xref: /linux/net/socket.c (revision cb787f4ac0c2e439ea8d7e6387b925f74576bdf8)
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 #ifdef CONFIG_NET_RX_BUSY_POLL
114 unsigned int sysctl_net_busy_read __read_mostly;
115 unsigned int sysctl_net_busy_poll __read_mostly;
116 #endif
117 
118 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
119 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 
122 static int sock_close(struct inode *inode, struct file *file);
123 static __poll_t sock_poll(struct file *file,
124 			      struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 #ifdef CONFIG_COMPAT
127 static long compat_sock_ioctl(struct file *file,
128 			      unsigned int cmd, unsigned long arg);
129 #endif
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 				struct pipe_inode_info *pipe, size_t len,
133 				unsigned int flags);
134 static void sock_splice_eof(struct file *file);
135 
136 #ifdef CONFIG_PROC_FS
137 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
138 {
139 	struct socket *sock = f->private_data;
140 	const struct proto_ops *ops = READ_ONCE(sock->ops);
141 
142 	if (ops->show_fdinfo)
143 		ops->show_fdinfo(m, sock);
144 }
145 #else
146 #define sock_show_fdinfo NULL
147 #endif
148 
149 /*
150  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
151  *	in the operation structures but are done directly via the socketcall() multiplexor.
152  */
153 
154 static const struct file_operations socket_file_ops = {
155 	.owner =	THIS_MODULE,
156 	.read_iter =	sock_read_iter,
157 	.write_iter =	sock_write_iter,
158 	.poll =		sock_poll,
159 	.unlocked_ioctl = sock_ioctl,
160 #ifdef CONFIG_COMPAT
161 	.compat_ioctl = compat_sock_ioctl,
162 #endif
163 	.uring_cmd =    io_uring_cmd_sock,
164 	.mmap =		sock_mmap,
165 	.release =	sock_close,
166 	.fasync =	sock_fasync,
167 	.splice_write = splice_to_socket,
168 	.splice_read =	sock_splice_read,
169 	.splice_eof =	sock_splice_eof,
170 	.show_fdinfo =	sock_show_fdinfo,
171 };
172 
173 static const char * const pf_family_names[] = {
174 	[PF_UNSPEC]	= "PF_UNSPEC",
175 	[PF_UNIX]	= "PF_UNIX/PF_LOCAL",
176 	[PF_INET]	= "PF_INET",
177 	[PF_AX25]	= "PF_AX25",
178 	[PF_IPX]	= "PF_IPX",
179 	[PF_APPLETALK]	= "PF_APPLETALK",
180 	[PF_NETROM]	= "PF_NETROM",
181 	[PF_BRIDGE]	= "PF_BRIDGE",
182 	[PF_ATMPVC]	= "PF_ATMPVC",
183 	[PF_X25]	= "PF_X25",
184 	[PF_INET6]	= "PF_INET6",
185 	[PF_ROSE]	= "PF_ROSE",
186 	[PF_DECnet]	= "PF_DECnet",
187 	[PF_NETBEUI]	= "PF_NETBEUI",
188 	[PF_SECURITY]	= "PF_SECURITY",
189 	[PF_KEY]	= "PF_KEY",
190 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
191 	[PF_PACKET]	= "PF_PACKET",
192 	[PF_ASH]	= "PF_ASH",
193 	[PF_ECONET]	= "PF_ECONET",
194 	[PF_ATMSVC]	= "PF_ATMSVC",
195 	[PF_RDS]	= "PF_RDS",
196 	[PF_SNA]	= "PF_SNA",
197 	[PF_IRDA]	= "PF_IRDA",
198 	[PF_PPPOX]	= "PF_PPPOX",
199 	[PF_WANPIPE]	= "PF_WANPIPE",
200 	[PF_LLC]	= "PF_LLC",
201 	[PF_IB]		= "PF_IB",
202 	[PF_MPLS]	= "PF_MPLS",
203 	[PF_CAN]	= "PF_CAN",
204 	[PF_TIPC]	= "PF_TIPC",
205 	[PF_BLUETOOTH]	= "PF_BLUETOOTH",
206 	[PF_IUCV]	= "PF_IUCV",
207 	[PF_RXRPC]	= "PF_RXRPC",
208 	[PF_ISDN]	= "PF_ISDN",
209 	[PF_PHONET]	= "PF_PHONET",
210 	[PF_IEEE802154]	= "PF_IEEE802154",
211 	[PF_CAIF]	= "PF_CAIF",
212 	[PF_ALG]	= "PF_ALG",
213 	[PF_NFC]	= "PF_NFC",
214 	[PF_VSOCK]	= "PF_VSOCK",
215 	[PF_KCM]	= "PF_KCM",
216 	[PF_QIPCRTR]	= "PF_QIPCRTR",
217 	[PF_SMC]	= "PF_SMC",
218 	[PF_XDP]	= "PF_XDP",
219 	[PF_MCTP]	= "PF_MCTP",
220 };
221 
222 /*
223  *	The protocol list. Each protocol is registered in here.
224  */
225 
226 static DEFINE_SPINLOCK(net_family_lock);
227 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
228 
229 /*
230  * Support routines.
231  * Move socket addresses back and forth across the kernel/user
232  * divide and look after the messy bits.
233  */
234 
235 /**
236  *	move_addr_to_kernel	-	copy a socket address into kernel space
237  *	@uaddr: Address in user space
238  *	@kaddr: Address in kernel space
239  *	@ulen: Length in user space
240  *
241  *	The address is copied into kernel space. If the provided address is
242  *	too long an error code of -EINVAL is returned. If the copy gives
243  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
244  */
245 
246 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
247 {
248 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
249 		return -EINVAL;
250 	if (ulen == 0)
251 		return 0;
252 	if (copy_from_user(kaddr, uaddr, ulen))
253 		return -EFAULT;
254 	return audit_sockaddr(ulen, kaddr);
255 }
256 
257 /**
258  *	move_addr_to_user	-	copy an address to user space
259  *	@kaddr: kernel space address
260  *	@klen: length of address in kernel
261  *	@uaddr: user space address
262  *	@ulen: pointer to user length field
263  *
264  *	The value pointed to by ulen on entry is the buffer length available.
265  *	This is overwritten with the buffer space used. -EINVAL is returned
266  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
267  *	is returned if either the buffer or the length field are not
268  *	accessible.
269  *	After copying the data up to the limit the user specifies, the true
270  *	length of the data is written over the length limit the user
271  *	specified. Zero is returned for a success.
272  */
273 
274 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
275 			     void __user *uaddr, int __user *ulen)
276 {
277 	int err;
278 	int len;
279 
280 	BUG_ON(klen > sizeof(struct sockaddr_storage));
281 	err = get_user(len, ulen);
282 	if (err)
283 		return err;
284 	if (len > klen)
285 		len = klen;
286 	if (len < 0)
287 		return -EINVAL;
288 	if (len) {
289 		if (audit_sockaddr(klen, kaddr))
290 			return -ENOMEM;
291 		if (copy_to_user(uaddr, kaddr, len))
292 			return -EFAULT;
293 	}
294 	/*
295 	 *      "fromlen shall refer to the value before truncation.."
296 	 *                      1003.1g
297 	 */
298 	return __put_user(klen, ulen);
299 }
300 
301 static struct kmem_cache *sock_inode_cachep __ro_after_init;
302 
303 static struct inode *sock_alloc_inode(struct super_block *sb)
304 {
305 	struct socket_alloc *ei;
306 
307 	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
308 	if (!ei)
309 		return NULL;
310 	init_waitqueue_head(&ei->socket.wq.wait);
311 	ei->socket.wq.fasync_list = NULL;
312 	ei->socket.wq.flags = 0;
313 
314 	ei->socket.state = SS_UNCONNECTED;
315 	ei->socket.flags = 0;
316 	ei->socket.ops = NULL;
317 	ei->socket.sk = NULL;
318 	ei->socket.file = NULL;
319 
320 	return &ei->vfs_inode;
321 }
322 
323 static void sock_free_inode(struct inode *inode)
324 {
325 	struct socket_alloc *ei;
326 
327 	ei = container_of(inode, struct socket_alloc, vfs_inode);
328 	kmem_cache_free(sock_inode_cachep, ei);
329 }
330 
331 static void init_once(void *foo)
332 {
333 	struct socket_alloc *ei = (struct socket_alloc *)foo;
334 
335 	inode_init_once(&ei->vfs_inode);
336 }
337 
338 static void init_inodecache(void)
339 {
340 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
341 					      sizeof(struct socket_alloc),
342 					      0,
343 					      (SLAB_HWCACHE_ALIGN |
344 					       SLAB_RECLAIM_ACCOUNT |
345 					       SLAB_ACCOUNT),
346 					      init_once);
347 	BUG_ON(sock_inode_cachep == NULL);
348 }
349 
350 static const struct super_operations sockfs_ops = {
351 	.alloc_inode	= sock_alloc_inode,
352 	.free_inode	= sock_free_inode,
353 	.statfs		= simple_statfs,
354 };
355 
356 /*
357  * sockfs_dname() is called from d_path().
358  */
359 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
360 {
361 	return dynamic_dname(buffer, buflen, "socket:[%lu]",
362 				d_inode(dentry)->i_ino);
363 }
364 
365 static const struct dentry_operations sockfs_dentry_operations = {
366 	.d_dname  = sockfs_dname,
367 };
368 
369 static int sockfs_xattr_get(const struct xattr_handler *handler,
370 			    struct dentry *dentry, struct inode *inode,
371 			    const char *suffix, void *value, size_t size)
372 {
373 	if (value) {
374 		if (dentry->d_name.len + 1 > size)
375 			return -ERANGE;
376 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
377 	}
378 	return dentry->d_name.len + 1;
379 }
380 
381 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
382 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
383 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
384 
385 static const struct xattr_handler sockfs_xattr_handler = {
386 	.name = XATTR_NAME_SOCKPROTONAME,
387 	.get = sockfs_xattr_get,
388 };
389 
390 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
391 				     struct mnt_idmap *idmap,
392 				     struct dentry *dentry, struct inode *inode,
393 				     const char *suffix, const void *value,
394 				     size_t size, int flags)
395 {
396 	/* Handled by LSM. */
397 	return -EAGAIN;
398 }
399 
400 static const struct xattr_handler sockfs_security_xattr_handler = {
401 	.prefix = XATTR_SECURITY_PREFIX,
402 	.set = sockfs_security_xattr_set,
403 };
404 
405 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
406 	&sockfs_xattr_handler,
407 	&sockfs_security_xattr_handler,
408 	NULL
409 };
410 
411 static int sockfs_init_fs_context(struct fs_context *fc)
412 {
413 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
414 	if (!ctx)
415 		return -ENOMEM;
416 	ctx->ops = &sockfs_ops;
417 	ctx->dops = &sockfs_dentry_operations;
418 	ctx->xattr = sockfs_xattr_handlers;
419 	return 0;
420 }
421 
422 static struct vfsmount *sock_mnt __read_mostly;
423 
424 static struct file_system_type sock_fs_type = {
425 	.name =		"sockfs",
426 	.init_fs_context = sockfs_init_fs_context,
427 	.kill_sb =	kill_anon_super,
428 };
429 
430 /*
431  *	Obtains the first available file descriptor and sets it up for use.
432  *
433  *	These functions create file structures and maps them to fd space
434  *	of the current process. On success it returns file descriptor
435  *	and file struct implicitly stored in sock->file.
436  *	Note that another thread may close file descriptor before we return
437  *	from this function. We use the fact that now we do not refer
438  *	to socket after mapping. If one day we will need it, this
439  *	function will increment ref. count on file by 1.
440  *
441  *	In any case returned fd MAY BE not valid!
442  *	This race condition is unavoidable
443  *	with shared fd spaces, we cannot solve it inside kernel,
444  *	but we take care of internal coherence yet.
445  */
446 
447 /**
448  *	sock_alloc_file - Bind a &socket to a &file
449  *	@sock: socket
450  *	@flags: file status flags
451  *	@dname: protocol name
452  *
453  *	Returns the &file bound with @sock, implicitly storing it
454  *	in sock->file. If dname is %NULL, sets to "".
455  *
456  *	On failure @sock is released, and an ERR pointer is returned.
457  *
458  *	This function uses GFP_KERNEL internally.
459  */
460 
461 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
462 {
463 	struct file *file;
464 
465 	if (!dname)
466 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
467 
468 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
469 				O_RDWR | (flags & O_NONBLOCK),
470 				&socket_file_ops);
471 	if (IS_ERR(file)) {
472 		sock_release(sock);
473 		return file;
474 	}
475 
476 	file->f_mode |= FMODE_NOWAIT;
477 	sock->file = file;
478 	file->private_data = sock;
479 	stream_open(SOCK_INODE(sock), file);
480 	return file;
481 }
482 EXPORT_SYMBOL(sock_alloc_file);
483 
484 static int sock_map_fd(struct socket *sock, int flags)
485 {
486 	struct file *newfile;
487 	int fd = get_unused_fd_flags(flags);
488 	if (unlikely(fd < 0)) {
489 		sock_release(sock);
490 		return fd;
491 	}
492 
493 	newfile = sock_alloc_file(sock, flags, NULL);
494 	if (!IS_ERR(newfile)) {
495 		fd_install(fd, newfile);
496 		return fd;
497 	}
498 
499 	put_unused_fd(fd);
500 	return PTR_ERR(newfile);
501 }
502 
503 /**
504  *	sock_from_file - Return the &socket bounded to @file.
