xref: /linux/net/socket.c (revision c3b999cad7ec39a5487b20e8b6e737d2ab0c5393)
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, <obz@Kodak.COM>
8  *		Ross Biro
9  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10  *
11  * Fixes:
12  *		Anonymous	:	NOTSOCK/BADF cleanup. Error fix in
13  *					shutdown()
14  *		Alan Cox	:	verify_area() fixes
15  *		Alan Cox	:	Removed DDI
16  *		Jonathan Kamens	:	SOCK_DGRAM reconnect bug
17  *		Alan Cox	:	Moved a load of checks to the very
18  *					top level.
19  *		Alan Cox	:	Move address structures to/from user
20  *					mode above the protocol layers.
21  *		Rob Janssen	:	Allow 0 length sends.
22  *		Alan Cox	:	Asynchronous I/O support (cribbed from the
23  *					tty drivers).
24  *		Niibe Yutaka	:	Asynchronous I/O for writes (4.4BSD style)
25  *		Jeff Uphoff	:	Made max number of sockets command-line
26  *					configurable.
27  *		Matti Aarnio	:	Made the number of sockets dynamic,
28  *					to be allocated when needed, and mr.
29  *					Uphoff's max is used as max to be
30  *					allowed to allocate.
31  *		Linus		:	Argh. removed all the socket allocation
32  *					altogether: it's in the inode now.
33  *		Alan Cox	:	Made sock_alloc()/sock_release() public
34  *					for NetROM and future kernel nfsd type
35  *					stuff.
36  *		Alan Cox	:	sendmsg/recvmsg basics.
37  *		Tom Dyas	:	Export net symbols.
38  *		Marcin Dalecki	:	Fixed problems with CONFIG_NET="n".
39  *		Alan Cox	:	Added thread locking to sys_* calls
40  *					for sockets. May have errors at the
41  *					moment.
42  *		Kevin Buhr	:	Fixed the dumb errors in the above.
43  *		Andi Kleen	:	Some small cleanups, optimizations,
44  *					and fixed a copy_from_user() bug.
45  *		Tigran Aivazian	:	sys_send(args) calls sys_sendto(args, NULL, 0)
46  *		Tigran Aivazian	:	Made listen(2) backlog sanity checks
47  *					protocol-independent
48  *
49  *	This module is effectively the top level interface to the BSD socket
50  *	paradigm.
51  *
52  *	Based upon Swansea University Computer Society NET3.039
53  */
54 
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
57 #include <linux/mm.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/splice.h>
61 #include <linux/net.h>
62 #include <linux/interrupt.h>
63 #include <linux/thread_info.h>
64 #include <linux/rcupdate.h>
65 #include <linux/netdevice.h>
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68 #include <linux/mutex.h>
69 #include <linux/if_bridge.h>
70 #include <linux/if_vlan.h>
71 #include <linux/ptp_classify.h>
72 #include <linux/init.h>
73 #include <linux/poll.h>
74 #include <linux/cache.h>
75 #include <linux/module.h>
76 #include <linux/highmem.h>
77 #include <linux/mount.h>
78 #include <linux/pseudo_fs.h>
79 #include <linux/security.h>
80 #include <linux/syscalls.h>
81 #include <linux/compat.h>
82 #include <linux/kmod.h>
83 #include <linux/audit.h>
84 #include <linux/wireless.h>
85 #include <linux/nsproxy.h>
86 #include <linux/magic.h>
87 #include <linux/slab.h>
88 #include <linux/xattr.h>
89 #include <linux/nospec.h>
90 #include <linux/indirect_call_wrapper.h>
91 #include <linux/io_uring/net.h>
92 
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
95 
96 #include <net/compat.h>
97 #include <net/wext.h>
98 #include <net/cls_cgroup.h>
99 
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102 
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/termios.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 #include <linux/ptp_clock_kernel.h>
111 #include <trace/events/sock.h>
112 
113 #include "core/dev.h"
114 
115 #ifdef CONFIG_NET_RX_BUSY_POLL
116 unsigned int sysctl_net_busy_read __read_mostly;
117 unsigned int sysctl_net_busy_poll __read_mostly;
118 #endif
119 
120 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
121 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
122 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
123 
124 static int sock_close(struct inode *inode, struct file *file);
125 static __poll_t sock_poll(struct file *file,
126 			      struct poll_table_struct *wait);
127 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
128 #ifdef CONFIG_COMPAT
129 static long compat_sock_ioctl(struct file *file,
130 			      unsigned int cmd, unsigned long arg);
131 #endif
132 static int sock_fasync(int fd, struct file *filp, int on);
133 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
134 				struct pipe_inode_info *pipe, size_t len,
135 				unsigned int flags);
136 static void sock_splice_eof(struct file *file);
137 
138 #ifdef CONFIG_PROC_FS
139 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
140 {
141 	struct socket *sock = f->private_data;
142 	const struct proto_ops *ops = READ_ONCE(sock->ops);
143 
144 	if (ops->show_fdinfo)
145 		ops->show_fdinfo(m, sock);
146 }
147 #else
148 #define sock_show_fdinfo NULL
149 #endif
150 
151 /*
152  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
153  *	in the operation structures but are done directly via the socketcall() multiplexor.
154  */
155 
156 static const struct file_operations socket_file_ops = {
157 	.owner =	THIS_MODULE,
158 	.read_iter =	sock_read_iter,
159 	.write_iter =	sock_write_iter,
160 	.poll =		sock_poll,
161 	.unlocked_ioctl = sock_ioctl,
162 #ifdef CONFIG_COMPAT
163 	.compat_ioctl = compat_sock_ioctl,
164 #endif
165 	.uring_cmd =    io_uring_cmd_sock,
166 	.mmap =		sock_mmap,
167 	.release =	sock_close,
168 	.fasync =	sock_fasync,
169 	.splice_write = splice_to_socket,
170 	.splice_read =	sock_splice_read,
171 	.splice_eof =	sock_splice_eof,
172 	.show_fdinfo =	sock_show_fdinfo,
173 };
174 
175 static const char * const pf_family_names[] = {
176 	[PF_UNSPEC]	= "PF_UNSPEC",
177 	[PF_UNIX]	= "PF_UNIX/PF_LOCAL",
178 	[PF_INET]	= "PF_INET",
179 	[PF_AX25]	= "PF_AX25",
180 	[PF_IPX]	= "PF_IPX",
181 	[PF_APPLETALK]	= "PF_APPLETALK",
182 	[PF_NETROM]	= "PF_NETROM",
183 	[PF_BRIDGE]	= "PF_BRIDGE",
184 	[PF_ATMPVC]	= "PF_ATMPVC",
185 	[PF_X25]	= "PF_X25",
186 	[PF_INET6]	= "PF_INET6",
187 	[PF_ROSE]	= "PF_ROSE",
188 	[PF_DECnet]	= "PF_DECnet",
189 	[PF_NETBEUI]	= "PF_NETBEUI",
190 	[PF_SECURITY]	= "PF_SECURITY",
191 	[PF_KEY]	= "PF_KEY",
192 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
193 	[PF_PACKET]	= "PF_PACKET",
194 	[PF_ASH]	= "PF_ASH",
195 	[PF_ECONET]	= "PF_ECONET",
196 	[PF_ATMSVC]	= "PF_ATMSVC",
197 	[PF_RDS]	= "PF_RDS",
198 	[PF_SNA]	= "PF_SNA",
199 	[PF_IRDA]	= "PF_IRDA",
200 	[PF_PPPOX]	= "PF_PPPOX",
201 	[PF_WANPIPE]	= "PF_WANPIPE",
202 	[PF_LLC]	= "PF_LLC",
203 	[PF_IB]		= "PF_IB",
204 	[PF_MPLS]	= "PF_MPLS",
205 	[PF_CAN]	= "PF_CAN",
206 	[PF_TIPC]	= "PF_TIPC",
207 	[PF_BLUETOOTH]	= "PF_BLUETOOTH",
208 	[PF_IUCV]	= "PF_IUCV",
209 	[PF_RXRPC]	= "PF_RXRPC",
210 	[PF_ISDN]	= "PF_ISDN",
211 	[PF_PHONET]	= "PF_PHONET",
212 	[PF_IEEE802154]	= "PF_IEEE802154",
213 	[PF_CAIF]	= "PF_CAIF",
214 	[PF_ALG]	= "PF_ALG",
215 	[PF_NFC]	= "PF_NFC",
216 	[PF_VSOCK]	= "PF_VSOCK",
217 	[PF_KCM]	= "PF_KCM",
218 	[PF_QIPCRTR]	= "PF_QIPCRTR",
219 	[PF_SMC]	= "PF_SMC",
220 	[PF_XDP]	= "PF_XDP",
221 	[PF_MCTP]	= "PF_MCTP",
222 };
223 
224 /*
225  *	The protocol list. Each protocol is registered in here.
226  */
227 
228 static DEFINE_SPINLOCK(net_family_lock);
229 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
230 
231 /*
232  * Support routines.
233  * Move socket addresses back and forth across the kernel/user
234  * divide and look after the messy bits.
235  */
236 
237 /**
238  *	move_addr_to_kernel	-	copy a socket address into kernel space
239  *	@uaddr: Address in user space
240  *	@kaddr: Address in kernel space
241  *	@ulen: Length in user space
242  *
243  *	The address is copied into kernel space. If the provided address is
244  *	too long an error code of -EINVAL is returned. If the copy gives
245  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
246  */
247 
248 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
249 {
250 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
251 		return -EINVAL;
252 	if (ulen == 0)
253 		return 0;
254 	if (copy_from_user(kaddr, uaddr, ulen))
255 		return -EFAULT;
256 	return audit_sockaddr(ulen, kaddr);
257 }
258 
259 /**
260  *	move_addr_to_user	-	copy an address to user space
261  *	@kaddr: kernel space address
262  *	@klen: length of address in kernel
263  *	@uaddr: user space address
264  *	@ulen: pointer to user length field
265  *
266  *	The value pointed to by ulen on entry is the buffer length available.
267  *	This is overwritten with the buffer space used. -EINVAL is returned
268  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
269  *	is returned if either the buffer or the length field are not
270  *	accessible.
271  *	After copying the data up to the limit the user specifies, the true
272  *	length of the data is written over the length limit the user
273  *	specified. Zero is returned for a success.
274  */
275 
276 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
277 			     void __user *uaddr, int __user *ulen)
278 {
279 	int err;
280 	int len;
281 
282 	BUG_ON(klen > sizeof(struct sockaddr_storage));
283 	err = get_user(len, ulen);
284 	if (err)
285 		return err;
286 	if (len > klen)
287 		len = klen;
288 	if (len < 0)
289 		return -EINVAL;
290 	if (len) {
291 		if (audit_sockaddr(klen, kaddr))
292 			return -ENOMEM;
293 		if (copy_to_user(uaddr, kaddr, len))
294 			return -EFAULT;
295 	}
296 	/*
297 	 *      "fromlen shall refer to the value before truncation.."
298 	 *                      1003.1g
299 	 */
300 	return __put_user(klen, ulen);
301 }
302 
303 static struct kmem_cache *sock_inode_cachep __ro_after_init;
304 
305 static struct inode *sock_alloc_inode(struct super_block *sb)
306 {
307 	struct socket_alloc *ei;
308 
309 	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
310 	if (!ei)
311 		return NULL;
312 	init_waitqueue_head(&ei->socket.wq.wait);
313 	ei->socket.wq.fasync_list = NULL;
314 	ei->socket.wq.flags = 0;
315 
316 	ei->socket.state = SS_UNCONNECTED;
317 	ei->socket.flags = 0;
318 	ei->socket.ops = NULL;
319 	ei->socket.sk = NULL;
320 	ei->socket.file = NULL;
321 
322 	return &ei->vfs_inode;
323 }
324 
325 static void sock_free_inode(struct inode *inode)
326 {
327 	struct socket_alloc *ei;
328 
329 	ei = container_of(inode, struct socket_alloc, vfs_inode);
330 	kmem_cache_free(sock_inode_cachep, ei);
331 }
332 
333 static void init_once(void *foo)
334 {
335 	struct socket_alloc *ei = (struct socket_alloc *)foo;
336 
337 	inode_init_once(&ei->vfs_inode);
338 }
339 
340 static void init_inodecache(void)
341 {
342 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
343 					      sizeof(struct socket_alloc),
344 					      0,
345 					      (SLAB_HWCACHE_ALIGN |
346 					       SLAB_RECLAIM_ACCOUNT |
347 					       SLAB_ACCOUNT),
348 					      init_once);
349 	BUG_ON(sock_inode_cachep == NULL);
350 }
351 
352 static const struct super_operations sockfs_ops = {
353 	.alloc_inode	= sock_alloc_inode,
354 	.free_inode	= sock_free_inode,
355 	.statfs		= simple_statfs,
356 };
357 
358 /*
359  * sockfs_dname() is called from d_path().
