blob: 851edcd6b0982d5a820da29e2145be75bddcff9c [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070096#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070097#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
Arnd Bergmann6b960182009-11-06 23:10:54 -0800102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700121 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800122#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700127 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800128 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
Arjan van de Venda7071d2007-02-12 00:55:36 -0800135static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 .owner = THIS_MODULE,
137 .llseek = no_llseek,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
140 .poll = sock_poll,
141 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800142#ifdef CONFIG_COMPAT
143 .compat_ioctl = compat_sock_ioctl,
144#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 .mmap = sock_mmap,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800151 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161/*
162 * Statistics counters of the socket lists
163 */
164
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700165static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
167/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700188 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700190 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100192 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700211
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000212static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000213 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 int err;
216 int len;
217
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218 err = get_user(len, ulen);
219 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700221 if (len > klen)
222 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700223 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return -EFAULT;
230 }
231 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235 return __put_user(klen, ulen);
236}
237
Christoph Lametere18b8902006-12-06 20:33:20 -0800238static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000243 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700244
Christoph Lametere94b1762006-12-06 20:33:17 -0800245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 if (!ei)
247 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000248 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
249 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000250 kmem_cache_free(sock_inode_cachep, ei);
251 return NULL;
252 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000253 init_waitqueue_head(&wq->wait);
254 wq->fasync_list = NULL;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700256
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
263 return &ei->vfs_inode;
264}
265
266static void sock_destroy_inode(struct inode *inode)
267{
Eric Dumazet43815482010-04-29 11:01:49 +0000268 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000269 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000270
271 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000272 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800273 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000274 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275}
276
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700277static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700279 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
Christoph Lametera35afb82007-05-16 22:10:57 -0700281 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284static int init_inodecache(void)
285{
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700287 sizeof(struct socket_alloc),
288 0,
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
291 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900292 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 if (sock_inode_cachep == NULL)
294 return -ENOMEM;
295 return 0;
296}
297
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700298static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700299 .alloc_inode = sock_alloc_inode,
300 .destroy_inode = sock_destroy_inode,
301 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302};
303
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700304/*
305 * sockfs_dname() is called from d_path().
306 */
307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
308{
309 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
310 dentry->d_inode->i_ino);
311}
312
Al Viro3ba13d12009-02-20 06:02:22 +0000313static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700314 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315};
316
Al Viroc74a1cb2011-01-12 16:59:34 -0500317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
318 int flags, const char *dev_name, void *data)
319{
320 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
321 &sockfs_dentry_operations, SOCKFS_MAGIC);
322}
323
324static struct vfsmount *sock_mnt __read_mostly;
325
326static struct file_system_type sock_fs_type = {
327 .name = "sockfs",
328 .mount = sockfs_mount,
329 .kill_sb = kill_anon_super,
330};
331
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332/*
333 * Obtains the first available file descriptor and sets it up for use.
334 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
342 *
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
347 */
348
Al Viro7cbe66b2009-08-05 19:59:08 +0400349static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
Al Viro7cbe66b2009-08-05 19:59:08 +0400351 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400352 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400353 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800355
Ulrich Dreppera677a032008-07-23 21:29:17 -0700356 fd = get_unused_fd_flags(flags);
Al Viro7cbe66b2009-08-05 19:59:08 +0400357 if (unlikely(fd < 0))
358 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800359
Nick Piggin4b936882011-01-07 17:50:07 +1100360 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro2c48b9c2009-08-09 00:52:35 +0400361 if (unlikely(!path.dentry)) {
Al Viro7cbe66b2009-08-05 19:59:08 +0400362 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800363 return -ENOMEM;
Al Viro7cbe66b2009-08-05 19:59:08 +0400364 }
Al Viro2c48b9c2009-08-09 00:52:35 +0400365 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366
Al Viro2c48b9c2009-08-09 00:52:35 +0400367 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400368 SOCK_INODE(sock)->i_fop = &socket_file_ops;
369
Al Viro2c48b9c2009-08-09 00:52:35 +0400370 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400371 &socket_file_ops);
372 if (unlikely(!file)) {
373 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400374 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400375 path_put(&path);
Al Virocc3808f2009-08-06 09:43:59 +0400376 put_unused_fd(fd);
377 return -ENFILE;
378 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800379
380 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700381 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800382 file->f_pos = 0;
383 file->private_data = sock;
384
Al Viro7cbe66b2009-08-05 19:59:08 +0400385 *f = file;
386 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387}
388
Ulrich Dreppera677a032008-07-23 21:29:17 -0700389int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800390{
391 struct file *newfile;
Al Viro7cbe66b2009-08-05 19:59:08 +0400392 int fd = sock_alloc_file(sock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393
Al Viro7cbe66b2009-08-05 19:59:08 +0400394 if (likely(fd >= 0))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800395 fd_install(fd, newfile);
Al Viro7cbe66b2009-08-05 19:59:08 +0400396
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 return fd;
398}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700399EXPORT_SYMBOL(sock_map_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800401static struct socket *sock_from_file(struct file *file, int *err)
402{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800403 if (file->f_op == &socket_file_ops)
404 return file->private_data; /* set in sock_map_fd */
405
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800406 *err = -ENOTSOCK;
407 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800408}
409
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700411 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 * @fd: file handle
413 * @err: pointer to an error code return
414 *
415 * The file handle passed in is locked and the socket it is bound
416 * too is returned. If an error occurs the err pointer is overwritten
417 * with a negative errno code and NULL is returned. The function checks
418 * for both invalid handles and passing a handle which is not a socket.
419 *
420 * On a success the socket object pointer is returned.
