xref: /linux/net/socket.c (revision f69e98a91a01fd7c5755dd710e94a17d6e9f583f)
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/net.h>
61 #include <linux/interrupt.h>
62 #include <linux/thread_info.h>
63 #include <linux/rcupdate.h>
64 #include <linux/netdevice.h>
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67 #include <linux/mutex.h>
68 #include <linux/if_bridge.h>
69 #include <linux/if_vlan.h>
70 #include <linux/ptp_classify.h>
71 #include <linux/init.h>
72 #include <linux/poll.h>
73 #include <linux/cache.h>
74 #include <linux/module.h>
75 #include <linux/highmem.h>
76 #include <linux/mount.h>
77 #include <linux/pseudo_fs.h>
78 #include <linux/security.h>
79 #include <linux/syscalls.h>
80 #include <linux/compat.h>
81 #include <linux/kmod.h>
82 #include <linux/audit.h>
83 #include <linux/wireless.h>
84 #include <linux/nsproxy.h>
85 #include <linux/magic.h>
86 #include <linux/slab.h>
87 #include <linux/xattr.h>
88 #include <linux/nospec.h>
89 #include <linux/indirect_call_wrapper.h>
90 
91 #include <linux/uaccess.h>
92 #include <asm/unistd.h>
93 
94 #include <net/compat.h>
95 #include <net/wext.h>
96 #include <net/cls_cgroup.h>
97 
98 #include <net/sock.h>
99 #include <linux/netfilter.h>
100 
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/termios.h>
105 #include <linux/sockios.h>
106 #include <net/busy_poll.h>
107 #include <linux/errqueue.h>
108 #include <linux/ptp_clock_kernel.h>
109 
110 #ifdef CONFIG_NET_RX_BUSY_POLL
111 unsigned int sysctl_net_busy_read __read_mostly;
112 unsigned int sysctl_net_busy_poll __read_mostly;
113 #endif
114 
115 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
116 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
117 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
118 
119 static int sock_close(struct inode *inode, struct file *file);
120 static __poll_t sock_poll(struct file *file,
121 			      struct poll_table_struct *wait);
122 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
123 #ifdef CONFIG_COMPAT
124 static long compat_sock_ioctl(struct file *file,
125 			      unsigned int cmd, unsigned long arg);
126 #endif
127 static int sock_fasync(int fd, struct file *filp, int on);
128 static ssize_t sock_sendpage(struct file *file, struct page *page,
129 			     int offset, size_t size, loff_t *ppos, int more);
130 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
131 				struct pipe_inode_info *pipe, size_t len,
132 				unsigned int flags);
133 
134 #ifdef CONFIG_PROC_FS
135 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
136 {
137 	struct socket *sock = f->private_data;
138 
139 	if (sock->ops->show_fdinfo)
140 		sock->ops->show_fdinfo(m, sock);
141 }
142 #else
143 #define sock_show_fdinfo NULL
144 #endif
145 
146 /*
147  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
148  *	in the operation structures but are done directly via the socketcall() multiplexor.
149  */
150 
151 static const struct file_operations socket_file_ops = {
152 	.owner =	THIS_MODULE,
153 	.llseek =	no_llseek,
154 	.read_iter =	sock_read_iter,
155 	.write_iter =	sock_write_iter,
156 	.poll =		sock_poll,
157 	.unlocked_ioctl = sock_ioctl,
158 #ifdef CONFIG_COMPAT
159 	.compat_ioctl = compat_sock_ioctl,
160 #endif
161 	.mmap =		sock_mmap,
162 	.release =	sock_close,
163 	.fasync =	sock_fasync,
164 	.sendpage =	sock_sendpage,
165 	.splice_write = generic_splice_sendpage,
166 	.splice_read =	sock_splice_read,
167 	.show_fdinfo =	sock_show_fdinfo,
168 };
169 
170 static const char * const pf_family_names[] = {
171 	[PF_UNSPEC]	= "PF_UNSPEC",
172 	[PF_UNIX]	= "PF_UNIX/PF_LOCAL",
173 	[PF_INET]	= "PF_INET",
174 	[PF_AX25]	= "PF_AX25",
175 	[PF_IPX]	= "PF_IPX",
176 	[PF_APPLETALK]	= "PF_APPLETALK",
177 	[PF_NETROM]	= "PF_NETROM",
178 	[PF_BRIDGE]	= "PF_BRIDGE",
179 	[PF_ATMPVC]	= "PF_ATMPVC",
180 	[PF_X25]	= "PF_X25",
181 	[PF_INET6]	= "PF_INET6",
182 	[PF_ROSE]	= "PF_ROSE",
183 	[PF_DECnet]	= "PF_DECnet",
184 	[PF_NETBEUI]	= "PF_NETBEUI",
185 	[PF_SECURITY]	= "PF_SECURITY",
186 	[PF_KEY]	= "PF_KEY",
187 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
188 	[PF_PACKET]	= "PF_PACKET",
189 	[PF_ASH]	= "PF_ASH",
190 	[PF_ECONET]	= "PF_ECONET",
191 	[PF_ATMSVC]	= "PF_ATMSVC",
192 	[PF_RDS]	= "PF_RDS",
193 	[PF_SNA]	= "PF_SNA",
194 	[PF_IRDA]	= "PF_IRDA",
195 	[PF_PPPOX]	= "PF_PPPOX",
196 	[PF_WANPIPE]	= "PF_WANPIPE",
197 	[PF_LLC]	= "PF_LLC",
198 	[PF_IB]		= "PF_IB",
199 	[PF_MPLS]	= "PF_MPLS",
200 	[PF_CAN]	= "PF_CAN",
201 	[PF_TIPC]	= "PF_TIPC",
202 	[PF_BLUETOOTH]	= "PF_BLUETOOTH",
203 	[PF_IUCV]	= "PF_IUCV",
204 	[PF_RXRPC]	= "PF_RXRPC",
205 	[PF_ISDN]	= "PF_ISDN",
206 	[PF_PHONET]	= "PF_PHONET",
207 	[PF_IEEE802154]	= "PF_IEEE802154",
208 	[PF_CAIF]	= "PF_CAIF",
209 	[PF_ALG]	= "PF_ALG",
210 	[PF_NFC]	= "PF_NFC",
211 	[PF_VSOCK]	= "PF_VSOCK",
212 	[PF_KCM]	= "PF_KCM",
213 	[PF_QIPCRTR]	= "PF_QIPCRTR",
214 	[PF_SMC]	= "PF_SMC",
215 	[PF_XDP]	= "PF_XDP",
216 	[PF_MCTP]	= "PF_MCTP",
217 };
218 
219 /*
220  *	The protocol list. Each protocol is registered in here.
221  */
222 
223 static DEFINE_SPINLOCK(net_family_lock);
224 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
225 
226 /*
227  * Support routines.
228  * Move socket addresses back and forth across the kernel/user
229  * divide and look after the messy bits.
230  */
231 
232 /**
233  *	move_addr_to_kernel	-	copy a socket address into kernel space
234  *	@uaddr: Address in user space
235  *	@kaddr: Address in kernel space
236  *	@ulen: Length in user space
237  *
238  *	The address is copied into kernel space. If the provided address is
239  *	too long an error code of -EINVAL is returned. If the copy gives
240  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
241  */
242 
243 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
244 {
245 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
246 		return -EINVAL;
247 	if (ulen == 0)
248 		return 0;
249 	if (copy_from_user(kaddr, uaddr, ulen))
250 		return -EFAULT;
251 	return audit_sockaddr(ulen, kaddr);
252 }
253 
254 /**
255  *	move_addr_to_user	-	copy an address to user space
256  *	@kaddr: kernel space address
257  *	@klen: length of address in kernel
258  *	@uaddr: user space address
259  *	@ulen: pointer to user length field
260  *
261  *	The value pointed to by ulen on entry is the buffer length available.
262  *	This is overwritten with the buffer space used. -EINVAL is returned
263  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
264  *	is returned if either the buffer or the length field are not
265  *	accessible.
266  *	After copying the data up to the limit the user specifies, the true
267  *	length of the data is written over the length limit the user
268  *	specified. Zero is returned for a success.
269  */
270 
271 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
272 			     void __user *uaddr, int __user *ulen)
273 {
274 	int err;
275 	int len;
276 
277 	BUG_ON(klen > sizeof(struct sockaddr_storage));
278 	err = get_user(len, ulen);
279 	if (err)
280 		return err;
281 	if (len > klen)
282 		len = klen;
283 	if (len < 0)
284 		return -EINVAL;
285 	if (len) {
286 		if (audit_sockaddr(klen, kaddr))
287 			return -ENOMEM;
288 		if (copy_to_user(uaddr, kaddr, len))
289 			return -EFAULT;
290 	}
291 	/*
292 	 *      "fromlen shall refer to the value before truncation.."
293 	 *                      1003.1g
294 	 */
295 	return __put_user(klen, ulen);
296 }
297 
298 static struct kmem_cache *sock_inode_cachep __ro_after_init;
299 
300 static struct inode *sock_alloc_inode(struct super_block *sb)
301 {
302 	struct socket_alloc *ei;
303 
304 	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
305 	if (!ei)
306 		return NULL;
307 	init_waitqueue_head(&ei->socket.wq.wait);
308 	ei->socket.wq.fasync_list = NULL;
309 	ei->socket.wq.flags = 0;
310 
311 	ei->socket.state = SS_UNCONNECTED;
312 	ei->socket.flags = 0;
313 	ei->socket.ops = NULL;
314 	ei->socket.sk = NULL;
315 	ei->socket.file = NULL;
316 
317 	return &ei->vfs_inode;
318 }
319 
320 static void sock_free_inode(struct inode *inode)
321 {
322 	struct socket_alloc *ei;
323 
324 	ei = container_of(inode, struct socket_alloc, vfs_inode);
325 	kmem_cache_free(sock_inode_cachep, ei);
326 }
327 
328 static void init_once(void *foo)
329 {
330 	struct socket_alloc *ei = (struct socket_alloc *)foo;
331 
332 	inode_init_once(&ei->vfs_inode);
333 }
334 
335 static void init_inodecache(void)
336 {
337 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
338 					      sizeof(struct socket_alloc),
339 					      0,
340 					      (SLAB_HWCACHE_ALIGN |
341 					       SLAB_RECLAIM_ACCOUNT |
342 					       SLAB_MEM_SPREAD | SLAB_ACCOUNT),
343 					      init_once);
344 	BUG_ON(sock_inode_cachep == NULL);
345 }
346 
347 static const struct super_operations sockfs_ops = {
348 	.alloc_inode	= sock_alloc_inode,
349 	.free_inode	= sock_free_inode,
350 	.statfs		= simple_statfs,
351 };
352 
353 /*
354  * sockfs_dname() is called from d_path().
355  */
356 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
357 {
358 	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
359 				d_inode(dentry)->i_ino);
360 }
361 
362 static const struct dentry_operations sockfs_dentry_operations = {
363 	.d_dname  = sockfs_dname,
364 };
365 
366 static int sockfs_xattr_get(const struct xattr_handler *handler,
367 			    struct dentry *dentry, struct inode *inode,
368 			    const char *suffix, void *value, size_t size)
369 {
370 	if (value) {
371 		if (dentry->d_name.len + 1 > size)
372 			return -ERANGE;
373 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
374 	}
375 	return dentry->d_name.len + 1;
376 }
377 
378 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
379 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
380 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
381 
382 static const struct xattr_handler sockfs_xattr_handler = {
383 	.name = XATTR_NAME_SOCKPROTONAME,
384 	.get = sockfs_xattr_get,
385 };
386 
387 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
388 				     struct user_namespace *mnt_userns,
389 				     struct dentry *dentry, struct inode *inode,
390 				     const char *suffix, const void *value,
391 				     size_t size, int flags)
392 {
393 	/* Handled by LSM. */
394 	return -EAGAIN;
395 }
396 
397 static const struct xattr_handler sockfs_security_xattr_handler = {
398 	.prefix = XATTR_SECURITY_PREFIX,
399 	.set = sockfs_security_xattr_set,
400 };
401 
402 static const struct xattr_handler *sockfs_xattr_handlers[] = {
403 	&sockfs_xattr_handler,
404 	&sockfs_security_xattr_handler,
405 	NULL
406 };
407 
408 static int sockfs_init_fs_context(struct fs_context *fc)
409 {
410 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
411 	if (!ctx)
412 		return -ENOMEM;
413 	ctx->ops = &sockfs_ops;
414 	ctx->dops = &sockfs_dentry_operations;
415 	ctx->xattr = sockfs_xattr_handlers;
416 	return 0;
417 }
418 
419 static struct vfsmount *sock_mnt __read_mostly;
420 
421 static struct file_system_type sock_fs_type = {
422 	.name =		"sockfs",
423 	.init_fs_context = sockfs_init_fs_context,
424 	.kill_sb =	kill_anon_super,
425 };
426 
427 /*
428  *	Obtains the first available file descriptor and sets it up for use.
429  *
430  *	These functions create file structures and maps them to fd space
431  *	of the current process. On success it returns file descriptor
432  *	and file struct implicitly stored in sock->file.
