FMODE_NONOTIFY_PERM);
549	return file;
550	}
551	EXPORT_SYMBOL(sock_alloc_file);
552
553	static int sock_map_fd(struct socket *sock, int flags)
554	{
555	struct file *newfile;
556	int fd = get_unused_fd_flags(flags);
557	if (unlikely(fd < 0)) {
558	sock_release(sock);
559	return fd;
560	}
561
562	newfile = sock_alloc_file(sock, flags, NULL);
563	if (!IS_ERR(newfile)) {
564	fd_install(fd, newfile);
565	return fd;
566	}
567
568	put_unused_fd(fd);
569	return PTR_ERR(newfile);
570	}
571
572	/**
573	* sock_from_file - Return the &socket bounded to @file.
574	* @file: file
575	*
576	* On failure returns %NULL.
577	*/
578
579	struct socket sock_from_file(struct file file)
580	{
581	if (likely(file->f_op == &socket_file_ops))
582	return file->private_data; /* set in sock_alloc_file */
583
584	return NULL;
585	}
586	EXPORT_SYMBOL(sock_from_file);
587
588	/**
589	* sockfd_lookup - Go from a file number to its socket slot
590	* @fd: file handle
591	* @err: pointer to an error code return
592	*
593	* The file handle passed in is locked and the socket it is bound
594	* to is returned. If an error occurs the err pointer is overwritten
595	* with a negative errno code and NULL is returned. The function checks
596	* for both invalid handles and passing a handle which is not a socket.
597	*
598	* On a success the socket object pointer is returned.
599	*/
600
601	struct socket sockfd_lookup(int fd, int err)
602	{
603	struct file *file;
604	struct socket *sock;
605
606	file = fget(fd);
607	if (!file) {
608	*err = -EBADF;
609	return NULL;
610	}
611
612	sock = sock_from_file(file);
613	if (!sock) {
614	*err = -ENOTSOCK;
615	fput(file);
616	}
617	return sock;
618	}
619	EXPORT_SYMBOL(sockfd_lookup);
620
621	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
622	size_t size)
623	{
624	struct sockfs_inode *si = SOCKFS_I(d_inode(dentry));
625	ssize_t len, used;
626
627	len = simple_xattr_list(d_inode(dentry), &si->xattrs, buffer, size);
628	if (len < 0)
629	return len;
630
631	used = len;
632	if (buffer) {
633	buffer += len;
634	size -= len;
635	}
636
637	len = XATTR_NAME_SOCKPROTONAME_LEN + 1;
638	used += len;
639	if (buffer) {
640	if (size < len)
641	return -ERANGE;
642	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
643	}
644
645	return used;
646	}
647
648	static int sockfs_setattr(struct mnt_idmap *idmap,
649	struct dentry dentry, struct iattr iattr)
650	{
651	int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
652
653	if (!err && (iattr->ia_valid & ATTR_UID)) {
654	struct socket *sock = SOCKET_I(d_inode(dentry));
655
656	if (sock->sk) {
657	/* Paired with READ_ONCE() in sk_uid() */
658	WRITE_ONCE(sock->sk->sk_uid, iattr->ia_uid);
659	} else {
660	err = -ENOENT;
661	}
662	}
663
664	return err;
665	}
666
667	static const struct inode_operations sockfs_inode_ops = {
668	.listxattr = sockfs_listxattr,
669	.setattr = sockfs_setattr,
670	};
671
672	/**
673	* sock_alloc - allocate a socket
674	*
675	* Allocate a new inode and socket object. The two are bound together
676	* and initialised. The socket is then returned. If we are out of inodes
677	* NULL is returned. This functions uses GFP_KERNEL internally.
678	*/
679
680	struct socket *sock_alloc(void)
681	{
682	struct inode *inode;
683	struct socket *sock;
684
685	inode = new_inode_pseudo(sock_mnt->mnt_sb);
686	if (!inode)
687	return NULL;
688
689	sock = SOCKET_I(inode);
690
691	inode->i_ino = get_next_ino();
692	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
693	inode->i_uid = current_fsuid();
694	inode->i_gid = current_fsgid();
695	inode->i_op = &sockfs_inode_ops;
696
697	return sock;
698	}
699	EXPORT_SYMBOL(sock_alloc);
700
701	static void __sock_release(struct socket sock, struct inode inode)
702	{
703	const struct proto_ops *ops = READ_ONCE(sock->ops);
704
705	if (ops) {
706	struct module *owner = ops->owner;
707
708	if (inode)
709	inode_lock(inode);
710	ops->release(sock);
711	sock->sk = NULL;
712	if (inode)
713	inode_unlock(inode);
714	sock->ops = NULL;
715	module_put(owner);
716	}
717
718	if (sock->wq.fasync_list)
719	pr_err("%s: fasync list not empty!\n", __func__);
720
721	if (!sock->file) {
722	iput(SOCK_INODE(sock));
723	return;
724	}
725	WRITE_ONCE(sock->file, NULL);
726	}
727
728	/**
729	* sock_release - close a socket
730	* @sock: socket to close
731	*
732	* The socket is released from the protocol stack if it has a release
733	* callback, and the inode is then released if the socket is bound to
734	* an inode not a file.
735	*/
736	void sock_release(struct socket *sock)
737	{
738	__sock_release(sock, NULL);
739	}
740	EXPORT_SYMBOL(sock_release);
741
742	void __sock_tx_timestamp(__u32 tsflags, __u8 *tx_flags)
743	{
744	u8 flags = *tx_flags;
745
746	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
747	flags \|= SKBTX_HW_TSTAMP_NOBPF;
748
749	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
750	flags \|= SKBTX_SW_TSTAMP;
751
752	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
753	flags \|= SKBTX_SCHED_TSTAMP;
754
755	if (tsflags & SOF_TIMESTAMPING_TX_COMPLETION)
756	flags \|= SKBTX_COMPLETION_TSTAMP;
757
758	*tx_flags = flags;
759	}
760	EXPORT_SYMBOL(__sock_tx_timestamp);
761
762	INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket , struct msghdr ,
763	size_t));
764	INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket , struct msghdr ,
765	size_t));
766
767	static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
768	int flags)
769	{
770	trace_sock_send_length(sk, ret, 0);
771	}
772
773	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
774	{
775	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
776	inet_sendmsg, sock, msg,
777	msg_data_left(msg));
778	BUG_ON(ret == -EIOCBQUEUED);
779
780	if (trace_sock_send_length_enabled())
781	call_trace_sock_send_length(sock->sk, ret, 0);
782	return ret;
783	}
784
785	static int __sock_sendmsg(struct socket sock, struct msghdr msg)
786	{
787	int err = security_socket_sendmsg(sock, msg,
788	msg_data_left(msg));
789
790	return err ?: sock_sendmsg_nosec(sock, msg);
791	}
792
793	/**
794	* sock_sendmsg - send a message through @sock
795	* @sock: socket
796	* @msg: message to send
797	*
798	* Sends @msg through @sock, passing through LSM.
799	* Returns the number of bytes sent, or an error code.
800	*/
801	int sock_sendmsg(struct socket sock, struct msghdr msg)
802	{
803	struct sockaddr_storage save_addr = (struct sockaddr_storage )msg->msg_name;
804	struct sockaddr_storage address;
805	int save_len = msg->msg_namelen;
806	int ret;
807
808	if (msg->msg_name) {
809	memcpy(&address, msg->msg_name, msg->msg_namelen);
810	msg->msg_name = &address;
811	}
812
813	ret = __sock_sendmsg(sock, msg);
814	msg->msg_name = save_addr;
815	msg->msg_namelen = save_len;
816
817	return ret;
818	}
819	EXPORT_SYMBOL(sock_sendmsg);
820
821	/**
822	* kernel_sendmsg - send a message through @sock (kernel-space)
823	* @sock: socket
824	* @msg: message header
825	* @vec: kernel vec
826	* @num: vec array length
827	* @size: total message data size
828	*
829	* Builds the message data with @vec and sends it through @sock.
830	* Returns the number of bytes sent, or an error code.
831	*/
832
833	int kernel_sendmsg(struct socket sock, struct msghdr msg,
834	struct kvec *vec, size_t num, size_t size)
835	{
836	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
837	return sock_sendmsg(sock, msg);
838	}
839	EXPORT_SYMBOL(kernel_sendmsg);
840
841	static bool skb_is_err_queue(const struct sk_buff *skb)
842	{
843	/* Error-queue skbs are marked as PACKET_OUTGOING in
844	* skb_set_err_queue() and use the destructor installed by
845	* sock_queue_err_skb(). PACKET_OUTGOING alone is not unique:
846	* AF_PACKET outgoing taps use the same pkt_type.
847	*/
848	return skb->pkt_type == PACKET_OUTGOING &&
849	skb->destructor == sock_rmem_free;
850	}
851
852	/* On transmit, software and hardware timestamps are returned independently.
853	* As the two skb clones share the hardware timestamp, which may be updated
854	* before the software timestamp is received, a hardware TX timestamp may be
855	* returned only if there is no software TX timestamp. Ignore false software
856	* timestamps, which may be made in the __sock_recv_timestamp() call when the
857	* option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
858	* hardware timestamp.