505  *	@file: file
506  *
507  *	On failure returns %NULL.
508  */
509 
510 struct socket *sock_from_file(struct file *file)
511 {
512 	if (file->f_op == &socket_file_ops)
513 		return file->private_data;	/* set in sock_alloc_file */
514 
515 	return NULL;
516 }
517 EXPORT_SYMBOL(sock_from_file);
518 
519 /**
520  *	sockfd_lookup - Go from a file number to its socket slot
521  *	@fd: file handle
522  *	@err: pointer to an error code return
523  *
524  *	The file handle passed in is locked and the socket it is bound
525  *	to is returned. If an error occurs the err pointer is overwritten
526  *	with a negative errno code and NULL is returned. The function checks
527  *	for both invalid handles and passing a handle which is not a socket.
528  *
529  *	On a success the socket object pointer is returned.
530  */
531 
532 struct socket *sockfd_lookup(int fd, int *err)
533 {
534 	struct file *file;
535 	struct socket *sock;
536 
537 	file = fget(fd);
538 	if (!file) {
539 		*err = -EBADF;
540 		return NULL;
541 	}
542 
543 	sock = sock_from_file(file);
544 	if (!sock) {
545 		*err = -ENOTSOCK;
546 		fput(file);
547 	}
548 	return sock;
549 }
550 EXPORT_SYMBOL(sockfd_lookup);
551 
552 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
553 {
554 	struct fd f = fdget(fd);
555 	struct socket *sock;
556 
557 	*err = -EBADF;
558 	if (fd_file(f)) {
559 		sock = sock_from_file(fd_file(f));
560 		if (likely(sock)) {
561 			*fput_needed = f.word & FDPUT_FPUT;
562 			return sock;
563 		}
564 		*err = -ENOTSOCK;
565 		fdput(f);
566 	}
567 	return NULL;
568 }
569 
570 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
571 				size_t size)
572 {
573 	ssize_t len;
574 	ssize_t used = 0;
575 
576 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
577 	if (len < 0)
578 		return len;
579 	used += len;
580 	if (buffer) {
581 		if (size < used)
582 			return -ERANGE;
583 		buffer += len;
584 	}
585 
586 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
587 	used += len;
588 	if (buffer) {
589 		if (size < used)
590 			return -ERANGE;
591 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
592 		buffer += len;
593 	}
594 
595 	return used;
596 }
597 
598 static int sockfs_setattr(struct mnt_idmap *idmap,
599 			  struct dentry *dentry, struct iattr *iattr)
600 {
601 	int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
602 
603 	if (!err && (iattr->ia_valid & ATTR_UID)) {
604 		struct socket *sock = SOCKET_I(d_inode(dentry));
605 
606 		if (sock->sk)
607 			sock->sk->sk_uid = iattr->ia_uid;
608 		else
609 			err = -ENOENT;
610 	}
611 
612 	return err;
613 }
614 
615 static const struct inode_operations sockfs_inode_ops = {
616 	.listxattr = sockfs_listxattr,
617 	.setattr = sockfs_setattr,
618 };
619 
620 /**
621  *	sock_alloc - allocate a socket
622  *
623  *	Allocate a new inode and socket object. The two are bound together
624  *	and initialised. The socket is then returned. If we are out of inodes
625  *	NULL is returned. This functions uses GFP_KERNEL internally.
626  */
627 
628 struct socket *sock_alloc(void)
629 {
630 	struct inode *inode;
631 	struct socket *sock;
632 
633 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
634 	if (!inode)
635 		return NULL;
636 
637 	sock = SOCKET_I(inode);
638 
639 	inode->i_ino = get_next_ino();
640 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
641 	inode->i_uid = current_fsuid();
642 	inode->i_gid = current_fsgid();
643 	inode->i_op = &sockfs_inode_ops;
644 
645 	return sock;
646 }
647 EXPORT_SYMBOL(sock_alloc);
648 
649 static void __sock_release(struct socket *sock, struct inode *inode)
650 {
651 	const struct proto_ops *ops = READ_ONCE(sock->ops);
652 
653 	if (ops) {
654 		struct module *owner = ops->owner;
655 
656 		if (inode)
657 			inode_lock(inode);
658 		ops->release(sock);
659 		sock->sk = NULL;
660 		if (inode)
661 			inode_unlock(inode);
662 		sock->ops = NULL;
663 		module_put(owner);
664 	}
665 
666 	if (sock->wq.fasync_list)
667 		pr_err("%s: fasync list not empty!\n", __func__);
668 
669 	if (!sock->file) {
670 		iput(SOCK_INODE(sock));
671 		return;
672 	}
673 	sock->file = NULL;
674 }
675 
676 /**
677  *	sock_release - close a socket
678  *	@sock: socket to close
679  *
680  *	The socket is released from the protocol stack if it has a release
681  *	callback, and the inode is then released if the socket is bound to
682  *	an inode not a file.
683  */
684 void sock_release(struct socket *sock)
685 {
686 	__sock_release(sock, NULL);
687 }
688 EXPORT_SYMBOL(sock_release);
689 
690 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
691 {
692 	u8 flags = *tx_flags;
693 
694 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
695 		flags |= SKBTX_HW_TSTAMP;
696 
697 		/* PTP hardware clocks can provide a free running cycle counter
698 		 * as a time base for virtual clocks. Tell driver to use the
699 		 * free running cycle counter for timestamp if socket is bound
700 		 * to virtual clock.
701 		 */
702 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
703 			flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
704 	}
705 
706 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
707 		flags |= SKBTX_SW_TSTAMP;
708 
709 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
710 		flags |= SKBTX_SCHED_TSTAMP;
711 
712 	*tx_flags = flags;
713 }
714 EXPORT_SYMBOL(__sock_tx_timestamp);
715 
716 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
717 					   size_t));
718 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
719 					    size_t));
720 
721 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
722 						 int flags)
723 {
724 	trace_sock_send_length(sk, ret, 0);
725 }
726 
727 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
728 {
729 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
730 				     inet_sendmsg, sock, msg,
731 				     msg_data_left(msg));
732 	BUG_ON(ret == -EIOCBQUEUED);
733 
734 	if (trace_sock_send_length_enabled())
735 		call_trace_sock_send_length(sock->sk, ret, 0);
736 	return ret;
737 }
738 
739 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
740 {
741 	int err = security_socket_sendmsg(sock, msg,
742 					  msg_data_left(msg));
743 
744 	return err ?: sock_sendmsg_nosec(sock, msg);
745 }
746 
747 /**
748  *	sock_sendmsg - send a message through @sock
749  *	@sock: socket
750  *	@msg: message to send
751  *
752  *	Sends @msg through @sock, passing through LSM.
753  *	Returns the number of bytes sent, or an error code.
754  */
755 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
756 {
757 	struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
758 	struct sockaddr_storage address;
759 	int save_len = msg->msg_namelen;
760 	int ret;
761 
762 	if (msg->msg_name) {
763 		memcpy(&address, msg->msg_name, msg->msg_namelen);
764 		msg->msg_name = &address;
765 	}
766 
767 	ret = __sock_sendmsg(sock, msg);
768 	msg->msg_name = save_addr;
769 	msg->msg_namelen = save_len;
770 
771 	return ret;
772 }
773 EXPORT_SYMBOL(sock_sendmsg);
774 
775 /**
776  *	kernel_sendmsg - send a message through @sock (kernel-space)
777  *	@sock: socket
778  *	@msg: message header
779  *	@vec: kernel vec
780  *	@num: vec array length
781  *	@size: total message data size
782  *
783  *	Builds the message data with @vec and sends it through @sock.
784  *	Returns the number of bytes sent, or an error code.
785  */
786 
787 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
788 		   struct kvec *vec, size_t num, size_t size)
789 {
790 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
791 	return sock_sendmsg(sock, msg);
792 }
793 EXPORT_SYMBOL(kernel_sendmsg);
794 
795 /**
796  *	kernel_sendmsg_locked - send a message through @sock (kernel-space)
797  *	@sk: sock
798  *	@msg: message header
799  *	@vec: output s/g array
800  *	@num: output s/g array length
801  *	@size: total message data size
802  *
803  *	Builds the message data with @vec and sends it through @sock.
804  *	Returns the number of bytes sent, or an error code.
805  *	Caller must hold @sk.
806  */
807 
808 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
809 			  struct kvec *vec, size_t num, size_t size)
810 {
811 	struct socket *sock = sk->sk_socket;
812 	const struct proto_ops *ops = READ_ONCE(sock->ops);
813 
814 	if (!ops->sendmsg_locked)
815 		return sock_no_sendmsg_locked(sk, msg, size);
816 
817 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
818 
819 	return ops->sendmsg_locked(sk, msg, msg_data_left(msg));
820 }
821 EXPORT_SYMBOL(kernel_sendmsg_locked);
822 
823 static bool skb_is_err_queue(const struct sk_buff *skb)
824 {
825 	/* pkt_type of skbs enqueued on the error queue are set to
826 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
827 	 * in recvmsg, since skbs received on a local socket will never
828 	 * have a pkt_type of PACKET_OUTGOING.
829 	 */
830 	return skb->pkt_type == PACKET_OUTGOING;
831 }
832 
833 /* On transmit, software and hardware timestamps are returned independently.
834  * As the two skb clones share the hardware timestamp, which may be updated
835  * before the software timestamp is received, a hardware TX timestamp may be
836  * returned only if there is no software TX timestamp. Ignore false software
837  * timestamps, which may be made in the __sock_recv_timestamp() call when the
838  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
839  * hardware timestamp.
840  */
841 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
842 {
843 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
844 }
845 
846 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
847 {
848 	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
849 	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
850 	struct net_device *orig_dev;
851 	ktime_t hwtstamp;
852 
853 	rcu_read_lock();
854 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
855 	if (orig_dev) {
856 		*if_index = orig_dev->ifindex;
857 		hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
858 	} else {
859 		hwtstamp = shhwtstamps->hwtstamp;
860 	}
861 	rcu_read_unlock();
862 
863 	return hwtstamp;
864 }
865 
866 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
867 			   int if_index)
868 {
869 	struct scm_ts_pktinfo ts_pktinfo;
870 	struct net_device *orig_dev;
871 
872 	if (!skb_mac_header_was_set(skb))
873 		return;
874 
875 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
876 
877 	if (!if_index) {
878 		rcu_read_lock();
879 		orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
880 		if (orig_dev)
881 			if_index = orig_dev->ifindex;
882 		rcu_read_unlock();
883 	}
884 	ts_pktinfo.if_index = if_index;
885 
886 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
887 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
888 		 sizeof(ts_pktinfo), &ts_pktinfo);
889 }
890 
891 /*
892  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
893  */
894 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
895 	struct sk_buff *skb)
896 {
897 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
898 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
899 	struct scm_timestamping_internal tss;
900 	int empty = 1, false_tstamp = 0;
901 	struct skb_shared_hwtstamps *shhwtstamps =
902 		skb_hwtstamps(skb);
903 	int if_index;
904 	ktime_t hwtstamp;
905 	u32 tsflags;
906 
907 	/* Race occurred between timestamp enabling and packet
908 	   receiving.  Fill in the current time for now. */
909 	if (need_software_tstamp && skb->tstamp == 0) {
910 		__net_timestamp(skb);
911 		false_tstamp = 1;
912 	}
913 
914 	if (need_software_tstamp) {
915 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
916 			if (new_tstamp) {
917 				struct __kernel_sock_timeval tv;
918 
919 				skb_get_new_timestamp(skb, &tv);
920 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
921 					 sizeof(tv), &tv);
922 			} else {
923 				struct __kernel_old_timeval tv;
924 
925 				skb_get_timestamp(skb, &tv);
926 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
927 					 sizeof(tv), &tv);
928 			}
929 		} else {
930 			if (new_tstamp) {
931 				struct __kernel_timespec ts;
932 
933 				skb_get_new_timestampns(skb, &ts);
934 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
935 					 sizeof(ts), &ts);
936 			} else {
937 				struct __kernel_old_timespec ts;
938 
939 				skb_get_timestampns(skb, &ts);
940 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
941 					 sizeof(ts), &ts);
942 			}
943 		}
944 	}
945 
946 	memset(&tss, 0, sizeof(tss));
947 	tsflags = READ_ONCE(sk->sk_tsflags);
948 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
949 	     (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE ||
950 	      skb_is_err_queue(skb) ||
951 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
952 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
953 		empty = 0;
954 	if (shhwtstamps &&
955 	    (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
956 	     (tsflags & SOF_TIMESTAMPING_RX_HARDWARE ||
957 	      skb_is_err_queue(skb) ||
958 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
959 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
960 		if_index = 0;
961 		if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
962 			hwtstamp = get_timestamp(sk, skb, &if_index);
963 		else
964 			hwtstamp = shhwtstamps->hwtstamp;
965 
966 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
967 			hwtstamp = ptp_convert_timestamp(&hwtstamp,
968 							 READ_ONCE(sk->sk_bind_phc));
969 
970 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
971 			empty = 0;
972 
973 			if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
974 			    !skb_is_err_queue(skb))
975 				put_ts_pktinfo(msg, skb, if_index);
976 		}
977 	}
978 	if (!empty) {
979 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
980 			put_cmsg_scm_timestamping64(msg, &tss);
981 		else
982 			put_cmsg_scm_timestamping(msg, &tss);
983 
984 		if (skb_is_err_queue(skb) && skb->len &&
985 		    SKB_EXT_ERR(skb)->opt_stats)
986 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
987 				 skb->len, skb->data);
988 	}
989 }
990 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
991 
992 #ifdef CONFIG_WIRELESS
993 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
994 	struct sk_buff *skb)
995 {
996 	int ack;
997 
998 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
999 		return;
1000 	if (!skb->wifi_acked_valid)
1001 		return;
1002 
1003 	ack = skb->wifi_acked;
1004 
1005 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1006 }
1007 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1008 #endif
1009 
1010 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1011 				   struct sk_buff *skb)
1012 {
1013 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1014 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1015 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1016 }
1017 
1018 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1019 			   struct sk_buff *skb)
1020 {
1021 	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1022 		/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1023 		__u32 mark = skb->mark;
1024 
1025 		put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1026 	}
1027 }
1028 
1029 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1030 		       struct sk_buff *skb)
1031 {
1032 	sock_recv_timestamp(msg, sk, skb);
1033 	sock_recv_drops(msg, sk, skb);
1034 	sock_recv_mark(msg, sk, skb);
1035 }
1036 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1037 
1038 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1039 					   size_t, int));
1040 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1041 					    size_t, int));
1042 
1043 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1044 {
1045 	trace_sock_recv_length(sk, ret, flags);
1046 }
1047 
1048 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1049 				     int flags)
1050 {
1051 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1052 				     inet6_recvmsg,
1053 				     inet_recvmsg, sock, msg,
1054 				     msg_data_left(msg), flags);
1055 	if (trace_sock_recv_length_enabled())
1056 		call_trace_sock_recv_length(sock->sk, ret, flags);
1057 	return ret;
1058 }
1059 
1060 /**
1061  *	sock_recvmsg - receive a message from @sock
1062  *	@sock: socket
1063  *	@msg: message to receive
1064  *	@flags: message flags
1065  *
1066  *	Receives @msg from @sock, passing through LSM. Returns the total number
1067  *	of bytes received, or an error.
