blob: 977c0f48c8d6c3efa655a91a69f7911445b313a9 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
Jesper Juhl02c30a82005-05-05 16:16:16 -07007 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
Stephen Hemminger89bddce2006-09-01 00:19:31 -070045 * Tigran Aivazian : Made listen(2) backlog sanity checks
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 * 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
Stephen Hemminger89bddce2006-09-01 00:19:31 -070056 * paradigm.
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070066#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080071#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030074#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070088#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070089#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090090#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000091#include <linux/xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93#include <asm/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070097#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070098#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
Arnd Bergmann6b960182009-11-06 23:10:54 -0800103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800106#include <linux/sockios.h>
107#include <linux/atalk.h>
108
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700110static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
111 unsigned long nr_segs, loff_t pos);
112static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
113 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700114static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
116static int sock_close(struct inode *inode, struct file *file);
117static unsigned int sock_poll(struct file *file,
118 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700119static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800120#ifdef CONFIG_COMPAT
121static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700122 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800123#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125static ssize_t sock_sendpage(struct file *file, struct page *page,
126 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800127static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700128 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800129 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/*
132 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
133 * in the operation structures but are done directly via the socketcall() multiplexor.
134 */
135
Arjan van de Venda7071d2007-02-12 00:55:36 -0800136static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 .owner = THIS_MODULE,
138 .llseek = no_llseek,
139 .aio_read = sock_aio_read,
140 .aio_write = sock_aio_write,
141 .poll = sock_poll,
142 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800143#ifdef CONFIG_COMPAT
144 .compat_ioctl = compat_sock_ioctl,
145#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 .mmap = sock_mmap,
147 .open = sock_no_open, /* special open code to disallow open via /proc */
148 .release = sock_close,
149 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200150 .sendpage = sock_sendpage,
151 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800152 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153};
154
155/*
156 * The protocol list. Each protocol is registered in here.
157 */
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000160static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162/*
163 * Statistics counters of the socket lists
164 */
165
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700166static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
168/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700169 * Support routines.
170 * Move socket addresses back and forth across the kernel/user
171 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 */
173
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174/**
175 * move_addr_to_kernel - copy a socket address into kernel space
176 * @uaddr: Address in user space
177 * @kaddr: Address in kernel space
178 * @ulen: Length in user space
179 *
180 * The address is copied into kernel space. If the provided address is
181 * too long an error code of -EINVAL is returned. If the copy gives
182 * invalid addresses -EFAULT is returned. On a success 0 is returned.
183 */
184
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000185int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700187 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700189 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700191 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100193 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194}
195
196/**
197 * move_addr_to_user - copy an address to user space
198 * @kaddr: kernel space address
199 * @klen: length of address in kernel
200 * @uaddr: user space address
201 * @ulen: pointer to user length field
202 *
203 * The value pointed to by ulen on entry is the buffer length available.
204 * This is overwritten with the buffer space used. -EINVAL is returned
205 * if an overlong buffer is specified or a negative buffer size. -EFAULT
206 * is returned if either the buffer or the length field are not
207 * accessible.
208 * After copying the data up to the limit the user specifies, the true
209 * length of the data is written over the length limit the user
210 * specified. Zero is returned for a success.
211 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700212
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000213static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000214 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 int err;
217 int len;
218
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700219 err = get_user(len, ulen);
220 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700222 if (len > klen)
223 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700224 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700226 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500227 if (audit_sockaddr(klen, kaddr))
228 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700229 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return -EFAULT;
231 }
232 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700233 * "fromlen shall refer to the value before truncation.."
234 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 */
236 return __put_user(klen, ulen);
237}
238
Christoph Lametere18b8902006-12-06 20:33:20 -0800239static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241static struct inode *sock_alloc_inode(struct super_block *sb)
242{
243 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000244 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700245
Christoph Lametere94b1762006-12-06 20:33:17 -0800246 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 if (!ei)
248 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000249 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
250 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000251 kmem_cache_free(sock_inode_cachep, ei);
252 return NULL;
253 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000254 init_waitqueue_head(&wq->wait);
255 wq->fasync_list = NULL;
256 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700257
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 ei->socket.state = SS_UNCONNECTED;
259 ei->socket.flags = 0;
260 ei->socket.ops = NULL;
261 ei->socket.sk = NULL;
262 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
264 return &ei->vfs_inode;
265}
266
267static void sock_destroy_inode(struct inode *inode)
268{
Eric Dumazet43815482010-04-29 11:01:49 +0000269 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000270 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000271
272 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000273 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800274 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000275 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276}
277
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700278static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700280 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
Christoph Lametera35afb82007-05-16 22:10:57 -0700282 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285static int init_inodecache(void)
286{
287 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700288 sizeof(struct socket_alloc),
289 0,
290 (SLAB_HWCACHE_ALIGN |
291 SLAB_RECLAIM_ACCOUNT |
292 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900293 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 if (sock_inode_cachep == NULL)
295 return -ENOMEM;
296 return 0;
297}
298
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700299static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700300 .alloc_inode = sock_alloc_inode,
301 .destroy_inode = sock_destroy_inode,
302 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303};
304
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700305/*
306 * sockfs_dname() is called from d_path().
307 */
308static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
309{
310 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
311 dentry->d_inode->i_ino);
312}
313
Al Viro3ba13d12009-02-20 06:02:22 +0000314static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700315 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316};
317
Al Viroc74a1cb2011-01-12 16:59:34 -0500318static struct dentry *sockfs_mount(struct file_system_type *fs_type,
319 int flags, const char *dev_name, void *data)
320{
321 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
322 &sockfs_dentry_operations, SOCKFS_MAGIC);
323}
324
325static struct vfsmount *sock_mnt __read_mostly;
326
327static struct file_system_type sock_fs_type = {
328 .name = "sockfs",
329 .mount = sockfs_mount,
330 .kill_sb = kill_anon_super,
331};
332
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333/*
334 * Obtains the first available file descriptor and sets it up for use.
335 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800336 * These functions create file structures and maps them to fd space
337 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 * and file struct implicitly stored in sock->file.
339 * Note that another thread may close file descriptor before we return
340 * from this function. We use the fact that now we do not refer
341 * to socket after mapping. If one day we will need it, this
342 * function will increment ref. count on file by 1.
343 *
344 * In any case returned fd MAY BE not valid!
345 * This race condition is unavoidable
346 * with shared fd spaces, we cannot solve it inside kernel,
347 * but we take care of internal coherence yet.
348 */
349
Masatake YAMATO600e1772012-08-29 10:44:29 +0000350static int sock_alloc_file(struct socket *sock, struct file **f, int flags,
351 const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352{
Al Viro7cbe66b2009-08-05 19:59:08 +0400353 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400354 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400355 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800357
Ulrich Dreppera677a032008-07-23 21:29:17 -0700358 fd = get_unused_fd_flags(flags);
Al Viro7cbe66b2009-08-05 19:59:08 +0400359 if (unlikely(fd < 0))
360 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800361
Masatake YAMATO600e1772012-08-29 10:44:29 +0000362 if (dname) {
363 name.name = dname;
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
368 }
Nick Piggin4b936882011-01-07 17:50:07 +1100369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro2c48b9c2009-08-09 00:52:35 +0400370 if (unlikely(!path.dentry)) {
Al Viro7cbe66b2009-08-05 19:59:08 +0400371 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800372 return -ENOMEM;
Al Viro7cbe66b2009-08-05 19:59:08 +0400373 }
Al Viro2c48b9c2009-08-09 00:52:35 +0400374 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
Al Viro2c48b9c2009-08-09 00:52:35 +0400376 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400377 SOCK_INODE(sock)->i_fop = &socket_file_ops;
378
Al Viro2c48b9c2009-08-09 00:52:35 +0400379 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400380 &socket_file_ops);
381 if (unlikely(!file)) {
382 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400383 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400384 path_put(&path);
Al Virocc3808f2009-08-06 09:43:59 +0400385 put_unused_fd(fd);
386 return -ENFILE;
387 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800388
389 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700390 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800391 file->f_pos = 0;
392 file->private_data = sock;
393
Al Viro7cbe66b2009-08-05 19:59:08 +0400394 *f = file;
395 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800396}
397
Ulrich Dreppera677a032008-07-23 21:29:17 -0700398int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800399{
400 struct file *newfile;
Masatake YAMATO600e1772012-08-29 10:44:29 +0000401 int fd = sock_alloc_file(sock, &newfile, flags, NULL);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800402
Al Viro7cbe66b2009-08-05 19:59:08 +0400403 if (likely(fd >= 0))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800404 fd_install(fd, newfile);
Al Viro7cbe66b2009-08-05 19:59:08 +0400405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 return fd;
407}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700408EXPORT_SYMBOL(sock_map_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
John Fastabend406a3c62012-07-20 10:39:25 +0000410struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800411{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800412 if (file->f_op == &socket_file_ops)
413 return file->private_data; /* set in sock_map_fd */
414
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800415 *err = -ENOTSOCK;
416 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800417}
John Fastabend406a3c62012-07-20 10:39:25 +0000418EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800419
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700421 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 * @fd: file handle
423 * @err: pointer to an error code return
424 *
425 * The file handle passed in is locked and the socket it is bound
426 * too is returned. If an error occurs the err pointer is overwritten
427 * with a negative errno code and NULL is returned. The function checks
428 * for both invalid handles and passing a handle which is not a socket.
429 *
430 * On a success the socket object pointer is returned.
