linux - net/socket.c

3390 lines · 83.3 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
93	#include <linux/uaccess.h>
94	#include <asm/unistd.h>
95
96	#include <net/compat.h>
97	#include <net/wext.h>
98	#include <net/cls_cgroup.h>
99
100	#include <net/sock.h>
101	#include <linux/netfilter.h>
102
103	#include <linux/if_tun.h>
104	#include <linux/ipv6_route.h>
105	#include <linux/route.h>
106	#include <linux/sockios.h>
107	#include <net/busy_poll.h>
108	#include <linux/errqueue.h>
109
110	#ifdef CONFIG_NET_RX_BUSY_POLL
111	unsigned int sysctl_net_busy_read __read_mostly;
112	unsigned int sysctl_net_busy_poll __read_mostly;
113	#endif
114
115	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to);
116	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from);
117	static int sock_mmap(struct file file, struct vm_area_struct vma);
118
119	static int sock_close(struct inode inode, struct file file);
120	static __poll_t sock_poll(struct file *file,
121	struct poll_table_struct *wait);
122	static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
123	#ifdef CONFIG_COMPAT
124	static long compat_sock_ioctl(struct file *file,
125	unsigned int cmd, unsigned long arg);
126	#endif
127	static int sock_fasync(int fd, struct file *filp, int on);
128	static ssize_t sock_sendpage(struct file file, struct page page,
129	int offset, size_t size, loff_t *ppos, int more);
130	static ssize_t sock_splice_read(struct file file, loff_t ppos,
131	struct pipe_inode_info *pipe, size_t len,
132	unsigned int flags);
133
134	/*
135	* Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
136	* in the operation structures but are done directly via the socketcall() multiplexor.
137	*/
138
139	static const struct file_operations socket_file_ops = {
140	.owner = THIS_MODULE,
141	.llseek = no_llseek,
142	.read_iter = sock_read_iter,
143	.write_iter = sock_write_iter,
144	.poll = sock_poll,
145	.unlocked_ioctl = sock_ioctl,
146	#ifdef CONFIG_COMPAT
147	.compat_ioctl = compat_sock_ioctl,
148	#endif
149	.mmap = sock_mmap,
150	.release = sock_close,
151	.fasync = sock_fasync,
152	.sendpage = sock_sendpage,
153	.splice_write = generic_splice_sendpage,
154	.splice_read = sock_splice_read,
155	};
156
157	/*
158	* The protocol list. Each protocol is registered in here.
159	*/
160
161	static DEFINE_SPINLOCK(net_family_lock);
162	static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
163
164	/*
165	* Support routines.
166	* Move socket addresses back and forth across the kernel/user
167	* divide and look after the messy bits.
168	*/
169
170	/**
171	* move_addr_to_kernel - copy a socket address into kernel space
172	* @uaddr: Address in user space
173	* @kaddr: Address in kernel space
174	* @ulen: Length in user space
175	*
176	* The address is copied into kernel space. If the provided address is
177	* too long an error code of -EINVAL is returned. If the copy gives
178	* invalid addresses -EFAULT is returned. On a success 0 is returned.
179	*/
180
181	int move_addr_to_kernel(void __user uaddr, int ulen, struct sockaddr_storage kaddr)
182	{
183	if (ulen < 0 \|\| ulen > sizeof(struct sockaddr_storage))
184	return -EINVAL;
185	if (ulen == 0)
186	return 0;
187	if (copy_from_user(kaddr, uaddr, ulen))
188	return -EFAULT;
189	return audit_sockaddr(ulen, kaddr);
190	}
191
192	/**
193	* move_addr_to_user - copy an address to user space
194	* @kaddr: kernel space address
195	* @klen: length of address in kernel
196	* @uaddr: user space address
197	* @ulen: pointer to user length field
198	*
199	* The value pointed to by ulen on entry is the buffer length available.
200	* This is overwritten with the buffer space used. -EINVAL is returned
201	* if an overlong buffer is specified or a negative buffer size. -EFAULT
202	* is returned if either the buffer or the length field are not
203	* accessible.
204	* After copying the data up to the limit the user specifies, the true
205	* length of the data is written over the length limit the user
206	* specified. Zero is returned for a success.
207	*/
208
209	static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
210	void __user uaddr, int __user ulen)
211	{
212	int err;
213	int len;
214
215	BUG_ON(klen > sizeof(struct sockaddr_storage));
216	err = get_user(len, ulen);
217	if (err)
218	return err;
219	if (len > klen)
220	len = klen;
221	if (len < 0)
222	return -EINVAL;
223	if (len) {
224	if (audit_sockaddr(klen, kaddr))
225	return -ENOMEM;
226	if (copy_to_user(uaddr, kaddr, len))
227	return -EFAULT;
228	}
229	/*
230	* "fromlen shall refer to the value before truncation.."
231	* 1003.1g
232	*/
233	return __put_user(klen, ulen);
234	}
235
236	static struct kmem_cache *sock_inode_cachep __ro_after_init;
237
238	static struct inode sock_alloc_inode(struct super_block sb)
239	{
240	struct socket_alloc *ei;
241	struct socket_wq *wq;
242
243	ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
244	if (!ei)
245	return NULL;
246	wq = kmalloc(sizeof(*wq), GFP_KERNEL);
247	if (!wq) {
248	kmem_cache_free(sock_inode_cachep, ei);
249	return NULL;
250	}
251	init_waitqueue_head(&wq->wait);
252	wq->fasync_list = NULL;
253	wq->flags = 0;
254	RCU_INIT_POINTER(ei->socket.wq, wq);
255
256	ei->socket.state = SS_UNCONNECTED;
257	ei->socket.flags = 0;
258	ei->socket.ops = NULL;
259	ei->socket.sk = NULL;
260	ei->socket.file = NULL;
261
262	return &ei->vfs_inode;
263	}
264
265	static void sock_destroy_inode(struct inode *inode)
266	{
267	struct socket_alloc *ei;
268	struct socket_wq *wq;
269
270	ei = container_of(inode, struct socket_alloc, vfs_inode);
271	wq = rcu_dereference_protected(ei->socket.wq, 1);
272	kfree_rcu(wq, rcu);
273	kmem_cache_free(sock_inode_cachep, ei);
274	}
275
276	static void init_once(void *foo)
277	{
278	struct socket_alloc ei = (struct socket_alloc )foo;
279
280	inode_init_once(&ei->vfs_inode);
281	}
282
283	static void init_inodecache(void)
284	{
285	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
286	sizeof(struct socket_alloc),
287	0,
288	(SLAB_HWCACHE_ALIGN \|
289	SLAB_RECLAIM_ACCOUNT \|
290	SLAB_MEM_SPREAD \| SLAB_ACCOUNT),
291	init_once);
292	BUG_ON(sock_inode_cachep == NULL);
293	}
294
295	static const struct super_operations sockfs_ops = {
296	.alloc_inode = sock_alloc_inode,
297	.destroy_inode = sock_destroy_inode,
298	.statfs = simple_statfs,
299	};
300
301	/*
302	* sockfs_dname() is called from d_path().
303	*/
304	static char sockfs_dname(struct dentry dentry, char *buffer, int buflen)
305	{
306	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
307	d_inode(dentry)->i_ino);
308	}
309
310	static const struct dentry_operations sockfs_dentry_operations = {
311	.d_dname = sockfs_dname,
312	};
313
314	static int sockfs_xattr_get(const struct xattr_handler *handler,
315	struct dentry dentry, struct inode inode,
316	const char suffix, void value, size_t size)
317	{
318	if (value) {
319	if (dentry->d_name.len + 1 > size)
320	return -ERANGE;
321	memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
322	}
323	return dentry->d_name.len + 1;
324	}
325
326	#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
327	#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
328	#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
329
330	static const struct xattr_handler sockfs_xattr_handler = {
331	.name = XATTR_NAME_SOCKPROTONAME,
332	.get = sockfs_xattr_get,
333	};
334
335	static int sockfs_security_xattr_set(const struct xattr_handler *handler,
336	struct dentry dentry, struct inode inode,
337	const char suffix, const void value,
338	size_t size, int flags)
339	{
340	/* Handled by LSM. */
341	return -EAGAIN;
342	}
343
344	static const struct xattr_handler sockfs_security_xattr_handler = {
345	.prefix = XATTR_SECURITY_PREFIX,
346	.set = sockfs_security_xattr_set,
347	};
348
349	static const struct xattr_handler *sockfs_xattr_handlers[] = {
350	&sockfs_xattr_handler,
351	&sockfs_security_xattr_handler,
352	NULL
353	};
354
355	static struct dentry sockfs_mount(struct file_system_type fs_type,
356	int flags, const char dev_name, void data)
357	{
358	return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
359	sockfs_xattr_handlers,
360	&sockfs_dentry_operations, SOCKFS_MAGIC);
361	}
362
363	static struct vfsmount *sock_mnt __read_mostly;
364
365	static struct file_system_type sock_fs_type = {
366	.name = "sockfs",
367	.mount = sockfs_mount,
368	.kill_sb = kill_anon_super,
369	};
370
371	/*
372	* Obtains the first available file descriptor and sets it up for use.
373	*
374	* These functions create file structures and maps them to fd space
375	* of the current process. On success it returns file descriptor
376	* and file struct implicitly stored in sock->file.
377	* Note that another thread may close file descriptor before we return
378	* from this function. We use the fact that now we do not refer
379	* to socket after mapping. If one day we will need it, this
380	* function will increment ref. count on file by 1.
381	*
382	* In any case returned fd MAY BE not valid!
383	* This race condition is unavoidable
384	* with shared fd spaces, we cannot solve it inside kernel,
385	* but we take care of internal coherence yet.
386	*/
387
388	struct file sock_alloc_file(struct socket sock, int flags, const char *dname)
389	{
390	struct qstr name = { .name = "" };
391	struct path path;
392	struct file *file;
393
394	if (dname) {
395	name.name = dname;
396	name.len = strlen(name.name);
397	} else if (sock->sk) {
398	name.name = sock->sk->sk_prot_creator->name;
399	name.len = strlen(name.name);
400	}
401	path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
402	if (unlikely(!path.dentry)) {
403	sock_release(sock);
404	return ERR_PTR(-ENOMEM);
405	}
406	path.mnt = mntget(sock_mnt);
407
408	d_instantiate(path.dentry, SOCK_INODE(sock));
409
410	file = alloc_file(&path, FMODE_READ \| FMODE_WRITE,
411	&socket_file_ops);
412	if (IS_ERR(file)) {
413	/* drop dentry, keep inode for a bit */
414	ihold(d_inode(path.dentry));
415	path_put(&path);
416	/* ... and now kill it properly */
417	sock_release(sock);
418	return file;
419	}
420
421	sock->file = file;
422	file->f_flags = O_RDWR \| (flags & O_NONBLOCK);
423	file->private_data = sock;
424	return file;
425	}
426	EXPORT_SYMBOL(sock_alloc_file);
427
428	static int sock_map_fd(struct socket *sock, int flags)
429	{
430	struct file *newfile;
431	int fd = get_unused_fd_flags(flags);
432	if (unlikely(fd < 0)) {
433	sock_release(sock);
434	return fd;
435	}
436
437	newfile = sock_alloc_file(sock, flags, NULL);
438	if (likely(!IS_ERR(newfile))) {
439	fd_install(fd, newfile);
440	return fd;
441	}
442
443	put_unused_fd(fd);
444	return PTR_ERR(newfile);
445	}
446
447	struct socket sock_from_file(struct file file, int *err)
448	{
449	if (file->f_op == &socket_file_ops)
450	return file->private_data; /* set in sock_map_fd */
451
452	*err = -ENOTSOCK;
453	return NULL;
454	}
455	EXPORT_SYMBOL(sock_from_file);
456
457	/**
458	* sockfd_lookup - Go from a file number to its socket slot
459	* @fd: file handle
460	* @err: pointer to an error code return
461	*
462	* The file handle passed in is locked and the socket it is bound
463	* to is returned. If an error occurs the err pointer is overwritten
464	* with a negative errno code and NULL is returned. The function checks
465	* for both invalid handles and passing a handle which is not a socket.
466	*
467	* On a success the socket object pointer is returned.