360  */
361 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
362 {
363 	return dynamic_dname(buffer, buflen, "socket:[%lu]",
364 				d_inode(dentry)->i_ino);
365 }
366 
367 static const struct dentry_operations sockfs_dentry_operations = {
368 	.d_dname  = sockfs_dname,
369 };
370 
371 static int sockfs_xattr_get(const struct xattr_handler *handler,
372 			    struct dentry *dentry, struct inode *inode,
373 			    const char *suffix, void *value, size_t size)
374 {
375 	if (value) {
376 		if (dentry->d_name.len + 1 > size)
377 			return -ERANGE;
378 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
379 	}
380 	return dentry->d_name.len + 1;
381 }
382 
383 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
384 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
385 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
386 
387 static const struct xattr_handler sockfs_xattr_handler = {
388 	.name = XATTR_NAME_SOCKPROTONAME,
389 	.get = sockfs_xattr_get,
390 };
391 
392 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
393 				     struct mnt_idmap *idmap,
394 				     struct dentry *dentry, struct inode *inode,
395 				     const char *suffix, const void *value,
396 				     size_t size, int flags)
397 {
398 	/* Handled by LSM. */
399 	return -EAGAIN;
400 }
401 
402 static const struct xattr_handler sockfs_security_xattr_handler = {
403 	.prefix = XATTR_SECURITY_PREFIX,
404 	.set = sockfs_security_xattr_set,
405 };
406 
407 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
408 	&sockfs_xattr_handler,
409 	&sockfs_security_xattr_handler,
410 	NULL
411 };
412 
413 static int sockfs_init_fs_context(struct fs_context *fc)
414 {
415 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
416 	if (!ctx)
417 		return -ENOMEM;
418 	ctx->ops = &sockfs_ops;
419 	ctx->dops = &sockfs_dentry_operations;
420 	ctx->xattr = sockfs_xattr_handlers;
421 	return 0;
422 }
423 
424 static struct vfsmount *sock_mnt __read_mostly;
425 
426 static struct file_system_type sock_fs_type = {
427 	.name =		"sockfs",
428 	.init_fs_context = sockfs_init_fs_context,
429 	.kill_sb =	kill_anon_super,
430 };
431 
432 /*
433  *	Obtains the first available file descriptor and sets it up for use.
434  *
435  *	These functions create file structures and maps them to fd space
436  *	of the current process. On success it returns file descriptor
437  *	and file struct implicitly stored in sock->file.
438  *	Note that another thread may close file descriptor before we return
439  *	from this function. We use the fact that now we do not refer
440  *	to socket after mapping. If one day we will need it, this
441  *	function will increment ref. count on file by 1.
442  *
443  *	In any case returned fd MAY BE not valid!
444  *	This race condition is unavoidable
445  *	with shared fd spaces, we cannot solve it inside kernel,
446  *	but we take care of internal coherence yet.
447  */
448 
449 /**
450  *	sock_alloc_file - Bind a &socket to a &file
451  *	@sock: socket
452  *	@flags: file status flags
453  *	@dname: protocol name
454  *
455  *	Returns the &file bound with @sock, implicitly storing it
456  *	in sock->file. If dname is %NULL, sets to "".
457  *
458  *	On failure @sock is released, and an ERR pointer is returned.
459  *
460  *	This function uses GFP_KERNEL internally.
461  */
462 
463 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
464 {
465 	struct file *file;
466 
467 	if (!dname)
468 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
469 
470 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
471 				O_RDWR | (flags & O_NONBLOCK),
472 				&socket_file_ops);
473 	if (IS_ERR(file)) {
474 		sock_release(sock);
475 		return file;
476 	}
477 
478 	file->f_mode |= FMODE_NOWAIT;
479 	sock->file = file;
480 	file->private_data = sock;
481 	stream_open(SOCK_INODE(sock), file);
482 	/*
483 	 * Disable permission and pre-content events, but enable legacy
484 	 * inotify events for legacy users.
485 	 */
486 	file_set_fsnotify_mode(file, FMODE_NONOTIFY_PERM);
487 	return file;
488 }
489 EXPORT_SYMBOL(sock_alloc_file);
490 
491 static int sock_map_fd(struct socket *sock, int flags)
492 {
493 	struct file *newfile;
494 	int fd = get_unused_fd_flags(flags);
495 	if (unlikely(fd < 0)) {
496 		sock_release(sock);
497 		return fd;
498 	}
499 
500 	newfile = sock_alloc_file(sock, flags, NULL);
501 	if (!IS_ERR(newfile)) {
502 		fd_install(fd, newfile);
503 		return fd;
504 	}
505 
506 	put_unused_fd(fd);
507 	return PTR_ERR(newfile);
508 }
509 
510 /**
511  *	sock_from_file - Return the &socket bounded to @file.
512  *	@file: file
513  *
514  *	On failure returns %NULL.
515  */
516 
517 struct socket *sock_from_file(struct file *file)
518 {
519 	if (likely(file->f_op == &socket_file_ops))
520 		return file->private_data;	/* set in sock_alloc_file */
521 
522 	return NULL;
523 }
524 EXPORT_SYMBOL(sock_from_file);
525 
526 /**
527  *	sockfd_lookup - Go from a file number to its socket slot
528  *	@fd: file handle
529  *	@err: pointer to an error code return
530  *
531  *	The file handle passed in is locked and the socket it is bound
532  *	to is returned. If an error occurs the err pointer is overwritten
533  *	with a negative errno code and NULL is returned. The function checks
534  *	for both invalid handles and passing a handle which is not a socket.
535  *
536  *	On a success the socket object pointer is returned.
537  */
538 
539 struct socket *sockfd_lookup(int fd, int *err)
540 {
541 	struct file *file;
542 	struct socket *sock;
543 
544 	file = fget(fd);
545 	if (!file) {
546 		*err = -EBADF;
547 		return NULL;
548 	}
549 
550 	sock = sock_from_file(file);
551 	if (!sock) {
552 		*err = -ENOTSOCK;
553 		fput(file);
554 	}
555 	return sock;
556 }
557 EXPORT_SYMBOL(sockfd_lookup);
558 
559 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
560 				size_t size)
561 {
562 	ssize_t len;
563 	ssize_t used = 0;
564 
565 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
566 	if (len < 0)
567 		return len;
568 	used += len;
569 	if (buffer) {
570 		if (size < used)
571 			return -ERANGE;
572 		buffer += len;
573 	}
574 
575 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
576 	used += len;
577 	if (buffer) {
578 		if (size < used)
579 			return -ERANGE;
580 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
581 		buffer += len;
582 	}
583 
584 	return used;
585 }
586 
587 static int sockfs_setattr(struct mnt_idmap *idmap,
588 			  struct dentry *dentry, struct iattr *iattr)
589 {
590 	int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
591 
592 	if (!err && (iattr->ia_valid & ATTR_UID)) {
593 		struct socket *sock = SOCKET_I(d_inode(dentry));
594 
595 		if (sock->sk)
596 			sock->sk->sk_uid = iattr->ia_uid;
597 		else
598 			err = -ENOENT;
599 	}
600 
601 	return err;
602 }
603 
604 static const struct inode_operations sockfs_inode_ops = {
605 	.listxattr = sockfs_listxattr,
606 	.setattr = sockfs_setattr,
607 };
608 
609 /**
610  *	sock_alloc - allocate a socket
611  *
612  *	Allocate a new inode and socket object. The two are bound together
613  *	and initialised. The socket is then returned. If we are out of inodes
614  *	NULL is returned. This functions uses GFP_KERNEL internally.
615  */
616 
617 struct socket *sock_alloc(void)
618 {
619 	struct inode *inode;
620 	struct socket *sock;
621 
622 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
623 	if (!inode)
624 		return NULL;
625 
626 	sock = SOCKET_I(inode);
627 
628 	inode->i_ino = get_next_ino();
629 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
630 	inode->i_uid = current_fsuid();
631 	inode->i_gid = current_fsgid();
632 	inode->i_op = &sockfs_inode_ops;
633 
634 	return sock;
635 }
636 EXPORT_SYMBOL(sock_alloc);
637 
638 static void __sock_release(struct socket *sock, struct inode *inode)
639 {
640 	const struct proto_ops *ops = READ_ONCE(sock->ops);
641 
642 	if (ops) {
643 		struct module *owner = ops->owner;
644 
645 		if (inode)
646 			inode_lock(inode);
647 		ops->release(sock);
648 		sock->sk = NULL;
649 		if (inode)
650 			inode_unlock(inode);
651 		sock->ops = NULL;
652 		module_put(owner);
653 	}
654 
655 	if (sock->wq.fasync_list)
656 		pr_err("%s: fasync list not empty!\n", __func__);
657 
658 	if (!sock->file) {
659 		iput(SOCK_INODE(sock));
660 		return;
661 	}
662 	sock->file = NULL;
663 }
664 
665 /**
666  *	sock_release - close a socket
667  *	@sock: socket to close
668  *
669  *	The socket is released from the protocol stack if it has a release
670  *	callback, and the inode is then released if the socket is bound to
671  *	an inode not a file.
672  */
673 void sock_release(struct socket *sock)
674 {
675 	__sock_release(sock, NULL);
676 }
677 EXPORT_SYMBOL(sock_release);
678 
679 void __sock_tx_timestamp(__u32 tsflags, __u8 *tx_flags)
680 {
681 	u8 flags = *tx_flags;
682 
683 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
684 		flags |= SKBTX_HW_TSTAMP_NOBPF;
685 
686 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
687 		flags |= SKBTX_SW_TSTAMP;
688 
689 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
690 		flags |= SKBTX_SCHED_TSTAMP;
691 
692 	*tx_flags = flags;
693 }
694 EXPORT_SYMBOL(__sock_tx_timestamp);
695 
696 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
697 					   size_t));
698 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
699 					    size_t));
700 
701 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
702 						 int flags)
703 {
704 	trace_sock_send_length(sk, ret, 0);
705 }
706 
707 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
708 {
709 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
710 				     inet_sendmsg, sock, msg,
711 				     msg_data_left(msg));
712 	BUG_ON(ret == -EIOCBQUEUED);
713 
714 	if (trace_sock_send_length_enabled())
715 		call_trace_sock_send_length(sock->sk, ret, 0);
716 	return ret;
717 }
718 
719 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
720 {
721 	int err = security_socket_sendmsg(sock, msg,
722 					  msg_data_left(msg));
723 
724 	return err ?: sock_sendmsg_nosec(sock, msg);
725 }
726 
727 /**
728  *	sock_sendmsg - send a message through @sock
729  *	@sock: socket
730  *	@msg: message to send
731  *
732  *	Sends @msg through @sock, passing through LSM.
733  *	Returns the number of bytes sent, or an error code.
734  */
735 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
736 {
737 	struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
738 	struct sockaddr_storage address;
739 	int save_len = msg->msg_namelen;
740 	int ret;
741 
742 	if (msg->msg_name) {
743 		memcpy(&address, msg->msg_name, msg->msg_namelen);
744 		msg->msg_name = &address;
745 	}
746 
747 	ret = __sock_sendmsg(sock, msg);
748 	msg->msg_name = save_addr;
749 	msg->msg_namelen = save_len;
750 
751 	return ret;
752 }
753 EXPORT_SYMBOL(sock_sendmsg);
754 
755 /**
756  *	kernel_sendmsg - send a message through @sock (kernel-space)
757  *	@sock: socket
758  *	@msg: message header
759  *	@vec: kernel vec
760  *	@num: vec array length
761  *	@size: total message data size
762  *
763  *	Builds the message data with @vec and sends it through @sock.
764  *	Returns the number of bytes sent, or an error code.
765  */
766 
767 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
768 		   struct kvec *vec, size_t num, size_t size)
769 {
770 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
771 	return sock_sendmsg(sock, msg);
772 }
773 EXPORT_SYMBOL(kernel_sendmsg);
774 
775 static bool skb_is_err_queue(const struct sk_buff *skb)
776 {
777 	/* pkt_type of skbs enqueued on the error queue are set to
778 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
779 	 * in recvmsg, since skbs received on a local socket will never
780 	 * have a pkt_type of PACKET_OUTGOING.
781 	 */
782 	return skb->pkt_type == PACKET_OUTGOING;
783 }
784 
785 /* On transmit, software and hardware timestamps are returned independently.
786  * As the two skb clones share the hardware timestamp, which may be updated
787  * before the software timestamp is received, a hardware TX timestamp may be
788  * returned only if there is no software TX timestamp. Ignore false software
789  * timestamps, which may be made in the __sock_recv_timestamp() call when the
790  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
791  * hardware timestamp.