421 */
422
423struct socket *sockfd_lookup(int fd, int *err)
424{
425 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 struct socket *sock;
427
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700428 file = fget(fd);
429 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 *err = -EBADF;
431 return NULL;
432 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700433
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800434 sock = sock_from_file(file, err);
435 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 return sock;
438}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700439EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800441static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
442{
443 struct file *file;
444 struct socket *sock;
445
Hua Zhong36725582006-04-19 15:25:02 -0700446 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800447 file = fget_light(fd, fput_needed);
448 if (file) {
449 sock = sock_from_file(file, err);
450 if (sock)
451 return sock;
452 fput_light(file, *fput_needed);
453 }
454 return NULL;
455}
456
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457/**
458 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700459 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 * Allocate a new inode and socket object. The two are bound together
461 * and initialised. The socket is then returned. If we are out of inodes
462 * NULL is returned.
463 */
464
465static struct socket *sock_alloc(void)
466{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700467 struct inode *inode;
468 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200470 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 if (!inode)
472 return NULL;
473
474 sock = SOCKET_I(inode);
475
Eric Dumazet29a020d2009-09-15 02:39:20 -0700476 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400477 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700478 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100479 inode->i_uid = current_fsuid();
480 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481
Eric Dumazet4e694892009-04-04 16:41:09 -0700482 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 return sock;
484}
485
486/*
487 * In theory you can't get an open on this inode, but /proc provides
488 * a back door. Remember to keep it shut otherwise you'll let the
489 * creepy crawlies in.
490 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700491
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
493{
494 return -ENXIO;
495}
496
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800497const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 .owner = THIS_MODULE,
499 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200500 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501};
502
503/**
504 * sock_release - close a socket
505 * @sock: socket to close
506 *
507 * The socket is released from the protocol stack if it has a release
508 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700509 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512void sock_release(struct socket *sock)
513{
514 if (sock->ops) {
515 struct module *owner = sock->ops->owner;
516
517 sock->ops->release(sock);
518 sock->ops = NULL;
519 module_put(owner);
520 }
521
Eric Dumazeteaefd112011-02-18 03:26:36 +0000522 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 printk(KERN_ERR "sock_release: fasync list not empty!\n");
524
Eric Dumazet4e694892009-04-04 16:41:09 -0700525 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 if (!sock->file) {
527 iput(SOCK_INODE(sock));
528 return;
529 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700530 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700532EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000534int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000535{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000536 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000537 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000538 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000539 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000540 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100541 if (sock_flag(sk, SOCK_WIFI_STATUS))
542 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000543 return 0;
544}
545EXPORT_SYMBOL(sock_tx_timestamp);
546
Anton Blanchard228e5482011-05-02 20:21:35 +0000547static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
548 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549{
550 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
Herbert Xuf8451722010-05-24 00:12:34 -0700552 sock_update_classid(sock->sk);
553
Neil Horman5bc14212011-11-22 05:10:51 +0000554 sock_update_netprioidx(sock->sk);
555
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 si->sock = sock;
557 si->scm = NULL;
558 si->msg = msg;
559 si->size = size;
560
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return sock->ops->sendmsg(iocb, sock, msg, size);
562}
563
Anton Blanchard228e5482011-05-02 20:21:35 +0000564static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
565 struct msghdr *msg, size_t size)
566{
567 int err = security_socket_sendmsg(sock, msg, size);
568
569 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
570}
571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
573{
574 struct kiocb iocb;
575 struct sock_iocb siocb;
576 int ret;
577
578 init_sync_kiocb(&iocb, NULL);
579 iocb.private = &siocb;
580 ret = __sock_sendmsg(&iocb, sock, msg, size);
581 if (-EIOCBQUEUED == ret)
582 ret = wait_on_sync_kiocb(&iocb);
583 return ret;
584}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700585EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586
Eric Dumazet894dc242011-07-26 02:39:41 +0000587static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000588{
589 struct kiocb iocb;
590 struct sock_iocb siocb;
591 int ret;
592
593 init_sync_kiocb(&iocb, NULL);
594 iocb.private = &siocb;
595 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
596 if (-EIOCBQUEUED == ret)
597 ret = wait_on_sync_kiocb(&iocb);
598 return ret;
599}
600
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
602 struct kvec *vec, size_t num, size_t size)
603{
604 mm_segment_t oldfs = get_fs();
605 int result;
606
607 set_fs(KERNEL_DS);
608 /*
609 * the following is safe, since for compiler definitions of kvec and
610 * iovec are identical, yielding the same in-core layout and alignment
611 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700612 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 msg->msg_iovlen = num;
614 result = sock_sendmsg(sock, msg, size);
615 set_fs(oldfs);
616 return result;
617}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700618EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619
Patrick Ohly20d49472009-02-12 05:03:38 +0000620static int ktime2ts(ktime_t kt, struct timespec *ts)
621{
622 if (kt.tv64) {
623 *ts = ktime_to_timespec(kt);
624 return 1;
625 } else {
626 return 0;
627 }
628}
629
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700630/*
631 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
632 */
633void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
634 struct sk_buff *skb)
635{
Patrick Ohly20d49472009-02-12 05:03:38 +0000636 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
637 struct timespec ts[3];
638 int empty = 1;
639 struct skb_shared_hwtstamps *shhwtstamps =
640 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700641
Patrick Ohly20d49472009-02-12 05:03:38 +0000642 /* Race occurred between timestamp enabling and packet
643 receiving. Fill in the current time for now. */
644 if (need_software_tstamp && skb->tstamp.tv64 == 0)
645 __net_timestamp(skb);
646
647 if (need_software_tstamp) {
648 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
649 struct timeval tv;
650 skb_get_timestamp(skb, &tv);
651 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
652 sizeof(tv), &tv);
653 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000654 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000655 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000656 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000657 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700658 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000659
660
661 memset(ts, 0, sizeof(ts));
662 if (skb->tstamp.tv64 &&
663 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
664 skb_get_timestampns(skb, ts + 0);
665 empty = 0;
666 }
667 if (shhwtstamps) {
668 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
669 ktime2ts(shhwtstamps->syststamp, ts + 1))
670 empty = 0;
671 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
672 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
673 empty = 0;
674 }
675 if (!empty)
676 put_cmsg(msg, SOL_SOCKET,
677 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700678}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300679EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
680