431 */
432
433struct socket *sockfd_lookup(int fd, int *err)
434{
435 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 struct socket *sock;
437
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700438 file = fget(fd);
439 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 *err = -EBADF;
441 return NULL;
442 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700443
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800444 sock = sock_from_file(file, err);
445 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 return sock;
448}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700449EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452{
453 struct file *file;
454 struct socket *sock;
455
Hua Zhong36725582006-04-19 15:25:02 -0700456 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800457 file = fget_light(fd, fput_needed);
458 if (file) {
459 sock = sock_from_file(file, err);
460 if (sock)
461 return sock;
462 fput_light(file, *fput_needed);
463 }
464 return NULL;
465}
466
Masatake YAMATO600e1772012-08-29 10:44:29 +0000467#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
472{
473 const char *proto_name;
474 size_t proto_size;
475 int error;
476
477 error = -ENODATA;
478 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 proto_name = dentry->d_name.name;
480 proto_size = strlen(proto_name);
481
482 if (value) {
483 error = -ERANGE;
484 if (proto_size + 1 > size)
485 goto out;
486
487 strncpy(value, proto_name, proto_size + 1);
488 }
489 error = proto_size + 1;
490 }
491
492out:
493 return error;
494}
495
496static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
497 size_t size)
498{
499 ssize_t len;
500 ssize_t used = 0;
501
502 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
503 if (len < 0)
504 return len;
505 used += len;
506 if (buffer) {
507 if (size < used)
508 return -ERANGE;
509 buffer += len;
510 }
511
512 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
513 used += len;
514 if (buffer) {
515 if (size < used)
516 return -ERANGE;
517 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
518 buffer += len;
519 }
520
521 return used;
522}
523
524static const struct inode_operations sockfs_inode_ops = {
525 .getxattr = sockfs_getxattr,
526 .listxattr = sockfs_listxattr,
527};
528
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529/**
530 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700531 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
534 * NULL is returned.
535 */
536
537static struct socket *sock_alloc(void)
538{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700539 struct inode *inode;
540 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200542 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 if (!inode)
544 return NULL;
545
546 sock = SOCKET_I(inode);
547
Eric Dumazet29a020d2009-09-15 02:39:20 -0700548 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400549 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700550 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100551 inode->i_uid = current_fsuid();
552 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000553 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
Alex Shi19e8d692012-05-14 14:15:31 -0700555 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 return sock;
557}
558
559/*
560 * In theory you can't get an open on this inode, but /proc provides
561 * a back door. Remember to keep it shut otherwise you'll let the
562 * creepy crawlies in.
563 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
566{
567 return -ENXIO;
568}
569
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800570const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 .owner = THIS_MODULE,
572 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200573 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574};
575
576/**
577 * sock_release - close a socket
578 * @sock: socket to close
579 *
580 * The socket is released from the protocol stack if it has a release
581 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700582 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585void sock_release(struct socket *sock)
586{
587 if (sock->ops) {
588 struct module *owner = sock->ops->owner;
589
590 sock->ops->release(sock);
591 sock->ops = NULL;
592 module_put(owner);
593 }
594
Eric Dumazeteaefd112011-02-18 03:26:36 +0000595 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 printk(KERN_ERR "sock_release: fasync list not empty!\n");
597
Mikulas Patockab09e7862012-07-19 06:13:36 +0000598 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
599 return;
600
Alex Shi19e8d692012-05-14 14:15:31 -0700601 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 if (!sock->file) {
603 iput(SOCK_INODE(sock));
604 return;
605 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700606 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700608EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000610int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000611{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000612 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000613 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000614 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000615 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000616 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100617 if (sock_flag(sk, SOCK_WIFI_STATUS))
618 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000619 return 0;
620}
621EXPORT_SYMBOL(sock_tx_timestamp);
622
Anton Blanchard228e5482011-05-02 20:21:35 +0000623static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
624 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625{
626 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Herbert Xuf8451722010-05-24 00:12:34 -0700628 sock_update_classid(sock->sk);
629
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 si->sock = sock;
631 si->scm = NULL;
632 si->msg = msg;
633 si->size = size;
634
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 return sock->ops->sendmsg(iocb, sock, msg, size);
636}
637
Anton Blanchard228e5482011-05-02 20:21:35 +0000638static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
639 struct msghdr *msg, size_t size)
640{
641 int err = security_socket_sendmsg(sock, msg, size);
642
643 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
644}
645
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
647{
648 struct kiocb iocb;
649 struct sock_iocb siocb;
650 int ret;
651
652 init_sync_kiocb(&iocb, NULL);
653 iocb.private = &siocb;
654 ret = __sock_sendmsg(&iocb, sock, msg, size);
655 if (-EIOCBQUEUED == ret)
656 ret = wait_on_sync_kiocb(&iocb);
657 return ret;
658}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700659EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Eric Dumazet894dc242011-07-26 02:39:41 +0000661static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000662{
663 struct kiocb iocb;
664 struct sock_iocb siocb;
665 int ret;
666
667 init_sync_kiocb(&iocb, NULL);
668 iocb.private = &siocb;
669 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
670 if (-EIOCBQUEUED == ret)
671 ret = wait_on_sync_kiocb(&iocb);
672 return ret;
673}
674
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
676 struct kvec *vec, size_t num, size_t size)
677{
678 mm_segment_t oldfs = get_fs();
679 int result;
680
681 set_fs(KERNEL_DS);
682 /*
683 * the following is safe, since for compiler definitions of kvec and
684 * iovec are identical, yielding the same in-core layout and alignment
685 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700686 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 msg->msg_iovlen = num;
688 result = sock_sendmsg(sock, msg, size);
689 set_fs(oldfs);
690 return result;
691}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700692EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
Patrick Ohly20d49472009-02-12 05:03:38 +0000694static int ktime2ts(ktime_t kt, struct timespec *ts)
695{
696 if (kt.tv64) {
697 *ts = ktime_to_timespec(kt);
698 return 1;
699 } else {
700 return 0;
701 }
702}
703
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700704/*
705 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
706 */
707void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
708 struct sk_buff *skb)
709{
Patrick Ohly20d49472009-02-12 05:03:38 +0000710 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
711 struct timespec ts[3];
712 int empty = 1;
713 struct skb_shared_hwtstamps *shhwtstamps =
714 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700715
Patrick Ohly20d49472009-02-12 05:03:38 +0000716 /* Race occurred between timestamp enabling and packet
717 receiving. Fill in the current time for now. */
718 if (need_software_tstamp && skb->tstamp.tv64 == 0)
719 __net_timestamp(skb);
720
721 if (need_software_tstamp) {
722 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
723 struct timeval tv;
724 skb_get_timestamp(skb, &tv);
725 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
726 sizeof(tv), &tv);
727 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000728 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000729 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000730 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000731 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700732 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000733
734
735 memset(ts, 0, sizeof(ts));
736 if (skb->tstamp.tv64 &&
737 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
738 skb_get_timestampns(skb, ts + 0);
739 empty = 0;
740 }
741 if (shhwtstamps) {
742 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
743 ktime2ts(shhwtstamps->syststamp, ts + 1))
744 empty = 0;
745 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
746 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
747 empty = 0;
748 }
749 if (!empty)
750 put_cmsg(msg, SOL_SOCKET,
751 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700752}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300753EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
754
Johannes Berg6e3e9392011-11-09 10:15:42 +0100755void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
756 struct sk_buff *skb)
757{
758 int ack;
759
760 if (!sock_flag(sk, SOCK_WIFI_STATUS))
761 return;
762 if (!skb->wifi_acked_valid)
763 return;
764
765 ack = skb->wifi_acked;
766
767 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
768}
769EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
770
stephen hemminger11165f12010-10-18 14:27:29 +0000771static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
772 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700773{
774 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
775 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
776 sizeof(__u32), &skb->dropcount);
777}
778
Eric Dumazet767dd032010-04-28 19:14:43 +0000779void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700780 struct sk_buff *skb)
781{
782 sock_recv_timestamp(msg, sk, skb);
783 sock_recv_drops(msg, sk, skb);
784}
Eric Dumazet767dd032010-04-28 19:14:43 +0000785EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700786
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700787static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
788 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 struct sock_iocb *si = kiocb_to_siocb(iocb);
791
Herbert Xuf8451722010-05-24 00:12:34 -0700792 sock_update_classid(sock->sk);
793
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 si->sock = sock;
795 si->scm = NULL;
796 si->msg = msg;
797 si->size = size;
798 si->flags = flags;
799
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
801}
802
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700803static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
804 struct msghdr *msg, size_t size, int flags)
805{
806 int err = security_socket_recvmsg(sock, msg, size, flags);
807
808 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
809}
810
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700811int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 size_t size, int flags)
813{
814 struct kiocb iocb;
815 struct sock_iocb siocb;
816 int ret;
817
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700818 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 iocb.private = &siocb;
820 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
821 if (-EIOCBQUEUED == ret)
822 ret = wait_on_sync_kiocb(&iocb);
823 return ret;
824}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700825EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700827static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
828 size_t size, int flags)
829{
830 struct kiocb iocb;
831 struct sock_iocb siocb;
832 int ret;
833
834 init_sync_kiocb(&iocb, NULL);
835 iocb.private = &siocb;
836 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
837 if (-EIOCBQUEUED == ret)
838 ret = wait_on_sync_kiocb(&iocb);
839 return ret;
840}
841
Martin Lucinac1249c02010-12-10 00:04:05 +0000842/**
843 * kernel_recvmsg - Receive a message from a socket (kernel space)
844 * @sock: The socket to receive the message from
845 * @msg: Received message
846 * @vec: Input s/g array for message data
847 * @num: Size of input s/g array
848 * @size: Number of bytes to read
849 * @flags: Message flags (MSG_DONTWAIT, etc...)
850 *
851 * On return the msg structure contains the scatter/gather array passed in the
852 * vec argument. The array is modified so that it consists of the unfilled
853 * portion of the original array.
854 *
855 * The returned value is the total number of bytes received, or an error.
