linux - net/socket.c

3388 lines · 83.1 KiB

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, <[email protected]>
7	* Ross Biro
8	* Fred N. van Kempen, <[email protected]>
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	#include <linux/nospec.h>
93
94	#include <linux/uaccess.h>
95	#include <asm/unistd.h>
96
97	#include <net/compat.h>
98	#include <net/wext.h>
99	#include <net/cls_cgroup.h>
100
101	#include <net/sock.h>
102	#include <linux/netfilter.h>
103
104	#include <linux/if_tun.h>
105	#include <linux/ipv6_route.h>
106	#include <linux/route.h>
107	#include <linux/sockios.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 __poll_t 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	* Support routines.
167	* Move socket addresses back and forth across the kernel/user
168	* divide and look after the messy bits.
169	*/
170
171	/**
172	* move_addr_to_kernel - copy a socket address into kernel space
173	* @uaddr: Address in user space
174	* @kaddr: Address in kernel space
175	* @ulen: Length in user space
176	*
177	* The address is copied into kernel space. If the provided address is
178	* too long an error code of -EINVAL is returned. If the copy gives
179	* invalid addresses -EFAULT is returned. On a success 0 is returned.
180	*/
181
182	int move_addr_to_kernel(void __user uaddr, int ulen, struct sockaddr_storage kaddr)
183	{
184	if (ulen < 0 \|\| ulen > sizeof(struct sockaddr_storage))
185	return -EINVAL;
186	if (ulen == 0)
187	return 0;
188	if (copy_from_user(kaddr, uaddr, ulen))
189	return -EFAULT;
190	return audit_sockaddr(ulen, kaddr);
191	}
192
193	/**
194	* move_addr_to_user - copy an address to user space
195	* @kaddr: kernel space address
196	* @klen: length of address in kernel
197	* @uaddr: user space address
198	* @ulen: pointer to user length field
199	*
200	* The value pointed to by ulen on entry is the buffer length available.
201	* This is overwritten with the buffer space used. -EINVAL is returned
202	* if an overlong buffer is specified or a negative buffer size. -EFAULT
203	* is returned if either the buffer or the length field are not
204	* accessible.
205	* After copying the data up to the limit the user specifies, the true
206	* length of the data is written over the length limit the user
207	* specified. Zero is returned for a success.
208	*/
209
210	static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
211	void __user uaddr, int __user ulen)
212	{
213	int err;
214	int len;
215
216	BUG_ON(klen > sizeof(struct sockaddr_storage));
217	err = get_user(len, ulen);
218	if (err)
219	return err;
220	if (len > klen)
221	len = klen;
222	if (len < 0)
223	return -EINVAL;
224	if (len) {
225	if (audit_sockaddr(klen, kaddr))
226	return -ENOMEM;
227	if (copy_to_user(uaddr, kaddr, len))
228	return -EFAULT;
229	}
230	/*
231	* "fromlen shall refer to the value before truncation.."
232	* 1003.1g
233	*/
234	return __put_user(klen, ulen);
235	}
236
237	static struct kmem_cache *sock_inode_cachep __ro_after_init;
238
239	static struct inode sock_alloc_inode(struct super_block sb)
240	{
241	struct socket_alloc *ei;
242	struct socket_wq *wq;
243
244	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
245	if (!ei)
246	return NULL;
247	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
248	if (!wq) {
249	kmem_cache_free(sock_inode_cachep, ei);
250	return NULL;
251	}
252	init_waitqueue_head(&wq->wait);
253	wq->fasync_list = NULL;
254	wq->flags = 0;
255	ei->socket.wq = wq;
256
257	ei->socket.state = SS_UNCONNECTED;
258	ei->socket.flags = 0;
259	ei->socket.ops = NULL;
260	ei->socket.sk = NULL;
261	ei->socket.file = NULL;
262
263	return &ei->vfs_inode;
264	}
265
266	static void sock_destroy_inode(struct inode *inode)
267	{
268	struct socket_alloc *ei;
269
270	ei = container_of(inode, struct socket_alloc, vfs_inode);
271	kfree_rcu(ei->socket.wq, rcu);
272	kmem_cache_free(sock_inode_cachep, ei);
273	}
274
275	static void init_once(void *foo)
276	{
277	struct socket_alloc ei = (struct socket_alloc )foo;
278
279	inode_init_once(&ei->vfs_inode);
280	}
281
282	static void init_inodecache(void)
283	{
284	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
285	sizeof(struct socket_alloc),
286	0,
287	(SLAB_HWCACHE_ALIGN \|
288	SLAB_RECLAIM_ACCOUNT \|
289	SLAB_MEM_SPREAD \| SLAB_ACCOUNT),
290	init_once);
291	BUG_ON(sock_inode_cachep == NULL);
292	}
293
294	static const struct super_operations sockfs_ops = {
295	.alloc_inode = sock_alloc_inode,
296	.destroy_inode = sock_destroy_inode,
297	.statfs = simple_statfs,
298	};
299
300	/*
301	* sockfs_dname() is called from d_path().
302	*/
303	static char sockfs_dname(struct dentry dentry, char *buffer, int buflen)
304	{
305	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
306	d_inode(dentry)->i_ino);
307	}
308
309	static const struct dentry_operations sockfs_dentry_operations = {
310	.d_dname = sockfs_dname,
311	};
312
313	static int sockfs_xattr_get(const struct xattr_handler *handler,
314	struct dentry dentry, struct inode inode,
315	const char suffix, void value, size_t size)
316	{
317	if (value) {
318	if (dentry->d_name.len + 1 > size)
319	return -ERANGE;
320	memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
321	}
322	return dentry->d_name.len + 1;
323	}
324
325	#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
326	#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
327	#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
328
329	static const struct xattr_handler sockfs_xattr_handler = {
330	.name = XATTR_NAME_SOCKPROTONAME,
331	.get = sockfs_xattr_get,
332	};
333
334	static int sockfs_security_xattr_set(const struct xattr_handler *handler,
335	struct dentry dentry, struct inode inode,
336	const char suffix, const void value,
337	size_t size, int flags)
338	{
339	/* Handled by LSM. */
340	return -EAGAIN;
341	}
342
343	static const struct xattr_handler sockfs_security_xattr_handler = {
344	.prefix = XATTR_SECURITY_PREFIX,
345	.set = sockfs_security_xattr_set,
346	};
347
348	static const struct xattr_handler *sockfs_xattr_handlers[] = {
349	&sockfs_xattr_handler,
350	&sockfs_security_xattr_handler,
351	NULL
352	};
353
354	static struct dentry sockfs_mount(struct file_system_type fs_type,
355	int flags, const char dev_name, void data)
356	{
357	return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
358	sockfs_xattr_handlers,
359	&sockfs_dentry_operations, SOCKFS_MAGIC);
360	}
361
362	static struct vfsmount *sock_mnt __read_mostly;
363
364	static struct file_system_type sock_fs_type = {
365	.name = "sockfs",
366	.mount = sockfs_mount,
367	.kill_sb = kill_anon_super,
368	};
369
370	/*
371	* Obtains the first available file descriptor and sets it up for use.
372	*
373	* These functions create file structures and maps them to fd space
374	* of the current process. On success it returns file descriptor
375	* and file struct implicitly stored in sock->file.
376	* Note that another thread may close file descriptor before we return
377	* from this function. We use the fact that now we do not refer
378	* to socket after mapping. If one day we will need it, this
379	* function will increment ref. count on file by 1.
380	*
381	* In any case returned fd MAY BE not valid!
382	* This race condition is unavoidable
383	* with shared fd spaces, we cannot solve it inside kernel,
384	* but we take care of internal coherence yet.
385	*/
386
387	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
388	{
389	struct file *file;
390
391	if (!dname)
392	dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
393
394	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
395	O_RDWR \| (flags & O_NONBLOCK),
396	&socket_file_ops);
397	if (IS_ERR(file)) {
398	sock_release(sock);
399	return file;
400	}
401
402	sock->file = file;
403	file->private_data = sock;
404	return file;
405	}
406	EXPORT_SYMBOL(sock_alloc_file);
407
408	static int sock_map_fd(struct socket *sock, int flags)
409	{
410	struct file *newfile;
411	int fd = get_unused_fd_flags(flags);
412	if (unlikely(fd < 0)) {
413	sock_release(sock);
414	return fd;
415	}
416
417	newfile = sock_alloc_file(sock, flags, NULL);
418	if (likely(!IS_ERR(newfile))) {
419	fd_install(fd, newfile);
420	return fd;
421	}
422
423	put_unused_fd(fd);
424	return PTR_ERR(newfile);
425	}
426
427	struct socket sock_from_file(struct file file, int *err)
428	{
429	if (file->f_op == &socket_file_ops)
430	return file->private_data; /* set in sock_map_fd */
431
432	*err = -ENOTSOCK;
433	return NULL;
434	}
435	EXPORT_SYMBOL(sock_from_file);
436
437	/**
438	* sockfd_lookup - Go from a file number to its socket slot
439	* @fd: file handle
440	* @err: pointer to an error code return
441	*
442	* The file handle passed in is locked and the socket it is bound
443	* to is returned. If an error occurs the err pointer is overwritten
444	* with a negative errno code and NULL is returned. The function checks
445	* for both invalid handles and passing a handle which is not a socket.
446	*
447	* On a success the socket object pointer is returned.
448	*/
449
450	struct socket sockfd_lookup(int fd, int err)
451	{
452	struct file *file;
453	struct socket *sock;
454
455	file = fget(fd);
456	if (!file) {
457	*err = -EBADF;
458	return NULL;
459	}
460
461	sock = sock_from_file(file, err);
462	if (!sock)
463	fput(file);
464	return sock;
465	}
466	EXPORT_SYMBOL(sockfd_lookup);
467
468	static struct socket sockfd_lookup_light(int fd, int err, int *fput_needed)
469	{
470	struct fd f = fdget(fd);
471	struct socket *sock;
472
473	*err = -EBADF;
474	if (f.file) {
475	sock = sock_from_file(f.file, err);
476	if (likely(sock)) {
477	*fput_needed = f.flags;
478	return sock;
479	}
480	fdput(f);
481	}
482	return NULL;
483	}
484
485	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
486	size_t size)
487	{
488	ssize_t len;
489	ssize_t used = 0;
490
491	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
492	if (len < 0)
493	return len;
494	used += len;
495	if (buffer) {
496	if (size < used)
497	return -ERANGE;
498	buffer += len;
499	}
500
501	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
502	used += len;
503	if (buffer) {
504	if (size < used)
505	return -ERANGE;
506	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
507	buffer += len;
508	}
509
510	return used;
511	}
512
513	static int sockfs_setattr(struct dentry dentry, struct iattr iattr)
514	{
515	int err = simple_setattr(dentry, iattr);
516
517	if (!err && (iattr->ia_valid & ATTR_UID)) {
518	struct socket *sock = SOCKET_I(d_inode(dentry));
519
520	if (sock->sk)
521	sock->sk->sk_uid = iattr->ia_uid;
522	else
523	err = -ENOENT;
524	}
525
526	return err;
527	}
528
529	static const struct inode_operations sockfs_inode_ops = {
530	.listxattr = sockfs_listxattr,
531	.setattr = sockfs_setattr,
532	};
533
534	/**
535	* sock_alloc - allocate a socket
536	*
537	* Allocate a new inode and socket object. The two are bound together
538	* and initialised. The socket is then returned. If we are out of inodes
539	* NULL is returned.