Johannes Berg6e3e9392011-11-09 10:15:42 +0100681void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
682 struct sk_buff *skb)
683{
684 int ack;
685
686 if (!sock_flag(sk, SOCK_WIFI_STATUS))
687 return;
688 if (!skb->wifi_acked_valid)
689 return;
690
691 ack = skb->wifi_acked;
692
693 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
694}
695EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
696
stephen hemminger11165f12010-10-18 14:27:29 +0000697static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
698 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700699{
700 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
701 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
702 sizeof(__u32), &skb->dropcount);
703}
704
Eric Dumazet767dd032010-04-28 19:14:43 +0000705void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700706 struct sk_buff *skb)
707{
708 sock_recv_timestamp(msg, sk, skb);
709 sock_recv_drops(msg, sk, skb);
710}
Eric Dumazet767dd032010-04-28 19:14:43 +0000711EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700712
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700713static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
714 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 struct sock_iocb *si = kiocb_to_siocb(iocb);
717
Herbert Xuf8451722010-05-24 00:12:34 -0700718 sock_update_classid(sock->sk);
719
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 si->sock = sock;
721 si->scm = NULL;
722 si->msg = msg;
723 si->size = size;
724 si->flags = flags;
725
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
727}
728
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700729static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
730 struct msghdr *msg, size_t size, int flags)
731{
732 int err = security_socket_recvmsg(sock, msg, size, flags);
733
734 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
735}
736
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700737int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738 size_t size, int flags)
739{
740 struct kiocb iocb;
741 struct sock_iocb siocb;
742 int ret;
743
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700744 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 iocb.private = &siocb;
746 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
747 if (-EIOCBQUEUED == ret)
748 ret = wait_on_sync_kiocb(&iocb);
749 return ret;
750}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700751EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700753static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
754 size_t size, int flags)
755{
756 struct kiocb iocb;
757 struct sock_iocb siocb;
758 int ret;
759
760 init_sync_kiocb(&iocb, NULL);
761 iocb.private = &siocb;
762 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
763 if (-EIOCBQUEUED == ret)
764 ret = wait_on_sync_kiocb(&iocb);
765 return ret;
766}
767
Martin Lucinac1249c02010-12-10 00:04:05 +0000768/**
769 * kernel_recvmsg - Receive a message from a socket (kernel space)
770 * @sock: The socket to receive the message from
771 * @msg: Received message
772 * @vec: Input s/g array for message data
773 * @num: Size of input s/g array
774 * @size: Number of bytes to read
775 * @flags: Message flags (MSG_DONTWAIT, etc...)
776 *
777 * On return the msg structure contains the scatter/gather array passed in the
778 * vec argument. The array is modified so that it consists of the unfilled
779 * portion of the original array.
780 *
781 * The returned value is the total number of bytes received, or an error.
782 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700783int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
784 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785{
786 mm_segment_t oldfs = get_fs();
787 int result;
788
789 set_fs(KERNEL_DS);
790 /*
791 * the following is safe, since for compiler definitions of kvec and
792 * iovec are identical, yielding the same in-core layout and alignment
793 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700794 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 result = sock_recvmsg(sock, msg, size, flags);
796 set_fs(oldfs);
797 return result;
798}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700799EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800
801static void sock_aio_dtor(struct kiocb *iocb)
802{
803 kfree(iocb->private);
804}
805
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300806static ssize_t sock_sendpage(struct file *file, struct page *page,
807 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
809 struct socket *sock;
810 int flags;
811
Eric Dumazetb69aee02005-09-06 14:42:45 -0700812 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Eric Dumazet35f9c092012-04-05 03:05:35 +0000814 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
815 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
816 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
Linus Torvaldse6949582009-08-13 08:28:36 -0700818 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819}
820
Jens Axboe9c55e012007-11-06 23:30:13 -0800821static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700822 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800823 unsigned int flags)
824{
825 struct socket *sock = file->private_data;
826
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800827 if (unlikely(!sock->ops->splice_read))
828 return -EINVAL;
829
Herbert Xuf8451722010-05-24 00:12:34 -0700830 sock_update_classid(sock->sk);
831
Jens Axboe9c55e012007-11-06 23:30:13 -0800832 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
833}
834
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800835static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700836 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800837{
838 if (!is_sync_kiocb(iocb)) {
839 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
840 if (!siocb)
841 return NULL;
842 iocb->ki_dtor = sock_aio_dtor;
843 }
844
845 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800846 iocb->private = siocb;
847 return siocb;
848}
849
850static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700851 struct file *file, const struct iovec *iov,
852 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800853{
854 struct socket *sock = file->private_data;
855 size_t size = 0;
856 int i;
857
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700858 for (i = 0; i < nr_segs; i++)
859 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800860
861 msg->msg_name = NULL;
862 msg->msg_namelen = 0;
863 msg->msg_control = NULL;
864 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700865 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800866 msg->msg_iovlen = nr_segs;
867 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
868
869 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
870}
871
Badari Pulavarty027445c2006-09-30 23:28:46 -0700872static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
873 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800874{
875 struct sock_iocb siocb, *x;
876
877 if (pos != 0)
878 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700879
880 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800881 return 0;
882
Badari Pulavarty027445c2006-09-30 23:28:46 -0700883
884 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800885 if (!x)
886 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700887 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800888}
889
890static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700891 struct file *file, const struct iovec *iov,
892 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800893{
894 struct socket *sock = file->private_data;
895 size_t size = 0;
896 int i;
897
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700898 for (i = 0; i < nr_segs; i++)
899 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800900
901 msg->msg_name = NULL;
902 msg->msg_namelen = 0;
903 msg->msg_control = NULL;
904 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700905 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800906 msg->msg_iovlen = nr_segs;
907 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
908 if (sock->type == SOCK_SEQPACKET)
909 msg->msg_flags |= MSG_EOR;
910
911 return __sock_sendmsg(iocb, sock, msg, size);
912}
913
Badari Pulavarty027445c2006-09-30 23:28:46 -0700914static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
915 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800917 struct sock_iocb siocb, *x;
918
919 if (pos != 0)
920 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700921
Badari Pulavarty027445c2006-09-30 23:28:46 -0700922 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800923 if (!x)
924 return -ENOMEM;
925
Badari Pulavarty027445c2006-09-30 23:28:46 -0700926 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927}
928
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929/*
930 * Atomic setting of ioctl hooks to avoid race
931 * with module unload.