1068  */
1069 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1070 {
1071 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1072 
1073 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1074 }
1075 EXPORT_SYMBOL(sock_recvmsg);
1076 
1077 /**
1078  *	kernel_recvmsg - Receive a message from a socket (kernel space)
1079  *	@sock: The socket to receive the message from
1080  *	@msg: Received message
1081  *	@vec: Input s/g array for message data
1082  *	@num: Size of input s/g array
1083  *	@size: Number of bytes to read
1084  *	@flags: Message flags (MSG_DONTWAIT, etc...)
1085  *
1086  *	On return the msg structure contains the scatter/gather array passed in the
1087  *	vec argument. The array is modified so that it consists of the unfilled
1088  *	portion of the original array.
1089  *
1090  *	The returned value is the total number of bytes received, or an error.
1091  */
1092 
1093 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1094 		   struct kvec *vec, size_t num, size_t size, int flags)
1095 {
1096 	msg->msg_control_is_user = false;
1097 	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1098 	return sock_recvmsg(sock, msg, flags);
1099 }
1100 EXPORT_SYMBOL(kernel_recvmsg);
1101 
1102 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1103 				struct pipe_inode_info *pipe, size_t len,
1104 				unsigned int flags)
1105 {
1106 	struct socket *sock = file->private_data;
1107 	const struct proto_ops *ops;
1108 
1109 	ops = READ_ONCE(sock->ops);
1110 	if (unlikely(!ops->splice_read))
1111 		return copy_splice_read(file, ppos, pipe, len, flags);
1112 
1113 	return ops->splice_read(sock, ppos, pipe, len, flags);
1114 }
1115 
1116 static void sock_splice_eof(struct file *file)
1117 {
1118 	struct socket *sock = file->private_data;
1119 	const struct proto_ops *ops;
1120 
1121 	ops = READ_ONCE(sock->ops);
1122 	if (ops->splice_eof)
1123 		ops->splice_eof(sock);
1124 }
1125 
1126 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1127 {
1128 	struct file *file = iocb->ki_filp;
1129 	struct socket *sock = file->private_data;
1130 	struct msghdr msg = {.msg_iter = *to,
1131 			     .msg_iocb = iocb};
1132 	ssize_t res;
1133 
1134 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1135 		msg.msg_flags = MSG_DONTWAIT;
1136 
1137 	if (iocb->ki_pos != 0)
1138 		return -ESPIPE;
1139 
1140 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1141 		return 0;
1142 
1143 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1144 	*to = msg.msg_iter;
1145 	return res;
1146 }
1147 
1148 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1149 {
1150 	struct file *file = iocb->ki_filp;
1151 	struct socket *sock = file->private_data;
1152 	struct msghdr msg = {.msg_iter = *from,
1153 			     .msg_iocb = iocb};
1154 	ssize_t res;
1155 
1156 	if (iocb->ki_pos != 0)
1157 		return -ESPIPE;
1158 
1159 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1160 		msg.msg_flags = MSG_DONTWAIT;
1161 
1162 	if (sock->type == SOCK_SEQPACKET)
1163 		msg.msg_flags |= MSG_EOR;
1164 
1165 	res = __sock_sendmsg(sock, &msg);
1166 	*from = msg.msg_iter;
1167 	return res;
1168 }
1169 
1170 /*
1171  * Atomic setting of ioctl hooks to avoid race
1172  * with module unload.
1173  */
1174 
1175 static DEFINE_MUTEX(br_ioctl_mutex);
1176 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1177 			    unsigned int cmd, struct ifreq *ifr,
1178 			    void __user *uarg);
1179 
1180 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1181 			     unsigned int cmd, struct ifreq *ifr,
1182 			     void __user *uarg))
1183 {
1184 	mutex_lock(&br_ioctl_mutex);
1185 	br_ioctl_hook = hook;
1186 	mutex_unlock(&br_ioctl_mutex);
1187 }
1188 EXPORT_SYMBOL(brioctl_set);
1189 
1190 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1191 		  struct ifreq *ifr, void __user *uarg)
1192 {
1193 	int err = -ENOPKG;
1194 
1195 	if (!br_ioctl_hook)
1196 		request_module("bridge");
1197 
1198 	mutex_lock(&br_ioctl_mutex);
1199 	if (br_ioctl_hook)
1200 		err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1201 	mutex_unlock(&br_ioctl_mutex);
1202 
1203 	return err;
1204 }
1205 
1206 static DEFINE_MUTEX(vlan_ioctl_mutex);
1207 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1208 
1209 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1210 {
1211 	mutex_lock(&vlan_ioctl_mutex);
1212 	vlan_ioctl_hook = hook;
1213 	mutex_unlock(&vlan_ioctl_mutex);
1214 }
1215 EXPORT_SYMBOL(vlan_ioctl_set);
1216 
1217 static long sock_do_ioctl(struct net *net, struct socket *sock,
1218 			  unsigned int cmd, unsigned long arg)
1219 {
1220 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1221 	struct ifreq ifr;
1222 	bool need_copyout;
1223 	int err;
1224 	void __user *argp = (void __user *)arg;
1225 	void __user *data;
1226 
1227 	err = ops->ioctl(sock, cmd, arg);
1228 
1229 	/*
1230 	 * If this ioctl is unknown try to hand it down
1231 	 * to the NIC driver.
1232 	 */
1233 	if (err != -ENOIOCTLCMD)
1234 		return err;
1235 
1236 	if (!is_socket_ioctl_cmd(cmd))
1237 		return -ENOTTY;
1238 
1239 	if (get_user_ifreq(&ifr, &data, argp))
1240 		return -EFAULT;
1241 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1242 	if (!err && need_copyout)
1243 		if (put_user_ifreq(&ifr, argp))
1244 			return -EFAULT;
1245 
1246 	return err;
1247 }
1248 
1249 /*
1250  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1251  *	what to do with it - that's up to the protocol still.
1252  */
1253 
1254 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1255 {
1256 	const struct proto_ops  *ops;
1257 	struct socket *sock;
1258 	struct sock *sk;
1259 	void __user *argp = (void __user *)arg;
1260 	int pid, err;
1261 	struct net *net;
1262 
1263 	sock = file->private_data;
1264 	ops = READ_ONCE(sock->ops);
1265 	sk = sock->sk;
1266 	net = sock_net(sk);
1267 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1268 		struct ifreq ifr;
1269 		void __user *data;
1270 		bool need_copyout;
1271 		if (get_user_ifreq(&ifr, &data, argp))
1272 			return -EFAULT;
1273 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1274 		if (!err && need_copyout)
1275 			if (put_user_ifreq(&ifr, argp))
1276 				return -EFAULT;
1277 	} else
1278 #ifdef CONFIG_WEXT_CORE
1279 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1280 		err = wext_handle_ioctl(net, cmd, argp);
1281 	} else
1282 #endif
1283 		switch (cmd) {
1284 		case FIOSETOWN:
1285 		case SIOCSPGRP:
1286 			err = -EFAULT;
1287 			if (get_user(pid, (int __user *)argp))
1288 				break;
1289 			err = f_setown(sock->file, pid, 1);
1290 			break;
1291 		case FIOGETOWN:
1292 		case SIOCGPGRP:
1293 			err = put_user(f_getown(sock->file),
1294 				       (int __user *)argp);
1295 			break;
1296 		case SIOCGIFBR:
1297 		case SIOCSIFBR:
1298 		case SIOCBRADDBR:
1299 		case SIOCBRDELBR:
1300 			err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1301 			break;
1302 		case SIOCGIFVLAN:
1303 		case SIOCSIFVLAN:
1304 			err = -ENOPKG;
1305 			if (!vlan_ioctl_hook)
1306 				request_module("8021q");
1307 
1308 			mutex_lock(&vlan_ioctl_mutex);
1309 			if (vlan_ioctl_hook)
1310 				err = vlan_ioctl_hook(net, argp);
1311 			mutex_unlock(&vlan_ioctl_mutex);
1312 			break;
1313 		case SIOCGSKNS:
1314 			err = -EPERM;
1315 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1316 				break;
1317 
1318 			err = open_related_ns(&net->ns, get_net_ns);
1319 			break;
1320 		case SIOCGSTAMP_OLD:
1321 		case SIOCGSTAMPNS_OLD:
1322 			if (!ops->gettstamp) {
1323 				err = -ENOIOCTLCMD;
1324 				break;
1325 			}
1326 			err = ops->gettstamp(sock, argp,
1327 					     cmd == SIOCGSTAMP_OLD,
1328 					     !IS_ENABLED(CONFIG_64BIT));
1329 			break;
1330 		case SIOCGSTAMP_NEW:
1331 		case SIOCGSTAMPNS_NEW:
1332 			if (!ops->gettstamp) {
1333 				err = -ENOIOCTLCMD;
1334 				break;
1335 			}
1336 			err = ops->gettstamp(sock, argp,
1337 					     cmd == SIOCGSTAMP_NEW,
1338 					     false);
1339 			break;
1340 
1341 		case SIOCGIFCONF:
1342 			err = dev_ifconf(net, argp);
1343 			break;
1344 
1345 		default:
1346 			err = sock_do_ioctl(net, sock, cmd, arg);
1347 			break;
1348 		}
1349 	return err;
1350 }
1351 
1352 /**
1353  *	sock_create_lite - creates a socket
1354  *	@family: protocol family (AF_INET, ...)
1355  *	@type: communication type (SOCK_STREAM, ...)
1356  *	@protocol: protocol (0, ...)
1357  *	@res: new socket
1358  *
1359  *	Creates a new socket and assigns it to @res, passing through LSM.
1360  *	The new socket initialization is not complete, see kernel_accept().
1361  *	Returns 0 or an error. On failure @res is set to %NULL.
1362  *	This function internally uses GFP_KERNEL.
1363  */
1364 
1365 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1366 {
1367 	int err;
1368 	struct socket *sock = NULL;
1369 
1370 	err = security_socket_create(family, type, protocol, 1);
1371 	if (err)
1372 		goto out;
1373 
1374 	sock = sock_alloc();
1375 	if (!sock) {
1376 		err = -ENOMEM;
1377 		goto out;
1378 	}
1379 
1380 	sock->type = type;
1381 	err = security_socket_post_create(sock, family, type, protocol, 1);
1382 	if (err)
1383 		goto out_release;
1384 
1385 out:
1386 	*res = sock;
1387 	return err;
1388 out_release:
1389 	sock_release(sock);
1390 	sock = NULL;
1391 	goto out;
1392 }
1393 EXPORT_SYMBOL(sock_create_lite);
1394 
1395 /* No kernel lock held - perfect */
1396 static __poll_t sock_poll(struct file *file, poll_table *wait)
1397 {
1398 	struct socket *sock = file->private_data;
1399 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1400 	__poll_t events = poll_requested_events(wait), flag = 0;
1401 
1402 	if (!ops->poll)
1403 		return 0;
1404 
1405 	if (sk_can_busy_loop(sock->sk)) {
1406 		/* poll once if requested by the syscall */
1407 		if (events & POLL_BUSY_LOOP)
1408 			sk_busy_loop(sock->sk, 1);
1409 
1410 		/* if this socket can poll_ll, tell the system call */
1411 		flag = POLL_BUSY_LOOP;
1412 	}
1413 
1414 	return ops->poll(file, sock, wait) | flag;
1415 }
1416 
1417 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1418 {
1419 	struct socket *sock = file->private_data;
1420 
1421 	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1422 }
1423 
1424 static int sock_close(struct inode *inode, struct file *filp)
1425 {
1426 	__sock_release(SOCKET_I(inode), inode);
1427 	return 0;
1428 }
1429 
1430 /*
1431  *	Update the socket async list
1432  *
1433  *	Fasync_list locking strategy.