792  */
793 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
794 {
795 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
796 }
797 
798 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
799 {
800 	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
801 	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
802 	struct net_device *orig_dev;
803 	ktime_t hwtstamp;
804 
805 	rcu_read_lock();
806 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
807 	if (orig_dev) {
808 		*if_index = orig_dev->ifindex;
809 		hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
810 	} else {
811 		hwtstamp = shhwtstamps->hwtstamp;
812 	}
813 	rcu_read_unlock();
814 
815 	return hwtstamp;
816 }
817 
818 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
819 			   int if_index)
820 {
821 	struct scm_ts_pktinfo ts_pktinfo;
822 	struct net_device *orig_dev;
823 
824 	if (!skb_mac_header_was_set(skb))
825 		return;
826 
827 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
828 
829 	if (!if_index) {
830 		rcu_read_lock();
831 		orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
832 		if (orig_dev)
833 			if_index = orig_dev->ifindex;
834 		rcu_read_unlock();
835 	}
836 	ts_pktinfo.if_index = if_index;
837 
838 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
839 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
840 		 sizeof(ts_pktinfo), &ts_pktinfo);
841 }
842 
843 /*
844  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
845  */
846 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
847 	struct sk_buff *skb)
848 {
849 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
850 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
851 	struct scm_timestamping_internal tss;
852 	int empty = 1, false_tstamp = 0;
853 	struct skb_shared_hwtstamps *shhwtstamps =
854 		skb_hwtstamps(skb);
855 	int if_index;
856 	ktime_t hwtstamp;
857 	u32 tsflags;
858 
859 	/* Race occurred between timestamp enabling and packet
860 	   receiving.  Fill in the current time for now. */
861 	if (need_software_tstamp && skb->tstamp == 0) {
862 		__net_timestamp(skb);
863 		false_tstamp = 1;
864 	}
865 
866 	if (need_software_tstamp) {
867 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
868 			if (new_tstamp) {
869 				struct __kernel_sock_timeval tv;
870 
871 				skb_get_new_timestamp(skb, &tv);
872 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
873 					 sizeof(tv), &tv);
874 			} else {
875 				struct __kernel_old_timeval tv;
876 
877 				skb_get_timestamp(skb, &tv);
878 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
879 					 sizeof(tv), &tv);
880 			}
881 		} else {
882 			if (new_tstamp) {
883 				struct __kernel_timespec ts;
884 
885 				skb_get_new_timestampns(skb, &ts);
886 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
887 					 sizeof(ts), &ts);
888 			} else {
889 				struct __kernel_old_timespec ts;
890 
891 				skb_get_timestampns(skb, &ts);
892 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
893 					 sizeof(ts), &ts);
894 			}
895 		}
896 	}
897 
898 	memset(&tss, 0, sizeof(tss));
899 	tsflags = READ_ONCE(sk->sk_tsflags);
900 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
901 	     (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE ||
902 	      skb_is_err_queue(skb) ||
903 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
904 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
905 		empty = 0;
906 	if (shhwtstamps &&
907 	    (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
908 	     (tsflags & SOF_TIMESTAMPING_RX_HARDWARE ||
909 	      skb_is_err_queue(skb) ||
910 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
911 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
912 		if_index = 0;
913 		if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
914 			hwtstamp = get_timestamp(sk, skb, &if_index);
915 		else
916 			hwtstamp = shhwtstamps->hwtstamp;
917 
918 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
919 			hwtstamp = ptp_convert_timestamp(&hwtstamp,
920 							 READ_ONCE(sk->sk_bind_phc));
921 
922 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
923 			empty = 0;
924 
925 			if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
926 			    !skb_is_err_queue(skb))
927 				put_ts_pktinfo(msg, skb, if_index);
928 		}
929 	}
930 	if (!empty) {
931 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
932 			put_cmsg_scm_timestamping64(msg, &tss);
933 		else
934 			put_cmsg_scm_timestamping(msg, &tss);
935 
936 		if (skb_is_err_queue(skb) && skb->len &&
937 		    SKB_EXT_ERR(skb)->opt_stats)
938 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
939 				 skb->len, skb->data);
940 	}
941 }
942 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
943 
944 #ifdef CONFIG_WIRELESS
945 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
946 	struct sk_buff *skb)
947 {
948 	int ack;
949 
950 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
951 		return;
952 	if (!skb->wifi_acked_valid)
953 		return;
954 
955 	ack = skb->wifi_acked;
956 
957 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
958 }
959 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
960 #endif
961 
962 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
963 				   struct sk_buff *skb)
964 {
965 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
966 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
967 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
968 }
969 
970 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
971 			   struct sk_buff *skb)
972 {
973 	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
974 		/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
975 		__u32 mark = skb->mark;
976 
977 		put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
978 	}
979 }
980 
981 static void sock_recv_priority(struct msghdr *msg, struct sock *sk,
982 			       struct sk_buff *skb)
983 {
984 	if (sock_flag(sk, SOCK_RCVPRIORITY) && skb) {
985 		__u32 priority = skb->priority;
986 
987 		put_cmsg(msg, SOL_SOCKET, SO_PRIORITY, sizeof(__u32), &priority);
988 	}
989 }
990 
991 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
992 		       struct sk_buff *skb)
993 {
994 	sock_recv_timestamp(msg, sk, skb);
995 	sock_recv_drops(msg, sk, skb);
996 	sock_recv_mark(msg, sk, skb);
997 	sock_recv_priority(msg, sk, skb);
998 }
999 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1000 
1001 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1002 					   size_t, int));
1003 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1004 					    size_t, int));
1005 
1006 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1007 {
1008 	trace_sock_recv_length(sk, ret, flags);
1009 }
1010 
1011 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1012 				     int flags)
1013 {
1014 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1015 				     inet6_recvmsg,
1016 				     inet_recvmsg, sock, msg,
1017 				     msg_data_left(msg), flags);
1018 	if (trace_sock_recv_length_enabled())
1019 		call_trace_sock_recv_length(sock->sk, ret, flags);
1020 	return ret;
1021 }
1022 
1023 /**
1024  *	sock_recvmsg - receive a message from @sock
1025  *	@sock: socket
1026  *	@msg: message to receive
1027  *	@flags: message flags
1028  *
1029  *	Receives @msg from @sock, passing through LSM. Returns the total number
1030  *	of bytes received, or an error.
1031  */
1032 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1033 {
1034 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1035 
1036 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1037 }
1038 EXPORT_SYMBOL(sock_recvmsg);
1039 
1040 /**
1041  *	kernel_recvmsg - Receive a message from a socket (kernel space)
1042  *	@sock: The socket to receive the message from
1043  *	@msg: Received message
1044  *	@vec: Input s/g array for message data
1045  *	@num: Size of input s/g array
1046  *	@size: Number of bytes to read
1047  *	@flags: Message flags (MSG_DONTWAIT, etc...)
1048  *
1049  *	On return the msg structure contains the scatter/gather array passed in the
1050  *	vec argument. The array is modified so that it consists of the unfilled
1051  *	portion of the original array.
1052  *
1053  *	The returned value is the total number of bytes received, or an error.
1054  */
1055 
1056 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1057 		   struct kvec *vec, size_t num, size_t size, int flags)
1058 {
1059 	msg->msg_control_is_user = false;
1060 	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1061 	return sock_recvmsg(sock, msg, flags);
1062 }
1063 EXPORT_SYMBOL(kernel_recvmsg);
1064 
1065 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1066 				struct pipe_inode_info *pipe, size_t len,
1067 				unsigned int flags)
1068 {
1069 	struct socket *sock = file->private_data;
1070 	const struct proto_ops *ops;
1071 
1072 	ops = READ_ONCE(sock->ops);
1073 	if (unlikely(!ops->splice_read))
1074 		return copy_splice_read(file, ppos, pipe, len, flags);
1075 
1076 	return ops->splice_read(sock, ppos, pipe, len, flags);
1077 }
1078 
1079 static void sock_splice_eof(struct file *file)
1080 {
1081 	struct socket *sock = file->private_data;
1082 	const struct proto_ops *ops;
1083 
1084 	ops = READ_ONCE(sock->ops);
1085 	if (ops->splice_eof)
1086 		ops->splice_eof(sock);
1087 }
1088 
1089 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1090 {
1091 	struct file *file = iocb->ki_filp;
1092 	struct socket *sock = file->private_data;
1093 	struct msghdr msg = {.msg_iter = *to,
1094 			     .msg_iocb = iocb};
1095 	ssize_t res;
1096 
1097 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1098 		msg.msg_flags = MSG_DONTWAIT;
1099 
1100 	if (iocb->ki_pos != 0)
1101 		return -ESPIPE;
1102 
1103 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1104 		return 0;
1105 
1106 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1107 	*to = msg.msg_iter;
1108 	return res;
1109 }
1110 
1111 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1112 {
1113 	struct file *file = iocb->ki_filp;
1114 	struct socket *sock = file->private_data;
1115 	struct msghdr msg = {.msg_iter = *from,
1116 			     .msg_iocb = iocb};
1117 	ssize_t res;
1118 
1119 	if (iocb->ki_pos != 0)
1120 		return -ESPIPE;
1121 
1122 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1123 		msg.msg_flags = MSG_DONTWAIT;
1124 
1125 	if (sock->type == SOCK_SEQPACKET)
1126 		msg.msg_flags |= MSG_EOR;
1127 
1128 	res = __sock_sendmsg(sock, &msg);
1129 	*from = msg.msg_iter;
1130 	return res;
1131 }
1132 
1133 /*
1134  * Atomic setting of ioctl hooks to avoid race
1135  * with module unload.
1136  */
1137 
1138 static DEFINE_MUTEX(br_ioctl_mutex);
1139 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1140 			    unsigned int cmd, struct ifreq *ifr,
1141 			    void __user *uarg);
1142 
1143 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1144 			     unsigned int cmd, struct ifreq *ifr,
1145 			     void __user *uarg))
1146 {
1147 	mutex_lock(&br_ioctl_mutex);
1148 	br_ioctl_hook = hook;
1149 	mutex_unlock(&br_ioctl_mutex);
1150 }
1151 EXPORT_SYMBOL(brioctl_set);
1152 
1153 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1154 		  struct ifreq *ifr, void __user *uarg)
1155 {
1156 	int err = -ENOPKG;
1157 
1158 	if (!br_ioctl_hook)
1159 		request_module("bridge");
1160 
1161 	mutex_lock(&br_ioctl_mutex);
1162 	if (br_ioctl_hook)
1163 		err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1164 	mutex_unlock(&br_ioctl_mutex);
1165 
1166 	return err;
1167 }
1168 
1169 static DEFINE_MUTEX(vlan_ioctl_mutex);
1170 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1171 
1172 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1173 {
1174 	mutex_lock(&vlan_ioctl_mutex);
1175 	vlan_ioctl_hook = hook;
1176 	mutex_unlock(&vlan_ioctl_mutex);
1177 }
1178 EXPORT_SYMBOL(vlan_ioctl_set);
1179 
1180 static long sock_do_ioctl(struct net *net, struct socket *sock,
1181 			  unsigned int cmd, unsigned long arg)
1182 {
1183 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1184 	struct ifreq ifr;
1185 	bool need_copyout;
1186 	int err;
1187 	void __user *argp = (void __user *)arg;
1188 	void __user *data;
1189 
1190 	err = ops->ioctl(sock, cmd, arg);
1191 
1192 	/*
1193 	 * If this ioctl is unknown try to hand it down
1194 	 * to the NIC driver.
1195 	 */
1196 	if (err != -ENOIOCTLCMD)
1197 		return err;
1198 
1199 	if (!is_socket_ioctl_cmd(cmd))
1200 		return -ENOTTY;
1201 
1202 	if (get_user_ifreq(&ifr, &data, argp))
1203 		return -EFAULT;
1204 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1205 	if (!err && need_copyout)
1206 		if (put_user_ifreq(&ifr, argp))
1207 			return -EFAULT;
1208 
1209 	return err;
1210 }
1211 
1212 /*
1213  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1214  *	what to do with it - that's up to the protocol still.
1215  */
1216 
1217 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1218 {
1219 	const struct proto_ops  *ops;
1220 	struct socket *sock;
1221 	struct sock *sk;
1222 	void __user *argp = (void __user *)arg;
1223 	int pid, err;
1224 	struct net *net;
1225 
1226 	sock = file->private_data;
1227 	ops = READ_ONCE(sock->ops);
1228 	sk = sock->sk;
1229 	net = sock_net(sk);
1230 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1231 		struct ifreq ifr;
1232 		void __user *data;
1233 		bool need_copyout;
1234 		if (get_user_ifreq(&ifr, &data, argp))
1235 			return -EFAULT;
1236 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1237 		if (!err && need_copyout)
1238 			if (put_user_ifreq(&ifr, argp))
1239 				return -EFAULT;
1240 	} else
1241 #ifdef CONFIG_WEXT_CORE
1242 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1243 		err = wext_handle_ioctl(net, cmd, argp);
1244 	} else
1245 #endif
1246 		switch (cmd) {
1247 		case FIOSETOWN:
1248 		case SIOCSPGRP:
1249 			err = -EFAULT;
1250 			if (get_user(pid, (int __user *)argp))
1251 				break;
1252 			err = f_setown(sock->file, pid, 1);
1253 			break;
1254 		case FIOGETOWN:
1255 		case SIOCGPGRP:
1256 			err = put_user(f_getown(sock->file),
1257 				       (int __user *)argp);
1258 			break;
1259 		case SIOCGIFBR:
1260 		case SIOCSIFBR:
1261 		case SIOCBRADDBR:
1262 		case SIOCBRDELBR:
1263 			err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1264 			break;
1265 		case SIOCGIFVLAN:
1266 		case SIOCSIFVLAN:
1267 			err = -ENOPKG;
1268 			if (!vlan_ioctl_hook)
1269 				request_module("8021q");
1270 
1271 			mutex_lock(&vlan_ioctl_mutex);
1272 			if (vlan_ioctl_hook)
1273 				err = vlan_ioctl_hook(net, argp);
1274 			mutex_unlock(&vlan_ioctl_mutex);
1275 			break;
1276 		case SIOCGSKNS:
1277 			err = -EPERM;
1278 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1279 				break;
1280 
1281 			err = open_related_ns(&net->ns, get_net_ns);
1282 			break;
1283 		case SIOCGSTAMP_OLD:
1284 		case SIOCGSTAMPNS_OLD:
1285 			if (!ops->gettstamp) {
1286 				err = -ENOIOCTLCMD;
1287 				break;
1288 			}
1289 			err = ops->gettstamp(sock, argp,
1290 					     cmd == SIOCGSTAMP_OLD,
1291 					     !IS_ENABLED(CONFIG_64BIT));
1292 			break;
1293 		case SIOCGSTAMP_NEW:
1294 		case SIOCGSTAMPNS_NEW:
1295 			if (!ops->gettstamp) {
1296 				err = -ENOIOCTLCMD;
1297 				break;
1298 			}
1299 			err = ops->gettstamp(sock, argp,
1300 					     cmd == SIOCGSTAMP_NEW,
1301 					     false);
1302 			break;
1303 
1304 		case SIOCGIFCONF:
1305 			err = dev_ifconf(net, argp);
1306 			break;
1307 
1308 		default:
1309 			err = sock_do_ioctl(net, sock, cmd, arg);
1310 			break;
1311 		}
1312 	return err;
1313 }
1314 
1315 /**
1316  *	sock_create_lite - creates a socket
1317  *	@family: protocol family (AF_INET, ...)
1318  *	@type: communication type (SOCK_STREAM, ...)
1319  *	@protocol: protocol (0, ...)
1320  *	@res: new socket
1321  *
1322  *	Creates a new socket and assigns it to @res, passing through LSM.
1323  *	The new socket initialization is not complete, see kernel_accept().
1324  *	Returns 0 or an error. On failure @res is set to %NULL.
1325  *	This function internally uses GFP_KERNEL.
