xref: /openbmc/linux/net/socket.c (revision 612a462a)
1 /*
2  * NET		An implementation of the SOCKET network access protocol.
3  *
4  * Version:	@(#)socket.c	1.1.93	18/02/95
5  *
6  * Authors:	Orest Zborowski, <obz@Kodak.COM>
7  *		Ross Biro
8  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9  *
10  * Fixes:
11  *		Anonymous	:	NOTSOCK/BADF cleanup. Error fix in
12  *					shutdown()
13  *		Alan Cox	:	verify_area() fixes
14  *		Alan Cox	:	Removed DDI
15  *		Jonathan Kamens	:	SOCK_DGRAM reconnect bug
16  *		Alan Cox	:	Moved a load of checks to the very
17  *					top level.
18  *		Alan Cox	:	Move address structures to/from user
19  *					mode above the protocol layers.
20  *		Rob Janssen	:	Allow 0 length sends.
21  *		Alan Cox	:	Asynchronous I/O support (cribbed from the
22  *					tty drivers).
23  *		Niibe Yutaka	:	Asynchronous I/O for writes (4.4BSD style)
24  *		Jeff Uphoff	:	Made max number of sockets command-line
25  *					configurable.
26  *		Matti Aarnio	:	Made the number of sockets dynamic,
27  *					to be allocated when needed, and mr.
28  *					Uphoff's max is used as max to be
29  *					allowed to allocate.
30  *		Linus		:	Argh. removed all the socket allocation
31  *					altogether: it's in the inode now.
32  *		Alan Cox	:	Made sock_alloc()/sock_release() public
33  *					for NetROM and future kernel nfsd type
34  *					stuff.
35  *		Alan Cox	:	sendmsg/recvmsg basics.
36  *		Tom Dyas	:	Export net symbols.
37  *		Marcin Dalecki	:	Fixed problems with CONFIG_NET="n".
38  *		Alan Cox	:	Added thread locking to sys_* calls
39  *					for sockets. May have errors at the
40  *					moment.
41  *		Kevin Buhr	:	Fixed the dumb errors in the above.
42  *		Andi Kleen	:	Some small cleanups, optimizations,
43  *					and fixed a copy_from_user() bug.
44  *		Tigran Aivazian	:	sys_send(args) calls sys_sendto(args, NULL, 0)
45  *		Tigran Aivazian	:	Made listen(2) backlog sanity checks
46  *					protocol-independent
47  *
48  *
49  *		This program is free software; you can redistribute it and/or
50  *		modify it under the terms of the GNU General Public License
51  *		as published by the Free Software Foundation; either version
52  *		2 of the License, or (at your option) any later version.
53  *
54  *
55  *	This module is effectively the top level interface to the BSD socket
56  *	paradigm.
57  *
58  *	Based upon Swansea University Computer Society NET3.039
59  */
60 
61 #include <linux/mm.h>
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
92 
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
95 
96 #include <net/compat.h>
97 #include <net/wext.h>
98 #include <net/cls_cgroup.h>
99 
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102 
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115 
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119 
120 static int sock_close(struct inode *inode, struct file *file);
121 static unsigned int sock_poll(struct file *file,
122 			      struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124 #ifdef CONFIG_COMPAT
125 static long compat_sock_ioctl(struct file *file,
126 			      unsigned int cmd, unsigned long arg);
127 #endif
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 			     int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 				struct pipe_inode_info *pipe, size_t len,
133 				unsigned int flags);
134 
135 /*
136  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137  *	in the operation structures but are done directly via the socketcall() multiplexor.
138  */
139 
140 static const struct file_operations socket_file_ops = {
141 	.owner =	THIS_MODULE,
142 	.llseek =	no_llseek,
143 	.read_iter =	sock_read_iter,
144 	.write_iter =	sock_write_iter,
145 	.poll =		sock_poll,
146 	.unlocked_ioctl = sock_ioctl,
147 #ifdef CONFIG_COMPAT
148 	.compat_ioctl = compat_sock_ioctl,
149 #endif
150 	.mmap =		sock_mmap,
151 	.release =	sock_close,
152 	.fasync =	sock_fasync,
153 	.sendpage =	sock_sendpage,
154 	.splice_write = generic_splice_sendpage,
155 	.splice_read =	sock_splice_read,
156 };
157 
158 /*
159  *	The protocol list. Each protocol is registered in here.
160  */
161 
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164 
165 /*
166  *	Statistics counters of the socket lists
167  */
168 
169 static DEFINE_PER_CPU(int, sockets_in_use);
170 
171 /*
172  * Support routines.
173  * Move socket addresses back and forth across the kernel/user
174  * divide and look after the messy bits.
175  */
176 
177 /**
178  *	move_addr_to_kernel	-	copy a socket address into kernel space
179  *	@uaddr: Address in user space
180  *	@kaddr: Address in kernel space
181  *	@ulen: Length in user space
182  *
183  *	The address is copied into kernel space. If the provided address is
184  *	too long an error code of -EINVAL is returned. If the copy gives
185  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
186  */
187 
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
189 {
190 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191 		return -EINVAL;
192 	if (ulen == 0)
193 		return 0;
194 	if (copy_from_user(kaddr, uaddr, ulen))
195 		return -EFAULT;
196 	return audit_sockaddr(ulen, kaddr);
197 }
198 
199 /**
200  *	move_addr_to_user	-	copy an address to user space
201  *	@kaddr: kernel space address
202  *	@klen: length of address in kernel
203  *	@uaddr: user space address
204  *	@ulen: pointer to user length field
205  *
206  *	The value pointed to by ulen on entry is the buffer length available.
207  *	This is overwritten with the buffer space used. -EINVAL is returned
208  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
209  *	is returned if either the buffer or the length field are not
210  *	accessible.
211  *	After copying the data up to the limit the user specifies, the true
212  *	length of the data is written over the length limit the user
213  *	specified. Zero is returned for a success.
214  */
215 
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217 			     void __user *uaddr, int __user *ulen)
218 {
219 	int err;
220 	int len;
221 
222 	BUG_ON(klen > sizeof(struct sockaddr_storage));
223 	err = get_user(len, ulen);
224 	if (err)
225 		return err;
226 	if (len > klen)
227 		len = klen;
228 	if (len < 0)
229 		return -EINVAL;
230 	if (len) {
231 		if (audit_sockaddr(klen, kaddr))
232 			return -ENOMEM;
233 		if (copy_to_user(uaddr, kaddr, len))
234 			return -EFAULT;
235 	}
236 	/*
237 	 *      "fromlen shall refer to the value before truncation.."
238 	 *                      1003.1g
239 	 */
240 	return __put_user(klen, ulen);
241 }
242 
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
244 
245 static struct inode *sock_alloc_inode(struct super_block *sb)
246 {
247 	struct socket_alloc *ei;
248 	struct socket_wq *wq;
249 
250 	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
251 	if (!ei)
252 		return NULL;
253 	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
254 	if (!wq) {
255 		kmem_cache_free(sock_inode_cachep, ei);
256 		return NULL;
257 	}
258 	init_waitqueue_head(&wq->wait);
259 	wq->fasync_list = NULL;
260 	wq->flags = 0;
261 	RCU_INIT_POINTER(ei->socket.wq, wq);
262 
263 	ei->socket.state = SS_UNCONNECTED;
264 	ei->socket.flags = 0;
265 	ei->socket.ops = NULL;
266 	ei->socket.sk = NULL;
267 	ei->socket.file = NULL;
268 
269 	return &ei->vfs_inode;
270 }
271 
272 static void sock_destroy_inode(struct inode *inode)
273 {
274 	struct socket_alloc *ei;
275 	struct socket_wq *wq;
276 
277 	ei = container_of(inode, struct socket_alloc, vfs_inode);
278 	wq = rcu_dereference_protected(ei->socket.wq, 1);
279 	kfree_rcu(wq, rcu);
280 	kmem_cache_free(sock_inode_cachep, ei);
281 }
282 
283 static void init_once(void *foo)
284 {
285 	struct socket_alloc *ei = (struct socket_alloc *)foo;
286 
287 	inode_init_once(&ei->vfs_inode);
288 }
289 
290 static void init_inodecache(void)
291 {
292 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293 					      sizeof(struct socket_alloc),
294 					      0,
295 					      (SLAB_HWCACHE_ALIGN |
296 					       SLAB_RECLAIM_ACCOUNT |
297 					       SLAB_MEM_SPREAD | SLAB_ACCOUNT),
298 					      init_once);
299 	BUG_ON(sock_inode_cachep == NULL);
300 }
301 
302 static const struct super_operations sockfs_ops = {
303 	.alloc_inode	= sock_alloc_inode,
304 	.destroy_inode	= sock_destroy_inode,
305 	.statfs		= simple_statfs,
306 };
307 
308 /*
309  * sockfs_dname() is called from d_path().
310  */
311 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
312 {
313 	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
314 				d_inode(dentry)->i_ino);
315 }
316 
317 static const struct dentry_operations sockfs_dentry_operations = {
318 	.d_dname  = sockfs_dname,
319 };
320 
321 static int sockfs_xattr_get(const struct xattr_handler *handler,
322 			    struct dentry *dentry, struct inode *inode,
323 			    const char *suffix, void *value, size_t size)
324 {
325 	if (value) {
326 		if (dentry->d_name.len + 1 > size)
327 			return -ERANGE;
328 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
329 	}
330 	return dentry->d_name.len + 1;
331 }
332 
333 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
336 
337 static const struct xattr_handler sockfs_xattr_handler = {
338 	.name = XATTR_NAME_SOCKPROTONAME,
339 	.get = sockfs_xattr_get,
340 };
341 
342 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
343 				     struct dentry *dentry, struct inode *inode,
344 				     const char *suffix, const void *value,
345 				     size_t size, int flags)
346 {
347 	/* Handled by LSM. */
348 	return -EAGAIN;
349 }
350 
351 static const struct xattr_handler sockfs_security_xattr_handler = {
352 	.prefix = XATTR_SECURITY_PREFIX,
353 	.set = sockfs_security_xattr_set,
354 };
355 
356 static const struct xattr_handler *sockfs_xattr_handlers[] = {
357 	&sockfs_xattr_handler,
358 	&sockfs_security_xattr_handler,
359 	NULL
360 };
361 
362 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
363 			 int flags, const char *dev_name, void *data)
364 {
365 	return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
366 				  sockfs_xattr_handlers,
367 				  &sockfs_dentry_operations, SOCKFS_MAGIC);
368 }
369 
370 static struct vfsmount *sock_mnt __read_mostly;
371 
372 static struct file_system_type sock_fs_type = {
373 	.name =		"sockfs",
374 	.mount =	sockfs_mount,
375 	.kill_sb =	kill_anon_super,
376 };
377 
378 /*
379  *	Obtains the first available file descriptor and sets it up for use.
380  *
381  *	These functions create file structures and maps them to fd space
382  *	of the current process. On success it returns file descriptor
383  *	and file struct implicitly stored in sock->file.
384  *	Note that another thread may close file descriptor before we return
385  *	from this function. We use the fact that now we do not refer
386  *	to socket after mapping. If one day we will need it, this
387  *	function will increment ref. count on file by 1.
388  *
389  *	In any case returned fd MAY BE not valid!
390  *	This race condition is unavoidable
391  *	with shared fd spaces, we cannot solve it inside kernel,
392  *	but we take care of internal coherence yet.
393  */
394 
395 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
396 {
397 	struct qstr name = { .name = "" };
398 	struct path path;
399 	struct file *file;
400 
401 	if (dname) {
402 		name.name = dname;
403 		name.len = strlen(name.name);
404 	} else if (sock->sk) {
405 		name.name = sock->sk->sk_prot_creator->name;
406 		name.len = strlen(name.name);
407 	}
408 	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
409 	if (unlikely(!path.dentry))
410 		return ERR_PTR(-ENOMEM);
411 	path.mnt = mntget(sock_mnt);
412 
413 	d_instantiate(path.dentry, SOCK_INODE(sock));
414 
415 	file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
416 		  &socket_file_ops);
417 	if (IS_ERR(file)) {
418 		/* drop dentry, keep inode */
419 		ihold(d_inode(path.dentry));
420 		path_put(&path);
421 		return file;
422 	}
423 
424 	sock->file = file;
425 	file->f_flags = O_RDWR | (flags & O_NONBLOCK);
426 	file->private_data = sock;
427 	return file;
428 }
429 EXPORT_SYMBOL(sock_alloc_file);
430 
431 static int sock_map_fd(struct socket *sock, int flags)
432 {
433 	struct file *newfile;
434 	int fd = get_unused_fd_flags(flags);
435 	if (unlikely(fd < 0))
436 		return fd;
437 
438 	newfile = sock_alloc_file(sock, flags, NULL);
439 	if (likely(!IS_ERR(newfile))) {
440 		fd_install(fd, newfile);
441 		return fd;
442 	}
443 
444 	put_unused_fd(fd);
445 	return PTR_ERR(newfile);
446 }
447 
448 struct socket *sock_from_file(struct file *file, int *err)
449 {
450 	if (file->f_op == &socket_file_ops)
451 		return file->private_data;	/* set in sock_map_fd */
452 
453 	*err = -ENOTSOCK;
454 	return NULL;
455 }
456 EXPORT_SYMBOL(sock_from_file);
457 
458 /**
459  *	sockfd_lookup - Go from a file number to its socket slot
460  *	@fd: file handle
461  *	@err: pointer to an error code return
462  *
463  *	The file handle passed in is locked and the socket it is bound
464  *	to is returned. If an error occurs the err pointer is overwritten
465  *	with a negative errno code and NULL is returned. The function checks
466  *	for both invalid handles and passing a handle which is not a socket.