1434  *
1435  *	1. fasync_list is modified only under process context socket lock
1436  *	   i.e. under semaphore.
1437  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1438  *	   or under socket lock
1439  */
1440 
1441 static int sock_fasync(int fd, struct file *filp, int on)
1442 {
1443 	struct socket *sock = filp->private_data;
1444 	struct sock *sk = sock->sk;
1445 	struct socket_wq *wq = &sock->wq;
1446 
1447 	if (sk == NULL)
1448 		return -EINVAL;
1449 
1450 	lock_sock(sk);
1451 	fasync_helper(fd, filp, on, &wq->fasync_list);
1452 
1453 	if (!wq->fasync_list)
1454 		sock_reset_flag(sk, SOCK_FASYNC);
1455 	else
1456 		sock_set_flag(sk, SOCK_FASYNC);
1457 
1458 	release_sock(sk);
1459 	return 0;
1460 }
1461 
1462 /* This function may be called only under rcu_lock */
1463 
1464 int sock_wake_async(struct socket_wq *wq, int how, int band)
1465 {
1466 	if (!wq || !wq->fasync_list)
1467 		return -1;
1468 
1469 	switch (how) {
1470 	case SOCK_WAKE_WAITD:
1471 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1472 			break;
1473 		goto call_kill;
1474 	case SOCK_WAKE_SPACE:
1475 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1476 			break;
1477 		fallthrough;
1478 	case SOCK_WAKE_IO:
1479 call_kill:
1480 		kill_fasync(&wq->fasync_list, SIGIO, band);
1481 		break;
1482 	case SOCK_WAKE_URG:
1483 		kill_fasync(&wq->fasync_list, SIGURG, band);
1484 	}
1485 
1486 	return 0;
1487 }
1488 EXPORT_SYMBOL(sock_wake_async);
1489 
1490 /**
1491  *	__sock_create - creates a socket
1492  *	@net: net namespace
1493  *	@family: protocol family (AF_INET, ...)
1494  *	@type: communication type (SOCK_STREAM, ...)
1495  *	@protocol: protocol (0, ...)
1496  *	@res: new socket
1497  *	@kern: boolean for kernel space sockets
1498  *
1499  *	Creates a new socket and assigns it to @res, passing through LSM.
1500  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1501  *	be set to true if the socket resides in kernel space.
1502  *	This function internally uses GFP_KERNEL.
1503  */
1504 
1505 int __sock_create(struct net *net, int family, int type, int protocol,
1506 			 struct socket **res, int kern)
1507 {
1508 	int err;
1509 	struct socket *sock;
1510 	const struct net_proto_family *pf;
1511 
1512 	/*
1513 	 *      Check protocol is in range
1514 	 */
1515 	if (family < 0 || family >= NPROTO)
1516 		return -EAFNOSUPPORT;
1517 	if (type < 0 || type >= SOCK_MAX)
1518 		return -EINVAL;
1519 
1520 	/* Compatibility.
1521 
1522 	   This uglymoron is moved from INET layer to here to avoid
1523 	   deadlock in module load.
1524 	 */
1525 	if (family == PF_INET && type == SOCK_PACKET) {
1526 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1527 			     current->comm);
1528 		family = PF_PACKET;
1529 	}
1530 
1531 	err = security_socket_create(family, type, protocol, kern);
1532 	if (err)
1533 		return err;
1534 
1535 	/*
1536 	 *	Allocate the socket and allow the family to set things up. if
1537 	 *	the protocol is 0, the family is instructed to select an appropriate
1538 	 *	default.
1539 	 */
1540 	sock = sock_alloc();
1541 	if (!sock) {
1542 		net_warn_ratelimited("socket: no more sockets\n");
1543 		return -ENFILE;	/* Not exactly a match, but its the
1544 				   closest posix thing */
1545 	}
1546 
1547 	sock->type = type;
1548 
1549 #ifdef CONFIG_MODULES
1550 	/* Attempt to load a protocol module if the find failed.
1551 	 *
1552 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1553 	 * requested real, full-featured networking support upon configuration.
1554 	 * Otherwise module support will break!
1555 	 */
1556 	if (rcu_access_pointer(net_families[family]) == NULL)
1557 		request_module("net-pf-%d", family);
1558 #endif
1559 
1560 	rcu_read_lock();
1561 	pf = rcu_dereference(net_families[family]);
1562 	err = -EAFNOSUPPORT;
1563 	if (!pf)
1564 		goto out_release;
1565 
1566 	/*
1567 	 * We will call the ->create function, that possibly is in a loadable
1568 	 * module, so we have to bump that loadable module refcnt first.
1569 	 */
1570 	if (!try_module_get(pf->owner))
1571 		goto out_release;
1572 
1573 	/* Now protected by module ref count */
1574 	rcu_read_unlock();
1575 
1576 	err = pf->create(net, sock, protocol, kern);
1577 	if (err < 0)
1578 		goto out_module_put;
1579 
1580 	/*
1581 	 * Now to bump the refcnt of the [loadable] module that owns this
1582 	 * socket at sock_release time we decrement its refcnt.
1583 	 */
1584 	if (!try_module_get(sock->ops->owner))
1585 		goto out_module_busy;
1586 
1587 	/*
1588 	 * Now that we're done with the ->create function, the [loadable]
1589 	 * module can have its refcnt decremented
1590 	 */
1591 	module_put(pf->owner);
1592 	err = security_socket_post_create(sock, family, type, protocol, kern);
1593 	if (err)
1594 		goto out_sock_release;
1595 	*res = sock;
1596 
1597 	return 0;
1598 
1599 out_module_busy:
1600 	err = -EAFNOSUPPORT;
1601 out_module_put:
1602 	sock->ops = NULL;
1603 	module_put(pf->owner);
1604 out_sock_release:
1605 	sock_release(sock);
1606 	return err;
1607 
1608 out_release:
1609 	rcu_read_unlock();
1610 	goto out_sock_release;
1611 }
1612 EXPORT_SYMBOL(__sock_create);
1613 
1614 /**
1615  *	sock_create - creates a socket
1616  *	@family: protocol family (AF_INET, ...)
1617  *	@type: communication type (SOCK_STREAM, ...)
1618  *	@protocol: protocol (0, ...)
1619  *	@res: new socket
1620  *
1621  *	A wrapper around __sock_create().
1622  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1623  */
1624 
1625 int sock_create(int family, int type, int protocol, struct socket **res)
1626 {
1627 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1628 }
1629 EXPORT_SYMBOL(sock_create);
1630 
1631 /**
1632  *	sock_create_kern - creates a socket (kernel space)
1633  *	@net: net namespace
1634  *	@family: protocol family (AF_INET, ...)
1635  *	@type: communication type (SOCK_STREAM, ...)
1636  *	@protocol: protocol (0, ...)
1637  *	@res: new socket
1638  *
1639  *	A wrapper around __sock_create().
1640  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1641  */
1642 
1643 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1644 {
1645 	return __sock_create(net, family, type, protocol, res, 1);
1646 }
1647 EXPORT_SYMBOL(sock_create_kern);
1648 
1649 static struct socket *__sys_socket_create(int family, int type, int protocol)
1650 {
1651 	struct socket *sock;
1652 	int retval;
1653 
1654 	/* Check the SOCK_* constants for consistency.  */
1655 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1656 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1657 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1658 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1659 
1660 	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1661 		return ERR_PTR(-EINVAL);
1662 	type &= SOCK_TYPE_MASK;
1663 
1664 	retval = sock_create(family, type, protocol, &sock);
1665 	if (retval < 0)
1666 		return ERR_PTR(retval);
1667 
1668 	return sock;
1669 }
1670 
1671 struct file *__sys_socket_file(int family, int type, int protocol)
1672 {
1673 	struct socket *sock;
1674 	int flags;
1675 
1676 	sock = __sys_socket_create(family, type, protocol);
1677 	if (IS_ERR(sock))
1678 		return ERR_CAST(sock);
1679 
1680 	flags = type & ~SOCK_TYPE_MASK;
1681 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1682 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1683 
1684 	return sock_alloc_file(sock, flags, NULL);
1685 }
1686 
1687 /*	A hook for bpf progs to attach to and update socket protocol.
1688  *
1689  *	A static noinline declaration here could cause the compiler to
1690  *	optimize away the function. A global noinline declaration will
1691  *	keep the definition, but may optimize away the callsite.
1692  *	Therefore, __weak is needed to ensure that the call is still
1693  *	emitted, by telling the compiler that we don't know what the
1694  *	function might eventually be.
1695  */
1696 
1697 __bpf_hook_start();
1698 
1699 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1700 {
1701 	return protocol;
1702 }
1703 
1704 __bpf_hook_end();
1705 
1706 int __sys_socket(int family, int type, int protocol)
1707 {
1708 	struct socket *sock;
1709 	int flags;
1710 
1711 	sock = __sys_socket_create(family, type,
1712 				   update_socket_protocol(family, type, protocol));
1713 	if (IS_ERR(sock))
1714 		return PTR_ERR(sock);
1715 
1716 	flags = type & ~SOCK_TYPE_MASK;
1717 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1718 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1719 
1720 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1721 }
1722 
1723 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1724 {
1725 	return __sys_socket(family, type, protocol);
1726 }
1727 
1728 /*
1729  *	Create a pair of connected sockets.
1730  */
1731 
1732 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1733 {
1734 	struct socket *sock1, *sock2;
1735 	int fd1, fd2, err;
1736 	struct file *newfile1, *newfile2;
1737 	int flags;
1738 
1739 	flags = type & ~SOCK_TYPE_MASK;
1740 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1741 		return -EINVAL;
1742 	type &= SOCK_TYPE_MASK;
1743 
1744 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1745 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1746 
1747 	/*
1748 	 * reserve descriptors and make sure we won't fail
1749 	 * to return them to userland.
1750 	 */
1751 	fd1 = get_unused_fd_flags(flags);
1752 	if (unlikely(fd1 < 0))
1753 		return fd1;
1754 
1755 	fd2 = get_unused_fd_flags(flags);
1756 	if (unlikely(fd2 < 0)) {
1757 		put_unused_fd(fd1);
1758 		return fd2;
1759 	}
1760 
1761 	err = put_user(fd1, &usockvec[0]);
1762 	if (err)
1763 		goto out;
1764 
1765 	err = put_user(fd2, &usockvec[1]);
1766 	if (err)
1767 		goto out;
1768 
1769 	/*
1770 	 * Obtain the first socket and check if the underlying protocol
1771 	 * supports the socketpair call.
1772 	 */
1773 
1774 	err = sock_create(family, type, protocol, &sock1);
1775 	if (unlikely(err < 0))
1776 		goto out;
1777 
1778 	err = sock_create(family, type, protocol, &sock2);
1779 	if (unlikely(err < 0)) {
1780 		sock_release(sock1);
1781 		goto out;
1782 	}
1783 
1784 	err = security_socket_socketpair(sock1, sock2);
1785 	if (unlikely(err)) {
1786 		sock_release(sock2);
1787 		sock_release(sock1);
1788 		goto out;
1789 	}
1790 
1791 	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1792 	if (unlikely(err < 0)) {
1793 		sock_release(sock2);
1794 		sock_release(sock1);
1795 		goto out;
1796 	}
1797 
1798 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1799 	if (IS_ERR(newfile1)) {
1800 		err = PTR_ERR(newfile1);
1801 		sock_release(sock2);
1802 		goto out;
1803 	}
1804 
1805 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1806 	if (IS_ERR(newfile2)) {
1807 		err = PTR_ERR(newfile2);
1808 		fput(newfile1);
1809 		goto out;
1810 	}
1811 
1812 	audit_fd_pair(fd1, fd2);
1813 
1814 	fd_install(fd1, newfile1);
1815 	fd_install(fd2, newfile2);
1816 	return 0;
1817 
1818 out:
1819 	put_unused_fd(fd2);
1820 	put_unused_fd(fd1);
1821 	return err;
1822 }
1823 
1824 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1825 		int __user *, usockvec)
1826 {
1827 	return __sys_socketpair(family, type, protocol, usockvec);
1828 }
1829 
1830 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1831 		      int addrlen)
1832 {
1833 	int err;
1834 
1835 	err = security_socket_bind(sock, (struct sockaddr *)address,
1836 				   addrlen);
1837 	if (!err)
1838 		err = READ_ONCE(sock->ops)->bind(sock,
1839 						 (struct sockaddr *)address,
1840 						 addrlen);
1841 	return err;
1842 }
1843 
1844 /*
1845  *	Bind a name to a socket. Nothing much to do here since it's
1846  *	the protocol's responsibility to handle the local address.
1847  *
1848  *	We move the socket address to kernel space before we call
1849  *	the protocol layer (having also checked the address is ok).
1850  */
1851 
1852 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1853 {
1854 	struct socket *sock;
1855 	struct sockaddr_storage address;
1856 	int err, fput_needed;
1857 
1858 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1859 	if (sock) {
1860 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1861 		if (!err)
1862 			err = __sys_bind_socket(sock, &address, addrlen);
1863 		fput_light(sock->file, fput_needed);
1864 	}
1865 	return err;
1866 }
1867 
1868 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1869 {
1870 	return __sys_bind(fd, umyaddr, addrlen);
1871 }
1872 
1873 /*
1874  *	Perform a listen. Basically, we allow the protocol to do anything
1875  *	necessary for a listen, and if that works, we mark the socket as
1876  *	ready for listening.