433  *	Note that another thread may close file descriptor before we return
434  *	from this function. We use the fact that now we do not refer
435  *	to socket after mapping. If one day we will need it, this
436  *	function will increment ref. count on file by 1.
437  *
438  *	In any case returned fd MAY BE not valid!
439  *	This race condition is unavoidable
440  *	with shared fd spaces, we cannot solve it inside kernel,
441  *	but we take care of internal coherence yet.
442  */
443 
444 /**
445  *	sock_alloc_file - Bind a &socket to a &file
446  *	@sock: socket
447  *	@flags: file status flags
448  *	@dname: protocol name
449  *
450  *	Returns the &file bound with @sock, implicitly storing it
451  *	in sock->file. If dname is %NULL, sets to "".
452  *	On failure the return is a ERR pointer (see linux/err.h).
453  *	This function uses GFP_KERNEL internally.
454  */
455 
456 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
457 {
458 	struct file *file;
459 
460 	if (!dname)
461 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
462 
463 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
464 				O_RDWR | (flags & O_NONBLOCK),
465 				&socket_file_ops);
466 	if (IS_ERR(file)) {
467 		sock_release(sock);
468 		return file;
469 	}
470 
471 	sock->file = file;
472 	file->private_data = sock;
473 	stream_open(SOCK_INODE(sock), file);
474 	return file;
475 }
476 EXPORT_SYMBOL(sock_alloc_file);
477 
478 static int sock_map_fd(struct socket *sock, int flags)
479 {
480 	struct file *newfile;
481 	int fd = get_unused_fd_flags(flags);
482 	if (unlikely(fd < 0)) {
483 		sock_release(sock);
484 		return fd;
485 	}
486 
487 	newfile = sock_alloc_file(sock, flags, NULL);
488 	if (!IS_ERR(newfile)) {
489 		fd_install(fd, newfile);
490 		return fd;
491 	}
492 
493 	put_unused_fd(fd);
494 	return PTR_ERR(newfile);
495 }
496 
497 /**
498  *	sock_from_file - Return the &socket bounded to @file.
499  *	@file: file
500  *
501  *	On failure returns %NULL.
502  */
503 
504 struct socket *sock_from_file(struct file *file)
505 {
506 	if (file->f_op == &socket_file_ops)
507 		return file->private_data;	/* set in sock_map_fd */
508 
509 	return NULL;
510 }
511 EXPORT_SYMBOL(sock_from_file);
512 
513 /**
514  *	sockfd_lookup - Go from a file number to its socket slot
515  *	@fd: file handle
516  *	@err: pointer to an error code return
517  *
518  *	The file handle passed in is locked and the socket it is bound
519  *	to is returned. If an error occurs the err pointer is overwritten
520  *	with a negative errno code and NULL is returned. The function checks
521  *	for both invalid handles and passing a handle which is not a socket.
522  *
523  *	On a success the socket object pointer is returned.
524  */
525 
526 struct socket *sockfd_lookup(int fd, int *err)
527 {
528 	struct file *file;
529 	struct socket *sock;
530 
531 	file = fget(fd);
532 	if (!file) {
533 		*err = -EBADF;
534 		return NULL;
535 	}
536 
537 	sock = sock_from_file(file);
538 	if (!sock) {
539 		*err = -ENOTSOCK;
540 		fput(file);
541 	}
542 	return sock;
543 }
544 EXPORT_SYMBOL(sockfd_lookup);
545 
546 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
547 {
548 	struct fd f = fdget(fd);
549 	struct socket *sock;
550 
551 	*err = -EBADF;
552 	if (f.file) {
553 		sock = sock_from_file(f.file);
554 		if (likely(sock)) {
555 			*fput_needed = f.flags & FDPUT_FPUT;
556 			return sock;
557 		}
558 		*err = -ENOTSOCK;
559 		fdput(f);
560 	}
561 	return NULL;
562 }
563 
564 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
565 				size_t size)
566 {
567 	ssize_t len;
568 	ssize_t used = 0;
569 
570 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
571 	if (len < 0)
572 		return len;
573 	used += len;
574 	if (buffer) {
575 		if (size < used)
576 			return -ERANGE;
577 		buffer += len;
578 	}
579 
580 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
581 	used += len;
582 	if (buffer) {
583 		if (size < used)
584 			return -ERANGE;
585 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
586 		buffer += len;
587 	}
588 
589 	return used;
590 }
591 
592 static int sockfs_setattr(struct user_namespace *mnt_userns,
593 			  struct dentry *dentry, struct iattr *iattr)
594 {
595 	int err = simple_setattr(&init_user_ns, dentry, iattr);
596 
597 	if (!err && (iattr->ia_valid & ATTR_UID)) {
598 		struct socket *sock = SOCKET_I(d_inode(dentry));
599 
600 		if (sock->sk)
601 			sock->sk->sk_uid = iattr->ia_uid;
602 		else
603 			err = -ENOENT;
604 	}
605 
606 	return err;
607 }
608 
609 static const struct inode_operations sockfs_inode_ops = {
610 	.listxattr = sockfs_listxattr,
611 	.setattr = sockfs_setattr,
612 };
613 
614 /**
615  *	sock_alloc - allocate a socket
616  *
617  *	Allocate a new inode and socket object. The two are bound together
618  *	and initialised. The socket is then returned. If we are out of inodes
619  *	NULL is returned. This functions uses GFP_KERNEL internally.
620  */
621 
622 struct socket *sock_alloc(void)
623 {
624 	struct inode *inode;
625 	struct socket *sock;
626 
627 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
628 	if (!inode)
629 		return NULL;
630 
631 	sock = SOCKET_I(inode);
632 
633 	inode->i_ino = get_next_ino();
634 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
635 	inode->i_uid = current_fsuid();
636 	inode->i_gid = current_fsgid();
637 	inode->i_op = &sockfs_inode_ops;
638 
639 	return sock;
640 }
641 EXPORT_SYMBOL(sock_alloc);
642 
643 static void __sock_release(struct socket *sock, struct inode *inode)
644 {
645 	if (sock->ops) {
646 		struct module *owner = sock->ops->owner;
647 
648 		if (inode)
649 			inode_lock(inode);
650 		sock->ops->release(sock);
651 		sock->sk = NULL;
652 		if (inode)
653 			inode_unlock(inode);
654 		sock->ops = NULL;
655 		module_put(owner);
656 	}
657 
658 	if (sock->wq.fasync_list)
659 		pr_err("%s: fasync list not empty!\n", __func__);
660 
661 	if (!sock->file) {
662 		iput(SOCK_INODE(sock));
663 		return;
664 	}
665 	sock->file = NULL;
666 }
667 
668 /**
669  *	sock_release - close a socket
670  *	@sock: socket to close
671  *
672  *	The socket is released from the protocol stack if it has a release
673  *	callback, and the inode is then released if the socket is bound to
674  *	an inode not a file.
675  */
676 void sock_release(struct socket *sock)
677 {
678 	__sock_release(sock, NULL);
679 }
680 EXPORT_SYMBOL(sock_release);
681 
682 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
683 {
684 	u8 flags = *tx_flags;
685 
686 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
687 		flags |= SKBTX_HW_TSTAMP;
688 
689 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
690 		flags |= SKBTX_SW_TSTAMP;
691 
692 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
693 		flags |= SKBTX_SCHED_TSTAMP;
694 
695 	*tx_flags = flags;
696 }
697 EXPORT_SYMBOL(__sock_tx_timestamp);
698 
699 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
700 					   size_t));
701 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
702 					    size_t));
703 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
704 {
705 	int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
706 				     inet_sendmsg, sock, msg,
707 				     msg_data_left(msg));
708 	BUG_ON(ret == -EIOCBQUEUED);
709 	return ret;
710 }
711 
712 /**
713  *	sock_sendmsg - send a message through @sock
714  *	@sock: socket
715  *	@msg: message to send
716  *
717  *	Sends @msg through @sock, passing through LSM.
718  *	Returns the number of bytes sent, or an error code.
719  */
720 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
721 {
722 	int err = security_socket_sendmsg(sock, msg,
723 					  msg_data_left(msg));
724 
725 	return err ?: sock_sendmsg_nosec(sock, msg);
726 }
727 EXPORT_SYMBOL(sock_sendmsg);
728 
729 /**
730  *	kernel_sendmsg - send a message through @sock (kernel-space)
731  *	@sock: socket
732  *	@msg: message header
733  *	@vec: kernel vec
734  *	@num: vec array length
735  *	@size: total message data size
736  *
737  *	Builds the message data with @vec and sends it through @sock.
738  *	Returns the number of bytes sent, or an error code.
739  */
740 
741 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
742 		   struct kvec *vec, size_t num, size_t size)
743 {
744 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
745 	return sock_sendmsg(sock, msg);
746 }
747 EXPORT_SYMBOL(kernel_sendmsg);
748 
749 /**
750  *	kernel_sendmsg_locked - send a message through @sock (kernel-space)
751  *	@sk: sock
752  *	@msg: message header
753  *	@vec: output s/g array
754  *	@num: output s/g array length
755  *	@size: total message data size
756  *
757  *	Builds the message data with @vec and sends it through @sock.
758  *	Returns the number of bytes sent, or an error code.
759  *	Caller must hold @sk.
760  */
761 
762 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
763 			  struct kvec *vec, size_t num, size_t size)
764 {
765 	struct socket *sock = sk->sk_socket;
766 
767 	if (!sock->ops->sendmsg_locked)
768 		return sock_no_sendmsg_locked(sk, msg, size);
769 
770 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
771 
772 	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
773 }
774 EXPORT_SYMBOL(kernel_sendmsg_locked);
775 
776 static bool skb_is_err_queue(const struct sk_buff *skb)
777 {
778 	/* pkt_type of skbs enqueued on the error queue are set to
779 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
780 	 * in recvmsg, since skbs received on a local socket will never
781 	 * have a pkt_type of PACKET_OUTGOING.
782 	 */
783 	return skb->pkt_type == PACKET_OUTGOING;
784 }
785 
786 /* On transmit, software and hardware timestamps are returned independently.
787  * As the two skb clones share the hardware timestamp, which may be updated
788  * before the software timestamp is received, a hardware TX timestamp may be
789  * returned only if there is no software TX timestamp. Ignore false software
790  * timestamps, which may be made in the __sock_recv_timestamp() call when the
791  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
792  * hardware timestamp.
793  */
794 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
795 {
796 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
797 }
798 
799 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
800 {
801 	struct scm_ts_pktinfo ts_pktinfo;
802 	struct net_device *orig_dev;
803 
804 	if (!skb_mac_header_was_set(skb))
805 		return;
806 
807 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
808 
809 	rcu_read_lock();
810 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
811 	if (orig_dev)
812 		ts_pktinfo.if_index = orig_dev->ifindex;
813 	rcu_read_unlock();
814 
815 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
816 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
817 		 sizeof(ts_pktinfo), &ts_pktinfo);
818 }
819 
820 /*
821  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
822  */
823 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
824 	struct sk_buff *skb)
825 {
826 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
827 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
828 	struct scm_timestamping_internal tss;
829 
830 	int empty = 1, false_tstamp = 0;
831 	struct skb_shared_hwtstamps *shhwtstamps =
832 		skb_hwtstamps(skb);
833 	ktime_t hwtstamp;
834 
835 	/* Race occurred between timestamp enabling and packet
836 	   receiving.  Fill in the current time for now. */
837 	if (need_software_tstamp && skb->tstamp == 0) {
838 		__net_timestamp(skb);
839 		false_tstamp = 1;
840 	}
841 
842 	if (need_software_tstamp) {
843 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
844 			if (new_tstamp) {
845 				struct __kernel_sock_timeval tv;
846 
847 				skb_get_new_timestamp(skb, &tv);
848 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
849 					 sizeof(tv), &tv);
850 			} else {
851 				struct __kernel_old_timeval tv;
852 
853 				skb_get_timestamp(skb, &tv);
854 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
855 					 sizeof(tv), &tv);
856 			}
857 		} else {
858 			if (new_tstamp) {
859 				struct __kernel_timespec ts;
860 
861 				skb_get_new_timestampns(skb, &ts);
862 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
863 					 sizeof(ts), &ts);
864 			} else {
865 				struct __kernel_old_timespec ts;
866 
867 				skb_get_timestampns(skb, &ts);
868 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
869 					 sizeof(ts), &ts);
870 			}
871 		}
872 	}
873 
874 	memset(&tss, 0, sizeof(tss));
875 	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
876 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
877 		empty = 0;
878 	if (shhwtstamps &&
879 	    (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
880 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
881 		if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
882 			hwtstamp = ptp_convert_timestamp(shhwtstamps,
883 							 sk->sk_bind_phc);
884 		else
885 			hwtstamp = shhwtstamps->hwtstamp;
886 
887 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
888 			empty = 0;
889 
890 			if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
891 			    !skb_is_err_queue(skb))
892 				put_ts_pktinfo(msg, skb);
893 		}
894 	}
895 	if (!empty) {
896 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
897 			put_cmsg_scm_timestamping64(msg, &tss);
898 		else
899 			put_cmsg_scm_timestamping(msg, &tss);
900 
901 		if (skb_is_err_queue(skb) && skb->len &&
902 		    SKB_EXT_ERR(skb)->opt_stats)
903 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
904 				 skb->len, skb->data);
905 	}
906 }
907 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
908 
909 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
910 	struct sk_buff *skb)
911 {
912 	int ack;
913 
914 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
915 		return;
916 	if (!skb->wifi_acked_valid)
917 		return;
918 
919 	ack = skb->wifi_acked;
920 
921 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
922 }
923 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
924 
925 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
926 				   struct sk_buff *skb)
927 {
928 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
929 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
930 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
931 }
932 
933 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
934 	struct sk_buff *skb)
935 {
936 	sock_recv_timestamp(msg, sk, skb);
937 	sock_recv_drops(msg, sk, skb);
938 }
939 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
940 
941 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
942 					   size_t, int));
943 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
944 					    size_t, int));
945 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
946 				     int flags)
947 {
948 	return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
949 				  inet_recvmsg, sock, msg, msg_data_left(msg),
950 				  flags);
951 }
952 
953 /**
954  *	sock_recvmsg - receive a message from @sock
955  *	@sock: socket
956  *	@msg: message to receive
957  *	@flags: message flags
958  *
959  *	Receives @msg from @sock, passing through LSM. Returns the total number
960  *	of bytes received, or an error.