467  *
468  *	On a success the socket object pointer is returned.
469  */
470 
471 struct socket *sockfd_lookup(int fd, int *err)
472 {
473 	struct file *file;
474 	struct socket *sock;
475 
476 	file = fget(fd);
477 	if (!file) {
478 		*err = -EBADF;
479 		return NULL;
480 	}
481 
482 	sock = sock_from_file(file, err);
483 	if (!sock)
484 		fput(file);
485 	return sock;
486 }
487 EXPORT_SYMBOL(sockfd_lookup);
488 
489 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
490 {
491 	struct fd f = fdget(fd);
492 	struct socket *sock;
493 
494 	*err = -EBADF;
495 	if (f.file) {
496 		sock = sock_from_file(f.file, err);
497 		if (likely(sock)) {
498 			*fput_needed = f.flags;
499 			return sock;
500 		}
501 		fdput(f);
502 	}
503 	return NULL;
504 }
505 
506 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
507 				size_t size)
508 {
509 	ssize_t len;
510 	ssize_t used = 0;
511 
512 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
513 	if (len < 0)
514 		return len;
515 	used += len;
516 	if (buffer) {
517 		if (size < used)
518 			return -ERANGE;
519 		buffer += len;
520 	}
521 
522 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
523 	used += len;
524 	if (buffer) {
525 		if (size < used)
526 			return -ERANGE;
527 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
528 		buffer += len;
529 	}
530 
531 	return used;
532 }
533 
534 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
535 {
536 	int err = simple_setattr(dentry, iattr);
537 
538 	if (!err && (iattr->ia_valid & ATTR_UID)) {
539 		struct socket *sock = SOCKET_I(d_inode(dentry));
540 
541 		sock->sk->sk_uid = iattr->ia_uid;
542 	}
543 
544 	return err;
545 }
546 
547 static const struct inode_operations sockfs_inode_ops = {
548 	.listxattr = sockfs_listxattr,
549 	.setattr = sockfs_setattr,
550 };
551 
552 /**
553  *	sock_alloc	-	allocate a socket
554  *
555  *	Allocate a new inode and socket object. The two are bound together
556  *	and initialised. The socket is then returned. If we are out of inodes
557  *	NULL is returned.
558  */
559 
560 struct socket *sock_alloc(void)
561 {
562 	struct inode *inode;
563 	struct socket *sock;
564 
565 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
566 	if (!inode)
567 		return NULL;
568 
569 	sock = SOCKET_I(inode);
570 
571 	kmemcheck_annotate_bitfield(sock, type);
572 	inode->i_ino = get_next_ino();
573 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
574 	inode->i_uid = current_fsuid();
575 	inode->i_gid = current_fsgid();
576 	inode->i_op = &sockfs_inode_ops;
577 
578 	this_cpu_add(sockets_in_use, 1);
579 	return sock;
580 }
581 EXPORT_SYMBOL(sock_alloc);
582 
583 /**
584  *	sock_release	-	close a socket
585  *	@sock: socket to close
586  *
587  *	The socket is released from the protocol stack if it has a release
588  *	callback, and the inode is then released if the socket is bound to
589  *	an inode not a file.
590  */
591 
592 void sock_release(struct socket *sock)
593 {
594 	if (sock->ops) {
595 		struct module *owner = sock->ops->owner;
596 
597 		sock->ops->release(sock);
598 		sock->ops = NULL;
599 		module_put(owner);
600 	}
601 
602 	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
603 		pr_err("%s: fasync list not empty!\n", __func__);
604 
605 	this_cpu_sub(sockets_in_use, 1);
606 	if (!sock->file) {
607 		iput(SOCK_INODE(sock));
608 		return;
609 	}
610 	sock->file = NULL;
611 }
612 EXPORT_SYMBOL(sock_release);
613 
614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
615 {
616 	u8 flags = *tx_flags;
617 
618 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
619 		flags |= SKBTX_HW_TSTAMP;
620 
621 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
622 		flags |= SKBTX_SW_TSTAMP;
623 
624 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
625 		flags |= SKBTX_SCHED_TSTAMP;
626 
627 	*tx_flags = flags;
628 }
629 EXPORT_SYMBOL(__sock_tx_timestamp);
630 
631 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
632 {
633 	int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
634 	BUG_ON(ret == -EIOCBQUEUED);
635 	return ret;
636 }
637 
638 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
639 {
640 	int err = security_socket_sendmsg(sock, msg,
641 					  msg_data_left(msg));
642 
643 	return err ?: sock_sendmsg_nosec(sock, msg);
644 }
645 EXPORT_SYMBOL(sock_sendmsg);
646 
647 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648 		   struct kvec *vec, size_t num, size_t size)
649 {
650 	iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
651 	return sock_sendmsg(sock, msg);
652 }
653 EXPORT_SYMBOL(kernel_sendmsg);
654 
655 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
656 			  struct kvec *vec, size_t num, size_t size)
657 {
658 	struct socket *sock = sk->sk_socket;
659 
660 	if (!sock->ops->sendmsg_locked)
661 		return sock_no_sendmsg_locked(sk, msg, size);
662 
663 	iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
664 
665 	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
666 }
667 EXPORT_SYMBOL(kernel_sendmsg_locked);
668 
669 static bool skb_is_err_queue(const struct sk_buff *skb)
670 {
671 	/* pkt_type of skbs enqueued on the error queue are set to
672 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
673 	 * in recvmsg, since skbs received on a local socket will never
674 	 * have a pkt_type of PACKET_OUTGOING.
675 	 */
676 	return skb->pkt_type == PACKET_OUTGOING;
677 }
678 
679 /* On transmit, software and hardware timestamps are returned independently.
680  * As the two skb clones share the hardware timestamp, which may be updated
681  * before the software timestamp is received, a hardware TX timestamp may be
682  * returned only if there is no software TX timestamp. Ignore false software
683  * timestamps, which may be made in the __sock_recv_timestamp() call when the
684  * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
685  * hardware timestamp.
686  */
687 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
688 {
689 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
690 }
691 
692 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
693 {
694 	struct scm_ts_pktinfo ts_pktinfo;
695 	struct net_device *orig_dev;
696 
697 	if (!skb_mac_header_was_set(skb))
698 		return;
699 
700 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
701 
702 	rcu_read_lock();
703 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
704 	if (orig_dev)
705 		ts_pktinfo.if_index = orig_dev->ifindex;
706 	rcu_read_unlock();
707 
708 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
709 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
710 		 sizeof(ts_pktinfo), &ts_pktinfo);
711 }
712 
713 /*
714  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
715  */
716 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
717 	struct sk_buff *skb)
718 {
719 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
720 	struct scm_timestamping tss;
721 	int empty = 1, false_tstamp = 0;
722 	struct skb_shared_hwtstamps *shhwtstamps =
723 		skb_hwtstamps(skb);
724 
725 	/* Race occurred between timestamp enabling and packet
726 	   receiving.  Fill in the current time for now. */
727 	if (need_software_tstamp && skb->tstamp == 0) {
728 		__net_timestamp(skb);
729 		false_tstamp = 1;
730 	}
731 
732 	if (need_software_tstamp) {
733 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
734 			struct timeval tv;
735 			skb_get_timestamp(skb, &tv);
736 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
737 				 sizeof(tv), &tv);
738 		} else {
739 			struct timespec ts;
740 			skb_get_timestampns(skb, &ts);
741 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
742 				 sizeof(ts), &ts);
743 		}
744 	}
745 
746 	memset(&tss, 0, sizeof(tss));
747 	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
748 	    ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
749 		empty = 0;
750 	if (shhwtstamps &&
751 	    (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
752 	    !skb_is_swtx_tstamp(skb, false_tstamp) &&
753 	    ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
754 		empty = 0;
755 		if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
756 		    !skb_is_err_queue(skb))
757 			put_ts_pktinfo(msg, skb);
758 	}
759 	if (!empty) {
760 		put_cmsg(msg, SOL_SOCKET,
761 			 SCM_TIMESTAMPING, sizeof(tss), &tss);
762 
763 		if (skb_is_err_queue(skb) && skb->len &&
764 		    SKB_EXT_ERR(skb)->opt_stats)
765 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
766 				 skb->len, skb->data);
767 	}
768 }
769 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
770 
771 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
772 	struct sk_buff *skb)
773 {
774 	int ack;
775 
776 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
777 		return;
778 	if (!skb->wifi_acked_valid)
779 		return;
780 
781 	ack = skb->wifi_acked;
782 
783 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
784 }
785 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
786 
787 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
788 				   struct sk_buff *skb)
789 {
790 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
791 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
792 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
793 }
794 
795 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
796 	struct sk_buff *skb)
797 {
798 	sock_recv_timestamp(msg, sk, skb);
799 	sock_recv_drops(msg, sk, skb);
800 }
801 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
802 
803 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
804 				     int flags)
805 {
806 	return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
807 }
808 
809 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
810 {
811 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
812 
813 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
814 }
815 EXPORT_SYMBOL(sock_recvmsg);
816 
817 /**
818  * kernel_recvmsg - Receive a message from a socket (kernel space)
819  * @sock:       The socket to receive the message from
820  * @msg:        Received message
821  * @vec:        Input s/g array for message data
822  * @num:        Size of input s/g array
823  * @size:       Number of bytes to read
824  * @flags:      Message flags (MSG_DONTWAIT, etc...)
825  *
826  * On return the msg structure contains the scatter/gather array passed in the
827  * vec argument. The array is modified so that it consists of the unfilled
828  * portion of the original array.
829  *
830  * The returned value is the total number of bytes received, or an error.
831  */
832 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
833 		   struct kvec *vec, size_t num, size_t size, int flags)
834 {
835 	mm_segment_t oldfs = get_fs();
836 	int result;
837 
838 	iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
839 	set_fs(KERNEL_DS);
840 	result = sock_recvmsg(sock, msg, flags);
841 	set_fs(oldfs);
842 	return result;
843 }
844 EXPORT_SYMBOL(kernel_recvmsg);
845 
846 static ssize_t sock_sendpage(struct file *file, struct page *page,
847 			     int offset, size_t size, loff_t *ppos, int more)
848 {
849 	struct socket *sock;
850 	int flags;
851 
852 	sock = file->private_data;
853 
854 	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
855 	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
856 	flags |= more;
857 
858 	return kernel_sendpage(sock, page, offset, size, flags);
859 }
860 
861 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
862 				struct pipe_inode_info *pipe, size_t len,
863 				unsigned int flags)
864 {
865 	struct socket *sock = file->private_data;
866 
867 	if (unlikely(!sock->ops->splice_read))
868 		return -EINVAL;
869 
870 	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
871 }
872 
873 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
874 {
875 	struct file *file = iocb->ki_filp;
876 	struct socket *sock = file->private_data;
877 	struct msghdr msg = {.msg_iter = *to,
878 			     .msg_iocb = iocb};
879 	ssize_t res;
880 
881 	if (file->f_flags & O_NONBLOCK)
882 		msg.msg_flags = MSG_DONTWAIT;
883 
884 	if (iocb->ki_pos != 0)
885 		return -ESPIPE;
886 
887 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
888 		return 0;
889 
890 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
891 	*to = msg.msg_iter;
892 	return res;
893 }
894 
895 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
896 {
897 	struct file *file = iocb->ki_filp;
898 	struct socket *sock = file->private_data;
899 	struct msghdr msg = {.msg_iter = *from,
900 			     .msg_iocb = iocb};
901 	ssize_t res;
902 
903 	if (iocb->ki_pos != 0)
904 		return -ESPIPE;
905 
906 	if (file->f_flags & O_NONBLOCK)
907 		msg.msg_flags = MSG_DONTWAIT;
908 
909 	if (sock->type == SOCK_SEQPACKET)
910 		msg.msg_flags |= MSG_EOR;
911 
912 	res = sock_sendmsg(sock, &msg);
913 	*from = msg.msg_iter;
914 	return res;
915 }
916 
917 /*
918  * Atomic setting of ioctl hooks to avoid race
919  * with module unload.