856 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700857int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
858 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859{
860 mm_segment_t oldfs = get_fs();
861 int result;
862
863 set_fs(KERNEL_DS);
864 /*
865 * the following is safe, since for compiler definitions of kvec and
866 * iovec are identical, yielding the same in-core layout and alignment
867 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700868 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 result = sock_recvmsg(sock, msg, size, flags);
870 set_fs(oldfs);
871 return result;
872}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700873EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
875static void sock_aio_dtor(struct kiocb *iocb)
876{
877 kfree(iocb->private);
878}
879
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300880static ssize_t sock_sendpage(struct file *file, struct page *page,
881 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
883 struct socket *sock;
884 int flags;
885
Eric Dumazetb69aee02005-09-06 14:42:45 -0700886 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
Eric Dumazet35f9c092012-04-05 03:05:35 +0000888 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
889 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
890 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
Linus Torvaldse6949582009-08-13 08:28:36 -0700892 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893}
894
Jens Axboe9c55e012007-11-06 23:30:13 -0800895static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700896 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800897 unsigned int flags)
898{
899 struct socket *sock = file->private_data;
900
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800901 if (unlikely(!sock->ops->splice_read))
902 return -EINVAL;
903
Herbert Xuf8451722010-05-24 00:12:34 -0700904 sock_update_classid(sock->sk);
905
Jens Axboe9c55e012007-11-06 23:30:13 -0800906 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
907}
908
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800909static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700910 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800911{
912 if (!is_sync_kiocb(iocb)) {
913 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
914 if (!siocb)
915 return NULL;
916 iocb->ki_dtor = sock_aio_dtor;
917 }
918
919 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800920 iocb->private = siocb;
921 return siocb;
922}
923
924static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700925 struct file *file, const struct iovec *iov,
926 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800927{
928 struct socket *sock = file->private_data;
929 size_t size = 0;
930 int i;
931
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700932 for (i = 0; i < nr_segs; i++)
933 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800934
935 msg->msg_name = NULL;
936 msg->msg_namelen = 0;
937 msg->msg_control = NULL;
938 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700939 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800940 msg->msg_iovlen = nr_segs;
941 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
942
943 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
944}
945
Badari Pulavarty027445c2006-09-30 23:28:46 -0700946static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
947 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800948{
949 struct sock_iocb siocb, *x;
950
951 if (pos != 0)
952 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700953
954 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800955 return 0;
956
Badari Pulavarty027445c2006-09-30 23:28:46 -0700957
958 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800959 if (!x)
960 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700961 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800962}
963
964static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700965 struct file *file, const struct iovec *iov,
966 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800967{
968 struct socket *sock = file->private_data;
969 size_t size = 0;
970 int i;
971
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700972 for (i = 0; i < nr_segs; i++)
973 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800974
975 msg->msg_name = NULL;
976 msg->msg_namelen = 0;
977 msg->msg_control = NULL;
978 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700979 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800980 msg->msg_iovlen = nr_segs;
981 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
982 if (sock->type == SOCK_SEQPACKET)
983 msg->msg_flags |= MSG_EOR;
984
985 return __sock_sendmsg(iocb, sock, msg, size);
986}
987
Badari Pulavarty027445c2006-09-30 23:28:46 -0700988static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
989 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800991 struct sock_iocb siocb, *x;
992
993 if (pos != 0)
994 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700995
Badari Pulavarty027445c2006-09-30 23:28:46 -0700996 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800997 if (!x)
998 return -ENOMEM;
999
Badari Pulavarty027445c2006-09-30 23:28:46 -07001000 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001}
1002
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003/*
1004 * Atomic setting of ioctl hooks to avoid race
1005 * with module unload.
1006 */
1007
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001008static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001009static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Eric W. Biederman881d9662007-09-17 11:56:21 -07001011void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001013 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001015 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016}
1017EXPORT_SYMBOL(brioctl_set);
1018
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001019static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001020static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
Eric W. Biederman881d9662007-09-17 11:56:21 -07001022void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001024 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001026 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027}
1028EXPORT_SYMBOL(vlan_ioctl_set);
1029
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001030static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001031static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001033void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001035 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001037 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038}
1039EXPORT_SYMBOL(dlci_ioctl_set);
1040
Arnd Bergmann6b960182009-11-06 23:10:54 -08001041static long sock_do_ioctl(struct net *net, struct socket *sock,
1042 unsigned int cmd, unsigned long arg)
1043{
1044 int err;
1045 void __user *argp = (void __user *)arg;
1046
1047 err = sock->ops->ioctl(sock, cmd, arg);
1048
1049 /*
1050 * If this ioctl is unknown try to hand it down
1051 * to the NIC driver.
1052 */
1053 if (err == -ENOIOCTLCMD)
1054 err = dev_ioctl(net, cmd, argp);
1055
1056 return err;
1057}
1058
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059/*
1060 * With an ioctl, arg may well be a user mode pointer, but we don't know
1061 * what to do with it - that's up to the protocol still.
1062 */
1063
1064static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1065{
1066 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001067 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 void __user *argp = (void __user *)arg;
1069 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001070 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Eric Dumazetb69aee02005-09-06 14:42:45 -07001072 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001073 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001074 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001076 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001078#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001080 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001082#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001083 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 case FIOSETOWN:
1085 case SIOCSPGRP:
1086 err = -EFAULT;
1087 if (get_user(pid, (int __user *)argp))
1088 break;
1089 err = f_setown(sock->file, pid, 1);
1090 break;
1091 case FIOGETOWN:
1092 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001093 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001094 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 break;
1096 case SIOCGIFBR:
1097 case SIOCSIFBR:
1098 case SIOCBRADDBR:
1099 case SIOCBRDELBR:
1100 err = -ENOPKG;
1101 if (!br_ioctl_hook)
1102 request_module("bridge");
1103
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001104 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001105 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001106 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001107 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 break;
1109 case SIOCGIFVLAN:
1110 case SIOCSIFVLAN:
1111 err = -ENOPKG;
1112 if (!vlan_ioctl_hook)
1113 request_module("8021q");
1114
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001115 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001117 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001118 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 case SIOCADDDLCI:
1121 case SIOCDELDLCI:
1122 err = -ENOPKG;
1123 if (!dlci_ioctl_hook)
1124 request_module("dlci");
1125
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001126 mutex_lock(&dlci_ioctl_mutex);
1127 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001129 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 break;
1131 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001132 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001134 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 return err;
1136}
1137
1138int sock_create_lite(int family, int type, int protocol, struct socket **res)
1139{
1140 int err;
1141 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001142
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 err = security_socket_create(family, type, protocol, 1);
1144 if (err)
1145 goto out;
1146
1147 sock = sock_alloc();
1148 if (!sock) {
1149 err = -ENOMEM;
1150 goto out;
1151 }
1152
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001154 err = security_socket_post_create(sock, family, type, protocol, 1);
1155 if (err)
1156 goto out_release;
1157
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158out:
1159 *res = sock;
1160 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001161out_release:
1162 sock_release(sock);
1163 sock = NULL;
1164 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001166EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
1168/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001169static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170{
1171 struct socket *sock;
1172
1173 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001174 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001176 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 return sock->ops->poll(file, sock, wait);
1178}
1179
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001180static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001182 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
1184 return sock->ops->mmap(file, sock, vma);
1185}
1186
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001187static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188{
1189 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001190 * It was possible the inode is NULL we were
1191 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 */
1193
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001194 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 printk(KERN_DEBUG "sock_close: NULL inode\n");
1196 return 0;
1197 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 sock_release(SOCKET_I(inode));
1199 return 0;
1200}
1201
1202/*
1203 * Update the socket async list
1204 *
1205 * Fasync_list locking strategy.
1206 *
1207 * 1. fasync_list is modified only under process context socket lock
1208 * i.e. under semaphore.
1209 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001210 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 */
1212
1213static int sock_fasync(int fd, struct file *filp, int on)
1214{
Eric Dumazet989a2972010-04-14 09:55:35 +00001215 struct socket *sock = filp->private_data;
1216 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001217 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218
Eric Dumazet989a2972010-04-14 09:55:35 +00001219 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221
1222 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001223 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1224 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225
Eric Dumazeteaefd112011-02-18 03:26:36 +00001226 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001227 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001228 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001229 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230
Eric Dumazet989a2972010-04-14 09:55:35 +00001231 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 return 0;
1233}
1234
Eric Dumazet43815482010-04-29 11:01:49 +00001235/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236
1237int sock_wake_async(struct socket *sock, int how, int band)
1238{
Eric Dumazet43815482010-04-29 11:01:49 +00001239 struct socket_wq *wq;
1240
1241 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001243 rcu_read_lock();
1244 wq = rcu_dereference(sock->wq);
1245 if (!wq || !wq->fasync_list) {
1246 rcu_read_unlock();
1247 return -1;
1248 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001249 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001250 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1252 break;
1253 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001254 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1256 break;
1257 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001258 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001259call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001260 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001262 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001263 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 }
Eric Dumazet43815482010-04-29 11:01:49 +00001265 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 return 0;
1267}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001268EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269
Pavel Emelyanov721db932010-09-29 16:06:32 +04001270int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001271 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272{
1273 int err;
1274 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001275 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
1277 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001278 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 */
1280 if (family < 0 || family >= NPROTO)
1281 return -EAFNOSUPPORT;
1282 if (type < 0 || type >= SOCK_MAX)
1283 return -EINVAL;
1284
1285 /* Compatibility.
1286
1287 This uglymoron is moved from INET layer to here to avoid
1288 deadlock in module load.
1289 */
1290 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001291 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 if (!warned) {
1293 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001294 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1295 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 }
1297 family = PF_PACKET;
1298 }
1299
1300 err = security_socket_create(family, type, protocol, kern);
1301 if (err)
1302 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001303
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001304 /*
1305 * Allocate the socket and allow the family to set things up. if
1306 * the protocol is 0, the family is instructed to select an appropriate
1307 * default.