961  */
962 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
963 {
964 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
965 
966 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
967 }
968 EXPORT_SYMBOL(sock_recvmsg);
969 
970 /**
971  *	kernel_recvmsg - Receive a message from a socket (kernel space)
972  *	@sock: The socket to receive the message from
973  *	@msg: Received message
974  *	@vec: Input s/g array for message data
975  *	@num: Size of input s/g array
976  *	@size: Number of bytes to read
977  *	@flags: Message flags (MSG_DONTWAIT, etc...)
978  *
979  *	On return the msg structure contains the scatter/gather array passed in the
980  *	vec argument. The array is modified so that it consists of the unfilled
981  *	portion of the original array.
982  *
983  *	The returned value is the total number of bytes received, or an error.
984  */
985 
986 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
987 		   struct kvec *vec, size_t num, size_t size, int flags)
988 {
989 	msg->msg_control_is_user = false;
990 	iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
991 	return sock_recvmsg(sock, msg, flags);
992 }
993 EXPORT_SYMBOL(kernel_recvmsg);
994 
995 static ssize_t sock_sendpage(struct file *file, struct page *page,
996 			     int offset, size_t size, loff_t *ppos, int more)
997 {
998 	struct socket *sock;
999 	int flags;
1000 
1001 	sock = file->private_data;
1002 
1003 	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1004 	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1005 	flags |= more;
1006 
1007 	return kernel_sendpage(sock, page, offset, size, flags);
1008 }
1009 
1010 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1011 				struct pipe_inode_info *pipe, size_t len,
1012 				unsigned int flags)
1013 {
1014 	struct socket *sock = file->private_data;
1015 
1016 	if (unlikely(!sock->ops->splice_read))
1017 		return generic_file_splice_read(file, ppos, pipe, len, flags);
1018 
1019 	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1020 }
1021 
1022 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1023 {
1024 	struct file *file = iocb->ki_filp;
1025 	struct socket *sock = file->private_data;
1026 	struct msghdr msg = {.msg_iter = *to,
1027 			     .msg_iocb = iocb};
1028 	ssize_t res;
1029 
1030 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1031 		msg.msg_flags = MSG_DONTWAIT;
1032 
1033 	if (iocb->ki_pos != 0)
1034 		return -ESPIPE;
1035 
1036 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1037 		return 0;
1038 
1039 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1040 	*to = msg.msg_iter;
1041 	return res;
1042 }
1043 
1044 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1045 {
1046 	struct file *file = iocb->ki_filp;
1047 	struct socket *sock = file->private_data;
1048 	struct msghdr msg = {.msg_iter = *from,
1049 			     .msg_iocb = iocb};
1050 	ssize_t res;
1051 
1052 	if (iocb->ki_pos != 0)
1053 		return -ESPIPE;
1054 
1055 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1056 		msg.msg_flags = MSG_DONTWAIT;
1057 
1058 	if (sock->type == SOCK_SEQPACKET)
1059 		msg.msg_flags |= MSG_EOR;
1060 
1061 	res = sock_sendmsg(sock, &msg);
1062 	*from = msg.msg_iter;
1063 	return res;
1064 }
1065 
1066 /*
1067  * Atomic setting of ioctl hooks to avoid race
1068  * with module unload.
1069  */
1070 
1071 static DEFINE_MUTEX(br_ioctl_mutex);
1072 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1073 			    unsigned int cmd, struct ifreq *ifr,
1074 			    void __user *uarg);
1075 
1076 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1077 			     unsigned int cmd, struct ifreq *ifr,
1078 			     void __user *uarg))
1079 {
1080 	mutex_lock(&br_ioctl_mutex);
1081 	br_ioctl_hook = hook;
1082 	mutex_unlock(&br_ioctl_mutex);
1083 }
1084 EXPORT_SYMBOL(brioctl_set);
1085 
1086 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1087 		  struct ifreq *ifr, void __user *uarg)
1088 {
1089 	int err = -ENOPKG;
1090 
1091 	if (!br_ioctl_hook)
1092 		request_module("bridge");
1093 
1094 	mutex_lock(&br_ioctl_mutex);
1095 	if (br_ioctl_hook)
1096 		err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1097 	mutex_unlock(&br_ioctl_mutex);
1098 
1099 	return err;
1100 }
1101 
1102 static DEFINE_MUTEX(vlan_ioctl_mutex);
1103 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1104 
1105 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1106 {
1107 	mutex_lock(&vlan_ioctl_mutex);
1108 	vlan_ioctl_hook = hook;
1109 	mutex_unlock(&vlan_ioctl_mutex);
1110 }
1111 EXPORT_SYMBOL(vlan_ioctl_set);
1112 
1113 static long sock_do_ioctl(struct net *net, struct socket *sock,
1114 			  unsigned int cmd, unsigned long arg)
1115 {
1116 	struct ifreq ifr;
1117 	bool need_copyout;
1118 	int err;
1119 	void __user *argp = (void __user *)arg;
1120 	void __user *data;
1121 
1122 	err = sock->ops->ioctl(sock, cmd, arg);
1123 
1124 	/*
1125 	 * If this ioctl is unknown try to hand it down
1126 	 * to the NIC driver.
1127 	 */
1128 	if (err != -ENOIOCTLCMD)
1129 		return err;
1130 
1131 	if (!is_socket_ioctl_cmd(cmd))
1132 		return -ENOTTY;
1133 
1134 	if (get_user_ifreq(&ifr, &data, argp))
1135 		return -EFAULT;
1136 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1137 	if (!err && need_copyout)
1138 		if (put_user_ifreq(&ifr, argp))
1139 			return -EFAULT;
1140 
1141 	return err;
1142 }
1143 
1144 /*
1145  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1146  *	what to do with it - that's up to the protocol still.
1147  */
1148 
1149 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1150 {
1151 	struct socket *sock;
1152 	struct sock *sk;
1153 	void __user *argp = (void __user *)arg;
1154 	int pid, err;
1155 	struct net *net;
1156 
1157 	sock = file->private_data;
1158 	sk = sock->sk;
1159 	net = sock_net(sk);
1160 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1161 		struct ifreq ifr;
1162 		void __user *data;
1163 		bool need_copyout;
1164 		if (get_user_ifreq(&ifr, &data, argp))
1165 			return -EFAULT;
1166 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1167 		if (!err && need_copyout)
1168 			if (put_user_ifreq(&ifr, argp))
1169 				return -EFAULT;
1170 	} else
1171 #ifdef CONFIG_WEXT_CORE
1172 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1173 		err = wext_handle_ioctl(net, cmd, argp);
1174 	} else
1175 #endif
1176 		switch (cmd) {
1177 		case FIOSETOWN:
1178 		case SIOCSPGRP:
1179 			err = -EFAULT;
1180 			if (get_user(pid, (int __user *)argp))
1181 				break;
1182 			err = f_setown(sock->file, pid, 1);
1183 			break;
1184 		case FIOGETOWN:
1185 		case SIOCGPGRP:
1186 			err = put_user(f_getown(sock->file),
1187 				       (int __user *)argp);
1188 			break;
1189 		case SIOCGIFBR:
1190 		case SIOCSIFBR:
1191 		case SIOCBRADDBR:
1192 		case SIOCBRDELBR:
1193 			err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1194 			break;
1195 		case SIOCGIFVLAN:
1196 		case SIOCSIFVLAN:
1197 			err = -ENOPKG;
1198 			if (!vlan_ioctl_hook)
1199 				request_module("8021q");
1200 
1201 			mutex_lock(&vlan_ioctl_mutex);
1202 			if (vlan_ioctl_hook)
1203 				err = vlan_ioctl_hook(net, argp);
1204 			mutex_unlock(&vlan_ioctl_mutex);
1205 			break;
1206 		case SIOCGSKNS:
1207 			err = -EPERM;
1208 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1209 				break;
1210 
1211 			err = open_related_ns(&net->ns, get_net_ns);
1212 			break;
1213 		case SIOCGSTAMP_OLD:
1214 		case SIOCGSTAMPNS_OLD:
1215 			if (!sock->ops->gettstamp) {
1216 				err = -ENOIOCTLCMD;
1217 				break;
1218 			}
1219 			err = sock->ops->gettstamp(sock, argp,
1220 						   cmd == SIOCGSTAMP_OLD,
1221 						   !IS_ENABLED(CONFIG_64BIT));
1222 			break;
1223 		case SIOCGSTAMP_NEW:
1224 		case SIOCGSTAMPNS_NEW:
1225 			if (!sock->ops->gettstamp) {
1226 				err = -ENOIOCTLCMD;
1227 				break;
1228 			}
1229 			err = sock->ops->gettstamp(sock, argp,
1230 						   cmd == SIOCGSTAMP_NEW,
1231 						   false);
1232 			break;
1233 
1234 		case SIOCGIFCONF:
1235 			err = dev_ifconf(net, argp);
1236 			break;
1237 
1238 		default:
1239 			err = sock_do_ioctl(net, sock, cmd, arg);
1240 			break;
1241 		}
1242 	return err;
1243 }
1244 
1245 /**
1246  *	sock_create_lite - creates a socket
1247  *	@family: protocol family (AF_INET, ...)
1248  *	@type: communication type (SOCK_STREAM, ...)
1249  *	@protocol: protocol (0, ...)
1250  *	@res: new socket
1251  *
1252  *	Creates a new socket and assigns it to @res, passing through LSM.
1253  *	The new socket initialization is not complete, see kernel_accept().
1254  *	Returns 0 or an error. On failure @res is set to %NULL.
1255  *	This function internally uses GFP_KERNEL.
1256  */
1257 
1258 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1259 {
1260 	int err;
1261 	struct socket *sock = NULL;
1262 
1263 	err = security_socket_create(family, type, protocol, 1);
1264 	if (err)
1265 		goto out;
1266 
1267 	sock = sock_alloc();
1268 	if (!sock) {
1269 		err = -ENOMEM;
1270 		goto out;
1271 	}
1272 
1273 	sock->type = type;
1274 	err = security_socket_post_create(sock, family, type, protocol, 1);
1275 	if (err)
1276 		goto out_release;
1277 
1278 out:
1279 	*res = sock;
1280 	return err;
1281 out_release:
1282 	sock_release(sock);
1283 	sock = NULL;
1284 	goto out;
1285 }
1286 EXPORT_SYMBOL(sock_create_lite);
1287 
1288 /* No kernel lock held - perfect */
1289 static __poll_t sock_poll(struct file *file, poll_table *wait)
1290 {
1291 	struct socket *sock = file->private_data;
1292 	__poll_t events = poll_requested_events(wait), flag = 0;
1293 
1294 	if (!sock->ops->poll)
1295 		return 0;
1296 
1297 	if (sk_can_busy_loop(sock->sk)) {
1298 		/* poll once if requested by the syscall */
1299 		if (events & POLL_BUSY_LOOP)
1300 			sk_busy_loop(sock->sk, 1);
1301 
1302 		/* if this socket can poll_ll, tell the system call */
1303 		flag = POLL_BUSY_LOOP;
1304 	}
1305 
1306 	return sock->ops->poll(file, sock, wait) | flag;
1307 }
1308 
1309 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1310 {
1311 	struct socket *sock = file->private_data;
1312 
1313 	return sock->ops->mmap(file, sock, vma);
1314 }
1315 
1316 static int sock_close(struct inode *inode, struct file *filp)
1317 {
1318 	__sock_release(SOCKET_I(inode), inode);
1319 	return 0;
1320 }
1321 
1322 /*
1323  *	Update the socket async list
1324  *
1325  *	Fasync_list locking strategy.
1326  *
1327  *	1. fasync_list is modified only under process context socket lock
1328  *	   i.e. under semaphore.