920  */
921 
922 static DEFINE_MUTEX(br_ioctl_mutex);
923 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
924 
925 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
926 {
927 	mutex_lock(&br_ioctl_mutex);
928 	br_ioctl_hook = hook;
929 	mutex_unlock(&br_ioctl_mutex);
930 }
931 EXPORT_SYMBOL(brioctl_set);
932 
933 static DEFINE_MUTEX(vlan_ioctl_mutex);
934 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
935 
936 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
937 {
938 	mutex_lock(&vlan_ioctl_mutex);
939 	vlan_ioctl_hook = hook;
940 	mutex_unlock(&vlan_ioctl_mutex);
941 }
942 EXPORT_SYMBOL(vlan_ioctl_set);
943 
944 static DEFINE_MUTEX(dlci_ioctl_mutex);
945 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
946 
947 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
948 {
949 	mutex_lock(&dlci_ioctl_mutex);
950 	dlci_ioctl_hook = hook;
951 	mutex_unlock(&dlci_ioctl_mutex);
952 }
953 EXPORT_SYMBOL(dlci_ioctl_set);
954 
955 static long sock_do_ioctl(struct net *net, struct socket *sock,
956 				 unsigned int cmd, unsigned long arg)
957 {
958 	int err;
959 	void __user *argp = (void __user *)arg;
960 
961 	err = sock->ops->ioctl(sock, cmd, arg);
962 
963 	/*
964 	 * If this ioctl is unknown try to hand it down
965 	 * to the NIC driver.
966 	 */
967 	if (err == -ENOIOCTLCMD)
968 		err = dev_ioctl(net, cmd, argp);
969 
970 	return err;
971 }
972 
973 /*
974  *	With an ioctl, arg may well be a user mode pointer, but we don't know
975  *	what to do with it - that's up to the protocol still.
976  */
977 
978 static struct ns_common *get_net_ns(struct ns_common *ns)
979 {
980 	return &get_net(container_of(ns, struct net, ns))->ns;
981 }
982 
983 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
984 {
985 	struct socket *sock;
986 	struct sock *sk;
987 	void __user *argp = (void __user *)arg;
988 	int pid, err;
989 	struct net *net;
990 
991 	sock = file->private_data;
992 	sk = sock->sk;
993 	net = sock_net(sk);
994 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
995 		err = dev_ioctl(net, cmd, argp);
996 	} else
997 #ifdef CONFIG_WEXT_CORE
998 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
999 		err = dev_ioctl(net, cmd, argp);
1000 	} else
1001 #endif
1002 		switch (cmd) {
1003 		case FIOSETOWN:
1004 		case SIOCSPGRP:
1005 			err = -EFAULT;
1006 			if (get_user(pid, (int __user *)argp))
1007 				break;
1008 			err = f_setown(sock->file, pid, 1);
1009 			break;
1010 		case FIOGETOWN:
1011 		case SIOCGPGRP:
1012 			err = put_user(f_getown(sock->file),
1013 				       (int __user *)argp);
1014 			break;
1015 		case SIOCGIFBR:
1016 		case SIOCSIFBR:
1017 		case SIOCBRADDBR:
1018 		case SIOCBRDELBR:
1019 			err = -ENOPKG;
1020 			if (!br_ioctl_hook)
1021 				request_module("bridge");
1022 
1023 			mutex_lock(&br_ioctl_mutex);
1024 			if (br_ioctl_hook)
1025 				err = br_ioctl_hook(net, cmd, argp);
1026 			mutex_unlock(&br_ioctl_mutex);
1027 			break;
1028 		case SIOCGIFVLAN:
1029 		case SIOCSIFVLAN:
1030 			err = -ENOPKG;
1031 			if (!vlan_ioctl_hook)
1032 				request_module("8021q");
1033 
1034 			mutex_lock(&vlan_ioctl_mutex);
1035 			if (vlan_ioctl_hook)
1036 				err = vlan_ioctl_hook(net, argp);
1037 			mutex_unlock(&vlan_ioctl_mutex);
1038 			break;
1039 		case SIOCADDDLCI:
1040 		case SIOCDELDLCI:
1041 			err = -ENOPKG;
1042 			if (!dlci_ioctl_hook)
1043 				request_module("dlci");
1044 
1045 			mutex_lock(&dlci_ioctl_mutex);
1046 			if (dlci_ioctl_hook)
1047 				err = dlci_ioctl_hook(cmd, argp);
1048 			mutex_unlock(&dlci_ioctl_mutex);
1049 			break;
1050 		case SIOCGSKNS:
1051 			err = -EPERM;
1052 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1053 				break;
1054 
1055 			err = open_related_ns(&net->ns, get_net_ns);
1056 			break;
1057 		default:
1058 			err = sock_do_ioctl(net, sock, cmd, arg);
1059 			break;
1060 		}
1061 	return err;
1062 }
1063 
1064 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1065 {
1066 	int err;
1067 	struct socket *sock = NULL;
1068 
1069 	err = security_socket_create(family, type, protocol, 1);
1070 	if (err)
1071 		goto out;
1072 
1073 	sock = sock_alloc();
1074 	if (!sock) {
1075 		err = -ENOMEM;
1076 		goto out;
1077 	}
1078 
1079 	sock->type = type;
1080 	err = security_socket_post_create(sock, family, type, protocol, 1);
1081 	if (err)
1082 		goto out_release;
1083 
1084 out:
1085 	*res = sock;
1086 	return err;
1087 out_release:
1088 	sock_release(sock);
1089 	sock = NULL;
1090 	goto out;
1091 }
1092 EXPORT_SYMBOL(sock_create_lite);
1093 
1094 /* No kernel lock held - perfect */
1095 static unsigned int sock_poll(struct file *file, poll_table *wait)
1096 {
1097 	unsigned int busy_flag = 0;
1098 	struct socket *sock;
1099 
1100 	/*
1101 	 *      We can't return errors to poll, so it's either yes or no.
1102 	 */
1103 	sock = file->private_data;
1104 
1105 	if (sk_can_busy_loop(sock->sk)) {
1106 		/* this socket can poll_ll so tell the system call */
1107 		busy_flag = POLL_BUSY_LOOP;
1108 
1109 		/* once, only if requested by syscall */
1110 		if (wait && (wait->_key & POLL_BUSY_LOOP))
1111 			sk_busy_loop(sock->sk, 1);
1112 	}
1113 
1114 	return busy_flag | sock->ops->poll(file, sock, wait);
1115 }
1116 
1117 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1118 {
1119 	struct socket *sock = file->private_data;
1120 
1121 	return sock->ops->mmap(file, sock, vma);
1122 }
1123 
1124 static int sock_close(struct inode *inode, struct file *filp)
1125 {
1126 	sock_release(SOCKET_I(inode));
1127 	return 0;
1128 }
1129 
1130 /*
1131  *	Update the socket async list
1132  *
1133  *	Fasync_list locking strategy.
1134  *
1135  *	1. fasync_list is modified only under process context socket lock
1136  *	   i.e. under semaphore.
1137  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1138  *	   or under socket lock
1139  */
1140 
1141 static int sock_fasync(int fd, struct file *filp, int on)
1142 {
1143 	struct socket *sock = filp->private_data;
1144 	struct sock *sk = sock->sk;
1145 	struct socket_wq *wq;
1146 
1147 	if (sk == NULL)
1148 		return -EINVAL;
1149 
1150 	lock_sock(sk);
1151 	wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1152 	fasync_helper(fd, filp, on, &wq->fasync_list);
1153 
1154 	if (!wq->fasync_list)
1155 		sock_reset_flag(sk, SOCK_FASYNC);
1156 	else
1157 		sock_set_flag(sk, SOCK_FASYNC);
1158 
1159 	release_sock(sk);
1160 	return 0;
1161 }
1162 
1163 /* This function may be called only under rcu_lock */
1164 
1165 int sock_wake_async(struct socket_wq *wq, int how, int band)
1166 {
1167 	if (!wq || !wq->fasync_list)
1168 		return -1;
1169 
1170 	switch (how) {
1171 	case SOCK_WAKE_WAITD:
1172 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1173 			break;
1174 		goto call_kill;
1175 	case SOCK_WAKE_SPACE:
1176 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1177 			break;
1178 		/* fall through */
1179 	case SOCK_WAKE_IO:
1180 call_kill:
1181 		kill_fasync(&wq->fasync_list, SIGIO, band);
1182 		break;
1183 	case SOCK_WAKE_URG:
1184 		kill_fasync(&wq->fasync_list, SIGURG, band);
1185 	}
1186 
1187 	return 0;
1188 }
1189 EXPORT_SYMBOL(sock_wake_async);
1190 
1191 int __sock_create(struct net *net, int family, int type, int protocol,
1192 			 struct socket **res, int kern)
1193 {
1194 	int err;
1195 	struct socket *sock;
1196 	const struct net_proto_family *pf;
1197 
1198 	/*
1199 	 *      Check protocol is in range
1200 	 */
1201 	if (family < 0 || family >= NPROTO)
1202 		return -EAFNOSUPPORT;
1203 	if (type < 0 || type >= SOCK_MAX)
1204 		return -EINVAL;
1205 
1206 	/* Compatibility.
1207 
1208 	   This uglymoron is moved from INET layer to here to avoid
1209 	   deadlock in module load.
1210 	 */
1211 	if (family == PF_INET && type == SOCK_PACKET) {
1212 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1213 			     current->comm);
1214 		family = PF_PACKET;
1215 	}
1216 
1217 	err = security_socket_create(family, type, protocol, kern);
1218 	if (err)
1219 		return err;
1220 
1221 	/*
1222 	 *	Allocate the socket and allow the family to set things up. if
1223 	 *	the protocol is 0, the family is instructed to select an appropriate
1224 	 *	default.
1225 	 */
1226 	sock = sock_alloc();
1227 	if (!sock) {
1228 		net_warn_ratelimited("socket: no more sockets\n");
1229 		return -ENFILE;	/* Not exactly a match, but its the
1230 				   closest posix thing */
1231 	}
1232 
1233 	sock->type = type;
1234 
1235 #ifdef CONFIG_MODULES
1236 	/* Attempt to load a protocol module if the find failed.
1237 	 *
1238 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1239 	 * requested real, full-featured networking support upon configuration.
1240 	 * Otherwise module support will break!
1241 	 */
1242 	if (rcu_access_pointer(net_families[family]) == NULL)
1243 		request_module("net-pf-%d", family);
1244 #endif
1245 
1246 	rcu_read_lock();
1247 	pf = rcu_dereference(net_families[family]);
1248 	err = -EAFNOSUPPORT;
1249 	if (!pf)
1250 		goto out_release;
1251 
1252 	/*
1253 	 * We will call the ->create function, that possibly is in a loadable
1254 	 * module, so we have to bump that loadable module refcnt first.
1255 	 */
1256 	if (!try_module_get(pf->owner))
1257 		goto out_release;
1258 
1259 	/* Now protected by module ref count */
1260 	rcu_read_unlock();
1261 
1262 	err = pf->create(net, sock, protocol, kern);
1263 	if (err < 0)
1264 		goto out_module_put;
1265 
1266 	/*
1267 	 * Now to bump the refcnt of the [loadable] module that owns this
1268 	 * socket at sock_release time we decrement its refcnt.
1269 	 */
1270 	if (!try_module_get(sock->ops->owner))
1271 		goto out_module_busy;
1272 
1273 	/*
1274 	 * Now that we're done with the ->create function, the [loadable]
1275 	 * module can have its refcnt decremented
1276 	 */
1277 	module_put(pf->owner);
1278 	err = security_socket_post_create(sock, family, type, protocol, kern);
1279 	if (err)
1280 		goto out_sock_release;
1281 	*res = sock;
1282 
1283 	return 0;
1284 
1285 out_module_busy:
1286 	err = -EAFNOSUPPORT;
1287 out_module_put:
1288 	sock->ops = NULL;
1289 	module_put(pf->owner);
1290 out_sock_release:
1291 	sock_release(sock);
1292 	return err;
1293 
1294 out_release:
1295 	rcu_read_unlock();
1296 	goto out_sock_release;
1297 }
1298 EXPORT_SYMBOL(__sock_create);
1299 
1300 int sock_create(int family, int type, int protocol, struct socket **res)
1301 {
1302 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1303 }
1304 EXPORT_SYMBOL(sock_create);
1305 
1306 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1307 {
1308 	return __sock_create(net, family, type, protocol, res, 1);
1309 }
1310 EXPORT_SYMBOL(sock_create_kern);
1311 
1312 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1313 {
1314 	int retval;
1315 	struct socket *sock;
1316 	int flags;
1317 
1318 	/* Check the SOCK_* constants for consistency.  */
1319 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1320 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1321 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1322 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1323 
1324 	flags = type & ~SOCK_TYPE_MASK;
1325 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1326 		return -EINVAL;
1327 	type &= SOCK_TYPE_MASK;
1328 
1329 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1330 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1331 
1332 	retval = sock_create(family, type, protocol, &sock);
1333 	if (retval < 0)
1334 		goto out;
1335 
1336 	retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1337 	if (retval < 0)
1338 		goto out_release;
1339 
1340 out:
1341 	/* It may be already another descriptor 8) Not kernel problem. */
1342 	return retval;
1343 
1344 out_release:
1345 	sock_release(sock);
1346 	return retval;
1347 }
1348 
1349 /*
1350  *	Create a pair of connected sockets.