1877  */
1878 int __sys_listen_socket(struct socket *sock, int backlog)
1879 {
1880 	int somaxconn, err;
1881 
1882 	somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1883 	if ((unsigned int)backlog > somaxconn)
1884 		backlog = somaxconn;
1885 
1886 	err = security_socket_listen(sock, backlog);
1887 	if (!err)
1888 		err = READ_ONCE(sock->ops)->listen(sock, backlog);
1889 	return err;
1890 }
1891 
1892 int __sys_listen(int fd, int backlog)
1893 {
1894 	struct socket *sock;
1895 	int err, fput_needed;
1896 
1897 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1898 	if (sock) {
1899 		err = __sys_listen_socket(sock, backlog);
1900 		fput_light(sock->file, fput_needed);
1901 	}
1902 	return err;
1903 }
1904 
1905 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1906 {
1907 	return __sys_listen(fd, backlog);
1908 }
1909 
1910 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1911 		       struct sockaddr __user *upeer_sockaddr,
1912 		       int __user *upeer_addrlen, int flags)
1913 {
1914 	struct socket *sock, *newsock;
1915 	struct file *newfile;
1916 	int err, len;
1917 	struct sockaddr_storage address;
1918 	const struct proto_ops *ops;
1919 
1920 	sock = sock_from_file(file);
1921 	if (!sock)
1922 		return ERR_PTR(-ENOTSOCK);
1923 
1924 	newsock = sock_alloc();
1925 	if (!newsock)
1926 		return ERR_PTR(-ENFILE);
1927 	ops = READ_ONCE(sock->ops);
1928 
1929 	newsock->type = sock->type;
1930 	newsock->ops = ops;
1931 
1932 	/*
1933 	 * We don't need try_module_get here, as the listening socket (sock)
1934 	 * has the protocol module (sock->ops->owner) held.
1935 	 */
1936 	__module_get(ops->owner);
1937 
1938 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1939 	if (IS_ERR(newfile))
1940 		return newfile;
1941 
1942 	err = security_socket_accept(sock, newsock);
1943 	if (err)
1944 		goto out_fd;
1945 
1946 	arg->flags |= sock->file->f_flags;
1947 	err = ops->accept(sock, newsock, arg);
1948 	if (err < 0)
1949 		goto out_fd;
1950 
1951 	if (upeer_sockaddr) {
1952 		len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1953 		if (len < 0) {
1954 			err = -ECONNABORTED;
1955 			goto out_fd;
1956 		}
1957 		err = move_addr_to_user(&address,
1958 					len, upeer_sockaddr, upeer_addrlen);
1959 		if (err < 0)
1960 			goto out_fd;
1961 	}
1962 
1963 	/* File flags are not inherited via accept() unlike another OSes. */
1964 	return newfile;
1965 out_fd:
1966 	fput(newfile);
1967 	return ERR_PTR(err);
1968 }
1969 
1970 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1971 			      int __user *upeer_addrlen, int flags)
1972 {
1973 	struct proto_accept_arg arg = { };
1974 	struct file *newfile;
1975 	int newfd;
1976 
1977 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1978 		return -EINVAL;
1979 
1980 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1981 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1982 
1983 	newfd = get_unused_fd_flags(flags);
1984 	if (unlikely(newfd < 0))
1985 		return newfd;
1986 
1987 	newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1988 			    flags);
1989 	if (IS_ERR(newfile)) {
1990 		put_unused_fd(newfd);
1991 		return PTR_ERR(newfile);
1992 	}
1993 	fd_install(newfd, newfile);
1994 	return newfd;
1995 }
1996 
1997 /*
1998  *	For accept, we attempt to create a new socket, set up the link
1999  *	with the client, wake up the client, then return the new
2000  *	connected fd. We collect the address of the connector in kernel
2001  *	space and move it to user at the very end. This is unclean because
2002  *	we open the socket then return an error.
2003  *
2004  *	1003.1g adds the ability to recvmsg() to query connection pending
2005  *	status to recvmsg. We need to add that support in a way thats
2006  *	clean when we restructure accept also.
2007  */
2008 
2009 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2010 		  int __user *upeer_addrlen, int flags)
2011 {
2012 	int ret = -EBADF;
2013 	struct fd f;
2014 
2015 	f = fdget(fd);
2016 	if (fd_file(f)) {
2017 		ret = __sys_accept4_file(fd_file(f), upeer_sockaddr,
2018 					 upeer_addrlen, flags);
2019 		fdput(f);
2020 	}
2021 
2022 	return ret;
2023 }
2024 
2025 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2026 		int __user *, upeer_addrlen, int, flags)
2027 {
2028 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2029 }
2030 
2031 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2032 		int __user *, upeer_addrlen)
2033 {
2034 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2035 }
2036 
2037 /*
2038  *	Attempt to connect to a socket with the server address.  The address
2039  *	is in user space so we verify it is OK and move it to kernel space.
2040  *
2041  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2042  *	break bindings
2043  *
2044  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2045  *	other SEQPACKET protocols that take time to connect() as it doesn't
2046  *	include the -EINPROGRESS status for such sockets.
2047  */
2048 
2049 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2050 		       int addrlen, int file_flags)
2051 {
2052 	struct socket *sock;
2053 	int err;
2054 
2055 	sock = sock_from_file(file);
2056 	if (!sock) {
2057 		err = -ENOTSOCK;
2058 		goto out;
2059 	}
2060 
2061 	err =
2062 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2063 	if (err)
2064 		goto out;
2065 
2066 	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2067 				addrlen, sock->file->f_flags | file_flags);
2068 out:
2069 	return err;
2070 }
2071 
2072 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2073 {
2074 	int ret = -EBADF;
2075 	struct fd f;
2076 
2077 	f = fdget(fd);
2078 	if (fd_file(f)) {
2079 		struct sockaddr_storage address;
2080 
2081 		ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2082 		if (!ret)
2083 			ret = __sys_connect_file(fd_file(f), &address, addrlen, 0);
2084 		fdput(f);
2085 	}
2086 
2087 	return ret;
2088 }
2089 
2090 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2091 		int, addrlen)
2092 {
2093 	return __sys_connect(fd, uservaddr, addrlen);
2094 }
2095 
2096 /*
2097  *	Get the local address ('name') of a socket object. Move the obtained
2098  *	name to user space.
2099  */
2100 
2101 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2102 		      int __user *usockaddr_len)
2103 {
2104 	struct socket *sock;
2105 	struct sockaddr_storage address;
2106 	int err, fput_needed;
2107 
2108 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2109 	if (!sock)
2110 		goto out;
2111 
2112 	err = security_socket_getsockname(sock);
2113 	if (err)
2114 		goto out_put;
2115 
2116 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2117 	if (err < 0)
2118 		goto out_put;
2119 	/* "err" is actually length in this case */
2120 	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2121 
2122 out_put:
2123 	fput_light(sock->file, fput_needed);
2124 out:
2125 	return err;
2126 }
2127 
2128 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2129 		int __user *, usockaddr_len)
2130 {
2131 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2132 }
2133 
2134 /*
2135  *	Get the remote address ('name') of a socket object. Move the obtained
2136  *	name to user space.
2137  */
2138 
2139 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2140 		      int __user *usockaddr_len)
2141 {
2142 	struct socket *sock;
2143 	struct sockaddr_storage address;
2144 	int err, fput_needed;
2145 
2146 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2147 	if (sock != NULL) {
2148 		const struct proto_ops *ops = READ_ONCE(sock->ops);
2149 
2150 		err = security_socket_getpeername(sock);
2151 		if (err) {
2152 			fput_light(sock->file, fput_needed);
2153 			return err;
2154 		}
2155 
2156 		err = ops->getname(sock, (struct sockaddr *)&address, 1);
2157 		if (err >= 0)
2158 			/* "err" is actually length in this case */
2159 			err = move_addr_to_user(&address, err, usockaddr,
2160 						usockaddr_len);
2161 		fput_light(sock->file, fput_needed);
2162 	}
2163 	return err;
2164 }
2165 
2166 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2167 		int __user *, usockaddr_len)
2168 {
2169 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2170 }
2171 
2172 /*
2173  *	Send a datagram to a given address. We move the address into kernel
2174  *	space and check the user space data area is readable before invoking
2175  *	the protocol.
2176  */
2177 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2178 		 struct sockaddr __user *addr,  int addr_len)
2179 {
2180 	struct socket *sock;
2181 	struct sockaddr_storage address;
2182 	int err;
2183 	struct msghdr msg;
2184 	int fput_needed;
2185 
2186 	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2187 	if (unlikely(err))
2188 		return err;
2189 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2190 	if (!sock)
2191 		goto out;
2192 
2193 	msg.msg_name = NULL;
2194 	msg.msg_control = NULL;
2195 	msg.msg_controllen = 0;
2196 	msg.msg_namelen = 0;
2197 	msg.msg_ubuf = NULL;
2198 	if (addr) {
2199 		err = move_addr_to_kernel(addr, addr_len, &address);
2200 		if (err < 0)
2201 			goto out_put;
2202 		msg.msg_name = (struct sockaddr *)&address;
2203 		msg.msg_namelen = addr_len;
2204 	}
2205 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2206 	if (sock->file->f_flags & O_NONBLOCK)
2207 		flags |= MSG_DONTWAIT;
2208 	msg.msg_flags = flags;
2209 	err = __sock_sendmsg(sock, &msg);
2210 
2211 out_put:
2212 	fput_light(sock->file, fput_needed);
2213 out:
2214 	return err;
2215 }
2216 
2217 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2218 		unsigned int, flags, struct sockaddr __user *, addr,
2219 		int, addr_len)
2220 {
2221 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2222 }
2223 
2224 /*
2225  *	Send a datagram down a socket.
2226  */
2227 
2228 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2229 		unsigned int, flags)
2230 {
2231 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2232 }
2233 
2234 /*
2235  *	Receive a frame from the socket and optionally record the address of the
2236  *	sender. We verify the buffers are writable and if needed move the
2237  *	sender address from kernel to user space.
2238  */
2239 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2240 		   struct sockaddr __user *addr, int __user *addr_len)
2241 {
2242 	struct sockaddr_storage address;
2243 	struct msghdr msg = {
2244 		/* Save some cycles and don't copy the address if not needed */
2245 		.msg_name = addr ? (struct sockaddr *)&address : NULL,
2246 	};
2247 	struct socket *sock;
2248 	int err, err2;
2249 	int fput_needed;
2250 
2251 	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2252 	if (unlikely(err))
2253 		return err;
2254 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2255 	if (!sock)
2256 		goto out;
2257 
2258 	if (sock->file->f_flags & O_NONBLOCK)
2259 		flags |= MSG_DONTWAIT;
2260 	err = sock_recvmsg(sock, &msg, flags);
2261 
2262 	if (err >= 0 && addr != NULL) {
2263 		err2 = move_addr_to_user(&address,
2264 					 msg.msg_namelen, addr, addr_len);
2265 		if (err2 < 0)
2266 			err = err2;
2267 	}
2268 
2269 	fput_light(sock->file, fput_needed);
2270 out:
2271 	return err;
2272 }
2273 
2274 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2275 		unsigned int, flags, struct sockaddr __user *, addr,
2276 		int __user *, addr_len)
2277 {
2278 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2279 }
2280 
2281 /*
2282  *	Receive a datagram from a socket.
2283  */
2284 
2285 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2286 		unsigned int, flags)
2287 {
2288 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2289 }
2290 
2291 static bool sock_use_custom_sol_socket(const struct socket *sock)
2292 {
2293 	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2294 }
2295 
2296 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2297 		       int optname, sockptr_t optval, int optlen)
2298 {
2299 	const struct proto_ops *ops;
2300 	char *kernel_optval = NULL;
2301 	int err;
2302 
2303 	if (optlen < 0)
2304 		return -EINVAL;
2305 
2306 	err = security_socket_setsockopt(sock, level, optname);
2307 	if (err)
2308 		goto out_put;
2309 
2310 	if (!compat)
2311 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2312 						     optval, &optlen,
2313 						     &kernel_optval);
2314 	if (err < 0)
2315 		goto out_put;
2316 	if (err > 0) {
2317 		err = 0;
2318 		goto out_put;
2319 	}
2320 
2321 	if (kernel_optval)
2322 		optval = KERNEL_SOCKPTR(kernel_optval);
2323 	ops = READ_ONCE(sock->ops);
2324 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2325 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2326 	else if (unlikely(!ops->setsockopt))
2327 		err = -EOPNOTSUPP;
2328 	else
2329 		err = ops->setsockopt(sock, level, optname, optval,
2330 					    optlen);
2331 	kfree(kernel_optval);
2332 out_put:
2333 	return err;
2334 }
2335 EXPORT_SYMBOL(do_sock_setsockopt);
2336 
2337 /* Set a socket option. Because we don't know the option lengths we have
2338  * to pass the user mode parameter for the protocols to sort out.