540	*/
541
542	struct socket *sock_alloc(void)
543	{
544	struct inode *inode;
545	struct socket *sock;
546
547	inode = new_inode_pseudo(sock_mnt->mnt_sb);
548	if (!inode)
549	return NULL;
550
551	sock = SOCKET_I(inode);
552
553	inode->i_ino = get_next_ino();
554	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
555	inode->i_uid = current_fsuid();
556	inode->i_gid = current_fsgid();
557	inode->i_op = &sockfs_inode_ops;
558
559	return sock;
560	}
561	EXPORT_SYMBOL(sock_alloc);
562
563	/**
564	* sock_release - close a socket
565	* @sock: socket to close
566	*
567	* The socket is released from the protocol stack if it has a release
568	* callback, and the inode is then released if the socket is bound to
569	* an inode not a file.
570	*/
571
572	static void __sock_release(struct socket sock, struct inode inode)
573	{
574	if (sock->ops) {
575	struct module *owner = sock->ops->owner;
576
577	if (inode)
578	inode_lock(inode);
579	sock->ops->release(sock);
580	if (inode)
581	inode_unlock(inode);
582	sock->ops = NULL;
583	module_put(owner);
584	}
585
586	if (sock->wq->fasync_list)
587	pr_err("%s: fasync list not empty!\n", __func__);
588
589	if (!sock->file) {
590	iput(SOCK_INODE(sock));
591	return;
592	}
593	sock->file = NULL;
594	}
595
596	void sock_release(struct socket *sock)
597	{
598	__sock_release(sock, NULL);
599	}
600	EXPORT_SYMBOL(sock_release);
601
602	void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
603	{
604	u8 flags = *tx_flags;
605
606	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
607	flags \|= SKBTX_HW_TSTAMP;
608
609	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
610	flags \|= SKBTX_SW_TSTAMP;
611
612	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
613	flags \|= SKBTX_SCHED_TSTAMP;
614
615	*tx_flags = flags;
616	}
617	EXPORT_SYMBOL(__sock_tx_timestamp);
618
619	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
620	{
621	int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
622	BUG_ON(ret == -EIOCBQUEUED);
623	return ret;
624	}
625
626	int sock_sendmsg(struct socket sock, struct msghdr msg)
627	{
628	int err = security_socket_sendmsg(sock, msg,
629	msg_data_left(msg));
630
631	return err ?: sock_sendmsg_nosec(sock, msg);
632	}
633	EXPORT_SYMBOL(sock_sendmsg);
634
635	int kernel_sendmsg(struct socket sock, struct msghdr msg,
636	struct kvec *vec, size_t num, size_t size)
637	{
638	iov_iter_kvec(&msg->msg_iter, WRITE \| ITER_KVEC, vec, num, size);
639	return sock_sendmsg(sock, msg);
640	}
641	EXPORT_SYMBOL(kernel_sendmsg);
642
643	int kernel_sendmsg_locked(struct sock sk, struct msghdr msg,
644	struct kvec *vec, size_t num, size_t size)
645	{
646	struct socket *sock = sk->sk_socket;
647
648	if (!sock->ops->sendmsg_locked)
649	return sock_no_sendmsg_locked(sk, msg, size);
650
651	iov_iter_kvec(&msg->msg_iter, WRITE \| ITER_KVEC, vec, num, size);
652
653	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
654	}
655	EXPORT_SYMBOL(kernel_sendmsg_locked);
656
657	static bool skb_is_err_queue(const struct sk_buff *skb)
658	{
659	/* pkt_type of skbs enqueued on the error queue are set to
660	* PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
661	* in recvmsg, since skbs received on a local socket will never
662	* have a pkt_type of PACKET_OUTGOING.
663	*/
664	return skb->pkt_type == PACKET_OUTGOING;
665	}
666
667	/* On transmit, software and hardware timestamps are returned independently.
668	* As the two skb clones share the hardware timestamp, which may be updated
669	* before the software timestamp is received, a hardware TX timestamp may be
670	* returned only if there is no software TX timestamp. Ignore false software
671	* timestamps, which may be made in the __sock_recv_timestamp() call when the
672	* option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
673	* hardware timestamp.
674	*/
675	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
676	{
677	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
678	}
679
680	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb)
681	{
682	struct scm_ts_pktinfo ts_pktinfo;
683	struct net_device *orig_dev;
684
685	if (!skb_mac_header_was_set(skb))
686	return;
687
688	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
689
690	rcu_read_lock();
691	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
692	if (orig_dev)
693	ts_pktinfo.if_index = orig_dev->ifindex;
694	rcu_read_unlock();
695
696	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
697	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
698	sizeof(ts_pktinfo), &ts_pktinfo);
699	}
700
701	/*
702	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
703	*/
704	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
705	struct sk_buff *skb)
706	{
707	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
708	struct scm_timestamping tss;
709	int empty = 1, false_tstamp = 0;
710	struct skb_shared_hwtstamps *shhwtstamps =
711	skb_hwtstamps(skb);
712
713	/* Race occurred between timestamp enabling and packet
714	receiving. Fill in the current time for now. */
715	if (need_software_tstamp && skb->tstamp == 0) {
716	__net_timestamp(skb);
717	false_tstamp = 1;
718	}
719
720	if (need_software_tstamp) {
721	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
722	struct timeval tv;
723	skb_get_timestamp(skb, &tv);
724	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
725	sizeof(tv), &tv);
726	} else {
727	struct timespec ts;
728	skb_get_timestampns(skb, &ts);
729	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
730	sizeof(ts), &ts);
731	}
732	}
733
734	memset(&tss, 0, sizeof(tss));
735	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
736	ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
737	empty = 0;
738	if (shhwtstamps &&
739	(sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
740	!skb_is_swtx_tstamp(skb, false_tstamp) &&
741	ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
742	empty = 0;
743	if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
744	!skb_is_err_queue(skb))
745	put_ts_pktinfo(msg, skb);
746	}
747	if (!empty) {
748	put_cmsg(msg, SOL_SOCKET,
749	SCM_TIMESTAMPING, sizeof(tss), &tss);
750
751	if (skb_is_err_queue(skb) && skb->len &&
752	SKB_EXT_ERR(skb)->opt_stats)
753	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
754	skb->len, skb->data);
755	}
756	}
757	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
758
759	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
760	struct sk_buff *skb)
761	{
762	int ack;
763
764	if (!sock_flag(sk, SOCK_WIFI_STATUS))
765	return;
766	if (!skb->wifi_acked_valid)
767	return;
768
769	ack = skb->wifi_acked;
770
771	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
772	}
773	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
774
775	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
776	struct sk_buff *skb)
777	{
778	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
779	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
780	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
781	}
782
783	void __sock_recv_ts_and_drops(struct msghdr msg, struct sock sk,
784	struct sk_buff *skb)
785	{
786	sock_recv_timestamp(msg, sk, skb);
787	sock_recv_drops(msg, sk, skb);
788	}
789	EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
790
791	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
792	int flags)
793	{
794	return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
795	}
796
797	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
798	{
799	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
800
801	return err ?: sock_recvmsg_nosec(sock, msg, flags);
802	}
803	EXPORT_SYMBOL(sock_recvmsg);
804
805	/**
806	* kernel_recvmsg - Receive a message from a socket (kernel space)
807	* @sock: The socket to receive the message from
808	* @msg: Received message
809	* @vec: Input s/g array for message data
810	* @num: Size of input s/g array
811	* @size: Number of bytes to read
812	* @flags: Message flags (MSG_DONTWAIT, etc...)
813	*
814	* On return the msg structure contains the scatter/gather array passed in the
815	* vec argument. The array is modified so that it consists of the unfilled
816	* portion of the original array.
817	*
818	* The returned value is the total number of bytes received, or an error.
819	*/
820	int kernel_recvmsg(struct socket sock, struct msghdr msg,
821	struct kvec *vec, size_t num, size_t size, int flags)
822	{
823	mm_segment_t oldfs = get_fs();
824	int result;
825
826	iov_iter_kvec(&msg->msg_iter, READ \| ITER_KVEC, vec, num, size);
827	set_fs(KERNEL_DS);
828	result = sock_recvmsg(sock, msg, flags);
829	set_fs(oldfs);
830	return result;
831	}
832	EXPORT_SYMBOL(kernel_recvmsg);
833
834	static ssize_t sock_sendpage(struct file file, struct page page,
835	int offset, size_t size, loff_t *ppos, int more)
836	{
837	struct socket *sock;
838	int flags;
839
840	sock = file->private_data;
841
842	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
843	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
844	flags \|= more;
845
846	return kernel_sendpage(sock, page, offset, size, flags);
847	}
848
849	static ssize_t sock_splice_read(struct file file, loff_t ppos,
850	struct pipe_inode_info *pipe, size_t len,
851	unsigned int flags)
852	{
853	struct socket *sock = file->private_data;
854
855	if (unlikely(!sock->ops->splice_read))
856	return -EINVAL;
857
858	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
859	}
860
861	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
862	{
863	struct file *file = iocb->ki_filp;
864	struct socket *sock = file->private_data;
865	struct msghdr msg = {.msg_iter = *to,
866	.msg_iocb = iocb};
867	ssize_t res;
868
869	if (file->f_flags & O_NONBLOCK)
870	msg.msg_flags = MSG_DONTWAIT;
871
872	if (iocb->ki_pos != 0)
873	return -ESPIPE;
874
875	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
876	return 0;
877
878	res = sock_recvmsg(sock, &msg, msg.msg_flags);
879	*to = msg.msg_iter;
880	return res;
881	}
882
883	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
884	{
885	struct file *file = iocb->ki_filp;
886	struct socket *sock = file->private_data;
887	struct msghdr msg = {.msg_iter = *from,
888	.msg_iocb = iocb};
889	ssize_t res;
890
891	if (iocb->ki_pos != 0)
892	return -ESPIPE;
893
894	if (file->f_flags & O_NONBLOCK)
895	msg.msg_flags = MSG_DONTWAIT;
896
897	if (sock->type == SOCK_SEQPACKET)
898	msg.msg_flags \|= MSG_EOR;
899
900	res = sock_sendmsg(sock, &msg);
901	*from = msg.msg_iter;
902	return res;
903	}
904
905	/*
906	* Atomic setting of ioctl hooks to avoid race
907	* with module unload.
908	*/
909
910	static DEFINE_MUTEX(br_ioctl_mutex);
911	static int (br_ioctl_hook) (struct net , unsigned int cmd, void __user *arg);
912
913	void brioctl_set(int (hook) (struct net , unsigned int, void __user *))
914	{
915	mutex_lock(&br_ioctl_mutex);
916	br_ioctl_hook = hook;
917	mutex_unlock(&br_ioctl_mutex);
918	}
919	EXPORT_SYMBOL(brioctl_set);
920
921	static DEFINE_MUTEX(vlan_ioctl_mutex);
922	static int (vlan_ioctl_hook) (struct net , void __user *arg);
923
924	void vlan_ioctl_set(int (hook) (struct net , void __user *))
925	{
926	mutex_lock(&vlan_ioctl_mutex);
927	vlan_ioctl_hook = hook;
928	mutex_unlock(&vlan_ioctl_mutex);
929	}
930	EXPORT_SYMBOL(vlan_ioctl_set);
931
932	static DEFINE_MUTEX(dlci_ioctl_mutex);
933	static int (dlci_ioctl_hook) (unsigned int, void __user );
934
935	void dlci_ioctl_set(int (hook) (unsigned int, void __user ))
936	{
937	mutex_lock(&dlci_ioctl_mutex);
938	dlci_ioctl_hook = hook;
939	mutex_unlock(&dlci_ioctl_mutex);
940	}
941	EXPORT_SYMBOL(dlci_ioctl_set);
942
943	static long sock_do_ioctl(struct net net, struct socket sock,
944	unsigned int cmd, unsigned long arg)
945	{
946	int err;
947	void __user argp = (void __user )arg;
948
949	err = sock->ops->ioctl(sock, cmd, arg);
950
951	/*
952	* If this ioctl is unknown try to hand it down
953	* to the NIC driver.