1351  */
1352 
1353 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1354 		int __user *, usockvec)
1355 {
1356 	struct socket *sock1, *sock2;
1357 	int fd1, fd2, err;
1358 	struct file *newfile1, *newfile2;
1359 	int flags;
1360 
1361 	flags = type & ~SOCK_TYPE_MASK;
1362 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1363 		return -EINVAL;
1364 	type &= SOCK_TYPE_MASK;
1365 
1366 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1367 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1368 
1369 	/*
1370 	 * Obtain the first socket and check if the underlying protocol
1371 	 * supports the socketpair call.
1372 	 */
1373 
1374 	err = sock_create(family, type, protocol, &sock1);
1375 	if (err < 0)
1376 		goto out;
1377 
1378 	err = sock_create(family, type, protocol, &sock2);
1379 	if (err < 0)
1380 		goto out_release_1;
1381 
1382 	err = sock1->ops->socketpair(sock1, sock2);
1383 	if (err < 0)
1384 		goto out_release_both;
1385 
1386 	fd1 = get_unused_fd_flags(flags);
1387 	if (unlikely(fd1 < 0)) {
1388 		err = fd1;
1389 		goto out_release_both;
1390 	}
1391 
1392 	fd2 = get_unused_fd_flags(flags);
1393 	if (unlikely(fd2 < 0)) {
1394 		err = fd2;
1395 		goto out_put_unused_1;
1396 	}
1397 
1398 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1399 	if (IS_ERR(newfile1)) {
1400 		err = PTR_ERR(newfile1);
1401 		goto out_put_unused_both;
1402 	}
1403 
1404 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1405 	if (IS_ERR(newfile2)) {
1406 		err = PTR_ERR(newfile2);
1407 		goto out_fput_1;
1408 	}
1409 
1410 	err = put_user(fd1, &usockvec[0]);
1411 	if (err)
1412 		goto out_fput_both;
1413 
1414 	err = put_user(fd2, &usockvec[1]);
1415 	if (err)
1416 		goto out_fput_both;
1417 
1418 	audit_fd_pair(fd1, fd2);
1419 
1420 	fd_install(fd1, newfile1);
1421 	fd_install(fd2, newfile2);
1422 	/* fd1 and fd2 may be already another descriptors.
1423 	 * Not kernel problem.
1424 	 */
1425 
1426 	return 0;
1427 
1428 out_fput_both:
1429 	fput(newfile2);
1430 	fput(newfile1);
1431 	put_unused_fd(fd2);
1432 	put_unused_fd(fd1);
1433 	goto out;
1434 
1435 out_fput_1:
1436 	fput(newfile1);
1437 	put_unused_fd(fd2);
1438 	put_unused_fd(fd1);
1439 	sock_release(sock2);
1440 	goto out;
1441 
1442 out_put_unused_both:
1443 	put_unused_fd(fd2);
1444 out_put_unused_1:
1445 	put_unused_fd(fd1);
1446 out_release_both:
1447 	sock_release(sock2);
1448 out_release_1:
1449 	sock_release(sock1);
1450 out:
1451 	return err;
1452 }
1453 
1454 /*
1455  *	Bind a name to a socket. Nothing much to do here since it's
1456  *	the protocol's responsibility to handle the local address.
1457  *
1458  *	We move the socket address to kernel space before we call
1459  *	the protocol layer (having also checked the address is ok).
1460  */
1461 
1462 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1463 {
1464 	struct socket *sock;
1465 	struct sockaddr_storage address;
1466 	int err, fput_needed;
1467 
1468 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1469 	if (sock) {
1470 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1471 		if (err >= 0) {
1472 			err = security_socket_bind(sock,
1473 						   (struct sockaddr *)&address,
1474 						   addrlen);
1475 			if (!err)
1476 				err = sock->ops->bind(sock,
1477 						      (struct sockaddr *)
1478 						      &address, addrlen);
1479 		}
1480 		fput_light(sock->file, fput_needed);
1481 	}
1482 	return err;
1483 }
1484 
1485 /*
1486  *	Perform a listen. Basically, we allow the protocol to do anything
1487  *	necessary for a listen, and if that works, we mark the socket as
1488  *	ready for listening.
1489  */
1490 
1491 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1492 {
1493 	struct socket *sock;
1494 	int err, fput_needed;
1495 	int somaxconn;
1496 
1497 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1498 	if (sock) {
1499 		somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1500 		if ((unsigned int)backlog > somaxconn)
1501 			backlog = somaxconn;
1502 
1503 		err = security_socket_listen(sock, backlog);
1504 		if (!err)
1505 			err = sock->ops->listen(sock, backlog);
1506 
1507 		fput_light(sock->file, fput_needed);
1508 	}
1509 	return err;
1510 }
1511 
1512 /*
1513  *	For accept, we attempt to create a new socket, set up the link
1514  *	with the client, wake up the client, then return the new
1515  *	connected fd. We collect the address of the connector in kernel
1516  *	space and move it to user at the very end. This is unclean because
1517  *	we open the socket then return an error.
1518  *
1519  *	1003.1g adds the ability to recvmsg() to query connection pending
1520  *	status to recvmsg. We need to add that support in a way thats
1521  *	clean when we restucture accept also.
1522  */
1523 
1524 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1525 		int __user *, upeer_addrlen, int, flags)
1526 {
1527 	struct socket *sock, *newsock;
1528 	struct file *newfile;
1529 	int err, len, newfd, fput_needed;
1530 	struct sockaddr_storage address;
1531 
1532 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1533 		return -EINVAL;
1534 
1535 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1536 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1537 
1538 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1539 	if (!sock)
1540 		goto out;
1541 
1542 	err = -ENFILE;
1543 	newsock = sock_alloc();
1544 	if (!newsock)
1545 		goto out_put;
1546 
1547 	newsock->type = sock->type;
1548 	newsock->ops = sock->ops;
1549 
1550 	/*
1551 	 * We don't need try_module_get here, as the listening socket (sock)
1552 	 * has the protocol module (sock->ops->owner) held.
1553 	 */
1554 	__module_get(newsock->ops->owner);
1555 
1556 	newfd = get_unused_fd_flags(flags);
1557 	if (unlikely(newfd < 0)) {
1558 		err = newfd;
1559 		sock_release(newsock);
1560 		goto out_put;
1561 	}
1562 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1563 	if (IS_ERR(newfile)) {
1564 		err = PTR_ERR(newfile);
1565 		put_unused_fd(newfd);
1566 		sock_release(newsock);
1567 		goto out_put;
1568 	}
1569 
1570 	err = security_socket_accept(sock, newsock);
1571 	if (err)
1572 		goto out_fd;
1573 
1574 	err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1575 	if (err < 0)
1576 		goto out_fd;
1577 
1578 	if (upeer_sockaddr) {
1579 		if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1580 					  &len, 2) < 0) {
1581 			err = -ECONNABORTED;
1582 			goto out_fd;
1583 		}
1584 		err = move_addr_to_user(&address,
1585 					len, upeer_sockaddr, upeer_addrlen);
1586 		if (err < 0)
1587 			goto out_fd;
1588 	}
1589 
1590 	/* File flags are not inherited via accept() unlike another OSes. */
1591 
1592 	fd_install(newfd, newfile);
1593 	err = newfd;
1594 
1595 out_put:
1596 	fput_light(sock->file, fput_needed);
1597 out:
1598 	return err;
1599 out_fd:
1600 	fput(newfile);
1601 	put_unused_fd(newfd);
1602 	goto out_put;
1603 }
1604 
1605 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1606 		int __user *, upeer_addrlen)
1607 {
1608 	return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1609 }
1610 
1611 /*
1612  *	Attempt to connect to a socket with the server address.  The address
1613  *	is in user space so we verify it is OK and move it to kernel space.
1614  *
1615  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1616  *	break bindings
1617  *
1618  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1619  *	other SEQPACKET protocols that take time to connect() as it doesn't
1620  *	include the -EINPROGRESS status for such sockets.
1621  */
1622 
1623 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1624 		int, addrlen)
1625 {
1626 	struct socket *sock;
1627 	struct sockaddr_storage address;
1628 	int err, fput_needed;
1629 
1630 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1631 	if (!sock)
1632 		goto out;
1633 	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1634 	if (err < 0)
1635 		goto out_put;
1636 
1637 	err =
1638 	    security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1639 	if (err)
1640 		goto out_put;
1641 
1642 	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1643 				 sock->file->f_flags);
1644 out_put:
1645 	fput_light(sock->file, fput_needed);
1646 out:
1647 	return err;
1648 }
1649 
1650 /*
1651  *	Get the local address ('name') of a socket object. Move the obtained
1652  *	name to user space.
1653  */
1654 
1655 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1656 		int __user *, usockaddr_len)
1657 {
1658 	struct socket *sock;
1659 	struct sockaddr_storage address;
1660 	int len, err, fput_needed;
1661 
1662 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1663 	if (!sock)
1664 		goto out;
1665 
1666 	err = security_socket_getsockname(sock);
1667 	if (err)
1668 		goto out_put;
1669 
1670 	err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1671 	if (err)
1672 		goto out_put;
1673 	err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1674 
1675 out_put:
1676 	fput_light(sock->file, fput_needed);
1677 out:
1678 	return err;
1679 }
1680 
1681 /*
1682  *	Get the remote address ('name') of a socket object. Move the obtained
1683  *	name to user space.
1684  */
1685 
1686 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1687 		int __user *, usockaddr_len)
1688 {
1689 	struct socket *sock;
1690 	struct sockaddr_storage address;
1691 	int len, err, fput_needed;
1692 
1693 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1694 	if (sock != NULL) {
1695 		err = security_socket_getpeername(sock);
1696 		if (err) {
1697 			fput_light(sock->file, fput_needed);
1698 			return err;
1699 		}
1700 
1701 		err =
1702 		    sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1703 				       1);
1704 		if (!err)
1705 			err = move_addr_to_user(&address, len, usockaddr,
1706 						usockaddr_len);
1707 		fput_light(sock->file, fput_needed);
1708 	}
1709 	return err;
1710 }
1711 
1712 /*
1713  *	Send a datagram to a given address. We move the address into kernel
1714  *	space and check the user space data area is readable before invoking
1715  *	the protocol.
1716  */
1717 
1718 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1719 		unsigned int, flags, struct sockaddr __user *, addr,
1720 		int, addr_len)
1721 {
1722 	struct socket *sock;
1723 	struct sockaddr_storage address;
1724 	int err;
1725 	struct msghdr msg;
1726 	struct iovec iov;
1727 	int fput_needed;
1728 
1729 	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1730 	if (unlikely(err))
1731 		return err;
1732 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1733 	if (!sock)
1734 		goto out;
1735 
1736 	msg.msg_name = NULL;
1737 	msg.msg_control = NULL;
1738 	msg.msg_controllen = 0;
1739 	msg.msg_namelen = 0;
1740 	if (addr) {
1741 		err = move_addr_to_kernel(addr, addr_len, &address);
1742 		if (err < 0)
1743 			goto out_put;
1744 		msg.msg_name = (struct sockaddr *)&address;
1745 		msg.msg_namelen = addr_len;
1746 	}
1747 	if (sock->file->f_flags & O_NONBLOCK)
1748 		flags |= MSG_DONTWAIT;
1749 	msg.msg_flags = flags;
1750 	err = sock_sendmsg(sock, &msg);
1751 
1752 out_put:
1753 	fput_light(sock->file, fput_needed);
1754 out:
1755 	return err;
1756 }
1757 
1758 /*
1759  *	Send a datagram down a socket.
1760  */
1761 
1762 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1763 		unsigned int, flags)
1764 {
1765 	return sys_sendto(fd, buff, len, flags, NULL, 0);
1766 }
1767 
1768 /*
1769  *	Receive a frame from the socket and optionally record the address of the
1770  *	sender. We verify the buffers are writable and if needed move the
1771  *	sender address from kernel to user space.
1772  */
1773 
1774 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1775 		unsigned int, flags, struct sockaddr __user *, addr,
1776 		int __user *, addr_len)
1777 {
1778 	struct socket *sock;
1779 	struct iovec iov;
1780 	struct msghdr msg;
1781 	struct sockaddr_storage address;
1782 	int err, err2;
1783 	int fput_needed;
1784 
1785 	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1786 	if (unlikely(err))
1787 		return err;
1788 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1789 	if (!sock)
1790 		goto out;
1791 
1792 	msg.msg_control = NULL;
1793 	msg.msg_controllen = 0;
1794 	/* Save some cycles and don't copy the address if not needed */
1795 	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1796 	/* We assume all kernel code knows the size of sockaddr_storage */
1797 	msg.msg_namelen = 0;
1798 	msg.msg_iocb = NULL;
1799 	msg.msg_flags = 0;
1800 	if (sock->file->f_flags & O_NONBLOCK)
1801 		flags |= MSG_DONTWAIT;
1802 	err = sock_recvmsg(sock, &msg, flags);
1803 
1804 	if (err >= 0 && addr != NULL) {
1805 		err2 = move_addr_to_user(&address,
1806 					 msg.msg_namelen, addr, addr_len);
1807 		if (err2 < 0)
1808 			err = err2;
1809 	}
1810 
1811 	fput_light(sock->file, fput_needed);
1812 out:
1813 	return err;
1814 }
1815 
1816 /*
1817  *	Receive a datagram from a socket.