468	*/
469
470	struct socket sockfd_lookup(int fd, int err)
471	{
472	struct file *file;
473	struct socket *sock;
474
475	file = fget(fd);
476	if (!file) {
477	*err = -EBADF;
478	return NULL;
479	}
480
481	sock = sock_from_file(file, err);
482	if (!sock)
483	fput(file);
484	return sock;
485	}
486	EXPORT_SYMBOL(sockfd_lookup);
487
488	static struct socket sockfd_lookup_light(int fd, int err, int *fput_needed)
489	{
490	struct fd f = fdget(fd);
491	struct socket *sock;
492
493	*err = -EBADF;
494	if (f.file) {
495	sock = sock_from_file(f.file, err);
496	if (likely(sock)) {
497	*fput_needed = f.flags;
498	return sock;
499	}
500	fdput(f);
501	}
502	return NULL;
503	}
504
505	static ssize_t sockfs_listxattr(struct dentry dentry, char buffer,
506	size_t size)
507	{
508	ssize_t len;
509	ssize_t used = 0;
510
511	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
512	if (len < 0)
513	return len;
514	used += len;
515	if (buffer) {
516	if (size < used)
517	return -ERANGE;
518	buffer += len;
519	}
520
521	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
522	used += len;
523	if (buffer) {
524	if (size < used)
525	return -ERANGE;
526	memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
527	buffer += len;
528	}
529
530	return used;
531	}
532
533	static int sockfs_setattr(struct dentry dentry, struct iattr iattr)
534	{
535	int err = simple_setattr(dentry, iattr);
536
537	if (!err && (iattr->ia_valid & ATTR_UID)) {
538	struct socket *sock = SOCKET_I(d_inode(dentry));
539
540	sock->sk->sk_uid = iattr->ia_uid;
541	}
542
543	return err;
544	}
545
546	static const struct inode_operations sockfs_inode_ops = {
547	.listxattr = sockfs_listxattr,
548	.setattr = sockfs_setattr,
549	};
550
551	/**
552	* sock_alloc - allocate a socket
553	*
554	* Allocate a new inode and socket object. The two are bound together
555	* and initialised. The socket is then returned. If we are out of inodes
556	* NULL is returned.
557	*/
558
559	struct socket *sock_alloc(void)
560	{
561	struct inode *inode;
562	struct socket *sock;
563
564	inode = new_inode_pseudo(sock_mnt->mnt_sb);
565	if (!inode)
566	return NULL;
567
568	sock = SOCKET_I(inode);
569
570	inode->i_ino = get_next_ino();
571	inode->i_mode = S_IFSOCK \| S_IRWXUGO;
572	inode->i_uid = current_fsuid();
573	inode->i_gid = current_fsgid();
574	inode->i_op = &sockfs_inode_ops;
575
576	return sock;
577	}
578	EXPORT_SYMBOL(sock_alloc);
579
580	/**
581	* sock_release - close a socket
582	* @sock: socket to close
583	*
584	* The socket is released from the protocol stack if it has a release
585	* callback, and the inode is then released if the socket is bound to
586	* an inode not a file.
587	*/
588
589	void sock_release(struct socket *sock)
590	{
591	if (sock->ops) {
592	struct module *owner = sock->ops->owner;
593
594	sock->ops->release(sock);
595	sock->ops = NULL;
596	module_put(owner);
597	}
598
599	if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
600	pr_err("%s: fasync list not empty!\n", __func__);
601
602	if (!sock->file) {
603	iput(SOCK_INODE(sock));
604	return;
605	}
606	sock->file = NULL;
607	}
608	EXPORT_SYMBOL(sock_release);
609
610	void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
611	{
612	u8 flags = *tx_flags;
613
614	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
615	flags \|= SKBTX_HW_TSTAMP;
616
617	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
618	flags \|= SKBTX_SW_TSTAMP;
619
620	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
621	flags \|= SKBTX_SCHED_TSTAMP;
622
623	*tx_flags = flags;
624	}
625	EXPORT_SYMBOL(__sock_tx_timestamp);
626
627	static inline int sock_sendmsg_nosec(struct socket sock, struct msghdr msg)
628	{
629	int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
630	BUG_ON(ret == -EIOCBQUEUED);
631	return ret;
632	}
633
634	int sock_sendmsg(struct socket sock, struct msghdr msg)
635	{
636	int err = security_socket_sendmsg(sock, msg,
637	msg_data_left(msg));
638
639	return err ?: sock_sendmsg_nosec(sock, msg);
640	}
641	EXPORT_SYMBOL(sock_sendmsg);
642
643	int kernel_sendmsg(struct socket sock, struct msghdr msg,
644	struct kvec *vec, size_t num, size_t size)
645	{
646	iov_iter_kvec(&msg->msg_iter, WRITE \| ITER_KVEC, vec, num, size);
647	return sock_sendmsg(sock, msg);
648	}
649	EXPORT_SYMBOL(kernel_sendmsg);
650
651	int kernel_sendmsg_locked(struct sock sk, struct msghdr msg,
652	struct kvec *vec, size_t num, size_t size)
653	{
654	struct socket *sock = sk->sk_socket;
655
656	if (!sock->ops->sendmsg_locked)
657	return sock_no_sendmsg_locked(sk, msg, size);
658
659	iov_iter_kvec(&msg->msg_iter, WRITE \| ITER_KVEC, vec, num, size);
660
661	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
662	}
663	EXPORT_SYMBOL(kernel_sendmsg_locked);
664
665	static bool skb_is_err_queue(const struct sk_buff *skb)
666	{
667	/* pkt_type of skbs enqueued on the error queue are set to
668	* PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
669	* in recvmsg, since skbs received on a local socket will never
670	* have a pkt_type of PACKET_OUTGOING.
671	*/
672	return skb->pkt_type == PACKET_OUTGOING;
673	}
674
675	/* On transmit, software and hardware timestamps are returned independently.
676	* As the two skb clones share the hardware timestamp, which may be updated
677	* before the software timestamp is received, a hardware TX timestamp may be
678	* returned only if there is no software TX timestamp. Ignore false software
679	* timestamps, which may be made in the __sock_recv_timestamp() call when the
680	* option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
681	* hardware timestamp.
682	*/
683	static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
684	{
685	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
686	}
687
688	static void put_ts_pktinfo(struct msghdr msg, struct sk_buff skb)
689	{
690	struct scm_ts_pktinfo ts_pktinfo;
691	struct net_device *orig_dev;
692
693	if (!skb_mac_header_was_set(skb))
694	return;
695
696	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
697
698	rcu_read_lock();
699	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
700	if (orig_dev)
701	ts_pktinfo.if_index = orig_dev->ifindex;
702	rcu_read_unlock();
703
704	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
705	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
706	sizeof(ts_pktinfo), &ts_pktinfo);
707	}
708
709	/*
710	* called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
711	*/
712	void __sock_recv_timestamp(struct msghdr msg, struct sock sk,
713	struct sk_buff *skb)
714	{
715	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
716	struct scm_timestamping tss;
717	int empty = 1, false_tstamp = 0;
718	struct skb_shared_hwtstamps *shhwtstamps =
719	skb_hwtstamps(skb);
720
721	/* Race occurred between timestamp enabling and packet
722	receiving. Fill in the current time for now. */
723	if (need_software_tstamp && skb->tstamp == 0) {
724	__net_timestamp(skb);
725	false_tstamp = 1;
726	}
727
728	if (need_software_tstamp) {
729	if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
730	struct timeval tv;
731	skb_get_timestamp(skb, &tv);
732	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
733	sizeof(tv), &tv);
734	} else {
735	struct timespec ts;
736	skb_get_timestampns(skb, &ts);
737	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
738	sizeof(ts), &ts);
739	}
740	}
741
742	memset(&tss, 0, sizeof(tss));
743	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
744	ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
745	empty = 0;
746	if (shhwtstamps &&
747	(sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
748	!skb_is_swtx_tstamp(skb, false_tstamp) &&
749	ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
750	empty = 0;
751	if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
752	!skb_is_err_queue(skb))
753	put_ts_pktinfo(msg, skb);
754	}
755	if (!empty) {
756	put_cmsg(msg, SOL_SOCKET,
757	SCM_TIMESTAMPING, sizeof(tss), &tss);
758
759	if (skb_is_err_queue(skb) && skb->len &&
760	SKB_EXT_ERR(skb)->opt_stats)
761	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
762	skb->len, skb->data);
763	}
764	}
765	EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
766
767	void __sock_recv_wifi_status(struct msghdr msg, struct sock sk,
768	struct sk_buff *skb)
769	{
770	int ack;
771
772	if (!sock_flag(sk, SOCK_WIFI_STATUS))
773	return;
774	if (!skb->wifi_acked_valid)
775	return;
776
777	ack = skb->wifi_acked;
778
779	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
780	}
781	EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
782
783	static inline void sock_recv_drops(struct msghdr msg, struct sock sk,
784	struct sk_buff *skb)
785	{
786	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
787	put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
788	sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
789	}
790
791	void __sock_recv_ts_and_drops(struct msghdr msg, struct sock sk,
792	struct sk_buff *skb)
793	{
794	sock_recv_timestamp(msg, sk, skb);
795	sock_recv_drops(msg, sk, skb);
796	}
797	EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
798
799	static inline int sock_recvmsg_nosec(struct socket sock, struct msghdr msg,
800	int flags)
801	{
802	return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
803	}
804
805	int sock_recvmsg(struct socket sock, struct msghdr msg, int flags)
806	{
807	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
808
809	return err ?: sock_recvmsg_nosec(sock, msg, flags);
810	}
811	EXPORT_SYMBOL(sock_recvmsg);
812
813	/**
814	* kernel_recvmsg - Receive a message from a socket (kernel space)
815	* @sock: The socket to receive the message from
816	* @msg: Received message
817	* @vec: Input s/g array for message data
818	* @num: Size of input s/g array
819	* @size: Number of bytes to read
820	* @flags: Message flags (MSG_DONTWAIT, etc...)
821	*
822	* On return the msg structure contains the scatter/gather array passed in the
823	* vec argument. The array is modified so that it consists of the unfilled
824	* portion of the original array.
825	*
826	* The returned value is the total number of bytes received, or an error.
827	*/
828	int kernel_recvmsg(struct socket sock, struct msghdr msg,
829	struct kvec *vec, size_t num, size_t size, int flags)
830	{
831	mm_segment_t oldfs = get_fs();
832	int result;
833
834	iov_iter_kvec(&msg->msg_iter, READ \| ITER_KVEC, vec, num, size);
835	set_fs(KERNEL_DS);
836	result = sock_recvmsg(sock, msg, flags);
837	set_fs(oldfs);
838	return result;
839	}
840	EXPORT_SYMBOL(kernel_recvmsg);
841
842	static ssize_t sock_sendpage(struct file file, struct page page,
843	int offset, size_t size, loff_t *ppos, int more)
844	{
845	struct socket *sock;
846	int flags;
847
848	sock = file->private_data;
849
850	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
851	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
852	flags \|= more;
853
854	return kernel_sendpage(sock, page, offset, size, flags);
855	}
856
857	static ssize_t sock_splice_read(struct file file, loff_t ppos,
858	struct pipe_inode_info *pipe, size_t len,
859	unsigned int flags)
860	{
861	struct socket *sock = file->private_data;
862
863	if (unlikely(!sock->ops->splice_read))
864	return -EINVAL;
865
866	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
867	}
868
869	static ssize_t sock_read_iter(struct kiocb iocb, struct iov_iter to)
870	{
871	struct file *file = iocb->ki_filp;
872	struct socket *sock = file->private_data;
873	struct msghdr msg = {.msg_iter = *to,
874	.msg_iocb = iocb};
875	ssize_t res;
876
877	if (file->f_flags & O_NONBLOCK)
878	msg.msg_flags = MSG_DONTWAIT;
879
880	if (iocb->ki_pos != 0)
881	return -ESPIPE;
882
883	if (!iov_iter_count(to)) /* Match SYS5 behaviour */
884	return 0;
885
886	res = sock_recvmsg(sock, &msg, msg.msg_flags);
887	*to = msg.msg_iter;
888	return res;
889	}
890
891	static ssize_t sock_write_iter(struct kiocb iocb, struct iov_iter from)
892	{
893	struct file *file = iocb->ki_filp;
894	struct socket *sock = file->private_data;
895	struct msghdr msg = {.msg_iter = *from,
896	.msg_iocb = iocb};
897	ssize_t res;
898
899	if (iocb->ki_pos != 0)
900	return -ESPIPE;
901
902	if (file->f_flags & O_NONBLOCK)
903	msg.msg_flags = MSG_DONTWAIT;
904
905	if (sock->type == SOCK_SEQPACKET)
906	msg.msg_flags \|= MSG_EOR;
907
908	res = sock_sendmsg(sock, &msg);
909	*from = msg.msg_iter;
910	return res;
911	}
912
913	/*
914	* Atomic setting of ioctl hooks to avoid race
915	* with module unload.