859	*/
860	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
861	{
862	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
863	}
864
865	static ktime_t get_timestamp(struct sock sk, struct sk_buff skb, int *if_index)
866	{
867	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
868	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
869	struct net_device *orig_dev;
870	ktime_t hwtstamp;
871
872	rcu_read_lock();
873	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
874	if (orig_dev) {
875	*if_index = orig_dev->ifindex;
876	hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
877	} else {
878	hwtstamp = shhwtstamps->hwtstamp;
879	}
880	rcu_read_unlock();
881
882	return hwtstamp;
883	}
884
885	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb,
886	int if_index)
887	{
888	struct scm_ts_pktinfo ts_pktinfo;
889	struct net_device *orig_dev;
890
891	if (!skb_mac_header_was_set(skb))
892	return;
893
894	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
895
896	if (!if_index) {
897	rcu_read_lock();
898	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
899	if (orig_dev)
900	if_index = orig_dev->ifindex;
901	rcu_read_unlock();
902	}
903	ts_pktinfo.if_index = if_index;
904
905	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
906	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
907	sizeof(ts_pktinfo), &ts_pktinfo);
908	}
909
910	bool skb_has_tx_timestamp(struct sk_buff skb, const struct sock sk)
911	{
912	const struct sock_exterr_skb *serr = SKB_EXT_ERR(skb);
913	u32 tsflags = READ_ONCE(sk->sk_tsflags);
914
915	if (serr->ee.ee_errno != ENOMSG \|\|
916	serr->ee.ee_origin != SO_EE_ORIGIN_TIMESTAMPING)
917	return false;
918
919	/* software time stamp available and wanted */
920	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) && skb->tstamp)
921	return true;
922	/* hardware time stamps available and wanted */
923	return (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
924	skb_hwtstamps(skb)->hwtstamp;
925	}
926
927	int skb_get_tx_timestamp(struct sk_buff skb, struct sock sk,
928	struct timespec64 *ts)
929	{
930	u32 tsflags = READ_ONCE(sk->sk_tsflags);
931	ktime_t hwtstamp;
932	int if_index = 0;
933
934	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
935	ktime_to_timespec64_cond(skb->tstamp, ts))
936	return SOF_TIMESTAMPING_TX_SOFTWARE;
937
938	if (!(tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) \|\|
939	skb_is_swtx_tstamp(skb, false))
940	return -ENOENT;
941
942	if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
943	hwtstamp = get_timestamp(sk, skb, &if_index);
944	else
945	hwtstamp = skb_hwtstamps(skb)->hwtstamp;
946
947	if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
948	hwtstamp = ptp_convert_timestamp(&hwtstamp,
949	READ_ONCE(sk->sk_bind_phc));
950	if (!ktime_to_timespec64_cond(hwtstamp, ts))
951	return -ENOENT;
952
953	return SOF_TIMESTAMPING_TX_HARDWARE;
954	}
955
956	/*
957	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
958	*/
959	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
960	struct sk_buff *skb)
961	{
962	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
963	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
964	struct skb_shared_hwtstamps *shhwtstamps =
965	skb_hwtstamps(skb);
966	struct scm_timestamping_internal tss;
967	int if_index, false_tstamp = 0;
968	ktime_t hwtstamp;
969	u32 tsflags;
970
971	/* Race occurred between timestamp enabling and packet
972	receiving. Fill in the current time for now. */
973	if (need_software_tstamp && skb->tstamp == 0) {
974	__net_timestamp(skb);
975	false_tstamp = 1;
976	}
977
978	if (need_software_tstamp) {
979	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
980	if (new_tstamp) {
981	struct __kernel_sock_timeval tv;
982
983	skb_get_new_timestamp(skb, &tv);
984	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
985	sizeof(tv), &tv);
986	} else {
987	struct __kernel_old_timeval tv;
988
989	skb_get_timestamp(skb, &tv);
990	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
991	sizeof(tv), &tv);
992	}
993	} else {
994	if (new_tstamp) {
995	struct __kernel_timespec ts;
996
997	skb_get_new_timestampns(skb, &ts);
998	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
999	sizeof(ts), &ts);
1000	} else {
1001	struct __kernel_old_timespec ts;
1002
1003	skb_get_timestampns(skb, &ts);
1004	put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
1005	sizeof(ts), &ts);
1006	}
1007	}
1008	}
1009
1010	memset(&tss, 0, sizeof(tss));
1011	tsflags = READ_ONCE(sk->sk_tsflags);
1012	if (tsflags & SOF_TIMESTAMPING_SOFTWARE &&
1013	(tsflags & SOF_TIMESTAMPING_RX_SOFTWARE \|\|
1014	skb_is_err_queue(skb) \|\|
1015	!(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER)))
1016	tss.ts[0] = skb->tstamp;
1017
1018	if (shhwtstamps &&
1019	(tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
1020	(tsflags & SOF_TIMESTAMPING_RX_HARDWARE \|\|
1021	skb_is_err_queue(skb) \|\|
1022	!(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
1023	!skb_is_swtx_tstamp(skb, false_tstamp)) {
1024	if_index = 0;
1025	if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
1026	hwtstamp = get_timestamp(sk, skb, &if_index);
1027	else
1028	hwtstamp = shhwtstamps->hwtstamp;
1029
1030	if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
1031	hwtstamp = ptp_convert_timestamp(&hwtstamp,
1032	READ_ONCE(sk->sk_bind_phc));
1033
1034	if (hwtstamp) {
1035	tss.ts[2] = hwtstamp;
1036
1037	if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
1038	!skb_is_err_queue(skb))
1039	put_ts_pktinfo(msg, skb, if_index);
1040	}
1041	}
1042	if (tss.ts[0] \| tss.ts[2]) {
1043	if (sock_flag(sk, SOCK_TSTAMP_NEW))
1044	put_cmsg_scm_timestamping64(msg, &tss);
1045	else
1046	put_cmsg_scm_timestamping(msg, &tss);
1047
1048	if (skb_is_err_queue(skb) && skb->len &&
1049	SKB_EXT_ERR(skb)->opt_stats)
1050	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
1051	skb->len, skb->data);
1052	}
1053	}
1054	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
1055
1056	#ifdef CONFIG_WIRELESS
1057	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
1058	struct sk_buff *skb)
1059	{
1060	int ack;
1061
1062	if (!sock_flag(sk, SOCK_WIFI_STATUS))
1063	return;
1064	if (!skb->wifi_acked_valid)
1065	return;
1066
1067	ack = skb->wifi_acked;
1068
1069	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1070	}
1071	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1072	#endif
1073
1074	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
1075	struct sk_buff *skb)
1076	{
1077	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1078	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1079	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1080	}
1081
1082	static void sock_recv_mark(struct msghdr msg, struct sock sk,
1083	struct sk_buff *skb)
1084	{
1085	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1086	/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1087	__u32 mark = skb->mark;
1088
1089	put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1090	}
1091	}
1092
1093	static void sock_recv_priority(struct msghdr msg, struct sock sk,
1094	struct sk_buff *skb)
1095	{
1096	if (sock_flag(sk, SOCK_RCVPRIORITY) && skb) {
1097	__u32 priority = skb->priority;
1098
1099	put_cmsg(msg, SOL_SOCKET, SO_PRIORITY, sizeof(__u32), &priority);
1100	}
1101	}
1102
1103	void __sock_recv_cmsgs(struct msghdr msg, struct sock sk,
1104	struct sk_buff *skb)
1105	{
1106	sock_recv_timestamp(msg, sk, skb);
1107	sock_recv_drops(msg, sk, skb);
1108	sock_recv_mark(msg, sk, skb);
1109	sock_recv_priority(msg, sk, skb);
1110	}
1111	EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1112
1113	INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket , struct msghdr ,
1114	size_t, int));
1115	INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket , struct msghdr ,
1116	size_t, int));
1117
1118	static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1119	{
1120	trace_sock_recv_length(sk, ret, flags);
1121	}
1122
1123	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
1124	int flags)
1125	{
1126	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1127	inet6_recvmsg,
1128	inet_recvmsg, sock, msg,
1129	msg_data_left(msg), flags);
1130	if (trace_sock_recv_length_enabled())
1131	call_trace_sock_recv_length(sock->sk, ret, flags);
1132	return ret;
1133	}
1134
1135	/**
1136	* sock_recvmsg - receive a message from @sock
1137	* @sock: socket
1138	* @msg: message to receive
1139	* @flags: message flags
1140	*
1141	* Receives @msg from @sock, passing through LSM. Returns the total number
1142	* of bytes received, or an error.
1143	*/
1144	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
1145	{
1146	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1147
1148	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1149	}
1150	EXPORT_SYMBOL(sock_recvmsg);
1151
1152	/**
1153	* kernel_recvmsg - Receive a message from a socket (kernel space)
1154	* @sock: The socket to receive the message from
1155	* @msg: Received message
1156	* @vec: Input s/g array for message data
1157	* @num: Size of input s/g array
1158	* @size: Number of bytes to read
1159	* @flags: Message flags (MSG_DONTWAIT, etc...)
1160	*
1161	* On return the msg structure contains the scatter/gather array passed in the
1162	* vec argument. The array is modified so that it consists of the unfilled
1163	* portion of the original array.
1164	*
1165	* The returned value is the total number of bytes received, or an error.
1166	*/
1167
1168	int kernel_recvmsg(struct socket sock, struct msghdr msg,
1169	struct kvec *vec, size_t num, size_t size, int flags)
1170	{
1171	msg->msg_control_is_user = false;
1172	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1173	return sock_recvmsg(sock, msg, flags);
1174	}
1175	EXPORT_SYMBOL(kernel_recvmsg);
1176
1177	static ssize_t sock_splice_read(struct file file, loff_t ppos,
1178	struct pipe_inode_info *pipe, size_t len,
1179	unsigned int flags)
1180	{
1181	struct socket *sock = file->private_data;
1182	const struct proto_ops *ops;
1183
1184	ops = READ_ONCE(sock->ops);
1185	if (unlikely(!ops->splice_read))
1186	return copy_splice_read(file, ppos, pipe, len, flags);
1187
1188	return ops->splice_read(sock, ppos, pipe, len, flags);
1189	}
1190
1191	static void sock_splice_eof(struct file *file)
1192	{
1193	struct socket *sock = file->private_data;
1194	const struct proto_ops *ops;
1195
1196	ops = READ_ONCE(sock->ops);
1197	if (ops->splice_eof)
1198	ops->splice_eof(sock);
1199	}
1200
1201	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
1202	{
1203	struct file *file = iocb->ki_filp;
1204	struct socket *sock = file->private_data;
1205	struct msghdr msg = {.msg_iter = *to};
1206	ssize_t res;
1207
1208	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
1209	msg.msg_flags = MSG_DONTWAIT;
1210
1211	if (iocb->ki_pos != 0)
1212	return -ESPIPE;
1213
1214	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1215	return 0;
1216
1217	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1218	*to = msg.msg_iter;
1219	return res;
1220	}
1221
1222	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
1223	{
1224	struct file *file = iocb->ki_filp;
1225	struct socket *sock = file->private_data;
1226	struct msghdr msg = {.msg_iter = *from};
1227	ssize_t res;
1228
1229	if (iocb->ki_pos != 0)
1230	return -ESPIPE;
1231
1232	if (file->f_flags & O_NONBLOCK \|\| (iocb->ki_flags & IOCB_NOWAIT))
1233	msg.msg_flags = MSG_DONTWAIT;
1234
1235	if (sock->type == SOCK_SEQPACKET)
1236	msg.msg_flags \|= MSG_EOR;
1237
1238	if (iocb->ki_flags & IOCB_NOSIGNAL)
1239	msg.msg_flags \|= MSG_NOSIGNAL;
1240
1241	res = __sock_sendmsg(sock, &msg);
1242	*from = msg.msg_iter;
1243	return res;
1244	}
1245
1246	/*
1247	* Atomic setting of ioctl hooks to avoid race
1248	* with module unload.
1249	*/
1250
1251	static DEFINE_MUTEX(br_ioctl_mutex);
1252	static int (br_ioctl_hook)(struct net net, unsigned int cmd,
1253	void __user *uarg);
1254
1255	void brioctl_set(int (hook)(struct net net, unsigned int cmd,
1256	void __user *uarg))
1257	{
1258	mutex_lock(&br_ioctl_mutex);
1259	br_ioctl_hook = hook;
1260	mutex_unlock(&br_ioctl_mutex);
1261	}
1262	EXPORT_SYMBOL(brioctl_set);
1263
1264	int br_ioctl_call(struct net net, unsigned int cmd, void __user uarg)
1265	{
1266	int err = -ENOPKG;
1267
1268	if (!br_ioctl_hook)
1269	request_module("bridge");
1270
1271	mutex_lock(&br_ioctl_mutex);
1272	if (br_ioctl_hook)
1273	err = br_ioctl_hook(net, cmd, uarg);
1274	mutex_unlock(&br_ioctl_mutex);
1275
1276	return err;
1277	}
1278
1279	static DEFINE_MUTEX(vlan_ioctl_mutex);
1280	static int (vlan_ioctl_hook) (struct net , void __user *arg);
1281
1282	void vlan_ioctl_set(int (hook) (struct net , void __user *))
1283	{
1284	mutex_lock(&vlan_ioctl_mutex);
1285	vlan_ioctl_hook = hook;
1286	mutex_unlock(&vlan_ioctl_mutex);
1287	}
1288	EXPORT_SYMBOL(vlan_ioctl_set);
1289
1290	static long sock_do_ioctl(struct net net, struct socket sock,
1291	unsigned int cmd, unsigned long arg)
1292	{
1293	const struct proto_ops *ops = READ_ONCE(sock->ops);
1294	struct ifreq ifr;
1295	bool need_copyout;
1296	int err;
1297	void __user argp = (void __user )arg;
1298	void __user *data;
1299
1300	err = ops->ioctl(sock, cmd, arg);
1301
1302	/*
1303	* If this ioctl is unknown try to hand it down
1304	* to the NIC driver.
1305	*/
1306	if (err != -ENOIOCTLCMD)
1307	return err;
1308
1309	if (!is_socket_ioctl_cmd(cmd))
1310	return -ENOTTY;
1311
1312	if (get_user_ifreq(&ifr, &data, argp))
1313	return -EFAULT;
1314	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1315	if (!err && need_copyout)
1316	if (put_user_ifreq(&ifr, argp))
1317	return -EFAULT;
1318
1319	return err;
1320	}
1321
1322	/*
1323	* With an ioctl, arg may well be a user mode pointer, but we don't know
1324	* what to do with it - that's up to the protocol still.