1818  */
1819 
1820 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1821 		unsigned int, flags)
1822 {
1823 	return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1824 }
1825 
1826 /*
1827  *	Set a socket option. Because we don't know the option lengths we have
1828  *	to pass the user mode parameter for the protocols to sort out.
1829  */
1830 
1831 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1832 		char __user *, optval, int, optlen)
1833 {
1834 	int err, fput_needed;
1835 	struct socket *sock;
1836 
1837 	if (optlen < 0)
1838 		return -EINVAL;
1839 
1840 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1841 	if (sock != NULL) {
1842 		err = security_socket_setsockopt(sock, level, optname);
1843 		if (err)
1844 			goto out_put;
1845 
1846 		if (level == SOL_SOCKET)
1847 			err =
1848 			    sock_setsockopt(sock, level, optname, optval,
1849 					    optlen);
1850 		else
1851 			err =
1852 			    sock->ops->setsockopt(sock, level, optname, optval,
1853 						  optlen);
1854 out_put:
1855 		fput_light(sock->file, fput_needed);
1856 	}
1857 	return err;
1858 }
1859 
1860 /*
1861  *	Get a socket option. Because we don't know the option lengths we have
1862  *	to pass a user mode parameter for the protocols to sort out.
1863  */
1864 
1865 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1866 		char __user *, optval, int __user *, optlen)
1867 {
1868 	int err, fput_needed;
1869 	struct socket *sock;
1870 
1871 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1872 	if (sock != NULL) {
1873 		err = security_socket_getsockopt(sock, level, optname);
1874 		if (err)
1875 			goto out_put;
1876 
1877 		if (level == SOL_SOCKET)
1878 			err =
1879 			    sock_getsockopt(sock, level, optname, optval,
1880 					    optlen);
1881 		else
1882 			err =
1883 			    sock->ops->getsockopt(sock, level, optname, optval,
1884 						  optlen);
1885 out_put:
1886 		fput_light(sock->file, fput_needed);
1887 	}
1888 	return err;
1889 }
1890 
1891 /*
1892  *	Shutdown a socket.
1893  */
1894 
1895 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1896 {
1897 	int err, fput_needed;
1898 	struct socket *sock;
1899 
1900 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1901 	if (sock != NULL) {
1902 		err = security_socket_shutdown(sock, how);
1903 		if (!err)
1904 			err = sock->ops->shutdown(sock, how);
1905 		fput_light(sock->file, fput_needed);
1906 	}
1907 	return err;
1908 }
1909 
1910 /* A couple of helpful macros for getting the address of the 32/64 bit
1911  * fields which are the same type (int / unsigned) on our platforms.
1912  */
1913 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1914 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
1915 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
1916 
1917 struct used_address {
1918 	struct sockaddr_storage name;
1919 	unsigned int name_len;
1920 };
1921 
1922 static int copy_msghdr_from_user(struct msghdr *kmsg,
1923 				 struct user_msghdr __user *umsg,
1924 				 struct sockaddr __user **save_addr,
1925 				 struct iovec **iov)
1926 {
1927 	struct user_msghdr msg;
1928 	ssize_t err;
1929 
1930 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1931 		return -EFAULT;
1932 
1933 	kmsg->msg_control = (void __force *)msg.msg_control;
1934 	kmsg->msg_controllen = msg.msg_controllen;
1935 	kmsg->msg_flags = msg.msg_flags;
1936 
1937 	kmsg->msg_namelen = msg.msg_namelen;
1938 	if (!msg.msg_name)
1939 		kmsg->msg_namelen = 0;
1940 
1941 	if (kmsg->msg_namelen < 0)
1942 		return -EINVAL;
1943 
1944 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1945 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1946 
1947 	if (save_addr)
1948 		*save_addr = msg.msg_name;
1949 
1950 	if (msg.msg_name && kmsg->msg_namelen) {
1951 		if (!save_addr) {
1952 			err = move_addr_to_kernel(msg.msg_name,
1953 						  kmsg->msg_namelen,
1954 						  kmsg->msg_name);
1955 			if (err < 0)
1956 				return err;
1957 		}
1958 	} else {
1959 		kmsg->msg_name = NULL;
1960 		kmsg->msg_namelen = 0;
1961 	}
1962 
1963 	if (msg.msg_iovlen > UIO_MAXIOV)
1964 		return -EMSGSIZE;
1965 
1966 	kmsg->msg_iocb = NULL;
1967 
1968 	return import_iovec(save_addr ? READ : WRITE,
1969 			    msg.msg_iov, msg.msg_iovlen,
1970 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
1971 }
1972 
1973 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1974 			 struct msghdr *msg_sys, unsigned int flags,
1975 			 struct used_address *used_address,
1976 			 unsigned int allowed_msghdr_flags)
1977 {
1978 	struct compat_msghdr __user *msg_compat =
1979 	    (struct compat_msghdr __user *)msg;
1980 	struct sockaddr_storage address;
1981 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1982 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
1983 				__aligned(sizeof(__kernel_size_t));
1984 	/* 20 is size of ipv6_pktinfo */
1985 	unsigned char *ctl_buf = ctl;
1986 	int ctl_len;
1987 	ssize_t err;
1988 
1989 	msg_sys->msg_name = &address;
1990 
1991 	if (MSG_CMSG_COMPAT & flags)
1992 		err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1993 	else
1994 		err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1995 	if (err < 0)
1996 		return err;
1997 
1998 	err = -ENOBUFS;
1999 
2000 	if (msg_sys->msg_controllen > INT_MAX)
2001 		goto out_freeiov;
2002 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2003 	ctl_len = msg_sys->msg_controllen;
2004 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2005 		err =
2006 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2007 						     sizeof(ctl));
2008 		if (err)
2009 			goto out_freeiov;
2010 		ctl_buf = msg_sys->msg_control;
2011 		ctl_len = msg_sys->msg_controllen;
2012 	} else if (ctl_len) {
2013 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2014 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2015 		if (ctl_len > sizeof(ctl)) {
2016 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2017 			if (ctl_buf == NULL)
2018 				goto out_freeiov;
2019 		}
2020 		err = -EFAULT;
2021 		/*
2022 		 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2023 		 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2024 		 * checking falls down on this.
2025 		 */
2026 		if (copy_from_user(ctl_buf,
2027 				   (void __user __force *)msg_sys->msg_control,
2028 				   ctl_len))
2029 			goto out_freectl;
2030 		msg_sys->msg_control = ctl_buf;
2031 	}
2032 	msg_sys->msg_flags = flags;
2033 
2034 	if (sock->file->f_flags & O_NONBLOCK)
2035 		msg_sys->msg_flags |= MSG_DONTWAIT;
2036 	/*
2037 	 * If this is sendmmsg() and current destination address is same as
2038 	 * previously succeeded address, omit asking LSM's decision.
2039 	 * used_address->name_len is initialized to UINT_MAX so that the first
2040 	 * destination address never matches.
2041 	 */
2042 	if (used_address && msg_sys->msg_name &&
2043 	    used_address->name_len == msg_sys->msg_namelen &&
2044 	    !memcmp(&used_address->name, msg_sys->msg_name,
2045 		    used_address->name_len)) {
2046 		err = sock_sendmsg_nosec(sock, msg_sys);
2047 		goto out_freectl;
2048 	}
2049 	err = sock_sendmsg(sock, msg_sys);
2050 	/*
2051 	 * If this is sendmmsg() and sending to current destination address was
2052 	 * successful, remember it.
2053 	 */
2054 	if (used_address && err >= 0) {
2055 		used_address->name_len = msg_sys->msg_namelen;
2056 		if (msg_sys->msg_name)
2057 			memcpy(&used_address->name, msg_sys->msg_name,
2058 			       used_address->name_len);
2059 	}
2060 
2061 out_freectl:
2062 	if (ctl_buf != ctl)
2063 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2064 out_freeiov:
2065 	kfree(iov);
2066 	return err;
2067 }
2068 
2069 /*
2070  *	BSD sendmsg interface
2071  */
2072 
2073 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2074 {
2075 	int fput_needed, err;
2076 	struct msghdr msg_sys;
2077 	struct socket *sock;
2078 
2079 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2080 	if (!sock)
2081 		goto out;
2082 
2083 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2084 
2085 	fput_light(sock->file, fput_needed);
2086 out:
2087 	return err;
2088 }
2089 
2090 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2091 {
2092 	if (flags & MSG_CMSG_COMPAT)
2093 		return -EINVAL;
2094 	return __sys_sendmsg(fd, msg, flags);
2095 }
2096 
2097 /*
2098  *	Linux sendmmsg interface
2099  */
2100 
2101 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2102 		   unsigned int flags)
2103 {
2104 	int fput_needed, err, datagrams;
2105 	struct socket *sock;
2106 	struct mmsghdr __user *entry;
2107 	struct compat_mmsghdr __user *compat_entry;
2108 	struct msghdr msg_sys;
2109 	struct used_address used_address;
2110 	unsigned int oflags = flags;
2111 
2112 	if (vlen > UIO_MAXIOV)
2113 		vlen = UIO_MAXIOV;
2114 
2115 	datagrams = 0;
2116 
2117 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2118 	if (!sock)
2119 		return err;
2120 
2121 	used_address.name_len = UINT_MAX;
2122 	entry = mmsg;
2123 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2124 	err = 0;
2125 	flags |= MSG_BATCH;
2126 
2127 	while (datagrams < vlen) {
2128 		if (datagrams == vlen - 1)
2129 			flags = oflags;
2130 
2131 		if (MSG_CMSG_COMPAT & flags) {
2132 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2133 					     &msg_sys, flags, &used_address, MSG_EOR);
2134 			if (err < 0)
2135 				break;
2136 			err = __put_user(err, &compat_entry->msg_len);
2137 			++compat_entry;
2138 		} else {
2139 			err = ___sys_sendmsg(sock,
2140 					     (struct user_msghdr __user *)entry,
2141 					     &msg_sys, flags, &used_address, MSG_EOR);
2142 			if (err < 0)
2143 				break;
2144 			err = put_user(err, &entry->msg_len);
2145 			++entry;
2146 		}
2147 
2148 		if (err)
2149 			break;
2150 		++datagrams;
2151 		if (msg_data_left(&msg_sys))
2152 			break;
2153 		cond_resched();
2154 	}
2155 
2156 	fput_light(sock->file, fput_needed);
2157 
2158 	/* We only return an error if no datagrams were able to be sent */
2159 	if (datagrams != 0)
2160 		return datagrams;
2161 
2162 	return err;
2163 }
2164 
2165 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2166 		unsigned int, vlen, unsigned int, flags)
2167 {
2168 	if (flags & MSG_CMSG_COMPAT)
2169 		return -EINVAL;
2170 	return __sys_sendmmsg(fd, mmsg, vlen, flags);
2171 }
2172 
2173 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2174 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2175 {
2176 	struct compat_msghdr __user *msg_compat =
2177 	    (struct compat_msghdr __user *)msg;
2178 	struct iovec iovstack[UIO_FASTIOV];
2179 	struct iovec *iov = iovstack;
2180 	unsigned long cmsg_ptr;
2181 	int len;
2182 	ssize_t err;
2183 
2184 	/* kernel mode address */
2185 	struct sockaddr_storage addr;
2186 
2187 	/* user mode address pointers */
2188 	struct sockaddr __user *uaddr;
2189 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2190 
2191 	msg_sys->msg_name = &addr;
2192 
2193 	if (MSG_CMSG_COMPAT & flags)
2194 		err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2195 	else
2196 		err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2197 	if (err < 0)
2198 		return err;
2199 
2200 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2201 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2202 
2203 	/* We assume all kernel code knows the size of sockaddr_storage */
2204 	msg_sys->msg_namelen = 0;
2205 
2206 	if (sock->file->f_flags & O_NONBLOCK)
2207 		flags |= MSG_DONTWAIT;
2208 	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2209 	if (err < 0)
2210 		goto out_freeiov;
2211 	len = err;
2212 
2213 	if (uaddr != NULL) {
2214 		err = move_addr_to_user(&addr,
2215 					msg_sys->msg_namelen, uaddr,
2216 					uaddr_len);
2217 		if (err < 0)
2218 			goto out_freeiov;
2219 	}
2220 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2221 			 COMPAT_FLAGS(msg));
2222 	if (err)
2223 		goto out_freeiov;
2224 	if (MSG_CMSG_COMPAT & flags)
2225 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2226 				 &msg_compat->msg_controllen);
2227 	else
2228 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2229 				 &msg->msg_controllen);
2230 	if (err)
2231 		goto out_freeiov;
2232 	err = len;
2233 
2234 out_freeiov:
2235 	kfree(iov);
2236 	return err;
2237 }
2238 
2239 /*
2240  *	BSD recvmsg interface
2241  */
2242 
2243 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2244 {
2245 	int fput_needed, err;
2246 	struct msghdr msg_sys;
2247 	struct socket *sock;
2248 
2249 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2250 	if (!sock)
2251 		goto out;
2252 
2253 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2254 
2255 	fput_light(sock->file, fput_needed);
2256 out:
2257 	return err;
2258 }
2259 
2260 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2261 		unsigned int, flags)
2262 {
2263 	if (flags & MSG_CMSG_COMPAT)
2264 		return -EINVAL;
2265 	return __sys_recvmsg(fd, msg, flags);
2266 }
2267 
2268 /*
2269  *     Linux recvmmsg interface
2270  */
2271 
2272 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2273 		   unsigned int flags, struct timespec *timeout)
2274 {
2275 	int fput_needed, err, datagrams;
2276 	struct socket *sock;
2277 	struct mmsghdr __user *entry;
2278 	struct compat_mmsghdr __user *compat_entry;
2279 	struct msghdr msg_sys;
2280 	struct timespec64 end_time;
2281 	struct timespec64 timeout64;
2282 
2283 	if (timeout &&
2284 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2285 				    timeout->tv_nsec))
2286 		return -EINVAL;
2287 
2288 	datagrams = 0;
2289 
2290 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2291 	if (!sock)
2292 		return err;
2293 
2294 	err = sock_error(sock->sk);
2295 	if (err) {
2296 		datagrams = err;
2297 		goto out_put;
2298 	}
2299 
2300 	entry = mmsg;
2301 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2302 
2303 	while (datagrams < vlen) {
2304 		/*
2305 		 * No need to ask LSM for more than the first datagram.