1329  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1330  *	   or under socket lock
1331  */
1332 
1333 static int sock_fasync(int fd, struct file *filp, int on)
1334 {
1335 	struct socket *sock = filp->private_data;
1336 	struct sock *sk = sock->sk;
1337 	struct socket_wq *wq = &sock->wq;
1338 
1339 	if (sk == NULL)
1340 		return -EINVAL;
1341 
1342 	lock_sock(sk);
1343 	fasync_helper(fd, filp, on, &wq->fasync_list);
1344 
1345 	if (!wq->fasync_list)
1346 		sock_reset_flag(sk, SOCK_FASYNC);
1347 	else
1348 		sock_set_flag(sk, SOCK_FASYNC);
1349 
1350 	release_sock(sk);
1351 	return 0;
1352 }
1353 
1354 /* This function may be called only under rcu_lock */
1355 
1356 int sock_wake_async(struct socket_wq *wq, int how, int band)
1357 {
1358 	if (!wq || !wq->fasync_list)
1359 		return -1;
1360 
1361 	switch (how) {
1362 	case SOCK_WAKE_WAITD:
1363 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1364 			break;
1365 		goto call_kill;
1366 	case SOCK_WAKE_SPACE:
1367 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1368 			break;
1369 		fallthrough;
1370 	case SOCK_WAKE_IO:
1371 call_kill:
1372 		kill_fasync(&wq->fasync_list, SIGIO, band);
1373 		break;
1374 	case SOCK_WAKE_URG:
1375 		kill_fasync(&wq->fasync_list, SIGURG, band);
1376 	}
1377 
1378 	return 0;
1379 }
1380 EXPORT_SYMBOL(sock_wake_async);
1381 
1382 /**
1383  *	__sock_create - creates a socket
1384  *	@net: net namespace
1385  *	@family: protocol family (AF_INET, ...)
1386  *	@type: communication type (SOCK_STREAM, ...)
1387  *	@protocol: protocol (0, ...)
1388  *	@res: new socket
1389  *	@kern: boolean for kernel space sockets
1390  *
1391  *	Creates a new socket and assigns it to @res, passing through LSM.
1392  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1393  *	be set to true if the socket resides in kernel space.
1394  *	This function internally uses GFP_KERNEL.
1395  */
1396 
1397 int __sock_create(struct net *net, int family, int type, int protocol,
1398 			 struct socket **res, int kern)
1399 {
1400 	int err;
1401 	struct socket *sock;
1402 	const struct net_proto_family *pf;
1403 
1404 	/*
1405 	 *      Check protocol is in range
1406 	 */
1407 	if (family < 0 || family >= NPROTO)
1408 		return -EAFNOSUPPORT;
1409 	if (type < 0 || type >= SOCK_MAX)
1410 		return -EINVAL;
1411 
1412 	/* Compatibility.
1413 
1414 	   This uglymoron is moved from INET layer to here to avoid
1415 	   deadlock in module load.
1416 	 */
1417 	if (family == PF_INET && type == SOCK_PACKET) {
1418 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1419 			     current->comm);
1420 		family = PF_PACKET;
1421 	}
1422 
1423 	err = security_socket_create(family, type, protocol, kern);
1424 	if (err)
1425 		return err;
1426 
1427 	/*
1428 	 *	Allocate the socket and allow the family to set things up. if
1429 	 *	the protocol is 0, the family is instructed to select an appropriate
1430 	 *	default.
1431 	 */
1432 	sock = sock_alloc();
1433 	if (!sock) {
1434 		net_warn_ratelimited("socket: no more sockets\n");
1435 		return -ENFILE;	/* Not exactly a match, but its the
1436 				   closest posix thing */
1437 	}
1438 
1439 	sock->type = type;
1440 
1441 #ifdef CONFIG_MODULES
1442 	/* Attempt to load a protocol module if the find failed.
1443 	 *
1444 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1445 	 * requested real, full-featured networking support upon configuration.
1446 	 * Otherwise module support will break!
1447 	 */
1448 	if (rcu_access_pointer(net_families[family]) == NULL)
1449 		request_module("net-pf-%d", family);
1450 #endif
1451 
1452 	rcu_read_lock();
1453 	pf = rcu_dereference(net_families[family]);
1454 	err = -EAFNOSUPPORT;
1455 	if (!pf)
1456 		goto out_release;
1457 
1458 	/*
1459 	 * We will call the ->create function, that possibly is in a loadable
1460 	 * module, so we have to bump that loadable module refcnt first.
1461 	 */
1462 	if (!try_module_get(pf->owner))
1463 		goto out_release;
1464 
1465 	/* Now protected by module ref count */
1466 	rcu_read_unlock();
1467 
1468 	err = pf->create(net, sock, protocol, kern);
1469 	if (err < 0)
1470 		goto out_module_put;
1471 
1472 	/*
1473 	 * Now to bump the refcnt of the [loadable] module that owns this
1474 	 * socket at sock_release time we decrement its refcnt.
1475 	 */
1476 	if (!try_module_get(sock->ops->owner))
1477 		goto out_module_busy;
1478 
1479 	/*
1480 	 * Now that we're done with the ->create function, the [loadable]
1481 	 * module can have its refcnt decremented
1482 	 */
1483 	module_put(pf->owner);
1484 	err = security_socket_post_create(sock, family, type, protocol, kern);
1485 	if (err)
1486 		goto out_sock_release;
1487 	*res = sock;
1488 
1489 	return 0;
1490 
1491 out_module_busy:
1492 	err = -EAFNOSUPPORT;
1493 out_module_put:
1494 	sock->ops = NULL;
1495 	module_put(pf->owner);
1496 out_sock_release:
1497 	sock_release(sock);
1498 	return err;
1499 
1500 out_release:
1501 	rcu_read_unlock();
1502 	goto out_sock_release;
1503 }
1504 EXPORT_SYMBOL(__sock_create);
1505 
1506 /**
1507  *	sock_create - creates a socket
1508  *	@family: protocol family (AF_INET, ...)
1509  *	@type: communication type (SOCK_STREAM, ...)
1510  *	@protocol: protocol (0, ...)
1511  *	@res: new socket
1512  *
1513  *	A wrapper around __sock_create().
1514  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1515  */
1516 
1517 int sock_create(int family, int type, int protocol, struct socket **res)
1518 {
1519 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1520 }
1521 EXPORT_SYMBOL(sock_create);
1522 
1523 /**
1524  *	sock_create_kern - creates a socket (kernel space)
1525  *	@net: net namespace
1526  *	@family: protocol family (AF_INET, ...)
1527  *	@type: communication type (SOCK_STREAM, ...)
1528  *	@protocol: protocol (0, ...)
1529  *	@res: new socket
1530  *
1531  *	A wrapper around __sock_create().
1532  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1533  */
1534 
1535 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1536 {
1537 	return __sock_create(net, family, type, protocol, res, 1);
1538 }
1539 EXPORT_SYMBOL(sock_create_kern);
1540 
1541 int __sys_socket(int family, int type, int protocol)
1542 {
1543 	int retval;
1544 	struct socket *sock;
1545 	int flags;
1546 
1547 	/* Check the SOCK_* constants for consistency.  */
1548 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1549 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1550 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1551 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1552 
1553 	flags = type & ~SOCK_TYPE_MASK;
1554 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1555 		return -EINVAL;
1556 	type &= SOCK_TYPE_MASK;
1557 
1558 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1559 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1560 
1561 	retval = sock_create(family, type, protocol, &sock);
1562 	if (retval < 0)
1563 		return retval;
1564 
1565 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1566 }
1567 
1568 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1569 {
1570 	return __sys_socket(family, type, protocol);
1571 }
1572 
1573 /*
1574  *	Create a pair of connected sockets.
1575  */
1576 
1577 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1578 {
1579 	struct socket *sock1, *sock2;
1580 	int fd1, fd2, err;
1581 	struct file *newfile1, *newfile2;
1582 	int flags;
1583 
1584 	flags = type & ~SOCK_TYPE_MASK;
1585 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1586 		return -EINVAL;
1587 	type &= SOCK_TYPE_MASK;
1588 
1589 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1590 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1591 
1592 	/*
1593 	 * reserve descriptors and make sure we won't fail
1594 	 * to return them to userland.
1595 	 */
1596 	fd1 = get_unused_fd_flags(flags);
1597 	if (unlikely(fd1 < 0))
1598 		return fd1;
1599 
1600 	fd2 = get_unused_fd_flags(flags);
1601 	if (unlikely(fd2 < 0)) {
1602 		put_unused_fd(fd1);
1603 		return fd2;
1604 	}
1605 
1606 	err = put_user(fd1, &usockvec[0]);
1607 	if (err)
1608 		goto out;
1609 
1610 	err = put_user(fd2, &usockvec[1]);
1611 	if (err)
1612 		goto out;
1613 
1614 	/*
1615 	 * Obtain the first socket and check if the underlying protocol
1616 	 * supports the socketpair call.
1617 	 */
1618 
1619 	err = sock_create(family, type, protocol, &sock1);
1620 	if (unlikely(err < 0))
1621 		goto out;
1622 
1623 	err = sock_create(family, type, protocol, &sock2);
1624 	if (unlikely(err < 0)) {
1625 		sock_release(sock1);
1626 		goto out;
1627 	}
1628 
1629 	err = security_socket_socketpair(sock1, sock2);
1630 	if (unlikely(err)) {
1631 		sock_release(sock2);
1632 		sock_release(sock1);
1633 		goto out;
1634 	}
1635 
1636 	err = sock1->ops->socketpair(sock1, sock2);
1637 	if (unlikely(err < 0)) {
1638 		sock_release(sock2);
1639 		sock_release(sock1);
1640 		goto out;
1641 	}
1642 
1643 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1644 	if (IS_ERR(newfile1)) {
1645 		err = PTR_ERR(newfile1);
1646 		sock_release(sock2);
1647 		goto out;
1648 	}
1649 
1650 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1651 	if (IS_ERR(newfile2)) {
1652 		err = PTR_ERR(newfile2);
1653 		fput(newfile1);
1654 		goto out;
1655 	}
1656 
1657 	audit_fd_pair(fd1, fd2);
1658 
1659 	fd_install(fd1, newfile1);
1660 	fd_install(fd2, newfile2);
1661 	return 0;
1662 
1663 out:
1664 	put_unused_fd(fd2);
1665 	put_unused_fd(fd1);
1666 	return err;
1667 }
1668 
1669 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1670 		int __user *, usockvec)
1671 {
1672 	return __sys_socketpair(family, type, protocol, usockvec);
1673 }
1674 
1675 /*
1676  *	Bind a name to a socket. Nothing much to do here since it's
1677  *	the protocol's responsibility to handle the local address.
1678  *
1679  *	We move the socket address to kernel space before we call
1680  *	the protocol layer (having also checked the address is ok).
1681  */
1682 
1683 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1684 {
1685 	struct socket *sock;
1686 	struct sockaddr_storage address;
1687 	int err, fput_needed;
1688 
1689 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1690 	if (sock) {
1691 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1692 		if (!err) {
1693 			err = security_socket_bind(sock,
1694 						   (struct sockaddr *)&address,
1695 						   addrlen);
1696 			if (!err)
1697 				err = sock->ops->bind(sock,
1698 						      (struct sockaddr *)
1699 						      &address, addrlen);
1700 		}
1701 		fput_light(sock->file, fput_needed);
1702 	}
1703 	return err;
1704 }
1705 
1706 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1707 {
1708 	return __sys_bind(fd, umyaddr, addrlen);
1709 }
1710 
1711 /*
1712  *	Perform a listen. Basically, we allow the protocol to do anything
1713  *	necessary for a listen, and if that works, we mark the socket as
1714  *	ready for listening.
1715  */
1716 
1717 int __sys_listen(int fd, int backlog)
1718 {
1719 	struct socket *sock;
1720 	int err, fput_needed;
1721 	int somaxconn;
1722 
1723 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1724 	if (sock) {
1725 		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1726 		if ((unsigned int)backlog > somaxconn)
1727 			backlog = somaxconn;
1728 
1729 		err = security_socket_listen(sock, backlog);
1730 		if (!err)
1731 			err = sock->ops->listen(sock, backlog);
1732 
1733 		fput_light(sock->file, fput_needed);
1734 	}
1735 	return err;
1736 }
1737 
1738 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1739 {
1740 	return __sys_listen(fd, backlog);
1741 }
1742 
1743 struct file *do_accept(struct file *file, unsigned file_flags,
1744 		       struct sockaddr __user *upeer_sockaddr,
1745 		       int __user *upeer_addrlen, int flags)
1746 {
1747 	struct socket *sock, *newsock;
1748 	struct file *newfile;
1749 	int err, len;
1750 	struct sockaddr_storage address;
1751 
1752 	sock = sock_from_file(file);
1753 	if (!sock)
1754 		return ERR_PTR(-ENOTSOCK);
1755 
1756 	newsock = sock_alloc();
1757 	if (!newsock)
1758 		return ERR_PTR(-ENFILE);
1759 
1760 	newsock->type = sock->type;
1761 	newsock->ops = sock->ops;
1762 
1763 	/*
1764 	 * We don't need try_module_get here, as the listening socket (sock)
1765 	 * has the protocol module (sock->ops->owner) held.