954	*/
955	if (err != -ENOIOCTLCMD)
956	return err;
957
958	if (cmd == SIOCGIFCONF) {
959	struct ifconf ifc;
960	if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
961	return -EFAULT;
962	rtnl_lock();
963	err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
964	rtnl_unlock();
965	if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
966	err = -EFAULT;
967	} else {
968	struct ifreq ifr;
969	bool need_copyout;
970	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
971	return -EFAULT;
972	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
973	if (!err && need_copyout)
974	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
975	return -EFAULT;
976	}
977	return err;
978	}
979
980	/*
981	* With an ioctl, arg may well be a user mode pointer, but we don't know
982	* what to do with it - that's up to the protocol still.
983	*/
984
985	struct ns_common get_net_ns(struct ns_common ns)
986	{
987	return &get_net(container_of(ns, struct net, ns))->ns;
988	}
989	EXPORT_SYMBOL_GPL(get_net_ns);
990
991	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
992	{
993	struct socket *sock;
994	struct sock *sk;
995	void __user argp = (void __user )arg;
996	int pid, err;
997	struct net *net;
998
999	sock = file->private_data;
1000	sk = sock->sk;
1001	net = sock_net(sk);
1002	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1003	struct ifreq ifr;
1004	bool need_copyout;
1005	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1006	return -EFAULT;
1007	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1008	if (!err && need_copyout)
1009	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1010	return -EFAULT;
1011	} else
1012	#ifdef CONFIG_WEXT_CORE
1013	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1014	err = wext_handle_ioctl(net, cmd, argp);
1015	} else
1016	#endif
1017	switch (cmd) {
1018	case FIOSETOWN:
1019	case SIOCSPGRP:
1020	err = -EFAULT;
1021	if (get_user(pid, (int __user *)argp))
1022	break;
1023	err = f_setown(sock->file, pid, 1);
1024	break;
1025	case FIOGETOWN:
1026	case SIOCGPGRP:
1027	err = put_user(f_getown(sock->file),
1028	(int __user *)argp);
1029	break;
1030	case SIOCGIFBR:
1031	case SIOCSIFBR:
1032	case SIOCBRADDBR:
1033	case SIOCBRDELBR:
1034	err = -ENOPKG;
1035	if (!br_ioctl_hook)
1036	request_module("bridge");
1037
1038	mutex_lock(&br_ioctl_mutex);
1039	if (br_ioctl_hook)
1040	err = br_ioctl_hook(net, cmd, argp);
1041	mutex_unlock(&br_ioctl_mutex);
1042	break;
1043	case SIOCGIFVLAN:
1044	case SIOCSIFVLAN:
1045	err = -ENOPKG;
1046	if (!vlan_ioctl_hook)
1047	request_module("8021q");
1048
1049	mutex_lock(&vlan_ioctl_mutex);
1050	if (vlan_ioctl_hook)
1051	err = vlan_ioctl_hook(net, argp);
1052	mutex_unlock(&vlan_ioctl_mutex);
1053	break;
1054	case SIOCADDDLCI:
1055	case SIOCDELDLCI:
1056	err = -ENOPKG;
1057	if (!dlci_ioctl_hook)
1058	request_module("dlci");
1059
1060	mutex_lock(&dlci_ioctl_mutex);
1061	if (dlci_ioctl_hook)
1062	err = dlci_ioctl_hook(cmd, argp);
1063	mutex_unlock(&dlci_ioctl_mutex);
1064	break;
1065	case SIOCGSKNS:
1066	err = -EPERM;
1067	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1068	break;
1069
1070	err = open_related_ns(&net->ns, get_net_ns);
1071	break;
1072	default:
1073	err = sock_do_ioctl(net, sock, cmd, arg);
1074	break;
1075	}
1076	return err;
1077	}
1078
1079	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1080	{
1081	int err;
1082	struct socket *sock = NULL;
1083
1084	err = security_socket_create(family, type, protocol, 1);
1085	if (err)
1086	goto out;
1087
1088	sock = sock_alloc();
1089	if (!sock) {
1090	err = -ENOMEM;
1091	goto out;
1092	}
1093
1094	sock->type = type;
1095	err = security_socket_post_create(sock, family, type, protocol, 1);
1096	if (err)
1097	goto out_release;
1098
1099	out:
1100	*res = sock;
1101	return err;
1102	out_release:
1103	sock_release(sock);
1104	sock = NULL;
1105	goto out;
1106	}
1107	EXPORT_SYMBOL(sock_create_lite);
1108
1109	/* No kernel lock held - perfect */
1110	static __poll_t sock_poll(struct file file, poll_table wait)
1111	{
1112	struct socket *sock = file->private_data;
1113	__poll_t events = poll_requested_events(wait), flag = 0;
1114
1115	if (!sock->ops->poll)
1116	return 0;
1117
1118	if (sk_can_busy_loop(sock->sk)) {
1119	/* poll once if requested by the syscall */
1120	if (events & POLL_BUSY_LOOP)
1121	sk_busy_loop(sock->sk, 1);
1122
1123	/* if this socket can poll_ll, tell the system call */
1124	flag = POLL_BUSY_LOOP;
1125	}
1126
1127	return sock->ops->poll(file, sock, wait) \| flag;
1128	}
1129
1130	static int sock_mmap(struct file file, struct vm_area_struct vma)
1131	{
1132	struct socket *sock = file->private_data;
1133
1134	return sock->ops->mmap(file, sock, vma);
1135	}
1136
1137	static int sock_close(struct inode inode, struct file filp)
1138	{
1139	__sock_release(SOCKET_I(inode), inode);
1140	return 0;
1141	}
1142
1143	/*
1144	* Update the socket async list
1145	*
1146	* Fasync_list locking strategy.
1147	*
1148	* 1. fasync_list is modified only under process context socket lock
1149	* i.e. under semaphore.
1150	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1151	* or under socket lock
1152	*/
1153
1154	static int sock_fasync(int fd, struct file *filp, int on)
1155	{
1156	struct socket *sock = filp->private_data;
1157	struct sock *sk = sock->sk;
1158	struct socket_wq *wq;
1159
1160	if (sk == NULL)
1161	return -EINVAL;
1162
1163	lock_sock(sk);
1164	wq = sock->wq;
1165	fasync_helper(fd, filp, on, &wq->fasync_list);
1166
1167	if (!wq->fasync_list)
1168	sock_reset_flag(sk, SOCK_FASYNC);
1169	else
1170	sock_set_flag(sk, SOCK_FASYNC);
1171
1172	release_sock(sk);
1173	return 0;
1174	}
1175
1176	/* This function may be called only under rcu_lock */
1177
1178	int sock_wake_async(struct socket_wq *wq, int how, int band)
1179	{
1180	if (!wq \|\| !wq->fasync_list)
1181	return -1;
1182
1183	switch (how) {
1184	case SOCK_WAKE_WAITD:
1185	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1186	break;
1187	goto call_kill;
1188	case SOCK_WAKE_SPACE:
1189	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1190	break;
1191	/* fall through */
1192	case SOCK_WAKE_IO:
1193	call_kill:
1194	kill_fasync(&wq->fasync_list, SIGIO, band);
1195	break;
1196	case SOCK_WAKE_URG:
1197	kill_fasync(&wq->fasync_list, SIGURG, band);
1198	}
1199
1200	return 0;
1201	}
1202	EXPORT_SYMBOL(sock_wake_async);
1203
1204	int __sock_create(struct net *net, int family, int type, int protocol,
1205	struct socket **res, int kern)
1206	{
1207	int err;
1208	struct socket *sock;
1209	const struct net_proto_family *pf;
1210
1211	/*
1212	* Check protocol is in range
1213	*/
1214	if (family < 0 \|\| family >= NPROTO)
1215	return -EAFNOSUPPORT;
1216	if (type < 0 \|\| type >= SOCK_MAX)
1217	return -EINVAL;
1218
1219	/* Compatibility.
1220
1221	This uglymoron is moved from INET layer to here to avoid
1222	deadlock in module load.
1223	*/
1224	if (family == PF_INET && type == SOCK_PACKET) {
1225	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1226	current->comm);
1227	family = PF_PACKET;
1228	}
1229
1230	err = security_socket_create(family, type, protocol, kern);
1231	if (err)
1232	return err;
1233
1234	/*
1235	* Allocate the socket and allow the family to set things up. if
1236	* the protocol is 0, the family is instructed to select an appropriate
1237	* default.
1238	*/
1239	sock = sock_alloc();
1240	if (!sock) {
1241	net_warn_ratelimited("socket: no more sockets\n");
1242	return -ENFILE; /* Not exactly a match, but its the
1243	closest posix thing */
1244	}
1245
1246	sock->type = type;
1247
1248	#ifdef CONFIG_MODULES
1249	/* Attempt to load a protocol module if the find failed.
1250	*
1251	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1252	* requested real, full-featured networking support upon configuration.
1253	* Otherwise module support will break!
1254	*/
1255	if (rcu_access_pointer(net_families[family]) == NULL)
1256	request_module("net-pf-%d", family);
1257	#endif
1258
1259	rcu_read_lock();
1260	pf = rcu_dereference(net_families[family]);
1261	err = -EAFNOSUPPORT;
1262	if (!pf)
1263	goto out_release;
1264
1265	/*
1266	* We will call the ->create function, that possibly is in a loadable
1267	* module, so we have to bump that loadable module refcnt first.
1268	*/
1269	if (!try_module_get(pf->owner))
1270	goto out_release;
1271
1272	/* Now protected by module ref count */
1273	rcu_read_unlock();
1274
1275	err = pf->create(net, sock, protocol, kern);
1276	if (err < 0)
1277	goto out_module_put;
1278
1279	/*
1280	* Now to bump the refcnt of the [loadable] module that owns this
1281	* socket at sock_release time we decrement its refcnt.
1282	*/
1283	if (!try_module_get(sock->ops->owner))
1284	goto out_module_busy;
1285
1286	/*
1287	* Now that we're done with the ->create function, the [loadable]
1288	* module can have its refcnt decremented
1289	*/
1290	module_put(pf->owner);
1291	err = security_socket_post_create(sock, family, type, protocol, kern);
1292	if (err)
1293	goto out_sock_release;
1294	*res = sock;
1295
1296	return 0;
1297
1298	out_module_busy:
1299	err = -EAFNOSUPPORT;
1300	out_module_put:
1301	sock->ops = NULL;
1302	module_put(pf->owner);
1303	out_sock_release:
1304	sock_release(sock);
1305	return err;
1306
1307	out_release:
1308	rcu_read_unlock();
1309	goto out_sock_release;
1310	}
1311	EXPORT_SYMBOL(__sock_create);
1312
1313	int sock_create(int family, int type, int protocol, struct socket **res)
1314	{
1315	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1316	}
1317	EXPORT_SYMBOL(sock_create);
1318
1319	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1320	{
1321	return __sock_create(net, family, type, protocol, res, 1);
1322	}
1323	EXPORT_SYMBOL(sock_create_kern);
1324
1325	int __sys_socket(int family, int type, int protocol)
1326	{
1327	int retval;
1328	struct socket *sock;
1329	int flags;
1330
1331	/* Check the SOCK_* constants for consistency. */
1332	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1333	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1334	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1335	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1336
1337	flags = type & ~SOCK_TYPE_MASK;
1338	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1339	return -EINVAL;
1340	type &= SOCK_TYPE_MASK;
1341
1342	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1343	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1344
1345	retval = sock_create(family, type, protocol, &sock);
1346	if (retval < 0)
1347	return retval;
1348
1349	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1350	}
1351
1352	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1353	{
1354	return __sys_socket(family, type, protocol);
1355	}
1356
1357	/*
1358	* Create a pair of connected sockets.
1359	*/
1360
1361	int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1362	{
1363	struct socket sock1, sock2;
1364	int fd1, fd2, err;
1365	struct file newfile1, newfile2;
1366	int flags;
1367
1368	flags = type & ~SOCK_TYPE_MASK;
1369	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1370	return -EINVAL;
1371	type &= SOCK_TYPE_MASK;
1372
1373	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1374	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1375
1376	/*
1377	* reserve descriptors and make sure we won't fail
1378	* to return them to userland.