1308 */
1309 sock = sock_alloc();
1310 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001311 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001312 return -ENFILE; /* Not exactly a match, but its the
1313 closest posix thing */
1314 }
1315
1316 sock->type = type;
1317
Johannes Berg95a5afc2008-10-16 15:24:51 -07001318#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001319 /* Attempt to load a protocol module if the find failed.
1320 *
1321 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 * requested real, full-featured networking support upon configuration.
1323 * Otherwise module support will break!
1324 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001325 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001326 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327#endif
1328
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001329 rcu_read_lock();
1330 pf = rcu_dereference(net_families[family]);
1331 err = -EAFNOSUPPORT;
1332 if (!pf)
1333 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334
1335 /*
1336 * We will call the ->create function, that possibly is in a loadable
1337 * module, so we have to bump that loadable module refcnt first.
1338 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001339 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 goto out_release;
1341
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001342 /* Now protected by module ref count */
1343 rcu_read_unlock();
1344
Eric Paris3f378b62009-11-05 22:18:14 -08001345 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001346 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001348
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 /*
1350 * Now to bump the refcnt of the [loadable] module that owns this
1351 * socket at sock_release time we decrement its refcnt.
1352 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001353 if (!try_module_get(sock->ops->owner))
1354 goto out_module_busy;
1355
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 /*
1357 * Now that we're done with the ->create function, the [loadable]
1358 * module can have its refcnt decremented
1359 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001360 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001361 err = security_socket_post_create(sock, family, type, protocol, kern);
1362 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001363 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001364 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001366 return 0;
1367
1368out_module_busy:
1369 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001371 sock->ops = NULL;
1372 module_put(pf->owner);
1373out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001375 return err;
1376
1377out_release:
1378 rcu_read_unlock();
1379 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001381EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382
1383int sock_create(int family, int type, int protocol, struct socket **res)
1384{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001385 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001387EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388
1389int sock_create_kern(int family, int type, int protocol, struct socket **res)
1390{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001391 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001393EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001395SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396{
1397 int retval;
1398 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001399 int flags;
1400
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001401 /* Check the SOCK_* constants for consistency. */
1402 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1403 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1404 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1405 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1406
Ulrich Dreppera677a032008-07-23 21:29:17 -07001407 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001408 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001409 return -EINVAL;
1410 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001412 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1413 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1414
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 retval = sock_create(family, type, protocol, &sock);
1416 if (retval < 0)
1417 goto out;
1418
Ulrich Drepper77d27202008-07-23 21:29:35 -07001419 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 if (retval < 0)
1421 goto out_release;
1422
1423out:
1424 /* It may be already another descriptor 8) Not kernel problem. */
1425 return retval;
1426
1427out_release:
1428 sock_release(sock);
1429 return retval;
1430}
1431
1432/*
1433 * Create a pair of connected sockets.
1434 */
1435
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001436SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1437 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438{
1439 struct socket *sock1, *sock2;
1440 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001441 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001442 int flags;
1443
1444 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001445 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001446 return -EINVAL;
1447 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001449 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1450 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1451
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 /*
1453 * Obtain the first socket and check if the underlying protocol
1454 * supports the socketpair call.
1455 */
1456
1457 err = sock_create(family, type, protocol, &sock1);
1458 if (err < 0)
1459 goto out;
1460
1461 err = sock_create(family, type, protocol, &sock2);
1462 if (err < 0)
1463 goto out_release_1;
1464
1465 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001466 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 goto out_release_both;
1468
Masatake YAMATO600e1772012-08-29 10:44:29 +00001469 fd1 = sock_alloc_file(sock1, &newfile1, flags, NULL);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001470 if (unlikely(fd1 < 0)) {
1471 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001473 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474
Masatake YAMATO600e1772012-08-29 10:44:29 +00001475 fd2 = sock_alloc_file(sock2, &newfile2, flags, NULL);
Al Viro198de4d2009-08-05 19:29:23 +04001476 if (unlikely(fd2 < 0)) {
1477 err = fd2;
1478 fput(newfile1);
1479 put_unused_fd(fd1);
1480 sock_release(sock2);
1481 goto out;
Al Virodb349502007-02-07 01:48:00 -05001482 }
1483
Al Viro157cf642008-12-14 04:57:47 -05001484 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001485 fd_install(fd1, newfile1);
1486 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 /* fd1 and fd2 may be already another descriptors.
1488 * Not kernel problem.
1489 */
1490
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001491 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 if (!err)
1493 err = put_user(fd2, &usockvec[1]);
1494 if (!err)
1495 return 0;
1496
1497 sys_close(fd2);
1498 sys_close(fd1);
1499 return err;
1500
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001502 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001504 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505out:
1506 return err;
1507}
1508
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509/*
1510 * Bind a name to a socket. Nothing much to do here since it's
1511 * the protocol's responsibility to handle the local address.
1512 *
1513 * We move the socket address to kernel space before we call
1514 * the protocol layer (having also checked the address is ok).
1515 */
1516
Heiko Carstens20f37032009-01-14 14:14:23 +01001517SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518{
1519 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001520 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001521 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001523 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001524 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001525 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001526 if (err >= 0) {
1527 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001528 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001529 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001530 if (!err)
1531 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001532 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001533 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001535 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001536 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 return err;
1538}
1539
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540/*
1541 * Perform a listen. Basically, we allow the protocol to do anything
1542 * necessary for a listen, and if that works, we mark the socket as
1543 * ready for listening.
1544 */
1545
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001546SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547{
1548 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001549 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001550 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001551
1552 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1553 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001554 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001555 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001556 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557
1558 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001559 if (!err)
1560 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001562 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 }
1564 return err;
1565}
1566
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567/*
1568 * For accept, we attempt to create a new socket, set up the link
1569 * with the client, wake up the client, then return the new
1570 * connected fd. We collect the address of the connector in kernel
1571 * space and move it to user at the very end. This is unclean because
1572 * we open the socket then return an error.
1573 *
1574 * 1003.1g adds the ability to recvmsg() to query connection pending
1575 * status to recvmsg. We need to add that support in a way thats
1576 * clean when we restucture accept also.
1577 */
1578
Heiko Carstens20f37032009-01-14 14:14:23 +01001579SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1580 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581{
1582 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001583 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001584 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001585 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586
Ulrich Drepper77d27202008-07-23 21:29:35 -07001587 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001588 return -EINVAL;
1589
1590 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1591 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1592
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001593 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 if (!sock)
1595 goto out;
1596
1597 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001598 newsock = sock_alloc();
1599 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 goto out_put;
1601
1602 newsock->type = sock->type;
1603 newsock->ops = sock->ops;
1604
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 /*
1606 * We don't need try_module_get here, as the listening socket (sock)
1607 * has the protocol module (sock->ops->owner) held.
1608 */
1609 __module_get(newsock->ops->owner);
1610
Masatake YAMATO600e1772012-08-29 10:44:29 +00001611 newfd = sock_alloc_file(newsock, &newfile, flags,
1612 sock->sk->sk_prot_creator->name);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001613 if (unlikely(newfd < 0)) {
1614 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001615 sock_release(newsock);
1616 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001617 }
1618
Frank Filza79af592005-09-27 15:23:38 -07001619 err = security_socket_accept(sock, newsock);
1620 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001621 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001622
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1624 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001625 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626
1627 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001628 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001629 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001631 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001633 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001634 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001636 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 }
1638
1639 /* File flags are not inherited via accept() unlike another OSes. */
1640
David S. Miller39d8c1b2006-03-20 17:13:49 -08001641 fd_install(newfd, newfile);
1642 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001645 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646out:
1647 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001648out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001649 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001650 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 goto out_put;
1652}
1653
Heiko Carstens20f37032009-01-14 14:14:23 +01001654SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1655 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001656{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001657 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001658}
1659
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660/*
1661 * Attempt to connect to a socket with the server address. The address
1662 * is in user space so we verify it is OK and move it to kernel space.
1663 *
1664 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1665 * break bindings
1666 *
1667 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1668 * other SEQPACKET protocols that take time to connect() as it doesn't
1669 * include the -EINPROGRESS status for such sockets.
1670 */
1671
Heiko Carstens20f37032009-01-14 14:14:23 +01001672SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1673 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674{
1675 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001676 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001677 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001679 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 if (!sock)
1681 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001682 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (err < 0)
1684 goto out_put;
1685
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001686 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001687 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 if (err)
1689 goto out_put;
1690
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001691 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 sock->file->f_flags);
1693out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001694 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695out:
1696 return err;
1697}
1698
1699/*
1700 * Get the local address ('name') of a socket object. Move the obtained
1701 * name to user space.
1702 */
1703
Heiko Carstens20f37032009-01-14 14:14:23 +01001704SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1705 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706{
1707 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001708 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001709 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001710
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001711 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 if (!sock)
1713 goto out;
1714
1715 err = security_socket_getsockname(sock);
1716 if (err)
1717 goto out_put;
1718
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001719 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 if (err)
1721 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001722 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723
1724out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001725 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726out:
1727 return err;
1728}
1729
1730/*
1731 * Get the remote address ('name') of a socket object. Move the obtained
1732 * name to user space.
1733 */
1734
Heiko Carstens20f37032009-01-14 14:14:23 +01001735SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1736 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737{
1738 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001739 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001740 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001742 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1743 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 err = security_socket_getpeername(sock);
1745 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001746 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 return err;
1748 }
1749
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001750 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001751 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001752 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001754 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001755 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001756 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 }
1758 return err;
1759}
1760
1761/*
1762 * Send a datagram to a given address. We move the address into kernel
1763 * space and check the user space data area is readable before invoking
1764 * the protocol.