1325	*/
1326
1327	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1328	{
1329	const struct proto_ops *ops;
1330	struct socket *sock;
1331	struct sock *sk;
1332	void __user argp = (void __user )arg;
1333	int pid, err;
1334	struct net *net;
1335
1336	sock = file->private_data;
1337	ops = READ_ONCE(sock->ops);
1338	sk = sock->sk;
1339	net = sock_net(sk);
1340	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1341	struct ifreq ifr;
1342	void __user *data;
1343	bool need_copyout;
1344	if (get_user_ifreq(&ifr, &data, argp))
1345	return -EFAULT;
1346	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1347	if (!err && need_copyout)
1348	if (put_user_ifreq(&ifr, argp))
1349	return -EFAULT;
1350	} else
1351	#ifdef CONFIG_WEXT_CORE
1352	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1353	err = wext_handle_ioctl(net, cmd, argp);
1354	} else
1355	#endif
1356	switch (cmd) {
1357	case FIOSETOWN:
1358	case SIOCSPGRP:
1359	err = -EFAULT;
1360	if (get_user(pid, (int __user *)argp))
1361	break;
1362	err = f_setown(sock->file, pid, 1);
1363	break;
1364	case FIOGETOWN:
1365	case SIOCGPGRP:
1366	err = put_user(f_getown(sock->file),
1367	(int __user *)argp);
1368	break;
1369	case SIOCGIFBR:
1370	case SIOCSIFBR:
1371	case SIOCBRADDBR:
1372	case SIOCBRDELBR:
1373	case SIOCBRADDIF:
1374	case SIOCBRDELIF:
1375	err = br_ioctl_call(net, cmd, argp);
1376	break;
1377	case SIOCGIFVLAN:
1378	case SIOCSIFVLAN:
1379	err = -ENOPKG;
1380	if (!vlan_ioctl_hook)
1381	request_module("8021q");
1382
1383	mutex_lock(&vlan_ioctl_mutex);
1384	if (vlan_ioctl_hook)
1385	err = vlan_ioctl_hook(net, argp);
1386	mutex_unlock(&vlan_ioctl_mutex);
1387	break;
1388	case SIOCGSKNS:
1389	err = -EPERM;
1390	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1391	break;
1392
1393	err = open_related_ns(&net->ns, get_net_ns);
1394	break;
1395	case SIOCGSTAMP_OLD:
1396	case SIOCGSTAMPNS_OLD:
1397	if (!ops->gettstamp) {
1398	err = -ENOIOCTLCMD;
1399	break;
1400	}
1401	err = ops->gettstamp(sock, argp,
1402	cmd == SIOCGSTAMP_OLD,
1403	!IS_ENABLED(CONFIG_64BIT));
1404	break;
1405	case SIOCGSTAMP_NEW:
1406	case SIOCGSTAMPNS_NEW:
1407	if (!ops->gettstamp) {
1408	err = -ENOIOCTLCMD;
1409	break;
1410	}
1411	err = ops->gettstamp(sock, argp,
1412	cmd == SIOCGSTAMP_NEW,
1413	false);
1414	break;
1415
1416	case SIOCGIFCONF:
1417	err = dev_ifconf(net, argp);
1418	break;
1419
1420	default:
1421	err = sock_do_ioctl(net, sock, cmd, arg);
1422	break;
1423	}
1424	return err;
1425	}
1426
1427	/**
1428	* sock_create_lite - creates a socket
1429	* @family: protocol family (AF_INET, ...)
1430	* @type: communication type (SOCK_STREAM, ...)
1431	* @protocol: protocol (0, ...)
1432	* @res: new socket
1433	*
1434	* Creates a new socket and assigns it to @res, passing through LSM.
1435	* The new socket initialization is not complete, see kernel_accept().
1436	* Returns 0 or an error. On failure @res is set to %NULL.
1437	* This function internally uses GFP_KERNEL.
1438	*/
1439
1440	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1441	{
1442	int err;
1443	struct socket *sock = NULL;
1444
1445	err = security_socket_create(family, type, protocol, 1);
1446	if (err)
1447	goto out;
1448
1449	sock = sock_alloc();
1450	if (!sock) {
1451	err = -ENOMEM;
1452	goto out;
1453	}
1454
1455	sock->type = type;
1456	err = security_socket_post_create(sock, family, type, protocol, 1);
1457	if (err)
1458	goto out_release;
1459
1460	out:
1461	*res = sock;
1462	return err;
1463	out_release:
1464	sock_release(sock);
1465	sock = NULL;
1466	goto out;
1467	}
1468	EXPORT_SYMBOL(sock_create_lite);
1469
1470	/* No kernel lock held - perfect */
1471	static __poll_t sock_poll(struct file file, poll_table wait)
1472	{
1473	struct socket *sock = file->private_data;
1474	const struct proto_ops *ops = READ_ONCE(sock->ops);
1475	__poll_t events = poll_requested_events(wait), flag = 0;
1476
1477	if (!ops->poll)
1478	return 0;
1479
1480	if (sk_can_busy_loop(sock->sk)) {
1481	/* poll once if requested by the syscall */
1482	if (events & POLL_BUSY_LOOP)
1483	sk_busy_loop(sock->sk, 1);
1484
1485	/* if this socket can poll_ll, tell the system call */
1486	flag = POLL_BUSY_LOOP;
1487	}
1488
1489	return ops->poll(file, sock, wait) \| flag;
1490	}
1491
1492	static int sock_mmap(struct file file, struct vm_area_struct vma)
1493	{
1494	struct socket *sock = file->private_data;
1495
1496	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1497	}
1498
1499	static int sock_close(struct inode inode, struct file filp)
1500	{
1501	__sock_release(SOCKET_I(inode), inode);
1502	return 0;
1503	}
1504
1505	/*
1506	* Update the socket async list
1507	*
1508	* Fasync_list locking strategy.
1509	*
1510	* 1. fasync_list is modified only under process context socket lock
1511	* i.e. under semaphore.
1512	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1513	* or under socket lock
1514	*/
1515
1516	static int sock_fasync(int fd, struct file *filp, int on)
1517	{
1518	struct socket *sock = filp->private_data;
1519	struct sock *sk = sock->sk;
1520	struct socket_wq *wq = &sock->wq;
1521
1522	if (sk == NULL)
1523	return -EINVAL;
1524
1525	lock_sock(sk);
1526	fasync_helper(fd, filp, on, &wq->fasync_list);
1527
1528	if (!wq->fasync_list)
1529	sock_reset_flag(sk, SOCK_FASYNC);
1530	else
1531	sock_set_flag(sk, SOCK_FASYNC);
1532
1533	release_sock(sk);
1534	return 0;
1535	}
1536
1537	/* This function may be called only under rcu_lock */
1538
1539	int sock_wake_async(struct socket_wq *wq, int how, int band)
1540	{
1541	if (!wq \|\| !wq->fasync_list)
1542	return -1;
1543
1544	switch (how) {
1545	case SOCK_WAKE_WAITD:
1546	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1547	break;
1548	goto call_kill;
1549	case SOCK_WAKE_SPACE:
1550	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1551	break;
1552	fallthrough;
1553	case SOCK_WAKE_IO:
1554	call_kill:
1555	kill_fasync(&wq->fasync_list, SIGIO, band);
1556	break;
1557	case SOCK_WAKE_URG:
1558	kill_fasync(&wq->fasync_list, SIGURG, band);
1559	}
1560
1561	return 0;
1562	}
1563	EXPORT_SYMBOL(sock_wake_async);
1564
1565	/**
1566	* __sock_create - creates a socket
1567	* @net: net namespace
1568	* @family: protocol family (AF_INET, ...)
1569	* @type: communication type (SOCK_STREAM, ...)
1570	* @protocol: protocol (0, ...)
1571	* @res: new socket
1572	* @kern: boolean for kernel space sockets
1573	*
1574	* Creates a new socket and assigns it to @res, passing through LSM.
1575	* Returns 0 or an error. On failure @res is set to %NULL. @kern must
1576	* be set to true if the socket resides in kernel space.
1577	* This function internally uses GFP_KERNEL.
1578	*/
1579
1580	int __sock_create(struct net *net, int family, int type, int protocol,
1581	struct socket **res, int kern)
1582	{
1583	int err;
1584	struct socket *sock;
1585	const struct net_proto_family *pf;
1586
1587	/*
1588	* Check protocol is in range
1589	*/
1590	if (family < 0 \|\| family >= NPROTO)
1591	return -EAFNOSUPPORT;
1592	if (type < 0 \|\| type >= SOCK_MAX)
1593	return -EINVAL;
1594
1595	/* Compatibility.
1596
1597	This uglymoron is moved from INET layer to here to avoid
1598	deadlock in module load.
1599	*/
1600	if (family == PF_INET && type == SOCK_PACKET) {
1601	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1602	current->comm);
1603	family = PF_PACKET;
1604	}
1605
1606	err = security_socket_create(family, type, protocol, kern);
1607	if (err)
1608	return err;
1609
1610	/*
1611	* Allocate the socket and allow the family to set things up. if
1612	* the protocol is 0, the family is instructed to select an appropriate
1613	* default.
1614	*/
1615	sock = sock_alloc();
1616	if (!sock) {
1617	net_warn_ratelimited("socket: no more sockets\n");
1618	return -ENFILE; /* Not exactly a match, but its the
1619	closest posix thing */
1620	}
1621
1622	sock->type = type;
1623
1624	#ifdef CONFIG_MODULES
1625	/* Attempt to load a protocol module if the find failed.
1626	*
1627	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1628	* requested real, full-featured networking support upon configuration.
1629	* Otherwise module support will break!
1630	*/
1631	if (rcu_access_pointer(net_families[family]) == NULL)
1632	request_module("net-pf-%d", family);
1633	#endif
1634
1635	rcu_read_lock();
1636	pf = rcu_dereference(net_families[family]);
1637	err = -EAFNOSUPPORT;
1638	if (!pf)
1639	goto out_release;
1640
1641	/*
1642	* We will call the ->create function, that possibly is in a loadable
1643	* module, so we have to bump that loadable module refcnt first.
1644	*/
1645	if (!try_module_get(pf->owner))
1646	goto out_release;
1647
1648	/* Now protected by module ref count */
1649	rcu_read_unlock();
1650
1651	err = pf->create(net, sock, protocol, kern);
1652	if (err < 0) {
1653	/* ->create should release the allocated sock->sk object on error
1654	* and make sure sock->sk is set to NULL to avoid use-after-free
1655	*/
1656	DEBUG_NET_WARN_ONCE(sock->sk,
1657	"%ps must clear sock->sk on failure, family: %d, type: %d, protocol: %d\n",
1658	pf->create, family, type, protocol);
1659	goto out_module_put;
1660	}
1661
1662	/*
1663	* Now to bump the refcnt of the [loadable] module that owns this
1664	* socket at sock_release time we decrement its refcnt.
1665	*/
1666	if (!try_module_get(sock->ops->owner))
1667	goto out_module_busy;
1668
1669	/*
1670	* Now that we're done with the ->create function, the [loadable]
1671	* module can have its refcnt decremented
1672	*/
1673	module_put(pf->owner);
1674	err = security_socket_post_create(sock, family, type, protocol, kern);
1675	if (err)
1676	goto out_sock_release;
1677	*res = sock;
1678
1679	return 0;
1680
1681	out_module_busy:
1682	err = -EAFNOSUPPORT;
1683	out_module_put:
1684	sock->ops = NULL;
1685	module_put(pf->owner);
1686	out_sock_release:
1687	sock_release(sock);
1688	return err;
1689
1690	out_release:
1691	rcu_read_unlock();
1692	goto out_sock_release;
1693	}
1694	EXPORT_SYMBOL(__sock_create);
1695
1696	/**
1697	* sock_create - creates a socket
1698	* @family: protocol family (AF_INET, ...)
1699	* @type: communication type (SOCK_STREAM, ...)
1700	* @protocol: protocol (0, ...)
1701	* @res: new socket
1702	*
1703	* A wrapper around __sock_create().
1704	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1705	*/
1706
1707	int sock_create(int family, int type, int protocol, struct socket **res)
1708	{
1709	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1710	}
1711	EXPORT_SYMBOL(sock_create);
1712
1713	/**
1714	* sock_create_kern - creates a socket (kernel space)
1715	* @net: net namespace
1716	* @family: protocol family (AF_INET, ...)
1717	* @type: communication type (SOCK_STREAM, ...)
1718	* @protocol: protocol (0, ...)
1719	* @res: new socket
1720	*
1721	* A wrapper around __sock_create().
1722	* Returns 0 or an error. This function internally uses GFP_KERNEL.