1766 	 */
1767 	__module_get(newsock->ops->owner);
1768 
1769 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1770 	if (IS_ERR(newfile))
1771 		return newfile;
1772 
1773 	err = security_socket_accept(sock, newsock);
1774 	if (err)
1775 		goto out_fd;
1776 
1777 	err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1778 					false);
1779 	if (err < 0)
1780 		goto out_fd;
1781 
1782 	if (upeer_sockaddr) {
1783 		len = newsock->ops->getname(newsock,
1784 					(struct sockaddr *)&address, 2);
1785 		if (len < 0) {
1786 			err = -ECONNABORTED;
1787 			goto out_fd;
1788 		}
1789 		err = move_addr_to_user(&address,
1790 					len, upeer_sockaddr, upeer_addrlen);
1791 		if (err < 0)
1792 			goto out_fd;
1793 	}
1794 
1795 	/* File flags are not inherited via accept() unlike another OSes. */
1796 	return newfile;
1797 out_fd:
1798 	fput(newfile);
1799 	return ERR_PTR(err);
1800 }
1801 
1802 int __sys_accept4_file(struct file *file, unsigned file_flags,
1803 		       struct sockaddr __user *upeer_sockaddr,
1804 		       int __user *upeer_addrlen, int flags,
1805 		       unsigned long nofile)
1806 {
1807 	struct file *newfile;
1808 	int newfd;
1809 
1810 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1811 		return -EINVAL;
1812 
1813 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1814 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1815 
1816 	newfd = __get_unused_fd_flags(flags, nofile);
1817 	if (unlikely(newfd < 0))
1818 		return newfd;
1819 
1820 	newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen,
1821 			    flags);
1822 	if (IS_ERR(newfile)) {
1823 		put_unused_fd(newfd);
1824 		return PTR_ERR(newfile);
1825 	}
1826 	fd_install(newfd, newfile);
1827 	return newfd;
1828 }
1829 
1830 /*
1831  *	For accept, we attempt to create a new socket, set up the link
1832  *	with the client, wake up the client, then return the new
1833  *	connected fd. We collect the address of the connector in kernel
1834  *	space and move it to user at the very end. This is unclean because
1835  *	we open the socket then return an error.
1836  *
1837  *	1003.1g adds the ability to recvmsg() to query connection pending
1838  *	status to recvmsg. We need to add that support in a way thats
1839  *	clean when we restructure accept also.
1840  */
1841 
1842 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1843 		  int __user *upeer_addrlen, int flags)
1844 {
1845 	int ret = -EBADF;
1846 	struct fd f;
1847 
1848 	f = fdget(fd);
1849 	if (f.file) {
1850 		ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1851 						upeer_addrlen, flags,
1852 						rlimit(RLIMIT_NOFILE));
1853 		fdput(f);
1854 	}
1855 
1856 	return ret;
1857 }
1858 
1859 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1860 		int __user *, upeer_addrlen, int, flags)
1861 {
1862 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1863 }
1864 
1865 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1866 		int __user *, upeer_addrlen)
1867 {
1868 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1869 }
1870 
1871 /*
1872  *	Attempt to connect to a socket with the server address.  The address
1873  *	is in user space so we verify it is OK and move it to kernel space.
1874  *
1875  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1876  *	break bindings
1877  *
1878  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1879  *	other SEQPACKET protocols that take time to connect() as it doesn't
1880  *	include the -EINPROGRESS status for such sockets.
1881  */
1882 
1883 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1884 		       int addrlen, int file_flags)
1885 {
1886 	struct socket *sock;
1887 	int err;
1888 
1889 	sock = sock_from_file(file);
1890 	if (!sock) {
1891 		err = -ENOTSOCK;
1892 		goto out;
1893 	}
1894 
1895 	err =
1896 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1897 	if (err)
1898 		goto out;
1899 
1900 	err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1901 				 sock->file->f_flags | file_flags);
1902 out:
1903 	return err;
1904 }
1905 
1906 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1907 {
1908 	int ret = -EBADF;
1909 	struct fd f;
1910 
1911 	f = fdget(fd);
1912 	if (f.file) {
1913 		struct sockaddr_storage address;
1914 
1915 		ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1916 		if (!ret)
1917 			ret = __sys_connect_file(f.file, &address, addrlen, 0);
1918 		fdput(f);
1919 	}
1920 
1921 	return ret;
1922 }
1923 
1924 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1925 		int, addrlen)
1926 {
1927 	return __sys_connect(fd, uservaddr, addrlen);
1928 }
1929 
1930 /*
1931  *	Get the local address ('name') of a socket object. Move the obtained
1932  *	name to user space.
1933  */
1934 
1935 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1936 		      int __user *usockaddr_len)
1937 {
1938 	struct socket *sock;
1939 	struct sockaddr_storage address;
1940 	int err, fput_needed;
1941 
1942 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1943 	if (!sock)
1944 		goto out;
1945 
1946 	err = security_socket_getsockname(sock);
1947 	if (err)
1948 		goto out_put;
1949 
1950 	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1951 	if (err < 0)
1952 		goto out_put;
1953 	/* "err" is actually length in this case */
1954 	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1955 
1956 out_put:
1957 	fput_light(sock->file, fput_needed);
1958 out:
1959 	return err;
1960 }
1961 
1962 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1963 		int __user *, usockaddr_len)
1964 {
1965 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1966 }
1967 
1968 /*
1969  *	Get the remote address ('name') of a socket object. Move the obtained
1970  *	name to user space.
1971  */
1972 
1973 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1974 		      int __user *usockaddr_len)
1975 {
1976 	struct socket *sock;
1977 	struct sockaddr_storage address;
1978 	int err, fput_needed;
1979 
1980 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1981 	if (sock != NULL) {
1982 		err = security_socket_getpeername(sock);
1983 		if (err) {
1984 			fput_light(sock->file, fput_needed);
1985 			return err;
1986 		}
1987 
1988 		err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1989 		if (err >= 0)
1990 			/* "err" is actually length in this case */
1991 			err = move_addr_to_user(&address, err, usockaddr,
1992 						usockaddr_len);
1993 		fput_light(sock->file, fput_needed);
1994 	}
1995 	return err;
1996 }
1997 
1998 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1999 		int __user *, usockaddr_len)
2000 {
2001 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2002 }
2003 
2004 /*
2005  *	Send a datagram to a given address. We move the address into kernel
2006  *	space and check the user space data area is readable before invoking
2007  *	the protocol.
2008  */
2009 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2010 		 struct sockaddr __user *addr,  int addr_len)
2011 {
2012 	struct socket *sock;
2013 	struct sockaddr_storage address;
2014 	int err;
2015 	struct msghdr msg;
2016 	struct iovec iov;
2017 	int fput_needed;
2018 
2019 	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
2020 	if (unlikely(err))
2021 		return err;
2022 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2023 	if (!sock)
2024 		goto out;
2025 
2026 	msg.msg_name = NULL;
2027 	msg.msg_control = NULL;
2028 	msg.msg_controllen = 0;
2029 	msg.msg_namelen = 0;
2030 	if (addr) {
2031 		err = move_addr_to_kernel(addr, addr_len, &address);
2032 		if (err < 0)
2033 			goto out_put;
2034 		msg.msg_name = (struct sockaddr *)&address;
2035 		msg.msg_namelen = addr_len;
2036 	}
2037 	if (sock->file->f_flags & O_NONBLOCK)
2038 		flags |= MSG_DONTWAIT;
2039 	msg.msg_flags = flags;
2040 	err = sock_sendmsg(sock, &msg);
2041 
2042 out_put:
2043 	fput_light(sock->file, fput_needed);
2044 out:
2045 	return err;
2046 }
2047 
2048 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2049 		unsigned int, flags, struct sockaddr __user *, addr,
2050 		int, addr_len)
2051 {
2052 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2053 }
2054 
2055 /*
2056  *	Send a datagram down a socket.
2057  */
2058 
2059 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2060 		unsigned int, flags)
2061 {
2062 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2063 }
2064 
2065 /*
2066  *	Receive a frame from the socket and optionally record the address of the
2067  *	sender. We verify the buffers are writable and if needed move the
2068  *	sender address from kernel to user space.
2069  */
2070 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2071 		   struct sockaddr __user *addr, int __user *addr_len)
2072 {
2073 	struct socket *sock;
2074 	struct iovec iov;
2075 	struct msghdr msg;
2076 	struct sockaddr_storage address;
2077 	int err, err2;
2078 	int fput_needed;
2079 
2080 	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2081 	if (unlikely(err))
2082 		return err;
2083 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2084 	if (!sock)
2085 		goto out;
2086 
2087 	msg.msg_control = NULL;
2088 	msg.msg_controllen = 0;
2089 	/* Save some cycles and don't copy the address if not needed */
2090 	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2091 	/* We assume all kernel code knows the size of sockaddr_storage */
2092 	msg.msg_namelen = 0;
2093 	msg.msg_iocb = NULL;
2094 	msg.msg_flags = 0;
2095 	if (sock->file->f_flags & O_NONBLOCK)
2096 		flags |= MSG_DONTWAIT;
2097 	err = sock_recvmsg(sock, &msg, flags);
2098 
2099 	if (err >= 0 && addr != NULL) {
2100 		err2 = move_addr_to_user(&address,
2101 					 msg.msg_namelen, addr, addr_len);
2102 		if (err2 < 0)
2103 			err = err2;
2104 	}
2105 
2106 	fput_light(sock->file, fput_needed);
2107 out:
2108 	return err;
2109 }
2110 
2111 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2112 		unsigned int, flags, struct sockaddr __user *, addr,
2113 		int __user *, addr_len)
2114 {
2115 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2116 }
2117 
2118 /*
2119  *	Receive a datagram from a socket.
2120  */
2121 
2122 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2123 		unsigned int, flags)
2124 {
2125 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2126 }
2127 
2128 static bool sock_use_custom_sol_socket(const struct socket *sock)
2129 {
2130 	const struct sock *sk = sock->sk;
2131 
2132 	/* Use sock->ops->setsockopt() for MPTCP */
2133 	return IS_ENABLED(CONFIG_MPTCP) &&
2134 	       sk->sk_protocol == IPPROTO_MPTCP &&
2135 	       sk->sk_type == SOCK_STREAM &&
2136 	       (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2137 }
2138 
2139 /*
2140  *	Set a socket option. Because we don't know the option lengths we have
2141  *	to pass the user mode parameter for the protocols to sort out.
2142  */
2143 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2144 		int optlen)
2145 {
2146 	sockptr_t optval = USER_SOCKPTR(user_optval);
2147 	char *kernel_optval = NULL;
2148 	int err, fput_needed;
2149 	struct socket *sock;
2150 
2151 	if (optlen < 0)
2152 		return -EINVAL;
2153 
2154 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2155 	if (!sock)
2156 		return err;
2157 
2158 	err = security_socket_setsockopt(sock, level, optname);
2159 	if (err)
2160 		goto out_put;
2161 
2162 	if (!in_compat_syscall())
2163 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2164 						     user_optval, &optlen,
2165 						     &kernel_optval);
2166 	if (err < 0)
2167 		goto out_put;
2168 	if (err > 0) {
2169 		err = 0;
2170 		goto out_put;
2171 	}
2172 
2173 	if (kernel_optval)
2174 		optval = KERNEL_SOCKPTR(kernel_optval);
2175 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2176 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2177 	else if (unlikely(!sock->ops->setsockopt))
2178 		err = -EOPNOTSUPP;
2179 	else
2180 		err = sock->ops->setsockopt(sock, level, optname, optval,
2181 					    optlen);
2182 	kfree(kernel_optval);
2183 out_put:
2184 	fput_light(sock->file, fput_needed);
2185 	return err;
2186 }
2187 
2188 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2189 		char __user *, optval, int, optlen)
2190 {
2191 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2192 }
2193 
2194 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2195 							 int optname));
2196 
2197 /*
2198  *	Get a socket option. Because we don't know the option lengths we have
2199  *	to pass a user mode parameter for the protocols to sort out.
2200  */
2201 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2202 		int __user *optlen)
2203 {
2204 	int err, fput_needed;
2205 	struct socket *sock;
2206 	int max_optlen;
2207 
2208 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2209 	if (!sock)
2210 		return err;
2211 
2212 	err = security_socket_getsockopt(sock, level, optname);
2213 	if (err)
2214 		goto out_put;
2215 
2216 	if (!in_compat_syscall())
2217 		max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2218 
2219 	if (level == SOL_SOCKET)
2220 		err = sock_getsockopt(sock, level, optname, optval, optlen);
2221 	else if (unlikely(!sock->ops->getsockopt))
2222 		err = -EOPNOTSUPP;
2223 	else
2224 		err = sock->ops->getsockopt(sock, level, optname, optval,
2225 					    optlen);
2226 
2227 	if (!in_compat_syscall())
2228 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2229 						     optval, optlen, max_optlen,
2230 						     err);
2231 out_put:
2232 	fput_light(sock->file, fput_needed);
2233 	return err;
2234 }
2235 
2236 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2237 		char __user *, optval, int __user *, optlen)
2238 {
2239 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2240 }
2241 
2242 /*
2243  *	Shutdown a socket.