1379	*/
1380	fd1 = get_unused_fd_flags(flags);
1381	if (unlikely(fd1 < 0))
1382	return fd1;
1383
1384	fd2 = get_unused_fd_flags(flags);
1385	if (unlikely(fd2 < 0)) {
1386	put_unused_fd(fd1);
1387	return fd2;
1388	}
1389
1390	err = put_user(fd1, &usockvec[0]);
1391	if (err)
1392	goto out;
1393
1394	err = put_user(fd2, &usockvec[1]);
1395	if (err)
1396	goto out;
1397
1398	/*
1399	* Obtain the first socket and check if the underlying protocol
1400	* supports the socketpair call.
1401	*/
1402
1403	err = sock_create(family, type, protocol, &sock1);
1404	if (unlikely(err < 0))
1405	goto out;
1406
1407	err = sock_create(family, type, protocol, &sock2);
1408	if (unlikely(err < 0)) {
1409	sock_release(sock1);
1410	goto out;
1411	}
1412
1413	err = security_socket_socketpair(sock1, sock2);
1414	if (unlikely(err)) {
1415	sock_release(sock2);
1416	sock_release(sock1);
1417	goto out;
1418	}
1419
1420	err = sock1->ops->socketpair(sock1, sock2);
1421	if (unlikely(err < 0)) {
1422	sock_release(sock2);
1423	sock_release(sock1);
1424	goto out;
1425	}
1426
1427	newfile1 = sock_alloc_file(sock1, flags, NULL);
1428	if (IS_ERR(newfile1)) {
1429	err = PTR_ERR(newfile1);
1430	sock_release(sock2);
1431	goto out;
1432	}
1433
1434	newfile2 = sock_alloc_file(sock2, flags, NULL);
1435	if (IS_ERR(newfile2)) {
1436	err = PTR_ERR(newfile2);
1437	fput(newfile1);
1438	goto out;
1439	}
1440
1441	audit_fd_pair(fd1, fd2);
1442
1443	fd_install(fd1, newfile1);
1444	fd_install(fd2, newfile2);
1445	return 0;
1446
1447	out:
1448	put_unused_fd(fd2);
1449	put_unused_fd(fd1);
1450	return err;
1451	}
1452
1453	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1454	int __user *, usockvec)
1455	{
1456	return __sys_socketpair(family, type, protocol, usockvec);
1457	}
1458
1459	/*
1460	* Bind a name to a socket. Nothing much to do here since it's
1461	* the protocol's responsibility to handle the local address.
1462	*
1463	* We move the socket address to kernel space before we call
1464	* the protocol layer (having also checked the address is ok).
1465	*/
1466
1467	int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1468	{
1469	struct socket *sock;
1470	struct sockaddr_storage address;
1471	int err, fput_needed;
1472
1473	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1474	if (sock) {
1475	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1476	if (err >= 0) {
1477	err = security_socket_bind(sock,
1478	(struct sockaddr *)&address,
1479	addrlen);
1480	if (!err)
1481	err = sock->ops->bind(sock,
1482	(struct sockaddr *)
1483	&address, addrlen);
1484	}
1485	fput_light(sock->file, fput_needed);
1486	}
1487	return err;
1488	}
1489
1490	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1491	{
1492	return __sys_bind(fd, umyaddr, addrlen);
1493	}
1494
1495	/*
1496	* Perform a listen. Basically, we allow the protocol to do anything
1497	* necessary for a listen, and if that works, we mark the socket as
1498	* ready for listening.
1499	*/
1500
1501	int __sys_listen(int fd, int backlog)
1502	{
1503	struct socket *sock;
1504	int err, fput_needed;
1505	int somaxconn;
1506
1507	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1508	if (sock) {
1509	somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1510	if ((unsigned int)backlog > somaxconn)
1511	backlog = somaxconn;
1512
1513	err = security_socket_listen(sock, backlog);
1514	if (!err)
1515	err = sock->ops->listen(sock, backlog);
1516
1517	fput_light(sock->file, fput_needed);
1518	}
1519	return err;
1520	}
1521
1522	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1523	{
1524	return __sys_listen(fd, backlog);
1525	}
1526
1527	/*
1528	* For accept, we attempt to create a new socket, set up the link
1529	* with the client, wake up the client, then return the new
1530	* connected fd. We collect the address of the connector in kernel
1531	* space and move it to user at the very end. This is unclean because
1532	* we open the socket then return an error.
1533	*
1534	* 1003.1g adds the ability to recvmsg() to query connection pending
1535	* status to recvmsg. We need to add that support in a way thats
1536	* clean when we restructure accept also.
1537	*/
1538
1539	int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1540	int __user *upeer_addrlen, int flags)
1541	{
1542	struct socket sock, newsock;
1543	struct file *newfile;
1544	int err, len, newfd, fput_needed;
1545	struct sockaddr_storage address;
1546
1547	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1548	return -EINVAL;
1549
1550	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1551	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1552
1553	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1554	if (!sock)
1555	goto out;
1556
1557	err = -ENFILE;
1558	newsock = sock_alloc();
1559	if (!newsock)
1560	goto out_put;
1561
1562	newsock->type = sock->type;
1563	newsock->ops = sock->ops;
1564
1565	/*
1566	* We don't need try_module_get here, as the listening socket (sock)
1567	* has the protocol module (sock->ops->owner) held.
1568	*/
1569	__module_get(newsock->ops->owner);
1570
1571	newfd = get_unused_fd_flags(flags);
1572	if (unlikely(newfd < 0)) {
1573	err = newfd;
1574	sock_release(newsock);
1575	goto out_put;
1576	}
1577	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1578	if (IS_ERR(newfile)) {
1579	err = PTR_ERR(newfile);
1580	put_unused_fd(newfd);
1581	goto out_put;
1582	}
1583
1584	err = security_socket_accept(sock, newsock);
1585	if (err)
1586	goto out_fd;
1587
1588	err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1589	if (err < 0)
1590	goto out_fd;
1591
1592	if (upeer_sockaddr) {
1593	len = newsock->ops->getname(newsock,
1594	(struct sockaddr *)&address, 2);
1595	if (len < 0) {
1596	err = -ECONNABORTED;
1597	goto out_fd;
1598	}
1599	err = move_addr_to_user(&address,
1600	len, upeer_sockaddr, upeer_addrlen);
1601	if (err < 0)
1602	goto out_fd;
1603	}
1604
1605	/* File flags are not inherited via accept() unlike another OSes. */
1606
1607	fd_install(newfd, newfile);
1608	err = newfd;
1609
1610	out_put:
1611	fput_light(sock->file, fput_needed);
1612	out:
1613	return err;
1614	out_fd:
1615	fput(newfile);
1616	put_unused_fd(newfd);
1617	goto out_put;
1618	}
1619
1620	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1621	int __user *, upeer_addrlen, int, flags)
1622	{
1623	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1624	}
1625
1626	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1627	int __user *, upeer_addrlen)
1628	{
1629	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1630	}
1631
1632	/*
1633	* Attempt to connect to a socket with the server address. The address
1634	* is in user space so we verify it is OK and move it to kernel space.
1635	*
1636	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1637	* break bindings
1638	*
1639	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1640	* other SEQPACKET protocols that take time to connect() as it doesn't
1641	* include the -EINPROGRESS status for such sockets.
1642	*/
1643
1644	int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1645	{
1646	struct socket *sock;
1647	struct sockaddr_storage address;
1648	int err, fput_needed;
1649
1650	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1651	if (!sock)
1652	goto out;
1653	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1654	if (err < 0)
1655	goto out_put;
1656
1657	err =
1658	security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1659	if (err)
1660	goto out_put;
1661
1662	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1663	sock->file->f_flags);
1664	out_put:
1665	fput_light(sock->file, fput_needed);
1666	out:
1667	return err;
1668	}
1669
1670	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1671	int, addrlen)
1672	{
1673	return __sys_connect(fd, uservaddr, addrlen);
1674	}
1675
1676	/*
1677	* Get the local address ('name') of a socket object. Move the obtained
1678	* name to user space.
1679	*/
1680
1681	int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1682	int __user *usockaddr_len)
1683	{
1684	struct socket *sock;
1685	struct sockaddr_storage address;
1686	int err, fput_needed;
1687
1688	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1689	if (!sock)
1690	goto out;
1691
1692	err = security_socket_getsockname(sock);
1693	if (err)
1694	goto out_put;
1695
1696	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1697	if (err < 0)
1698	goto out_put;
1699	/* "err" is actually length in this case */
1700	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1701
1702	out_put:
1703	fput_light(sock->file, fput_needed);
1704	out:
1705	return err;
1706	}
1707
1708	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1709	int __user *, usockaddr_len)
1710	{
1711	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1712	}
1713
1714	/*
1715	* Get the remote address ('name') of a socket object. Move the obtained
1716	* name to user space.
1717	*/
1718
1719	int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1720	int __user *usockaddr_len)
1721	{
1722	struct socket *sock;
1723	struct sockaddr_storage address;
1724	int err, fput_needed;
1725
1726	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1727	if (sock != NULL) {
1728	err = security_socket_getpeername(sock);
1729	if (err) {
1730	fput_light(sock->file, fput_needed);
1731	return err;
1732	}
1733
1734	err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1735	if (err >= 0)
1736	/* "err" is actually length in this case */
1737	err = move_addr_to_user(&address, err, usockaddr,
1738	usockaddr_len);
1739	fput_light(sock->file, fput_needed);
1740	}
1741	return err;
1742	}
1743
1744	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1745	int __user *, usockaddr_len)
1746	{
1747	return __sys_getpeername(fd, usockaddr, usockaddr_len);
1748	}
1749
1750	/*
1751	* Send a datagram to a given address. We move the address into kernel
1752	* space and check the user space data area is readable before invoking
1753	* the protocol.
1754	*/
1755	int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1756	struct sockaddr __user *addr, int addr_len)
1757	{
1758	struct socket *sock;
1759	struct sockaddr_storage address;
1760	int err;
1761	struct msghdr msg;
1762	struct iovec iov;
1763	int fput_needed;
1764
1765	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1766	if (unlikely(err))
1767	return err;
1768	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769	if (!sock)
1770	goto out;
1771
1772	msg.msg_name = NULL;
1773	msg.msg_control = NULL;
1774	msg.msg_controllen = 0;
1775	msg.msg_namelen = 0;
1776	if (addr) {
1777	err = move_addr_to_kernel(addr, addr_len, &address);
1778	if (err < 0)
1779	goto out_put;
1780	msg.msg_name = (struct sockaddr *)&address;
1781	msg.msg_namelen = addr_len;
1782	}
1783	if (sock->file->f_flags & O_NONBLOCK)
1784	flags \|= MSG_DONTWAIT;
1785	msg.msg_flags = flags;
1786	err = sock_sendmsg(sock, &msg);
1787
1788	out_put:
1789	fput_light(sock->file, fput_needed);
1790	out:
1791	return err;
1792	}
1793
1794	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1795	unsigned int, flags, struct sockaddr __user *, addr,
1796	int, addr_len)
1797	{
1798	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1799	}
1800
1801	/*
1802	* Send a datagram down a socket.
1803	*/
1804
1805	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1806	unsigned int, flags)
1807	{
1808	return __sys_sendto(fd, buff, len, flags, NULL, 0);
1809	}
1810
1811	/*
1812	* Receive a frame from the socket and optionally record the address of the
1813	* sender. We verify the buffers are writable and if needed move the
1814	* sender address from kernel to user space.