2306 		 */
2307 		if (MSG_CMSG_COMPAT & flags) {
2308 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2309 					     &msg_sys, flags & ~MSG_WAITFORONE,
2310 					     datagrams);
2311 			if (err < 0)
2312 				break;
2313 			err = __put_user(err, &compat_entry->msg_len);
2314 			++compat_entry;
2315 		} else {
2316 			err = ___sys_recvmsg(sock,
2317 					     (struct user_msghdr __user *)entry,
2318 					     &msg_sys, flags & ~MSG_WAITFORONE,
2319 					     datagrams);
2320 			if (err < 0)
2321 				break;
2322 			err = put_user(err, &entry->msg_len);
2323 			++entry;
2324 		}
2325 
2326 		if (err)
2327 			break;
2328 		++datagrams;
2329 
2330 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2331 		if (flags & MSG_WAITFORONE)
2332 			flags |= MSG_DONTWAIT;
2333 
2334 		if (timeout) {
2335 			ktime_get_ts64(&timeout64);
2336 			*timeout = timespec64_to_timespec(
2337 					timespec64_sub(end_time, timeout64));
2338 			if (timeout->tv_sec < 0) {
2339 				timeout->tv_sec = timeout->tv_nsec = 0;
2340 				break;
2341 			}
2342 
2343 			/* Timeout, return less than vlen datagrams */
2344 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2345 				break;
2346 		}
2347 
2348 		/* Out of band data, return right away */
2349 		if (msg_sys.msg_flags & MSG_OOB)
2350 			break;
2351 		cond_resched();
2352 	}
2353 
2354 	if (err == 0)
2355 		goto out_put;
2356 
2357 	if (datagrams == 0) {
2358 		datagrams = err;
2359 		goto out_put;
2360 	}
2361 
2362 	/*
2363 	 * We may return less entries than requested (vlen) if the
2364 	 * sock is non block and there aren't enough datagrams...
2365 	 */
2366 	if (err != -EAGAIN) {
2367 		/*
2368 		 * ... or  if recvmsg returns an error after we
2369 		 * received some datagrams, where we record the
2370 		 * error to return on the next call or if the
2371 		 * app asks about it using getsockopt(SO_ERROR).
2372 		 */
2373 		sock->sk->sk_err = -err;
2374 	}
2375 out_put:
2376 	fput_light(sock->file, fput_needed);
2377 
2378 	return datagrams;
2379 }
2380 
2381 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2382 		unsigned int, vlen, unsigned int, flags,
2383 		struct timespec __user *, timeout)
2384 {
2385 	int datagrams;
2386 	struct timespec timeout_sys;
2387 
2388 	if (flags & MSG_CMSG_COMPAT)
2389 		return -EINVAL;
2390 
2391 	if (!timeout)
2392 		return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2393 
2394 	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2395 		return -EFAULT;
2396 
2397 	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2398 
2399 	if (datagrams > 0 &&
2400 	    copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2401 		datagrams = -EFAULT;
2402 
2403 	return datagrams;
2404 }
2405 
2406 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2407 /* Argument list sizes for sys_socketcall */
2408 #define AL(x) ((x) * sizeof(unsigned long))
2409 static const unsigned char nargs[21] = {
2410 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2411 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2412 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2413 	AL(4), AL(5), AL(4)
2414 };
2415 
2416 #undef AL
2417 
2418 /*
2419  *	System call vectors.
2420  *
2421  *	Argument checking cleaned up. Saved 20% in size.
2422  *  This function doesn't need to set the kernel lock because
2423  *  it is set by the callees.
2424  */
2425 
2426 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2427 {
2428 	unsigned long a[AUDITSC_ARGS];
2429 	unsigned long a0, a1;
2430 	int err;
2431 	unsigned int len;
2432 
2433 	if (call < 1 || call > SYS_SENDMMSG)
2434 		return -EINVAL;
2435 
2436 	len = nargs[call];
2437 	if (len > sizeof(a))
2438 		return -EINVAL;
2439 
2440 	/* copy_from_user should be SMP safe. */
2441 	if (copy_from_user(a, args, len))
2442 		return -EFAULT;
2443 
2444 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2445 	if (err)
2446 		return err;
2447 
2448 	a0 = a[0];
2449 	a1 = a[1];
2450 
2451 	switch (call) {
2452 	case SYS_SOCKET:
2453 		err = sys_socket(a0, a1, a[2]);
2454 		break;
2455 	case SYS_BIND:
2456 		err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2457 		break;
2458 	case SYS_CONNECT:
2459 		err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2460 		break;
2461 	case SYS_LISTEN:
2462 		err = sys_listen(a0, a1);
2463 		break;
2464 	case SYS_ACCEPT:
2465 		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2466 				  (int __user *)a[2], 0);
2467 		break;
2468 	case SYS_GETSOCKNAME:
2469 		err =
2470 		    sys_getsockname(a0, (struct sockaddr __user *)a1,
2471 				    (int __user *)a[2]);
2472 		break;
2473 	case SYS_GETPEERNAME:
2474 		err =
2475 		    sys_getpeername(a0, (struct sockaddr __user *)a1,
2476 				    (int __user *)a[2]);
2477 		break;
2478 	case SYS_SOCKETPAIR:
2479 		err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2480 		break;
2481 	case SYS_SEND:
2482 		err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2483 		break;
2484 	case SYS_SENDTO:
2485 		err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2486 				 (struct sockaddr __user *)a[4], a[5]);
2487 		break;
2488 	case SYS_RECV:
2489 		err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2490 		break;
2491 	case SYS_RECVFROM:
2492 		err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2493 				   (struct sockaddr __user *)a[4],
2494 				   (int __user *)a[5]);
2495 		break;
2496 	case SYS_SHUTDOWN:
2497 		err = sys_shutdown(a0, a1);
2498 		break;
2499 	case SYS_SETSOCKOPT:
2500 		err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2501 		break;
2502 	case SYS_GETSOCKOPT:
2503 		err =
2504 		    sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2505 				   (int __user *)a[4]);
2506 		break;
2507 	case SYS_SENDMSG:
2508 		err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2509 		break;
2510 	case SYS_SENDMMSG:
2511 		err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2512 		break;
2513 	case SYS_RECVMSG:
2514 		err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2515 		break;
2516 	case SYS_RECVMMSG:
2517 		err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2518 				   (struct timespec __user *)a[4]);
2519 		break;
2520 	case SYS_ACCEPT4:
2521 		err = sys_accept4(a0, (struct sockaddr __user *)a1,
2522 				  (int __user *)a[2], a[3]);
2523 		break;
2524 	default:
2525 		err = -EINVAL;
2526 		break;
2527 	}
2528 	return err;
2529 }
2530 
2531 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
2532 
2533 /**
2534  *	sock_register - add a socket protocol handler
2535  *	@ops: description of protocol
2536  *
2537  *	This function is called by a protocol handler that wants to
2538  *	advertise its address family, and have it linked into the
2539  *	socket interface. The value ops->family corresponds to the
2540  *	socket system call protocol family.
2541  */
2542 int sock_register(const struct net_proto_family *ops)
2543 {
2544 	int err;
2545 
2546 	if (ops->family >= NPROTO) {
2547 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2548 		return -ENOBUFS;
2549 	}
2550 
2551 	spin_lock(&net_family_lock);
2552 	if (rcu_dereference_protected(net_families[ops->family],
2553 				      lockdep_is_held(&net_family_lock)))
2554 		err = -EEXIST;
2555 	else {
2556 		rcu_assign_pointer(net_families[ops->family], ops);
2557 		err = 0;
2558 	}
2559 	spin_unlock(&net_family_lock);
2560 
2561 	pr_info("NET: Registered protocol family %d\n", ops->family);
2562 	return err;
2563 }
2564 EXPORT_SYMBOL(sock_register);
2565 
2566 /**
2567  *	sock_unregister - remove a protocol handler
2568  *	@family: protocol family to remove
2569  *
2570  *	This function is called by a protocol handler that wants to
2571  *	remove its address family, and have it unlinked from the
2572  *	new socket creation.
2573  *
2574  *	If protocol handler is a module, then it can use module reference
2575  *	counts to protect against new references. If protocol handler is not
2576  *	a module then it needs to provide its own protection in
2577  *	the ops->create routine.
2578  */
2579 void sock_unregister(int family)
2580 {
2581 	BUG_ON(family < 0 || family >= NPROTO);
2582 
2583 	spin_lock(&net_family_lock);
2584 	RCU_INIT_POINTER(net_families[family], NULL);
2585 	spin_unlock(&net_family_lock);
2586 
2587 	synchronize_rcu();
2588 
2589 	pr_info("NET: Unregistered protocol family %d\n", family);
2590 }
2591 EXPORT_SYMBOL(sock_unregister);
2592 
2593 static int __init sock_init(void)
2594 {
2595 	int err;
2596 	/*
2597 	 *      Initialize the network sysctl infrastructure.
2598 	 */
2599 	err = net_sysctl_init();
2600 	if (err)
2601 		goto out;
2602 
2603 	/*
2604 	 *      Initialize skbuff SLAB cache
2605 	 */
2606 	skb_init();
2607 
2608 	/*
2609 	 *      Initialize the protocols module.
2610 	 */
2611 
2612 	init_inodecache();
2613 
2614 	err = register_filesystem(&sock_fs_type);
2615 	if (err)
2616 		goto out_fs;
2617 	sock_mnt = kern_mount(&sock_fs_type);
2618 	if (IS_ERR(sock_mnt)) {
2619 		err = PTR_ERR(sock_mnt);
2620 		goto out_mount;
2621 	}
2622 
2623 	/* The real protocol initialization is performed in later initcalls.