2339  */
2340 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2341 		     int optlen)
2342 {
2343 	sockptr_t optval = USER_SOCKPTR(user_optval);
2344 	bool compat = in_compat_syscall();
2345 	int err, fput_needed;
2346 	struct socket *sock;
2347 
2348 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2349 	if (!sock)
2350 		return err;
2351 
2352 	err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2353 
2354 	fput_light(sock->file, fput_needed);
2355 	return err;
2356 }
2357 
2358 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2359 		char __user *, optval, int, optlen)
2360 {
2361 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2362 }
2363 
2364 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2365 							 int optname));
2366 
2367 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2368 		       int optname, sockptr_t optval, sockptr_t optlen)
2369 {
2370 	int max_optlen __maybe_unused = 0;
2371 	const struct proto_ops *ops;
2372 	int err;
2373 
2374 	err = security_socket_getsockopt(sock, level, optname);
2375 	if (err)
2376 		return err;
2377 
2378 	if (!compat)
2379 		copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2380 
2381 	ops = READ_ONCE(sock->ops);
2382 	if (level == SOL_SOCKET) {
2383 		err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2384 	} else if (unlikely(!ops->getsockopt)) {
2385 		err = -EOPNOTSUPP;
2386 	} else {
2387 		if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2388 			      "Invalid argument type"))
2389 			return -EOPNOTSUPP;
2390 
2391 		err = ops->getsockopt(sock, level, optname, optval.user,
2392 				      optlen.user);
2393 	}
2394 
2395 	if (!compat)
2396 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2397 						     optval, optlen, max_optlen,
2398 						     err);
2399 
2400 	return err;
2401 }
2402 EXPORT_SYMBOL(do_sock_getsockopt);
2403 
2404 /*
2405  *	Get a socket option. Because we don't know the option lengths we have
2406  *	to pass a user mode parameter for the protocols to sort out.
2407  */
2408 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2409 		int __user *optlen)
2410 {
2411 	int err, fput_needed;
2412 	struct socket *sock;
2413 	bool compat;
2414 
2415 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2416 	if (!sock)
2417 		return err;
2418 
2419 	compat = in_compat_syscall();
2420 	err = do_sock_getsockopt(sock, compat, level, optname,
2421 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2422 
2423 	fput_light(sock->file, fput_needed);
2424 	return err;
2425 }
2426 
2427 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2428 		char __user *, optval, int __user *, optlen)
2429 {
2430 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2431 }
2432 
2433 /*
2434  *	Shutdown a socket.
2435  */
2436 
2437 int __sys_shutdown_sock(struct socket *sock, int how)
2438 {
2439 	int err;
2440 
2441 	err = security_socket_shutdown(sock, how);
2442 	if (!err)
2443 		err = READ_ONCE(sock->ops)->shutdown(sock, how);
2444 
2445 	return err;
2446 }
2447 
2448 int __sys_shutdown(int fd, int how)
2449 {
2450 	int err, fput_needed;
2451 	struct socket *sock;
2452 
2453 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2454 	if (sock != NULL) {
2455 		err = __sys_shutdown_sock(sock, how);
2456 		fput_light(sock->file, fput_needed);
2457 	}
2458 	return err;
2459 }
2460 
2461 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2462 {
2463 	return __sys_shutdown(fd, how);
2464 }
2465 
2466 /* A couple of helpful macros for getting the address of the 32/64 bit
2467  * fields which are the same type (int / unsigned) on our platforms.
2468  */
2469 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2470 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2471 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2472 
2473 struct used_address {
2474 	struct sockaddr_storage name;
2475 	unsigned int name_len;
2476 };
2477 
2478 int __copy_msghdr(struct msghdr *kmsg,
2479 		  struct user_msghdr *msg,
2480 		  struct sockaddr __user **save_addr)
2481 {
2482 	ssize_t err;
2483 
2484 	kmsg->msg_control_is_user = true;
2485 	kmsg->msg_get_inq = 0;
2486 	kmsg->msg_control_user = msg->msg_control;
2487 	kmsg->msg_controllen = msg->msg_controllen;
2488 	kmsg->msg_flags = msg->msg_flags;
2489 
2490 	kmsg->msg_namelen = msg->msg_namelen;
2491 	if (!msg->msg_name)
2492 		kmsg->msg_namelen = 0;
2493 
2494 	if (kmsg->msg_namelen < 0)
2495 		return -EINVAL;
2496 
2497 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2498 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2499 
2500 	if (save_addr)
2501 		*save_addr = msg->msg_name;
2502 
2503 	if (msg->msg_name && kmsg->msg_namelen) {
2504 		if (!save_addr) {
2505 			err = move_addr_to_kernel(msg->msg_name,
2506 						  kmsg->msg_namelen,
2507 						  kmsg->msg_name);
2508 			if (err < 0)
2509 				return err;
2510 		}
2511 	} else {
2512 		kmsg->msg_name = NULL;
2513 		kmsg->msg_namelen = 0;
2514 	}
2515 
2516 	if (msg->msg_iovlen > UIO_MAXIOV)
2517 		return -EMSGSIZE;
2518 
2519 	kmsg->msg_iocb = NULL;
2520 	kmsg->msg_ubuf = NULL;
2521 	return 0;
2522 }
2523 
2524 static int copy_msghdr_from_user(struct msghdr *kmsg,
2525 				 struct user_msghdr __user *umsg,
2526 				 struct sockaddr __user **save_addr,
2527 				 struct iovec **iov)
2528 {
2529 	struct user_msghdr msg;
2530 	ssize_t err;
2531 
2532 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2533 		return -EFAULT;
2534 
2535 	err = __copy_msghdr(kmsg, &msg, save_addr);
2536 	if (err)
2537 		return err;
2538 
2539 	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2540 			    msg.msg_iov, msg.msg_iovlen,
2541 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2542 	return err < 0 ? err : 0;
2543 }
2544 
2545 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2546 			   unsigned int flags, struct used_address *used_address,
2547 			   unsigned int allowed_msghdr_flags)
2548 {
2549 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2550 				__aligned(sizeof(__kernel_size_t));
2551 	/* 20 is size of ipv6_pktinfo */
2552 	unsigned char *ctl_buf = ctl;
2553 	int ctl_len;
2554 	ssize_t err;
2555 
2556 	err = -ENOBUFS;
2557 
2558 	if (msg_sys->msg_controllen > INT_MAX)
2559 		goto out;
2560 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2561 	ctl_len = msg_sys->msg_controllen;
2562 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2563 		err =
2564 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2565 						     sizeof(ctl));
2566 		if (err)
2567 			goto out;
2568 		ctl_buf = msg_sys->msg_control;
2569 		ctl_len = msg_sys->msg_controllen;
2570 	} else if (ctl_len) {
2571 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2572 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2573 		if (ctl_len > sizeof(ctl)) {
2574 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2575 			if (ctl_buf == NULL)
2576 				goto out;
2577 		}
2578 		err = -EFAULT;
2579 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2580 			goto out_freectl;
2581 		msg_sys->msg_control = ctl_buf;
2582 		msg_sys->msg_control_is_user = false;
2583 	}
2584 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2585 	msg_sys->msg_flags = flags;
2586 
2587 	if (sock->file->f_flags & O_NONBLOCK)
2588 		msg_sys->msg_flags |= MSG_DONTWAIT;
2589 	/*
2590 	 * If this is sendmmsg() and current destination address is same as
2591 	 * previously succeeded address, omit asking LSM's decision.
2592 	 * used_address->name_len is initialized to UINT_MAX so that the first
2593 	 * destination address never matches.
2594 	 */
2595 	if (used_address && msg_sys->msg_name &&
2596 	    used_address->name_len == msg_sys->msg_namelen &&
2597 	    !memcmp(&used_address->name, msg_sys->msg_name,
2598 		    used_address->name_len)) {
2599 		err = sock_sendmsg_nosec(sock, msg_sys);
2600 		goto out_freectl;
2601 	}
2602 	err = __sock_sendmsg(sock, msg_sys);
2603 	/*
2604 	 * If this is sendmmsg() and sending to current destination address was
2605 	 * successful, remember it.
2606 	 */
2607 	if (used_address && err >= 0) {
2608 		used_address->name_len = msg_sys->msg_namelen;
2609 		if (msg_sys->msg_name)
2610 			memcpy(&used_address->name, msg_sys->msg_name,
2611 			       used_address->name_len);
2612 	}
2613 
2614 out_freectl:
2615 	if (ctl_buf != ctl)
2616 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2617 out:
2618 	return err;
2619 }
2620 
2621 static int sendmsg_copy_msghdr(struct msghdr *msg,
2622 			       struct user_msghdr __user *umsg, unsigned flags,
2623 			       struct iovec **iov)
2624 {
2625 	int err;
2626 
2627 	if (flags & MSG_CMSG_COMPAT) {
2628 		struct compat_msghdr __user *msg_compat;
2629 
2630 		msg_compat = (struct compat_msghdr __user *) umsg;
2631 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2632 	} else {
2633 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2634 	}
2635 	if (err < 0)
2636 		return err;
2637 
2638 	return 0;
2639 }
2640 
2641 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2642 			 struct msghdr *msg_sys, unsigned int flags,
2643 			 struct used_address *used_address,
2644 			 unsigned int allowed_msghdr_flags)
2645 {
2646 	struct sockaddr_storage address;
2647 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2648 	ssize_t err;
2649 
2650 	msg_sys->msg_name = &address;
2651 
2652 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2653 	if (err < 0)
2654 		return err;
2655 
2656 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2657 				allowed_msghdr_flags);
2658 	kfree(iov);
2659 	return err;
2660 }
2661 
2662 /*
2663  *	BSD sendmsg interface
2664  */
2665 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2666 			unsigned int flags)
2667 {
2668 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2669 }
2670 
2671 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2672 		   bool forbid_cmsg_compat)
2673 {
2674 	int fput_needed, err;
2675 	struct msghdr msg_sys;
2676 	struct socket *sock;
2677 
2678 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2679 		return -EINVAL;
2680 
2681 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2682 	if (!sock)
2683 		goto out;
2684 
2685 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2686 
2687 	fput_light(sock->file, fput_needed);
2688 out:
2689 	return err;
2690 }
2691 
2692 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2693 {
2694 	return __sys_sendmsg(fd, msg, flags, true);
2695 }
2696 
2697 /*
2698  *	Linux sendmmsg interface
2699  */
2700 
2701 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2702 		   unsigned int flags, bool forbid_cmsg_compat)
2703 {
2704 	int fput_needed, err, datagrams;
2705 	struct socket *sock;
2706 	struct mmsghdr __user *entry;
2707 	struct compat_mmsghdr __user *compat_entry;
2708 	struct msghdr msg_sys;
2709 	struct used_address used_address;
2710 	unsigned int oflags = flags;
2711 
2712 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2713 		return -EINVAL;
2714 
2715 	if (vlen > UIO_MAXIOV)
2716 		vlen = UIO_MAXIOV;
2717 
2718 	datagrams = 0;
2719 
2720 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2721 	if (!sock)
2722 		return err;
2723 
2724 	used_address.name_len = UINT_MAX;
2725 	entry = mmsg;
2726 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2727 	err = 0;
2728 	flags |= MSG_BATCH;
2729 
2730 	while (datagrams < vlen) {
2731 		if (datagrams == vlen - 1)
2732 			flags = oflags;
2733 
2734 		if (MSG_CMSG_COMPAT & flags) {
2735 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2736 					     &msg_sys, flags, &used_address, MSG_EOR);
2737 			if (err < 0)
2738 				break;
2739 			err = __put_user(err, &compat_entry->msg_len);
2740 			++compat_entry;
2741 		} else {
2742 			err = ___sys_sendmsg(sock,
2743 					     (struct user_msghdr __user *)entry,
2744 					     &msg_sys, flags, &used_address, MSG_EOR);
2745 			if (err < 0)
2746 				break;
2747 			err = put_user(err, &entry->msg_len);
2748 			++entry;
2749 		}
2750 
2751 		if (err)
2752 			break;
2753 		++datagrams;
2754 		if (msg_data_left(&msg_sys))
2755 			break;
2756 		cond_resched();
2757 	}
2758 
2759 	fput_light(sock->file, fput_needed);
2760 
2761 	/* We only return an error if no datagrams were able to be sent */
2762 	if (datagrams != 0)
2763 		return datagrams;
2764 
2765 	return err;
2766 }
2767 
2768 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2769 		unsigned int, vlen, unsigned int, flags)
2770 {
2771 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2772 }
2773 
2774 static int recvmsg_copy_msghdr(struct msghdr *msg,
2775 			       struct user_msghdr __user *umsg, unsigned flags,
2776 			       struct sockaddr __user **uaddr,
2777 			       struct iovec **iov)
2778 {
2779 	ssize_t err;
2780 
2781 	if (MSG_CMSG_COMPAT & flags) {
2782 		struct compat_msghdr __user *msg_compat;
2783 
2784 		msg_compat = (struct compat_msghdr __user *) umsg;
2785 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2786 	} else {
2787 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2788 	}
2789 	if (err < 0)
2790 		return err;
2791 
2792 	return 0;
2793 }
2794 
2795 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2796 			   struct user_msghdr __user *msg,
2797 			   struct sockaddr __user *uaddr,
2798 			   unsigned int flags, int nosec)
2799 {
2800 	struct compat_msghdr __user *msg_compat =
2801 					(struct compat_msghdr __user *) msg;
2802 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2803 	struct sockaddr_storage addr;
2804 	unsigned long cmsg_ptr;
2805 	int len;
2806 	ssize_t err;
2807 
2808 	msg_sys->msg_name = &addr;
2809 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2810 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2811 
2812 	/* We assume all kernel code knows the size of sockaddr_storage */
2813 	msg_sys->msg_namelen = 0;
2814 
2815 	if (sock->file->f_flags & O_NONBLOCK)
2816 		flags |= MSG_DONTWAIT;
2817 
2818 	if (unlikely(nosec))
2819 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2820 	else
2821 		err = sock_recvmsg(sock, msg_sys, flags);
2822 
2823 	if (err < 0)
2824 		goto out;
2825 	len = err;
2826 
2827 	if (uaddr != NULL) {
2828 		err = move_addr_to_user(&addr,
2829 					msg_sys->msg_namelen, uaddr,
2830 					uaddr_len);
2831 		if (err < 0)
2832 			goto out;
2833 	}
2834 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2835 			 COMPAT_FLAGS(msg));
2836 	if (err)
2837 		goto out;
2838 	if (MSG_CMSG_COMPAT & flags)
2839 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2840 				 &msg_compat->msg_controllen);
2841 	else
2842 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2843 				 &msg->msg_controllen);
2844 	if (err)
2845 		goto out;
2846 	err = len;
2847 out:
2848 	return err;
2849 }
2850 
2851 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2852 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2853 {
2854 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2855 	/* user mode address pointers */
2856 	struct sockaddr __user *uaddr;
2857 	ssize_t err;
2858 
2859 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2860 	if (err < 0)
2861 		return err;
2862 
2863 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2864 	kfree(iov);
2865 	return err;
2866 }
2867 
2868 /*
2869  *	BSD recvmsg interface
2870  */
2871 
2872 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2873 			struct user_msghdr __user *umsg,
2874 			struct sockaddr __user *uaddr, unsigned int flags)
2875 {
2876 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2877 }
2878 
2879 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2880 		   bool forbid_cmsg_compat)
2881 {
2882 	int fput_needed, err;
2883 	struct msghdr msg_sys;
2884 	struct socket *sock;
2885 
2886 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2887 		return -EINVAL;
2888 
2889 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2890 	if (!sock)
2891 		goto out;
2892 
2893 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2894 
2895 	fput_light(sock->file, fput_needed);
2896 out:
2897 	return err;
2898 }
2899 
2900 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2901 		unsigned int, flags)
2902 {
2903 	return __sys_recvmsg(fd, msg, flags, true);
2904 }
2905 
2906 /*
2907  *     Linux recvmmsg interface
2908  */
2909 
2910 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2911 			  unsigned int vlen, unsigned int flags,
2912 			  struct timespec64 *timeout)
2913 {
2914 	int fput_needed, err, datagrams;
2915 	struct socket *sock;
2916 	struct mmsghdr __user *entry;
2917 	struct compat_mmsghdr __user *compat_entry;
2918 	struct msghdr msg_sys;
2919 	struct timespec64 end_time;
2920 	struct timespec64 timeout64;
2921 
2922 	if (timeout &&
2923 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2924 				    timeout->tv_nsec))
2925 		return -EINVAL;
2926 
2927 	datagrams = 0;
2928 
2929 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2930 	if (!sock)
2931 		return err;
2932 
2933 	if (likely(!(flags & MSG_ERRQUEUE))) {
2934 		err = sock_error(sock->sk);
2935 		if (err) {
2936 			datagrams = err;
2937 			goto out_put;
2938 		}
2939 	}
2940 
2941 	entry = mmsg;
2942 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2943 
2944 	while (datagrams < vlen) {
2945 		/*
2946 		 * No need to ask LSM for more than the first datagram.