1815	*/
1816	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1817	struct sockaddr __user addr, int __user addr_len)
1818	{
1819	struct socket *sock;
1820	struct iovec iov;
1821	struct msghdr msg;
1822	struct sockaddr_storage address;
1823	int err, err2;
1824	int fput_needed;
1825
1826	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1827	if (unlikely(err))
1828	return err;
1829	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1830	if (!sock)
1831	goto out;
1832
1833	msg.msg_control = NULL;
1834	msg.msg_controllen = 0;
1835	/* Save some cycles and don't copy the address if not needed */
1836	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1837	/* We assume all kernel code knows the size of sockaddr_storage */
1838	msg.msg_namelen = 0;
1839	msg.msg_iocb = NULL;
1840	msg.msg_flags = 0;
1841	if (sock->file->f_flags & O_NONBLOCK)
1842	flags \|= MSG_DONTWAIT;
1843	err = sock_recvmsg(sock, &msg, flags);
1844
1845	if (err >= 0 && addr != NULL) {
1846	err2 = move_addr_to_user(&address,
1847	msg.msg_namelen, addr, addr_len);
1848	if (err2 < 0)
1849	err = err2;
1850	}
1851
1852	fput_light(sock->file, fput_needed);
1853	out:
1854	return err;
1855	}
1856
1857	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1858	unsigned int, flags, struct sockaddr __user *, addr,
1859	int __user *, addr_len)
1860	{
1861	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1862	}
1863
1864	/*
1865	* Receive a datagram from a socket.
1866	*/
1867
1868	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1869	unsigned int, flags)
1870	{
1871	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1872	}
1873
1874	/*
1875	* Set a socket option. Because we don't know the option lengths we have
1876	* to pass the user mode parameter for the protocols to sort out.
1877	*/
1878
1879	static int __sys_setsockopt(int fd, int level, int optname,
1880	char __user *optval, int optlen)
1881	{
1882	int err, fput_needed;
1883	struct socket *sock;
1884
1885	if (optlen < 0)
1886	return -EINVAL;
1887
1888	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1889	if (sock != NULL) {
1890	err = security_socket_setsockopt(sock, level, optname);
1891	if (err)
1892	goto out_put;
1893
1894	if (level == SOL_SOCKET)
1895	err =
1896	sock_setsockopt(sock, level, optname, optval,
1897	optlen);
1898	else
1899	err =
1900	sock->ops->setsockopt(sock, level, optname, optval,
1901	optlen);
1902	out_put:
1903	fput_light(sock->file, fput_needed);
1904	}
1905	return err;
1906	}
1907
1908	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1909	char __user *, optval, int, optlen)
1910	{
1911	return __sys_setsockopt(fd, level, optname, optval, optlen);
1912	}
1913
1914	/*
1915	* Get a socket option. Because we don't know the option lengths we have
1916	* to pass a user mode parameter for the protocols to sort out.
1917	*/
1918
1919	static int __sys_getsockopt(int fd, int level, int optname,
1920	char __user optval, int __user optlen)
1921	{
1922	int err, fput_needed;
1923	struct socket *sock;
1924
1925	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1926	if (sock != NULL) {
1927	err = security_socket_getsockopt(sock, level, optname);
1928	if (err)
1929	goto out_put;
1930
1931	if (level == SOL_SOCKET)
1932	err =
1933	sock_getsockopt(sock, level, optname, optval,
1934	optlen);
1935	else
1936	err =
1937	sock->ops->getsockopt(sock, level, optname, optval,
1938	optlen);
1939	out_put:
1940	fput_light(sock->file, fput_needed);
1941	}
1942	return err;
1943	}
1944
1945	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1946	char __user , optval, int __user , optlen)
1947	{
1948	return __sys_getsockopt(fd, level, optname, optval, optlen);
1949	}
1950
1951	/*
1952	* Shutdown a socket.
1953	*/
1954
1955	int __sys_shutdown(int fd, int how)
1956	{
1957	int err, fput_needed;
1958	struct socket *sock;
1959
1960	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1961	if (sock != NULL) {
1962	err = security_socket_shutdown(sock, how);
1963	if (!err)
1964	err = sock->ops->shutdown(sock, how);
1965	fput_light(sock->file, fput_needed);
1966	}
1967	return err;
1968	}
1969
1970	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1971	{
1972	return __sys_shutdown(fd, how);
1973	}
1974
1975	/* A couple of helpful macros for getting the address of the 32/64 bit
1976	* fields which are the same type (int / unsigned) on our platforms.
1977	*/
1978	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1979	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1980	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1981
1982	struct used_address {
1983	struct sockaddr_storage name;
1984	unsigned int name_len;
1985	};
1986
1987	static int copy_msghdr_from_user(struct msghdr *kmsg,
1988	struct user_msghdr __user *umsg,
1989	struct sockaddr __user **save_addr,
1990	struct iovec **iov)
1991	{
1992	struct user_msghdr msg;
1993	ssize_t err;
1994
1995	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1996	return -EFAULT;
1997
1998	kmsg->msg_control = (void __force *)msg.msg_control;
1999	kmsg->msg_controllen = msg.msg_controllen;
2000	kmsg->msg_flags = msg.msg_flags;
2001
2002	kmsg->msg_namelen = msg.msg_namelen;
2003	if (!msg.msg_name)
2004	kmsg->msg_namelen = 0;
2005
2006	if (kmsg->msg_namelen < 0)
2007	return -EINVAL;
2008
2009	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2010	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2011
2012	if (save_addr)
2013	*save_addr = msg.msg_name;
2014
2015	if (msg.msg_name && kmsg->msg_namelen) {
2016	if (!save_addr) {
2017	err = move_addr_to_kernel(msg.msg_name,
2018	kmsg->msg_namelen,
2019	kmsg->msg_name);
2020	if (err < 0)
2021	return err;
2022	}
2023	} else {
2024	kmsg->msg_name = NULL;
2025	kmsg->msg_namelen = 0;
2026	}
2027
2028	if (msg.msg_iovlen > UIO_MAXIOV)
2029	return -EMSGSIZE;
2030
2031	kmsg->msg_iocb = NULL;
2032
2033	return import_iovec(save_addr ? READ : WRITE,
2034	msg.msg_iov, msg.msg_iovlen,
2035	UIO_FASTIOV, iov, &kmsg->msg_iter);
2036	}
2037
2038	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
2039	struct msghdr *msg_sys, unsigned int flags,
2040	struct used_address *used_address,
2041	unsigned int allowed_msghdr_flags)
2042	{
2043	struct compat_msghdr __user *msg_compat =
2044	(struct compat_msghdr __user *)msg;
2045	struct sockaddr_storage address;
2046	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2047	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2048	__aligned(sizeof(__kernel_size_t));
2049	/* 20 is size of ipv6_pktinfo */
2050	unsigned char *ctl_buf = ctl;
2051	int ctl_len;
2052	ssize_t err;
2053
2054	msg_sys->msg_name = &address;
2055
2056	if (MSG_CMSG_COMPAT & flags)
2057	err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2058	else
2059	err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2060	if (err < 0)
2061	return err;
2062
2063	err = -ENOBUFS;
2064
2065	if (msg_sys->msg_controllen > INT_MAX)
2066	goto out_freeiov;
2067	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2068	ctl_len = msg_sys->msg_controllen;
2069	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2070	err =
2071	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2072	sizeof(ctl));
2073	if (err)
2074	goto out_freeiov;
2075	ctl_buf = msg_sys->msg_control;
2076	ctl_len = msg_sys->msg_controllen;
2077	} else if (ctl_len) {
2078	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2079	CMSG_ALIGN(sizeof(struct cmsghdr)));
2080	if (ctl_len > sizeof(ctl)) {
2081	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2082	if (ctl_buf == NULL)
2083	goto out_freeiov;
2084	}
2085	err = -EFAULT;
2086	/*
2087	* Careful! Before this, msg_sys->msg_control contains a user pointer.
2088	* Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2089	* checking falls down on this.
2090	*/
2091	if (copy_from_user(ctl_buf,
2092	(void __user __force *)msg_sys->msg_control,
2093	ctl_len))
2094	goto out_freectl;
2095	msg_sys->msg_control = ctl_buf;
2096	}
2097	msg_sys->msg_flags = flags;
2098
2099	if (sock->file->f_flags & O_NONBLOCK)
2100	msg_sys->msg_flags \|= MSG_DONTWAIT;
2101	/*
2102	* If this is sendmmsg() and current destination address is same as
2103	* previously succeeded address, omit asking LSM's decision.
2104	* used_address->name_len is initialized to UINT_MAX so that the first
2105	* destination address never matches.
2106	*/
2107	if (used_address && msg_sys->msg_name &&
2108	used_address->name_len == msg_sys->msg_namelen &&
2109	!memcmp(&used_address->name, msg_sys->msg_name,
2110	used_address->name_len)) {
2111	err = sock_sendmsg_nosec(sock, msg_sys);
2112	goto out_freectl;
2113	}
2114	err = sock_sendmsg(sock, msg_sys);
2115	/*
2116	* If this is sendmmsg() and sending to current destination address was
2117	* successful, remember it.