2624 	 */
2625 
2626 #ifdef CONFIG_NETFILTER
2627 	err = netfilter_init();
2628 	if (err)
2629 		goto out;
2630 #endif
2631 
2632 	ptp_classifier_init();
2633 
2634 out:
2635 	return err;
2636 
2637 out_mount:
2638 	unregister_filesystem(&sock_fs_type);
2639 out_fs:
2640 	goto out;
2641 }
2642 
2643 core_initcall(sock_init);	/* early initcall */
2644 
2645 #ifdef CONFIG_PROC_FS
2646 void socket_seq_show(struct seq_file *seq)
2647 {
2648 	int cpu;
2649 	int counter = 0;
2650 
2651 	for_each_possible_cpu(cpu)
2652 	    counter += per_cpu(sockets_in_use, cpu);
2653 
2654 	/* It can be negative, by the way. 8) */
2655 	if (counter < 0)
2656 		counter = 0;
2657 
2658 	seq_printf(seq, "sockets: used %d\n", counter);
2659 }
2660 #endif				/* CONFIG_PROC_FS */
2661 
2662 #ifdef CONFIG_COMPAT
2663 static int do_siocgstamp(struct net *net, struct socket *sock,
2664 			 unsigned int cmd, void __user *up)
2665 {
2666 	mm_segment_t old_fs = get_fs();
2667 	struct timeval ktv;
2668 	int err;
2669 
2670 	set_fs(KERNEL_DS);
2671 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2672 	set_fs(old_fs);
2673 	if (!err)
2674 		err = compat_put_timeval(&ktv, up);
2675 
2676 	return err;
2677 }
2678 
2679 static int do_siocgstampns(struct net *net, struct socket *sock,
2680 			   unsigned int cmd, void __user *up)
2681 {
2682 	mm_segment_t old_fs = get_fs();
2683 	struct timespec kts;
2684 	int err;
2685 
2686 	set_fs(KERNEL_DS);
2687 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2688 	set_fs(old_fs);
2689 	if (!err)
2690 		err = compat_put_timespec(&kts, up);
2691 
2692 	return err;
2693 }
2694 
2695 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2696 {
2697 	struct ifreq __user *uifr;
2698 	int err;
2699 
2700 	uifr = compat_alloc_user_space(sizeof(struct ifreq));
2701 	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2702 		return -EFAULT;
2703 
2704 	err = dev_ioctl(net, SIOCGIFNAME, uifr);
2705 	if (err)
2706 		return err;
2707 
2708 	if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2709 		return -EFAULT;
2710 
2711 	return 0;
2712 }
2713 
2714 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2715 {
2716 	struct compat_ifconf ifc32;
2717 	struct ifconf ifc;
2718 	struct ifconf __user *uifc;
2719 	struct compat_ifreq __user *ifr32;
2720 	struct ifreq __user *ifr;
2721 	unsigned int i, j;
2722 	int err;
2723 
2724 	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2725 		return -EFAULT;
2726 
2727 	memset(&ifc, 0, sizeof(ifc));
2728 	if (ifc32.ifcbuf == 0) {
2729 		ifc32.ifc_len = 0;
2730 		ifc.ifc_len = 0;
2731 		ifc.ifc_req = NULL;
2732 		uifc = compat_alloc_user_space(sizeof(struct ifconf));
2733 	} else {
2734 		size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2735 			sizeof(struct ifreq);
2736 		uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2737 		ifc.ifc_len = len;
2738 		ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2739 		ifr32 = compat_ptr(ifc32.ifcbuf);
2740 		for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2741 			if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2742 				return -EFAULT;
2743 			ifr++;
2744 			ifr32++;
2745 		}
2746 	}
2747 	if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2748 		return -EFAULT;
2749 
2750 	err = dev_ioctl(net, SIOCGIFCONF, uifc);
2751 	if (err)
2752 		return err;
2753 
2754 	if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2755 		return -EFAULT;
2756 
2757 	ifr = ifc.ifc_req;
2758 	ifr32 = compat_ptr(ifc32.ifcbuf);
2759 	for (i = 0, j = 0;
2760 	     i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2761 	     i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2762 		if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2763 			return -EFAULT;
2764 		ifr32++;
2765 		ifr++;
2766 	}
2767 
2768 	if (ifc32.ifcbuf == 0) {
2769 		/* Translate from 64-bit structure multiple to
2770 		 * a 32-bit one.
2771 		 */
2772 		i = ifc.ifc_len;
2773 		i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2774 		ifc32.ifc_len = i;
2775 	} else {
2776 		ifc32.ifc_len = i;
2777 	}
2778 	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2779 		return -EFAULT;
2780 
2781 	return 0;
2782 }
2783 
2784 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2785 {
2786 	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2787 	bool convert_in = false, convert_out = false;
2788 	size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2789 	struct ethtool_rxnfc __user *rxnfc;
2790 	struct ifreq __user *ifr;
2791 	u32 rule_cnt = 0, actual_rule_cnt;
2792 	u32 ethcmd;
2793 	u32 data;
2794 	int ret;
2795 
2796 	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2797 		return -EFAULT;
2798 
2799 	compat_rxnfc = compat_ptr(data);
2800 
2801 	if (get_user(ethcmd, &compat_rxnfc->cmd))
2802 		return -EFAULT;
2803 
2804 	/* Most ethtool structures are defined without padding.
2805 	 * Unfortunately struct ethtool_rxnfc is an exception.
2806 	 */
2807 	switch (ethcmd) {
2808 	default:
2809 		break;
2810 	case ETHTOOL_GRXCLSRLALL:
2811 		/* Buffer size is variable */
2812 		if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2813 			return -EFAULT;
2814 		if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2815 			return -ENOMEM;
2816 		buf_size += rule_cnt * sizeof(u32);
2817 		/* fall through */
2818 	case ETHTOOL_GRXRINGS:
2819 	case ETHTOOL_GRXCLSRLCNT:
2820 	case ETHTOOL_GRXCLSRULE:
2821 	case ETHTOOL_SRXCLSRLINS:
2822 		convert_out = true;
2823 		/* fall through */
2824 	case ETHTOOL_SRXCLSRLDEL:
2825 		buf_size += sizeof(struct ethtool_rxnfc);
2826 		convert_in = true;
2827 		break;
2828 	}
2829 
2830 	ifr = compat_alloc_user_space(buf_size);
2831 	rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2832 
2833 	if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2834 		return -EFAULT;
2835 
2836 	if (put_user(convert_in ? rxnfc : compat_ptr(data),
2837 		     &ifr->ifr_ifru.ifru_data))
2838 		return -EFAULT;
2839 
2840 	if (convert_in) {
2841 		/* We expect there to be holes between fs.m_ext and
2842 		 * fs.ring_cookie and at the end of fs, but nowhere else.
2843 		 */
2844 		BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2845 			     sizeof(compat_rxnfc->fs.m_ext) !=
2846 			     offsetof(struct ethtool_rxnfc, fs.m_ext) +
2847 			     sizeof(rxnfc->fs.m_ext));
2848 		BUILD_BUG_ON(
2849 			offsetof(struct compat_ethtool_rxnfc, fs.location) -
2850 			offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2851 			offsetof(struct ethtool_rxnfc, fs.location) -
2852 			offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2853 
2854 		if (copy_in_user(rxnfc, compat_rxnfc,
2855 				 (void __user *)(&rxnfc->fs.m_ext + 1) -
2856 				 (void __user *)rxnfc) ||
2857 		    copy_in_user(&rxnfc->fs.ring_cookie,
2858 				 &compat_rxnfc->fs.ring_cookie,
2859 				 (void __user *)(&rxnfc->fs.location + 1) -
2860 				 (void __user *)&rxnfc->fs.ring_cookie) ||
2861 		    copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2862 				 sizeof(rxnfc->rule_cnt)))
2863 			return -EFAULT;
2864 	}
2865 
2866 	ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2867 	if (ret)
2868 		return ret;
2869 
2870 	if (convert_out) {
2871 		if (copy_in_user(compat_rxnfc, rxnfc,
2872 				 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2873 				 (const void __user *)rxnfc) ||
2874 		    copy_in_user(&compat_rxnfc->fs.ring_cookie,
2875 				 &rxnfc->fs.ring_cookie,
2876 				 (const void __user *)(&rxnfc->fs.location + 1) -
2877 				 (const void __user *)&rxnfc->fs.ring_cookie) ||
2878 		    copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2879 				 sizeof(rxnfc->rule_cnt)))
2880 			return -EFAULT;
2881 
2882 		if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2883 			/* As an optimisation, we only copy the actual
2884 			 * number of rules that the underlying
2885 			 * function returned.  Since Mallory might
2886 			 * change the rule count in user memory, we
2887 			 * check that it is less than the rule count
2888 			 * originally given (as the user buffer size),
2889 			 * which has been range-checked.
2890 			 */
2891 			if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2892 				return -EFAULT;
2893 			if (actual_rule_cnt < rule_cnt)
2894 				rule_cnt = actual_rule_cnt;
2895 			if (copy_in_user(&compat_rxnfc->rule_locs[0],
2896 					 &rxnfc->rule_locs[0],
2897 					 rule_cnt * sizeof(u32)))
2898 				return -EFAULT;
2899 		}
2900 	}
2901 
2902 	return 0;
2903 }
2904 
2905 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2906 {
2907 	void __user *uptr;
2908 	compat_uptr_t uptr32;
2909 	struct ifreq __user *uifr;
2910 
2911 	uifr = compat_alloc_user_space(sizeof(*uifr));
2912 	if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2913 		return -EFAULT;
2914 
2915 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2916 		return -EFAULT;
2917 
2918 	uptr = compat_ptr(uptr32);
2919 
2920 	if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2921 		return -EFAULT;
2922 
2923 	return dev_ioctl(net, SIOCWANDEV, uifr);
2924 }
2925 
2926 static int bond_ioctl(struct net *net, unsigned int cmd,
2927 			 struct compat_ifreq __user *ifr32)
2928 {
2929 	struct ifreq kifr;
2930 	mm_segment_t old_fs;
2931 	int err;
2932 
2933 	switch (cmd) {
2934 	case SIOCBONDENSLAVE:
2935 	case SIOCBONDRELEASE:
2936 	case SIOCBONDSETHWADDR:
2937 	case SIOCBONDCHANGEACTIVE:
2938 		if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2939 			return -EFAULT;
2940 
2941 		old_fs = get_fs();
2942 		set_fs(KERNEL_DS);
2943 		err = dev_ioctl(net, cmd,
2944 				(struct ifreq __user __force *) &kifr);
2945 		set_fs(old_fs);
2946 
2947 		return err;
2948 	default:
2949 		return -ENOIOCTLCMD;
2950 	}
2951 }
2952 
2953 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2954 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2955 				 struct compat_ifreq __user *u_ifreq32)
2956 {
2957 	struct ifreq __user *u_ifreq64;
2958 	char tmp_buf[IFNAMSIZ];
2959 	void __user *data64;
2960 	u32 data32;
2961 
2962 	if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2963 			   IFNAMSIZ))
2964 		return -EFAULT;
2965 	if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2966 		return -EFAULT;
2967 	data64 = compat_ptr(data32);
2968 
2969 	u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2970 
2971 	if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2972 			 IFNAMSIZ))
2973 		return -EFAULT;
2974 	if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2975 		return -EFAULT;
2976 
2977 	return dev_ioctl(net, cmd, u_ifreq64);
2978 }
2979 
2980 static int dev_ifsioc(struct net *net, struct socket *sock,
2981 			 unsigned int cmd, struct compat_ifreq __user *uifr32)
2982 {
2983 	struct ifreq __user *uifr;
2984 	int err;
2985 
2986 	uifr = compat_alloc_user_space(sizeof(*uifr));
2987 	if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2988 		return -EFAULT;
2989 
2990 	err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2991 
2992 	if (!err) {
2993 		switch (cmd) {
2994 		case SIOCGIFFLAGS:
2995 		case SIOCGIFMETRIC:
2996 		case SIOCGIFMTU:
2997 		case SIOCGIFMEM:
2998 		case SIOCGIFHWADDR:
2999 		case SIOCGIFINDEX:
3000 		case SIOCGIFADDR:
3001 		case SIOCGIFBRDADDR:
3002 		case SIOCGIFDSTADDR:
3003 		case SIOCGIFNETMASK:
3004 		case SIOCGIFPFLAGS:
3005 		case SIOCGIFTXQLEN:
3006 		case SIOCGMIIPHY:
3007 		case SIOCGMIIREG:
3008 			if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3009 				err = -EFAULT;
3010 			break;
3011 		}
3012 	}
3013 	return err;
3014 }
3015 
3016 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3017 			struct compat_ifreq __user *uifr32)
3018 {
3019 	struct ifreq ifr;
3020 	struct compat_ifmap __user *uifmap32;
3021 	mm_segment_t old_fs;
3022 	int err;
3023 
3024 	uifmap32 = &uifr32->ifr_ifru.ifru_map;
3025 	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3026 	err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3027 	err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3028 	err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3029 	err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3030 	err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3031 	err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3032 	if (err)
3033 		return -EFAULT;
3034 
3035 	old_fs = get_fs();
3036 	set_fs(KERNEL_DS);
3037 	err = dev_ioctl(net, cmd, (void  __user __force *)&ifr);
3038 	set_fs(old_fs);
3039 
3040 	if (cmd == SIOCGIFMAP && !err) {