1765 */
1766
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001767SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001768 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001769 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770{
1771 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001772 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 int err;
1774 struct msghdr msg;
1775 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001776 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001777
Linus Torvalds253eacc2010-10-30 16:43:10 -07001778 if (len > INT_MAX)
1779 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001780 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1781 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001782 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001783
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001784 iov.iov_base = buff;
1785 iov.iov_len = len;
1786 msg.msg_name = NULL;
1787 msg.msg_iov = &iov;
1788 msg.msg_iovlen = 1;
1789 msg.msg_control = NULL;
1790 msg.msg_controllen = 0;
1791 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001792 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001793 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 if (err < 0)
1795 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001796 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 }
1799 if (sock->file->f_flags & O_NONBLOCK)
1800 flags |= MSG_DONTWAIT;
1801 msg.msg_flags = flags;
1802 err = sock_sendmsg(sock, &msg, len);
1803
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001804out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001805 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001806out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 return err;
1808}
1809
1810/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001811 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 */
1813
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001814SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001815 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816{
1817 return sys_sendto(fd, buff, len, flags, NULL, 0);
1818}
1819
1820/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001821 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822 * sender. We verify the buffers are writable and if needed move the
1823 * sender address from kernel to user space.
1824 */
1825
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001826SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001827 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001828 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829{
1830 struct socket *sock;
1831 struct iovec iov;
1832 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001833 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001834 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001835 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836
Linus Torvalds253eacc2010-10-30 16:43:10 -07001837 if (size > INT_MAX)
1838 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001839 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001841 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001843 msg.msg_control = NULL;
1844 msg.msg_controllen = 0;
1845 msg.msg_iovlen = 1;
1846 msg.msg_iov = &iov;
1847 iov.iov_len = size;
1848 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001849 msg.msg_name = (struct sockaddr *)&address;
1850 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 if (sock->file->f_flags & O_NONBLOCK)
1852 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001853 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001855 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001856 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001857 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001858 if (err2 < 0)
1859 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001861
1862 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001863out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 return err;
1865}
1866
1867/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001868 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 */
1870
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001871asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001872 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873{
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
Heiko Carstens20f37032009-01-14 14:14:23 +01001882SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1883 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001885 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 struct socket *sock;
1887
1888 if (optlen < 0)
1889 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890
1891 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1892 if (sock != NULL) {
1893 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001894 if (err)
1895 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896
1897 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001898 err =
1899 sock_setsockopt(sock, level, optname, optval,
1900 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001902 err =
1903 sock->ops->setsockopt(sock, level, optname, optval,
1904 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001905out_put:
1906 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 }
1908 return err;
1909}
1910
1911/*
1912 * Get a socket option. Because we don't know the option lengths we have
1913 * to pass a user mode parameter for the protocols to sort out.
1914 */
1915
Heiko Carstens20f37032009-01-14 14:14:23 +01001916SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1917 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001919 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 struct socket *sock;
1921
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001922 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1923 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001924 err = security_socket_getsockopt(sock, level, optname);
1925 if (err)
1926 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927
1928 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001929 err =
1930 sock_getsockopt(sock, level, optname, optval,
1931 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001933 err =
1934 sock->ops->getsockopt(sock, level, optname, optval,
1935 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001936out_put:
1937 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 }
1939 return err;
1940}
1941
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942/*
1943 * Shutdown a socket.
1944 */
1945
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001946SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001948 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 struct socket *sock;
1950
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001951 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1952 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001954 if (!err)
1955 err = sock->ops->shutdown(sock, how);
1956 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 }
1958 return err;
1959}
1960
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001961/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962 * fields which are the same type (int / unsigned) on our platforms.
1963 */
1964#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1965#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1966#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1967
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001968struct used_address {
1969 struct sockaddr_storage name;
1970 unsigned int name_len;
1971};
1972
Anton Blanchard228e5482011-05-02 20:21:35 +00001973static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001974 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001975 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001977 struct compat_msghdr __user *msg_compat =
1978 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001979 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001981 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001982 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1983 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001985 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001986
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 err = -EFAULT;
1988 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001989 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001991 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 return -EFAULT;
1993
Anton Blanchard228e5482011-05-02 20:21:35 +00001994 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02001995 err = -EMSGSIZE;
1996 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1997 goto out;
1998 err = -ENOMEM;
1999 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2000 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00002002 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 }
2004
2005 /* This will also move the address data into kernel space */
2006 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002007 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002009 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002010 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 goto out_freeiov;
2012 total_len = err;
2013
2014 err = -ENOBUFS;
2015
Anton Blanchard228e5482011-05-02 20:21:35 +00002016 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002018 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002021 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002022 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 if (err)
2024 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002025 ctl_buf = msg_sys->msg_control;
2026 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002028 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002030 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 goto out_freeiov;
2032 }
2033 err = -EFAULT;
2034 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002035 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2037 * checking falls down on this.
2038 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002039 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002040 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002041 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002043 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002045 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046
2047 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002048 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002049 /*
2050 * If this is sendmmsg() and current destination address is same as
2051 * previously succeeded address, omit asking LSM's decision.
2052 * used_address->name_len is initialized to UINT_MAX so that the first
2053 * destination address never matches.
2054 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002055 if (used_address && msg_sys->msg_name &&
2056 used_address->name_len == msg_sys->msg_namelen &&
2057 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002058 used_address->name_len)) {
2059 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2060 goto out_freectl;
2061 }
2062 err = sock_sendmsg(sock, msg_sys, total_len);
2063 /*
2064 * If this is sendmmsg() and sending to current destination address was
2065 * successful, remember it.
2066 */
2067 if (used_address && err >= 0) {
2068 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002069 if (msg_sys->msg_name)
2070 memcpy(&used_address->name, msg_sys->msg_name,
2071 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002072 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
2074out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002075 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2077out_freeiov:
2078 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002079 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002080out:
2081 return err;
2082}
2083
2084/*
2085 * BSD sendmsg interface
2086 */
2087
Eric Dumazet95c96172012-04-15 05:58:06 +00002088SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002089{
2090 int fput_needed, err;
2091 struct msghdr msg_sys;
2092 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2093
2094 if (!sock)
2095 goto out;
2096
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002097 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002098
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002099 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002100out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 return err;
2102}
2103
Anton Blanchard228e5482011-05-02 20:21:35 +00002104/*
2105 * Linux sendmmsg interface
2106 */
2107
2108int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2109 unsigned int flags)
2110{
2111 int fput_needed, err, datagrams;
2112 struct socket *sock;
2113 struct mmsghdr __user *entry;
2114 struct compat_mmsghdr __user *compat_entry;
2115 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002116 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002117
Anton Blanchard98382f42011-08-04 14:07:39 +00002118 if (vlen > UIO_MAXIOV)
2119 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002120
2121 datagrams = 0;
2122
2123 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2124 if (!sock)
2125 return err;
2126
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002127 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002128 entry = mmsg;
2129 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002130 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002131
2132 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002133 if (MSG_CMSG_COMPAT & flags) {
2134 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002135 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002136 if (err < 0)
2137 break;
2138 err = __put_user(err, &compat_entry->msg_len);
2139 ++compat_entry;
2140 } else {
2141 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002142 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002143 if (err < 0)
2144 break;
2145 err = put_user(err, &entry->msg_len);
2146 ++entry;
2147 }
2148
2149 if (err)
2150 break;
2151 ++datagrams;
2152 }
2153
Anton Blanchard228e5482011-05-02 20:21:35 +00002154 fput_light(sock->file, fput_needed);
2155
Anton Blanchard728ffb82011-08-04 14:07:38 +00002156 /* We only return an error if no datagrams were able to be sent */
2157 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002158 return datagrams;
2159
Anton Blanchard228e5482011-05-02 20:21:35 +00002160 return err;
2161}
2162
2163SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2164 unsigned int, vlen, unsigned int, flags)
2165{
2166 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2167}
2168
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002169static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002170 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002172 struct compat_msghdr __user *msg_compat =
2173 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002175 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002177 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178
2179 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002180 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
2182 /* user mode address pointers */
2183 struct sockaddr __user *uaddr;
2184 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002185
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002187 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002189 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002190 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002192 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002193 err = -EMSGSIZE;
2194 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2195 goto out;
2196 err = -ENOMEM;
2197 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2198 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002200 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201 }
2202
2203 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002204 * Save the user-mode address (verify_iovec will change the
2205 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002207
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002208 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 uaddr_len = COMPAT_NAMELEN(msg);
2210 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002211 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002213 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 if (err < 0)
2215 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002216 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002218 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2219 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002220
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 if (sock->file->f_flags & O_NONBLOCK)
2222 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002223 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2224 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225 if (err < 0)
2226 goto out_freeiov;
2227 len = err;
2228
2229 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002230 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002231 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002232 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 if (err < 0)
2234 goto out_freeiov;
2235 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002236 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002237 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238 if (err)
2239 goto out_freeiov;
2240 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002241 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 &msg_compat->msg_controllen);
2243 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002244 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245 &msg->msg_controllen);
2246 if (err)
2247 goto out_freeiov;
2248 err = len;
2249
2250out_freeiov:
2251 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002252 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002253out:
2254 return err;
2255}
2256
2257/*
2258 * BSD recvmsg interface
2259 */
2260
2261SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2262 unsigned int, flags)
2263{
2264 int fput_needed, err;
2265 struct msghdr msg_sys;
2266 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2267
2268 if (!sock)
2269 goto out;
2270
2271 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2272
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002273 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274out:
2275 return err;
2276}
2277
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002278/*
2279 * Linux recvmmsg interface
2280 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002282int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2283 unsigned int flags, struct timespec *timeout)
2284{
2285 int fput_needed, err, datagrams;
2286 struct socket *sock;
2287 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002288 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002289 struct msghdr msg_sys;
2290 struct timespec end_time;
2291
2292 if (timeout &&
2293 poll_select_set_timeout(&end_time, timeout->tv_sec,
2294 timeout->tv_nsec))
2295 return -EINVAL;
2296
2297 datagrams = 0;
2298
2299 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2300 if (!sock)
2301 return err;
2302
2303 err = sock_error(sock->sk);
2304 if (err)
2305 goto out_put;
2306
2307 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002308 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002309
2310 while (datagrams < vlen) {
2311 /*
2312 * No need to ask LSM for more than the first datagram.