932 */
933
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800934static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700935static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936
Eric W. Biederman881d9662007-09-17 11:56:21 -0700937void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800939 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800941 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942}
943EXPORT_SYMBOL(brioctl_set);
944
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800945static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700946static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947
Eric W. Biederman881d9662007-09-17 11:56:21 -0700948void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800950 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800952 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953}
954EXPORT_SYMBOL(vlan_ioctl_set);
955
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800956static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700957static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700959void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800961 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800963 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964}
965EXPORT_SYMBOL(dlci_ioctl_set);
966
Arnd Bergmann6b960182009-11-06 23:10:54 -0800967static long sock_do_ioctl(struct net *net, struct socket *sock,
968 unsigned int cmd, unsigned long arg)
969{
970 int err;
971 void __user *argp = (void __user *)arg;
972
973 err = sock->ops->ioctl(sock, cmd, arg);
974
975 /*
976 * If this ioctl is unknown try to hand it down
977 * to the NIC driver.
978 */
979 if (err == -ENOIOCTLCMD)
980 err = dev_ioctl(net, cmd, argp);
981
982 return err;
983}
984
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985/*
986 * With an ioctl, arg may well be a user mode pointer, but we don't know
987 * what to do with it - that's up to the protocol still.
988 */
989
990static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
991{
992 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700993 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 void __user *argp = (void __user *)arg;
995 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700996 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
Eric Dumazetb69aee02005-09-06 14:42:45 -0700998 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700999 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001000 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001002 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001004#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001006 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001008#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001009 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 case FIOSETOWN:
1011 case SIOCSPGRP:
1012 err = -EFAULT;
1013 if (get_user(pid, (int __user *)argp))
1014 break;
1015 err = f_setown(sock->file, pid, 1);
1016 break;
1017 case FIOGETOWN:
1018 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001019 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001020 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 break;
1022 case SIOCGIFBR:
1023 case SIOCSIFBR:
1024 case SIOCBRADDBR:
1025 case SIOCBRDELBR:
1026 err = -ENOPKG;
1027 if (!br_ioctl_hook)
1028 request_module("bridge");
1029
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001030 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001031 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001032 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001033 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 break;
1035 case SIOCGIFVLAN:
1036 case SIOCSIFVLAN:
1037 err = -ENOPKG;
1038 if (!vlan_ioctl_hook)
1039 request_module("8021q");
1040
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001041 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001043 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001044 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 case SIOCADDDLCI:
1047 case SIOCDELDLCI:
1048 err = -ENOPKG;
1049 if (!dlci_ioctl_hook)
1050 request_module("dlci");
1051
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001052 mutex_lock(&dlci_ioctl_mutex);
1053 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001055 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 break;
1057 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001058 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001060 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 return err;
1062}
1063
1064int sock_create_lite(int family, int type, int protocol, struct socket **res)
1065{
1066 int err;
1067 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001068
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 err = security_socket_create(family, type, protocol, 1);
1070 if (err)
1071 goto out;
1072
1073 sock = sock_alloc();
1074 if (!sock) {
1075 err = -ENOMEM;
1076 goto out;
1077 }
1078
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001080 err = security_socket_post_create(sock, family, type, protocol, 1);
1081 if (err)
1082 goto out_release;
1083
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084out:
1085 *res = sock;
1086 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001087out_release:
1088 sock_release(sock);
1089 sock = NULL;
1090 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001092EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
1094/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001095static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096{
1097 struct socket *sock;
1098
1099 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001100 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001102 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 return sock->ops->poll(file, sock, wait);
1104}
1105
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001106static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001108 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109
1110 return sock->ops->mmap(file, sock, vma);
1111}
1112
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001113static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114{
1115 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001116 * It was possible the inode is NULL we were
1117 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 */
1119
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001120 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 printk(KERN_DEBUG "sock_close: NULL inode\n");
1122 return 0;
1123 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 sock_release(SOCKET_I(inode));
1125 return 0;
1126}
1127
1128/*
1129 * Update the socket async list
1130 *
1131 * Fasync_list locking strategy.
1132 *
1133 * 1. fasync_list is modified only under process context socket lock
1134 * i.e. under semaphore.
1135 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001136 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 */
1138
1139static int sock_fasync(int fd, struct file *filp, int on)
1140{
Eric Dumazet989a2972010-04-14 09:55:35 +00001141 struct socket *sock = filp->private_data;
1142 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001143 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
Eric Dumazet989a2972010-04-14 09:55:35 +00001145 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001149 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1150 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
Eric Dumazeteaefd112011-02-18 03:26:36 +00001152 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001153 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001154 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001155 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
Eric Dumazet989a2972010-04-14 09:55:35 +00001157 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 return 0;
1159}
1160
Eric Dumazet43815482010-04-29 11:01:49 +00001161/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
1163int sock_wake_async(struct socket *sock, int how, int band)
1164{
Eric Dumazet43815482010-04-29 11:01:49 +00001165 struct socket_wq *wq;
1166
1167 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001169 rcu_read_lock();
1170 wq = rcu_dereference(sock->wq);
1171 if (!wq || !wq->fasync_list) {
1172 rcu_read_unlock();
1173 return -1;
1174 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001175 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001176 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1178 break;
1179 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001180 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1182 break;
1183 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001184 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001185call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001186 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001188 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001189 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 }
Eric Dumazet43815482010-04-29 11:01:49 +00001191 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 return 0;
1193}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001194EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195
Pavel Emelyanov721db932010-09-29 16:06:32 +04001196int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001197 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198{
1199 int err;
1200 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001201 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202
1203 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001204 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 */
1206 if (family < 0 || family >= NPROTO)
1207 return -EAFNOSUPPORT;
1208 if (type < 0 || type >= SOCK_MAX)
1209 return -EINVAL;
1210
1211 /* Compatibility.