1326  */
1327 
1328 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1329 {
1330 	int err;
1331 	struct socket *sock = NULL;
1332 
1333 	err = security_socket_create(family, type, protocol, 1);
1334 	if (err)
1335 		goto out;
1336 
1337 	sock = sock_alloc();
1338 	if (!sock) {
1339 		err = -ENOMEM;
1340 		goto out;
1341 	}
1342 
1343 	sock->type = type;
1344 	err = security_socket_post_create(sock, family, type, protocol, 1);
1345 	if (err)
1346 		goto out_release;
1347 
1348 out:
1349 	*res = sock;
1350 	return err;
1351 out_release:
1352 	sock_release(sock);
1353 	sock = NULL;
1354 	goto out;
1355 }
1356 EXPORT_SYMBOL(sock_create_lite);
1357 
1358 /* No kernel lock held - perfect */
1359 static __poll_t sock_poll(struct file *file, poll_table *wait)
1360 {
1361 	struct socket *sock = file->private_data;
1362 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1363 	__poll_t events = poll_requested_events(wait), flag = 0;
1364 
1365 	if (!ops->poll)
1366 		return 0;
1367 
1368 	if (sk_can_busy_loop(sock->sk)) {
1369 		/* poll once if requested by the syscall */
1370 		if (events & POLL_BUSY_LOOP)
1371 			sk_busy_loop(sock->sk, 1);
1372 
1373 		/* if this socket can poll_ll, tell the system call */
1374 		flag = POLL_BUSY_LOOP;
1375 	}
1376 
1377 	return ops->poll(file, sock, wait) | flag;
1378 }
1379 
1380 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1381 {
1382 	struct socket *sock = file->private_data;
1383 
1384 	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1385 }
1386 
1387 static int sock_close(struct inode *inode, struct file *filp)
1388 {
1389 	__sock_release(SOCKET_I(inode), inode);
1390 	return 0;
1391 }
1392 
1393 /*
1394  *	Update the socket async list
1395  *
1396  *	Fasync_list locking strategy.
1397  *
1398  *	1. fasync_list is modified only under process context socket lock
1399  *	   i.e. under semaphore.
1400  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1401  *	   or under socket lock
1402  */
1403 
1404 static int sock_fasync(int fd, struct file *filp, int on)
1405 {
1406 	struct socket *sock = filp->private_data;
1407 	struct sock *sk = sock->sk;
1408 	struct socket_wq *wq = &sock->wq;
1409 
1410 	if (sk == NULL)
1411 		return -EINVAL;
1412 
1413 	lock_sock(sk);
1414 	fasync_helper(fd, filp, on, &wq->fasync_list);
1415 
1416 	if (!wq->fasync_list)
1417 		sock_reset_flag(sk, SOCK_FASYNC);
1418 	else
1419 		sock_set_flag(sk, SOCK_FASYNC);
1420 
1421 	release_sock(sk);
1422 	return 0;
1423 }
1424 
1425 /* This function may be called only under rcu_lock */
1426 
1427 int sock_wake_async(struct socket_wq *wq, int how, int band)
1428 {
1429 	if (!wq || !wq->fasync_list)
1430 		return -1;
1431 
1432 	switch (how) {
1433 	case SOCK_WAKE_WAITD:
1434 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1435 			break;
1436 		goto call_kill;
1437 	case SOCK_WAKE_SPACE:
1438 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1439 			break;
1440 		fallthrough;
1441 	case SOCK_WAKE_IO:
1442 call_kill:
1443 		kill_fasync(&wq->fasync_list, SIGIO, band);
1444 		break;
1445 	case SOCK_WAKE_URG:
1446 		kill_fasync(&wq->fasync_list, SIGURG, band);
1447 	}
1448 
1449 	return 0;
1450 }
1451 EXPORT_SYMBOL(sock_wake_async);
1452 
1453 /**
1454  *	__sock_create - creates a socket
1455  *	@net: net namespace
1456  *	@family: protocol family (AF_INET, ...)
1457  *	@type: communication type (SOCK_STREAM, ...)
1458  *	@protocol: protocol (0, ...)
1459  *	@res: new socket
1460  *	@kern: boolean for kernel space sockets
1461  *
1462  *	Creates a new socket and assigns it to @res, passing through LSM.
1463  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1464  *	be set to true if the socket resides in kernel space.
1465  *	This function internally uses GFP_KERNEL.
1466  */
1467 
1468 int __sock_create(struct net *net, int family, int type, int protocol,
1469 			 struct socket **res, int kern)
1470 {
1471 	int err;
1472 	struct socket *sock;
1473 	const struct net_proto_family *pf;
1474 
1475 	/*
1476 	 *      Check protocol is in range
1477 	 */
1478 	if (family < 0 || family >= NPROTO)
1479 		return -EAFNOSUPPORT;
1480 	if (type < 0 || type >= SOCK_MAX)
1481 		return -EINVAL;
1482 
1483 	/* Compatibility.
1484 
1485 	   This uglymoron is moved from INET layer to here to avoid
1486 	   deadlock in module load.
1487 	 */
1488 	if (family == PF_INET && type == SOCK_PACKET) {
1489 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1490 			     current->comm);
1491 		family = PF_PACKET;
1492 	}
1493 
1494 	err = security_socket_create(family, type, protocol, kern);
1495 	if (err)
1496 		return err;
1497 
1498 	/*
1499 	 *	Allocate the socket and allow the family to set things up. if
1500 	 *	the protocol is 0, the family is instructed to select an appropriate
1501 	 *	default.
1502 	 */
1503 	sock = sock_alloc();
1504 	if (!sock) {
1505 		net_warn_ratelimited("socket: no more sockets\n");
1506 		return -ENFILE;	/* Not exactly a match, but its the
1507 				   closest posix thing */
1508 	}
1509 
1510 	sock->type = type;
1511 
1512 #ifdef CONFIG_MODULES
1513 	/* Attempt to load a protocol module if the find failed.
1514 	 *
1515 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1516 	 * requested real, full-featured networking support upon configuration.
1517 	 * Otherwise module support will break!
1518 	 */
1519 	if (rcu_access_pointer(net_families[family]) == NULL)
1520 		request_module("net-pf-%d", family);
1521 #endif
1522 
1523 	rcu_read_lock();
1524 	pf = rcu_dereference(net_families[family]);
1525 	err = -EAFNOSUPPORT;
1526 	if (!pf)
1527 		goto out_release;
1528 
1529 	/*
1530 	 * We will call the ->create function, that possibly is in a loadable
1531 	 * module, so we have to bump that loadable module refcnt first.
1532 	 */
1533 	if (!try_module_get(pf->owner))
1534 		goto out_release;
1535 
1536 	/* Now protected by module ref count */
1537 	rcu_read_unlock();
1538 
1539 	err = pf->create(net, sock, protocol, kern);
1540 	if (err < 0) {
1541 		/* ->create should release the allocated sock->sk object on error
1542 		 * and make sure sock->sk is set to NULL to avoid use-after-free
1543 		 */
1544 		DEBUG_NET_WARN_ONCE(sock->sk,
1545 				    "%ps must clear sock->sk on failure, family: %d, type: %d, protocol: %d\n",
1546 				    pf->create, family, type, protocol);
1547 		goto out_module_put;
1548 	}
1549 
1550 	/*
1551 	 * Now to bump the refcnt of the [loadable] module that owns this
1552 	 * socket at sock_release time we decrement its refcnt.
1553 	 */
1554 	if (!try_module_get(sock->ops->owner))
1555 		goto out_module_busy;
1556 
1557 	/*
1558 	 * Now that we're done with the ->create function, the [loadable]
1559 	 * module can have its refcnt decremented
1560 	 */
1561 	module_put(pf->owner);
1562 	err = security_socket_post_create(sock, family, type, protocol, kern);
1563 	if (err)
1564 		goto out_sock_release;
1565 	*res = sock;
1566 
1567 	return 0;
1568 
1569 out_module_busy:
1570 	err = -EAFNOSUPPORT;
1571 out_module_put:
1572 	sock->ops = NULL;
1573 	module_put(pf->owner);
1574 out_sock_release:
1575 	sock_release(sock);
1576 	return err;
1577 
1578 out_release:
1579 	rcu_read_unlock();
1580 	goto out_sock_release;
1581 }
1582 EXPORT_SYMBOL(__sock_create);
1583 
1584 /**
1585  *	sock_create - creates a socket
1586  *	@family: protocol family (AF_INET, ...)
1587  *	@type: communication type (SOCK_STREAM, ...)
1588  *	@protocol: protocol (0, ...)
1589  *	@res: new socket
1590  *
1591  *	A wrapper around __sock_create().
1592  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1593  */
1594 
1595 int sock_create(int family, int type, int protocol, struct socket **res)
1596 {
1597 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1598 }
1599 EXPORT_SYMBOL(sock_create);
1600 
1601 /**
1602  *	sock_create_kern - creates a socket (kernel space)
1603  *	@net: net namespace
1604  *	@family: protocol family (AF_INET, ...)
1605  *	@type: communication type (SOCK_STREAM, ...)
1606  *	@protocol: protocol (0, ...)
1607  *	@res: new socket
1608  *
1609  *	A wrapper around __sock_create().
1610  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1611  */
1612 
1613 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1614 {
1615 	return __sock_create(net, family, type, protocol, res, 1);
1616 }
1617 EXPORT_SYMBOL(sock_create_kern);
1618 
1619 static struct socket *__sys_socket_create(int family, int type, int protocol)
1620 {
1621 	struct socket *sock;
1622 	int retval;
1623 
1624 	/* Check the SOCK_* constants for consistency.  */
1625 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1626 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1627 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1628 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1629 
1630 	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1631 		return ERR_PTR(-EINVAL);
1632 	type &= SOCK_TYPE_MASK;
1633 
1634 	retval = sock_create(family, type, protocol, &sock);
1635 	if (retval < 0)
1636 		return ERR_PTR(retval);
1637 
1638 	return sock;
1639 }
1640 
1641 struct file *__sys_socket_file(int family, int type, int protocol)
1642 {
1643 	struct socket *sock;
1644 	int flags;
1645 
1646 	sock = __sys_socket_create(family, type, protocol);
1647 	if (IS_ERR(sock))
1648 		return ERR_CAST(sock);
1649 
1650 	flags = type & ~SOCK_TYPE_MASK;
1651 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1652 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1653 
1654 	return sock_alloc_file(sock, flags, NULL);
1655 }
1656 
1657 /*	A hook for bpf progs to attach to and update socket protocol.
1658  *
1659  *	A static noinline declaration here could cause the compiler to
1660  *	optimize away the function. A global noinline declaration will
1661  *	keep the definition, but may optimize away the callsite.
1662  *	Therefore, __weak is needed to ensure that the call is still
1663  *	emitted, by telling the compiler that we don't know what the
1664  *	function might eventually be.
1665  */
1666 
1667 __bpf_hook_start();
1668 
1669 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1670 {
1671 	return protocol;
1672 }
1673 
1674 __bpf_hook_end();
1675 
1676 int __sys_socket(int family, int type, int protocol)
1677 {
1678 	struct socket *sock;
1679 	int flags;
1680 
1681 	sock = __sys_socket_create(family, type,
1682 				   update_socket_protocol(family, type, protocol));
1683 	if (IS_ERR(sock))
1684 		return PTR_ERR(sock);
1685 
1686 	flags = type & ~SOCK_TYPE_MASK;
1687 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1688 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1689 
1690 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1691 }
1692 
1693 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1694 {
1695 	return __sys_socket(family, type, protocol);
1696 }
1697 
1698 /*
1699  *	Create a pair of connected sockets.
1700  */
1701 
1702 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1703 {
1704 	struct socket *sock1, *sock2;
1705 	int fd1, fd2, err;
1706 	struct file *newfile1, *newfile2;
1707 	int flags;
1708 
1709 	flags = type & ~SOCK_TYPE_MASK;
1710 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1711 		return -EINVAL;
1712 	type &= SOCK_TYPE_MASK;
1713 
1714 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1715 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1716 
1717 	/*
1718 	 * reserve descriptors and make sure we won't fail
1719 	 * to return them to userland.
1720 	 */
1721 	fd1 = get_unused_fd_flags(flags);
1722 	if (unlikely(fd1 < 0))
1723 		return fd1;
1724 
1725 	fd2 = get_unused_fd_flags(flags);
1726 	if (unlikely(fd2 < 0)) {
1727 		put_unused_fd(fd1);
1728 		return fd2;
1729 	}
1730 
1731 	err = put_user(fd1, &usockvec[0]);
1732 	if (err)
1733 		goto out;
1734 
1735 	err = put_user(fd2, &usockvec[1]);
1736 	if (err)
1737 		goto out;
1738 
1739 	/*
1740 	 * Obtain the first socket and check if the underlying protocol
1741 	 * supports the socketpair call.
1742 	 */
1743 
1744 	err = sock_create(family, type, protocol, &sock1);
1745 	if (unlikely(err < 0))
1746 		goto out;
1747 
1748 	err = sock_create(family, type, protocol, &sock2);
1749 	if (unlikely(err < 0)) {
1750 		sock_release(sock1);
1751 		goto out;
1752 	}
1753 
1754 	err = security_socket_socketpair(sock1, sock2);
1755 	if (unlikely(err)) {
1756 		sock_release(sock2);
1757 		sock_release(sock1);
1758 		goto out;
1759 	}
1760 
1761 	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1762 	if (unlikely(err < 0)) {
1763 		sock_release(sock2);
1764 		sock_release(sock1);
1765 		goto out;
1766 	}
1767 
1768 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1769 	if (IS_ERR(newfile1)) {
1770 		err = PTR_ERR(newfile1);
1771 		sock_release(sock2);
1772 		goto out;
1773 	}
1774 
1775 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1776 	if (IS_ERR(newfile2)) {
1777 		err = PTR_ERR(newfile2);
1778 		fput(newfile1);
1779 		goto out;
1780 	}
1781 
1782 	audit_fd_pair(fd1, fd2);
1783 
1784 	fd_install(fd1, newfile1);
1785 	fd_install(fd2, newfile2);
1786 	return 0;
1787 
1788 out:
1789 	put_unused_fd(fd2);
1790 	put_unused_fd(fd1);
1791 	return err;
1792 }
1793 
1794 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1795 		int __user *, usockvec)
1796 {
1797 	return __sys_socketpair(family, type, protocol, usockvec);
1798 }
1799 
1800 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1801 		      int addrlen)
1802 {
1803 	int err;
1804 
1805 	err = security_socket_bind(sock, (struct sockaddr *)address,
1806 				   addrlen);
1807 	if (!err)
1808 		err = READ_ONCE(sock->ops)->bind(sock,
1809 						 (struct sockaddr *)address,
1810 						 addrlen);
1811 	return err;
1812 }
1813 
1814 /*
1815  *	Bind a name to a socket. Nothing much to do here since it's
1816  *	the protocol's responsibility to handle the local address.