916	*/
917
918	static DEFINE_MUTEX(br_ioctl_mutex);
919	static int (br_ioctl_hook) (struct net , unsigned int cmd, void __user *arg);
920
921	void brioctl_set(int (hook) (struct net , unsigned int, void __user *))
922	{
923	mutex_lock(&br_ioctl_mutex);
924	br_ioctl_hook = hook;
925	mutex_unlock(&br_ioctl_mutex);
926	}
927	EXPORT_SYMBOL(brioctl_set);
928
929	static DEFINE_MUTEX(vlan_ioctl_mutex);
930	static int (vlan_ioctl_hook) (struct net , void __user *arg);
931
932	void vlan_ioctl_set(int (hook) (struct net , void __user *))
933	{
934	mutex_lock(&vlan_ioctl_mutex);
935	vlan_ioctl_hook = hook;
936	mutex_unlock(&vlan_ioctl_mutex);
937	}
938	EXPORT_SYMBOL(vlan_ioctl_set);
939
940	static DEFINE_MUTEX(dlci_ioctl_mutex);
941	static int (dlci_ioctl_hook) (unsigned int, void __user );
942
943	void dlci_ioctl_set(int (hook) (unsigned int, void __user ))
944	{
945	mutex_lock(&dlci_ioctl_mutex);
946	dlci_ioctl_hook = hook;
947	mutex_unlock(&dlci_ioctl_mutex);
948	}
949	EXPORT_SYMBOL(dlci_ioctl_set);
950
951	static long sock_do_ioctl(struct net net, struct socket sock,
952	unsigned int cmd, unsigned long arg)
953	{
954	int err;
955	void __user argp = (void __user )arg;
956
957	err = sock->ops->ioctl(sock, cmd, arg);
958
959	/*
960	* If this ioctl is unknown try to hand it down
961	* to the NIC driver.
962	*/
963	if (err != -ENOIOCTLCMD)
964	return err;
965
966	if (cmd == SIOCGIFCONF) {
967	struct ifconf ifc;
968	if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
969	return -EFAULT;
970	rtnl_lock();
971	err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
972	rtnl_unlock();
973	if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
974	err = -EFAULT;
975	} else {
976	struct ifreq ifr;
977	bool need_copyout;
978	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
979	return -EFAULT;
980	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
981	if (!err && need_copyout)
982	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
983	return -EFAULT;
984	}
985	return err;
986	}
987
988	/*
989	* With an ioctl, arg may well be a user mode pointer, but we don't know
990	* what to do with it - that's up to the protocol still.
991	*/
992
993	struct ns_common get_net_ns(struct ns_common ns)
994	{
995	return &get_net(container_of(ns, struct net, ns))->ns;
996	}
997	EXPORT_SYMBOL_GPL(get_net_ns);
998
999	static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1000	{
1001	struct socket *sock;
1002	struct sock *sk;
1003	void __user argp = (void __user )arg;
1004	int pid, err;
1005	struct net *net;
1006
1007	sock = file->private_data;
1008	sk = sock->sk;
1009	net = sock_net(sk);
1010	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1011	struct ifreq ifr;
1012	bool need_copyout;
1013	if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1014	return -EFAULT;
1015	err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1016	if (!err && need_copyout)
1017	if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1018	return -EFAULT;
1019	} else
1020	#ifdef CONFIG_WEXT_CORE
1021	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1022	err = wext_handle_ioctl(net, cmd, argp);
1023	} else
1024	#endif
1025	switch (cmd) {
1026	case FIOSETOWN:
1027	case SIOCSPGRP:
1028	err = -EFAULT;
1029	if (get_user(pid, (int __user *)argp))
1030	break;
1031	err = f_setown(sock->file, pid, 1);
1032	break;
1033	case FIOGETOWN:
1034	case SIOCGPGRP:
1035	err = put_user(f_getown(sock->file),
1036	(int __user *)argp);
1037	break;
1038	case SIOCGIFBR:
1039	case SIOCSIFBR:
1040	case SIOCBRADDBR:
1041	case SIOCBRDELBR:
1042	err = -ENOPKG;
1043	if (!br_ioctl_hook)
1044	request_module("bridge");
1045
1046	mutex_lock(&br_ioctl_mutex);
1047	if (br_ioctl_hook)
1048	err = br_ioctl_hook(net, cmd, argp);
1049	mutex_unlock(&br_ioctl_mutex);
1050	break;
1051	case SIOCGIFVLAN:
1052	case SIOCSIFVLAN:
1053	err = -ENOPKG;
1054	if (!vlan_ioctl_hook)
1055	request_module("8021q");
1056
1057	mutex_lock(&vlan_ioctl_mutex);
1058	if (vlan_ioctl_hook)
1059	err = vlan_ioctl_hook(net, argp);
1060	mutex_unlock(&vlan_ioctl_mutex);
1061	break;
1062	case SIOCADDDLCI:
1063	case SIOCDELDLCI:
1064	err = -ENOPKG;
1065	if (!dlci_ioctl_hook)
1066	request_module("dlci");
1067
1068	mutex_lock(&dlci_ioctl_mutex);
1069	if (dlci_ioctl_hook)
1070	err = dlci_ioctl_hook(cmd, argp);
1071	mutex_unlock(&dlci_ioctl_mutex);
1072	break;
1073	case SIOCGSKNS:
1074	err = -EPERM;
1075	if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1076	break;
1077
1078	err = open_related_ns(&net->ns, get_net_ns);
1079	break;
1080	default:
1081	err = sock_do_ioctl(net, sock, cmd, arg);
1082	break;
1083	}
1084	return err;
1085	}
1086
1087	int sock_create_lite(int family, int type, int protocol, struct socket **res)
1088	{
1089	int err;
1090	struct socket *sock = NULL;
1091
1092	err = security_socket_create(family, type, protocol, 1);
1093	if (err)
1094	goto out;
1095
1096	sock = sock_alloc();
1097	if (!sock) {
1098	err = -ENOMEM;
1099	goto out;
1100	}
1101
1102	sock->type = type;
1103	err = security_socket_post_create(sock, family, type, protocol, 1);
1104	if (err)
1105	goto out_release;
1106
1107	out:
1108	*res = sock;
1109	return err;
1110	out_release:
1111	sock_release(sock);
1112	sock = NULL;
1113	goto out;
1114	}
1115	EXPORT_SYMBOL(sock_create_lite);
1116
1117	/* No kernel lock held - perfect */
1118	static __poll_t sock_poll(struct file file, poll_table wait)
1119	{
1120	__poll_t busy_flag = 0;
1121	struct socket *sock;
1122
1123	/*
1124	* We can't return errors to poll, so it's either yes or no.
1125	*/
1126	sock = file->private_data;
1127
1128	if (sk_can_busy_loop(sock->sk)) {
1129	/* this socket can poll_ll so tell the system call */
1130	busy_flag = POLL_BUSY_LOOP;
1131
1132	/* once, only if requested by syscall */
1133	if (wait && (wait->_key & POLL_BUSY_LOOP))
1134	sk_busy_loop(sock->sk, 1);
1135	}
1136
1137	return busy_flag \| sock->ops->poll(file, sock, wait);
1138	}
1139
1140	static int sock_mmap(struct file file, struct vm_area_struct vma)
1141	{
1142	struct socket *sock = file->private_data;
1143
1144	return sock->ops->mmap(file, sock, vma);
1145	}
1146
1147	static int sock_close(struct inode inode, struct file filp)
1148	{
1149	sock_release(SOCKET_I(inode));
1150	return 0;
1151	}
1152
1153	/*
1154	* Update the socket async list
1155	*
1156	* Fasync_list locking strategy.
1157	*
1158	* 1. fasync_list is modified only under process context socket lock
1159	* i.e. under semaphore.
1160	* 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1161	* or under socket lock
1162	*/
1163
1164	static int sock_fasync(int fd, struct file *filp, int on)
1165	{
1166	struct socket *sock = filp->private_data;
1167	struct sock *sk = sock->sk;
1168	struct socket_wq *wq;
1169
1170	if (sk == NULL)
1171	return -EINVAL;
1172
1173	lock_sock(sk);
1174	wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1175	fasync_helper(fd, filp, on, &wq->fasync_list);
1176
1177	if (!wq->fasync_list)
1178	sock_reset_flag(sk, SOCK_FASYNC);
1179	else
1180	sock_set_flag(sk, SOCK_FASYNC);
1181
1182	release_sock(sk);
1183	return 0;
1184	}
1185
1186	/* This function may be called only under rcu_lock */
1187
1188	int sock_wake_async(struct socket_wq *wq, int how, int band)
1189	{
1190	if (!wq \|\| !wq->fasync_list)
1191	return -1;
1192
1193	switch (how) {
1194	case SOCK_WAKE_WAITD:
1195	if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1196	break;
1197	goto call_kill;
1198	case SOCK_WAKE_SPACE:
1199	if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1200	break;
1201	/* fall through */
1202	case SOCK_WAKE_IO:
1203	call_kill:
1204	kill_fasync(&wq->fasync_list, SIGIO, band);
1205	break;
1206	case SOCK_WAKE_URG:
1207	kill_fasync(&wq->fasync_list, SIGURG, band);
1208	}
1209
1210	return 0;
1211	}
1212	EXPORT_SYMBOL(sock_wake_async);
1213
1214	int __sock_create(struct net *net, int family, int type, int protocol,
1215	struct socket **res, int kern)
1216	{
1217	int err;
1218	struct socket *sock;
1219	const struct net_proto_family *pf;
1220
1221	/*
1222	* Check protocol is in range
1223	*/
1224	if (family < 0 \|\| family >= NPROTO)
1225	return -EAFNOSUPPORT;
1226	if (type < 0 \|\| type >= SOCK_MAX)
1227	return -EINVAL;
1228
1229	/* Compatibility.
1230
1231	This uglymoron is moved from INET layer to here to avoid
1232	deadlock in module load.
1233	*/
1234	if (family == PF_INET && type == SOCK_PACKET) {
1235	pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1236	current->comm);
1237	family = PF_PACKET;
1238	}
1239
1240	err = security_socket_create(family, type, protocol, kern);
1241	if (err)
1242	return err;
1243
1244	/*
1245	* Allocate the socket and allow the family to set things up. if
1246	* the protocol is 0, the family is instructed to select an appropriate
1247	* default.
1248	*/
1249	sock = sock_alloc();
1250	if (!sock) {
1251	net_warn_ratelimited("socket: no more sockets\n");
1252	return -ENFILE; /* Not exactly a match, but its the
1253	closest posix thing */
1254	}
1255
1256	sock->type = type;
1257
1258	#ifdef CONFIG_MODULES
1259	/* Attempt to load a protocol module if the find failed.
1260	*
1261	* 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1262	* requested real, full-featured networking support upon configuration.
1263	* Otherwise module support will break!
1264	*/
1265	if (rcu_access_pointer(net_families[family]) == NULL)
1266	request_module("net-pf-%d", family);
1267	#endif
1268
1269	rcu_read_lock();
1270	pf = rcu_dereference(net_families[family]);
1271	err = -EAFNOSUPPORT;
1272	if (!pf)
1273	goto out_release;
1274
1275	/*
1276	* We will call the ->create function, that possibly is in a loadable
1277	* module, so we have to bump that loadable module refcnt first.
1278	*/
1279	if (!try_module_get(pf->owner))
1280	goto out_release;
1281
1282	/* Now protected by module ref count */
1283	rcu_read_unlock();
1284
1285	err = pf->create(net, sock, protocol, kern);
1286	if (err < 0)
1287	goto out_module_put;
1288
1289	/*
1290	* Now to bump the refcnt of the [loadable] module that owns this
1291	* socket at sock_release time we decrement its refcnt.