1212
1213 This uglymoron is moved from INET layer to here to avoid
1214 deadlock in module load.
1215 */
1216 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001217 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 if (!warned) {
1219 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001220 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1221 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 }
1223 family = PF_PACKET;
1224 }
1225
1226 err = security_socket_create(family, type, protocol, kern);
1227 if (err)
1228 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001229
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001230 /*
1231 * Allocate the socket and allow the family to set things up. if
1232 * the protocol is 0, the family is instructed to select an appropriate
1233 * default.
1234 */
1235 sock = sock_alloc();
1236 if (!sock) {
1237 if (net_ratelimit())
1238 printk(KERN_WARNING "socket: no more sockets\n");
1239 return -ENFILE; /* Not exactly a match, but its the
1240 closest posix thing */
1241 }
1242
1243 sock->type = type;
1244
Johannes Berg95a5afc2008-10-16 15:24:51 -07001245#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001246 /* Attempt to load a protocol module if the find failed.
1247 *
1248 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 * requested real, full-featured networking support upon configuration.
1250 * Otherwise module support will break!
1251 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001252 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001253 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254#endif
1255
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001256 rcu_read_lock();
1257 pf = rcu_dereference(net_families[family]);
1258 err = -EAFNOSUPPORT;
1259 if (!pf)
1260 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261
1262 /*
1263 * We will call the ->create function, that possibly is in a loadable
1264 * module, so we have to bump that loadable module refcnt first.
1265 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001266 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 goto out_release;
1268
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001269 /* Now protected by module ref count */
1270 rcu_read_unlock();
1271
Eric Paris3f378b62009-11-05 22:18:14 -08001272 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001273 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001275
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 /*
1277 * Now to bump the refcnt of the [loadable] module that owns this
1278 * socket at sock_release time we decrement its refcnt.
1279 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001280 if (!try_module_get(sock->ops->owner))
1281 goto out_module_busy;
1282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 /*
1284 * Now that we're done with the ->create function, the [loadable]
1285 * module can have its refcnt decremented
1286 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001287 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001288 err = security_socket_post_create(sock, family, type, protocol, kern);
1289 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001290 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001291 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001293 return 0;
1294
1295out_module_busy:
1296 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001298 sock->ops = NULL;
1299 module_put(pf->owner);
1300out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001302 return err;
1303
1304out_release:
1305 rcu_read_unlock();
1306 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001308EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309
1310int sock_create(int family, int type, int protocol, struct socket **res)
1311{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001312 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001314EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
1316int sock_create_kern(int family, int type, int protocol, struct socket **res)
1317{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001318 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001320EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001322SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323{
1324 int retval;
1325 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001326 int flags;
1327
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001328 /* Check the SOCK_* constants for consistency. */
1329 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1330 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1331 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1332 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1333
Ulrich Dreppera677a032008-07-23 21:29:17 -07001334 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001335 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001336 return -EINVAL;
1337 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001339 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1340 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1341
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 retval = sock_create(family, type, protocol, &sock);
1343 if (retval < 0)
1344 goto out;
1345
Ulrich Drepper77d27202008-07-23 21:29:35 -07001346 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347 if (retval < 0)
1348 goto out_release;
1349
1350out:
1351 /* It may be already another descriptor 8) Not kernel problem. */
1352 return retval;
1353
1354out_release:
1355 sock_release(sock);
1356 return retval;
1357}
1358
1359/*
1360 * Create a pair of connected sockets.
1361 */
1362
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001363SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1364 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365{
1366 struct socket *sock1, *sock2;
1367 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001368 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001369 int flags;
1370
1371 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001372 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001373 return -EINVAL;
1374 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001376 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1377 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1378
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 /*
1380 * Obtain the first socket and check if the underlying protocol
1381 * supports the socketpair call.
1382 */
1383
1384 err = sock_create(family, type, protocol, &sock1);
1385 if (err < 0)
1386 goto out;
1387
1388 err = sock_create(family, type, protocol, &sock2);
1389 if (err < 0)
1390 goto out_release_1;
1391
1392 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001393 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 goto out_release_both;
1395
Al Viro7cbe66b2009-08-05 19:59:08 +04001396 fd1 = sock_alloc_file(sock1, &newfile1, flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001397 if (unlikely(fd1 < 0)) {
1398 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001400 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401
Al Viro7cbe66b2009-08-05 19:59:08 +04001402 fd2 = sock_alloc_file(sock2, &newfile2, flags);
Al Viro198de4d2009-08-05 19:29:23 +04001403 if (unlikely(fd2 < 0)) {
1404 err = fd2;
1405 fput(newfile1);
1406 put_unused_fd(fd1);
1407 sock_release(sock2);
1408 goto out;
Al Virodb349502007-02-07 01:48:00 -05001409 }
1410
Al Viro157cf642008-12-14 04:57:47 -05001411 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001412 fd_install(fd1, newfile1);
1413 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 /* fd1 and fd2 may be already another descriptors.
1415 * Not kernel problem.
1416 */
1417
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001418 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 if (!err)
1420 err = put_user(fd2, &usockvec[1]);
1421 if (!err)
1422 return 0;
1423
1424 sys_close(fd2);
1425 sys_close(fd1);
1426 return err;
1427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001429 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001431 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432out:
1433 return err;
1434}
1435
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436/*
1437 * Bind a name to a socket. Nothing much to do here since it's
1438 * the protocol's responsibility to handle the local address.
1439 *
1440 * We move the socket address to kernel space before we call
1441 * the protocol layer (having also checked the address is ok).