2118	*/
2119	if (used_address && err >= 0) {
2120	used_address->name_len = msg_sys->msg_namelen;
2121	if (msg_sys->msg_name)
2122	memcpy(&used_address->name, msg_sys->msg_name,
2123	used_address->name_len);
2124	}
2125
2126	out_freectl:
2127	if (ctl_buf != ctl)
2128	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2129	out_freeiov:
2130	kfree(iov);
2131	return err;
2132	}
2133
2134	/*
2135	* BSD sendmsg interface
2136	*/
2137
2138	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2139	bool forbid_cmsg_compat)
2140	{
2141	int fput_needed, err;
2142	struct msghdr msg_sys;
2143	struct socket *sock;
2144
2145	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2146	return -EINVAL;
2147
2148	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2149	if (!sock)
2150	goto out;
2151
2152	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2153
2154	fput_light(sock->file, fput_needed);
2155	out:
2156	return err;
2157	}
2158
2159	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2160	{
2161	return __sys_sendmsg(fd, msg, flags, true);
2162	}
2163
2164	/*
2165	* Linux sendmmsg interface
2166	*/
2167
2168	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2169	unsigned int flags, bool forbid_cmsg_compat)
2170	{
2171	int fput_needed, err, datagrams;
2172	struct socket *sock;
2173	struct mmsghdr __user *entry;
2174	struct compat_mmsghdr __user *compat_entry;
2175	struct msghdr msg_sys;
2176	struct used_address used_address;
2177	unsigned int oflags = flags;
2178
2179	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2180	return -EINVAL;
2181
2182	if (vlen > UIO_MAXIOV)
2183	vlen = UIO_MAXIOV;
2184
2185	datagrams = 0;
2186
2187	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2188	if (!sock)
2189	return err;
2190
2191	used_address.name_len = UINT_MAX;
2192	entry = mmsg;
2193	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2194	err = 0;
2195	flags \|= MSG_BATCH;
2196
2197	while (datagrams < vlen) {
2198	if (datagrams == vlen - 1)
2199	flags = oflags;
2200
2201	if (MSG_CMSG_COMPAT & flags) {
2202	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2203	&msg_sys, flags, &used_address, MSG_EOR);
2204	if (err < 0)
2205	break;
2206	err = __put_user(err, &compat_entry->msg_len);
2207	++compat_entry;
2208	} else {
2209	err = ___sys_sendmsg(sock,
2210	(struct user_msghdr __user *)entry,
2211	&msg_sys, flags, &used_address, MSG_EOR);
2212	if (err < 0)
2213	break;
2214	err = put_user(err, &entry->msg_len);
2215	++entry;
2216	}
2217
2218	if (err)
2219	break;
2220	++datagrams;
2221	if (msg_data_left(&msg_sys))
2222	break;
2223	cond_resched();
2224	}
2225
2226	fput_light(sock->file, fput_needed);
2227
2228	/* We only return an error if no datagrams were able to be sent */
2229	if (datagrams != 0)
2230	return datagrams;
2231
2232	return err;
2233	}
2234
2235	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2236	unsigned int, vlen, unsigned int, flags)
2237	{
2238	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2239	}
2240
2241	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2242	struct msghdr *msg_sys, unsigned int flags, int nosec)
2243	{
2244	struct compat_msghdr __user *msg_compat =
2245	(struct compat_msghdr __user *)msg;
2246	struct iovec iovstack[UIO_FASTIOV];
2247	struct iovec *iov = iovstack;
2248	unsigned long cmsg_ptr;
2249	int len;
2250	ssize_t err;
2251
2252	/* kernel mode address */
2253	struct sockaddr_storage addr;
2254
2255	/* user mode address pointers */
2256	struct sockaddr __user *uaddr;
2257	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2258
2259	msg_sys->msg_name = &addr;
2260
2261	if (MSG_CMSG_COMPAT & flags)
2262	err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2263	else
2264	err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2265	if (err < 0)
2266	return err;
2267
2268	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2269	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2270
2271	/* We assume all kernel code knows the size of sockaddr_storage */
2272	msg_sys->msg_namelen = 0;
2273
2274	if (sock->file->f_flags & O_NONBLOCK)
2275	flags \|= MSG_DONTWAIT;
2276	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2277	if (err < 0)
2278	goto out_freeiov;
2279	len = err;
2280
2281	if (uaddr != NULL) {
2282	err = move_addr_to_user(&addr,
2283	msg_sys->msg_namelen, uaddr,
2284	uaddr_len);
2285	if (err < 0)
2286	goto out_freeiov;
2287	}
2288	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2289	COMPAT_FLAGS(msg));
2290	if (err)
2291	goto out_freeiov;
2292	if (MSG_CMSG_COMPAT & flags)
2293	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2294	&msg_compat->msg_controllen);
2295	else
2296	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2297	&msg->msg_controllen);
2298	if (err)
2299	goto out_freeiov;
2300	err = len;
2301
2302	out_freeiov:
2303	kfree(iov);
2304	return err;
2305	}
2306
2307	/*
2308	* BSD recvmsg interface
2309	*/
2310
2311	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2312	bool forbid_cmsg_compat)
2313	{
2314	int fput_needed, err;
2315	struct msghdr msg_sys;
2316	struct socket *sock;
2317
2318	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2319	return -EINVAL;
2320
2321	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2322	if (!sock)
2323	goto out;
2324
2325	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2326
2327	fput_light(sock->file, fput_needed);
2328	out:
2329	return err;
2330	}
2331
2332	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2333	unsigned int, flags)
2334	{
2335	return __sys_recvmsg(fd, msg, flags, true);
2336	}
2337
2338	/*
2339	* Linux recvmmsg interface
2340	*/
2341
2342	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2343	unsigned int flags, struct timespec *timeout)
2344	{
2345	int fput_needed, err, datagrams;
2346	struct socket *sock;
2347	struct mmsghdr __user *entry;
2348	struct compat_mmsghdr __user *compat_entry;
2349	struct msghdr msg_sys;
2350	struct timespec64 end_time;
2351	struct timespec64 timeout64;
2352
2353	if (timeout &&
2354	poll_select_set_timeout(&end_time, timeout->tv_sec,
2355	timeout->tv_nsec))
2356	return -EINVAL;
2357
2358	datagrams = 0;
2359
2360	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2361	if (!sock)
2362	return err;
2363
2364	if (likely(!(flags & MSG_ERRQUEUE))) {
2365	err = sock_error(sock->sk);
2366	if (err) {
2367	datagrams = err;
2368	goto out_put;
2369	}
2370	}
2371
2372	entry = mmsg;
2373	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2374
2375	while (datagrams < vlen) {
2376	/*
2377	* No need to ask LSM for more than the first datagram.
2378	*/
2379	if (MSG_CMSG_COMPAT & flags) {
2380	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2381	&msg_sys, flags & ~MSG_WAITFORONE,
2382	datagrams);
2383	if (err < 0)
2384	break;
2385	err = __put_user(err, &compat_entry->msg_len);
2386	++compat_entry;
2387	} else {
2388	err = ___sys_recvmsg(sock,
2389	(struct user_msghdr __user *)entry,
2390	&msg_sys, flags & ~MSG_WAITFORONE,
2391	datagrams);
2392	if (err < 0)
2393	break;
2394	err = put_user(err, &entry->msg_len);
2395	++entry;
2396	}
2397
2398	if (err)
2399	break;
2400	++datagrams;
2401
2402	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2403	if (flags & MSG_WAITFORONE)
2404	flags \|= MSG_DONTWAIT;
2405
2406	if (timeout) {
2407	ktime_get_ts64(&timeout64);
2408	*timeout = timespec64_to_timespec(
2409	timespec64_sub(end_time, timeout64));
2410	if (timeout->tv_sec < 0) {
2411	timeout->tv_sec = timeout->tv_nsec = 0;
2412	break;
2413	}
2414
2415	/* Timeout, return less than vlen datagrams */
2416	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2417	break;
2418	}
2419
2420	/* Out of band data, return right away */
2421	if (msg_sys.msg_flags & MSG_OOB)
2422	break;
2423	cond_resched();
2424	}
2425
2426	if (err == 0)
2427	goto out_put;
2428
2429	if (datagrams == 0) {
2430	datagrams = err;
2431	goto out_put;
2432	}
2433
2434	/*
2435	* We may return less entries than requested (vlen) if the
2436	* sock is non block and there aren't enough datagrams...
2437	*/
2438	if (err != -EAGAIN) {
2439	/*
2440	* ... or if recvmsg returns an error after we
2441	* received some datagrams, where we record the
2442	* error to return on the next call or if the
2443	* app asks about it using getsockopt(SO_ERROR).
2444	*/
2445	sock->sk->sk_err = -err;
2446	}
2447	out_put:
2448	fput_light(sock->file, fput_needed);
2449
2450	return datagrams;
2451	}
2452
2453	static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2454	unsigned int vlen, unsigned int flags,
2455	struct timespec __user *timeout)
2456	{
2457	int datagrams;
2458	struct timespec timeout_sys;
2459
2460	if (flags & MSG_CMSG_COMPAT)
2461	return -EINVAL;
2462
2463	if (!timeout)
2464	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2465
2466	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2467	return -EFAULT;
2468
2469	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2470
2471	if (datagrams > 0 &&
2472	copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2473	datagrams = -EFAULT;
2474
2475	return datagrams;
2476	}
2477
2478	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2479	unsigned int, vlen, unsigned int, flags,
2480	struct timespec __user *, timeout)
2481	{
2482	return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2483	}
2484
2485	#ifdef __ARCH_WANT_SYS_SOCKETCALL
2486	/* Argument list sizes for sys_socketcall */
2487	#define AL(x) ((x) * sizeof(unsigned long))
2488	static const unsigned char nargs[21] = {
2489	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2490	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2491	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2492	AL(4), AL(5), AL(4)
2493	};
2494
2495	#undef AL
2496
2497	/*
2498	* System call vectors.
2499	*
2500	* Argument checking cleaned up. Saved 20% in size.
2501	* This function doesn't need to set the kernel lock because
2502	* it is set by the callees.
2503	*/
2504
2505	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2506	{
2507	unsigned long a[AUDITSC_ARGS];
2508	unsigned long a0, a1;
2509	int err;
2510	unsigned int len;
2511
2512	if (call < 1 \|\| call > SYS_SENDMMSG)
2513	return -EINVAL;
2514	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2515
2516	len = nargs[call];
2517	if (len > sizeof(a))
2518	return -EINVAL;
2519
2520	/* copy_from_user should be SMP safe. */
2521	if (copy_from_user(a, args, len))
2522	return -EFAULT;
2523
2524	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2525	if (err)
2526	return err;
2527
2528	a0 = a[0];
2529	a1 = a[1];
2530
2531	switch (call) {
2532	case SYS_SOCKET:
2533	err = __sys_socket(a0, a1, a[2]);
2534	break;
2535	case SYS_BIND:
2536	err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2537	break;
2538	case SYS_CONNECT:
2539	err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2540	break;
2541	case SYS_LISTEN:
2542	err = __sys_listen(a0, a1);
2543	break;
2544	case SYS_ACCEPT:
2545	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2546	(int __user *)a[2], 0);
2547	break;
2548	case SYS_GETSOCKNAME:
2549	err =
2550	__sys_getsockname(a0, (struct sockaddr __user *)a1,
2551	(int __user *)a[2]);
2552	break;
2553	case SYS_GETPEERNAME:
2554	err =
2555	__sys_getpeername(a0, (struct sockaddr __user *)a1,
2556	(int __user *)a[2]);
2557	break;
2558	case SYS_SOCKETPAIR:
2559	err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2560	break;
2561	case SYS_SEND:
2562	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2563	NULL, 0);
2564	break;
2565	case SYS_SENDTO:
2566	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2567	(struct sockaddr __user *)a[4], a[5]);
2568	break;
2569	case SYS_RECV:
2570	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2571	NULL, NULL);
2572	break;
2573	case SYS_RECVFROM:
2574	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2575	(struct sockaddr __user *)a[4],
2576	(int __user *)a[5]);
2577	break;
2578	case SYS_SHUTDOWN:
2579	err = __sys_shutdown(a0, a1);
2580	break;
2581	case SYS_SETSOCKOPT:
2582	err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2583	a[4]);
2584	break;
2585	case SYS_GETSOCKOPT:
2586	err =
2587	__sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2588	(int __user *)a[4]);
2589	break;
2590	case SYS_SENDMSG:
2591	err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2592	a[2], true);
2593	break;
2594	case SYS_SENDMMSG:
2595	err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2596	a[3], true);
2597	break;
2598	case SYS_RECVMSG:
2599	err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2600	a[2], true);
2601	break;
2602	case SYS_RECVMMSG:
2603	err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2604	a[3], (struct timespec __user *)a[4]);
2605	break;
2606	case SYS_ACCEPT4:
2607	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2608	(int __user *)a[2], a[3]);
2609	break;
2610	default:
2611	err = -EINVAL;
2612	break;
2613	}
2614	return err;
2615	}
2616
2617	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2618
2619	/**
2620	* sock_register - add a socket protocol handler
2621	* @ops: description of protocol
2622	*
2623	* This function is called by a protocol handler that wants to
2624	* advertise its address family, and have it linked into the
2625	* socket interface. The value ops->family corresponds to the
2626	* socket system call protocol family.
2627	*/
2628	int sock_register(const struct net_proto_family *ops)
2629	{
2630	int err;
2631
2632	if (ops->family >= NPROTO) {
2633	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2634	return -ENOBUFS;
2635	}
2636
2637	spin_lock(&net_family_lock);
2638	if (rcu_dereference_protected(net_families[ops->family],
2639	lockdep_is_held(&net_family_lock)))
2640	err = -EEXIST;
2641	else {
2642	rcu_assign_pointer(net_families[ops->family], ops);
2643	err = 0;
2644	}
2645	spin_unlock(&net_family_lock);
2646
2647	pr_info("NET: Registered protocol family %d\n", ops->family);
2648	return err;
2649	}
2650	EXPORT_SYMBOL(sock_register);
2651
2652	/**
2653	* sock_unregister - remove a protocol handler
2654	* @family: protocol family to remove
2655	*
2656	* This function is called by a protocol handler that wants to
2657	* remove its address family, and have it unlinked from the
2658	* new socket creation.
2659	*
2660	* If protocol handler is a module, then it can use module reference
2661	* counts to protect against new references. If protocol handler is not
2662	* a module then it needs to provide its own protection in
2663	* the ops->create routine.