1292	*/
1293	if (!try_module_get(sock->ops->owner))
1294	goto out_module_busy;
1295
1296	/*
1297	* Now that we're done with the ->create function, the [loadable]
1298	* module can have its refcnt decremented
1299	*/
1300	module_put(pf->owner);
1301	err = security_socket_post_create(sock, family, type, protocol, kern);
1302	if (err)
1303	goto out_sock_release;
1304	*res = sock;
1305
1306	return 0;
1307
1308	out_module_busy:
1309	err = -EAFNOSUPPORT;
1310	out_module_put:
1311	sock->ops = NULL;
1312	module_put(pf->owner);
1313	out_sock_release:
1314	sock_release(sock);
1315	return err;
1316
1317	out_release:
1318	rcu_read_unlock();
1319	goto out_sock_release;
1320	}
1321	EXPORT_SYMBOL(__sock_create);
1322
1323	int sock_create(int family, int type, int protocol, struct socket **res)
1324	{
1325	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1326	}
1327	EXPORT_SYMBOL(sock_create);
1328
1329	int sock_create_kern(struct net net, int family, int type, int protocol, struct socket *res)
1330	{
1331	return __sock_create(net, family, type, protocol, res, 1);
1332	}
1333	EXPORT_SYMBOL(sock_create_kern);
1334
1335	int __sys_socket(int family, int type, int protocol)
1336	{
1337	int retval;
1338	struct socket *sock;
1339	int flags;
1340
1341	/* Check the SOCK_* constants for consistency. */
1342	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1343	BUILD_BUG_ON((SOCK_MAX \| SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1344	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1345	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1346
1347	flags = type & ~SOCK_TYPE_MASK;
1348	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1349	return -EINVAL;
1350	type &= SOCK_TYPE_MASK;
1351
1352	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1353	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1354
1355	retval = sock_create(family, type, protocol, &sock);
1356	if (retval < 0)
1357	return retval;
1358
1359	return sock_map_fd(sock, flags & (O_CLOEXEC \| O_NONBLOCK));
1360	}
1361
1362	SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1363	{
1364	return __sys_socket(family, type, protocol);
1365	}
1366
1367	/*
1368	* Create a pair of connected sockets.
1369	*/
1370
1371	int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1372	{
1373	struct socket sock1, sock2;
1374	int fd1, fd2, err;
1375	struct file newfile1, newfile2;
1376	int flags;
1377
1378	flags = type & ~SOCK_TYPE_MASK;
1379	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1380	return -EINVAL;
1381	type &= SOCK_TYPE_MASK;
1382
1383	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1384	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1385
1386	/*
1387	* reserve descriptors and make sure we won't fail
1388	* to return them to userland.
1389	*/
1390	fd1 = get_unused_fd_flags(flags);
1391	if (unlikely(fd1 < 0))
1392	return fd1;
1393
1394	fd2 = get_unused_fd_flags(flags);
1395	if (unlikely(fd2 < 0)) {
1396	put_unused_fd(fd1);
1397	return fd2;
1398	}
1399
1400	err = put_user(fd1, &usockvec[0]);
1401	if (err)
1402	goto out;
1403
1404	err = put_user(fd2, &usockvec[1]);
1405	if (err)
1406	goto out;
1407
1408	/*
1409	* Obtain the first socket and check if the underlying protocol
1410	* supports the socketpair call.
1411	*/
1412
1413	err = sock_create(family, type, protocol, &sock1);
1414	if (unlikely(err < 0))
1415	goto out;
1416
1417	err = sock_create(family, type, protocol, &sock2);
1418	if (unlikely(err < 0)) {
1419	sock_release(sock1);
1420	goto out;
1421	}
1422
1423	err = sock1->ops->socketpair(sock1, sock2);
1424	if (unlikely(err < 0)) {
1425	sock_release(sock2);
1426	sock_release(sock1);
1427	goto out;
1428	}
1429
1430	newfile1 = sock_alloc_file(sock1, flags, NULL);
1431	if (IS_ERR(newfile1)) {
1432	err = PTR_ERR(newfile1);
1433	sock_release(sock2);
1434	goto out;
1435	}
1436
1437	newfile2 = sock_alloc_file(sock2, flags, NULL);
1438	if (IS_ERR(newfile2)) {
1439	err = PTR_ERR(newfile2);
1440	fput(newfile1);
1441	goto out;
1442	}
1443
1444	audit_fd_pair(fd1, fd2);
1445
1446	fd_install(fd1, newfile1);
1447	fd_install(fd2, newfile2);
1448	return 0;
1449
1450	out:
1451	put_unused_fd(fd2);
1452	put_unused_fd(fd1);
1453	return err;
1454	}
1455
1456	SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1457	int __user *, usockvec)
1458	{
1459	return __sys_socketpair(family, type, protocol, usockvec);
1460	}
1461
1462	/*
1463	* Bind a name to a socket. Nothing much to do here since it's
1464	* the protocol's responsibility to handle the local address.
1465	*
1466	* We move the socket address to kernel space before we call
1467	* the protocol layer (having also checked the address is ok).
1468	*/
1469
1470	int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1471	{
1472	struct socket *sock;
1473	struct sockaddr_storage address;
1474	int err, fput_needed;
1475
1476	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1477	if (sock) {
1478	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1479	if (err >= 0) {
1480	err = security_socket_bind(sock,
1481	(struct sockaddr *)&address,
1482	addrlen);
1483	if (!err)
1484	err = sock->ops->bind(sock,
1485	(struct sockaddr *)
1486	&address, addrlen);
1487	}
1488	fput_light(sock->file, fput_needed);
1489	}
1490	return err;
1491	}
1492
1493	SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1494	{
1495	return __sys_bind(fd, umyaddr, addrlen);
1496	}
1497
1498	/*
1499	* Perform a listen. Basically, we allow the protocol to do anything
1500	* necessary for a listen, and if that works, we mark the socket as
1501	* ready for listening.
1502	*/
1503
1504	int __sys_listen(int fd, int backlog)
1505	{
1506	struct socket *sock;
1507	int err, fput_needed;
1508	int somaxconn;
1509
1510	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1511	if (sock) {
1512	somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1513	if ((unsigned int)backlog > somaxconn)
1514	backlog = somaxconn;
1515
1516	err = security_socket_listen(sock, backlog);
1517	if (!err)
1518	err = sock->ops->listen(sock, backlog);
1519
1520	fput_light(sock->file, fput_needed);
1521	}
1522	return err;
1523	}
1524
1525	SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1526	{
1527	return __sys_listen(fd, backlog);
1528	}
1529
1530	/*
1531	* For accept, we attempt to create a new socket, set up the link
1532	* with the client, wake up the client, then return the new
1533	* connected fd. We collect the address of the connector in kernel
1534	* space and move it to user at the very end. This is unclean because
1535	* we open the socket then return an error.
1536	*
1537	* 1003.1g adds the ability to recvmsg() to query connection pending
1538	* status to recvmsg. We need to add that support in a way thats
1539	* clean when we restructure accept also.
1540	*/
1541
1542	int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1543	int __user *upeer_addrlen, int flags)
1544	{
1545	struct socket sock, newsock;
1546	struct file *newfile;
1547	int err, len, newfd, fput_needed;
1548	struct sockaddr_storage address;
1549
1550	if (flags & ~(SOCK_CLOEXEC \| SOCK_NONBLOCK))
1551	return -EINVAL;
1552
1553	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1554	flags = (flags & ~SOCK_NONBLOCK) \| O_NONBLOCK;
1555
1556	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1557	if (!sock)
1558	goto out;
1559
1560	err = -ENFILE;
1561	newsock = sock_alloc();
1562	if (!newsock)
1563	goto out_put;
1564
1565	newsock->type = sock->type;
1566	newsock->ops = sock->ops;
1567
1568	/*
1569	* We don't need try_module_get here, as the listening socket (sock)
1570	* has the protocol module (sock->ops->owner) held.
1571	*/
1572	__module_get(newsock->ops->owner);
1573
1574	newfd = get_unused_fd_flags(flags);
1575	if (unlikely(newfd < 0)) {
1576	err = newfd;
1577	sock_release(newsock);
1578	goto out_put;
1579	}
1580	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1581	if (IS_ERR(newfile)) {
1582	err = PTR_ERR(newfile);
1583	put_unused_fd(newfd);
1584	goto out_put;
1585	}
1586
1587	err = security_socket_accept(sock, newsock);
1588	if (err)
1589	goto out_fd;
1590
1591	err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1592	if (err < 0)
1593	goto out_fd;
1594
1595	if (upeer_sockaddr) {
1596	len = newsock->ops->getname(newsock,
1597	(struct sockaddr *)&address, 2);
1598	if (len < 0) {
1599	err = -ECONNABORTED;
1600	goto out_fd;
1601	}
1602	err = move_addr_to_user(&address,
1603	len, upeer_sockaddr, upeer_addrlen);
1604	if (err < 0)
1605	goto out_fd;
1606	}
1607
1608	/* File flags are not inherited via accept() unlike another OSes. */
1609
1610	fd_install(newfd, newfile);
1611	err = newfd;
1612
1613	out_put:
1614	fput_light(sock->file, fput_needed);
1615	out:
1616	return err;
1617	out_fd:
1618	fput(newfile);
1619	put_unused_fd(newfd);
1620	goto out_put;
1621	}
1622
1623	SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1624	int __user *, upeer_addrlen, int, flags)
1625	{
1626	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1627	}
1628
1629	SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1630	int __user *, upeer_addrlen)
1631	{
1632	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1633	}
1634
1635	/*
1636	* Attempt to connect to a socket with the server address. The address
1637	* is in user space so we verify it is OK and move it to kernel space.
1638	*
1639	* For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1640	* break bindings
1641	*
1642	* NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1643	* other SEQPACKET protocols that take time to connect() as it doesn't
1644	* include the -EINPROGRESS status for such sockets.
1645	*/
1646
1647	int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1648	{
1649	struct socket *sock;
1650	struct sockaddr_storage address;
1651	int err, fput_needed;
1652
1653	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1654	if (!sock)
1655	goto out;
1656	err = move_addr_to_kernel(uservaddr, addrlen, &address);
1657	if (err < 0)
1658	goto out_put;
1659
1660	err =
1661	security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1662	if (err)
1663	goto out_put;
1664
1665	err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1666	sock->file->f_flags);
1667	out_put:
1668	fput_light(sock->file, fput_needed);
1669	out:
1670	return err;
1671	}
1672
1673	SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1674	int, addrlen)
1675	{
1676	return __sys_connect(fd, uservaddr, addrlen);
1677	}
1678
1679	/*
1680	* Get the local address ('name') of a socket object. Move the obtained
1681	* name to user space.
1682	*/
1683
1684	int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1685	int __user *usockaddr_len)
1686	{
1687	struct socket *sock;
1688	struct sockaddr_storage address;
1689	int err, fput_needed;
1690
1691	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1692	if (!sock)
1693	goto out;
1694
1695	err = security_socket_getsockname(sock);
1696	if (err)
1697	goto out_put;
1698
1699	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1700	if (err < 0)
1701	goto out_put;
1702	/* "err" is actually length in this case */
1703	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1704
1705	out_put:
1706	fput_light(sock->file, fput_needed);
1707	out:
1708	return err;
1709	}
1710
1711	SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1712	int __user *, usockaddr_len)
1713	{
1714	return __sys_getsockname(fd, usockaddr, usockaddr_len);
1715	}
1716
1717	/*
1718	* Get the remote address ('name') of a socket object. Move the obtained
1719	* name to user space.
1720	*/
1721
1722	int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1723	int __user *usockaddr_len)
1724	{
1725	struct socket *sock;
1726	struct sockaddr_storage address;
1727	int err, fput_needed;
1728
1729	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1730	if (sock != NULL) {
1731	err = security_socket_getpeername(sock);
1732	if (err) {
1733	fput_light(sock->file, fput_needed);
1734	return err;
1735	}
1736
1737	err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1738	if (err >= 0)
1739	/* "err" is actually length in this case */
1740	err = move_addr_to_user(&address, err, usockaddr,
1741	usockaddr_len);
1742	fput_light(sock->file, fput_needed);
1743	}
1744	return err;
1745	}
1746
1747	SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1748	int __user *, usockaddr_len)
1749	{
1750	return __sys_getpeername(fd, usockaddr, usockaddr_len);
1751	}
1752
1753	/*
1754	* Send a datagram to a given address. We move the address into kernel
1755	* space and check the user space data area is readable before invoking
1756	* the protocol.