1442 */
1443
Heiko Carstens20f37032009-01-14 14:14:23 +01001444SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445{
1446 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001447 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001448 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001450 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001451 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001452 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001453 if (err >= 0) {
1454 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001455 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001456 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001457 if (!err)
1458 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001459 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001460 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001462 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001463 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 return err;
1465}
1466
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467/*
1468 * Perform a listen. Basically, we allow the protocol to do anything
1469 * necessary for a listen, and if that works, we mark the socket as
1470 * ready for listening.
1471 */
1472
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001473SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474{
1475 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001476 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001477 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001478
1479 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1480 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001481 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001482 if ((unsigned)backlog > somaxconn)
1483 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484
1485 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001486 if (!err)
1487 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001489 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 }
1491 return err;
1492}
1493
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494/*
1495 * For accept, we attempt to create a new socket, set up the link
1496 * with the client, wake up the client, then return the new
1497 * connected fd. We collect the address of the connector in kernel
1498 * space and move it to user at the very end. This is unclean because
1499 * we open the socket then return an error.
1500 *
1501 * 1003.1g adds the ability to recvmsg() to query connection pending
1502 * status to recvmsg. We need to add that support in a way thats
1503 * clean when we restucture accept also.
1504 */
1505
Heiko Carstens20f37032009-01-14 14:14:23 +01001506SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1507 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508{
1509 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001510 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001511 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001512 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513
Ulrich Drepper77d27202008-07-23 21:29:35 -07001514 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001515 return -EINVAL;
1516
1517 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1518 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1519
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001520 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 if (!sock)
1522 goto out;
1523
1524 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001525 newsock = sock_alloc();
1526 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 goto out_put;
1528
1529 newsock->type = sock->type;
1530 newsock->ops = sock->ops;
1531
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 /*
1533 * We don't need try_module_get here, as the listening socket (sock)
1534 * has the protocol module (sock->ops->owner) held.
1535 */
1536 __module_get(newsock->ops->owner);
1537
Al Viro7cbe66b2009-08-05 19:59:08 +04001538 newfd = sock_alloc_file(newsock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001539 if (unlikely(newfd < 0)) {
1540 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001541 sock_release(newsock);
1542 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001543 }
1544
Frank Filza79af592005-09-27 15:23:38 -07001545 err = security_socket_accept(sock, newsock);
1546 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001547 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001548
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1550 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001551 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552
1553 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001554 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001555 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001557 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001559 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001560 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001562 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 }
1564
1565 /* File flags are not inherited via accept() unlike another OSes. */
1566
David S. Miller39d8c1b2006-03-20 17:13:49 -08001567 fd_install(newfd, newfile);
1568 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001571 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572out:
1573 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001574out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001575 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001576 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 goto out_put;
1578}
1579
Heiko Carstens20f37032009-01-14 14:14:23 +01001580SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1581 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001582{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001583 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001584}
1585
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586/*
1587 * Attempt to connect to a socket with the server address. The address
1588 * is in user space so we verify it is OK and move it to kernel space.
1589 *
1590 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1591 * break bindings
1592 *
1593 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1594 * other SEQPACKET protocols that take time to connect() as it doesn't
1595 * include the -EINPROGRESS status for such sockets.
1596 */
1597
Heiko Carstens20f37032009-01-14 14:14:23 +01001598SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1599 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600{
1601 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001602 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001603 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001605 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 if (!sock)
1607 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001608 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 if (err < 0)
1610 goto out_put;
1611
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001612 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001613 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 if (err)
1615 goto out_put;
1616
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001617 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 sock->file->f_flags);
1619out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001620 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621out:
1622 return err;
1623}
1624
1625/*
1626 * Get the local address ('name') of a socket object. Move the obtained
1627 * name to user space.
1628 */
1629
Heiko Carstens20f37032009-01-14 14:14:23 +01001630SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1631 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632{
1633 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001634 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001635 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001636
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001637 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 if (!sock)
1639 goto out;
1640
1641 err = security_socket_getsockname(sock);
1642 if (err)
1643 goto out_put;
1644
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001645 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 if (err)
1647 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001648 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649
1650out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001651 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652out:
1653 return err;
1654}
1655
1656/*
1657 * Get the remote address ('name') of a socket object. Move the obtained
1658 * name to user space.
1659 */
1660
Heiko Carstens20f37032009-01-14 14:14:23 +01001661SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1662 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663{
1664 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001665 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001666 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001668 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1669 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 err = security_socket_getpeername(sock);
1671 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001672 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 return err;
1674 }
1675
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001676 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001677 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001678 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001680 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001681 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001682 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 }
1684 return err;
1685}
1686
1687/*
1688 * Send a datagram to a given address. We move the address into kernel
1689 * space and check the user space data area is readable before invoking
1690 * the protocol.
1691 */
1692
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001693SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1694 unsigned, flags, struct sockaddr __user *, addr,
1695 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696{
1697 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001698 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 int err;
1700 struct msghdr msg;
1701 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001702 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001703
Linus Torvalds253eacc2010-10-30 16:43:10 -07001704 if (len > INT_MAX)
1705 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001706 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1707 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001708 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001709
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001710 iov.iov_base = buff;
1711 iov.iov_len = len;
1712 msg.msg_name = NULL;
1713 msg.msg_iov = &iov;
1714 msg.msg_iovlen = 1;
1715 msg.msg_control = NULL;
1716 msg.msg_controllen = 0;
1717 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001718 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001719 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 if (err < 0)
1721 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001722 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001723 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 }
1725 if (sock->file->f_flags & O_NONBLOCK)
1726 flags |= MSG_DONTWAIT;
1727 msg.msg_flags = flags;
1728 err = sock_sendmsg(sock, &msg, len);
1729
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001730out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001731 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001732out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 return err;
1734}
1735
1736/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001737 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 */
1739
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001740SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1741 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742{
1743 return sys_sendto(fd, buff, len, flags, NULL, 0);
1744}
1745
1746/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001747 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 * sender. We verify the buffers are writable and if needed move the
1749 * sender address from kernel to user space.