2244  */
2245 
2246 int __sys_shutdown_sock(struct socket *sock, int how)
2247 {
2248 	int err;
2249 
2250 	err = security_socket_shutdown(sock, how);
2251 	if (!err)
2252 		err = sock->ops->shutdown(sock, how);
2253 
2254 	return err;
2255 }
2256 
2257 int __sys_shutdown(int fd, int how)
2258 {
2259 	int err, fput_needed;
2260 	struct socket *sock;
2261 
2262 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2263 	if (sock != NULL) {
2264 		err = __sys_shutdown_sock(sock, how);
2265 		fput_light(sock->file, fput_needed);
2266 	}
2267 	return err;
2268 }
2269 
2270 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2271 {
2272 	return __sys_shutdown(fd, how);
2273 }
2274 
2275 /* A couple of helpful macros for getting the address of the 32/64 bit
2276  * fields which are the same type (int / unsigned) on our platforms.
2277  */
2278 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2279 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2280 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2281 
2282 struct used_address {
2283 	struct sockaddr_storage name;
2284 	unsigned int name_len;
2285 };
2286 
2287 int __copy_msghdr_from_user(struct msghdr *kmsg,
2288 			    struct user_msghdr __user *umsg,
2289 			    struct sockaddr __user **save_addr,
2290 			    struct iovec __user **uiov, size_t *nsegs)
2291 {
2292 	struct user_msghdr msg;
2293 	ssize_t err;
2294 
2295 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2296 		return -EFAULT;
2297 
2298 	kmsg->msg_control_is_user = true;
2299 	kmsg->msg_control_user = msg.msg_control;
2300 	kmsg->msg_controllen = msg.msg_controllen;
2301 	kmsg->msg_flags = msg.msg_flags;
2302 
2303 	kmsg->msg_namelen = msg.msg_namelen;
2304 	if (!msg.msg_name)
2305 		kmsg->msg_namelen = 0;
2306 
2307 	if (kmsg->msg_namelen < 0)
2308 		return -EINVAL;
2309 
2310 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2311 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2312 
2313 	if (save_addr)
2314 		*save_addr = msg.msg_name;
2315 
2316 	if (msg.msg_name && kmsg->msg_namelen) {
2317 		if (!save_addr) {
2318 			err = move_addr_to_kernel(msg.msg_name,
2319 						  kmsg->msg_namelen,
2320 						  kmsg->msg_name);
2321 			if (err < 0)
2322 				return err;
2323 		}
2324 	} else {
2325 		kmsg->msg_name = NULL;
2326 		kmsg->msg_namelen = 0;
2327 	}
2328 
2329 	if (msg.msg_iovlen > UIO_MAXIOV)
2330 		return -EMSGSIZE;
2331 
2332 	kmsg->msg_iocb = NULL;
2333 	*uiov = msg.msg_iov;
2334 	*nsegs = msg.msg_iovlen;
2335 	return 0;
2336 }
2337 
2338 static int copy_msghdr_from_user(struct msghdr *kmsg,
2339 				 struct user_msghdr __user *umsg,
2340 				 struct sockaddr __user **save_addr,
2341 				 struct iovec **iov)
2342 {
2343 	struct user_msghdr msg;
2344 	ssize_t err;
2345 
2346 	err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2347 					&msg.msg_iovlen);
2348 	if (err)
2349 		return err;
2350 
2351 	err = import_iovec(save_addr ? READ : WRITE,
2352 			    msg.msg_iov, msg.msg_iovlen,
2353 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2354 	return err < 0 ? err : 0;
2355 }
2356 
2357 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2358 			   unsigned int flags, struct used_address *used_address,
2359 			   unsigned int allowed_msghdr_flags)
2360 {
2361 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2362 				__aligned(sizeof(__kernel_size_t));
2363 	/* 20 is size of ipv6_pktinfo */
2364 	unsigned char *ctl_buf = ctl;
2365 	int ctl_len;
2366 	ssize_t err;
2367 
2368 	err = -ENOBUFS;
2369 
2370 	if (msg_sys->msg_controllen > INT_MAX)
2371 		goto out;
2372 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2373 	ctl_len = msg_sys->msg_controllen;
2374 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2375 		err =
2376 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2377 						     sizeof(ctl));
2378 		if (err)
2379 			goto out;
2380 		ctl_buf = msg_sys->msg_control;
2381 		ctl_len = msg_sys->msg_controllen;
2382 	} else if (ctl_len) {
2383 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2384 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2385 		if (ctl_len > sizeof(ctl)) {
2386 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2387 			if (ctl_buf == NULL)
2388 				goto out;
2389 		}
2390 		err = -EFAULT;
2391 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2392 			goto out_freectl;
2393 		msg_sys->msg_control = ctl_buf;
2394 		msg_sys->msg_control_is_user = false;
2395 	}
2396 	msg_sys->msg_flags = flags;
2397 
2398 	if (sock->file->f_flags & O_NONBLOCK)
2399 		msg_sys->msg_flags |= MSG_DONTWAIT;
2400 	/*
2401 	 * If this is sendmmsg() and current destination address is same as
2402 	 * previously succeeded address, omit asking LSM's decision.
2403 	 * used_address->name_len is initialized to UINT_MAX so that the first
2404 	 * destination address never matches.
2405 	 */
2406 	if (used_address && msg_sys->msg_name &&
2407 	    used_address->name_len == msg_sys->msg_namelen &&
2408 	    !memcmp(&used_address->name, msg_sys->msg_name,
2409 		    used_address->name_len)) {
2410 		err = sock_sendmsg_nosec(sock, msg_sys);
2411 		goto out_freectl;
2412 	}
2413 	err = sock_sendmsg(sock, msg_sys);
2414 	/*
2415 	 * If this is sendmmsg() and sending to current destination address was
2416 	 * successful, remember it.
2417 	 */
2418 	if (used_address && err >= 0) {
2419 		used_address->name_len = msg_sys->msg_namelen;
2420 		if (msg_sys->msg_name)
2421 			memcpy(&used_address->name, msg_sys->msg_name,
2422 			       used_address->name_len);
2423 	}
2424 
2425 out_freectl:
2426 	if (ctl_buf != ctl)
2427 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2428 out:
2429 	return err;
2430 }
2431 
2432 int sendmsg_copy_msghdr(struct msghdr *msg,
2433 			struct user_msghdr __user *umsg, unsigned flags,
2434 			struct iovec **iov)
2435 {
2436 	int err;
2437 
2438 	if (flags & MSG_CMSG_COMPAT) {
2439 		struct compat_msghdr __user *msg_compat;
2440 
2441 		msg_compat = (struct compat_msghdr __user *) umsg;
2442 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2443 	} else {
2444 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2445 	}
2446 	if (err < 0)
2447 		return err;
2448 
2449 	return 0;
2450 }
2451 
2452 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2453 			 struct msghdr *msg_sys, unsigned int flags,
2454 			 struct used_address *used_address,
2455 			 unsigned int allowed_msghdr_flags)
2456 {
2457 	struct sockaddr_storage address;
2458 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2459 	ssize_t err;
2460 
2461 	msg_sys->msg_name = &address;
2462 
2463 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2464 	if (err < 0)
2465 		return err;
2466 
2467 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2468 				allowed_msghdr_flags);
2469 	kfree(iov);
2470 	return err;
2471 }
2472 
2473 /*
2474  *	BSD sendmsg interface
2475  */
2476 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2477 			unsigned int flags)
2478 {
2479 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2480 }
2481 
2482 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2483 		   bool forbid_cmsg_compat)
2484 {
2485 	int fput_needed, err;
2486 	struct msghdr msg_sys;
2487 	struct socket *sock;
2488 
2489 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2490 		return -EINVAL;
2491 
2492 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2493 	if (!sock)
2494 		goto out;
2495 
2496 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2497 
2498 	fput_light(sock->file, fput_needed);
2499 out:
2500 	return err;
2501 }
2502 
2503 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2504 {
2505 	return __sys_sendmsg(fd, msg, flags, true);
2506 }
2507 
2508 /*
2509  *	Linux sendmmsg interface
2510  */
2511 
2512 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2513 		   unsigned int flags, bool forbid_cmsg_compat)
2514 {
2515 	int fput_needed, err, datagrams;
2516 	struct socket *sock;
2517 	struct mmsghdr __user *entry;
2518 	struct compat_mmsghdr __user *compat_entry;
2519 	struct msghdr msg_sys;
2520 	struct used_address used_address;
2521 	unsigned int oflags = flags;
2522 
2523 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2524 		return -EINVAL;
2525 
2526 	if (vlen > UIO_MAXIOV)
2527 		vlen = UIO_MAXIOV;
2528 
2529 	datagrams = 0;
2530 
2531 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2532 	if (!sock)
2533 		return err;
2534 
2535 	used_address.name_len = UINT_MAX;
2536 	entry = mmsg;
2537 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2538 	err = 0;
2539 	flags |= MSG_BATCH;
2540 
2541 	while (datagrams < vlen) {
2542 		if (datagrams == vlen - 1)
2543 			flags = oflags;
2544 
2545 		if (MSG_CMSG_COMPAT & flags) {
2546 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2547 					     &msg_sys, flags, &used_address, MSG_EOR);
2548 			if (err < 0)
2549 				break;
2550 			err = __put_user(err, &compat_entry->msg_len);
2551 			++compat_entry;
2552 		} else {
2553 			err = ___sys_sendmsg(sock,
2554 					     (struct user_msghdr __user *)entry,
2555 					     &msg_sys, flags, &used_address, MSG_EOR);
2556 			if (err < 0)
2557 				break;
2558 			err = put_user(err, &entry->msg_len);
2559 			++entry;
2560 		}
2561 
2562 		if (err)
2563 			break;
2564 		++datagrams;
2565 		if (msg_data_left(&msg_sys))
2566 			break;
2567 		cond_resched();
2568 	}
2569 
2570 	fput_light(sock->file, fput_needed);
2571 
2572 	/* We only return an error if no datagrams were able to be sent */
2573 	if (datagrams != 0)
2574 		return datagrams;
2575 
2576 	return err;
2577 }
2578 
2579 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2580 		unsigned int, vlen, unsigned int, flags)
2581 {
2582 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2583 }
2584 
2585 int recvmsg_copy_msghdr(struct msghdr *msg,
2586 			struct user_msghdr __user *umsg, unsigned flags,
2587 			struct sockaddr __user **uaddr,
2588 			struct iovec **iov)
2589 {
2590 	ssize_t err;
2591 
2592 	if (MSG_CMSG_COMPAT & flags) {
2593 		struct compat_msghdr __user *msg_compat;
2594 
2595 		msg_compat = (struct compat_msghdr __user *) umsg;
2596 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2597 	} else {
2598 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2599 	}
2600 	if (err < 0)
2601 		return err;
2602 
2603 	return 0;
2604 }
2605 
2606 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2607 			   struct user_msghdr __user *msg,
2608 			   struct sockaddr __user *uaddr,
2609 			   unsigned int flags, int nosec)
2610 {
2611 	struct compat_msghdr __user *msg_compat =
2612 					(struct compat_msghdr __user *) msg;
2613 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2614 	struct sockaddr_storage addr;
2615 	unsigned long cmsg_ptr;
2616 	int len;
2617 	ssize_t err;
2618 
2619 	msg_sys->msg_name = &addr;
2620 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2621 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2622 
2623 	/* We assume all kernel code knows the size of sockaddr_storage */
2624 	msg_sys->msg_namelen = 0;
2625 
2626 	if (sock->file->f_flags & O_NONBLOCK)
2627 		flags |= MSG_DONTWAIT;
2628 
2629 	if (unlikely(nosec))
2630 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2631 	else
2632 		err = sock_recvmsg(sock, msg_sys, flags);
2633 
2634 	if (err < 0)
2635 		goto out;
2636 	len = err;
2637 
2638 	if (uaddr != NULL) {
2639 		err = move_addr_to_user(&addr,
2640 					msg_sys->msg_namelen, uaddr,
2641 					uaddr_len);
2642 		if (err < 0)
2643 			goto out;
2644 	}
2645 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2646 			 COMPAT_FLAGS(msg));
2647 	if (err)
2648 		goto out;
2649 	if (MSG_CMSG_COMPAT & flags)
2650 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2651 				 &msg_compat->msg_controllen);
2652 	else
2653 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2654 				 &msg->msg_controllen);
2655 	if (err)
2656 		goto out;
2657 	err = len;
2658 out:
2659 	return err;
2660 }
2661 
2662 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2663 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2664 {
2665 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2666 	/* user mode address pointers */
2667 	struct sockaddr __user *uaddr;
2668 	ssize_t err;
2669 
2670 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2671 	if (err < 0)
2672 		return err;
2673 
2674 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2675 	kfree(iov);
2676 	return err;
2677 }
2678 
2679 /*
2680  *	BSD recvmsg interface
2681  */
2682 
2683 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2684 			struct user_msghdr __user *umsg,
2685 			struct sockaddr __user *uaddr, unsigned int flags)
2686 {
2687 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2688 }
2689 
2690 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2691 		   bool forbid_cmsg_compat)
2692 {
2693 	int fput_needed, err;
2694 	struct msghdr msg_sys;
2695 	struct socket *sock;
2696 
2697 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2698 		return -EINVAL;
2699 
2700 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2701 	if (!sock)
2702 		goto out;
2703 
2704 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2705 
2706 	fput_light(sock->file, fput_needed);
2707 out:
2708 	return err;
2709 }
2710 
2711 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2712 		unsigned int, flags)
2713 {
2714 	return __sys_recvmsg(fd, msg, flags, true);
2715 }
2716 
2717 /*
2718  *     Linux recvmmsg interface
2719  */
2720 
2721 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2722 			  unsigned int vlen, unsigned int flags,
2723 			  struct timespec64 *timeout)
2724 {
2725 	int fput_needed, err, datagrams;
2726 	struct socket *sock;
2727 	struct mmsghdr __user *entry;
2728 	struct compat_mmsghdr __user *compat_entry;
2729 	struct msghdr msg_sys;
2730 	struct timespec64 end_time;
2731 	struct timespec64 timeout64;
2732 
2733 	if (timeout &&
2734 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2735 				    timeout->tv_nsec))
2736 		return -EINVAL;
2737 
2738 	datagrams = 0;
2739 
2740 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2741 	if (!sock)
2742 		return err;
2743 
2744 	if (likely(!(flags & MSG_ERRQUEUE))) {
2745 		err = sock_error(sock->sk);
2746 		if (err) {
2747 			datagrams = err;
2748 			goto out_put;
2749 		}
2750 	}
2751 
2752 	entry = mmsg;
2753 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2754 
2755 	while (datagrams < vlen) {
2756 		/*
2757 		 * No need to ask LSM for more than the first datagram.