1723	*/
1724
1725	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1726	{
1727	return __sock_create(net, family, type, protocol, res, 1);
1728	}
1729	EXPORT_SYMBOL(sock_create_kern);
1730
1731	static struct socket *__sys_socket_create(int family, int type, int protocol)
1732	{
1733	struct socket *sock;
1734	int retval;
1735
1736	/* Check the SOCK_* constants for consistency. */
1737	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1738	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1739	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1740	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1741
1742	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1743	return ERR_PTR(-EINVAL);
1744	type &= SOCK_TYPE_MASK;
1745
1746	retval = sock_create(family, type, protocol, &sock);
1747	if (retval < 0)
1748	return ERR_PTR(retval);
1749
1750	return sock;
1751	}
1752
1753	struct file *__sys_socket_file(int family, int type, int protocol)
1754	{
1755	struct socket *sock;
1756	int flags;
1757
1758	sock = __sys_socket_create(family, type, protocol);
1759	if (IS_ERR(sock))
1760	return ERR_CAST(sock);
1761
1762	flags = type & ~SOCK_TYPE_MASK;
1763	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1764	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1765
1766	return sock_alloc_file(sock, flags, NULL);
1767	}
1768
1769	/* A hook for bpf progs to attach to and update socket protocol.
1770	*
1771	* A static noinline declaration here could cause the compiler to
1772	* optimize away the function. A global noinline declaration will
1773	* keep the definition, but may optimize away the callsite.
1774	* Therefore, __weak is needed to ensure that the call is still
1775	* emitted, by telling the compiler that we don't know what the
1776	* function might eventually be.
1777	*/
1778
1779	__bpf_hook_start();
1780
1781	__weak noinline int update_socket_protocol(int family, int type, int protocol)
1782	{
1783	return protocol;
1784	}
1785
1786	__bpf_hook_end();
1787
1788	int __sys_socket(int family, int type, int protocol)
1789	{
1790	struct socket *sock;
1791	int flags;
1792
1793	sock = __sys_socket_create(family, type,
1794	update_socket_protocol(family, type, protocol));
1795	if (IS_ERR(sock))
1796	return PTR_ERR(sock);
1797
1798	flags = type & ~SOCK_TYPE_MASK;
1799	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1800	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1801
1802	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1803	}
1804
1805	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1806	{
1807	return __sys_socket(family, type, protocol);
1808	}
1809
1810	/*
1811	* Create a pair of connected sockets.
1812	*/
1813
1814	int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1815	{
1816	struct socket sock1, sock2;
1817	int fd1, fd2, err;
1818	struct file newfile1, newfile2;
1819	int flags;
1820
1821	flags = type & ~SOCK_TYPE_MASK;
1822	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1823	return -EINVAL;
1824	type &= SOCK_TYPE_MASK;
1825
1826	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1827	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1828
1829	/*
1830	* reserve descriptors and make sure we won't fail
1831	* to return them to userland.
1832	*/
1833	fd1 = get_unused_fd_flags(flags);
1834	if (unlikely(fd1 < 0))
1835	return fd1;
1836
1837	fd2 = get_unused_fd_flags(flags);
1838	if (unlikely(fd2 < 0)) {
1839	put_unused_fd(fd1);
1840	return fd2;
1841	}
1842
1843	err = put_user(fd1, &usockvec[0]);
1844	if (err)
1845	goto out;
1846
1847	err = put_user(fd2, &usockvec[1]);
1848	if (err)
1849	goto out;
1850
1851	/*
1852	* Obtain the first socket and check if the underlying protocol
1853	* supports the socketpair call.
1854	*/
1855
1856	err = sock_create(family, type, protocol, &sock1);
1857	if (unlikely(err < 0))
1858	goto out;
1859
1860	err = sock_create(family, type, protocol, &sock2);
1861	if (unlikely(err < 0)) {
1862	sock_release(sock1);
1863	goto out;
1864	}
1865
1866	err = security_socket_socketpair(sock1, sock2);
1867	if (unlikely(err)) {
1868	sock_release(sock2);
1869	sock_release(sock1);
1870	goto out;
1871	}
1872
1873	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1874	if (unlikely(err < 0)) {
1875	sock_release(sock2);
1876	sock_release(sock1);
1877	goto out;
1878	}
1879
1880	newfile1 = sock_alloc_file(sock1, flags, NULL);
1881	if (IS_ERR(newfile1)) {
1882	err = PTR_ERR(newfile1);
1883	sock_release(sock2);
1884	goto out;
1885	}
1886
1887	newfile2 = sock_alloc_file(sock2, flags, NULL);
1888	if (IS_ERR(newfile2)) {
1889	err = PTR_ERR(newfile2);
1890	fput(newfile1);
1891	goto out;
1892	}
1893
1894	audit_fd_pair(fd1, fd2);
1895
1896	fd_install(fd1, newfile1);
1897	fd_install(fd2, newfile2);
1898	return 0;
1899
1900	out:
1901	put_unused_fd(fd2);
1902	put_unused_fd(fd1);
1903	return err;
1904	}
1905
1906	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1907	int __user *, usockvec)
1908	{
1909	return __sys_socketpair(family, type, protocol, usockvec);
1910	}
1911
1912	int __sys_bind_socket(struct socket sock, struct sockaddr_storage address,
1913	int addrlen)
1914	{
1915	int err;
1916
1917	err = security_socket_bind(sock, (struct sockaddr *)address,
1918	addrlen);
1919	if (!err)
1920	err = READ_ONCE(sock->ops)->bind(sock,
1921	(struct sockaddr_unsized *)address,
1922	addrlen);
1923	return err;
1924	}
1925
1926	/*
1927	* Bind a name to a socket. Nothing much to do here since it's
1928	* the protocol's responsibility to handle the local address.
1929	*
1930	* We move the socket address to kernel space before we call
1931	* the protocol layer (having also checked the address is ok).
1932	*/
1933
1934	int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1935	{
1936	struct socket *sock;
1937	struct sockaddr_storage address;
1938	CLASS(fd, f)(fd);
1939	int err;
1940
1941	if (fd_empty(f))
1942	return -EBADF;
1943	sock = sock_from_file(fd_file(f));
1944	if (unlikely(!sock))
1945	return -ENOTSOCK;
1946
1947	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1948	if (unlikely(err))
1949	return err;
1950
1951	return __sys_bind_socket(sock, &address, addrlen);
1952	}
1953
1954	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1955	{
1956	return __sys_bind(fd, umyaddr, addrlen);
1957	}
1958
1959	/*
1960	* Perform a listen. Basically, we allow the protocol to do anything
1961	* necessary for a listen, and if that works, we mark the socket as
1962	* ready for listening.
1963	*/
1964	int __sys_listen_socket(struct socket *sock, int backlog)
1965	{
1966	int somaxconn, err;
1967
1968	somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1969	if ((unsigned int)backlog > somaxconn)
1970	backlog = somaxconn;
1971
1972	err = security_socket_listen(sock, backlog);
1973	if (!err)
1974	err = READ_ONCE(sock->ops)->listen(sock, backlog);
1975	return err;
1976	}
1977
1978	int __sys_listen(int fd, int backlog)
1979	{
1980	CLASS(fd, f)(fd);
1981	struct socket *sock;
1982
1983	if (fd_empty(f))
1984	return -EBADF;
1985	sock = sock_from_file(fd_file(f));
1986	if (unlikely(!sock))
1987	return -ENOTSOCK;
1988
1989	return __sys_listen_socket(sock, backlog);
1990	}
1991
1992	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1993	{
1994	return __sys_listen(fd, backlog);
1995	}
1996
1997	struct file do_accept(struct file file, struct proto_accept_arg *arg,
1998	struct sockaddr __user *upeer_sockaddr,
1999	int __user *upeer_addrlen, int flags)
2000	{
2001	struct socket sock, newsock;
2002	struct file *newfile;
2003	int err, len;
2004	struct sockaddr_storage address;
2005	const struct proto_ops *ops;
2006
2007	sock = sock_from_file(file);
2008	if (!sock)
2009	return ERR_PTR(-ENOTSOCK);
2010
2011	newsock = sock_alloc();
2012	if (!newsock)
2013	return ERR_PTR(-ENFILE);
2014	ops = READ_ONCE(sock->ops);
2015
2016	newsock->type = sock->type;
2017	newsock->ops = ops;
2018
2019	/*
2020	* We don't need try_module_get here, as the listening socket (sock)
2021	* has the protocol module (sock->ops->owner) held.
2022	*/
2023	__module_get(ops->owner);
2024
2025	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
2026	if (IS_ERR(newfile))
2027	return newfile;
2028
2029	err = security_socket_accept(sock, newsock);
2030	if (err)
2031	goto out_fd;
2032
2033	arg->flags \|= sock->file->f_flags;
2034	err = ops->accept(sock, newsock, arg);
2035	if (err < 0)
2036	goto out_fd;
2037
2038	if (upeer_sockaddr) {
2039	len = ops->getname(newsock, (struct sockaddr *)&address, 2);
2040	if (len < 0) {
2041	err = -ECONNABORTED;
2042	goto out_fd;
2043	}
2044	err = move_addr_to_user(&address,
2045	len, upeer_sockaddr, upeer_addrlen);
2046	if (err < 0)
2047	goto out_fd;
2048	}
2049
2050	/* File flags are not inherited via accept() unlike another OSes. */
2051	return newfile;
2052	out_fd:
2053	fput(newfile);
2054	return ERR_PTR(err);
2055	}
2056
2057	static int __sys_accept4_file(struct file file, struct sockaddr __user upeer_sockaddr,
2058	int __user *upeer_addrlen, int flags)
2059	{
2060	struct proto_accept_arg arg = { };
2061
2062	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
2063	return -EINVAL;
2064
2065	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
2066	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
2067
2068	return FD_ADD(flags, do_accept(file, &arg, upeer_sockaddr, upeer_addrlen, flags));
2069	}
2070
2071	/*
2072	* For accept, we attempt to create a new socket, set up the link
2073	* with the client, wake up the client, then return the new
2074	* connected fd. We collect the address of the connector in kernel
2075	* space and move it to user at the very end. This is unclean because
2076	* we open the socket then return an error.
2077	*
2078	* 1003.1g adds the ability to recvmsg() to query connection pending
2079	* status. We need to add that support in a way that's
2080	* clean when we restructure accept also.
2081	*/
2082
2083	int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2084	int __user *upeer_addrlen, int flags)
2085	{
2086	CLASS(fd, f)(fd);
2087
2088	if (fd_empty(f))
2089	return -EBADF;
2090	return __sys_accept4_file(fd_file(f), upeer_sockaddr,
2091	upeer_addrlen, flags);
2092	}
2093
2094	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2095	int __user *, upeer_addrlen, int, flags)
2096	{
2097	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2098	}
2099
2100	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2101	int __user *, upeer_addrlen)
2102	{
2103	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2104	}
2105
2106	/*
2107	* Attempt to connect to a socket with the server address. The address
2108	* is in user space so we verify it is OK and move it to kernel space.
2109	*
2110	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2111	* break bindings
2112	*
2113	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2114	* other SEQPACKET protocols that take time to connect() as it doesn't
2115	* include the -EINPROGRESS status for such sockets.
2116	*/
2117
2118	int __sys_connect_file(struct file file, struct sockaddr_storage address,
2119	int addrlen, int file_flags)
2120	{
2121	struct socket *sock;
2122	int err;
2123
2124	sock = sock_from_file(file);
2125	if (!sock) {
2126	err = -ENOTSOCK;
2127	goto out;
2128	}
2129
2130	err =
2131	security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2132	if (err)
2133	goto out;
2134
2135	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr_unsized *)address,
2136	addrlen, sock->file->f_flags \| file_flags);
2137	out:
2138	return err;
2139	}
2140
2141	int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2142	{
2143	struct sockaddr_storage address;
2144	CLASS(fd, f)(fd);
2145	int ret;
2146
2147	if (fd_empty(f))
2148	return -EBADF;
2149
2150	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2151	if (ret)
2152	return ret;
2153
2154	return __sys_connect_file(fd_file(f), &address, addrlen, 0);
2155	}
2156
2157	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2158	int, addrlen)
2159	{
2160	return __sys_connect(fd, uservaddr, addrlen);
2161	}
2162
2163	int do_getsockname(struct socket *sock, int peer,
2164	struct sockaddr __user usockaddr, int __user usockaddr_len)
2165	{
2166	struct sockaddr_storage address;
2167	int err;
2168
2169	if (peer)
2170	err = security_socket_getpeername(sock);
2171	else
2172	err = security_socket_getsockname(sock);
2173	if (err)
2174	return err;
2175	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, peer);
2176	if (err < 0)
2177	return err;
2178	/* "err" is actually length in this case */
2179	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2180	}
2181
2182	/*
2183	* Get the remote or local address ('name') of a socket object. Move the
2184	* obtained name to user space.