1750 */
1751
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001752SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1753 unsigned, flags, struct sockaddr __user *, addr,
1754 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755{
1756 struct socket *sock;
1757 struct iovec iov;
1758 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001759 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001760 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001761 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
Linus Torvalds253eacc2010-10-30 16:43:10 -07001763 if (size > INT_MAX)
1764 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001765 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001767 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001769 msg.msg_control = NULL;
1770 msg.msg_controllen = 0;
1771 msg.msg_iovlen = 1;
1772 msg.msg_iov = &iov;
1773 iov.iov_len = size;
1774 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001775 msg.msg_name = (struct sockaddr *)&address;
1776 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 if (sock->file->f_flags & O_NONBLOCK)
1778 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001779 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001781 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001782 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001783 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001784 if (err2 < 0)
1785 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001787
1788 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001789out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 return err;
1791}
1792
1793/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001794 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 */
1796
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1798 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799{
1800 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1801}
1802
1803/*
1804 * Set a socket option. Because we don't know the option lengths we have
1805 * to pass the user mode parameter for the protocols to sort out.
1806 */
1807
Heiko Carstens20f37032009-01-14 14:14:23 +01001808SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1809 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001811 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 struct socket *sock;
1813
1814 if (optlen < 0)
1815 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001816
1817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1818 if (sock != NULL) {
1819 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001820 if (err)
1821 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
1823 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001824 err =
1825 sock_setsockopt(sock, level, optname, optval,
1826 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 err =
1829 sock->ops->setsockopt(sock, level, optname, optval,
1830 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001831out_put:
1832 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 }
1834 return err;
1835}
1836
1837/*
1838 * Get a socket option. Because we don't know the option lengths we have
1839 * to pass a user mode parameter for the protocols to sort out.
1840 */
1841
Heiko Carstens20f37032009-01-14 14:14:23 +01001842SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1843 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001845 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 struct socket *sock;
1847
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001848 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1849 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001850 err = security_socket_getsockopt(sock, level, optname);
1851 if (err)
1852 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853
1854 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001855 err =
1856 sock_getsockopt(sock, level, optname, optval,
1857 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001859 err =
1860 sock->ops->getsockopt(sock, level, optname, optval,
1861 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001862out_put:
1863 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 }
1865 return err;
1866}
1867
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868/*
1869 * Shutdown a socket.
1870 */
1871
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001872SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001874 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 struct socket *sock;
1876
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001877 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1878 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001880 if (!err)
1881 err = sock->ops->shutdown(sock, how);
1882 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 }
1884 return err;
1885}
1886
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001887/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 * fields which are the same type (int / unsigned) on our platforms.
1889 */
1890#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1891#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1892#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1893
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001894struct used_address {
1895 struct sockaddr_storage name;
1896 unsigned int name_len;
1897};
1898
Anton Blanchard228e5482011-05-02 20:21:35 +00001899static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001900 struct msghdr *msg_sys, unsigned flags,
1901 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001903 struct compat_msghdr __user *msg_compat =
1904 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001905 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001907 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1909 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 unsigned char *ctl_buf = ctl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 int err, ctl_len, iov_size, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001912
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 err = -EFAULT;
1914 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001915 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001917 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 return -EFAULT;
1919
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 /* do not move before msg_sys is valid */
1921 err = -EMSGSIZE;
Anton Blanchard228e5482011-05-02 20:21:35 +00001922 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1923 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001925 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 err = -ENOMEM;
Anton Blanchard228e5482011-05-02 20:21:35 +00001927 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1928 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1930 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00001931 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 }
1933
1934 /* This will also move the address data into kernel space */
1935 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001936 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001938 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001939 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 goto out_freeiov;
1941 total_len = err;
1942
1943 err = -ENOBUFS;
1944
Anton Blanchard228e5482011-05-02 20:21:35 +00001945 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001947 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001949 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001950 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001951 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 if (err)
1953 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001954 ctl_buf = msg_sys->msg_control;
1955 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001957 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001959 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 goto out_freeiov;
1961 }
1962 err = -EFAULT;
1963 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00001964 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1966 * checking falls down on this.
1967 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001968 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00001969 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001970 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00001972 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 }
Anton Blanchard228e5482011-05-02 20:21:35 +00001974 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975
1976 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00001977 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001978 /*
1979 * If this is sendmmsg() and current destination address is same as
1980 * previously succeeded address, omit asking LSM's decision.
1981 * used_address->name_len is initialized to UINT_MAX so that the first
1982 * destination address never matches.
1983 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001984 if (used_address && msg_sys->msg_name &&
1985 used_address->name_len == msg_sys->msg_namelen &&
1986 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001987 used_address->name_len)) {
1988 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1989 goto out_freectl;
1990 }
1991 err = sock_sendmsg(sock, msg_sys, total_len);
1992 /*
1993 * If this is sendmmsg() and sending to current destination address was
1994 * successful, remember it.