1757	*/
1758	int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1759	struct sockaddr __user *addr, int addr_len)
1760	{
1761	struct socket *sock;
1762	struct sockaddr_storage address;
1763	int err;
1764	struct msghdr msg;
1765	struct iovec iov;
1766	int fput_needed;
1767
1768	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1769	if (unlikely(err))
1770	return err;
1771	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1772	if (!sock)
1773	goto out;
1774
1775	msg.msg_name = NULL;
1776	msg.msg_control = NULL;
1777	msg.msg_controllen = 0;
1778	msg.msg_namelen = 0;
1779	if (addr) {
1780	err = move_addr_to_kernel(addr, addr_len, &address);
1781	if (err < 0)
1782	goto out_put;
1783	msg.msg_name = (struct sockaddr *)&address;
1784	msg.msg_namelen = addr_len;
1785	}
1786	if (sock->file->f_flags & O_NONBLOCK)
1787	flags \|= MSG_DONTWAIT;
1788	msg.msg_flags = flags;
1789	err = sock_sendmsg(sock, &msg);
1790
1791	out_put:
1792	fput_light(sock->file, fput_needed);
1793	out:
1794	return err;
1795	}
1796
1797	SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1798	unsigned int, flags, struct sockaddr __user *, addr,
1799	int, addr_len)
1800	{
1801	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1802	}
1803
1804	/*
1805	* Send a datagram down a socket.
1806	*/
1807
1808	SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1809	unsigned int, flags)
1810	{
1811	return __sys_sendto(fd, buff, len, flags, NULL, 0);
1812	}
1813
1814	/*
1815	* Receive a frame from the socket and optionally record the address of the
1816	* sender. We verify the buffers are writable and if needed move the
1817	* sender address from kernel to user space.
1818	*/
1819	int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1820	struct sockaddr __user addr, int __user addr_len)
1821	{
1822	struct socket *sock;
1823	struct iovec iov;
1824	struct msghdr msg;
1825	struct sockaddr_storage address;
1826	int err, err2;
1827	int fput_needed;
1828
1829	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1830	if (unlikely(err))
1831	return err;
1832	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1833	if (!sock)
1834	goto out;
1835
1836	msg.msg_control = NULL;
1837	msg.msg_controllen = 0;
1838	/* Save some cycles and don't copy the address if not needed */
1839	msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1840	/* We assume all kernel code knows the size of sockaddr_storage */
1841	msg.msg_namelen = 0;
1842	msg.msg_iocb = NULL;
1843	msg.msg_flags = 0;
1844	if (sock->file->f_flags & O_NONBLOCK)
1845	flags \|= MSG_DONTWAIT;
1846	err = sock_recvmsg(sock, &msg, flags);
1847
1848	if (err >= 0 && addr != NULL) {
1849	err2 = move_addr_to_user(&address,
1850	msg.msg_namelen, addr, addr_len);
1851	if (err2 < 0)
1852	err = err2;
1853	}
1854
1855	fput_light(sock->file, fput_needed);
1856	out:
1857	return err;
1858	}
1859
1860	SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1861	unsigned int, flags, struct sockaddr __user *, addr,
1862	int __user *, addr_len)
1863	{
1864	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1865	}
1866
1867	/*
1868	* Receive a datagram from a socket.
1869	*/
1870
1871	SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1872	unsigned int, flags)
1873	{
1874	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1875	}
1876
1877	/*
1878	* Set a socket option. Because we don't know the option lengths we have
1879	* to pass the user mode parameter for the protocols to sort out.
1880	*/
1881
1882	static int __sys_setsockopt(int fd, int level, int optname,
1883	char __user *optval, int optlen)
1884	{
1885	int err, fput_needed;
1886	struct socket *sock;
1887
1888	if (optlen < 0)
1889	return -EINVAL;
1890
1891	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1892	if (sock != NULL) {
1893	err = security_socket_setsockopt(sock, level, optname);
1894	if (err)
1895	goto out_put;
1896
1897	if (level == SOL_SOCKET)
1898	err =
1899	sock_setsockopt(sock, level, optname, optval,
1900	optlen);
1901	else
1902	err =
1903	sock->ops->setsockopt(sock, level, optname, optval,
1904	optlen);
1905	out_put:
1906	fput_light(sock->file, fput_needed);
1907	}
1908	return err;
1909	}
1910
1911	SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1912	char __user *, optval, int, optlen)
1913	{
1914	return __sys_setsockopt(fd, level, optname, optval, optlen);
1915	}
1916
1917	/*
1918	* Get a socket option. Because we don't know the option lengths we have
1919	* to pass a user mode parameter for the protocols to sort out.
1920	*/
1921
1922	static int __sys_getsockopt(int fd, int level, int optname,
1923	char __user optval, int __user optlen)
1924	{
1925	int err, fput_needed;
1926	struct socket *sock;
1927
1928	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1929	if (sock != NULL) {
1930	err = security_socket_getsockopt(sock, level, optname);
1931	if (err)
1932	goto out_put;
1933
1934	if (level == SOL_SOCKET)
1935	err =
1936	sock_getsockopt(sock, level, optname, optval,
1937	optlen);
1938	else
1939	err =
1940	sock->ops->getsockopt(sock, level, optname, optval,
1941	optlen);
1942	out_put:
1943	fput_light(sock->file, fput_needed);
1944	}
1945	return err;
1946	}
1947
1948	SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1949	char __user , optval, int __user , optlen)
1950	{
1951	return __sys_getsockopt(fd, level, optname, optval, optlen);
1952	}
1953
1954	/*
1955	* Shutdown a socket.
1956	*/
1957
1958	int __sys_shutdown(int fd, int how)
1959	{
1960	int err, fput_needed;
1961	struct socket *sock;
1962
1963	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1964	if (sock != NULL) {
1965	err = security_socket_shutdown(sock, how);
1966	if (!err)
1967	err = sock->ops->shutdown(sock, how);
1968	fput_light(sock->file, fput_needed);
1969	}
1970	return err;
1971	}
1972
1973	SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1974	{
1975	return __sys_shutdown(fd, how);
1976	}
1977
1978	/* A couple of helpful macros for getting the address of the 32/64 bit
1979	* fields which are the same type (int / unsigned) on our platforms.
1980	*/
1981	#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1982	#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1983	#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1984
1985	struct used_address {
1986	struct sockaddr_storage name;
1987	unsigned int name_len;
1988	};
1989
1990	static int copy_msghdr_from_user(struct msghdr *kmsg,
1991	struct user_msghdr __user *umsg,
1992	struct sockaddr __user **save_addr,
1993	struct iovec **iov)
1994	{
1995	struct user_msghdr msg;
1996	ssize_t err;
1997
1998	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
1999	return -EFAULT;
2000
2001	kmsg->msg_control = (void __force *)msg.msg_control;
2002	kmsg->msg_controllen = msg.msg_controllen;
2003	kmsg->msg_flags = msg.msg_flags;
2004
2005	kmsg->msg_namelen = msg.msg_namelen;
2006	if (!msg.msg_name)
2007	kmsg->msg_namelen = 0;
2008
2009	if (kmsg->msg_namelen < 0)
2010	return -EINVAL;
2011
2012	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2013	kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2014
2015	if (save_addr)
2016	*save_addr = msg.msg_name;
2017
2018	if (msg.msg_name && kmsg->msg_namelen) {
2019	if (!save_addr) {
2020	err = move_addr_to_kernel(msg.msg_name,
2021	kmsg->msg_namelen,
2022	kmsg->msg_name);
2023	if (err < 0)
2024	return err;
2025	}
2026	} else {
2027	kmsg->msg_name = NULL;
2028	kmsg->msg_namelen = 0;
2029	}
2030
2031	if (msg.msg_iovlen > UIO_MAXIOV)
2032	return -EMSGSIZE;
2033
2034	kmsg->msg_iocb = NULL;
2035
2036	return import_iovec(save_addr ? READ : WRITE,
2037	msg.msg_iov, msg.msg_iovlen,
2038	UIO_FASTIOV, iov, &kmsg->msg_iter);
2039	}
2040
2041	static int ___sys_sendmsg(struct socket sock, struct user_msghdr __user msg,
2042	struct msghdr *msg_sys, unsigned int flags,
2043	struct used_address *used_address,
2044	unsigned int allowed_msghdr_flags)
2045	{
2046	struct compat_msghdr __user *msg_compat =
2047	(struct compat_msghdr __user *)msg;
2048	struct sockaddr_storage address;
2049	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2050	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2051	__aligned(sizeof(__kernel_size_t));
2052	/* 20 is size of ipv6_pktinfo */
2053	unsigned char *ctl_buf = ctl;
2054	int ctl_len;
2055	ssize_t err;
2056
2057	msg_sys->msg_name = &address;
2058
2059	if (MSG_CMSG_COMPAT & flags)
2060	err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2061	else
2062	err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2063	if (err < 0)
2064	return err;
2065
2066	err = -ENOBUFS;
2067
2068	if (msg_sys->msg_controllen > INT_MAX)
2069	goto out_freeiov;
2070	flags \|= (msg_sys->msg_flags & allowed_msghdr_flags);
2071	ctl_len = msg_sys->msg_controllen;
2072	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2073	err =
2074	cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2075	sizeof(ctl));
2076	if (err)
2077	goto out_freeiov;
2078	ctl_buf = msg_sys->msg_control;
2079	ctl_len = msg_sys->msg_controllen;
2080	} else if (ctl_len) {
2081	BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2082	CMSG_ALIGN(sizeof(struct cmsghdr)));
2083	if (ctl_len > sizeof(ctl)) {
2084	ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2085	if (ctl_buf == NULL)
2086	goto out_freeiov;
2087	}
2088	err = -EFAULT;
2089	/*
2090	* Careful! Before this, msg_sys->msg_control contains a user pointer.
2091	* Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2092	* checking falls down on this.
2093	*/
2094	if (copy_from_user(ctl_buf,
2095	(void __user __force *)msg_sys->msg_control,
2096	ctl_len))
2097	goto out_freectl;
2098	msg_sys->msg_control = ctl_buf;
2099	}
2100	msg_sys->msg_flags = flags;
2101
2102	if (sock->file->f_flags & O_NONBLOCK)
2103	msg_sys->msg_flags \|= MSG_DONTWAIT;
2104	/*
2105	* If this is sendmmsg() and current destination address is same as
2106	* previously succeeded address, omit asking LSM's decision.
2107	* used_address->name_len is initialized to UINT_MAX so that the first
2108	* destination address never matches.
2109	*/
2110	if (used_address && msg_sys->msg_name &&
2111	used_address->name_len == msg_sys->msg_namelen &&
2112	!memcmp(&used_address->name, msg_sys->msg_name,
2113	used_address->name_len)) {
2114	err = sock_sendmsg_nosec(sock, msg_sys);
2115	goto out_freectl;
2116	}
2117	err = sock_sendmsg(sock, msg_sys);
2118	/*
2119	* If this is sendmmsg() and sending to current destination address was
2120	* successful, remember it.