1817  *
1818  *	We move the socket address to kernel space before we call
1819  *	the protocol layer (having also checked the address is ok).
1820  */
1821 
1822 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1823 {
1824 	struct socket *sock;
1825 	struct sockaddr_storage address;
1826 	CLASS(fd, f)(fd);
1827 	int err;
1828 
1829 	if (fd_empty(f))
1830 		return -EBADF;
1831 	sock = sock_from_file(fd_file(f));
1832 	if (unlikely(!sock))
1833 		return -ENOTSOCK;
1834 
1835 	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1836 	if (unlikely(err))
1837 		return err;
1838 
1839 	return __sys_bind_socket(sock, &address, addrlen);
1840 }
1841 
1842 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1843 {
1844 	return __sys_bind(fd, umyaddr, addrlen);
1845 }
1846 
1847 /*
1848  *	Perform a listen. Basically, we allow the protocol to do anything
1849  *	necessary for a listen, and if that works, we mark the socket as
1850  *	ready for listening.
1851  */
1852 int __sys_listen_socket(struct socket *sock, int backlog)
1853 {
1854 	int somaxconn, err;
1855 
1856 	somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1857 	if ((unsigned int)backlog > somaxconn)
1858 		backlog = somaxconn;
1859 
1860 	err = security_socket_listen(sock, backlog);
1861 	if (!err)
1862 		err = READ_ONCE(sock->ops)->listen(sock, backlog);
1863 	return err;
1864 }
1865 
1866 int __sys_listen(int fd, int backlog)
1867 {
1868 	CLASS(fd, f)(fd);
1869 	struct socket *sock;
1870 
1871 	if (fd_empty(f))
1872 		return -EBADF;
1873 	sock = sock_from_file(fd_file(f));
1874 	if (unlikely(!sock))
1875 		return -ENOTSOCK;
1876 
1877 	return __sys_listen_socket(sock, backlog);
1878 }
1879 
1880 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1881 {
1882 	return __sys_listen(fd, backlog);
1883 }
1884 
1885 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1886 		       struct sockaddr __user *upeer_sockaddr,
1887 		       int __user *upeer_addrlen, int flags)
1888 {
1889 	struct socket *sock, *newsock;
1890 	struct file *newfile;
1891 	int err, len;
1892 	struct sockaddr_storage address;
1893 	const struct proto_ops *ops;
1894 
1895 	sock = sock_from_file(file);
1896 	if (!sock)
1897 		return ERR_PTR(-ENOTSOCK);
1898 
1899 	newsock = sock_alloc();
1900 	if (!newsock)
1901 		return ERR_PTR(-ENFILE);
1902 	ops = READ_ONCE(sock->ops);
1903 
1904 	newsock->type = sock->type;
1905 	newsock->ops = ops;
1906 
1907 	/*
1908 	 * We don't need try_module_get here, as the listening socket (sock)
1909 	 * has the protocol module (sock->ops->owner) held.
1910 	 */
1911 	__module_get(ops->owner);
1912 
1913 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1914 	if (IS_ERR(newfile))
1915 		return newfile;
1916 
1917 	err = security_socket_accept(sock, newsock);
1918 	if (err)
1919 		goto out_fd;
1920 
1921 	arg->flags |= sock->file->f_flags;
1922 	err = ops->accept(sock, newsock, arg);
1923 	if (err < 0)
1924 		goto out_fd;
1925 
1926 	if (upeer_sockaddr) {
1927 		len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1928 		if (len < 0) {
1929 			err = -ECONNABORTED;
1930 			goto out_fd;
1931 		}
1932 		err = move_addr_to_user(&address,
1933 					len, upeer_sockaddr, upeer_addrlen);
1934 		if (err < 0)
1935 			goto out_fd;
1936 	}
1937 
1938 	/* File flags are not inherited via accept() unlike another OSes. */
1939 	return newfile;
1940 out_fd:
1941 	fput(newfile);
1942 	return ERR_PTR(err);
1943 }
1944 
1945 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1946 			      int __user *upeer_addrlen, int flags)
1947 {
1948 	struct proto_accept_arg arg = { };
1949 	struct file *newfile;
1950 	int newfd;
1951 
1952 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1953 		return -EINVAL;
1954 
1955 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1956 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1957 
1958 	newfd = get_unused_fd_flags(flags);
1959 	if (unlikely(newfd < 0))
1960 		return newfd;
1961 
1962 	newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1963 			    flags);
1964 	if (IS_ERR(newfile)) {
1965 		put_unused_fd(newfd);
1966 		return PTR_ERR(newfile);
1967 	}
1968 	fd_install(newfd, newfile);
1969 	return newfd;
1970 }
1971 
1972 /*
1973  *	For accept, we attempt to create a new socket, set up the link
1974  *	with the client, wake up the client, then return the new
1975  *	connected fd. We collect the address of the connector in kernel
1976  *	space and move it to user at the very end. This is unclean because
1977  *	we open the socket then return an error.
1978  *
1979  *	1003.1g adds the ability to recvmsg() to query connection pending
1980  *	status to recvmsg. We need to add that support in a way thats
1981  *	clean when we restructure accept also.
1982  */
1983 
1984 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1985 		  int __user *upeer_addrlen, int flags)
1986 {
1987 	CLASS(fd, f)(fd);
1988 
1989 	if (fd_empty(f))
1990 		return -EBADF;
1991 	return __sys_accept4_file(fd_file(f), upeer_sockaddr,
1992 					 upeer_addrlen, flags);
1993 }
1994 
1995 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1996 		int __user *, upeer_addrlen, int, flags)
1997 {
1998 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1999 }
2000 
2001 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2002 		int __user *, upeer_addrlen)
2003 {
2004 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2005 }
2006 
2007 /*
2008  *	Attempt to connect to a socket with the server address.  The address
2009  *	is in user space so we verify it is OK and move it to kernel space.
2010  *
2011  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2012  *	break bindings
2013  *
2014  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2015  *	other SEQPACKET protocols that take time to connect() as it doesn't
2016  *	include the -EINPROGRESS status for such sockets.
2017  */
2018 
2019 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2020 		       int addrlen, int file_flags)
2021 {
2022 	struct socket *sock;
2023 	int err;
2024 
2025 	sock = sock_from_file(file);
2026 	if (!sock) {
2027 		err = -ENOTSOCK;
2028 		goto out;
2029 	}
2030 
2031 	err =
2032 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2033 	if (err)
2034 		goto out;
2035 
2036 	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2037 				addrlen, sock->file->f_flags | file_flags);
2038 out:
2039 	return err;
2040 }
2041 
2042 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2043 {
2044 	struct sockaddr_storage address;
2045 	CLASS(fd, f)(fd);
2046 	int ret;
2047 
2048 	if (fd_empty(f))
2049 		return -EBADF;
2050 
2051 	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2052 	if (ret)
2053 		return ret;
2054 
2055 	return __sys_connect_file(fd_file(f), &address, addrlen, 0);
2056 }
2057 
2058 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2059 		int, addrlen)
2060 {
2061 	return __sys_connect(fd, uservaddr, addrlen);
2062 }
2063 
2064 /*
2065  *	Get the local address ('name') of a socket object. Move the obtained
2066  *	name to user space.
2067  */
2068 
2069 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2070 		      int __user *usockaddr_len)
2071 {
2072 	struct socket *sock;
2073 	struct sockaddr_storage address;
2074 	CLASS(fd, f)(fd);
2075 	int err;
2076 
2077 	if (fd_empty(f))
2078 		return -EBADF;
2079 	sock = sock_from_file(fd_file(f));
2080 	if (unlikely(!sock))
2081 		return -ENOTSOCK;
2082 
2083 	err = security_socket_getsockname(sock);
2084 	if (err)
2085 		return err;
2086 
2087 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2088 	if (err < 0)
2089 		return err;
2090 
2091 	/* "err" is actually length in this case */
2092 	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2093 }
2094 
2095 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2096 		int __user *, usockaddr_len)
2097 {
2098 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2099 }
2100 
2101 /*
2102  *	Get the remote address ('name') of a socket object. Move the obtained
2103  *	name to user space.
2104  */
2105 
2106 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2107 		      int __user *usockaddr_len)
2108 {
2109 	struct socket *sock;
2110 	struct sockaddr_storage address;
2111 	CLASS(fd, f)(fd);
2112 	int err;
2113 
2114 	if (fd_empty(f))
2115 		return -EBADF;
2116 	sock = sock_from_file(fd_file(f));
2117 	if (unlikely(!sock))
2118 		return -ENOTSOCK;
2119 
2120 	err = security_socket_getpeername(sock);
2121 	if (err)
2122 		return err;
2123 
2124 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 1);
2125 	if (err < 0)
2126 		return err;
2127 
2128 	/* "err" is actually length in this case */
2129 	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2130 }
2131 
2132 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2133 		int __user *, usockaddr_len)
2134 {
2135 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2136 }
2137 
2138 /*
2139  *	Send a datagram to a given address. We move the address into kernel
2140  *	space and check the user space data area is readable before invoking
2141  *	the protocol.
2142  */
2143 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2144 		 struct sockaddr __user *addr,  int addr_len)
2145 {
2146 	struct socket *sock;
2147 	struct sockaddr_storage address;
2148 	int err;
2149 	struct msghdr msg;
2150 
2151 	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2152 	if (unlikely(err))
2153 		return err;
2154 
2155 	CLASS(fd, f)(fd);
2156 	if (fd_empty(f))
2157 		return -EBADF;
2158 	sock = sock_from_file(fd_file(f));
2159 	if (unlikely(!sock))
2160 		return -ENOTSOCK;
2161 
2162 	msg.msg_name = NULL;
2163 	msg.msg_control = NULL;
2164 	msg.msg_controllen = 0;
2165 	msg.msg_namelen = 0;
2166 	msg.msg_ubuf = NULL;
2167 	if (addr) {
2168 		err = move_addr_to_kernel(addr, addr_len, &address);
2169 		if (err < 0)
2170 			return err;
2171 		msg.msg_name = (struct sockaddr *)&address;
2172 		msg.msg_namelen = addr_len;
2173 	}
2174 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2175 	if (sock->file->f_flags & O_NONBLOCK)
2176 		flags |= MSG_DONTWAIT;
2177 	msg.msg_flags = flags;
2178 	return __sock_sendmsg(sock, &msg);
2179 }
2180 
2181 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2182 		unsigned int, flags, struct sockaddr __user *, addr,
2183 		int, addr_len)
2184 {
2185 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2186 }
2187 
2188 /*
2189  *	Send a datagram down a socket.
2190  */
2191 
2192 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2193 		unsigned int, flags)
2194 {
2195 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2196 }
2197 
2198 /*
2199  *	Receive a frame from the socket and optionally record the address of the
2200  *	sender. We verify the buffers are writable and if needed move the
2201  *	sender address from kernel to user space.
2202  */
2203 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2204 		   struct sockaddr __user *addr, int __user *addr_len)
2205 {
2206 	struct sockaddr_storage address;
2207 	struct msghdr msg = {
2208 		/* Save some cycles and don't copy the address if not needed */
2209 		.msg_name = addr ? (struct sockaddr *)&address : NULL,
2210 	};
2211 	struct socket *sock;
2212 	int err, err2;
2213 
2214 	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2215 	if (unlikely(err))
2216 		return err;
2217 
2218 	CLASS(fd, f)(fd);
2219 
2220 	if (fd_empty(f))
2221 		return -EBADF;
2222 	sock = sock_from_file(fd_file(f));
2223 	if (unlikely(!sock))
2224 		return -ENOTSOCK;
2225 
2226 	if (sock->file->f_flags & O_NONBLOCK)
2227 		flags |= MSG_DONTWAIT;
2228 	err = sock_recvmsg(sock, &msg, flags);
2229 
2230 	if (err >= 0 && addr != NULL) {
2231 		err2 = move_addr_to_user(&address,
2232 					 msg.msg_namelen, addr, addr_len);
2233 		if (err2 < 0)
2234 			err = err2;
2235 	}
2236 	return err;
2237 }
2238 
2239 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2240 		unsigned int, flags, struct sockaddr __user *, addr,
2241 		int __user *, addr_len)
2242 {
2243 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2244 }
2245 
2246 /*
2247  *	Receive a datagram from a socket.
2248  */
2249 
2250 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2251 		unsigned int, flags)
2252 {
2253 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2254 }
2255 
2256 static bool sock_use_custom_sol_socket(const struct socket *sock)
2257 {
2258 	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2259 }
2260 
2261 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2262 		       int optname, sockptr_t optval, int optlen)
2263 {
2264 	const struct proto_ops *ops;
2265 	char *kernel_optval = NULL;
2266 	int err;
2267 
2268 	if (optlen < 0)
2269 		return -EINVAL;
2270 
2271 	err = security_socket_setsockopt(sock, level, optname);
2272 	if (err)
2273 		goto out_put;
2274 
2275 	if (!compat)
2276 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2277 						     optval, &optlen,
2278 						     &kernel_optval);
2279 	if (err < 0)
2280 		goto out_put;
2281 	if (err > 0) {
2282 		err = 0;
2283 		goto out_put;
2284 	}
2285 
2286 	if (kernel_optval)
2287 		optval = KERNEL_SOCKPTR(kernel_optval);
2288 	ops = READ_ONCE(sock->ops);
2289 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2290 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2291 	else if (unlikely(!ops->setsockopt))
2292 		err = -EOPNOTSUPP;
2293 	else
2294 		err = ops->setsockopt(sock, level, optname, optval,
2295 					    optlen);
2296 	kfree(kernel_optval);
2297 out_put:
2298 	return err;
2299 }
2300 EXPORT_SYMBOL(do_sock_setsockopt);
2301 
2302 /* Set a socket option. Because we don't know the option lengths we have
2303  * to pass the user mode parameter for the protocols to sort out.