2664	*/
2665	void sock_unregister(int family)
2666	{
2667	BUG_ON(family < 0 \|\| family >= NPROTO);
2668
2669	spin_lock(&net_family_lock);
2670	RCU_INIT_POINTER(net_families[family], NULL);
2671	spin_unlock(&net_family_lock);
2672
2673	synchronize_rcu();
2674
2675	pr_info("NET: Unregistered protocol family %d\n", family);
2676	}
2677	EXPORT_SYMBOL(sock_unregister);
2678
2679	bool sock_is_registered(int family)
2680	{
2681	return family < NPROTO && rcu_access_pointer(net_families[family]);
2682	}
2683
2684	static int __init sock_init(void)
2685	{
2686	int err;
2687	/*
2688	* Initialize the network sysctl infrastructure.
2689	*/
2690	err = net_sysctl_init();
2691	if (err)
2692	goto out;
2693
2694	/*
2695	* Initialize skbuff SLAB cache
2696	*/
2697	skb_init();
2698
2699	/*
2700	* Initialize the protocols module.
2701	*/
2702
2703	init_inodecache();
2704
2705	err = register_filesystem(&sock_fs_type);
2706	if (err)
2707	goto out_fs;
2708	sock_mnt = kern_mount(&sock_fs_type);
2709	if (IS_ERR(sock_mnt)) {
2710	err = PTR_ERR(sock_mnt);
2711	goto out_mount;
2712	}
2713
2714	/* The real protocol initialization is performed in later initcalls.
2715	*/
2716
2717	#ifdef CONFIG_NETFILTER
2718	err = netfilter_init();
2719	if (err)
2720	goto out;
2721	#endif
2722
2723	ptp_classifier_init();
2724
2725	out:
2726	return err;
2727
2728	out_mount:
2729	unregister_filesystem(&sock_fs_type);
2730	out_fs:
2731	goto out;
2732	}
2733
2734	core_initcall(sock_init); /* early initcall */
2735
2736	#ifdef CONFIG_PROC_FS
2737	void socket_seq_show(struct seq_file *seq)
2738	{
2739	seq_printf(seq, "sockets: used %d\n",
2740	sock_inuse_get(seq->private));
2741	}
2742	#endif /* CONFIG_PROC_FS */
2743
2744	#ifdef CONFIG_COMPAT
2745	static int do_siocgstamp(struct net net, struct socket sock,
2746	unsigned int cmd, void __user *up)
2747	{
2748	mm_segment_t old_fs = get_fs();
2749	struct timeval ktv;
2750	int err;
2751
2752	set_fs(KERNEL_DS);
2753	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2754	set_fs(old_fs);
2755	if (!err)
2756	err = compat_put_timeval(&ktv, up);
2757
2758	return err;
2759	}
2760
2761	static int do_siocgstampns(struct net net, struct socket sock,
2762	unsigned int cmd, void __user *up)
2763	{
2764	mm_segment_t old_fs = get_fs();
2765	struct timespec kts;
2766	int err;
2767
2768	set_fs(KERNEL_DS);
2769	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2770	set_fs(old_fs);
2771	if (!err)
2772	err = compat_put_timespec(&kts, up);
2773
2774	return err;
2775	}
2776
2777	static int compat_dev_ifconf(struct net net, struct compat_ifconf __user uifc32)
2778	{
2779	struct compat_ifconf ifc32;
2780	struct ifconf ifc;
2781	int err;
2782
2783	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2784	return -EFAULT;
2785
2786	ifc.ifc_len = ifc32.ifc_len;
2787	ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2788
2789	rtnl_lock();
2790	err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2791	rtnl_unlock();
2792	if (err)
2793	return err;
2794
2795	ifc32.ifc_len = ifc.ifc_len;
2796	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2797	return -EFAULT;
2798
2799	return 0;
2800	}
2801
2802	static int ethtool_ioctl(struct net net, struct compat_ifreq __user ifr32)
2803	{
2804	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2805	bool convert_in = false, convert_out = false;
2806	size_t buf_size = 0;
2807	struct ethtool_rxnfc __user *rxnfc = NULL;
2808	struct ifreq ifr;
2809	u32 rule_cnt = 0, actual_rule_cnt;
2810	u32 ethcmd;
2811	u32 data;
2812	int ret;
2813
2814	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2815	return -EFAULT;
2816
2817	compat_rxnfc = compat_ptr(data);
2818
2819	if (get_user(ethcmd, &compat_rxnfc->cmd))
2820	return -EFAULT;
2821
2822	/* Most ethtool structures are defined without padding.
2823	* Unfortunately struct ethtool_rxnfc is an exception.
2824	*/
2825	switch (ethcmd) {
2826	default:
2827	break;
2828	case ETHTOOL_GRXCLSRLALL:
2829	/* Buffer size is variable */
2830	if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2831	return -EFAULT;
2832	if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2833	return -ENOMEM;
2834	buf_size += rule_cnt * sizeof(u32);
2835	/* fall through */
2836	case ETHTOOL_GRXRINGS:
2837	case ETHTOOL_GRXCLSRLCNT:
2838	case ETHTOOL_GRXCLSRULE:
2839	case ETHTOOL_SRXCLSRLINS:
2840	convert_out = true;
2841	/* fall through */
2842	case ETHTOOL_SRXCLSRLDEL:
2843	buf_size += sizeof(struct ethtool_rxnfc);
2844	convert_in = true;
2845	rxnfc = compat_alloc_user_space(buf_size);
2846	break;
2847	}
2848
2849	if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2850	return -EFAULT;
2851
2852	ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2853
2854	if (convert_in) {
2855	/* We expect there to be holes between fs.m_ext and
2856	* fs.ring_cookie and at the end of fs, but nowhere else.
2857	*/
2858	BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2859	sizeof(compat_rxnfc->fs.m_ext) !=
2860	offsetof(struct ethtool_rxnfc, fs.m_ext) +
2861	sizeof(rxnfc->fs.m_ext));
2862	BUILD_BUG_ON(
2863	offsetof(struct compat_ethtool_rxnfc, fs.location) -
2864	offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2865	offsetof(struct ethtool_rxnfc, fs.location) -
2866	offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2867
2868	if (copy_in_user(rxnfc, compat_rxnfc,
2869	(void __user *)(&rxnfc->fs.m_ext + 1) -
2870	(void __user *)rxnfc) \|\|
2871	copy_in_user(&rxnfc->fs.ring_cookie,
2872	&compat_rxnfc->fs.ring_cookie,
2873	(void __user *)(&rxnfc->fs.location + 1) -
2874	(void __user *)&rxnfc->fs.ring_cookie) \|\|
2875	copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2876	sizeof(rxnfc->rule_cnt)))
2877	return -EFAULT;
2878	}
2879
2880	ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2881	if (ret)
2882	return ret;
2883
2884	if (convert_out) {
2885	if (copy_in_user(compat_rxnfc, rxnfc,
2886	(const void __user *)(&rxnfc->fs.m_ext + 1) -
2887	(const void __user *)rxnfc) \|\|
2888	copy_in_user(&compat_rxnfc->fs.ring_cookie,
2889	&rxnfc->fs.ring_cookie,
2890	(const void __user *)(&rxnfc->fs.location + 1) -
2891	(const void __user *)&rxnfc->fs.ring_cookie) \|\|
2892	copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2893	sizeof(rxnfc->rule_cnt)))
2894	return -EFAULT;
2895
2896	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2897	/* As an optimisation, we only copy the actual
2898	* number of rules that the underlying
2899	* function returned. Since Mallory might
2900	* change the rule count in user memory, we
2901	* check that it is less than the rule count
2902	* originally given (as the user buffer size),
2903	* which has been range-checked.
2904	*/
2905	if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2906	return -EFAULT;
2907	if (actual_rule_cnt < rule_cnt)
2908	rule_cnt = actual_rule_cnt;
2909	if (copy_in_user(&compat_rxnfc->rule_locs[0],
2910	&rxnfc->rule_locs[0],
2911	rule_cnt * sizeof(u32)))
2912	return -EFAULT;
2913	}
2914	}
2915
2916	return 0;
2917	}
2918
2919	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
2920	{
2921	compat_uptr_t uptr32;
2922	struct ifreq ifr;
2923	void __user *saved;
2924	int err;
2925
2926	if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2927	return -EFAULT;
2928
2929	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2930	return -EFAULT;
2931
2932	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2933	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2934
2935	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2936	if (!err) {
2937	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2938	if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2939	err = -EFAULT;
2940	}
2941	return err;
2942	}
2943
2944	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2945	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2946	struct compat_ifreq __user *u_ifreq32)
2947	{
2948	struct ifreq ifreq;
2949	u32 data32;
2950
2951	if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2952	return -EFAULT;
2953	if (get_user(data32, &u_ifreq32->ifr_data))
2954	return -EFAULT;
2955	ifreq.ifr_data = compat_ptr(data32);
2956
2957	return dev_ioctl(net, cmd, &ifreq, NULL);
2958	}
2959
2960	static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2961	struct compat_ifreq __user *uifr32)
2962	{
2963	struct ifreq ifr;
2964	struct compat_ifmap __user *uifmap32;
2965	int err;
2966
2967	uifmap32 = &uifr32->ifr_ifru.ifru_map;
2968	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2969	err \|= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2970	err \|= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2971	err \|= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2972	err \|= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2973	err \|= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2974	err \|= get_user(ifr.ifr_map.port, &uifmap32->port);
2975	if (err)
2976	return -EFAULT;
2977
2978	err = dev_ioctl(net, cmd, &ifr, NULL);
2979
2980	if (cmd == SIOCGIFMAP && !err) {
2981	err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2982	err \|= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2983	err \|= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2984	err \|= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2985	err \|= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2986	err \|= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2987	err \|= put_user(ifr.ifr_map.port, &uifmap32->port);
2988	if (err)
2989	err = -EFAULT;
2990	}
2991	return err;
2992	}
2993
2994	struct rtentry32 {
2995	u32 rt_pad1;
2996	struct sockaddr rt_dst; /* target address */
2997	struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2998	struct sockaddr rt_genmask; /* target network mask (IP) */
2999	unsigned short rt_flags;
3000	short rt_pad2;
3001	u32 rt_pad3;
3002	unsigned char rt_tos;
3003	unsigned char rt_class;
3004	short rt_pad4;
3005	short rt_metric; /* +1 for binary compatibility! */
3006	/* char * / u32 rt_dev; / forcing the device at add */
3007	u32 rt_mtu; /* per route MTU/Window */
3008	u32 rt_window; /* Window clamping */
3009	unsigned short rt_irtt; /* Initial RTT */
3010	};
3011
3012	struct in6_rtmsg32 {
3013	struct in6_addr rtmsg_dst;
3014	struct in6_addr rtmsg_src;
3015	struct in6_addr rtmsg_gateway;
3016	u32 rtmsg_type;
3017	u16 rtmsg_dst_len;
3018	u16 rtmsg_src_len;
3019	u32 rtmsg_metric;
3020	u32 rtmsg_info;
3021	u32 rtmsg_flags;
3022	s32 rtmsg_ifindex;