3041 		err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3042 		err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3043 		err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3044 		err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3045 		err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3046 		err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3047 		err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3048 		if (err)
3049 			err = -EFAULT;
3050 	}
3051 	return err;
3052 }
3053 
3054 struct rtentry32 {
3055 	u32		rt_pad1;
3056 	struct sockaddr rt_dst;         /* target address               */
3057 	struct sockaddr rt_gateway;     /* gateway addr (RTF_GATEWAY)   */
3058 	struct sockaddr rt_genmask;     /* target network mask (IP)     */
3059 	unsigned short	rt_flags;
3060 	short		rt_pad2;
3061 	u32		rt_pad3;
3062 	unsigned char	rt_tos;
3063 	unsigned char	rt_class;
3064 	short		rt_pad4;
3065 	short		rt_metric;      /* +1 for binary compatibility! */
3066 	/* char * */ u32 rt_dev;        /* forcing the device at add    */
3067 	u32		rt_mtu;         /* per route MTU/Window         */
3068 	u32		rt_window;      /* Window clamping              */
3069 	unsigned short  rt_irtt;        /* Initial RTT                  */
3070 };
3071 
3072 struct in6_rtmsg32 {
3073 	struct in6_addr		rtmsg_dst;
3074 	struct in6_addr		rtmsg_src;
3075 	struct in6_addr		rtmsg_gateway;
3076 	u32			rtmsg_type;
3077 	u16			rtmsg_dst_len;
3078 	u16			rtmsg_src_len;
3079 	u32			rtmsg_metric;
3080 	u32			rtmsg_info;
3081 	u32			rtmsg_flags;
3082 	s32			rtmsg_ifindex;
3083 };
3084 
3085 static int routing_ioctl(struct net *net, struct socket *sock,
3086 			 unsigned int cmd, void __user *argp)
3087 {
3088 	int ret;
3089 	void *r = NULL;
3090 	struct in6_rtmsg r6;
3091 	struct rtentry r4;
3092 	char devname[16];
3093 	u32 rtdev;
3094 	mm_segment_t old_fs = get_fs();
3095 
3096 	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3097 		struct in6_rtmsg32 __user *ur6 = argp;
3098 		ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3099 			3 * sizeof(struct in6_addr));
3100 		ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3101 		ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3102 		ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3103 		ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3104 		ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3105 		ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3106 		ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3107 
3108 		r = (void *) &r6;
3109 	} else { /* ipv4 */
3110 		struct rtentry32 __user *ur4 = argp;
3111 		ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3112 					3 * sizeof(struct sockaddr));
3113 		ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3114 		ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3115 		ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3116 		ret |= get_user(r4.rt_window, &(ur4->rt_window));
3117 		ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3118 		ret |= get_user(rtdev, &(ur4->rt_dev));
3119 		if (rtdev) {
3120 			ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3121 			r4.rt_dev = (char __user __force *)devname;
3122 			devname[15] = 0;
3123 		} else
3124 			r4.rt_dev = NULL;
3125 
3126 		r = (void *) &r4;
3127 	}
3128 
3129 	if (ret) {
3130 		ret = -EFAULT;
3131 		goto out;
3132 	}
3133 
3134 	set_fs(KERNEL_DS);
3135 	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3136 	set_fs(old_fs);
3137 
3138 out:
3139 	return ret;
3140 }
3141 
3142 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3143  * for some operations; this forces use of the newer bridge-utils that
3144  * use compatible ioctls
3145  */
3146 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3147 {
3148 	compat_ulong_t tmp;
3149 
3150 	if (get_user(tmp, argp))
3151 		return -EFAULT;
3152 	if (tmp == BRCTL_GET_VERSION)
3153 		return BRCTL_VERSION + 1;
3154 	return -EINVAL;
3155 }
3156 
3157 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3158 			 unsigned int cmd, unsigned long arg)
3159 {
3160 	void __user *argp = compat_ptr(arg);
3161 	struct sock *sk = sock->sk;
3162 	struct net *net = sock_net(sk);
3163 
3164 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3165 		return compat_ifr_data_ioctl(net, cmd, argp);
3166 
3167 	switch (cmd) {
3168 	case SIOCSIFBR:
3169 	case SIOCGIFBR:
3170 		return old_bridge_ioctl(argp);
3171 	case SIOCGIFNAME:
3172 		return dev_ifname32(net, argp);
3173 	case SIOCGIFCONF:
3174 		return dev_ifconf(net, argp);
3175 	case SIOCETHTOOL:
3176 		return ethtool_ioctl(net, argp);
3177 	case SIOCWANDEV:
3178 		return compat_siocwandev(net, argp);
3179 	case SIOCGIFMAP:
3180 	case SIOCSIFMAP:
3181 		return compat_sioc_ifmap(net, cmd, argp);
3182 	case SIOCBONDENSLAVE:
3183 	case SIOCBONDRELEASE:
3184 	case SIOCBONDSETHWADDR:
3185 	case SIOCBONDCHANGEACTIVE:
3186 		return bond_ioctl(net, cmd, argp);
3187 	case SIOCADDRT:
3188 	case SIOCDELRT:
3189 		return routing_ioctl(net, sock, cmd, argp);
3190 	case SIOCGSTAMP:
3191 		return do_siocgstamp(net, sock, cmd, argp);
3192 	case SIOCGSTAMPNS:
3193 		return do_siocgstampns(net, sock, cmd, argp);
3194 	case SIOCBONDSLAVEINFOQUERY:
3195 	case SIOCBONDINFOQUERY:
3196 	case SIOCSHWTSTAMP:
3197 	case SIOCGHWTSTAMP:
3198 		return compat_ifr_data_ioctl(net, cmd, argp);
3199 
3200 	case FIOSETOWN:
3201 	case SIOCSPGRP:
3202 	case FIOGETOWN:
3203 	case SIOCGPGRP:
3204 	case SIOCBRADDBR:
3205 	case SIOCBRDELBR:
3206 	case SIOCGIFVLAN:
3207 	case SIOCSIFVLAN:
3208 	case SIOCADDDLCI:
3209 	case SIOCDELDLCI:
3210 	case SIOCGSKNS:
3211 		return sock_ioctl(file, cmd, arg);
3212 
3213 	case SIOCGIFFLAGS:
3214 	case SIOCSIFFLAGS:
3215 	case SIOCGIFMETRIC:
3216 	case SIOCSIFMETRIC:
3217 	case SIOCGIFMTU:
3218 	case SIOCSIFMTU:
3219 	case SIOCGIFMEM:
3220 	case SIOCSIFMEM:
3221 	case SIOCGIFHWADDR:
3222 	case SIOCSIFHWADDR:
3223 	case SIOCADDMULTI:
3224 	case SIOCDELMULTI:
3225 	case SIOCGIFINDEX:
3226 	case SIOCGIFADDR:
3227 	case SIOCSIFADDR:
3228 	case SIOCSIFHWBROADCAST:
3229 	case SIOCDIFADDR:
3230 	case SIOCGIFBRDADDR:
3231 	case SIOCSIFBRDADDR:
3232 	case SIOCGIFDSTADDR:
3233 	case SIOCSIFDSTADDR:
3234 	case SIOCGIFNETMASK:
3235 	case SIOCSIFNETMASK:
3236 	case SIOCSIFPFLAGS:
3237 	case SIOCGIFPFLAGS:
3238 	case SIOCGIFTXQLEN:
3239 	case SIOCSIFTXQLEN:
3240 	case SIOCBRADDIF:
3241 	case SIOCBRDELIF:
3242 	case SIOCSIFNAME:
3243 	case SIOCGMIIPHY:
3244 	case SIOCGMIIREG:
3245 	case SIOCSMIIREG:
3246 		return dev_ifsioc(net, sock, cmd, argp);
3247 
3248 	case SIOCSARP:
3249 	case SIOCGARP:
3250 	case SIOCDARP:
3251 	case SIOCATMARK:
3252 		return sock_do_ioctl(net, sock, cmd, arg);
3253 	}
3254 
3255 	return -ENOIOCTLCMD;
3256 }
3257 
3258 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3259 			      unsigned long arg)
3260 {
3261 	struct socket *sock = file->private_data;
3262 	int ret = -ENOIOCTLCMD;
3263 	struct sock *sk;
3264 	struct net *net;
3265 
3266 	sk = sock->sk;
3267 	net = sock_net(sk);
3268 
3269 	if (sock->ops->compat_ioctl)
3270 		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3271 
3272 	if (ret == -ENOIOCTLCMD &&
3273 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3274 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3275 
3276 	if (ret == -ENOIOCTLCMD)
3277 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3278 
3279 	return ret;
3280 }
3281 #endif
3282 
3283 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3284 {
3285 	return sock->ops->bind(sock, addr, addrlen);
3286 }
3287 EXPORT_SYMBOL(kernel_bind);
3288 
3289 int kernel_listen(struct socket *sock, int backlog)
3290 {
3291 	return sock->ops->listen(sock, backlog);
3292 }
3293 EXPORT_SYMBOL(kernel_listen);
3294 
3295 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3296 {
3297 	struct sock *sk = sock->sk;
3298 	int err;
3299 
3300 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3301 			       newsock);
3302 	if (err < 0)
3303 		goto done;
3304 
3305 	err = sock->ops->accept(sock, *newsock, flags, true);
3306 	if (err < 0) {
3307 		sock_release(*newsock);
3308 		*newsock = NULL;
3309 		goto done;
3310 	}
3311 
3312 	(*newsock)->ops = sock->ops;
3313 	__module_get((*newsock)->ops->owner);
3314 
3315 done:
3316 	return err;
3317 }
3318 EXPORT_SYMBOL(kernel_accept);
3319 
3320 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3321 		   int flags)
3322 {
3323 	return sock->ops->connect(sock, addr, addrlen, flags);
3324 }
3325 EXPORT_SYMBOL(kernel_connect);
3326 
3327 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3328 			 int *addrlen)
3329 {
3330 	return sock->ops->getname(sock, addr, addrlen, 0);
3331 }
3332 EXPORT_SYMBOL(kernel_getsockname);
3333 
3334 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3335 			 int *addrlen)
3336 {
3337 	return sock->ops->getname(sock, addr, addrlen, 1);
3338 }
3339 EXPORT_SYMBOL(kernel_getpeername);
3340 
3341 int kernel_getsockopt(struct socket *sock, int level, int optname,
3342 			char *optval, int *optlen)
3343 {
3344 	mm_segment_t oldfs = get_fs();
3345 	char __user *uoptval;
3346 	int __user *uoptlen;
3347 	int err;
3348 
3349 	uoptval = (char __user __force *) optval;
3350 	uoptlen = (int __user __force *) optlen;
3351 
3352 	set_fs(KERNEL_DS);
3353 	if (level == SOL_SOCKET)
3354 		err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3355 	else
3356 		err = sock->ops->getsockopt(sock, level, optname, uoptval,
3357 					    uoptlen);
3358 	set_fs(oldfs);
3359 	return err;
3360 }
3361 EXPORT_SYMBOL(kernel_getsockopt);
3362 
3363 int kernel_setsockopt(struct socket *sock, int level, int optname,
3364 			char *optval, unsigned int optlen)
3365 {
3366 	mm_segment_t oldfs = get_fs();
3367 	char __user *uoptval;
3368 	int err;
3369 
3370 	uoptval = (char __user __force *) optval;
3371 
3372 	set_fs(KERNEL_DS);
3373 	if (level == SOL_SOCKET)
3374 		err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3375 	else
3376 		err = sock->ops->setsockopt(sock, level, optname, uoptval,
3377 					    optlen);
3378 	set_fs(oldfs);
3379 	return err;
3380 }
3381 EXPORT_SYMBOL(kernel_setsockopt);
3382 
3383 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3384 		    size_t size, int flags)
3385 {
3386 	if (sock->ops->sendpage)
3387 		return sock->ops->sendpage(sock, page, offset, size, flags);
3388 
3389 	return sock_no_sendpage(sock, page, offset, size, flags);
3390 }
3391 EXPORT_SYMBOL(kernel_sendpage);
3392 
3393 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3394 			   size_t size, int flags)
3395 {
3396 	struct socket *sock = sk->sk_socket;
3397 
3398 	if (sock->ops->sendpage_locked)
3399 		return sock->ops->sendpage_locked(sk, page, offset, size,
3400 						  flags);
3401 
3402 	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3403 }
3404 EXPORT_SYMBOL(kernel_sendpage_locked);
3405 
3406 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3407 {
3408 	mm_segment_t oldfs = get_fs();
3409 	int err;
3410 
3411 	set_fs(KERNEL_DS);
3412 	err = sock->ops->ioctl(sock, cmd, arg);
3413 	set_fs(oldfs);
3414 
3415 	return err;
3416 }
3417 EXPORT_SYMBOL(kernel_sock_ioctl);
3418 
3419 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3420 {
3421 	return sock->ops->shutdown(sock, how);
3422 }
3423 EXPORT_SYMBOL(kernel_sock_shutdown);
3424 
3425 /* This routine returns the IP overhead imposed by a socket i.e.
3426  * the length of the underlying IP header, depending on whether
3427  * this is an IPv4 or IPv6 socket and the length from IP options turned
3428  * on at the socket. Assumes that the caller has a lock on the socket.
3429  */
3430 u32 kernel_sock_ip_overhead(struct sock *sk)
3431 {
3432 	struct inet_sock *inet;
3433 	struct ip_options_rcu *opt;
3434 	u32 overhead = 0;
3435 #if IS_ENABLED(CONFIG_IPV6)
3436 	struct ipv6_pinfo *np;
3437 	struct ipv6_txoptions *optv6 = NULL;
3438 #endif /* IS_ENABLED(CONFIG_IPV6) */
3439 
3440 	if (!sk)
3441 		return overhead;
3442 
3443 	switch (sk->sk_family) {
3444 	case AF_INET:
3445 		inet = inet_sk(sk);
3446 		overhead += sizeof(struct iphdr);
3447 		opt = rcu_dereference_protected(inet->inet_opt,
3448 						sock_owned_by_user(sk));
3449 		if (opt)
3450 			overhead += opt->opt.optlen;
3451 		return overhead;
3452 #if IS_ENABLED(CONFIG_IPV6)
3453 	case AF_INET6:
3454 		np = inet6_sk(sk);
3455 		overhead += sizeof(struct ipv6hdr);
3456 		if (np)
3457 			optv6 = rcu_dereference_protected(np->opt,
3458 							  sock_owned_by_user(sk));
3459 		if (optv6)
3460 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3461 		return overhead;
3462 #endif /* IS_ENABLED(CONFIG_IPV6) */
3463 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3464 		return overhead;
3465 	}
3466 }
3467 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3468