2304  */
2305 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2306 		     int optlen)
2307 {
2308 	sockptr_t optval = USER_SOCKPTR(user_optval);
2309 	bool compat = in_compat_syscall();
2310 	struct socket *sock;
2311 	CLASS(fd, f)(fd);
2312 
2313 	if (fd_empty(f))
2314 		return -EBADF;
2315 	sock = sock_from_file(fd_file(f));
2316 	if (unlikely(!sock))
2317 		return -ENOTSOCK;
2318 
2319 	return do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2320 }
2321 
2322 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2323 		char __user *, optval, int, optlen)
2324 {
2325 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2326 }
2327 
2328 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2329 							 int optname));
2330 
2331 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2332 		       int optname, sockptr_t optval, sockptr_t optlen)
2333 {
2334 	int max_optlen __maybe_unused = 0;
2335 	const struct proto_ops *ops;
2336 	int err;
2337 
2338 	err = security_socket_getsockopt(sock, level, optname);
2339 	if (err)
2340 		return err;
2341 
2342 	if (!compat)
2343 		copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2344 
2345 	ops = READ_ONCE(sock->ops);
2346 	if (level == SOL_SOCKET) {
2347 		err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2348 	} else if (unlikely(!ops->getsockopt)) {
2349 		err = -EOPNOTSUPP;
2350 	} else {
2351 		if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2352 			      "Invalid argument type"))
2353 			return -EOPNOTSUPP;
2354 
2355 		err = ops->getsockopt(sock, level, optname, optval.user,
2356 				      optlen.user);
2357 	}
2358 
2359 	if (!compat)
2360 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2361 						     optval, optlen, max_optlen,
2362 						     err);
2363 
2364 	return err;
2365 }
2366 EXPORT_SYMBOL(do_sock_getsockopt);
2367 
2368 /*
2369  *	Get a socket option. Because we don't know the option lengths we have
2370  *	to pass a user mode parameter for the protocols to sort out.
2371  */
2372 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2373 		int __user *optlen)
2374 {
2375 	struct socket *sock;
2376 	CLASS(fd, f)(fd);
2377 
2378 	if (fd_empty(f))
2379 		return -EBADF;
2380 	sock = sock_from_file(fd_file(f));
2381 	if (unlikely(!sock))
2382 		return -ENOTSOCK;
2383 
2384 	return do_sock_getsockopt(sock, in_compat_syscall(), level, optname,
2385 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2386 }
2387 
2388 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2389 		char __user *, optval, int __user *, optlen)
2390 {
2391 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2392 }
2393 
2394 /*
2395  *	Shutdown a socket.
2396  */
2397 
2398 int __sys_shutdown_sock(struct socket *sock, int how)
2399 {
2400 	int err;
2401 
2402 	err = security_socket_shutdown(sock, how);
2403 	if (!err)
2404 		err = READ_ONCE(sock->ops)->shutdown(sock, how);
2405 
2406 	return err;
2407 }
2408 
2409 int __sys_shutdown(int fd, int how)
2410 {
2411 	struct socket *sock;
2412 	CLASS(fd, f)(fd);
2413 
2414 	if (fd_empty(f))
2415 		return -EBADF;
2416 	sock = sock_from_file(fd_file(f));
2417 	if (unlikely(!sock))
2418 		return -ENOTSOCK;
2419 
2420 	return __sys_shutdown_sock(sock, how);
2421 }
2422 
2423 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2424 {
2425 	return __sys_shutdown(fd, how);
2426 }
2427 
2428 /* A couple of helpful macros for getting the address of the 32/64 bit
2429  * fields which are the same type (int / unsigned) on our platforms.
2430  */
2431 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2432 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2433 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2434 
2435 struct used_address {
2436 	struct sockaddr_storage name;
2437 	unsigned int name_len;
2438 };
2439 
2440 int __copy_msghdr(struct msghdr *kmsg,
2441 		  struct user_msghdr *msg,
2442 		  struct sockaddr __user **save_addr)
2443 {
2444 	ssize_t err;
2445 
2446 	kmsg->msg_control_is_user = true;
2447 	kmsg->msg_get_inq = 0;
2448 	kmsg->msg_control_user = msg->msg_control;
2449 	kmsg->msg_controllen = msg->msg_controllen;
2450 	kmsg->msg_flags = msg->msg_flags;
2451 
2452 	kmsg->msg_namelen = msg->msg_namelen;
2453 	if (!msg->msg_name)
2454 		kmsg->msg_namelen = 0;
2455 
2456 	if (kmsg->msg_namelen < 0)
2457 		return -EINVAL;
2458 
2459 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2460 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2461 
2462 	if (save_addr)
2463 		*save_addr = msg->msg_name;
2464 
2465 	if (msg->msg_name && kmsg->msg_namelen) {
2466 		if (!save_addr) {
2467 			err = move_addr_to_kernel(msg->msg_name,
2468 						  kmsg->msg_namelen,
2469 						  kmsg->msg_name);
2470 			if (err < 0)
2471 				return err;
2472 		}
2473 	} else {
2474 		kmsg->msg_name = NULL;
2475 		kmsg->msg_namelen = 0;
2476 	}
2477 
2478 	if (msg->msg_iovlen > UIO_MAXIOV)
2479 		return -EMSGSIZE;
2480 
2481 	kmsg->msg_iocb = NULL;
2482 	kmsg->msg_ubuf = NULL;
2483 	return 0;
2484 }
2485 
2486 static int copy_msghdr_from_user(struct msghdr *kmsg,
2487 				 struct user_msghdr __user *umsg,
2488 				 struct sockaddr __user **save_addr,
2489 				 struct iovec **iov)
2490 {
2491 	struct user_msghdr msg;
2492 	ssize_t err;
2493 
2494 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2495 		return -EFAULT;
2496 
2497 	err = __copy_msghdr(kmsg, &msg, save_addr);
2498 	if (err)
2499 		return err;
2500 
2501 	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2502 			    msg.msg_iov, msg.msg_iovlen,
2503 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2504 	return err < 0 ? err : 0;
2505 }
2506 
2507 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2508 			   unsigned int flags, struct used_address *used_address,
2509 			   unsigned int allowed_msghdr_flags)
2510 {
2511 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2512 				__aligned(sizeof(__kernel_size_t));
2513 	/* 20 is size of ipv6_pktinfo */
2514 	unsigned char *ctl_buf = ctl;
2515 	int ctl_len;
2516 	ssize_t err;
2517 
2518 	err = -ENOBUFS;
2519 
2520 	if (msg_sys->msg_controllen > INT_MAX)
2521 		goto out;
2522 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2523 	ctl_len = msg_sys->msg_controllen;
2524 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2525 		err =
2526 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2527 						     sizeof(ctl));
2528 		if (err)
2529 			goto out;
2530 		ctl_buf = msg_sys->msg_control;
2531 		ctl_len = msg_sys->msg_controllen;
2532 	} else if (ctl_len) {
2533 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2534 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2535 		if (ctl_len > sizeof(ctl)) {
2536 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2537 			if (ctl_buf == NULL)
2538 				goto out;
2539 		}
2540 		err = -EFAULT;
2541 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2542 			goto out_freectl;
2543 		msg_sys->msg_control = ctl_buf;
2544 		msg_sys->msg_control_is_user = false;
2545 	}
2546 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2547 	msg_sys->msg_flags = flags;
2548 
2549 	if (sock->file->f_flags & O_NONBLOCK)
2550 		msg_sys->msg_flags |= MSG_DONTWAIT;
2551 	/*
2552 	 * If this is sendmmsg() and current destination address is same as
2553 	 * previously succeeded address, omit asking LSM's decision.
2554 	 * used_address->name_len is initialized to UINT_MAX so that the first
2555 	 * destination address never matches.
2556 	 */
2557 	if (used_address && msg_sys->msg_name &&
2558 	    used_address->name_len == msg_sys->msg_namelen &&
2559 	    !memcmp(&used_address->name, msg_sys->msg_name,
2560 		    used_address->name_len)) {
2561 		err = sock_sendmsg_nosec(sock, msg_sys);
2562 		goto out_freectl;
2563 	}
2564 	err = __sock_sendmsg(sock, msg_sys);
2565 	/*
2566 	 * If this is sendmmsg() and sending to current destination address was
2567 	 * successful, remember it.
2568 	 */
2569 	if (used_address && err >= 0) {
2570 		used_address->name_len = msg_sys->msg_namelen;
2571 		if (msg_sys->msg_name)
2572 			memcpy(&used_address->name, msg_sys->msg_name,
2573 			       used_address->name_len);
2574 	}
2575 
2576 out_freectl:
2577 	if (ctl_buf != ctl)
2578 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2579 out:
2580 	return err;
2581 }
2582 
2583 static int sendmsg_copy_msghdr(struct msghdr *msg,
2584 			       struct user_msghdr __user *umsg, unsigned flags,
2585 			       struct iovec **iov)
2586 {
2587 	int err;
2588 
2589 	if (flags & MSG_CMSG_COMPAT) {
2590 		struct compat_msghdr __user *msg_compat;
2591 
2592 		msg_compat = (struct compat_msghdr __user *) umsg;
2593 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2594 	} else {
2595 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2596 	}
2597 	if (err < 0)
2598 		return err;
2599 
2600 	return 0;
2601 }
2602 
2603 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2604 			 struct msghdr *msg_sys, unsigned int flags,
2605 			 struct used_address *used_address,
2606 			 unsigned int allowed_msghdr_flags)
2607 {
2608 	struct sockaddr_storage address;
2609 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2610 	ssize_t err;
2611 
2612 	msg_sys->msg_name = &address;
2613 
2614 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2615 	if (err < 0)
2616 		return err;
2617 
2618 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2619 				allowed_msghdr_flags);
2620 	kfree(iov);
2621 	return err;
2622 }
2623 
2624 /*
2625  *	BSD sendmsg interface
2626  */
2627 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2628 			unsigned int flags)
2629 {
2630 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2631 }
2632 
2633 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2634 		   bool forbid_cmsg_compat)
2635 {
2636 	struct msghdr msg_sys;
2637 	struct socket *sock;
2638 
2639 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2640 		return -EINVAL;
2641 
2642 	CLASS(fd, f)(fd);
2643 
2644 	if (fd_empty(f))
2645 		return -EBADF;
2646 	sock = sock_from_file(fd_file(f));
2647 	if (unlikely(!sock))
2648 		return -ENOTSOCK;
2649 
2650 	return ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2651 }
2652 
2653 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2654 {
2655 	return __sys_sendmsg(fd, msg, flags, true);
2656 }
2657 
2658 /*
2659  *	Linux sendmmsg interface
2660  */
2661 
2662 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2663 		   unsigned int flags, bool forbid_cmsg_compat)
2664 {
2665 	int err, datagrams;
2666 	struct socket *sock;
2667 	struct mmsghdr __user *entry;
2668 	struct compat_mmsghdr __user *compat_entry;
2669 	struct msghdr msg_sys;
2670 	struct used_address used_address;
2671 	unsigned int oflags = flags;
2672 
2673 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2674 		return -EINVAL;
2675 
2676 	if (vlen > UIO_MAXIOV)
2677 		vlen = UIO_MAXIOV;
2678 
2679 	datagrams = 0;
2680 
2681 	CLASS(fd, f)(fd);
2682 
2683 	if (fd_empty(f))
2684 		return -EBADF;
2685 	sock = sock_from_file(fd_file(f));
2686 	if (unlikely(!sock))
2687 		return -ENOTSOCK;
2688 
2689 	used_address.name_len = UINT_MAX;
2690 	entry = mmsg;
2691 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2692 	err = 0;
2693 	flags |= MSG_BATCH;
2694 
2695 	while (datagrams < vlen) {
2696 		if (datagrams == vlen - 1)
2697 			flags = oflags;
2698 
2699 		if (MSG_CMSG_COMPAT & flags) {
2700 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2701 					     &msg_sys, flags, &used_address, MSG_EOR);
2702 			if (err < 0)
2703 				break;
2704 			err = __put_user(err, &compat_entry->msg_len);
2705 			++compat_entry;
2706 		} else {
2707 			err = ___sys_sendmsg(sock,
2708 					     (struct user_msghdr __user *)entry,
2709 					     &msg_sys, flags, &used_address, MSG_EOR);
2710 			if (err < 0)
2711 				break;
2712 			err = put_user(err, &entry->msg_len);
2713 			++entry;
2714 		}
2715 
2716 		if (err)
2717 			break;
2718 		++datagrams;
2719 		if (msg_data_left(&msg_sys))
2720 			break;
2721 		cond_resched();
2722 	}
2723 
2724 	/* We only return an error if no datagrams were able to be sent */
2725 	if (datagrams != 0)
2726 		return datagrams;
2727 
2728 	return err;
2729 }
2730 
2731 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2732 		unsigned int, vlen, unsigned int, flags)
2733 {
2734 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2735 }
2736 
2737 static int recvmsg_copy_msghdr(struct msghdr *msg,
2738 			       struct user_msghdr __user *umsg, unsigned flags,
2739 			       struct sockaddr __user **uaddr,
2740 			       struct iovec **iov)
2741 {
2742 	ssize_t err;
2743 
2744 	if (MSG_CMSG_COMPAT & flags) {
2745 		struct compat_msghdr __user *msg_compat;
2746 
2747 		msg_compat = (struct compat_msghdr __user *) umsg;
2748 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2749 	} else {
2750 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2751 	}
2752 	if (err < 0)
2753 		return err;
2754 
2755 	return 0;
2756 }
2757 
2758 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2759 			   struct user_msghdr __user *msg,
2760 			   struct sockaddr __user *uaddr,
2761 			   unsigned int flags, int nosec)
2762 {
2763 	struct compat_msghdr __user *msg_compat =
2764 					(struct compat_msghdr __user *) msg;
2765 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2766 	struct sockaddr_storage addr;
2767 	unsigned long cmsg_ptr;
2768 	int len;
2769 	ssize_t err;
2770 
2771 	msg_sys->msg_name = &addr;
2772 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2773 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2774 
2775 	/* We assume all kernel code knows the size of sockaddr_storage */
2776 	msg_sys->msg_namelen = 0;
2777 
2778 	if (sock->file->f_flags & O_NONBLOCK)
2779 		flags |= MSG_DONTWAIT;
2780 
2781 	if (unlikely(nosec))
2782 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2783 	else
2784 		err = sock_recvmsg(sock, msg_sys, flags);
2785 
2786 	if (err < 0)
2787 		goto out;
2788 	len = err;
2789 
2790 	if (uaddr != NULL) {
2791 		err = move_addr_to_user(&addr,
2792 					msg_sys->msg_namelen, uaddr,
2793 					uaddr_len);
2794 		if (err < 0)
2795 			goto out;
2796 	}
2797 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2798 			 COMPAT_FLAGS(msg));
2799 	if (err)
2800 		goto out;
2801 	if (MSG_CMSG_COMPAT & flags)
2802 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2803 				 &msg_compat->msg_controllen);
2804 	else
2805 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2806 				 &msg->msg_controllen);
2807 	if (err)
2808 		goto out;
2809 	err = len;
2810 out:
2811 	return err;
2812 }
2813 
2814 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2815 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2816 {
2817 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2818 	/* user mode address pointers */
2819 	struct sockaddr __user *uaddr;
2820 	ssize_t err;
2821 
2822 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2823 	if (err < 0)
2824 		return err;
2825 
2826 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2827 	kfree(iov);
2828 	return err;
2829 }
2830 
2831 /*
2832  *	BSD recvmsg interface
2833  */
2834 
2835 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2836 			struct user_msghdr __user *umsg,
2837 			struct sockaddr __user *uaddr, unsigned int flags)
2838 {
2839 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2840 }
2841 
2842 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2843 		   bool forbid_cmsg_compat)
2844 {
2845 	struct msghdr msg_sys;
2846 	struct socket *sock;
2847 
2848 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2849 		return -EINVAL;
2850 
2851 	CLASS(fd, f)(fd);
2852 
2853 	if (fd_empty(f))
2854 		return -EBADF;
2855 	sock = sock_from_file(fd_file(f));
2856 	if (unlikely(!sock))
2857 		return -ENOTSOCK;
2858 
2859 	return ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2860 }
2861 
2862 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2863 		unsigned int, flags)
2864 {
2865 	return __sys_recvmsg(fd, msg, flags, true);
2866 }
2867 
2868 /*
2869  *     Linux recvmmsg interface
2870  */
2871 
2872 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2873 			  unsigned int vlen, unsigned int flags,
2874 			  struct timespec64 *timeout)
2875 {
2876 	int err = 0, datagrams;
2877 	struct socket *sock;
2878 	struct mmsghdr __user *entry;
2879 	struct compat_mmsghdr __user *compat_entry;
2880 	struct msghdr msg_sys;
2881 	struct timespec64 end_time;
2882 	struct timespec64 timeout64;
2883 
2884 	if (timeout &&
2885 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2886 				    timeout->tv_nsec))
2887 		return -EINVAL;
2888 
2889 	datagrams = 0;
2890 
2891 	CLASS(fd, f)(fd);
2892 
2893 	if (fd_empty(f))
2894 		return -EBADF;
2895 	sock = sock_from_file(fd_file(f));
2896 	if (unlikely(!sock))
2897 		return -ENOTSOCK;
2898 
2899 	if (likely(!(flags & MSG_ERRQUEUE))) {
2900 		err = sock_error(sock->sk);
2901 		if (err)
2902 			return err;
2903 	}
2904 
2905 	entry = mmsg;
2906 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2907 
2908 	while (datagrams < vlen) {
2909 		/*
2910 		 * No need to ask LSM for more than the first datagram.