2313 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002314 if (MSG_CMSG_COMPAT & flags) {
2315 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002316 &msg_sys, flags & ~MSG_WAITFORONE,
2317 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002318 if (err < 0)
2319 break;
2320 err = __put_user(err, &compat_entry->msg_len);
2321 ++compat_entry;
2322 } else {
2323 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002324 &msg_sys, flags & ~MSG_WAITFORONE,
2325 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002326 if (err < 0)
2327 break;
2328 err = put_user(err, &entry->msg_len);
2329 ++entry;
2330 }
2331
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002332 if (err)
2333 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002334 ++datagrams;
2335
Brandon L Black71c5c152010-03-26 16:18:03 +00002336 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2337 if (flags & MSG_WAITFORONE)
2338 flags |= MSG_DONTWAIT;
2339
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002340 if (timeout) {
2341 ktime_get_ts(timeout);
2342 *timeout = timespec_sub(end_time, *timeout);
2343 if (timeout->tv_sec < 0) {
2344 timeout->tv_sec = timeout->tv_nsec = 0;
2345 break;
2346 }
2347
2348 /* Timeout, return less than vlen datagrams */
2349 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2350 break;
2351 }
2352
2353 /* Out of band data, return right away */
2354 if (msg_sys.msg_flags & MSG_OOB)
2355 break;
2356 }
2357
2358out_put:
2359 fput_light(sock->file, fput_needed);
2360
2361 if (err == 0)
2362 return datagrams;
2363
2364 if (datagrams != 0) {
2365 /*
2366 * We may return less entries than requested (vlen) if the
2367 * sock is non block and there aren't enough datagrams...
2368 */
2369 if (err != -EAGAIN) {
2370 /*
2371 * ... or if recvmsg returns an error after we
2372 * received some datagrams, where we record the
2373 * error to return on the next call or if the
2374 * app asks about it using getsockopt(SO_ERROR).
2375 */
2376 sock->sk->sk_err = -err;
2377 }
2378
2379 return datagrams;
2380 }
2381
2382 return err;
2383}
2384
2385SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2386 unsigned int, vlen, unsigned int, flags,
2387 struct timespec __user *, timeout)
2388{
2389 int datagrams;
2390 struct timespec timeout_sys;
2391
2392 if (!timeout)
2393 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2394
2395 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2396 return -EFAULT;
2397
2398 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2399
2400 if (datagrams > 0 &&
2401 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2402 datagrams = -EFAULT;
2403
2404 return datagrams;
2405}
2406
2407#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408/* Argument list sizes for sys_socketcall */
2409#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002410static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002411 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2412 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2413 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002414 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002415};
2416
Linus Torvalds1da177e2005-04-16 15:20:36 -07002417#undef AL
2418
2419/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002420 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421 *
2422 * Argument checking cleaned up. Saved 20% in size.
2423 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002424 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425 */
2426
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002427SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428{
2429 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002430 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002432 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433
Anton Blanchard228e5482011-05-02 20:21:35 +00002434 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 return -EINVAL;
2436
Arjan van de Ven47379052009-09-28 12:57:44 -07002437 len = nargs[call];
2438 if (len > sizeof(a))
2439 return -EINVAL;
2440
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002442 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002444
Al Virof3298dc2008-12-10 03:16:51 -05002445 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002446
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002447 a0 = a[0];
2448 a1 = a[1];
2449
2450 switch (call) {
2451 case SYS_SOCKET:
2452 err = sys_socket(a0, a1, a[2]);
2453 break;
2454 case SYS_BIND:
2455 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2456 break;
2457 case SYS_CONNECT:
2458 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2459 break;
2460 case SYS_LISTEN:
2461 err = sys_listen(a0, a1);
2462 break;
2463 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002464 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2465 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002466 break;
2467 case SYS_GETSOCKNAME:
2468 err =
2469 sys_getsockname(a0, (struct sockaddr __user *)a1,
2470 (int __user *)a[2]);
2471 break;
2472 case SYS_GETPEERNAME:
2473 err =
2474 sys_getpeername(a0, (struct sockaddr __user *)a1,
2475 (int __user *)a[2]);
2476 break;
2477 case SYS_SOCKETPAIR:
2478 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2479 break;
2480 case SYS_SEND:
2481 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2482 break;
2483 case SYS_SENDTO:
2484 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2485 (struct sockaddr __user *)a[4], a[5]);
2486 break;
2487 case SYS_RECV:
2488 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2489 break;
2490 case SYS_RECVFROM:
2491 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2492 (struct sockaddr __user *)a[4],
2493 (int __user *)a[5]);
2494 break;
2495 case SYS_SHUTDOWN:
2496 err = sys_shutdown(a0, a1);
2497 break;
2498 case SYS_SETSOCKOPT:
2499 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2500 break;
2501 case SYS_GETSOCKOPT:
2502 err =
2503 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2504 (int __user *)a[4]);
2505 break;
2506 case SYS_SENDMSG:
2507 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2508 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002509 case SYS_SENDMMSG:
2510 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2511 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002512 case SYS_RECVMSG:
2513 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2514 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002515 case SYS_RECVMMSG:
2516 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2517 (struct timespec __user *)a[4]);
2518 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002519 case SYS_ACCEPT4:
2520 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2521 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002522 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002523 default:
2524 err = -EINVAL;
2525 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002526 }
2527 return err;
2528}
2529
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002530#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002531
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002532/**
2533 * sock_register - add a socket protocol handler
2534 * @ops: description of protocol
2535 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536 * This function is called by a protocol handler that wants to
2537 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002538 * socket interface. The value ops->family coresponds to the
2539 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002541int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542{
2543 int err;
2544
2545 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002546 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2547 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548 return -ENOBUFS;
2549 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002550
2551 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002552 if (rcu_dereference_protected(net_families[ops->family],
2553 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002554 err = -EEXIST;
2555 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002556 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557 err = 0;
2558 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002559 spin_unlock(&net_family_lock);
2560
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002561 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002562 return err;
2563}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002564EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002566/**
2567 * sock_unregister - remove a protocol handler
2568 * @family: protocol family to remove
2569 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570 * This function is called by a protocol handler that wants to
2571 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002572 * new socket creation.
2573 *
2574 * If protocol handler is a module, then it can use module reference
2575 * counts to protect against new references. If protocol handler is not
2576 * a module then it needs to provide its own protection in
2577 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002579void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002581 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002582
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002583 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002584 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002585 spin_unlock(&net_family_lock);
2586
2587 synchronize_rcu();
2588
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002589 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002591EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592
Andi Kleen77d76ea2005-12-22 12:43:42 -08002593static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002594{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002595 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002596 /*
2597 * Initialize the network sysctl infrastructure.
2598 */
2599 err = net_sysctl_init();
2600 if (err)
2601 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002602
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002604 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002605 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002606
Linus Torvalds1da177e2005-04-16 15:20:36 -07002607 sk_init();
2608
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002610 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611 */
2612 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002613
2614 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002615 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002616 */
2617
2618 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002619
2620 err = register_filesystem(&sock_fs_type);
2621 if (err)
2622 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002623 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002624 if (IS_ERR(sock_mnt)) {
2625 err = PTR_ERR(sock_mnt);
2626 goto out_mount;
2627 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002628
2629 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002630 */
2631
2632#ifdef CONFIG_NETFILTER
2633 netfilter_init();
2634#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002635
Richard Cochranc1f19b52010-07-17 08:49:36 +00002636#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2637 skb_timestamping_init();
2638#endif
2639
Nick Pigginb3e19d92011-01-07 17:50:11 +11002640out:
2641 return err;
2642
2643out_mount:
2644 unregister_filesystem(&sock_fs_type);
2645out_fs:
2646 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002647}
2648
Andi Kleen77d76ea2005-12-22 12:43:42 -08002649core_initcall(sock_init); /* early initcall */
2650
Linus Torvalds1da177e2005-04-16 15:20:36 -07002651#ifdef CONFIG_PROC_FS
2652void socket_seq_show(struct seq_file *seq)
2653{
2654 int cpu;
2655 int counter = 0;
2656
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002657 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002658 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659
2660 /* It can be negative, by the way. 8) */
2661 if (counter < 0)
2662 counter = 0;
2663
2664 seq_printf(seq, "sockets: used %d\n", counter);
2665}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002666#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002667
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002668#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002669static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002670 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002671{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002672 mm_segment_t old_fs = get_fs();
2673 struct timeval ktv;
2674 int err;
2675
2676 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002677 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002678 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002679 if (!err)
2680 err = compat_put_timeval(up, &ktv);
2681
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002682 return err;
2683}
2684
Arnd Bergmann6b960182009-11-06 23:10:54 -08002685static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002686 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002687{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002688 mm_segment_t old_fs = get_fs();
2689 struct timespec kts;
2690 int err;
2691
2692 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002693 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002694 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002695 if (!err)
2696 err = compat_put_timespec(up, &kts);
2697
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002698 return err;
2699}
2700
Arnd Bergmann6b960182009-11-06 23:10:54 -08002701static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002702{
2703 struct ifreq __user *uifr;
2704 int err;
2705
2706 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002707 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002708 return -EFAULT;
2709
Arnd Bergmann6b960182009-11-06 23:10:54 -08002710 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002711 if (err)
2712 return err;
2713
Arnd Bergmann6b960182009-11-06 23:10:54 -08002714 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002715 return -EFAULT;
2716
2717 return 0;
2718}
2719
Arnd Bergmann6b960182009-11-06 23:10:54 -08002720static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002721{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002722 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002723 struct ifconf ifc;
2724 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002725 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002726 struct ifreq __user *ifr;
2727 unsigned int i, j;
2728 int err;
2729
Arnd Bergmann6b960182009-11-06 23:10:54 -08002730 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002731 return -EFAULT;
2732
Mathias Krause43da5f22012-08-15 11:31:57 +00002733 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002734 if (ifc32.ifcbuf == 0) {
2735 ifc32.ifc_len = 0;
2736 ifc.ifc_len = 0;
2737 ifc.ifc_req = NULL;
2738 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2739 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002740 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2741 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002742 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2743 ifc.ifc_len = len;
2744 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2745 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002746 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002747 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002748 return -EFAULT;
2749 ifr++;
2750 ifr32++;
2751 }
2752 }
2753 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2754 return -EFAULT;
2755
Arnd Bergmann6b960182009-11-06 23:10:54 -08002756 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002757 if (err)
2758 return err;
2759
2760 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2761 return -EFAULT;
2762
2763 ifr = ifc.ifc_req;
2764 ifr32 = compat_ptr(ifc32.ifcbuf);
2765 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002766 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2767 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2768 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002769 return -EFAULT;
2770 ifr32++;
2771 ifr++;
2772 }
2773
2774 if (ifc32.ifcbuf == 0) {
2775 /* Translate from 64-bit structure multiple to
2776 * a 32-bit one.