2121	*/
2122	if (used_address && err >= 0) {
2123	used_address->name_len = msg_sys->msg_namelen;
2124	if (msg_sys->msg_name)
2125	memcpy(&used_address->name, msg_sys->msg_name,
2126	used_address->name_len);
2127	}
2128
2129	out_freectl:
2130	if (ctl_buf != ctl)
2131	sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2132	out_freeiov:
2133	kfree(iov);
2134	return err;
2135	}
2136
2137	/*
2138	* BSD sendmsg interface
2139	*/
2140
2141	long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2142	bool forbid_cmsg_compat)
2143	{
2144	int fput_needed, err;
2145	struct msghdr msg_sys;
2146	struct socket *sock;
2147
2148	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2149	return -EINVAL;
2150
2151	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2152	if (!sock)
2153	goto out;
2154
2155	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2156
2157	fput_light(sock->file, fput_needed);
2158	out:
2159	return err;
2160	}
2161
2162	SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2163	{
2164	return __sys_sendmsg(fd, msg, flags, true);
2165	}
2166
2167	/*
2168	* Linux sendmmsg interface
2169	*/
2170
2171	int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2172	unsigned int flags, bool forbid_cmsg_compat)
2173	{
2174	int fput_needed, err, datagrams;
2175	struct socket *sock;
2176	struct mmsghdr __user *entry;
2177	struct compat_mmsghdr __user *compat_entry;
2178	struct msghdr msg_sys;
2179	struct used_address used_address;
2180	unsigned int oflags = flags;
2181
2182	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2183	return -EINVAL;
2184
2185	if (vlen > UIO_MAXIOV)
2186	vlen = UIO_MAXIOV;
2187
2188	datagrams = 0;
2189
2190	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2191	if (!sock)
2192	return err;
2193
2194	used_address.name_len = UINT_MAX;
2195	entry = mmsg;
2196	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2197	err = 0;
2198	flags \|= MSG_BATCH;
2199
2200	while (datagrams < vlen) {
2201	if (datagrams == vlen - 1)
2202	flags = oflags;
2203
2204	if (MSG_CMSG_COMPAT & flags) {
2205	err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2206	&msg_sys, flags, &used_address, MSG_EOR);
2207	if (err < 0)
2208	break;
2209	err = __put_user(err, &compat_entry->msg_len);
2210	++compat_entry;
2211	} else {
2212	err = ___sys_sendmsg(sock,
2213	(struct user_msghdr __user *)entry,
2214	&msg_sys, flags, &used_address, MSG_EOR);
2215	if (err < 0)
2216	break;
2217	err = put_user(err, &entry->msg_len);
2218	++entry;
2219	}
2220
2221	if (err)
2222	break;
2223	++datagrams;
2224	if (msg_data_left(&msg_sys))
2225	break;
2226	cond_resched();
2227	}
2228
2229	fput_light(sock->file, fput_needed);
2230
2231	/* We only return an error if no datagrams were able to be sent */
2232	if (datagrams != 0)
2233	return datagrams;
2234
2235	return err;
2236	}
2237
2238	SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2239	unsigned int, vlen, unsigned int, flags)
2240	{
2241	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2242	}
2243
2244	static int ___sys_recvmsg(struct socket sock, struct user_msghdr __user msg,
2245	struct msghdr *msg_sys, unsigned int flags, int nosec)
2246	{
2247	struct compat_msghdr __user *msg_compat =
2248	(struct compat_msghdr __user *)msg;
2249	struct iovec iovstack[UIO_FASTIOV];
2250	struct iovec *iov = iovstack;
2251	unsigned long cmsg_ptr;
2252	int len;
2253	ssize_t err;
2254
2255	/* kernel mode address */
2256	struct sockaddr_storage addr;
2257
2258	/* user mode address pointers */
2259	struct sockaddr __user *uaddr;
2260	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2261
2262	msg_sys->msg_name = &addr;
2263
2264	if (MSG_CMSG_COMPAT & flags)
2265	err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2266	else
2267	err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2268	if (err < 0)
2269	return err;
2270
2271	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2272	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC\|MSG_CMSG_COMPAT);
2273
2274	/* We assume all kernel code knows the size of sockaddr_storage */
2275	msg_sys->msg_namelen = 0;
2276
2277	if (sock->file->f_flags & O_NONBLOCK)
2278	flags \|= MSG_DONTWAIT;
2279	err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2280	if (err < 0)
2281	goto out_freeiov;
2282	len = err;
2283
2284	if (uaddr != NULL) {
2285	err = move_addr_to_user(&addr,
2286	msg_sys->msg_namelen, uaddr,
2287	uaddr_len);
2288	if (err < 0)
2289	goto out_freeiov;
2290	}
2291	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2292	COMPAT_FLAGS(msg));
2293	if (err)
2294	goto out_freeiov;
2295	if (MSG_CMSG_COMPAT & flags)
2296	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2297	&msg_compat->msg_controllen);
2298	else
2299	err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2300	&msg->msg_controllen);
2301	if (err)
2302	goto out_freeiov;
2303	err = len;
2304
2305	out_freeiov:
2306	kfree(iov);
2307	return err;
2308	}
2309
2310	/*
2311	* BSD recvmsg interface
2312	*/
2313
2314	long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2315	bool forbid_cmsg_compat)
2316	{
2317	int fput_needed, err;
2318	struct msghdr msg_sys;
2319	struct socket *sock;
2320
2321	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2322	return -EINVAL;
2323
2324	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2325	if (!sock)
2326	goto out;
2327
2328	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2329
2330	fput_light(sock->file, fput_needed);
2331	out:
2332	return err;
2333	}
2334
2335	SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2336	unsigned int, flags)
2337	{
2338	return __sys_recvmsg(fd, msg, flags, true);
2339	}
2340
2341	/*
2342	* Linux recvmmsg interface
2343	*/
2344
2345	int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2346	unsigned int flags, struct timespec *timeout)
2347	{
2348	int fput_needed, err, datagrams;
2349	struct socket *sock;
2350	struct mmsghdr __user *entry;
2351	struct compat_mmsghdr __user *compat_entry;
2352	struct msghdr msg_sys;
2353	struct timespec64 end_time;
2354	struct timespec64 timeout64;
2355
2356	if (timeout &&
2357	poll_select_set_timeout(&end_time, timeout->tv_sec,
2358	timeout->tv_nsec))
2359	return -EINVAL;
2360
2361	datagrams = 0;
2362
2363	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2364	if (!sock)
2365	return err;
2366
2367	if (likely(!(flags & MSG_ERRQUEUE))) {
2368	err = sock_error(sock->sk);
2369	if (err) {
2370	datagrams = err;
2371	goto out_put;
2372	}
2373	}
2374
2375	entry = mmsg;
2376	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2377
2378	while (datagrams < vlen) {
2379	/*
2380	* No need to ask LSM for more than the first datagram.
2381	*/
2382	if (MSG_CMSG_COMPAT & flags) {
2383	err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2384	&msg_sys, flags & ~MSG_WAITFORONE,
2385	datagrams);
2386	if (err < 0)
2387	break;
2388	err = __put_user(err, &compat_entry->msg_len);
2389	++compat_entry;
2390	} else {
2391	err = ___sys_recvmsg(sock,
2392	(struct user_msghdr __user *)entry,
2393	&msg_sys, flags & ~MSG_WAITFORONE,
2394	datagrams);
2395	if (err < 0)
2396	break;
2397	err = put_user(err, &entry->msg_len);
2398	++entry;
2399	}
2400
2401	if (err)
2402	break;
2403	++datagrams;
2404
2405	/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2406	if (flags & MSG_WAITFORONE)
2407	flags \|= MSG_DONTWAIT;
2408
2409	if (timeout) {
2410	ktime_get_ts64(&timeout64);
2411	*timeout = timespec64_to_timespec(
2412	timespec64_sub(end_time, timeout64));
2413	if (timeout->tv_sec < 0) {
2414	timeout->tv_sec = timeout->tv_nsec = 0;
2415	break;
2416	}
2417
2418	/* Timeout, return less than vlen datagrams */
2419	if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2420	break;
2421	}
2422
2423	/* Out of band data, return right away */
2424	if (msg_sys.msg_flags & MSG_OOB)
2425	break;
2426	cond_resched();
2427	}
2428
2429	if (err == 0)
2430	goto out_put;
2431
2432	if (datagrams == 0) {
2433	datagrams = err;
2434	goto out_put;
2435	}
2436
2437	/*
2438	* We may return less entries than requested (vlen) if the
2439	* sock is non block and there aren't enough datagrams...
2440	*/
2441	if (err != -EAGAIN) {
2442	/*
2443	* ... or if recvmsg returns an error after we
2444	* received some datagrams, where we record the
2445	* error to return on the next call or if the
2446	* app asks about it using getsockopt(SO_ERROR).
2447	*/
2448	sock->sk->sk_err = -err;
2449	}
2450	out_put:
2451	fput_light(sock->file, fput_needed);
2452
2453	return datagrams;
2454	}
2455
2456	static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2457	unsigned int vlen, unsigned int flags,
2458	struct timespec __user *timeout)
2459	{
2460	int datagrams;
2461	struct timespec timeout_sys;
2462
2463	if (flags & MSG_CMSG_COMPAT)
2464	return -EINVAL;
2465
2466	if (!timeout)
2467	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2468
2469	if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2470	return -EFAULT;
2471
2472	datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2473
2474	if (datagrams > 0 &&
2475	copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2476	datagrams = -EFAULT;
2477
2478	return datagrams;
2479	}
2480
2481	SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2482	unsigned int, vlen, unsigned int, flags,
2483	struct timespec __user *, timeout)
2484	{
2485	return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2486	}
2487
2488	#ifdef __ARCH_WANT_SYS_SOCKETCALL
2489	/* Argument list sizes for sys_socketcall */
2490	#define AL(x) ((x) * sizeof(unsigned long))
2491	static const unsigned char nargs[21] = {
2492	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2493	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2494	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2495	AL(4), AL(5), AL(4)
2496	};
2497
2498	#undef AL
2499
2500	/*
2501	* System call vectors.
2502	*
2503	* Argument checking cleaned up. Saved 20% in size.
2504	* This function doesn't need to set the kernel lock because
2505	* it is set by the callees.
2506	*/
2507
2508	SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2509	{
2510	unsigned long a[AUDITSC_ARGS];
2511	unsigned long a0, a1;
2512	int err;
2513	unsigned int len;
2514
2515	if (call < 1 \|\| call > SYS_SENDMMSG)
2516	return -EINVAL;
2517
2518	len = nargs[call];
2519	if (len > sizeof(a))
2520	return -EINVAL;
2521
2522	/* copy_from_user should be SMP safe. */
2523	if (copy_from_user(a, args, len))
2524	return -EFAULT;
2525
2526	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2527	if (err)
2528	return err;
2529
2530	a0 = a[0];
2531	a1 = a[1];
2532
2533	switch (call) {
2534	case SYS_SOCKET:
2535	err = __sys_socket(a0, a1, a[2]);
2536	break;
2537	case SYS_BIND:
2538	err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2539	break;
2540	case SYS_CONNECT:
2541	err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2542	break;
2543	case SYS_LISTEN:
2544	err = __sys_listen(a0, a1);
2545	break;
2546	case SYS_ACCEPT:
2547	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2548	(int __user *)a[2], 0);
2549	break;
2550	case SYS_GETSOCKNAME:
2551	err =
2552	__sys_getsockname(a0, (struct sockaddr __user *)a1,
2553	(int __user *)a[2]);
2554	break;
2555	case SYS_GETPEERNAME:
2556	err =
2557	__sys_getpeername(a0, (struct sockaddr __user *)a1,
2558	(int __user *)a[2]);
2559	break;
2560	case SYS_SOCKETPAIR:
2561	err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2562	break;
2563	case SYS_SEND:
2564	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2565	NULL, 0);
2566	break;
2567	case SYS_SENDTO:
2568	err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2569	(struct sockaddr __user *)a[4], a[5]);
2570	break;
2571	case SYS_RECV:
2572	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2573	NULL, NULL);
2574	break;
2575	case SYS_RECVFROM:
2576	err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2577	(struct sockaddr __user *)a[4],
2578	(int __user *)a[5]);
2579	break;
2580	case SYS_SHUTDOWN:
2581	err = __sys_shutdown(a0, a1);
2582	break;
2583	case SYS_SETSOCKOPT:
2584	err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2585	a[4]);
2586	break;
2587	case SYS_GETSOCKOPT:
2588	err =
2589	__sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2590	(int __user *)a[4]);
2591	break;
2592	case SYS_SENDMSG:
2593	err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2594	a[2], true);
2595	break;
2596	case SYS_SENDMMSG:
2597	err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2598	a[3], true);
2599	break;
2600	case SYS_RECVMSG:
2601	err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2602	a[2], true);
2603	break;
2604	case SYS_RECVMMSG:
2605	err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2606	a[3], (struct timespec __user *)a[4]);
2607	break;
2608	case SYS_ACCEPT4:
2609	err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2610	(int __user *)a[2], a[3]);
2611	break;
2612	default:
2613	err = -EINVAL;
2614	break;
2615	}
2616	return err;
2617	}
2618
2619	#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2620
2621	/**
2622	* sock_register - add a socket protocol handler
2623	* @ops: description of protocol
2624	*
2625	* This function is called by a protocol handler that wants to
2626	* advertise its address family, and have it linked into the
2627	* socket interface. The value ops->family corresponds to the
2628	* socket system call protocol family.
2629	*/
2630	int sock_register(const struct net_proto_family *ops)
2631	{
2632	int err;
2633
2634	if (ops->family >= NPROTO) {
2635	pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2636	return -ENOBUFS;
2637	}
2638
2639	spin_lock(&net_family_lock);
2640	if (rcu_dereference_protected(net_families[ops->family],
2641	lockdep_is_held(&net_family_lock)))
2642	err = -EEXIST;
2643	else {
2644	rcu_assign_pointer(net_families[ops->family], ops);
2645	err = 0;
2646	}
2647	spin_unlock(&net_family_lock);
2648
2649	pr_info("NET: Registered protocol family %d\n", ops->family);
2650	return err;
2651	}
2652	EXPORT_SYMBOL(sock_register);
2653
2654	/**
2655	* sock_unregister - remove a protocol handler
2656	* @family: protocol family to remove
2657	*
2658	* This function is called by a protocol handler that wants to
2659	* remove its address family, and have it unlinked from the
2660	* new socket creation.