2911 		 */
2912 		if (MSG_CMSG_COMPAT & flags) {
2913 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2914 					     &msg_sys, flags & ~MSG_WAITFORONE,
2915 					     datagrams);
2916 			if (err < 0)
2917 				break;
2918 			err = __put_user(err, &compat_entry->msg_len);
2919 			++compat_entry;
2920 		} else {
2921 			err = ___sys_recvmsg(sock,
2922 					     (struct user_msghdr __user *)entry,
2923 					     &msg_sys, flags & ~MSG_WAITFORONE,
2924 					     datagrams);
2925 			if (err < 0)
2926 				break;
2927 			err = put_user(err, &entry->msg_len);
2928 			++entry;
2929 		}
2930 
2931 		if (err)
2932 			break;
2933 		++datagrams;
2934 
2935 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2936 		if (flags & MSG_WAITFORONE)
2937 			flags |= MSG_DONTWAIT;
2938 
2939 		if (timeout) {
2940 			ktime_get_ts64(&timeout64);
2941 			*timeout = timespec64_sub(end_time, timeout64);
2942 			if (timeout->tv_sec < 0) {
2943 				timeout->tv_sec = timeout->tv_nsec = 0;
2944 				break;
2945 			}
2946 
2947 			/* Timeout, return less than vlen datagrams */
2948 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2949 				break;
2950 		}
2951 
2952 		/* Out of band data, return right away */
2953 		if (msg_sys.msg_flags & MSG_OOB)
2954 			break;
2955 		cond_resched();
2956 	}
2957 
2958 	if (err == 0)
2959 		return datagrams;
2960 
2961 	if (datagrams == 0)
2962 		return err;
2963 
2964 	/*
2965 	 * We may return less entries than requested (vlen) if the
2966 	 * sock is non block and there aren't enough datagrams...
2967 	 */
2968 	if (err != -EAGAIN) {
2969 		/*
2970 		 * ... or  if recvmsg returns an error after we
2971 		 * received some datagrams, where we record the
2972 		 * error to return on the next call or if the
2973 		 * app asks about it using getsockopt(SO_ERROR).
2974 		 */
2975 		WRITE_ONCE(sock->sk->sk_err, -err);
2976 	}
2977 	return datagrams;
2978 }
2979 
2980 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2981 		   unsigned int vlen, unsigned int flags,
2982 		   struct __kernel_timespec __user *timeout,
2983 		   struct old_timespec32 __user *timeout32)
2984 {
2985 	int datagrams;
2986 	struct timespec64 timeout_sys;
2987 
2988 	if (timeout && get_timespec64(&timeout_sys, timeout))
2989 		return -EFAULT;
2990 
2991 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2992 		return -EFAULT;
2993 
2994 	if (!timeout && !timeout32)
2995 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2996 
2997 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2998 
2999 	if (datagrams <= 0)
3000 		return datagrams;
3001 
3002 	if (timeout && put_timespec64(&timeout_sys, timeout))
3003 		datagrams = -EFAULT;
3004 
3005 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3006 		datagrams = -EFAULT;
3007 
3008 	return datagrams;
3009 }
3010 
3011 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3012 		unsigned int, vlen, unsigned int, flags,
3013 		struct __kernel_timespec __user *, timeout)
3014 {
3015 	if (flags & MSG_CMSG_COMPAT)
3016 		return -EINVAL;
3017 
3018 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3019 }
3020 
3021 #ifdef CONFIG_COMPAT_32BIT_TIME
3022 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3023 		unsigned int, vlen, unsigned int, flags,
3024 		struct old_timespec32 __user *, timeout)
3025 {
3026 	if (flags & MSG_CMSG_COMPAT)
3027 		return -EINVAL;
3028 
3029 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3030 }
3031 #endif
3032 
3033 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3034 /* Argument list sizes for sys_socketcall */
3035 #define AL(x) ((x) * sizeof(unsigned long))
3036 static const unsigned char nargs[21] = {
3037 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3038 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3039 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3040 	AL(4), AL(5), AL(4)
3041 };
3042 
3043 #undef AL
3044 
3045 /*
3046  *	System call vectors.
3047  *
3048  *	Argument checking cleaned up. Saved 20% in size.
3049  *  This function doesn't need to set the kernel lock because
3050  *  it is set by the callees.
3051  */
3052 
3053 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3054 {
3055 	unsigned long a[AUDITSC_ARGS];
3056 	unsigned long a0, a1;
3057 	int err;
3058 	unsigned int len;
3059 
3060 	if (call < 1 || call > SYS_SENDMMSG)
3061 		return -EINVAL;
3062 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3063 
3064 	len = nargs[call];
3065 	if (len > sizeof(a))
3066 		return -EINVAL;
3067 
3068 	/* copy_from_user should be SMP safe. */
3069 	if (copy_from_user(a, args, len))
3070 		return -EFAULT;
3071 
3072 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3073 	if (err)
3074 		return err;
3075 
3076 	a0 = a[0];
3077 	a1 = a[1];
3078 
3079 	switch (call) {
3080 	case SYS_SOCKET:
3081 		err = __sys_socket(a0, a1, a[2]);
3082 		break;
3083 	case SYS_BIND:
3084 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3085 		break;
3086 	case SYS_CONNECT:
3087 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3088 		break;
3089 	case SYS_LISTEN:
3090 		err = __sys_listen(a0, a1);
3091 		break;
3092 	case SYS_ACCEPT:
3093 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3094 				    (int __user *)a[2], 0);
3095 		break;
3096 	case SYS_GETSOCKNAME:
3097 		err =
3098 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
3099 				      (int __user *)a[2]);
3100 		break;
3101 	case SYS_GETPEERNAME:
3102 		err =
3103 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
3104 				      (int __user *)a[2]);
3105 		break;
3106 	case SYS_SOCKETPAIR:
3107 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3108 		break;
3109 	case SYS_SEND:
3110 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3111 				   NULL, 0);
3112 		break;
3113 	case SYS_SENDTO:
3114 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3115 				   (struct sockaddr __user *)a[4], a[5]);
3116 		break;
3117 	case SYS_RECV:
3118 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3119 				     NULL, NULL);
3120 		break;
3121 	case SYS_RECVFROM:
3122 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3123 				     (struct sockaddr __user *)a[4],
3124 				     (int __user *)a[5]);
3125 		break;
3126 	case SYS_SHUTDOWN:
3127 		err = __sys_shutdown(a0, a1);
3128 		break;
3129 	case SYS_SETSOCKOPT:
3130 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3131 				       a[4]);
3132 		break;
3133 	case SYS_GETSOCKOPT:
3134 		err =
3135 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3136 				     (int __user *)a[4]);
3137 		break;
3138 	case SYS_SENDMSG:
3139 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3140 				    a[2], true);
3141 		break;
3142 	case SYS_SENDMMSG:
3143 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3144 				     a[3], true);
3145 		break;
3146 	case SYS_RECVMSG:
3147 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3148 				    a[2], true);
3149 		break;
3150 	case SYS_RECVMMSG:
3151 		if (IS_ENABLED(CONFIG_64BIT))
3152 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3153 					     a[2], a[3],
3154 					     (struct __kernel_timespec __user *)a[4],
3155 					     NULL);
3156 		else
3157 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3158 					     a[2], a[3], NULL,
3159 					     (struct old_timespec32 __user *)a[4]);
3160 		break;
3161 	case SYS_ACCEPT4:
3162 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3163 				    (int __user *)a[2], a[3]);
3164 		break;
3165 	default:
3166 		err = -EINVAL;
3167 		break;
3168 	}
3169 	return err;
3170 }
3171 
3172 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3173 
3174 /**
3175  *	sock_register - add a socket protocol handler
3176  *	@ops: description of protocol
3177  *
3178  *	This function is called by a protocol handler that wants to
3179  *	advertise its address family, and have it linked into the
3180  *	socket interface. The value ops->family corresponds to the
3181  *	socket system call protocol family.
3182  */
3183 int sock_register(const struct net_proto_family *ops)
3184 {
3185 	int err;
3186 
3187 	if (ops->family >= NPROTO) {
3188 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3189 		return -ENOBUFS;
3190 	}
3191 
3192 	spin_lock(&net_family_lock);
3193 	if (rcu_dereference_protected(net_families[ops->family],
3194 				      lockdep_is_held(&net_family_lock)))
3195 		err = -EEXIST;
3196 	else {
3197 		rcu_assign_pointer(net_families[ops->family], ops);
3198 		err = 0;
3199 	}
3200 	spin_unlock(&net_family_lock);
3201 
3202 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3203 	return err;
3204 }
3205 EXPORT_SYMBOL(sock_register);
3206 
3207 /**
3208  *	sock_unregister - remove a protocol handler
3209  *	@family: protocol family to remove
3210  *
3211  *	This function is called by a protocol handler that wants to
3212  *	remove its address family, and have it unlinked from the
3213  *	new socket creation.
3214  *
3215  *	If protocol handler is a module, then it can use module reference
3216  *	counts to protect against new references. If protocol handler is not
3217  *	a module then it needs to provide its own protection in
3218  *	the ops->create routine.
3219  */
3220 void sock_unregister(int family)
3221 {
3222 	BUG_ON(family < 0 || family >= NPROTO);
3223 
3224 	spin_lock(&net_family_lock);
3225 	RCU_INIT_POINTER(net_families[family], NULL);
3226 	spin_unlock(&net_family_lock);
3227 
3228 	synchronize_rcu();
3229 
3230 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3231 }
3232 EXPORT_SYMBOL(sock_unregister);
3233 
3234 bool sock_is_registered(int family)
3235 {
3236 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3237 }
3238 
3239 static int __init sock_init(void)
3240 {
3241 	int err;
3242 	/*
3243 	 *      Initialize the network sysctl infrastructure.
3244 	 */
3245 	err = net_sysctl_init();
3246 	if (err)
3247 		goto out;
3248 
3249 	/*
3250 	 *      Initialize skbuff SLAB cache
3251 	 */
3252 	skb_init();
3253 
3254 	/*
3255 	 *      Initialize the protocols module.