2947 		 */
2948 		if (MSG_CMSG_COMPAT & flags) {
2949 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2950 					     &msg_sys, flags & ~MSG_WAITFORONE,
2951 					     datagrams);
2952 			if (err < 0)
2953 				break;
2954 			err = __put_user(err, &compat_entry->msg_len);
2955 			++compat_entry;
2956 		} else {
2957 			err = ___sys_recvmsg(sock,
2958 					     (struct user_msghdr __user *)entry,
2959 					     &msg_sys, flags & ~MSG_WAITFORONE,
2960 					     datagrams);
2961 			if (err < 0)
2962 				break;
2963 			err = put_user(err, &entry->msg_len);
2964 			++entry;
2965 		}
2966 
2967 		if (err)
2968 			break;
2969 		++datagrams;
2970 
2971 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2972 		if (flags & MSG_WAITFORONE)
2973 			flags |= MSG_DONTWAIT;
2974 
2975 		if (timeout) {
2976 			ktime_get_ts64(&timeout64);
2977 			*timeout = timespec64_sub(end_time, timeout64);
2978 			if (timeout->tv_sec < 0) {
2979 				timeout->tv_sec = timeout->tv_nsec = 0;
2980 				break;
2981 			}
2982 
2983 			/* Timeout, return less than vlen datagrams */
2984 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2985 				break;
2986 		}
2987 
2988 		/* Out of band data, return right away */
2989 		if (msg_sys.msg_flags & MSG_OOB)
2990 			break;
2991 		cond_resched();
2992 	}
2993 
2994 	if (err == 0)
2995 		goto out_put;
2996 
2997 	if (datagrams == 0) {
2998 		datagrams = err;
2999 		goto out_put;
3000 	}
3001 
3002 	/*
3003 	 * We may return less entries than requested (vlen) if the
3004 	 * sock is non block and there aren't enough datagrams...
3005 	 */
3006 	if (err != -EAGAIN) {
3007 		/*
3008 		 * ... or  if recvmsg returns an error after we
3009 		 * received some datagrams, where we record the
3010 		 * error to return on the next call or if the
3011 		 * app asks about it using getsockopt(SO_ERROR).
3012 		 */
3013 		WRITE_ONCE(sock->sk->sk_err, -err);
3014 	}
3015 out_put:
3016 	fput_light(sock->file, fput_needed);
3017 
3018 	return datagrams;
3019 }
3020 
3021 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3022 		   unsigned int vlen, unsigned int flags,
3023 		   struct __kernel_timespec __user *timeout,
3024 		   struct old_timespec32 __user *timeout32)
3025 {
3026 	int datagrams;
3027 	struct timespec64 timeout_sys;
3028 
3029 	if (timeout && get_timespec64(&timeout_sys, timeout))
3030 		return -EFAULT;
3031 
3032 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3033 		return -EFAULT;
3034 
3035 	if (!timeout && !timeout32)
3036 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3037 
3038 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3039 
3040 	if (datagrams <= 0)
3041 		return datagrams;
3042 
3043 	if (timeout && put_timespec64(&timeout_sys, timeout))
3044 		datagrams = -EFAULT;
3045 
3046 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3047 		datagrams = -EFAULT;
3048 
3049 	return datagrams;
3050 }
3051 
3052 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3053 		unsigned int, vlen, unsigned int, flags,
3054 		struct __kernel_timespec __user *, timeout)
3055 {
3056 	if (flags & MSG_CMSG_COMPAT)
3057 		return -EINVAL;
3058 
3059 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3060 }
3061 
3062 #ifdef CONFIG_COMPAT_32BIT_TIME
3063 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3064 		unsigned int, vlen, unsigned int, flags,
3065 		struct old_timespec32 __user *, timeout)
3066 {
3067 	if (flags & MSG_CMSG_COMPAT)
3068 		return -EINVAL;
3069 
3070 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3071 }
3072 #endif
3073 
3074 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3075 /* Argument list sizes for sys_socketcall */
3076 #define AL(x) ((x) * sizeof(unsigned long))
3077 static const unsigned char nargs[21] = {
3078 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3079 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3080 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3081 	AL(4), AL(5), AL(4)
3082 };
3083 
3084 #undef AL
3085 
3086 /*
3087  *	System call vectors.
3088  *
3089  *	Argument checking cleaned up. Saved 20% in size.
3090  *  This function doesn't need to set the kernel lock because
3091  *  it is set by the callees.
3092  */
3093 
3094 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3095 {
3096 	unsigned long a[AUDITSC_ARGS];
3097 	unsigned long a0, a1;
3098 	int err;
3099 	unsigned int len;
3100 
3101 	if (call < 1 || call > SYS_SENDMMSG)
3102 		return -EINVAL;
3103 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3104 
3105 	len = nargs[call];
3106 	if (len > sizeof(a))
3107 		return -EINVAL;
3108 
3109 	/* copy_from_user should be SMP safe. */
3110 	if (copy_from_user(a, args, len))
3111 		return -EFAULT;
3112 
3113 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3114 	if (err)
3115 		return err;
3116 
3117 	a0 = a[0];
3118 	a1 = a[1];
3119 
3120 	switch (call) {
3121 	case SYS_SOCKET:
3122 		err = __sys_socket(a0, a1, a[2]);
3123 		break;
3124 	case SYS_BIND:
3125 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3126 		break;
3127 	case SYS_CONNECT:
3128 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3129 		break;
3130 	case SYS_LISTEN:
3131 		err = __sys_listen(a0, a1);
3132 		break;
3133 	case SYS_ACCEPT:
3134 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3135 				    (int __user *)a[2], 0);
3136 		break;
3137 	case SYS_GETSOCKNAME:
3138 		err =
3139 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
3140 				      (int __user *)a[2]);
3141 		break;
3142 	case SYS_GETPEERNAME:
3143 		err =
3144 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
3145 				      (int __user *)a[2]);
3146 		break;
3147 	case SYS_SOCKETPAIR:
3148 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3149 		break;
3150 	case SYS_SEND:
3151 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3152 				   NULL, 0);
3153 		break;
3154 	case SYS_SENDTO:
3155 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3156 				   (struct sockaddr __user *)a[4], a[5]);
3157 		break;
3158 	case SYS_RECV:
3159 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3160 				     NULL, NULL);
3161 		break;
3162 	case SYS_RECVFROM:
3163 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3164 				     (struct sockaddr __user *)a[4],
3165 				     (int __user *)a[5]);
3166 		break;
3167 	case SYS_SHUTDOWN:
3168 		err = __sys_shutdown(a0, a1);
3169 		break;
3170 	case SYS_SETSOCKOPT:
3171 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3172 				       a[4]);
3173 		break;
3174 	case SYS_GETSOCKOPT:
3175 		err =
3176 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3177 				     (int __user *)a[4]);
3178 		break;
3179 	case SYS_SENDMSG:
3180 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3181 				    a[2], true);
3182 		break;
3183 	case SYS_SENDMMSG:
3184 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3185 				     a[3], true);
3186 		break;
3187 	case SYS_RECVMSG:
3188 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3189 				    a[2], true);
3190 		break;
3191 	case SYS_RECVMMSG:
3192 		if (IS_ENABLED(CONFIG_64BIT))
3193 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3194 					     a[2], a[3],
3195 					     (struct __kernel_timespec __user *)a[4],
3196 					     NULL);
3197 		else
3198 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3199 					     a[2], a[3], NULL,
3200 					     (struct old_timespec32 __user *)a[4]);
3201 		break;
3202 	case SYS_ACCEPT4:
3203 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3204 				    (int __user *)a[2], a[3]);
3205 		break;
3206 	default:
3207 		err = -EINVAL;
3208 		break;
3209 	}
3210 	return err;
3211 }
3212 
3213 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3214 
3215 /**
3216  *	sock_register - add a socket protocol handler
3217  *	@ops: description of protocol
3218  *
3219  *	This function is called by a protocol handler that wants to
3220  *	advertise its address family, and have it linked into the
3221  *	socket interface. The value ops->family corresponds to the
3222  *	socket system call protocol family.
3223  */
3224 int sock_register(const struct net_proto_family *ops)
3225 {
3226 	int err;
3227 
3228 	if (ops->family >= NPROTO) {
3229 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3230 		return -ENOBUFS;
3231 	}
3232 
3233 	spin_lock(&net_family_lock);
3234 	if (rcu_dereference_protected(net_families[ops->family],
3235 				      lockdep_is_held(&net_family_lock)))
3236 		err = -EEXIST;
3237 	else {
3238 		rcu_assign_pointer(net_families[ops->family], ops);
3239 		err = 0;
3240 	}
3241 	spin_unlock(&net_family_lock);
3242 
3243 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3244 	return err;
3245 }
3246 EXPORT_SYMBOL(sock_register);
3247 
3248 /**
3249  *	sock_unregister - remove a protocol handler
3250  *	@family: protocol family to remove
3251  *
3252  *	This function is called by a protocol handler that wants to
3253  *	remove its address family, and have it unlinked from the
3254  *	new socket creation.
3255  *
3256  *	If protocol handler is a module, then it can use module reference
3257  *	counts to protect against new references. If protocol handler is not
3258  *	a module then it needs to provide its own protection in
3259  *	the ops->create routine.
3260  */
3261 void sock_unregister(int family)
3262 {
3263 	BUG_ON(family < 0 || family >= NPROTO);
3264 
3265 	spin_lock(&net_family_lock);
3266 	RCU_INIT_POINTER(net_families[family], NULL);
3267 	spin_unlock(&net_family_lock);
3268 
3269 	synchronize_rcu();
3270 
3271 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3272 }
3273 EXPORT_SYMBOL(sock_unregister);
3274 
3275 bool sock_is_registered(int family)
3276 {
3277 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3278 }
3279 
3280 static int __init sock_init(void)
3281 {
3282 	int err;
3283 	/*
3284 	 *      Initialize the network sysctl infrastructure.
3285 	 */
3286 	err = net_sysctl_init();
3287 	if (err)
3288 		goto out;
3289 
3290 	/*
3291 	 *      Initialize skbuff SLAB cache
3292 	 */
3293 	skb_init();
3294 
3295 	/*
3296 	 *      Initialize the protocols module.
3297 	 */
3298 
3299 	init_inodecache();
3300 
3301 	err = register_filesystem(&sock_fs_type);
3302 	if (err)
3303 		goto out;
3304 	sock_mnt = kern_mount(&sock_fs_type);
3305 	if (IS_ERR(sock_mnt)) {
3306 		err = PTR_ERR(sock_mnt);
3307 		goto out_mount;
3308 	}
3309 
3310 	/* The real protocol initialization is performed in later initcalls.