2758 		 */
2759 		if (MSG_CMSG_COMPAT & flags) {
2760 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2761 					     &msg_sys, flags & ~MSG_WAITFORONE,
2762 					     datagrams);
2763 			if (err < 0)
2764 				break;
2765 			err = __put_user(err, &compat_entry->msg_len);
2766 			++compat_entry;
2767 		} else {
2768 			err = ___sys_recvmsg(sock,
2769 					     (struct user_msghdr __user *)entry,
2770 					     &msg_sys, flags & ~MSG_WAITFORONE,
2771 					     datagrams);
2772 			if (err < 0)
2773 				break;
2774 			err = put_user(err, &entry->msg_len);
2775 			++entry;
2776 		}
2777 
2778 		if (err)
2779 			break;
2780 		++datagrams;
2781 
2782 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2783 		if (flags & MSG_WAITFORONE)
2784 			flags |= MSG_DONTWAIT;
2785 
2786 		if (timeout) {
2787 			ktime_get_ts64(&timeout64);
2788 			*timeout = timespec64_sub(end_time, timeout64);
2789 			if (timeout->tv_sec < 0) {
2790 				timeout->tv_sec = timeout->tv_nsec = 0;
2791 				break;
2792 			}
2793 
2794 			/* Timeout, return less than vlen datagrams */
2795 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2796 				break;
2797 		}
2798 
2799 		/* Out of band data, return right away */
2800 		if (msg_sys.msg_flags & MSG_OOB)
2801 			break;
2802 		cond_resched();
2803 	}
2804 
2805 	if (err == 0)
2806 		goto out_put;
2807 
2808 	if (datagrams == 0) {
2809 		datagrams = err;
2810 		goto out_put;
2811 	}
2812 
2813 	/*
2814 	 * We may return less entries than requested (vlen) if the
2815 	 * sock is non block and there aren't enough datagrams...
2816 	 */
2817 	if (err != -EAGAIN) {
2818 		/*
2819 		 * ... or  if recvmsg returns an error after we
2820 		 * received some datagrams, where we record the
2821 		 * error to return on the next call or if the
2822 		 * app asks about it using getsockopt(SO_ERROR).
2823 		 */
2824 		sock->sk->sk_err = -err;
2825 	}
2826 out_put:
2827 	fput_light(sock->file, fput_needed);
2828 
2829 	return datagrams;
2830 }
2831 
2832 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2833 		   unsigned int vlen, unsigned int flags,
2834 		   struct __kernel_timespec __user *timeout,
2835 		   struct old_timespec32 __user *timeout32)
2836 {
2837 	int datagrams;
2838 	struct timespec64 timeout_sys;
2839 
2840 	if (timeout && get_timespec64(&timeout_sys, timeout))
2841 		return -EFAULT;
2842 
2843 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2844 		return -EFAULT;
2845 
2846 	if (!timeout && !timeout32)
2847 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2848 
2849 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2850 
2851 	if (datagrams <= 0)
2852 		return datagrams;
2853 
2854 	if (timeout && put_timespec64(&timeout_sys, timeout))
2855 		datagrams = -EFAULT;
2856 
2857 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2858 		datagrams = -EFAULT;
2859 
2860 	return datagrams;
2861 }
2862 
2863 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2864 		unsigned int, vlen, unsigned int, flags,
2865 		struct __kernel_timespec __user *, timeout)
2866 {
2867 	if (flags & MSG_CMSG_COMPAT)
2868 		return -EINVAL;
2869 
2870 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2871 }
2872 
2873 #ifdef CONFIG_COMPAT_32BIT_TIME
2874 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2875 		unsigned int, vlen, unsigned int, flags,
2876 		struct old_timespec32 __user *, timeout)
2877 {
2878 	if (flags & MSG_CMSG_COMPAT)
2879 		return -EINVAL;
2880 
2881 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2882 }
2883 #endif
2884 
2885 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2886 /* Argument list sizes for sys_socketcall */
2887 #define AL(x) ((x) * sizeof(unsigned long))
2888 static const unsigned char nargs[21] = {
2889 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2890 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2891 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2892 	AL(4), AL(5), AL(4)
2893 };
2894 
2895 #undef AL
2896 
2897 /*
2898  *	System call vectors.
2899  *
2900  *	Argument checking cleaned up. Saved 20% in size.
2901  *  This function doesn't need to set the kernel lock because
2902  *  it is set by the callees.
2903  */
2904 
2905 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2906 {
2907 	unsigned long a[AUDITSC_ARGS];
2908 	unsigned long a0, a1;
2909 	int err;
2910 	unsigned int len;
2911 
2912 	if (call < 1 || call > SYS_SENDMMSG)
2913 		return -EINVAL;
2914 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2915 
2916 	len = nargs[call];
2917 	if (len > sizeof(a))
2918 		return -EINVAL;
2919 
2920 	/* copy_from_user should be SMP safe. */
2921 	if (copy_from_user(a, args, len))
2922 		return -EFAULT;
2923 
2924 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2925 	if (err)
2926 		return err;
2927 
2928 	a0 = a[0];
2929 	a1 = a[1];
2930 
2931 	switch (call) {
2932 	case SYS_SOCKET:
2933 		err = __sys_socket(a0, a1, a[2]);
2934 		break;
2935 	case SYS_BIND:
2936 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2937 		break;
2938 	case SYS_CONNECT:
2939 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2940 		break;
2941 	case SYS_LISTEN:
2942 		err = __sys_listen(a0, a1);
2943 		break;
2944 	case SYS_ACCEPT:
2945 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2946 				    (int __user *)a[2], 0);
2947 		break;
2948 	case SYS_GETSOCKNAME:
2949 		err =
2950 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
2951 				      (int __user *)a[2]);
2952 		break;
2953 	case SYS_GETPEERNAME:
2954 		err =
2955 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
2956 				      (int __user *)a[2]);
2957 		break;
2958 	case SYS_SOCKETPAIR:
2959 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2960 		break;
2961 	case SYS_SEND:
2962 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2963 				   NULL, 0);
2964 		break;
2965 	case SYS_SENDTO:
2966 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2967 				   (struct sockaddr __user *)a[4], a[5]);
2968 		break;
2969 	case SYS_RECV:
2970 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2971 				     NULL, NULL);
2972 		break;
2973 	case SYS_RECVFROM:
2974 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2975 				     (struct sockaddr __user *)a[4],
2976 				     (int __user *)a[5]);
2977 		break;
2978 	case SYS_SHUTDOWN:
2979 		err = __sys_shutdown(a0, a1);
2980 		break;
2981 	case SYS_SETSOCKOPT:
2982 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2983 				       a[4]);
2984 		break;
2985 	case SYS_GETSOCKOPT:
2986 		err =
2987 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2988 				     (int __user *)a[4]);
2989 		break;
2990 	case SYS_SENDMSG:
2991 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2992 				    a[2], true);
2993 		break;
2994 	case SYS_SENDMMSG:
2995 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2996 				     a[3], true);
2997 		break;
2998 	case SYS_RECVMSG:
2999 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3000 				    a[2], true);
3001 		break;
3002 	case SYS_RECVMMSG:
3003 		if (IS_ENABLED(CONFIG_64BIT))
3004 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3005 					     a[2], a[3],
3006 					     (struct __kernel_timespec __user *)a[4],
3007 					     NULL);
3008 		else
3009 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3010 					     a[2], a[3], NULL,
3011 					     (struct old_timespec32 __user *)a[4]);
3012 		break;
3013 	case SYS_ACCEPT4:
3014 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3015 				    (int __user *)a[2], a[3]);
3016 		break;
3017 	default:
3018 		err = -EINVAL;
3019 		break;
3020 	}
3021 	return err;
3022 }
3023 
3024 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3025 
3026 /**
3027  *	sock_register - add a socket protocol handler
3028  *	@ops: description of protocol
3029  *
3030  *	This function is called by a protocol handler that wants to
3031  *	advertise its address family, and have it linked into the
3032  *	socket interface. The value ops->family corresponds to the
3033  *	socket system call protocol family.
3034  */
3035 int sock_register(const struct net_proto_family *ops)
3036 {
3037 	int err;
3038 
3039 	if (ops->family >= NPROTO) {
3040 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3041 		return -ENOBUFS;
3042 	}
3043 
3044 	spin_lock(&net_family_lock);
3045 	if (rcu_dereference_protected(net_families[ops->family],
3046 				      lockdep_is_held(&net_family_lock)))
3047 		err = -EEXIST;
3048 	else {
3049 		rcu_assign_pointer(net_families[ops->family], ops);
3050 		err = 0;
3051 	}
3052 	spin_unlock(&net_family_lock);
3053 
3054 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3055 	return err;
3056 }
3057 EXPORT_SYMBOL(sock_register);
3058 
3059 /**
3060  *	sock_unregister - remove a protocol handler
3061  *	@family: protocol family to remove
3062  *
3063  *	This function is called by a protocol handler that wants to
3064  *	remove its address family, and have it unlinked from the
3065  *	new socket creation.
3066  *
3067  *	If protocol handler is a module, then it can use module reference
3068  *	counts to protect against new references. If protocol handler is not
3069  *	a module then it needs to provide its own protection in
3070  *	the ops->create routine.
3071  */
3072 void sock_unregister(int family)
3073 {
3074 	BUG_ON(family < 0 || family >= NPROTO);
3075 
3076 	spin_lock(&net_family_lock);
3077 	RCU_INIT_POINTER(net_families[family], NULL);
3078 	spin_unlock(&net_family_lock);
3079 
3080 	synchronize_rcu();
3081 
3082 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3083 }
3084 EXPORT_SYMBOL(sock_unregister);
3085 
3086 bool sock_is_registered(int family)
3087 {
3088 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3089 }
3090 
3091 static int __init sock_init(void)
3092 {
3093 	int err;
3094 	/*
3095 	 *      Initialize the network sysctl infrastructure.
3096 	 */
3097 	err = net_sysctl_init();
3098 	if (err)
3099 		goto out;
3100 
3101 	/*
3102 	 *      Initialize skbuff SLAB cache
3103 	 */
3104 	skb_init();
3105 
3106 	/*
3107 	 *      Initialize the protocols module.
3108 	 */
3109 
3110 	init_inodecache();
3111 
3112 	err = register_filesystem(&sock_fs_type);
3113 	if (err)
3114 		goto out;
3115 	sock_mnt = kern_mount(&sock_fs_type);
3116 	if (IS_ERR(sock_mnt)) {
3117 		err = PTR_ERR(sock_mnt);
3118 		goto out_mount;
3119 	}
3120 
3121 	/* The real protocol initialization is performed in later initcalls.
3122 	 */
3123 
3124 #ifdef CONFIG_NETFILTER
3125 	err = netfilter_init();
3126 	if (err)
3127 		goto out;
3128 #endif
3129 
3130 	ptp_classifier_init();
3131 
3132 out:
3133 	return err;
3134 
3135 out_mount:
3136 	unregister_filesystem(&sock_fs_type);
3137 	goto out;
3138 }
3139 
3140 core_initcall(sock_init);	/* early initcall */
3141 
3142 #ifdef CONFIG_PROC_FS
3143 void socket_seq_show(struct seq_file *seq)
3144 {
3145 	seq_printf(seq, "sockets: used %d\n",
3146 		   sock_inuse_get(seq->private));
3147 }
3148 #endif				/* CONFIG_PROC_FS */
3149 
3150 /* Handle the fact that while struct ifreq has the same *layout* on
3151  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3152  * which are handled elsewhere, it still has different *size* due to
3153  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3154  * resulting in struct ifreq being 32 and 40 bytes respectively).
3155  * As a result, if the struct happens to be at the end of a page and
3156  * the next page isn't readable/writable, we get a fault. To prevent
3157  * that, copy back and forth to the full size.