2185	*/
2186	int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2187	int __user *usockaddr_len, int peer)
2188	{
2189	struct socket *sock;
2190	CLASS(fd, f)(fd);
2191
2192	if (fd_empty(f))
2193	return -EBADF;
2194	sock = sock_from_file(fd_file(f));
2195	if (unlikely(!sock))
2196	return -ENOTSOCK;
2197	return do_getsockname(sock, peer, usockaddr, usockaddr_len);
2198	}
2199
2200	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2201	int __user *, usockaddr_len)
2202	{
2203	return __sys_getsockname(fd, usockaddr, usockaddr_len, 0);
2204	}
2205
2206	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2207	int __user *, usockaddr_len)
2208	{
2209	return __sys_getsockname(fd, usockaddr, usockaddr_len, 1);
2210	}
2211
2212	/*
2213	* Send a datagram to a given address. We move the address into kernel
2214	* space and check the user space data area is readable before invoking
2215	* the protocol.
2216	*/
2217	int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2218	struct sockaddr __user *addr, int addr_len)
2219	{
2220	struct socket *sock;
2221	struct sockaddr_storage address;
2222	int err;
2223	struct msghdr msg;
2224
2225	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2226	if (unlikely(err))
2227	return err;
2228
2229	CLASS(fd, f)(fd);
2230	if (fd_empty(f))
2231	return -EBADF;
2232	sock = sock_from_file(fd_file(f));
2233	if (unlikely(!sock))
2234	return -ENOTSOCK;
2235
2236	msg.msg_name = NULL;
2237	msg.msg_control = NULL;
2238	msg.msg_controllen = 0;
2239	msg.msg_namelen = 0;
2240	msg.msg_ubuf = NULL;
2241	if (addr) {
2242	err = move_addr_to_kernel(addr, addr_len, &address);
2243	if (err < 0)
2244	return err;
2245	msg.msg_name = (struct sockaddr *)&address;
2246	msg.msg_namelen = addr_len;
2247	}
2248	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2249	if (sock->file->f_flags & O_NONBLOCK)
2250	flags \|= MSG_DONTWAIT;
2251	msg.msg_flags = flags;
2252	return __sock_sendmsg(sock, &msg);
2253	}
2254
2255	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2256	unsigned int, flags, struct sockaddr __user *, addr,
2257	int, addr_len)
2258	{
2259	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2260	}
2261
2262	/*
2263	* Send a datagram down a socket.
2264	*/
2265
2266	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2267	unsigned int, flags)
2268	{
2269	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2270	}
2271
2272	/*
2273	* Receive a frame from the socket and optionally record the address of the
2274	* sender. We verify the buffers are writable and if needed move the
2275	* sender address from kernel to user space.
2276	*/
2277	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2278	struct sockaddr __user addr, int __user addr_len)
2279	{
2280	struct sockaddr_storage address;
2281	struct msghdr msg = {
2282	/* Save some cycles and don't copy the address if not needed */
2283	.msg_name = addr ? (struct sockaddr *)&address : NULL,
2284	};
2285	struct socket *sock;
2286	int err, err2;
2287
2288	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2289	if (unlikely(err))
2290	return err;
2291
2292	CLASS(fd, f)(fd);
2293
2294	if (fd_empty(f))
2295	return -EBADF;
2296	sock = sock_from_file(fd_file(f));
2297	if (unlikely(!sock))
2298	return -ENOTSOCK;
2299
2300	if (sock->file->f_flags & O_NONBLOCK)
2301	flags \|= MSG_DONTWAIT;
2302	err = sock_recvmsg(sock, &msg, flags);
2303
2304	if (err >= 0 && addr != NULL) {
2305	err2 = move_addr_to_user(&address,
2306	msg.msg_namelen, addr, addr_len);
2307	if (err2 < 0)
2308	err = err2;
2309	}
2310	return err;
2311	}
2312
2313	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2314	unsigned int, flags, struct sockaddr __user *, addr,
2315	int __user *, addr_len)
2316	{
2317	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2318	}
2319
2320	/*
2321	* Receive a datagram from a socket.
2322	*/
2323
2324	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2325	unsigned int, flags)
2326	{
2327	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2328	}
2329
2330	static bool sock_use_custom_sol_socket(const struct socket *sock)
2331	{
2332	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2333	}
2334
2335	int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2336	int optname, sockptr_t optval, int optlen)
2337	{
2338	const struct proto_ops *ops;
2339	char *kernel_optval = NULL;
2340	int err;
2341
2342	if (optlen < 0)
2343	return -EINVAL;
2344
2345	err = security_socket_setsockopt(sock, level, optname);
2346	if (err)
2347	goto out_put;
2348
2349	if (!compat)
2350	err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2351	optval, &optlen,
2352	&kernel_optval);
2353	if (err < 0)
2354	goto out_put;
2355	if (err > 0) {
2356	err = 0;
2357	goto out_put;
2358	}
2359
2360	if (kernel_optval)
2361	optval = KERNEL_SOCKPTR(kernel_optval);
2362	ops = READ_ONCE(sock->ops);
2363	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2364	err = sock_setsockopt(sock, level, optname, optval, optlen);
2365	else if (unlikely(!ops->setsockopt))
2366	err = -EOPNOTSUPP;
2367	else
2368	err = ops->setsockopt(sock, level, optname, optval,
2369	optlen);
2370	kfree(kernel_optval);
2371	out_put:
2372	return err;
2373	}
2374	EXPORT_SYMBOL(do_sock_setsockopt);
2375
2376	/* Set a socket option. Because we don't know the option lengths we have
2377	* to pass the user mode parameter for the protocols to sort out.
2378	*/
2379	int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2380	int optlen)
2381	{
2382	sockptr_t optval = USER_SOCKPTR(user_optval);
2383	bool compat = in_compat_syscall();
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_setsockopt(sock, compat, level, optname, optval, optlen);
2394	}
2395
2396	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2397	char __user *, optval, int, optlen)
2398	{
2399	return __sys_setsockopt(fd, level, optname, optval, optlen);
2400	}
2401
2402	INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2403	int optname));
2404
2405	/*
2406	* Initialize a sockopt_t from sockptr optval/optlen, setting up iov_iter
2407	* for both input and output directions.
2408	* It is important to remember that both iov points to the same data, but,
2409	* .iter_in is read-only and .iter_out is write-only by the protocol callbacks
2410	*/
2411	static int sockptr_to_sockopt(sockopt_t *opt, sockptr_t optval,
2412	sockptr_t optlen, struct kvec *kvec)
2413	{
2414	int koptlen;
2415
2416	if (copy_from_sockptr(&koptlen, optlen, sizeof(int)))
2417	return -EFAULT;
2418
2419	if (koptlen < 0)
2420	return -EINVAL;
2421
2422	if (optval.is_kernel) {
2423	kvec->iov_base = optval.kernel;
2424	kvec->iov_len = koptlen;
2425	iov_iter_kvec(&opt->iter_out, ITER_DEST, kvec, 1, koptlen);
2426	iov_iter_kvec(&opt->iter_in, ITER_SOURCE, kvec, 1, koptlen);
2427	} else {
2428	iov_iter_ubuf(&opt->iter_out, ITER_DEST, optval.user, koptlen);
2429	iov_iter_ubuf(&opt->iter_in, ITER_SOURCE, optval.user,
2430	koptlen);
2431	}
2432	opt->optlen = koptlen;
2433
2434	return 0;
2435	}
2436
2437	int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2438	int optname, sockptr_t optval, sockptr_t optlen)
2439	{
2440	int max_optlen __maybe_unused = 0;
2441	const struct proto_ops *ops;
2442	struct kvec kvec;
2443	sockopt_t opt;
2444	int err;
2445
2446	err = security_socket_getsockopt(sock, level, optname);
2447	if (err)
2448	return err;
2449
2450	if (!compat)
2451	copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2452
2453	ops = READ_ONCE(sock->ops);
2454	if (level == SOL_SOCKET) {
2455	err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2456	} else if (ops->getsockopt_iter) {
2457	err = sockptr_to_sockopt(&opt, optval, optlen, &kvec);
2458	if (err)
2459	return err;
2460
2461	err = ops->getsockopt_iter(sock, level, optname, &opt);
2462
2463	/* Always write back optlen, even on failure. Some protocols
2464	* (e.g. CAN raw) return -ERANGE and set optlen to the
2465	* required buffer size so userspace can discover it.
2466	*/
2467	if (copy_to_sockptr(optlen, &opt.optlen, sizeof(int)))
2468	return -EFAULT;
2469	} else if (ops->getsockopt) {
2470	if (WARN_ONCE(optval.is_kernel \|\| optlen.is_kernel,
2471	"Invalid argument type"))
2472	return -EOPNOTSUPP;
2473
2474	err = ops->getsockopt(sock, level, optname, optval.user,
2475	optlen.user);
2476	} else {
2477	err = -EOPNOTSUPP;
2478	}
2479
2480	if (!compat)
2481	err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2482	optval, optlen, max_optlen,
2483	err);
2484
2485	return err;
2486	}
2487	EXPORT_SYMBOL(do_sock_getsockopt);
2488
2489	/*
2490	* Get a socket option. Because we don't know the option lengths we have
2491	* to pass a user mode parameter for the protocols to sort out.
2492	*/
2493	int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2494	int __user *optlen)
2495	{
2496	struct socket *sock;
2497	CLASS(fd, f)(fd);
2498
2499	if (fd_empty(f))
2500	return -EBADF;
2501	sock = sock_from_file(fd_file(f));
2502	if (unlikely(!sock))
2503	return -ENOTSOCK;
2504
2505	return do_sock_getsockopt(sock, in_compat_syscall(), level, optname,
2506	USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2507	}
2508
2509	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2510	char __user , optval, int __user , optlen)
2511	{
2512	return __sys_getsockopt(fd, level, optname, optval, optlen);
2513	}
2514
2515	/*
2516	* Shutdown a socket.
2517	*/
2518
2519	int __sys_shutdown_sock(struct socket *sock, int how)
2520	{
2521	int err;
2522
2523	err = security_socket_shutdown(sock, how);
2524	if (!err)
2525	err = READ_ONCE(sock->ops)->shutdown(sock, how);
2526
2527	return err;
2528	}
2529
2530	int __sys_shutdown(int fd, int how)
2531	{
2532	struct socket *sock;
2533	CLASS(fd, f)(fd);
2534
2535	if (fd_empty(f))
2536	return -EBADF;
2537	sock = sock_from_file(fd_file(f));
2538	if (unlikely(!sock))
2539	return -ENOTSOCK;
2540
2541	return __sys_shutdown_sock(sock, how);
2542	}
2543
2544	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2545	{
2546	return __sys_shutdown(fd, how);
2547	}
2548
2549	/* A couple of helpful macros for getting the address of the 32/64 bit
2550	* fields which are the same type (int / unsigned) on our platforms.