3311 	 */
3312 
3313 #ifdef CONFIG_NETFILTER
3314 	err = netfilter_init();
3315 	if (err)
3316 		goto out;
3317 #endif
3318 
3319 	ptp_classifier_init();
3320 
3321 out:
3322 	return err;
3323 
3324 out_mount:
3325 	unregister_filesystem(&sock_fs_type);
3326 	goto out;
3327 }
3328 
3329 core_initcall(sock_init);	/* early initcall */
3330 
3331 #ifdef CONFIG_PROC_FS
3332 void socket_seq_show(struct seq_file *seq)
3333 {
3334 	seq_printf(seq, "sockets: used %d\n",
3335 		   sock_inuse_get(seq->private));
3336 }
3337 #endif				/* CONFIG_PROC_FS */
3338 
3339 /* Handle the fact that while struct ifreq has the same *layout* on
3340  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3341  * which are handled elsewhere, it still has different *size* due to
3342  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3343  * resulting in struct ifreq being 32 and 40 bytes respectively).
3344  * As a result, if the struct happens to be at the end of a page and
3345  * the next page isn't readable/writable, we get a fault. To prevent
3346  * that, copy back and forth to the full size.
3347  */
3348 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3349 {
3350 	if (in_compat_syscall()) {
3351 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3352 
3353 		memset(ifr, 0, sizeof(*ifr));
3354 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3355 			return -EFAULT;
3356 
3357 		if (ifrdata)
3358 			*ifrdata = compat_ptr(ifr32->ifr_data);
3359 
3360 		return 0;
3361 	}
3362 
3363 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3364 		return -EFAULT;
3365 
3366 	if (ifrdata)
3367 		*ifrdata = ifr->ifr_data;
3368 
3369 	return 0;
3370 }
3371 EXPORT_SYMBOL(get_user_ifreq);
3372 
3373 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3374 {
3375 	size_t size = sizeof(*ifr);
3376 
3377 	if (in_compat_syscall())
3378 		size = sizeof(struct compat_ifreq);
3379 
3380 	if (copy_to_user(arg, ifr, size))
3381 		return -EFAULT;
3382 
3383 	return 0;
3384 }
3385 EXPORT_SYMBOL(put_user_ifreq);
3386 
3387 #ifdef CONFIG_COMPAT
3388 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3389 {
3390 	compat_uptr_t uptr32;
3391 	struct ifreq ifr;
3392 	void __user *saved;
3393 	int err;
3394 
3395 	if (get_user_ifreq(&ifr, NULL, uifr32))
3396 		return -EFAULT;
3397 
3398 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3399 		return -EFAULT;
3400 
3401 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3402 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3403 
3404 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3405 	if (!err) {
3406 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3407 		if (put_user_ifreq(&ifr, uifr32))
3408 			err = -EFAULT;
3409 	}
3410 	return err;
3411 }
3412 
3413 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3414 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3415 				 struct compat_ifreq __user *u_ifreq32)
3416 {
3417 	struct ifreq ifreq;
3418 	void __user *data;
3419 
3420 	if (!is_socket_ioctl_cmd(cmd))
3421 		return -ENOTTY;
3422 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3423 		return -EFAULT;
3424 	ifreq.ifr_data = data;
3425 
3426 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3427 }
3428 
3429 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3430 			 unsigned int cmd, unsigned long arg)
3431 {
3432 	void __user *argp = compat_ptr(arg);
3433 	struct sock *sk = sock->sk;
3434 	struct net *net = sock_net(sk);
3435 	const struct proto_ops *ops;
3436 
3437 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3438 		return sock_ioctl(file, cmd, (unsigned long)argp);
3439 
3440 	switch (cmd) {
3441 	case SIOCWANDEV:
3442 		return compat_siocwandev(net, argp);
3443 	case SIOCGSTAMP_OLD:
3444 	case SIOCGSTAMPNS_OLD:
3445 		ops = READ_ONCE(sock->ops);
3446 		if (!ops->gettstamp)
3447 			return -ENOIOCTLCMD;
3448 		return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3449 				      !COMPAT_USE_64BIT_TIME);
3450 
3451 	case SIOCETHTOOL:
3452 	case SIOCBONDSLAVEINFOQUERY:
3453 	case SIOCBONDINFOQUERY:
3454 	case SIOCSHWTSTAMP:
3455 	case SIOCGHWTSTAMP:
3456 		return compat_ifr_data_ioctl(net, cmd, argp);
3457 
3458 	case FIOSETOWN:
3459 	case SIOCSPGRP:
3460 	case FIOGETOWN:
3461 	case SIOCGPGRP:
3462 	case SIOCBRADDBR:
3463 	case SIOCBRDELBR:
3464 	case SIOCGIFVLAN:
3465 	case SIOCSIFVLAN:
3466 	case SIOCGSKNS:
3467 	case SIOCGSTAMP_NEW:
3468 	case SIOCGSTAMPNS_NEW:
3469 	case SIOCGIFCONF:
3470 	case SIOCSIFBR:
3471 	case SIOCGIFBR:
3472 		return sock_ioctl(file, cmd, arg);
3473 
3474 	case SIOCGIFFLAGS:
3475 	case SIOCSIFFLAGS:
3476 	case SIOCGIFMAP:
3477 	case SIOCSIFMAP:
3478 	case SIOCGIFMETRIC:
3479 	case SIOCSIFMETRIC:
3480 	case SIOCGIFMTU:
3481 	case SIOCSIFMTU:
3482 	case SIOCGIFMEM:
3483 	case SIOCSIFMEM:
3484 	case SIOCGIFHWADDR:
3485 	case SIOCSIFHWADDR:
3486 	case SIOCADDMULTI:
3487 	case SIOCDELMULTI:
3488 	case SIOCGIFINDEX:
3489 	case SIOCGIFADDR:
3490 	case SIOCSIFADDR:
3491 	case SIOCSIFHWBROADCAST:
3492 	case SIOCDIFADDR:
3493 	case SIOCGIFBRDADDR:
3494 	case SIOCSIFBRDADDR:
3495 	case SIOCGIFDSTADDR:
3496 	case SIOCSIFDSTADDR:
3497 	case SIOCGIFNETMASK:
3498 	case SIOCSIFNETMASK:
3499 	case SIOCSIFPFLAGS:
3500 	case SIOCGIFPFLAGS:
3501 	case SIOCGIFTXQLEN:
3502 	case SIOCSIFTXQLEN:
3503 	case SIOCBRADDIF:
3504 	case SIOCBRDELIF:
3505 	case SIOCGIFNAME:
3506 	case SIOCSIFNAME:
3507 	case SIOCGMIIPHY:
3508 	case SIOCGMIIREG:
3509 	case SIOCSMIIREG:
3510 	case SIOCBONDENSLAVE:
3511 	case SIOCBONDRELEASE:
3512 	case SIOCBONDSETHWADDR:
3513 	case SIOCBONDCHANGEACTIVE:
3514 	case SIOCSARP:
3515 	case SIOCGARP:
3516 	case SIOCDARP:
3517 	case SIOCOUTQ:
3518 	case SIOCOUTQNSD:
3519 	case SIOCATMARK:
3520 		return sock_do_ioctl(net, sock, cmd, arg);
3521 	}
3522 
3523 	return -ENOIOCTLCMD;
3524 }
3525 
3526 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3527 			      unsigned long arg)
3528 {
3529 	struct socket *sock = file->private_data;
3530 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3531 	int ret = -ENOIOCTLCMD;
3532 	struct sock *sk;
3533 	struct net *net;
3534 
3535 	sk = sock->sk;
3536 	net = sock_net(sk);
3537 
3538 	if (ops->compat_ioctl)
3539 		ret = ops->compat_ioctl(sock, cmd, arg);
3540 
3541 	if (ret == -ENOIOCTLCMD &&
3542 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3543 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3544 
3545 	if (ret == -ENOIOCTLCMD)
3546 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3547 
3548 	return ret;
3549 }
3550 #endif
3551 
3552 /**
3553  *	kernel_bind - bind an address to a socket (kernel space)
3554  *	@sock: socket
3555  *	@addr: address
3556  *	@addrlen: length of address
3557  *
3558  *	Returns 0 or an error.
3559  */
3560 
3561 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3562 {
3563 	struct sockaddr_storage address;
3564 
3565 	memcpy(&address, addr, addrlen);
3566 
3567 	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3568 					  addrlen);
3569 }
3570 EXPORT_SYMBOL(kernel_bind);
3571 
3572 /**
3573  *	kernel_listen - move socket to listening state (kernel space)
3574  *	@sock: socket
3575  *	@backlog: pending connections queue size
3576  *
3577  *	Returns 0 or an error.
3578  */
3579 
3580 int kernel_listen(struct socket *sock, int backlog)
3581 {
3582 	return READ_ONCE(sock->ops)->listen(sock, backlog);
3583 }
3584 EXPORT_SYMBOL(kernel_listen);
3585 
3586 /**
3587  *	kernel_accept - accept a connection (kernel space)
3588  *	@sock: listening socket
3589  *	@newsock: new connected socket
3590  *	@flags: flags
3591  *
3592  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3593  *	If it fails, @newsock is guaranteed to be %NULL.
3594  *	Returns 0 or an error.
3595  */
3596 
3597 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3598 {
3599 	struct sock *sk = sock->sk;
3600 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3601 	struct proto_accept_arg arg = {
3602 		.flags = flags,
3603 		.kern = true,
3604 	};
3605 	int err;
3606 
3607 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3608 			       newsock);
3609 	if (err < 0)
3610 		goto done;
3611 
3612 	err = ops->accept(sock, *newsock, &arg);
3613 	if (err < 0) {
3614 		sock_release(*newsock);
3615 		*newsock = NULL;
3616 		goto done;
3617 	}
3618 
3619 	(*newsock)->ops = ops;
3620 	__module_get(ops->owner);
3621 
3622 done:
3623 	return err;
3624 }
3625 EXPORT_SYMBOL(kernel_accept);
3626 
3627 /**
3628  *	kernel_connect - connect a socket (kernel space)
3629  *	@sock: socket
3630  *	@addr: address
3631  *	@addrlen: address length
3632  *	@flags: flags (O_NONBLOCK, ...)
3633  *
3634  *	For datagram sockets, @addr is the address to which datagrams are sent
3635  *	by default, and the only address from which datagrams are received.
3636  *	For stream sockets, attempts to connect to @addr.
3637  *	Returns 0 or an error code.
3638  */
3639 
3640 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3641 		   int flags)
3642 {
3643 	struct sockaddr_storage address;
3644 
3645 	memcpy(&address, addr, addrlen);
3646 
3647 	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3648 					     addrlen, flags);
3649 }
3650 EXPORT_SYMBOL(kernel_connect);
3651 
3652 /**
3653  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3654  *	@sock: socket
3655  *	@addr: address holder
3656  *
3657  * 	Fills the @addr pointer with the address which the socket is bound.
3658  *	Returns the length of the address in bytes or an error code.
3659  */
3660 
3661 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3662 {
3663 	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3664 }
3665 EXPORT_SYMBOL(kernel_getsockname);
3666 
3667 /**
3668  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3669  *	@sock: socket
3670  *	@addr: address holder
3671  *
3672  * 	Fills the @addr pointer with the address which the socket is connected.
3673  *	Returns the length of the address in bytes or an error code.
3674  */
3675 
3676 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3677 {
3678 	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3679 }
3680 EXPORT_SYMBOL(kernel_getpeername);
3681 
3682 /**
3683  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3684  *	@sock: socket
3685  *	@how: connection part
3686  *
3687  *	Returns 0 or an error.
3688  */
3689 
3690 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3691 {
3692 	return READ_ONCE(sock->ops)->shutdown(sock, how);
3693 }
3694 EXPORT_SYMBOL(kernel_sock_shutdown);
3695 
3696 /**
3697  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3698  *	@sk: socket
3699  *
3700  *	This routine returns the IP overhead imposed by a socket i.e.
3701  *	the length of the underlying IP header, depending on whether
3702  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3703  *	on at the socket. Assumes that the caller has a lock on the socket.
3704  */
3705 
3706 u32 kernel_sock_ip_overhead(struct sock *sk)
3707 {
3708 	struct inet_sock *inet;
3709 	struct ip_options_rcu *opt;
3710 	u32 overhead = 0;
3711 #if IS_ENABLED(CONFIG_IPV6)
3712 	struct ipv6_pinfo *np;
3713 	struct ipv6_txoptions *optv6 = NULL;
3714 #endif /* IS_ENABLED(CONFIG_IPV6) */
3715 
3716 	if (!sk)
3717 		return overhead;
3718 
3719 	switch (sk->sk_family) {
3720 	case AF_INET:
3721 		inet = inet_sk(sk);
3722 		overhead += sizeof(struct iphdr);
3723 		opt = rcu_dereference_protected(inet->inet_opt,
3724 						sock_owned_by_user(sk));
3725 		if (opt)
3726 			overhead += opt->opt.optlen;
3727 		return overhead;
3728 #if IS_ENABLED(CONFIG_IPV6)
3729 	case AF_INET6:
3730 		np = inet6_sk(sk);
3731 		overhead += sizeof(struct ipv6hdr);
3732 		if (np)
3733 			optv6 = rcu_dereference_protected(np->opt,
3734 							  sock_owned_by_user(sk));
3735 		if (optv6)
3736 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3737 		return overhead;
3738 #endif /* IS_ENABLED(CONFIG_IPV6) */
3739 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3740 		return overhead;
3741 	}
3742 }
3743 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3744