2661	*
2662	* If protocol handler is a module, then it can use module reference
2663	* counts to protect against new references. If protocol handler is not
2664	* a module then it needs to provide its own protection in
2665	* the ops->create routine.
2666	*/
2667	void sock_unregister(int family)
2668	{
2669	BUG_ON(family < 0 \|\| family >= NPROTO);
2670
2671	spin_lock(&net_family_lock);
2672	RCU_INIT_POINTER(net_families[family], NULL);
2673	spin_unlock(&net_family_lock);
2674
2675	synchronize_rcu();
2676
2677	pr_info("NET: Unregistered protocol family %d\n", family);
2678	}
2679	EXPORT_SYMBOL(sock_unregister);
2680
2681	bool sock_is_registered(int family)
2682	{
2683	return family < NPROTO && rcu_access_pointer(net_families[family]);
2684	}
2685
2686	static int __init sock_init(void)
2687	{
2688	int err;
2689	/*
2690	* Initialize the network sysctl infrastructure.
2691	*/
2692	err = net_sysctl_init();
2693	if (err)
2694	goto out;
2695
2696	/*
2697	* Initialize skbuff SLAB cache
2698	*/
2699	skb_init();
2700
2701	/*
2702	* Initialize the protocols module.
2703	*/
2704
2705	init_inodecache();
2706
2707	err = register_filesystem(&sock_fs_type);
2708	if (err)
2709	goto out_fs;
2710	sock_mnt = kern_mount(&sock_fs_type);
2711	if (IS_ERR(sock_mnt)) {
2712	err = PTR_ERR(sock_mnt);
2713	goto out_mount;
2714	}
2715
2716	/* The real protocol initialization is performed in later initcalls.
2717	*/
2718
2719	#ifdef CONFIG_NETFILTER
2720	err = netfilter_init();
2721	if (err)
2722	goto out;
2723	#endif
2724
2725	ptp_classifier_init();
2726
2727	out:
2728	return err;
2729
2730	out_mount:
2731	unregister_filesystem(&sock_fs_type);
2732	out_fs:
2733	goto out;
2734	}
2735
2736	core_initcall(sock_init); /* early initcall */
2737
2738	#ifdef CONFIG_PROC_FS
2739	void socket_seq_show(struct seq_file *seq)
2740	{
2741	seq_printf(seq, "sockets: used %d\n",
2742	sock_inuse_get(seq->private));
2743	}
2744	#endif /* CONFIG_PROC_FS */
2745
2746	#ifdef CONFIG_COMPAT
2747	static int do_siocgstamp(struct net net, struct socket sock,
2748	unsigned int cmd, void __user *up)
2749	{
2750	mm_segment_t old_fs = get_fs();
2751	struct timeval ktv;
2752	int err;
2753
2754	set_fs(KERNEL_DS);
2755	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2756	set_fs(old_fs);
2757	if (!err)
2758	err = compat_put_timeval(&ktv, up);
2759
2760	return err;
2761	}
2762
2763	static int do_siocgstampns(struct net net, struct socket sock,
2764	unsigned int cmd, void __user *up)
2765	{
2766	mm_segment_t old_fs = get_fs();
2767	struct timespec kts;
2768	int err;
2769
2770	set_fs(KERNEL_DS);
2771	err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2772	set_fs(old_fs);
2773	if (!err)
2774	err = compat_put_timespec(&kts, up);
2775
2776	return err;
2777	}
2778
2779	static int compat_dev_ifconf(struct net net, struct compat_ifconf __user uifc32)
2780	{
2781	struct compat_ifconf ifc32;
2782	struct ifconf ifc;
2783	int err;
2784
2785	if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2786	return -EFAULT;
2787
2788	ifc.ifc_len = ifc32.ifc_len;
2789	ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2790
2791	rtnl_lock();
2792	err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2793	rtnl_unlock();
2794	if (err)
2795	return err;
2796
2797	ifc32.ifc_len = ifc.ifc_len;
2798	if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2799	return -EFAULT;
2800
2801	return 0;
2802	}
2803
2804	static int ethtool_ioctl(struct net net, struct compat_ifreq __user ifr32)
2805	{
2806	struct compat_ethtool_rxnfc __user *compat_rxnfc;
2807	bool convert_in = false, convert_out = false;
2808	size_t buf_size = 0;
2809	struct ethtool_rxnfc __user *rxnfc = NULL;
2810	struct ifreq ifr;
2811	u32 rule_cnt = 0, actual_rule_cnt;
2812	u32 ethcmd;
2813	u32 data;
2814	int ret;
2815
2816	if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2817	return -EFAULT;
2818
2819	compat_rxnfc = compat_ptr(data);
2820
2821	if (get_user(ethcmd, &compat_rxnfc->cmd))
2822	return -EFAULT;
2823
2824	/* Most ethtool structures are defined without padding.
2825	* Unfortunately struct ethtool_rxnfc is an exception.
2826	*/
2827	switch (ethcmd) {
2828	default:
2829	break;
2830	case ETHTOOL_GRXCLSRLALL:
2831	/* Buffer size is variable */
2832	if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2833	return -EFAULT;
2834	if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2835	return -ENOMEM;
2836	buf_size += rule_cnt * sizeof(u32);
2837	/* fall through */
2838	case ETHTOOL_GRXRINGS:
2839	case ETHTOOL_GRXCLSRLCNT:
2840	case ETHTOOL_GRXCLSRULE:
2841	case ETHTOOL_SRXCLSRLINS:
2842	convert_out = true;
2843	/* fall through */
2844	case ETHTOOL_SRXCLSRLDEL:
2845	buf_size += sizeof(struct ethtool_rxnfc);
2846	convert_in = true;
2847	rxnfc = compat_alloc_user_space(buf_size);
2848	break;
2849	}
2850
2851	if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2852	return -EFAULT;
2853
2854	ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2855
2856	if (convert_in) {
2857	/* We expect there to be holes between fs.m_ext and
2858	* fs.ring_cookie and at the end of fs, but nowhere else.
2859	*/
2860	BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2861	sizeof(compat_rxnfc->fs.m_ext) !=
2862	offsetof(struct ethtool_rxnfc, fs.m_ext) +
2863	sizeof(rxnfc->fs.m_ext));
2864	BUILD_BUG_ON(
2865	offsetof(struct compat_ethtool_rxnfc, fs.location) -
2866	offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2867	offsetof(struct ethtool_rxnfc, fs.location) -
2868	offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2869
2870	if (copy_in_user(rxnfc, compat_rxnfc,
2871	(void __user *)(&rxnfc->fs.m_ext + 1) -
2872	(void __user *)rxnfc) \|\|
2873	copy_in_user(&rxnfc->fs.ring_cookie,
2874	&compat_rxnfc->fs.ring_cookie,
2875	(void __user *)(&rxnfc->fs.location + 1) -
2876	(void __user *)&rxnfc->fs.ring_cookie) \|\|
2877	copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2878	sizeof(rxnfc->rule_cnt)))
2879	return -EFAULT;
2880	}
2881
2882	ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2883	if (ret)
2884	return ret;
2885
2886	if (convert_out) {
2887	if (copy_in_user(compat_rxnfc, rxnfc,
2888	(const void __user *)(&rxnfc->fs.m_ext + 1) -
2889	(const void __user *)rxnfc) \|\|
2890	copy_in_user(&compat_rxnfc->fs.ring_cookie,
2891	&rxnfc->fs.ring_cookie,
2892	(const void __user *)(&rxnfc->fs.location + 1) -
2893	(const void __user *)&rxnfc->fs.ring_cookie) \|\|
2894	copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2895	sizeof(rxnfc->rule_cnt)))
2896	return -EFAULT;
2897
2898	if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2899	/* As an optimisation, we only copy the actual
2900	* number of rules that the underlying
2901	* function returned. Since Mallory might
2902	* change the rule count in user memory, we
2903	* check that it is less than the rule count
2904	* originally given (as the user buffer size),
2905	* which has been range-checked.
2906	*/
2907	if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2908	return -EFAULT;
2909	if (actual_rule_cnt < rule_cnt)
2910	rule_cnt = actual_rule_cnt;
2911	if (copy_in_user(&compat_rxnfc->rule_locs[0],
2912	&rxnfc->rule_locs[0],
2913	rule_cnt * sizeof(u32)))
2914	return -EFAULT;
2915	}
2916	}
2917
2918	return 0;
2919	}
2920
2921	static int compat_siocwandev(struct net net, struct compat_ifreq __user uifr32)
2922	{
2923	compat_uptr_t uptr32;
2924	struct ifreq ifr;
2925	void __user *saved;
2926	int err;
2927
2928	if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2929	return -EFAULT;
2930
2931	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2932	return -EFAULT;
2933
2934	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2935	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2936
2937	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2938	if (!err) {
2939	ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2940	if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2941	err = -EFAULT;
2942	}
2943	return err;
2944	}
2945
2946	/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2947	static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2948	struct compat_ifreq __user *u_ifreq32)
2949	{
2950	struct ifreq ifreq;
2951	u32 data32;
2952
2953	if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2954	return -EFAULT;
2955	if (get_user(data32, &u_ifreq32->ifr_data))
2956	return -EFAULT;
2957	ifreq.ifr_data = compat_ptr(data32);
2958
2959	return dev_ioctl(net, cmd, &ifreq, NULL);
2960	}
2961
2962	static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2963	struct compat_ifreq __user *uifr32)
2964	{
2965	struct ifreq ifr;
2966	struct compat_ifmap __user *uifmap32;
2967	int err;
2968
2969	uifmap32 = &uifr32->ifr_ifru.ifru_map;
2970	err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2971	err \|= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2972	err \|= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2973	err \|= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2974	err \|= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2975	err \|= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2976	err \|= get_user(ifr.ifr_map.port, &uifmap32->port);
2977	if (err)
2978	return -EFAULT;
2979
2980	err = dev_ioctl(net, cmd, &ifr, NULL);
2981
2982	if (cmd == SIOCGIFMAP && !err) {
2983	err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2984	err \|= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2985	err \|= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2986	err \|= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2987	err \|= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2988	err \|= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2989	err \|= put_user(ifr.ifr_map.port, &uifmap32->port);
2990	if (err)
2991	err = -EFAULT;
2992	}
2993	return err;
2994	}
2995
2996	struct rtentry32 {
2997	u32 rt_pad1;
2998	struct sockaddr rt_dst; /* target address */
2999	struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3000	struct sockaddr rt_genmask; /* target network mask (IP) */
3001	unsigned short rt_flags;
3002	short rt_pad2;
3003	u32 rt_pad3;
3004	unsigned char rt_tos;
3005	unsigned char rt_class;
3006	short rt_pad4;
3007	short rt_metric; /* +1 for binary compatibility! */
3008	/* char * / u32 rt_dev; / forcing the device at add */
3009	u32 rt_mtu; /* per route MTU/Window */
3010	u32 rt_window; /* Window clamping */
3011	unsigned short rt_irtt; /* Initial RTT */
3012	};
3013
3014	struct in6_rtmsg32 {
3015	struct in6_addr rtmsg_dst;
3016	struct in6_addr rtmsg_src;
3017	struct in6_addr rtmsg_gateway;
3018	u32 rtmsg_type;
3019	u16 rtmsg_dst_len;