2777 */
2778 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002779 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002780 ifc32.ifc_len = i;
2781 } else {
2782 ifc32.ifc_len = i;
2783 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002784 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002785 return -EFAULT;
2786
2787 return 0;
2788}
2789
Arnd Bergmann6b960182009-11-06 23:10:54 -08002790static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002791{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002792 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2793 bool convert_in = false, convert_out = false;
2794 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2795 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002796 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002797 u32 rule_cnt = 0, actual_rule_cnt;
2798 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002799 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002800 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002801
2802 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2803 return -EFAULT;
2804
Ben Hutchings3a7da392011-03-17 07:34:32 +00002805 compat_rxnfc = compat_ptr(data);
2806
2807 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002808 return -EFAULT;
2809
Ben Hutchings3a7da392011-03-17 07:34:32 +00002810 /* Most ethtool structures are defined without padding.
2811 * Unfortunately struct ethtool_rxnfc is an exception.
2812 */
2813 switch (ethcmd) {
2814 default:
2815 break;
2816 case ETHTOOL_GRXCLSRLALL:
2817 /* Buffer size is variable */
2818 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2819 return -EFAULT;
2820 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2821 return -ENOMEM;
2822 buf_size += rule_cnt * sizeof(u32);
2823 /* fall through */
2824 case ETHTOOL_GRXRINGS:
2825 case ETHTOOL_GRXCLSRLCNT:
2826 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002827 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002828 convert_out = true;
2829 /* fall through */
2830 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002831 buf_size += sizeof(struct ethtool_rxnfc);
2832 convert_in = true;
2833 break;
2834 }
2835
2836 ifr = compat_alloc_user_space(buf_size);
2837 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2838
2839 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2840 return -EFAULT;
2841
2842 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2843 &ifr->ifr_ifru.ifru_data))
2844 return -EFAULT;
2845
2846 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002847 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002848 * fs.ring_cookie and at the end of fs, but nowhere else.
2849 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002850 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2851 sizeof(compat_rxnfc->fs.m_ext) !=
2852 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2853 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002854 BUILD_BUG_ON(
2855 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2856 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2857 offsetof(struct ethtool_rxnfc, fs.location) -
2858 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2859
2860 if (copy_in_user(rxnfc, compat_rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002861 (void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002862 (void *)rxnfc) ||
2863 copy_in_user(&rxnfc->fs.ring_cookie,
2864 &compat_rxnfc->fs.ring_cookie,
2865 (void *)(&rxnfc->fs.location + 1) -
2866 (void *)&rxnfc->fs.ring_cookie) ||
2867 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2868 sizeof(rxnfc->rule_cnt)))
2869 return -EFAULT;
2870 }
2871
2872 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2873 if (ret)
2874 return ret;
2875
2876 if (convert_out) {
2877 if (copy_in_user(compat_rxnfc, rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002878 (const void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002879 (const void *)rxnfc) ||
2880 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2881 &rxnfc->fs.ring_cookie,
2882 (const void *)(&rxnfc->fs.location + 1) -
2883 (const void *)&rxnfc->fs.ring_cookie) ||
2884 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2885 sizeof(rxnfc->rule_cnt)))
2886 return -EFAULT;
2887
2888 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2889 /* As an optimisation, we only copy the actual
2890 * number of rules that the underlying
2891 * function returned. Since Mallory might
2892 * change the rule count in user memory, we
2893 * check that it is less than the rule count
2894 * originally given (as the user buffer size),
2895 * which has been range-checked.
2896 */
2897 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2898 return -EFAULT;
2899 if (actual_rule_cnt < rule_cnt)
2900 rule_cnt = actual_rule_cnt;
2901 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2902 &rxnfc->rule_locs[0],
2903 rule_cnt * sizeof(u32)))
2904 return -EFAULT;
2905 }
2906 }
2907
2908 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002909}
2910
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002911static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2912{
2913 void __user *uptr;
2914 compat_uptr_t uptr32;
2915 struct ifreq __user *uifr;
2916
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002917 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002918 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2919 return -EFAULT;
2920
2921 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2922 return -EFAULT;
2923
2924 uptr = compat_ptr(uptr32);
2925
2926 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2927 return -EFAULT;
2928
2929 return dev_ioctl(net, SIOCWANDEV, uifr);
2930}
2931
Arnd Bergmann6b960182009-11-06 23:10:54 -08002932static int bond_ioctl(struct net *net, unsigned int cmd,
2933 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002934{
2935 struct ifreq kifr;
2936 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002937 mm_segment_t old_fs;
2938 int err;
2939 u32 data;
2940 void __user *datap;
2941
2942 switch (cmd) {
2943 case SIOCBONDENSLAVE:
2944 case SIOCBONDRELEASE:
2945 case SIOCBONDSETHWADDR:
2946 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002947 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002948 return -EFAULT;
2949
2950 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002951 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002952 err = dev_ioctl(net, cmd,
2953 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002954 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002955
2956 return err;
2957 case SIOCBONDSLAVEINFOQUERY:
2958 case SIOCBONDINFOQUERY:
2959 uifr = compat_alloc_user_space(sizeof(*uifr));
2960 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2961 return -EFAULT;
2962
2963 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2964 return -EFAULT;
2965
2966 datap = compat_ptr(data);
2967 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2968 return -EFAULT;
2969
Arnd Bergmann6b960182009-11-06 23:10:54 -08002970 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002971 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002972 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002973 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002974}
2975
Arnd Bergmann6b960182009-11-06 23:10:54 -08002976static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2977 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002978{
2979 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002980 char tmp_buf[IFNAMSIZ];
2981 void __user *data64;
2982 u32 data32;
2983
2984 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2985 IFNAMSIZ))
2986 return -EFAULT;
2987 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2988 return -EFAULT;
2989 data64 = compat_ptr(data32);
2990
2991 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2992
2993 /* Don't check these user accesses, just let that get trapped
2994 * in the ioctl handler instead.