2551	*/
2552	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2553	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2554	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2555
2556	struct used_address {
2557	struct sockaddr_storage name;
2558	unsigned int name_len;
2559	};
2560
2561	int __copy_msghdr(struct msghdr *kmsg,
2562	struct user_msghdr *msg,
2563	struct sockaddr __user **save_addr)
2564	{
2565	ssize_t err;
2566
2567	kmsg->msg_control_is_user = true;
2568	kmsg->msg_get_inq = 0;
2569	kmsg->msg_control_user = msg->msg_control;
2570	kmsg->msg_controllen = msg->msg_controllen;
2571	kmsg->msg_flags = msg->msg_flags;
2572
2573	kmsg->msg_namelen = msg->msg_namelen;
2574	if (!msg->msg_name)
2575	kmsg->msg_namelen = 0;
2576
2577	if (kmsg->msg_namelen < 0)
2578	return -EINVAL;
2579
2580	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2581	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2582
2583	if (save_addr)
2584	*save_addr = msg->msg_name;
2585
2586	if (msg->msg_name && kmsg->msg_namelen) {
2587	if (!save_addr) {
2588	err = move_addr_to_kernel(msg->msg_name,
2589	kmsg->msg_namelen,
2590	kmsg->msg_name);
2591	if (err < 0)
2592	return err;
2593	}
2594	} else {
2595	kmsg->msg_name = NULL;
2596	kmsg->msg_namelen = 0;
2597	}
2598
2599	if (msg->msg_iovlen > UIO_MAXIOV)
2600	return -EMSGSIZE;
2601
2602	kmsg->msg_ubuf = NULL;
2603	return 0;
2604	}
2605
2606	static int copy_msghdr_from_user(struct msghdr *kmsg,
2607	struct user_msghdr __user *umsg,
2608	struct sockaddr __user **save_addr,
2609	struct iovec **iov)
2610	{
2611	struct user_msghdr msg;
2612	ssize_t err;
2613
2614	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2615	return -EFAULT;
2616
2617	err = __copy_msghdr(kmsg, &msg, save_addr);
2618	if (err)
2619	return err;
2620
2621	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2622	msg.msg_iov, msg.msg_iovlen,
2623	UIO_FASTIOV, iov, &kmsg->msg_iter);
2624	return err < 0 ? err : 0;
2625	}
2626
2627	static int ____sys_sendmsg(struct socket sock, struct msghdr msg_sys,
2628	unsigned int flags, struct used_address *used_address,
2629	unsigned int allowed_msghdr_flags)
2630	{
2631	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2632	__aligned(sizeof(__kernel_size_t));
2633	/* 20 is size of ipv6_pktinfo */
2634	unsigned char *ctl_buf = ctl;
2635	int ctl_len;
2636	ssize_t err;
2637
2638	err = -ENOBUFS;
2639
2640	if (msg_sys->msg_controllen > INT_MAX)
2641	goto out;
2642	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2643	ctl_len = msg_sys->msg_controllen;
2644	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2645	err =
2646	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2647	sizeof(ctl));
2648	if (err)
2649	goto out;
2650	ctl_buf = msg_sys->msg_control;
2651	ctl_len = msg_sys->msg_controllen;
2652	} else if (ctl_len) {
2653	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2654	CMSG_ALIGN(sizeof(struct cmsghdr)));
2655	if (ctl_len > sizeof(ctl)) {
2656	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2657	if (ctl_buf == NULL)
2658	goto out;
2659	}
2660	err = -EFAULT;
2661	if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2662	goto out_freectl;
2663	msg_sys->msg_control = ctl_buf;
2664	msg_sys->msg_control_is_user = false;
2665	}
2666	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2667	msg_sys->msg_flags = flags;
2668
2669	if (sock->file->f_flags & O_NONBLOCK)
2670	msg_sys->msg_flags \|= MSG_DONTWAIT;
2671	/*
2672	* If this is sendmmsg() and current destination address is same as
2673	* previously succeeded address, omit asking LSM's decision.
2674	* used_address->name_len is initialized to UINT_MAX so that the first
2675	* destination address never matches.
2676	*/
2677	if (used_address && msg_sys->msg_name &&
2678	used_address->name_len == msg_sys->msg_namelen &&
2679	!memcmp(&used_address->name, msg_sys->msg_name,
2680	used_address->name_len)) {
2681	err = sock_sendmsg_nosec(sock, msg_sys);
2682	goto out_freectl;
2683	}
2684	err = __sock_sendmsg(sock, msg_sys);
2685	/*
2686	* If this is sendmmsg() and sending to current destination address was
2687	* successful, remember it.
2688	*/
2689	if (used_address && err >= 0) {
2690	used_address->name_len = msg_sys->msg_namelen;
2691	if (msg_sys->msg_name)
2692	memcpy(&used_address->name, msg_sys->msg_name,
2693	used_address->name_len);
2694	}
2695
2696	out_freectl:
2697	if (ctl_buf != ctl)
2698	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2699	out:
2700	return err;
2701	}
2702
2703	static int sendmsg_copy_msghdr(struct msghdr *msg,
2704	struct user_msghdr __user *umsg, unsigned flags,
2705	struct iovec **iov)
2706	{
2707	int err;
2708
2709	if (flags & MSG_CMSG_COMPAT) {
2710	struct compat_msghdr __user *msg_compat;
2711
2712	msg_compat = (struct compat_msghdr __user *) umsg;
2713	err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2714	} else {
2715	err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2716	}
2717	if (err < 0)
2718	return err;
2719
2720	return 0;
2721	}
2722
2723	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
2724	struct msghdr *msg_sys, unsigned int flags,
2725	struct used_address *used_address,
2726	unsigned int allowed_msghdr_flags)
2727	{
2728	struct sockaddr_storage address;
2729	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2730	ssize_t err;
2731
2732	msg_sys->msg_name = &address;
2733
2734	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2735	if (err < 0)
2736	return err;
2737
2738	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2739	allowed_msghdr_flags);
2740	kfree(iov);
2741	return err;
2742	}
2743
2744	/*
2745	* BSD sendmsg interface
2746	*/
2747	long __sys_sendmsg_sock(struct socket sock, struct msghdr msg,
2748	unsigned int flags)
2749	{
2750	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2751	}
2752
2753	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2754	bool forbid_cmsg_compat)
2755	{
2756	struct msghdr msg_sys;
2757	struct socket *sock;
2758
2759	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2760	return -EINVAL;
2761
2762	CLASS(fd, f)(fd);
2763
2764	if (fd_empty(f))
2765	return -EBADF;
2766	sock = sock_from_file(fd_file(f));
2767	if (unlikely(!sock))
2768	return -ENOTSOCK;
2769
2770	return ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2771	}
2772
2773	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2774	{
2775	return __sys_sendmsg(fd, msg, flags, true);
2776	}
2777
2778	/*
2779	* Linux sendmmsg interface
2780	*/
2781
2782	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2783	unsigned int flags, bool forbid_cmsg_compat)
2784	{
2785	int err, datagrams;
2786	struct socket *sock;
2787	struct mmsghdr __user *entry;
2788	struct compat_mmsghdr __user *compat_entry;
2789	struct msghdr msg_sys;
2790	struct used_address used_address;
2791	unsigned int oflags = flags;
2792
2793	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2794	return -EINVAL;
2795
2796	if (vlen > UIO_MAXIOV)
2797	vlen = UIO_MAXIOV;
2798
2799	datagrams = 0;
2800
2801	CLASS(fd, f)(fd);
2802
2803	if (fd_empty(f))
2804	return -EBADF;
2805	sock = sock_from_file(fd_file(f));
2806	if (unlikely(!sock))
2807	return -ENOTSOCK;
2808
2809	used_address.name_len = UINT_MAX;
2810	entry = mmsg;
2811	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2812	err = 0;
2813	flags \|= MSG_BATCH;
2814
2815	while (datagrams < vlen) {
2816	if (datagrams == vlen - 1)
2817	flags = oflags;
2818
2819	if (MSG_CMSG_COMPAT & flags) {
2820	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2821	&msg_sys, flags, &used_address, MSG_EOR);
2822	if (err < 0)
2823	break;
2824	err = __put_user(err, &compat_entry->msg_len);
2825	++compat_entry;
2826	} else {
2827	err = ___sys_sendmsg(sock,
2828	(struct user_msghdr __user *)entry,
2829	&msg_sys, flags, &used_address, MSG_EOR);
2830	if (err < 0)
2831	break;
2832	err = put_user(err, &entry->msg_len);
2833	++entry;
2834	}
2835
2836	if (err)
2837	break;
2838	++datagrams;
2839	if (msg_data_left(&msg_sys))
2840	break;
2841	cond_resched();
2842	}
2843
2844	/* We only return an error if no datagrams were able to be sent */
2845	if (datagrams != 0)
2846	return datagrams;
2847
2848	return err;
2849	}
2850
2851	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2852	unsigned int, vlen, unsigned int, flags)
2853	{
2854	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2855	}
2856
2857	static int recvmsg_copy_msghdr(struct msghdr *msg,
2858	struct user_msghdr __user *umsg, unsigned flags,
2859	struct sockaddr __user **uaddr,
2860	struct iovec **iov)
2861	{
2862	ssize_t err;
2863
2864	if (MSG_CMSG_COMPAT & flags) {
2865	struct compat_msghdr __user *msg_compat;
2866
2867	msg_compat = (struct compat_msghdr __user *) umsg;
2868	err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2869	} else {
2870	err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2871	}
2872	if (err < 0)
2873	return err;
2874
2875	return 0;
2876	}
2877
2878	static int ____sys_recvmsg(struct socket sock, struct msghdr msg_sys,
2879	struct user_msghdr __user *msg,
2880	struct sockaddr __user *uaddr,
2881	unsigned int flags, int nosec)
2882	{
2883	struct compat_msghdr __user *msg_compat =
2884	(struct compat_msghdr __user *) msg;
2885	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2886	struct sockaddr_storage addr;
2887	unsigned long cmsg_ptr;
2888	int len;
2889	ssize_t err;
2890
2891	msg_sys->msg_name = &addr;
2892	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2893	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2894
2895	/* We assume all kernel code knows the size of sockaddr_storage */
2896	msg_sys->msg_namelen = 0;
2897
2898	if (sock->file->f_flags & O_NONBLOCK)
2899	flags \|= MSG_DONTWAIT;
2900
2901	if (unlikely(nosec))
2902	err = sock_recvmsg_nosec(sock, msg_sys, flags);
2903	else
2904	err = sock_recvmsg(sock, msg_sys, flags);
2905
2906	if (err < 0)
2907	goto out;
2908	len = err;
2909
2910	if (uaddr != NULL) {
2911	err = move_addr_to_user(&addr,
2912	msg_sys->msg_namelen, uaddr,
2913	uaddr_len);
2914	if (err < 0)
2915	goto out;
2916	}
2917	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2918	COMPAT_FLAGS(msg));
2919	if (err)
2920	goto out;
2921	if (MSG_CMSG_COMPAT & flags)
2922	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2923	&msg_compat->msg_controllen);
2924	else
2925	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2926	&msg->msg_controllen);
2927	if (err)
2928	goto out;
2929	err = len;
2930	out:
2931	return err;
2932	}
2933
2934	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2935	struct msghdr *msg_sys, unsigned int flags, int nosec)
2936	{
2937	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2938	/* user mode address pointers */
2939	struct sockaddr __user *uaddr;
2940	ssize_t err;
2941
2942	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2943	if (err < 0)
2944	return err;
2945
2946	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2947	kfree(iov);
2948	return err;
2949	}
2950
2951	/*
2952	* BSD recvmsg interface
2953	*/
2954
2955	long __sys_recvmsg_sock(struct socket sock, struct msghdr msg,
2956	struct user_msghdr __user *umsg,
2957	struct sockaddr __user *uaddr, unsigned int flags)
2958	{
2959	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2960	}
2961
2962	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2963	bool forbid_cmsg_compat)
2964	{
2965	struct msghdr msg_sys;
2966	struct socket *sock;
2967
2968	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2969	return -EINVAL;
2970
2971	CLASS(fd, f)(fd);
2972
2973	if (fd_empty(f))
2974	return -EBADF;
2975	sock = sock_from_file(fd_file(f));
2976	if (unlikely(!sock))
2977	return -ENOTSOCK;
2978
2979	return ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2980	}
2981
2982	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2983	unsigned int, flags)
2984	{
2985	return __sys_recvmsg(fd, msg, flags, true);
2986	}
2987
2988	/*
2989	* Linux recvmmsg interface
2990	*/
2991
2992	static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2993	unsigned int vlen, unsigned int flags,
2994	struct timespec64 *timeout)
2995	{
2996	int err = 0, datagrams;
2997	struct socket *sock;
2998	struct mmsghdr __user *entry;
2999	struct compat_mmsghdr __user *compat_entry;
3000	struct msghdr msg_sys;
3001	struct timespec64 end_time;
3002	struct timespec64 timeout64;
3003
3004	if (timeout &&
3005	poll_select_set_timeout(&end_time, timeout->tv_sec,
3006	timeout->tv_nsec))
3007	return -EINVAL;
3008
3009	datagrams = 0;
3010
3011	CLASS(fd, f)(fd);
3012
3013	if (fd_empty(f))
3014	return -EBADF;
3015	sock = sock_from_file(fd_file(f));
3016	if (unlikely(!sock))
3017	return -ENOTSOCK;
3018
3019	if (likely(!(flags & MSG_ERRQUEUE))) {
3020	err = sock_error(sock->sk);
3021	if (err)
3022	return err;
3023	}
3024
3025	entry = mmsg;
3026	compat_entry = (struct compat_mmsghdr __user *)mmsg;
3027
3028	while (datagrams < vlen) {
3029	/*
3030	* No need to ask LSM for more than the first datagram.