3020	u16 rtmsg_src_len;
3021	u32 rtmsg_metric;
3022	u32 rtmsg_info;
3023	u32 rtmsg_flags;
3024	s32 rtmsg_ifindex;
3025	};
3026
3027	static int routing_ioctl(struct net net, struct socket sock,
3028	unsigned int cmd, void __user *argp)
3029	{
3030	int ret;
3031	void *r = NULL;
3032	struct in6_rtmsg r6;
3033	struct rtentry r4;
3034	char devname[16];
3035	u32 rtdev;
3036	mm_segment_t old_fs = get_fs();
3037
3038	if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3039	struct in6_rtmsg32 __user *ur6 = argp;
3040	ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3041	3 * sizeof(struct in6_addr));
3042	ret \|= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3043	ret \|= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3044	ret \|= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3045	ret \|= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3046	ret \|= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3047	ret \|= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3048	ret \|= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3049
3050	r = (void *) &r6;
3051	} else { /* ipv4 */
3052	struct rtentry32 __user *ur4 = argp;
3053	ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3054	3 * sizeof(struct sockaddr));
3055	ret \|= get_user(r4.rt_flags, &(ur4->rt_flags));
3056	ret \|= get_user(r4.rt_metric, &(ur4->rt_metric));
3057	ret \|= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3058	ret \|= get_user(r4.rt_window, &(ur4->rt_window));
3059	ret \|= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3060	ret \|= get_user(rtdev, &(ur4->rt_dev));
3061	if (rtdev) {
3062	ret \|= copy_from_user(devname, compat_ptr(rtdev), 15);
3063	r4.rt_dev = (char __user __force *)devname;
3064	devname[15] = 0;
3065	} else
3066	r4.rt_dev = NULL;
3067
3068	r = (void *) &r4;
3069	}
3070
3071	if (ret) {
3072	ret = -EFAULT;
3073	goto out;
3074	}
3075
3076	set_fs(KERNEL_DS);
3077	ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3078	set_fs(old_fs);
3079
3080	out:
3081	return ret;
3082	}
3083
3084	/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3085	* for some operations; this forces use of the newer bridge-utils that
3086	* use compatible ioctls
3087	*/
3088	static int old_bridge_ioctl(compat_ulong_t __user *argp)
3089	{
3090	compat_ulong_t tmp;
3091
3092	if (get_user(tmp, argp))
3093	return -EFAULT;
3094	if (tmp == BRCTL_GET_VERSION)
3095	return BRCTL_VERSION + 1;
3096	return -EINVAL;
3097	}
3098
3099	static int compat_sock_ioctl_trans(struct file file, struct socket sock,
3100	unsigned int cmd, unsigned long arg)
3101	{
3102	void __user *argp = compat_ptr(arg);
3103	struct sock *sk = sock->sk;
3104	struct net *net = sock_net(sk);
3105
3106	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3107	return compat_ifr_data_ioctl(net, cmd, argp);
3108
3109	switch (cmd) {
3110	case SIOCSIFBR:
3111	case SIOCGIFBR:
3112	return old_bridge_ioctl(argp);
3113	case SIOCGIFCONF:
3114	return compat_dev_ifconf(net, argp);
3115	case SIOCETHTOOL:
3116	return ethtool_ioctl(net, argp);
3117	case SIOCWANDEV:
3118	return compat_siocwandev(net, argp);
3119	case SIOCGIFMAP:
3120	case SIOCSIFMAP:
3121	return compat_sioc_ifmap(net, cmd, argp);
3122	case SIOCADDRT:
3123	case SIOCDELRT:
3124	return routing_ioctl(net, sock, cmd, argp);
3125	case SIOCGSTAMP:
3126	return do_siocgstamp(net, sock, cmd, argp);
3127	case SIOCGSTAMPNS:
3128	return do_siocgstampns(net, sock, cmd, argp);
3129	case SIOCBONDSLAVEINFOQUERY:
3130	case SIOCBONDINFOQUERY:
3131	case SIOCSHWTSTAMP:
3132	case SIOCGHWTSTAMP:
3133	return compat_ifr_data_ioctl(net, cmd, argp);
3134
3135	case FIOSETOWN:
3136	case SIOCSPGRP:
3137	case FIOGETOWN:
3138	case SIOCGPGRP:
3139	case SIOCBRADDBR:
3140	case SIOCBRDELBR:
3141	case SIOCGIFVLAN:
3142	case SIOCSIFVLAN:
3143	case SIOCADDDLCI:
3144	case SIOCDELDLCI:
3145	case SIOCGSKNS:
3146	return sock_ioctl(file, cmd, arg);
3147
3148	case SIOCGIFFLAGS:
3149	case SIOCSIFFLAGS:
3150	case SIOCGIFMETRIC:
3151	case SIOCSIFMETRIC:
3152	case SIOCGIFMTU:
3153	case SIOCSIFMTU:
3154	case SIOCGIFMEM:
3155	case SIOCSIFMEM:
3156	case SIOCGIFHWADDR:
3157	case SIOCSIFHWADDR:
3158	case SIOCADDMULTI:
3159	case SIOCDELMULTI:
3160	case SIOCGIFINDEX:
3161	case SIOCGIFADDR:
3162	case SIOCSIFADDR:
3163	case SIOCSIFHWBROADCAST:
3164	case SIOCDIFADDR:
3165	case SIOCGIFBRDADDR:
3166	case SIOCSIFBRDADDR:
3167	case SIOCGIFDSTADDR:
3168	case SIOCSIFDSTADDR:
3169	case SIOCGIFNETMASK:
3170	case SIOCSIFNETMASK:
3171	case SIOCSIFPFLAGS:
3172	case SIOCGIFPFLAGS:
3173	case SIOCGIFTXQLEN:
3174	case SIOCSIFTXQLEN:
3175	case SIOCBRADDIF:
3176	case SIOCBRDELIF:
3177	case SIOCSIFNAME:
3178	case SIOCGMIIPHY:
3179	case SIOCGMIIREG:
3180	case SIOCSMIIREG:
3181	case SIOCSARP:
3182	case SIOCGARP:
3183	case SIOCDARP:
3184	case SIOCATMARK:
3185	case SIOCBONDENSLAVE:
3186	case SIOCBONDRELEASE:
3187	case SIOCBONDSETHWADDR:
3188	case SIOCBONDCHANGEACTIVE:
3189	case SIOCGIFNAME:
3190	return sock_do_ioctl(net, sock, cmd, arg);
3191	}
3192
3193	return -ENOIOCTLCMD;
3194	}
3195
3196	static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3197	unsigned long arg)
3198	{
3199	struct socket *sock = file->private_data;
3200	int ret = -ENOIOCTLCMD;
3201	struct sock *sk;
3202	struct net *net;
3203
3204	sk = sock->sk;
3205	net = sock_net(sk);
3206
3207	if (sock->ops->compat_ioctl)
3208	ret = sock->ops->compat_ioctl(sock, cmd, arg);
3209
3210	if (ret == -ENOIOCTLCMD &&
3211	(cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3212	ret = compat_wext_handle_ioctl(net, cmd, arg);
3213
3214	if (ret == -ENOIOCTLCMD)
3215	ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3216
3217	return ret;
3218	}
3219	#endif
3220
3221	int kernel_bind(struct socket sock, struct sockaddr addr, int addrlen)
3222	{
3223	return sock->ops->bind(sock, addr, addrlen);
3224	}
3225	EXPORT_SYMBOL(kernel_bind);
3226
3227	int kernel_listen(struct socket *sock, int backlog)
3228	{
3229	return sock->ops->listen(sock, backlog);
3230	}
3231	EXPORT_SYMBOL(kernel_listen);
3232
3233	int kernel_accept(struct socket sock, struct socket *newsock, int flags)
3234	{
3235	struct sock *sk = sock->sk;
3236	int err;
3237
3238	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3239	newsock);
3240	if (err < 0)
3241	goto done;
3242
3243	err = sock->ops->accept(sock, *newsock, flags, true);
3244	if (err < 0) {
3245	sock_release(*newsock);
3246	*newsock = NULL;
3247	goto done;
3248	}
3249
3250	(*newsock)->ops = sock->ops;
3251	__module_get((*newsock)->ops->owner);
3252
3253	done:
3254	return err;
3255	}
3256	EXPORT_SYMBOL(kernel_accept);
3257
3258	int kernel_connect(struct socket sock, struct sockaddr addr, int addrlen,
3259	int flags)
3260	{
3261	return sock->ops->connect(sock, addr, addrlen, flags);
3262	}
3263	EXPORT_SYMBOL(kernel_connect);
3264
3265	int kernel_getsockname(struct socket sock, struct sockaddr addr)
3266	{
3267	return sock->ops->getname(sock, addr, 0);
3268	}
3269	EXPORT_SYMBOL(kernel_getsockname);
3270
3271	int kernel_getpeername(struct socket sock, struct sockaddr addr)
3272	{
3273	return sock->ops->getname(sock, addr, 1);
3274	}
3275	EXPORT_SYMBOL(kernel_getpeername);
3276
3277	int kernel_getsockopt(struct socket *sock, int level, int optname,
3278	char optval, int optlen)
3279	{
3280	mm_segment_t oldfs = get_fs();
3281	char __user *uoptval;
3282	int __user *uoptlen;
3283	int err;
3284
3285	uoptval = (char __user __force *) optval;
3286	uoptlen = (int __user __force *) optlen;
3287
3288	set_fs(KERNEL_DS);
3289	if (level == SOL_SOCKET)
3290	err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3291	else
3292	err = sock->ops->getsockopt(sock, level, optname, uoptval,
3293	uoptlen);
3294	set_fs(oldfs);
3295	return err;
3296	}
3297	EXPORT_SYMBOL(kernel_getsockopt);
3298
3299	int kernel_setsockopt(struct socket *sock, int level, int optname,
3300	char *optval, unsigned int optlen)
3301	{
3302	mm_segment_t oldfs = get_fs();
3303	char __user *uoptval;
3304	int err;
3305
3306	uoptval = (char __user __force *) optval;
3307
3308	set_fs(KERNEL_DS);
3309	if (level == SOL_SOCKET)
3310	err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3311	else
3312	err = sock->ops->setsockopt(sock, level, optname, uoptval,
3313	optlen);
3314	set_fs(oldfs);
3315	return err;
3316	}
3317	EXPORT_SYMBOL(kernel_setsockopt);
3318
3319	int kernel_sendpage(struct socket sock, struct page page, int offset,
3320	size_t size, int flags)
3321	{
3322	if (sock->ops->sendpage)
3323	return sock->ops->sendpage(sock, page, offset, size, flags);
3324
3325	return sock_no_sendpage(sock, page, offset, size, flags);
3326	}
3327	EXPORT_SYMBOL(kernel_sendpage);
3328
3329	int kernel_sendpage_locked(struct sock sk, struct page page, int offset,
3330	size_t size, int flags)
3331	{
3332	struct socket *sock = sk->sk_socket;
3333
3334	if (sock->ops->sendpage_locked)
3335	return sock->ops->sendpage_locked(sk, page, offset, size,
3336	flags);
3337
3338	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3339	}
3340	EXPORT_SYMBOL(kernel_sendpage_locked);
3341
3342	int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3343	{
3344	return sock->ops->shutdown(sock, how);
3345	}
3346	EXPORT_SYMBOL(kernel_sock_shutdown);
3347
3348	/* This routine returns the IP overhead imposed by a socket i.e.
3349	* the length of the underlying IP header, depending on whether
3350	* this is an IPv4 or IPv6 socket and the length from IP options turned
3351	* on at the socket. Assumes that the caller has a lock on the socket.
3352	*/
3353	u32 kernel_sock_ip_overhead(struct sock *sk)
3354	{
3355	struct inet_sock *inet;
3356	struct ip_options_rcu *opt;
3357	u32 overhead = 0;
3358	#if IS_ENABLED(CONFIG_IPV6)
3359	struct ipv6_pinfo *np;
3360	struct ipv6_txoptions *optv6 = NULL;
3361	#endif /* IS_ENABLED(CONFIG_IPV6) */
3362
3363	if (!sk)
3364	return overhead;
3365
3366	switch (sk->sk_family) {
3367	case AF_INET:
3368	inet = inet_sk(sk);
3369	overhead += sizeof(struct iphdr);
3370	opt = rcu_dereference_protected(inet->inet_opt,
3371	sock_owned_by_user(sk));
3372	if (opt)
3373	overhead += opt->opt.optlen;
3374	return overhead;
3375	#if IS_ENABLED(CONFIG_IPV6)
3376	case AF_INET6:
3377	np = inet6_sk(sk);
3378	overhead += sizeof(struct ipv6hdr);
3379	if (np)
3380	optv6 = rcu_dereference_protected(np->opt,
3381	sock_owned_by_user(sk));
3382	if (optv6)
3383	overhead += (optv6->opt_flen + optv6->opt_nflen);
3384	return overhead;
3385	#endif /* IS_ENABLED(CONFIG_IPV6) */
3386	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3387	return overhead;
3388	}
3389	}
3390	EXPORT_SYMBOL(kernel_sock_ip_overhead);