3256 	 */
3257 
3258 	init_inodecache();
3259 
3260 	err = register_filesystem(&sock_fs_type);
3261 	if (err)
3262 		goto out;
3263 	sock_mnt = kern_mount(&sock_fs_type);
3264 	if (IS_ERR(sock_mnt)) {
3265 		err = PTR_ERR(sock_mnt);
3266 		goto out_mount;
3267 	}
3268 
3269 	/* The real protocol initialization is performed in later initcalls.
3270 	 */
3271 
3272 #ifdef CONFIG_NETFILTER
3273 	err = netfilter_init();
3274 	if (err)
3275 		goto out;
3276 #endif
3277 
3278 	ptp_classifier_init();
3279 
3280 out:
3281 	return err;
3282 
3283 out_mount:
3284 	unregister_filesystem(&sock_fs_type);
3285 	goto out;
3286 }
3287 
3288 core_initcall(sock_init);	/* early initcall */
3289 
3290 #ifdef CONFIG_PROC_FS
3291 void socket_seq_show(struct seq_file *seq)
3292 {
3293 	seq_printf(seq, "sockets: used %d\n",
3294 		   sock_inuse_get(seq->private));
3295 }
3296 #endif				/* CONFIG_PROC_FS */
3297 
3298 /* Handle the fact that while struct ifreq has the same *layout* on
3299  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3300  * which are handled elsewhere, it still has different *size* due to
3301  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3302  * resulting in struct ifreq being 32 and 40 bytes respectively).
3303  * As a result, if the struct happens to be at the end of a page and
3304  * the next page isn't readable/writable, we get a fault. To prevent
3305  * that, copy back and forth to the full size.
3306  */
3307 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3308 {
3309 	if (in_compat_syscall()) {
3310 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3311 
3312 		memset(ifr, 0, sizeof(*ifr));
3313 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3314 			return -EFAULT;
3315 
3316 		if (ifrdata)
3317 			*ifrdata = compat_ptr(ifr32->ifr_data);
3318 
3319 		return 0;
3320 	}
3321 
3322 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3323 		return -EFAULT;
3324 
3325 	if (ifrdata)
3326 		*ifrdata = ifr->ifr_data;
3327 
3328 	return 0;
3329 }
3330 EXPORT_SYMBOL(get_user_ifreq);
3331 
3332 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3333 {
3334 	size_t size = sizeof(*ifr);
3335 
3336 	if (in_compat_syscall())
3337 		size = sizeof(struct compat_ifreq);
3338 
3339 	if (copy_to_user(arg, ifr, size))
3340 		return -EFAULT;
3341 
3342 	return 0;
3343 }
3344 EXPORT_SYMBOL(put_user_ifreq);
3345 
3346 #ifdef CONFIG_COMPAT
3347 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3348 {
3349 	compat_uptr_t uptr32;
3350 	struct ifreq ifr;
3351 	void __user *saved;
3352 	int err;
3353 
3354 	if (get_user_ifreq(&ifr, NULL, uifr32))
3355 		return -EFAULT;
3356 
3357 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3358 		return -EFAULT;
3359 
3360 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3361 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3362 
3363 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3364 	if (!err) {
3365 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3366 		if (put_user_ifreq(&ifr, uifr32))
3367 			err = -EFAULT;
3368 	}
3369 	return err;
3370 }
3371 
3372 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3373 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3374 				 struct compat_ifreq __user *u_ifreq32)
3375 {
3376 	struct ifreq ifreq;
3377 	void __user *data;
3378 
3379 	if (!is_socket_ioctl_cmd(cmd))
3380 		return -ENOTTY;
3381 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3382 		return -EFAULT;
3383 	ifreq.ifr_data = data;
3384 
3385 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3386 }
3387 
3388 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3389 			 unsigned int cmd, unsigned long arg)
3390 {
3391 	void __user *argp = compat_ptr(arg);
3392 	struct sock *sk = sock->sk;
3393 	struct net *net = sock_net(sk);
3394 	const struct proto_ops *ops;
3395 
3396 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3397 		return sock_ioctl(file, cmd, (unsigned long)argp);
3398 
3399 	switch (cmd) {
3400 	case SIOCWANDEV:
3401 		return compat_siocwandev(net, argp);
3402 	case SIOCGSTAMP_OLD:
3403 	case SIOCGSTAMPNS_OLD:
3404 		ops = READ_ONCE(sock->ops);
3405 		if (!ops->gettstamp)
3406 			return -ENOIOCTLCMD;
3407 		return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3408 				      !COMPAT_USE_64BIT_TIME);
3409 
3410 	case SIOCETHTOOL:
3411 	case SIOCBONDSLAVEINFOQUERY:
3412 	case SIOCBONDINFOQUERY:
3413 	case SIOCSHWTSTAMP:
3414 	case SIOCGHWTSTAMP:
3415 		return compat_ifr_data_ioctl(net, cmd, argp);
3416 
3417 	case FIOSETOWN:
3418 	case SIOCSPGRP:
3419 	case FIOGETOWN:
3420 	case SIOCGPGRP:
3421 	case SIOCBRADDBR:
3422 	case SIOCBRDELBR:
3423 	case SIOCGIFVLAN:
3424 	case SIOCSIFVLAN:
3425 	case SIOCGSKNS:
3426 	case SIOCGSTAMP_NEW:
3427 	case SIOCGSTAMPNS_NEW:
3428 	case SIOCGIFCONF:
3429 	case SIOCSIFBR:
3430 	case SIOCGIFBR:
3431 		return sock_ioctl(file, cmd, arg);
3432 
3433 	case SIOCGIFFLAGS:
3434 	case SIOCSIFFLAGS:
3435 	case SIOCGIFMAP:
3436 	case SIOCSIFMAP:
3437 	case SIOCGIFMETRIC:
3438 	case SIOCSIFMETRIC:
3439 	case SIOCGIFMTU:
3440 	case SIOCSIFMTU:
3441 	case SIOCGIFMEM:
3442 	case SIOCSIFMEM:
3443 	case SIOCGIFHWADDR:
3444 	case SIOCSIFHWADDR:
3445 	case SIOCADDMULTI:
3446 	case SIOCDELMULTI:
3447 	case SIOCGIFINDEX:
3448 	case SIOCGIFADDR:
3449 	case SIOCSIFADDR:
3450 	case SIOCSIFHWBROADCAST:
3451 	case SIOCDIFADDR:
3452 	case SIOCGIFBRDADDR:
3453 	case SIOCSIFBRDADDR:
3454 	case SIOCGIFDSTADDR:
3455 	case SIOCSIFDSTADDR:
3456 	case SIOCGIFNETMASK:
3457 	case SIOCSIFNETMASK:
3458 	case SIOCSIFPFLAGS:
3459 	case SIOCGIFPFLAGS:
3460 	case SIOCGIFTXQLEN:
3461 	case SIOCSIFTXQLEN:
3462 	case SIOCBRADDIF:
3463 	case SIOCBRDELIF:
3464 	case SIOCGIFNAME:
3465 	case SIOCSIFNAME:
3466 	case SIOCGMIIPHY:
3467 	case SIOCGMIIREG:
3468 	case SIOCSMIIREG:
3469 	case SIOCBONDENSLAVE:
3470 	case SIOCBONDRELEASE:
3471 	case SIOCBONDSETHWADDR:
3472 	case SIOCBONDCHANGEACTIVE:
3473 	case SIOCSARP:
3474 	case SIOCGARP:
3475 	case SIOCDARP:
3476 	case SIOCOUTQ:
3477 	case SIOCOUTQNSD:
3478 	case SIOCATMARK:
3479 		return sock_do_ioctl(net, sock, cmd, arg);
3480 	}
3481 
3482 	return -ENOIOCTLCMD;
3483 }
3484 
3485 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3486 			      unsigned long arg)
3487 {
3488 	struct socket *sock = file->private_data;
3489 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3490 	int ret = -ENOIOCTLCMD;
3491 	struct sock *sk;
3492 	struct net *net;
3493 
3494 	sk = sock->sk;
3495 	net = sock_net(sk);
3496 
3497 	if (ops->compat_ioctl)
3498 		ret = ops->compat_ioctl(sock, cmd, arg);
3499 
3500 	if (ret == -ENOIOCTLCMD &&
3501 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3502 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3503 
3504 	if (ret == -ENOIOCTLCMD)
3505 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3506 
3507 	return ret;
3508 }
3509 #endif
3510 
3511 /**
3512  *	kernel_bind - bind an address to a socket (kernel space)
3513  *	@sock: socket
3514  *	@addr: address
3515  *	@addrlen: length of address
3516  *
3517  *	Returns 0 or an error.
3518  */
3519 
3520 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3521 {
3522 	struct sockaddr_storage address;
3523 
3524 	memcpy(&address, addr, addrlen);
3525 
3526 	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3527 					  addrlen);
3528 }
3529 EXPORT_SYMBOL(kernel_bind);
3530 
3531 /**
3532  *	kernel_listen - move socket to listening state (kernel space)
3533  *	@sock: socket
3534  *	@backlog: pending connections queue size
3535  *
3536  *	Returns 0 or an error.
3537  */
3538 
3539 int kernel_listen(struct socket *sock, int backlog)
3540 {
3541 	return READ_ONCE(sock->ops)->listen(sock, backlog);
3542 }
3543 EXPORT_SYMBOL(kernel_listen);
3544 
3545 /**
3546  *	kernel_accept - accept a connection (kernel space)
3547  *	@sock: listening socket
3548  *	@newsock: new connected socket
3549  *	@flags: flags
3550  *
3551  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3552  *	If it fails, @newsock is guaranteed to be %NULL.
3553  *	Returns 0 or an error.
3554  */
3555 
3556 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3557 {
3558 	struct sock *sk = sock->sk;
3559 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3560 	struct proto_accept_arg arg = {
3561 		.flags = flags,
3562 		.kern = true,
3563 	};
3564 	int err;
3565 
3566 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3567 			       newsock);
3568 	if (err < 0)
3569 		goto done;
3570 
3571 	err = ops->accept(sock, *newsock, &arg);
3572 	if (err < 0) {
3573 		sock_release(*newsock);
3574 		*newsock = NULL;
3575 		goto done;
3576 	}
3577 
3578 	(*newsock)->ops = ops;
3579 	__module_get(ops->owner);
3580 
3581 done:
3582 	return err;
3583 }
3584 EXPORT_SYMBOL(kernel_accept);
3585 
3586 /**
3587  *	kernel_connect - connect a socket (kernel space)
3588  *	@sock: socket
3589  *	@addr: address
3590  *	@addrlen: address length
3591  *	@flags: flags (O_NONBLOCK, ...)
3592  *
3593  *	For datagram sockets, @addr is the address to which datagrams are sent
3594  *	by default, and the only address from which datagrams are received.
3595  *	For stream sockets, attempts to connect to @addr.
3596  *	Returns 0 or an error code.
3597  */
3598 
3599 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3600 		   int flags)
3601 {
3602 	struct sockaddr_storage address;
3603 
3604 	memcpy(&address, addr, addrlen);
3605 
3606 	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3607 					     addrlen, flags);
3608 }
3609 EXPORT_SYMBOL(kernel_connect);
3610 
3611 /**
3612  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3613  *	@sock: socket
3614  *	@addr: address holder
3615  *
3616  * 	Fills the @addr pointer with the address which the socket is bound.
3617  *	Returns the length of the address in bytes or an error code.
3618  */
3619 
3620 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3621 {
3622 	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3623 }
3624 EXPORT_SYMBOL(kernel_getsockname);
3625 
3626 /**
3627  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3628  *	@sock: socket
3629  *	@addr: address holder
3630  *
3631  * 	Fills the @addr pointer with the address which the socket is connected.
3632  *	Returns the length of the address in bytes or an error code.
3633  */
3634 
3635 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3636 {
3637 	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3638 }
3639 EXPORT_SYMBOL(kernel_getpeername);
3640 
3641 /**
3642  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3643  *	@sock: socket
3644  *	@how: connection part
3645  *
3646  *	Returns 0 or an error.
3647  */
3648 
3649 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3650 {
3651 	return READ_ONCE(sock->ops)->shutdown(sock, how);
3652 }
3653 EXPORT_SYMBOL(kernel_sock_shutdown);
3654 
3655 /**
3656  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3657  *	@sk: socket
3658  *
3659  *	This routine returns the IP overhead imposed by a socket i.e.
3660  *	the length of the underlying IP header, depending on whether
3661  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3662  *	on at the socket. Assumes that the caller has a lock on the socket.
3663  */
3664 
3665 u32 kernel_sock_ip_overhead(struct sock *sk)
3666 {
3667 	struct inet_sock *inet;
3668 	struct ip_options_rcu *opt;
3669 	u32 overhead = 0;
3670 #if IS_ENABLED(CONFIG_IPV6)
3671 	struct ipv6_pinfo *np;
3672 	struct ipv6_txoptions *optv6 = NULL;
3673 #endif /* IS_ENABLED(CONFIG_IPV6) */
3674 
3675 	if (!sk)
3676 		return overhead;
3677 
3678 	switch (sk->sk_family) {
3679 	case AF_INET:
3680 		inet = inet_sk(sk);
3681 		overhead += sizeof(struct iphdr);
3682 		opt = rcu_dereference_protected(inet->inet_opt,
3683 						sock_owned_by_user(sk));
3684 		if (opt)
3685 			overhead += opt->opt.optlen;
3686 		return overhead;
3687 #if IS_ENABLED(CONFIG_IPV6)
3688 	case AF_INET6:
3689 		np = inet6_sk(sk);
3690 		overhead += sizeof(struct ipv6hdr);
3691 		if (np)
3692 			optv6 = rcu_dereference_protected(np->opt,
3693 							  sock_owned_by_user(sk));
3694 		if (optv6)
3695 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3696 		return overhead;
3697 #endif /* IS_ENABLED(CONFIG_IPV6) */
3698 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3699 		return overhead;
3700 	}
3701 }
3702 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3703