3031	*/
3032	if (MSG_CMSG_COMPAT & flags) {
3033	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
3034	&msg_sys, flags & ~MSG_WAITFORONE,
3035	datagrams);
3036	if (err < 0)
3037	break;
3038	err = __put_user(err, &compat_entry->msg_len);
3039	++compat_entry;
3040	} else {
3041	err = ___sys_recvmsg(sock,
3042	(struct user_msghdr __user *)entry,
3043	&msg_sys, flags & ~MSG_WAITFORONE,
3044	datagrams);
3045	if (err < 0)
3046	break;
3047	err = put_user(err, &entry->msg_len);
3048	++entry;
3049	}
3050
3051	if (err)
3052	break;
3053	++datagrams;
3054
3055	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
3056	if (flags & MSG_WAITFORONE)
3057	flags \|= MSG_DONTWAIT;
3058
3059	if (timeout) {
3060	ktime_get_ts64(&timeout64);
3061	*timeout = timespec64_sub(end_time, timeout64);
3062	if (timeout->tv_sec < 0) {
3063	timeout->tv_sec = timeout->tv_nsec = 0;
3064	break;
3065	}
3066
3067	/* Timeout, return less than vlen datagrams */
3068	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
3069	break;
3070	}
3071
3072	/* Out of band data, return right away */
3073	if (msg_sys.msg_flags & MSG_OOB)
3074	break;
3075	cond_resched();
3076	}
3077
3078	if (err == 0)
3079	return datagrams;
3080
3081	if (datagrams == 0)
3082	return err;
3083
3084	/*
3085	* We may return less entries than requested (vlen) if the
3086	* sock is non block and there aren't enough datagrams...
3087	*/
3088	if (err != -EAGAIN) {
3089	/*
3090	* ... or if recvmsg returns an error after we
3091	* received some datagrams, where we record the
3092	* error to return on the next call or if the
3093	* app asks about it using getsockopt(SO_ERROR).
3094	*/
3095	WRITE_ONCE(sock->sk->sk_err, -err);
3096	}
3097	return datagrams;
3098	}
3099
3100	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3101	unsigned int vlen, unsigned int flags,
3102	struct __kernel_timespec __user *timeout,
3103	struct old_timespec32 __user *timeout32)
3104	{
3105	int datagrams;
3106	struct timespec64 timeout_sys;
3107
3108	if (timeout && get_timespec64(&timeout_sys, timeout))
3109	return -EFAULT;
3110
3111	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3112	return -EFAULT;
3113
3114	if (!timeout && !timeout32)
3115	return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3116
3117	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3118
3119	if (datagrams <= 0)
3120	return datagrams;
3121
3122	if (timeout && put_timespec64(&timeout_sys, timeout))
3123	datagrams = -EFAULT;
3124
3125	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3126	datagrams = -EFAULT;
3127
3128	return datagrams;
3129	}
3130
3131	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3132	unsigned int, vlen, unsigned int, flags,
3133	struct __kernel_timespec __user *, timeout)
3134	{
3135	if (flags & MSG_CMSG_COMPAT)
3136	return -EINVAL;
3137
3138	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3139	}
3140
3141	#ifdef CONFIG_COMPAT_32BIT_TIME
3142	SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3143	unsigned int, vlen, unsigned int, flags,
3144	struct old_timespec32 __user *, timeout)
3145	{
3146	if (flags & MSG_CMSG_COMPAT)
3147	return -EINVAL;
3148
3149	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3150	}
3151	#endif
3152
3153	#ifdef __ARCH_WANT_SYS_SOCKETCALL
3154	/* Argument list sizes for sys_socketcall */
3155	#define AL(x) ((x) * sizeof(unsigned long))
3156	static const unsigned char nargs[21] = {
3157	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3158	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3159	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3160	AL(4), AL(5), AL(4)
3161	};
3162
3163	#undef AL
3164
3165	/*
3166	* System call vectors.
3167	*
3168	* Argument checking cleaned up. Saved 20% in size.
3169	* This function doesn't need to set the kernel lock because
3170	* it is set by the callees.
3171	*/
3172
3173	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3174	{
3175	unsigned long a[AUDITSC_ARGS];
3176	unsigned long a0, a1;
3177	int err;
3178	unsigned int len;
3179
3180	if (call < 1 \|\| call > SYS_SENDMMSG)
3181	return -EINVAL;
3182	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3183
3184	len = nargs[call];
3185	if (len > sizeof(a))
3186	return -EINVAL;
3187
3188	/* copy_from_user should be SMP safe. */
3189	if (copy_from_user(a, args, len))
3190	return -EFAULT;
3191
3192	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3193	if (err)
3194	return err;
3195
3196	a0 = a[0];
3197	a1 = a[1];
3198
3199	switch (call) {
3200	case SYS_SOCKET:
3201	err = __sys_socket(a0, a1, a[2]);
3202	break;
3203	case SYS_BIND:
3204	err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3205	break;
3206	case SYS_CONNECT:
3207	err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3208	break;
3209	case SYS_LISTEN:
3210	err = __sys_listen(a0, a1);
3211	break;
3212	case SYS_ACCEPT:
3213	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3214	(int __user *)a[2], 0);
3215	break;
3216	case SYS_GETSOCKNAME:
3217	err =
3218	__sys_getsockname(a0, (struct sockaddr __user *)a1,
3219	(int __user *)a[2], 0);
3220	break;
3221	case SYS_GETPEERNAME:
3222	err =
3223	__sys_getsockname(a0, (struct sockaddr __user *)a1,
3224	(int __user *)a[2], 1);
3225	break;
3226	case SYS_SOCKETPAIR:
3227	err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3228	break;
3229	case SYS_SEND:
3230	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3231	NULL, 0);
3232	break;
3233	case SYS_SENDTO:
3234	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3235	(struct sockaddr __user *)a[4], a[5]);
3236	break;
3237	case SYS_RECV:
3238	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3239	NULL, NULL);
3240	break;
3241	case SYS_RECVFROM:
3242	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3243	(struct sockaddr __user *)a[4],
3244	(int __user *)a[5]);
3245	break;
3246	case SYS_SHUTDOWN:
3247	err = __sys_shutdown(a0, a1);
3248	break;
3249	case SYS_SETSOCKOPT:
3250	err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3251	a[4]);
3252	break;
3253	case SYS_GETSOCKOPT:
3254	err =
3255	__sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3256	(int __user *)a[4]);
3257	break;
3258	case SYS_SENDMSG:
3259	err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3260	a[2], true);
3261	break;
3262	case SYS_SENDMMSG:
3263	err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3264	a[3], true);
3265	break;
3266	case SYS_RECVMSG:
3267	err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3268	a[2], true);
3269	break;
3270	case SYS_RECVMMSG:
3271	if (IS_ENABLED(CONFIG_64BIT))
3272	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3273	a[2], a[3],
3274	(struct __kernel_timespec __user *)a[4],
3275	NULL);
3276	else
3277	err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3278	a[2], a[3], NULL,
3279	(struct old_timespec32 __user *)a[4]);
3280	break;
3281	case SYS_ACCEPT4:
3282	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3283	(int __user *)a[2], a[3]);
3284	break;
3285	default:
3286	err = -EINVAL;
3287	break;
3288	}
3289	return err;
3290	}
3291
3292	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
3293
3294	/**
3295	* sock_register - add a socket protocol handler
3296	* @ops: description of protocol
3297	*
3298	* This function is called by a protocol handler that wants to
3299	* advertise its address family, and have it linked into the
3300	* socket interface. The value ops->family corresponds to the
3301	* socket system call protocol family.
3302	*/
3303	int sock_register(const struct net_proto_family *ops)
3304	{
3305	int err;
3306
3307	if (ops->family >= NPROTO) {
3308	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3309	return -ENOBUFS;
3310	}
3311
3312	spin_lock(&net_family_lock);
3313	if (rcu_dereference_protected(net_families[ops->family],
3314	lockdep_is_held(&net_family_lock)))
3315	err = -EEXIST;
3316	else {
3317	rcu_assign_pointer(net_families[ops->family], ops);
3318	err = 0;
3319	}
3320	spin_unlock(&net_family_lock);
3321
3322	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3323	return err;
3324	}
3325	EXPORT_SYMBOL(sock_register);
3326
3327	/**
3328	* sock_unregister - remove a protocol handler
3329	* @family: protocol family to remove
3330	*
3331	* This function is called by a protocol handler that wants to
3332	* remove its address family, and have it unlinked from the
3333	* new socket creation.
3334	*
3335	* If protocol handler is a module, then it can use module reference
3336	* counts to protect against new references. If protocol handler is not
3337	* a module then it needs to provide its own protection in
3338	* the ops->create routine.
3339	*/
3340	void sock_unregister(int family)
3341	{
3342	BUG_ON(family < 0 \|\| family >= NPROTO);
3343
3344	spin_lock(&net_family_lock);
3345	RCU_INIT_POINTER(net_families[family], NULL);
3346	spin_unlock(&net_family_lock);
3347
3348	synchronize_rcu();
3349
3350	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3351	}
3352	EXPORT_SYMBOL(sock_unregister);
3353
3354	bool sock_is_registered(int family)
3355	{
3356	return family < NPROTO && rcu_access_pointer(net_families[family]);
3357	}
3358
3359	static int __init sock_init(void)
3360	{
3361	int err;
3362	/*
3363	* Initialize the network sysctl infrastructure.
3364	*/
3365	err = net_sysctl_init();
3366	if (err)
3367	goto out;
3368
3369	/*
3370	* Initialize skbuff SLAB cache
3371	*/
3372	skb_init();
3373
3374	/*
3375	* Initialize the protocols module.
3376	*/
3377
3378	init_inodecache();
3379
3380	err = register_filesystem(&sock_fs_type);
3381	if (err)
3382	goto out;
3383	sock_mnt = kern_mount(&sock_fs_type);
3384	if (IS_ERR(sock_mnt)) {
3385	err = PTR_ERR(sock_mnt);
3386	goto out_mount;
3387	}
3388
3389	/* The real protocol initialization is performed in later initcalls.
3390	*/
3391
3392	#ifdef CONFIG_NETFILTER
3393	err = netfilter_init();
3394	if (err)
3395	goto out;
3396	#endif
3397
3398	ptp_classifier_init();
3399
3400	out:
3401	return err;
3402
3403	out_mount:
3404	unregister_filesystem(&sock_fs_type);
3405	goto out;
3406	}
3407
3408	core_initcall(sock_init); /* early initcall */
3409
3410	#ifdef CONFIG_PROC_FS
3411	void socket_seq_show(struct seq_file *seq)
3412	{
3413	seq_printf(seq, "sockets: used %d\n",
3414	sock_inuse_get(seq->private));
3415	}
3416	#endif /* CONFIG_PROC_FS */
3417
3418	/* Handle the fact that while struct ifreq has the same layout on
3419	* 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3420	* which are handled elsewhere, it still has different size due to
3421	* ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3422	* resulting in struct ifreq being 32 and 40 bytes respectively).
3423	* As a result, if the struct happens to be at the end of a page and
3424	* the next page isn't readable/writable, we get a fault. To prevent
3425	* that, copy back and forth to the full size.