3023	};
3024
3025	static int routing_ioctl(struct net net, struct socket sock,
3026	unsigned int cmd, void __user *argp)
3027	{
3028	int ret;
3029	void *r = NULL;
3030	struct in6_rtmsg r6;
3031	struct rtentry r4;
3032	char devname[16];
3033	u32 rtdev;
3034	mm_segment_t old_fs = get_fs();
3035
3036	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3037	struct in6_rtmsg32 __user *ur6 = argp;
3038	ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3039	3 * sizeof(struct in6_addr));
3040	ret \|= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3041	ret \|= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3042	ret \|= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3043	ret \|= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3044	ret \|= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3045	ret \|= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3046	ret \|= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3047
3048	r = (void *) &r6;
3049	} else { /* ipv4 */
3050	struct rtentry32 __user *ur4 = argp;
3051	ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3052	3 * sizeof(struct sockaddr));
3053	ret \|= get_user(r4.rt_flags, &(ur4->rt_flags));
3054	ret \|= get_user(r4.rt_metric, &(ur4->rt_metric));
3055	ret \|= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3056	ret \|= get_user(r4.rt_window, &(ur4->rt_window));
3057	ret \|= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3058	ret \|= get_user(rtdev, &(ur4->rt_dev));
3059	if (rtdev) {
3060	ret \|= copy_from_user(devname, compat_ptr(rtdev), 15);
3061	r4.rt_dev = (char __user __force *)devname;
3062	devname[15] = 0;
3063	} else
3064	r4.rt_dev = NULL;
3065
3066	r = (void *) &r4;
3067	}
3068
3069	if (ret) {
3070	ret = -EFAULT;
3071	goto out;
3072	}
3073
3074	set_fs(KERNEL_DS);
3075	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3076	set_fs(old_fs);
3077
3078	out:
3079	return ret;
3080	}
3081
3082	/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3083	* for some operations; this forces use of the newer bridge-utils that
3084	* use compatible ioctls
3085	*/
3086	static int old_bridge_ioctl(compat_ulong_t __user *argp)
3087	{
3088	compat_ulong_t tmp;
3089
3090	if (get_user(tmp, argp))
3091	return -EFAULT;
3092	if (tmp == BRCTL_GET_VERSION)
3093	return BRCTL_VERSION + 1;
3094	return -EINVAL;
3095	}
3096
3097	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3098	unsigned int cmd, unsigned long arg)
3099	{
3100	void __user *argp = compat_ptr(arg);
3101	struct sock *sk = sock->sk;
3102	struct net *net = sock_net(sk);
3103
3104	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3105	return compat_ifr_data_ioctl(net, cmd, argp);
3106
3107	switch (cmd) {
3108	case SIOCSIFBR:
3109	case SIOCGIFBR:
3110	return old_bridge_ioctl(argp);
3111	case SIOCGIFCONF:
3112	return compat_dev_ifconf(net, argp);
3113	case SIOCETHTOOL:
3114	return ethtool_ioctl(net, argp);
3115	case SIOCWANDEV:
3116	return compat_siocwandev(net, argp);
3117	case SIOCGIFMAP:
3118	case SIOCSIFMAP:
3119	return compat_sioc_ifmap(net, cmd, argp);
3120	case SIOCADDRT:
3121	case SIOCDELRT:
3122	return routing_ioctl(net, sock, cmd, argp);
3123	case SIOCGSTAMP:
3124	return do_siocgstamp(net, sock, cmd, argp);
3125	case SIOCGSTAMPNS:
3126	return do_siocgstampns(net, sock, cmd, argp);
3127	case SIOCBONDSLAVEINFOQUERY:
3128	case SIOCBONDINFOQUERY:
3129	case SIOCSHWTSTAMP:
3130	case SIOCGHWTSTAMP:
3131	return compat_ifr_data_ioctl(net, cmd, argp);
3132
3133	case FIOSETOWN:
3134	case SIOCSPGRP:
3135	case FIOGETOWN:
3136	case SIOCGPGRP:
3137	case SIOCBRADDBR:
3138	case SIOCBRDELBR:
3139	case SIOCGIFVLAN:
3140	case SIOCSIFVLAN:
3141	case SIOCADDDLCI:
3142	case SIOCDELDLCI:
3143	case SIOCGSKNS:
3144	return sock_ioctl(file, cmd, arg);
3145
3146	case SIOCGIFFLAGS:
3147	case SIOCSIFFLAGS:
3148	case SIOCGIFMETRIC:
3149	case SIOCSIFMETRIC:
3150	case SIOCGIFMTU:
3151	case SIOCSIFMTU:
3152	case SIOCGIFMEM:
3153	case SIOCSIFMEM:
3154	case SIOCGIFHWADDR:
3155	case SIOCSIFHWADDR:
3156	case SIOCADDMULTI:
3157	case SIOCDELMULTI:
3158	case SIOCGIFINDEX:
3159	case SIOCGIFADDR:
3160	case SIOCSIFADDR:
3161	case SIOCSIFHWBROADCAST:
3162	case SIOCDIFADDR:
3163	case SIOCGIFBRDADDR:
3164	case SIOCSIFBRDADDR:
3165	case SIOCGIFDSTADDR:
3166	case SIOCSIFDSTADDR:
3167	case SIOCGIFNETMASK:
3168	case SIOCSIFNETMASK:
3169	case SIOCSIFPFLAGS:
3170	case SIOCGIFPFLAGS:
3171	case SIOCGIFTXQLEN:
3172	case SIOCSIFTXQLEN:
3173	case SIOCBRADDIF:
3174	case SIOCBRDELIF:
3175	case SIOCSIFNAME:
3176	case SIOCGMIIPHY:
3177	case SIOCGMIIREG:
3178	case SIOCSMIIREG:
3179	case SIOCSARP:
3180	case SIOCGARP:
3181	case SIOCDARP:
3182	case SIOCATMARK:
3183	case SIOCBONDENSLAVE:
3184	case SIOCBONDRELEASE:
3185	case SIOCBONDSETHWADDR:
3186	case SIOCBONDCHANGEACTIVE:
3187	case SIOCGIFNAME:
3188	return sock_do_ioctl(net, sock, cmd, arg);
3189	}
3190
3191	return -ENOIOCTLCMD;
3192	}
3193
3194	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3195	unsigned long arg)
3196	{
3197	struct socket *sock = file->private_data;
3198	int ret = -ENOIOCTLCMD;
3199	struct sock *sk;
3200	struct net *net;
3201
3202	sk = sock->sk;
3203	net = sock_net(sk);
3204
3205	if (sock->ops->compat_ioctl)
3206	ret = sock->ops->compat_ioctl(sock, cmd, arg);
3207
3208	if (ret == -ENOIOCTLCMD &&
3209	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3210	ret = compat_wext_handle_ioctl(net, cmd, arg);
3211
3212	if (ret == -ENOIOCTLCMD)
3213	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3214
3215	return ret;
3216	}
3217	#endif
3218
3219	int kernel_bind(struct socket sock, struct sockaddr addr, int addrlen)
3220	{
3221	return sock->ops->bind(sock, addr, addrlen);
3222	}
3223	EXPORT_SYMBOL(kernel_bind);
3224
3225	int kernel_listen(struct socket *sock, int backlog)
3226	{
3227	return sock->ops->listen(sock, backlog);
3228	}
3229	EXPORT_SYMBOL(kernel_listen);
3230
3231	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3232	{
3233	struct sock *sk = sock->sk;
3234	int err;
3235
3236	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3237	newsock);
3238	if (err < 0)
3239	goto done;
3240
3241	err = sock->ops->accept(sock, *newsock, flags, true);
3242	if (err < 0) {
3243	sock_release(*newsock);
3244	*newsock = NULL;
3245	goto done;
3246	}
3247
3248	(*newsock)->ops = sock->ops;
3249	__module_get((*newsock)->ops->owner);
3250
3251	done:
3252	return err;
3253	}
3254	EXPORT_SYMBOL(kernel_accept);
3255
3256	int kernel_connect(struct socket sock, struct sockaddr addr, int addrlen,
3257	int flags)
3258	{
3259	return sock->ops->connect(sock, addr, addrlen, flags);
3260	}
3261	EXPORT_SYMBOL(kernel_connect);
3262
3263	int kernel_getsockname(struct socket sock, struct sockaddr addr)
3264	{
3265	return sock->ops->getname(sock, addr, 0);
3266	}
3267	EXPORT_SYMBOL(kernel_getsockname);
3268
3269	int kernel_getpeername(struct socket sock, struct sockaddr addr)
3270	{
3271	return sock->ops->getname(sock, addr, 1);
3272	}
3273	EXPORT_SYMBOL(kernel_getpeername);
3274
3275	int kernel_getsockopt(struct socket *sock, int level, int optname,
3276	char optval, int optlen)
3277	{
3278	mm_segment_t oldfs = get_fs();
3279	char __user *uoptval;
3280	int __user *uoptlen;
3281	int err;
3282
3283	uoptval = (char __user __force *) optval;
3284	uoptlen = (int __user __force *) optlen;
3285
3286	set_fs(KERNEL_DS);
3287	if (level == SOL_SOCKET)
3288	err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3289	else
3290	err = sock->ops->getsockopt(sock, level, optname, uoptval,
3291	uoptlen);
3292	set_fs(oldfs);
3293	return err;
3294	}
3295	EXPORT_SYMBOL(kernel_getsockopt);
3296
3297	int kernel_setsockopt(struct socket *sock, int level, int optname,
3298	char *optval, unsigned int optlen)
3299	{
3300	mm_segment_t oldfs = get_fs();
3301	char __user *uoptval;
3302	int err;
3303
3304	uoptval = (char __user __force *) optval;
3305
3306	set_fs(KERNEL_DS);
3307	if (level == SOL_SOCKET)
3308	err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3309	else
3310	err = sock->ops->setsockopt(sock, level, optname, uoptval,
3311	optlen);
3312	set_fs(oldfs);
3313	return err;
3314	}
3315	EXPORT_SYMBOL(kernel_setsockopt);
3316
3317	int kernel_sendpage(struct socket sock, struct page page, int offset,
3318	size_t size, int flags)
3319	{
3320	if (sock->ops->sendpage)
3321	return sock->ops->sendpage(sock, page, offset, size, flags);
3322
3323	return sock_no_sendpage(sock, page, offset, size, flags);
3324	}
3325	EXPORT_SYMBOL(kernel_sendpage);
3326
3327	int kernel_sendpage_locked(struct sock sk, struct page page, int offset,
3328	size_t size, int flags)
3329	{
3330	struct socket *sock = sk->sk_socket;
3331
3332	if (sock->ops->sendpage_locked)
3333	return sock->ops->sendpage_locked(sk, page, offset, size,
3334	flags);
3335
3336	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3337	}
3338	EXPORT_SYMBOL(kernel_sendpage_locked);
3339
3340	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3341	{
3342	return sock->ops->shutdown(sock, how);
3343	}
3344	EXPORT_SYMBOL(kernel_sock_shutdown);
3345
3346	/* This routine returns the IP overhead imposed by a socket i.e.
3347	* the length of the underlying IP header, depending on whether
3348	* this is an IPv4 or IPv6 socket and the length from IP options turned
3349	* on at the socket. Assumes that the caller has a lock on the socket.
3350	*/
3351	u32 kernel_sock_ip_overhead(struct sock *sk)
3352	{
3353	struct inet_sock *inet;
3354	struct ip_options_rcu *opt;
3355	u32 overhead = 0;
3356	#if IS_ENABLED(CONFIG_IPV6)
3357	struct ipv6_pinfo *np;
3358	struct ipv6_txoptions *optv6 = NULL;
3359	#endif /* IS_ENABLED(CONFIG_IPV6) */
3360
3361	if (!sk)
3362	return overhead;
3363
3364	switch (sk->sk_family) {
3365	case AF_INET:
3366	inet = inet_sk(sk);
3367	overhead += sizeof(struct iphdr);
3368	opt = rcu_dereference_protected(inet->inet_opt,
3369	sock_owned_by_user(sk));
3370	if (opt)
3371	overhead += opt->opt.optlen;
3372	return overhead;
3373	#if IS_ENABLED(CONFIG_IPV6)
3374	case AF_INET6:
3375	np = inet6_sk(sk);
3376	overhead += sizeof(struct ipv6hdr);
3377	if (np)
3378	optv6 = rcu_dereference_protected(np->opt,
3379	sock_owned_by_user(sk));
3380	if (optv6)
3381	overhead += (optv6->opt_flen + optv6->opt_nflen);
3382	return overhead;
3383	#endif /* IS_ENABLED(CONFIG_IPV6) */
3384	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3385	return overhead;
3386	}
3387	}
3388	EXPORT_SYMBOL(kernel_sock_ip_overhead);