3158  */
3159 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3160 {
3161 	if (in_compat_syscall()) {
3162 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3163 
3164 		memset(ifr, 0, sizeof(*ifr));
3165 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3166 			return -EFAULT;
3167 
3168 		if (ifrdata)
3169 			*ifrdata = compat_ptr(ifr32->ifr_data);
3170 
3171 		return 0;
3172 	}
3173 
3174 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3175 		return -EFAULT;
3176 
3177 	if (ifrdata)
3178 		*ifrdata = ifr->ifr_data;
3179 
3180 	return 0;
3181 }
3182 EXPORT_SYMBOL(get_user_ifreq);
3183 
3184 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3185 {
3186 	size_t size = sizeof(*ifr);
3187 
3188 	if (in_compat_syscall())
3189 		size = sizeof(struct compat_ifreq);
3190 
3191 	if (copy_to_user(arg, ifr, size))
3192 		return -EFAULT;
3193 
3194 	return 0;
3195 }
3196 EXPORT_SYMBOL(put_user_ifreq);
3197 
3198 #ifdef CONFIG_COMPAT
3199 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3200 {
3201 	compat_uptr_t uptr32;
3202 	struct ifreq ifr;
3203 	void __user *saved;
3204 	int err;
3205 
3206 	if (get_user_ifreq(&ifr, NULL, uifr32))
3207 		return -EFAULT;
3208 
3209 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3210 		return -EFAULT;
3211 
3212 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3213 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3214 
3215 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3216 	if (!err) {
3217 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3218 		if (put_user_ifreq(&ifr, uifr32))
3219 			err = -EFAULT;
3220 	}
3221 	return err;
3222 }
3223 
3224 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3225 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3226 				 struct compat_ifreq __user *u_ifreq32)
3227 {
3228 	struct ifreq ifreq;
3229 	void __user *data;
3230 
3231 	if (!is_socket_ioctl_cmd(cmd))
3232 		return -ENOTTY;
3233 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3234 		return -EFAULT;
3235 	ifreq.ifr_data = data;
3236 
3237 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3238 }
3239 
3240 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3241 			 unsigned int cmd, unsigned long arg)
3242 {
3243 	void __user *argp = compat_ptr(arg);
3244 	struct sock *sk = sock->sk;
3245 	struct net *net = sock_net(sk);
3246 
3247 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3248 		return sock_ioctl(file, cmd, (unsigned long)argp);
3249 
3250 	switch (cmd) {
3251 	case SIOCWANDEV:
3252 		return compat_siocwandev(net, argp);
3253 	case SIOCGSTAMP_OLD:
3254 	case SIOCGSTAMPNS_OLD:
3255 		if (!sock->ops->gettstamp)
3256 			return -ENOIOCTLCMD;
3257 		return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3258 					    !COMPAT_USE_64BIT_TIME);
3259 
3260 	case SIOCETHTOOL:
3261 	case SIOCBONDSLAVEINFOQUERY:
3262 	case SIOCBONDINFOQUERY:
3263 	case SIOCSHWTSTAMP:
3264 	case SIOCGHWTSTAMP:
3265 		return compat_ifr_data_ioctl(net, cmd, argp);
3266 
3267 	case FIOSETOWN:
3268 	case SIOCSPGRP:
3269 	case FIOGETOWN:
3270 	case SIOCGPGRP:
3271 	case SIOCBRADDBR:
3272 	case SIOCBRDELBR:
3273 	case SIOCGIFVLAN:
3274 	case SIOCSIFVLAN:
3275 	case SIOCGSKNS:
3276 	case SIOCGSTAMP_NEW:
3277 	case SIOCGSTAMPNS_NEW:
3278 	case SIOCGIFCONF:
3279 	case SIOCSIFBR:
3280 	case SIOCGIFBR:
3281 		return sock_ioctl(file, cmd, arg);
3282 
3283 	case SIOCGIFFLAGS:
3284 	case SIOCSIFFLAGS:
3285 	case SIOCGIFMAP:
3286 	case SIOCSIFMAP:
3287 	case SIOCGIFMETRIC:
3288 	case SIOCSIFMETRIC:
3289 	case SIOCGIFMTU:
3290 	case SIOCSIFMTU:
3291 	case SIOCGIFMEM:
3292 	case SIOCSIFMEM:
3293 	case SIOCGIFHWADDR:
3294 	case SIOCSIFHWADDR:
3295 	case SIOCADDMULTI:
3296 	case SIOCDELMULTI:
3297 	case SIOCGIFINDEX:
3298 	case SIOCGIFADDR:
3299 	case SIOCSIFADDR:
3300 	case SIOCSIFHWBROADCAST:
3301 	case SIOCDIFADDR:
3302 	case SIOCGIFBRDADDR:
3303 	case SIOCSIFBRDADDR:
3304 	case SIOCGIFDSTADDR:
3305 	case SIOCSIFDSTADDR:
3306 	case SIOCGIFNETMASK:
3307 	case SIOCSIFNETMASK:
3308 	case SIOCSIFPFLAGS:
3309 	case SIOCGIFPFLAGS:
3310 	case SIOCGIFTXQLEN:
3311 	case SIOCSIFTXQLEN:
3312 	case SIOCBRADDIF:
3313 	case SIOCBRDELIF:
3314 	case SIOCGIFNAME:
3315 	case SIOCSIFNAME:
3316 	case SIOCGMIIPHY:
3317 	case SIOCGMIIREG:
3318 	case SIOCSMIIREG:
3319 	case SIOCBONDENSLAVE:
3320 	case SIOCBONDRELEASE:
3321 	case SIOCBONDSETHWADDR:
3322 	case SIOCBONDCHANGEACTIVE:
3323 	case SIOCSARP:
3324 	case SIOCGARP:
3325 	case SIOCDARP:
3326 	case SIOCOUTQ:
3327 	case SIOCOUTQNSD:
3328 	case SIOCATMARK:
3329 		return sock_do_ioctl(net, sock, cmd, arg);
3330 	}
3331 
3332 	return -ENOIOCTLCMD;
3333 }
3334 
3335 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3336 			      unsigned long arg)
3337 {
3338 	struct socket *sock = file->private_data;
3339 	int ret = -ENOIOCTLCMD;
3340 	struct sock *sk;
3341 	struct net *net;
3342 
3343 	sk = sock->sk;
3344 	net = sock_net(sk);
3345 
3346 	if (sock->ops->compat_ioctl)
3347 		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3348 
3349 	if (ret == -ENOIOCTLCMD &&
3350 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3351 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3352 
3353 	if (ret == -ENOIOCTLCMD)
3354 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3355 
3356 	return ret;
3357 }
3358 #endif
3359 
3360 /**
3361  *	kernel_bind - bind an address to a socket (kernel space)
3362  *	@sock: socket
3363  *	@addr: address
3364  *	@addrlen: length of address
3365  *
3366  *	Returns 0 or an error.
3367  */
3368 
3369 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3370 {
3371 	return sock->ops->bind(sock, addr, addrlen);
3372 }
3373 EXPORT_SYMBOL(kernel_bind);
3374 
3375 /**
3376  *	kernel_listen - move socket to listening state (kernel space)
3377  *	@sock: socket
3378  *	@backlog: pending connections queue size
3379  *
3380  *	Returns 0 or an error.
3381  */
3382 
3383 int kernel_listen(struct socket *sock, int backlog)
3384 {
3385 	return sock->ops->listen(sock, backlog);
3386 }
3387 EXPORT_SYMBOL(kernel_listen);
3388 
3389 /**
3390  *	kernel_accept - accept a connection (kernel space)
3391  *	@sock: listening socket
3392  *	@newsock: new connected socket
3393  *	@flags: flags
3394  *
3395  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3396  *	If it fails, @newsock is guaranteed to be %NULL.
3397  *	Returns 0 or an error.
3398  */
3399 
3400 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3401 {
3402 	struct sock *sk = sock->sk;
3403 	int err;
3404 
3405 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3406 			       newsock);
3407 	if (err < 0)
3408 		goto done;
3409 
3410 	err = sock->ops->accept(sock, *newsock, flags, true);
3411 	if (err < 0) {
3412 		sock_release(*newsock);
3413 		*newsock = NULL;
3414 		goto done;
3415 	}
3416 
3417 	(*newsock)->ops = sock->ops;
3418 	__module_get((*newsock)->ops->owner);
3419 
3420 done:
3421 	return err;
3422 }
3423 EXPORT_SYMBOL(kernel_accept);
3424 
3425 /**
3426  *	kernel_connect - connect a socket (kernel space)
3427  *	@sock: socket
3428  *	@addr: address
3429  *	@addrlen: address length
3430  *	@flags: flags (O_NONBLOCK, ...)
3431  *
3432  *	For datagram sockets, @addr is the address to which datagrams are sent
3433  *	by default, and the only address from which datagrams are received.
3434  *	For stream sockets, attempts to connect to @addr.
3435  *	Returns 0 or an error code.
3436  */
3437 
3438 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3439 		   int flags)
3440 {
3441 	return sock->ops->connect(sock, addr, addrlen, flags);
3442 }
3443 EXPORT_SYMBOL(kernel_connect);
3444 
3445 /**
3446  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3447  *	@sock: socket
3448  *	@addr: address holder
3449  *
3450  * 	Fills the @addr pointer with the address which the socket is bound.
3451  *	Returns the length of the address in bytes or an error code.
3452  */
3453 
3454 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3455 {
3456 	return sock->ops->getname(sock, addr, 0);
3457 }
3458 EXPORT_SYMBOL(kernel_getsockname);
3459 
3460 /**
3461  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3462  *	@sock: socket
3463  *	@addr: address holder
3464  *
3465  * 	Fills the @addr pointer with the address which the socket is connected.
3466  *	Returns the length of the address in bytes or an error code.
3467  */
3468 
3469 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3470 {
3471 	return sock->ops->getname(sock, addr, 1);
3472 }
3473 EXPORT_SYMBOL(kernel_getpeername);
3474 
3475 /**
3476  *	kernel_sendpage - send a &page through a socket (kernel space)
3477  *	@sock: socket
3478  *	@page: page
3479  *	@offset: page offset
3480  *	@size: total size in bytes
3481  *	@flags: flags (MSG_DONTWAIT, ...)
3482  *
3483  *	Returns the total amount sent in bytes or an error.
3484  */
3485 
3486 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3487 		    size_t size, int flags)
3488 {
3489 	if (sock->ops->sendpage) {
3490 		/* Warn in case the improper page to zero-copy send */
3491 		WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3492 		return sock->ops->sendpage(sock, page, offset, size, flags);
3493 	}
3494 	return sock_no_sendpage(sock, page, offset, size, flags);
3495 }
3496 EXPORT_SYMBOL(kernel_sendpage);
3497 
3498 /**
3499  *	kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3500  *	@sk: sock
3501  *	@page: page
3502  *	@offset: page offset
3503  *	@size: total size in bytes
3504  *	@flags: flags (MSG_DONTWAIT, ...)
3505  *
3506  *	Returns the total amount sent in bytes or an error.
3507  *	Caller must hold @sk.
3508  */
3509 
3510 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3511 			   size_t size, int flags)
3512 {
3513 	struct socket *sock = sk->sk_socket;
3514 
3515 	if (sock->ops->sendpage_locked)
3516 		return sock->ops->sendpage_locked(sk, page, offset, size,
3517 						  flags);
3518 
3519 	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3520 }
3521 EXPORT_SYMBOL(kernel_sendpage_locked);
3522 
3523 /**
3524  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3525  *	@sock: socket
3526  *	@how: connection part
3527  *
3528  *	Returns 0 or an error.
3529  */
3530 
3531 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3532 {
3533 	return sock->ops->shutdown(sock, how);
3534 }
3535 EXPORT_SYMBOL(kernel_sock_shutdown);
3536 
3537 /**
3538  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3539  *	@sk: socket
3540  *
3541  *	This routine returns the IP overhead imposed by a socket i.e.
3542  *	the length of the underlying IP header, depending on whether
3543  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3544  *	on at the socket. Assumes that the caller has a lock on the socket.
3545  */
3546 
3547 u32 kernel_sock_ip_overhead(struct sock *sk)
3548 {
3549 	struct inet_sock *inet;
3550 	struct ip_options_rcu *opt;
3551 	u32 overhead = 0;
3552 #if IS_ENABLED(CONFIG_IPV6)
3553 	struct ipv6_pinfo *np;
3554 	struct ipv6_txoptions *optv6 = NULL;
3555 #endif /* IS_ENABLED(CONFIG_IPV6) */
3556 
3557 	if (!sk)
3558 		return overhead;
3559 
3560 	switch (sk->sk_family) {
3561 	case AF_INET:
3562 		inet = inet_sk(sk);
3563 		overhead += sizeof(struct iphdr);
3564 		opt = rcu_dereference_protected(inet->inet_opt,
3565 						sock_owned_by_user(sk));
3566 		if (opt)
3567 			overhead += opt->opt.optlen;
3568 		return overhead;
3569 #if IS_ENABLED(CONFIG_IPV6)
3570 	case AF_INET6:
3571 		np = inet6_sk(sk);
3572 		overhead += sizeof(struct ipv6hdr);
3573 		if (np)
3574 			optv6 = rcu_dereference_protected(np->opt,
3575 							  sock_owned_by_user(sk));
3576 		if (optv6)
3577 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3578 		return overhead;
3579 #endif /* IS_ENABLED(CONFIG_IPV6) */
3580 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3581 		return overhead;
3582 	}
3583 }
3584 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3585