3426	*/
3427	int get_user_ifreq(struct ifreq ifr, void __user ifrdata, void __user arg)
3428	{
3429	if (in_compat_syscall()) {
3430	struct compat_ifreq ifr32 = (struct compat_ifreq )ifr;
3431
3432	memset(ifr, 0, sizeof(*ifr));
3433	if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3434	return -EFAULT;
3435
3436	if (ifrdata)
3437	*ifrdata = compat_ptr(ifr32->ifr_data);
3438
3439	return 0;
3440	}
3441
3442	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3443	return -EFAULT;
3444
3445	if (ifrdata)
3446	*ifrdata = ifr->ifr_data;
3447
3448	return 0;
3449	}
3450	EXPORT_SYMBOL(get_user_ifreq);
3451
3452	int put_user_ifreq(struct ifreq ifr, void __user arg)
3453	{
3454	size_t size = sizeof(*ifr);
3455
3456	if (in_compat_syscall())
3457	size = sizeof(struct compat_ifreq);
3458
3459	if (copy_to_user(arg, ifr, size))
3460	return -EFAULT;
3461
3462	return 0;
3463	}
3464	EXPORT_SYMBOL(put_user_ifreq);
3465
3466	#ifdef CONFIG_COMPAT
3467	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
3468	{
3469	compat_uptr_t uptr32;
3470	struct ifreq ifr;
3471	void __user *saved;
3472	int err;
3473
3474	if (get_user_ifreq(&ifr, NULL, uifr32))
3475	return -EFAULT;
3476
3477	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3478	return -EFAULT;
3479
3480	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3481	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3482
3483	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3484	if (!err) {
3485	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3486	if (put_user_ifreq(&ifr, uifr32))
3487	err = -EFAULT;
3488	}
3489	return err;
3490	}
3491
3492	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3493	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3494	struct compat_ifreq __user *u_ifreq32)
3495	{
3496	struct ifreq ifreq;
3497	void __user *data;
3498
3499	if (!is_socket_ioctl_cmd(cmd))
3500	return -ENOTTY;
3501	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3502	return -EFAULT;
3503	ifreq.ifr_data = data;
3504
3505	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3506	}
3507
3508	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3509	unsigned int cmd, unsigned long arg)
3510	{
3511	void __user *argp = compat_ptr(arg);
3512	struct sock *sk = sock->sk;
3513	struct net *net = sock_net(sk);
3514	const struct proto_ops *ops;
3515
3516	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3517	return sock_ioctl(file, cmd, (unsigned long)argp);
3518
3519	switch (cmd) {
3520	case SIOCWANDEV:
3521	return compat_siocwandev(net, argp);
3522	case SIOCGSTAMP_OLD:
3523	case SIOCGSTAMPNS_OLD:
3524	ops = READ_ONCE(sock->ops);
3525	if (!ops->gettstamp)
3526	return -ENOIOCTLCMD;
3527	return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3528	!COMPAT_USE_64BIT_TIME);
3529
3530	case SIOCETHTOOL:
3531	case SIOCBONDSLAVEINFOQUERY:
3532	case SIOCBONDINFOQUERY:
3533	case SIOCSHWTSTAMP:
3534	case SIOCGHWTSTAMP:
3535	return compat_ifr_data_ioctl(net, cmd, argp);
3536
3537	case FIOSETOWN:
3538	case SIOCSPGRP:
3539	case FIOGETOWN:
3540	case SIOCGPGRP:
3541	case SIOCBRADDBR:
3542	case SIOCBRDELBR:
3543	case SIOCBRADDIF:
3544	case SIOCBRDELIF:
3545	case SIOCGIFVLAN:
3546	case SIOCSIFVLAN:
3547	case SIOCGSKNS:
3548	case SIOCGSTAMP_NEW:
3549	case SIOCGSTAMPNS_NEW:
3550	case SIOCGIFCONF:
3551	case SIOCSIFBR:
3552	case SIOCGIFBR:
3553	return sock_ioctl(file, cmd, arg);
3554
3555	case SIOCGIFFLAGS:
3556	case SIOCSIFFLAGS:
3557	case SIOCGIFMAP:
3558	case SIOCSIFMAP:
3559	case SIOCGIFMETRIC:
3560	case SIOCSIFMETRIC:
3561	case SIOCGIFMTU:
3562	case SIOCSIFMTU:
3563	case SIOCGIFMEM:
3564	case SIOCSIFMEM:
3565	case SIOCGIFHWADDR:
3566	case SIOCSIFHWADDR:
3567	case SIOCADDMULTI:
3568	case SIOCDELMULTI:
3569	case SIOCGIFINDEX:
3570	case SIOCGIFADDR:
3571	case SIOCSIFADDR:
3572	case SIOCSIFHWBROADCAST:
3573	case SIOCDIFADDR:
3574	case SIOCGIFBRDADDR:
3575	case SIOCSIFBRDADDR:
3576	case SIOCGIFDSTADDR:
3577	case SIOCSIFDSTADDR:
3578	case SIOCGIFNETMASK:
3579	case SIOCSIFNETMASK:
3580	case SIOCSIFPFLAGS:
3581	case SIOCGIFPFLAGS:
3582	case SIOCGIFTXQLEN:
3583	case SIOCSIFTXQLEN:
3584	case SIOCGIFNAME:
3585	case SIOCSIFNAME:
3586	case SIOCGMIIPHY:
3587	case SIOCGMIIREG:
3588	case SIOCSMIIREG:
3589	case SIOCBONDENSLAVE:
3590	case SIOCBONDRELEASE:
3591	case SIOCBONDSETHWADDR:
3592	case SIOCBONDCHANGEACTIVE:
3593	case SIOCSARP:
3594	case SIOCGARP:
3595	case SIOCDARP:
3596	case SIOCOUTQ:
3597	case SIOCOUTQNSD:
3598	case SIOCATMARK:
3599	return sock_do_ioctl(net, sock, cmd, arg);
3600	}
3601
3602	return -ENOIOCTLCMD;
3603	}
3604
3605	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3606	unsigned long arg)
3607	{
3608	struct socket *sock = file->private_data;
3609	const struct proto_ops *ops = READ_ONCE(sock->ops);
3610	int ret = -ENOIOCTLCMD;
3611	struct sock *sk;
3612	struct net *net;
3613
3614	sk = sock->sk;
3615	net = sock_net(sk);
3616
3617	if (ops->compat_ioctl)
3618	ret = ops->compat_ioctl(sock, cmd, arg);
3619
3620	if (ret == -ENOIOCTLCMD &&
3621	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3622	ret = compat_wext_handle_ioctl(net, cmd, arg);
3623
3624	if (ret == -ENOIOCTLCMD)
3625	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3626
3627	return ret;
3628	}
3629	#endif
3630
3631	/**
3632	* kernel_bind - bind an address to a socket (kernel space)
3633	* @sock: socket
3634	* @addr: address
3635	* @addrlen: length of address
3636	*
3637	* Returns 0 or an error.
3638	*/
3639
3640	int kernel_bind(struct socket sock, struct sockaddr_unsized addr, int addrlen)
3641	{
3642	struct sockaddr_storage address;
3643
3644	memcpy(&address, addr, addrlen);
3645
3646	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr_unsized *)&address,
3647	addrlen);
3648	}
3649	EXPORT_SYMBOL(kernel_bind);
3650
3651	/**
3652	* kernel_listen - move socket to listening state (kernel space)
3653	* @sock: socket
3654	* @backlog: pending connections queue size
3655	*
3656	* Returns 0 or an error.
3657	*/
3658
3659	int kernel_listen(struct socket *sock, int backlog)
3660	{
3661	return READ_ONCE(sock->ops)->listen(sock, backlog);
3662	}
3663	EXPORT_SYMBOL(kernel_listen);
3664
3665	/**
3666	* kernel_accept - accept a connection (kernel space)
3667	* @sock: listening socket
3668	* @newsock: new connected socket
3669	* @flags: flags
3670	*
3671	* @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3672	* If it fails, @newsock is guaranteed to be %NULL.
3673	* Returns 0 or an error.
3674	*/
3675
3676	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3677	{
3678	struct sock *sk = sock->sk;
3679	const struct proto_ops *ops = READ_ONCE(sock->ops);
3680	struct proto_accept_arg arg = {
3681	.flags = flags,
3682	.kern = true,
3683	};
3684	int err;
3685
3686	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3687	newsock);
3688	if (err < 0)
3689	goto done;
3690
3691	err = ops->accept(sock, *newsock, &arg);
3692	if (err < 0) {
3693	sock_release(*newsock);
3694	*newsock = NULL;
3695	goto done;
3696	}
3697
3698	(*newsock)->ops = ops;
3699	__module_get(ops->owner);
3700
3701	done:
3702	return err;
3703	}
3704	EXPORT_SYMBOL(kernel_accept);
3705
3706	/**
3707	* kernel_connect - connect a socket (kernel space)
3708	* @sock: socket
3709	* @addr: address
3710	* @addrlen: address length
3711	* @flags: flags (O_NONBLOCK, ...)
3712	*
3713	* For datagram sockets, @addr is the address to which datagrams are sent
3714	* by default, and the only address from which datagrams are received.
3715	* For stream sockets, attempts to connect to @addr.
3716	* Returns 0 or an error code.
3717	*/
3718
3719	int kernel_connect(struct socket sock, struct sockaddr_unsized addr, int addrlen,
3720	int flags)
3721	{
3722	struct sockaddr_storage address;
3723
3724	memcpy(&address, addr, addrlen);
3725
3726	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr_unsized *)&address,
3727	addrlen, flags);
3728	}
3729	EXPORT_SYMBOL(kernel_connect);
3730
3731	/**
3732	* kernel_getsockname - get the address which the socket is bound (kernel space)
3733	* @sock: socket
3734	* @addr: address holder
3735	*
3736	* Fills the @addr pointer with the address which the socket is bound.
3737	* Returns the length of the address in bytes or an error code.
3738	*/
3739
3740	int kernel_getsockname(struct socket sock, struct sockaddr addr)
3741	{
3742	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3743	}
3744	EXPORT_SYMBOL(kernel_getsockname);
3745
3746	/**
3747	* kernel_getpeername - get the address which the socket is connected (kernel space)
3748	* @sock: socket
3749	* @addr: address holder
3750	*
3751	* Fills the @addr pointer with the address which the socket is connected.
3752	* Returns the length of the address in bytes or an error code.
3753	*/
3754
3755	int kernel_getpeername(struct socket sock, struct sockaddr addr)
3756	{
3757	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3758	}
3759	EXPORT_SYMBOL(kernel_getpeername);
3760
3761	/**
3762	* kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3763	* @sock: socket
3764	* @how: connection part
3765	*
3766	* Returns 0 or an error.
3767	*/
3768
3769	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3770	{
3771	return READ_ONCE(sock->ops)->shutdown(sock, how);
3772	}
3773	EXPORT_SYMBOL(kernel_sock_shutdown);
3774
3775	/**
3776	* kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3777	* @sk: socket
3778	*
3779	* This routine returns the IP overhead imposed by a socket i.e.
3780	* the length of the underlying IP header, depending on whether
3781	* this is an IPv4 or IPv6 socket and the length from IP options turned
3782	* on at the socket. Assumes that the caller has a lock on the socket.
3783	*/
3784
3785	u32 kernel_sock_ip_overhead(struct sock *sk)
3786	{
3787	struct inet_sock *inet;
3788	struct ip_options_rcu *opt;
3789	u32 overhead = 0;
3790	#if IS_ENABLED(CONFIG_IPV6)
3791	struct ipv6_pinfo *np;
3792	struct ipv6_txoptions *optv6 = NULL;
3793	#endif /* IS_ENABLED(CONFIG_IPV6) */
3794
3795	if (!sk)
3796	return overhead;
3797
3798	switch (sk->sk_family) {
3799	case AF_INET:
3800	inet = inet_sk(sk);
3801	overhead += sizeof(struct iphdr);
3802	opt = rcu_dereference_protected(inet->inet_opt,
3803	sock_owned_by_user(sk));
3804	if (opt)
3805	overhead += opt->opt.optlen;
3806	return overhead;
3807	#if IS_ENABLED(CONFIG_IPV6)
3808	case AF_INET6:
3809	np = inet6_sk(sk);
3810	overhead += sizeof(struct ipv6hdr);
3811	if (np)
3812	optv6 = rcu_dereference_protected(np->opt,
3813	sock_owned_by_user(sk));
3814	if (optv6)
3815	overhead += (optv6->opt_flen + optv6->opt_nflen);
3816	return overhead;
3817	#endif /* IS_ENABLED(CONFIG_IPV6) */
3818	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3819	return overhead;
3820	}
3821	}
3822	EXPORT_SYMBOL(kernel_sock_ip_overhead);