2995 */
2996 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2997 IFNAMSIZ))
2998 return -EFAULT;
2999 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3000 return -EFAULT;
3001
Arnd Bergmann6b960182009-11-06 23:10:54 -08003002 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003003}
3004
Arnd Bergmann6b960182009-11-06 23:10:54 -08003005static int dev_ifsioc(struct net *net, struct socket *sock,
3006 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003007{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003008 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003009 int err;
3010
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003011 uifr = compat_alloc_user_space(sizeof(*uifr));
3012 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3013 return -EFAULT;
3014
3015 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3016
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003017 if (!err) {
3018 switch (cmd) {
3019 case SIOCGIFFLAGS:
3020 case SIOCGIFMETRIC:
3021 case SIOCGIFMTU:
3022 case SIOCGIFMEM:
3023 case SIOCGIFHWADDR:
3024 case SIOCGIFINDEX:
3025 case SIOCGIFADDR:
3026 case SIOCGIFBRDADDR:
3027 case SIOCGIFDSTADDR:
3028 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003029 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003030 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003031 case SIOCGMIIPHY:
3032 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003033 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003034 err = -EFAULT;
3035 break;
3036 }
3037 }
3038 return err;
3039}
3040
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003041static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3042 struct compat_ifreq __user *uifr32)
3043{
3044 struct ifreq ifr;
3045 struct compat_ifmap __user *uifmap32;
3046 mm_segment_t old_fs;
3047 int err;
3048
3049 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3050 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3051 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3052 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3053 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3054 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3055 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3056 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3057 if (err)
3058 return -EFAULT;
3059
3060 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003061 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003062 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003063 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003064
3065 if (cmd == SIOCGIFMAP && !err) {
3066 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3067 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3068 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3069 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3070 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3071 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3072 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3073 if (err)
3074 err = -EFAULT;
3075 }
3076 return err;
3077}
3078
3079static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3080{
3081 void __user *uptr;
3082 compat_uptr_t uptr32;
3083 struct ifreq __user *uifr;
3084
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003085 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003086 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3087 return -EFAULT;
3088
3089 if (get_user(uptr32, &uifr32->ifr_data))
3090 return -EFAULT;
3091
3092 uptr = compat_ptr(uptr32);
3093
3094 if (put_user(uptr, &uifr->ifr_data))
3095 return -EFAULT;
3096
3097 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3098}
3099
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003100struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003101 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003102 struct sockaddr rt_dst; /* target address */
3103 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3104 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003105 unsigned short rt_flags;
3106 short rt_pad2;
3107 u32 rt_pad3;
3108 unsigned char rt_tos;
3109 unsigned char rt_class;
3110 short rt_pad4;
3111 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003112 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003113 u32 rt_mtu; /* per route MTU/Window */
3114 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003115 unsigned short rt_irtt; /* Initial RTT */
3116};
3117
3118struct in6_rtmsg32 {
3119 struct in6_addr rtmsg_dst;
3120 struct in6_addr rtmsg_src;
3121 struct in6_addr rtmsg_gateway;
3122 u32 rtmsg_type;
3123 u16 rtmsg_dst_len;
3124 u16 rtmsg_src_len;
3125 u32 rtmsg_metric;
3126 u32 rtmsg_info;
3127 u32 rtmsg_flags;
3128 s32 rtmsg_ifindex;
3129};
3130
Arnd Bergmann6b960182009-11-06 23:10:54 -08003131static int routing_ioctl(struct net *net, struct socket *sock,
3132 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003133{
3134 int ret;
3135 void *r = NULL;
3136 struct in6_rtmsg r6;
3137 struct rtentry r4;
3138 char devname[16];
3139 u32 rtdev;
3140 mm_segment_t old_fs = get_fs();
3141
Arnd Bergmann6b960182009-11-06 23:10:54 -08003142 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3143 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003144 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003145 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003146 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3147 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3148 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3149 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3150 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3151 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3152 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003153
3154 r = (void *) &r6;
3155 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003156 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003157 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003158 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003159 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3160 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3161 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3162 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3163 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3164 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003165 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003166 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003167 r4.rt_dev = (char __user __force *)devname;
3168 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003169 } else
3170 r4.rt_dev = NULL;
3171
3172 r = (void *) &r4;
3173 }
3174
3175 if (ret) {
3176 ret = -EFAULT;
3177 goto out;
3178 }
3179
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003180 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003181 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003182 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003183
3184out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003185 return ret;
3186}
3187
3188/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3189 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003190 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003191 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003192static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003193{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003194 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003195
Arnd Bergmann6b960182009-11-06 23:10:54 -08003196 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003197 return -EFAULT;
3198 if (tmp == BRCTL_GET_VERSION)
3199 return BRCTL_VERSION + 1;
3200 return -EINVAL;
3201}
3202
Arnd Bergmann6b960182009-11-06 23:10:54 -08003203static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3204 unsigned int cmd, unsigned long arg)
3205{
3206 void __user *argp = compat_ptr(arg);
3207 struct sock *sk = sock->sk;
3208 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003209
Arnd Bergmann6b960182009-11-06 23:10:54 -08003210 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3211 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003212
Arnd Bergmann6b960182009-11-06 23:10:54 -08003213 switch (cmd) {
3214 case SIOCSIFBR:
3215 case SIOCGIFBR:
3216 return old_bridge_ioctl(argp);
3217 case SIOCGIFNAME:
3218 return dev_ifname32(net, argp);
3219 case SIOCGIFCONF:
3220 return dev_ifconf(net, argp);
3221 case SIOCETHTOOL:
3222 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003223 case SIOCWANDEV:
3224 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003225 case SIOCGIFMAP:
3226 case SIOCSIFMAP:
3227 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003228 case SIOCBONDENSLAVE:
3229 case SIOCBONDRELEASE:
3230 case SIOCBONDSETHWADDR:
3231 case SIOCBONDSLAVEINFOQUERY:
3232 case SIOCBONDINFOQUERY:
3233 case SIOCBONDCHANGEACTIVE:
3234 return bond_ioctl(net, cmd, argp);
3235 case SIOCADDRT:
3236 case SIOCDELRT:
3237 return routing_ioctl(net, sock, cmd, argp);
3238 case SIOCGSTAMP:
3239 return do_siocgstamp(net, sock, cmd, argp);
3240 case SIOCGSTAMPNS:
3241 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003242 case SIOCSHWTSTAMP:
3243 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003244
Arnd Bergmann6b960182009-11-06 23:10:54 -08003245 case FIOSETOWN:
3246 case SIOCSPGRP:
3247 case FIOGETOWN:
3248 case SIOCGPGRP:
3249 case SIOCBRADDBR:
3250 case SIOCBRDELBR:
3251 case SIOCGIFVLAN:
3252 case SIOCSIFVLAN:
3253 case SIOCADDDLCI:
3254 case SIOCDELDLCI:
3255 return sock_ioctl(file, cmd, arg);
3256
3257 case SIOCGIFFLAGS:
3258 case SIOCSIFFLAGS:
3259 case SIOCGIFMETRIC:
3260 case SIOCSIFMETRIC:
3261 case SIOCGIFMTU:
3262 case SIOCSIFMTU:
3263 case SIOCGIFMEM:
3264 case SIOCSIFMEM:
3265 case SIOCGIFHWADDR:
3266 case SIOCSIFHWADDR:
3267 case SIOCADDMULTI:
3268 case SIOCDELMULTI:
3269 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003270 case SIOCGIFADDR:
3271 case SIOCSIFADDR:
3272 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003273 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003274 case SIOCGIFBRDADDR:
3275 case SIOCSIFBRDADDR:
3276 case SIOCGIFDSTADDR:
3277 case SIOCSIFDSTADDR:
3278 case SIOCGIFNETMASK:
3279 case SIOCSIFNETMASK:
3280 case SIOCSIFPFLAGS:
3281 case SIOCGIFPFLAGS:
3282 case SIOCGIFTXQLEN:
3283 case SIOCSIFTXQLEN:
3284 case SIOCBRADDIF:
3285 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003286 case SIOCSIFNAME:
3287 case SIOCGMIIPHY:
3288 case SIOCGMIIREG:
3289 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003290 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003291
Arnd Bergmann6b960182009-11-06 23:10:54 -08003292 case SIOCSARP:
3293 case SIOCGARP:
3294 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003295 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003296 return sock_do_ioctl(net, sock, cmd, arg);
3297 }
3298
Arnd Bergmann6b960182009-11-06 23:10:54 -08003299 return -ENOIOCTLCMD;
3300}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003301
Eric Dumazet95c96172012-04-15 05:58:06 +00003302static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003303 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003304{
3305 struct socket *sock = file->private_data;
3306 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003307 struct sock *sk;
3308 struct net *net;
3309
3310 sk = sock->sk;
3311 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003312
3313 if (sock->ops->compat_ioctl)
3314 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3315
David S. Miller87de87d2008-06-03 09:14:03 -07003316 if (ret == -ENOIOCTLCMD &&
3317 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3318 ret = compat_wext_handle_ioctl(net, cmd, arg);
3319
Arnd Bergmann6b960182009-11-06 23:10:54 -08003320 if (ret == -ENOIOCTLCMD)
3321 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3322
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003323 return ret;
3324}
3325#endif
3326
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003327int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3328{
3329 return sock->ops->bind(sock, addr, addrlen);
3330}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003331EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003332
3333int kernel_listen(struct socket *sock, int backlog)
3334{
3335 return sock->ops->listen(sock, backlog);
3336}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003337EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003338
3339int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3340{
3341 struct sock *sk = sock->sk;
3342 int err;
3343
3344 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3345 newsock);
3346 if (err < 0)
3347 goto done;
3348
3349 err = sock->ops->accept(sock, *newsock, flags);
3350 if (err < 0) {
3351 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003352 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003353 goto done;
3354 }
3355
3356 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003357 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003358
3359done:
3360 return err;
3361}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003362EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003363
3364int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003365 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003366{
3367 return sock->ops->connect(sock, addr, addrlen, flags);
3368}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003369EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003370
3371int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3372 int *addrlen)
3373{
3374 return sock->ops->getname(sock, addr, addrlen, 0);
3375}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003376EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003377
3378int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3379 int *addrlen)
3380{
3381 return sock->ops->getname(sock, addr, addrlen, 1);
3382}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003383EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003384
3385int kernel_getsockopt(struct socket *sock, int level, int optname,
3386 char *optval, int *optlen)
3387{
3388 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003389 char __user *uoptval;
3390 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003391 int err;
3392
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003393 uoptval = (char __user __force *) optval;
3394 uoptlen = (int __user __force *) optlen;
3395
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003396 set_fs(KERNEL_DS);
3397 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003398 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003399 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003400 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3401 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003402 set_fs(oldfs);
3403 return err;
3404}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003405EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003406
3407int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003408 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003409{
3410 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003411 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003412 int err;
3413
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003414 uoptval = (char __user __force *) optval;
3415
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003416 set_fs(KERNEL_DS);
3417 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003418 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003419 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003420 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003421 optlen);
3422 set_fs(oldfs);
3423 return err;
3424}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003425EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003426
3427int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3428 size_t size, int flags)
3429{
Herbert Xuf8451722010-05-24 00:12:34 -07003430 sock_update_classid(sock->sk);
3431
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003432 if (sock->ops->sendpage)
3433 return sock->ops->sendpage(sock, page, offset, size, flags);
3434
3435 return sock_no_sendpage(sock, page, offset, size, flags);
3436}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003437EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003438
3439int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3440{
3441 mm_segment_t oldfs = get_fs();
3442 int err;
3443
3444 set_fs(KERNEL_DS);
3445 err = sock->ops->ioctl(sock, cmd, arg);
3446 set_fs(oldfs);
3447
3448 return err;
3449}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003450EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003451
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003452int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3453{
3454 return sock->ops->shutdown(sock, how);
3455}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003456EXPORT_SYMBOL(kernel_sock_shutdown);