1995 */
1996 if (used_address && err >= 0) {
1997 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001998 if (msg_sys->msg_name)
1999 memcpy(&used_address->name, msg_sys->msg_name,
2000 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002001 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002
2003out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002004 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2006out_freeiov:
2007 if (iov != iovstack)
2008 sock_kfree_s(sock->sk, iov, iov_size);
Anton Blanchard228e5482011-05-02 20:21:35 +00002009out:
2010 return err;
2011}
2012
2013/*
2014 * BSD sendmsg interface
2015 */
2016
2017SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
2018{
2019 int fput_needed, err;
2020 struct msghdr msg_sys;
2021 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2022
2023 if (!sock)
2024 goto out;
2025
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002026 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002027
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002028 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002029out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 return err;
2031}
2032
Anton Blanchard228e5482011-05-02 20:21:35 +00002033/*
2034 * Linux sendmmsg interface
2035 */
2036
2037int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2038 unsigned int flags)
2039{
2040 int fput_needed, err, datagrams;
2041 struct socket *sock;
2042 struct mmsghdr __user *entry;
2043 struct compat_mmsghdr __user *compat_entry;
2044 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002045 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002046
Anton Blanchard98382f42011-08-04 14:07:39 +00002047 if (vlen > UIO_MAXIOV)
2048 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002049
2050 datagrams = 0;
2051
2052 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2053 if (!sock)
2054 return err;
2055
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002056 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002057 entry = mmsg;
2058 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002059 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002060
2061 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002062 if (MSG_CMSG_COMPAT & flags) {
2063 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002064 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002065 if (err < 0)
2066 break;
2067 err = __put_user(err, &compat_entry->msg_len);
2068 ++compat_entry;
2069 } else {
2070 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002071 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002072 if (err < 0)
2073 break;
2074 err = put_user(err, &entry->msg_len);
2075 ++entry;
2076 }
2077
2078 if (err)
2079 break;
2080 ++datagrams;
2081 }
2082
Anton Blanchard228e5482011-05-02 20:21:35 +00002083 fput_light(sock->file, fput_needed);
2084
Anton Blanchard728ffb82011-08-04 14:07:38 +00002085 /* We only return an error if no datagrams were able to be sent */
2086 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002087 return datagrams;
2088
Anton Blanchard228e5482011-05-02 20:21:35 +00002089 return err;
2090}
2091
2092SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2093 unsigned int, vlen, unsigned int, flags)
2094{
2095 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2096}
2097
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002098static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2099 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002101 struct compat_msghdr __user *msg_compat =
2102 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002104 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 unsigned long cmsg_ptr;
2106 int err, iov_size, total_len, len;
2107
2108 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002109 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110
2111 /* user mode address pointers */
2112 struct sockaddr __user *uaddr;
2113 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002114
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002116 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002118 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002119 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002122 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2123 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002124
2125 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002127 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2128 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2130 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002131 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132 }
2133
2134 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002135 * Save the user-mode address (verify_iovec will change the
2136 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002138
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002139 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 uaddr_len = COMPAT_NAMELEN(msg);
2141 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002142 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002144 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 if (err < 0)
2146 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002147 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002149 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2150 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002151
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 if (sock->file->f_flags & O_NONBLOCK)
2153 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002154 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2155 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 if (err < 0)
2157 goto out_freeiov;
2158 len = err;
2159
2160 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002161 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002162 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002163 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164 if (err < 0)
2165 goto out_freeiov;
2166 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002167 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002168 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002169 if (err)
2170 goto out_freeiov;
2171 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002172 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173 &msg_compat->msg_controllen);
2174 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002175 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 &msg->msg_controllen);
2177 if (err)
2178 goto out_freeiov;
2179 err = len;
2180
2181out_freeiov:
2182 if (iov != iovstack)
2183 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002184out:
2185 return err;
2186}
2187
2188/*
2189 * BSD recvmsg interface
2190 */
2191
2192SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2193 unsigned int, flags)
2194{
2195 int fput_needed, err;
2196 struct msghdr msg_sys;
2197 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2198
2199 if (!sock)
2200 goto out;
2201
2202 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2203
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002204 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205out:
2206 return err;
2207}
2208
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002209/*
2210 * Linux recvmmsg interface
2211 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002213int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2214 unsigned int flags, struct timespec *timeout)
2215{
2216 int fput_needed, err, datagrams;
2217 struct socket *sock;
2218 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002219 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002220 struct msghdr msg_sys;
2221 struct timespec end_time;
2222
2223 if (timeout &&
2224 poll_select_set_timeout(&end_time, timeout->tv_sec,
2225 timeout->tv_nsec))
2226 return -EINVAL;
2227
2228 datagrams = 0;
2229
2230 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2231 if (!sock)
2232 return err;
2233
2234 err = sock_error(sock->sk);
2235 if (err)
2236 goto out_put;
2237
2238 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002239 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002240
2241 while (datagrams < vlen) {
2242 /*
2243 * No need to ask LSM for more than the first datagram.
2244 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002245 if (MSG_CMSG_COMPAT & flags) {
2246 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002247 &msg_sys, flags & ~MSG_WAITFORONE,
2248 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002249 if (err < 0)
2250 break;
2251 err = __put_user(err, &compat_entry->msg_len);
2252 ++compat_entry;
2253 } else {
2254 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002255 &msg_sys, flags & ~MSG_WAITFORONE,
2256 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002257 if (err < 0)
2258 break;
2259 err = put_user(err, &entry->msg_len);
2260 ++entry;
2261 }
2262
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002263 if (err)
2264 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002265 ++datagrams;
2266
Brandon L Black71c5c152010-03-26 16:18:03 +00002267 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2268 if (flags & MSG_WAITFORONE)
2269 flags |= MSG_DONTWAIT;
2270
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002271 if (timeout) {
2272 ktime_get_ts(timeout);
2273 *timeout = timespec_sub(end_time, *timeout);
2274 if (timeout->tv_sec < 0) {
2275 timeout->tv_sec = timeout->tv_nsec = 0;
2276 break;
2277 }
2278
2279 /* Timeout, return less than vlen datagrams */
2280 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2281 break;
2282 }
2283
2284 /* Out of band data, return right away */
2285 if (msg_sys.msg_flags & MSG_OOB)
2286 break;
2287 }
2288
2289out_put:
2290 fput_light(sock->file, fput_needed);
2291
2292 if (err == 0)
2293 return datagrams;
2294
2295 if (datagrams != 0) {
2296 /*
2297 * We may return less entries than requested (vlen) if the
2298 * sock is non block and there aren't enough datagrams...
2299 */
2300 if (err != -EAGAIN) {
2301 /*
2302 * ... or if recvmsg returns an error after we
2303 * received some datagrams, where we record the
2304 * error to return on the next call or if the
2305 * app asks about it using getsockopt(SO_ERROR).
2306 */
2307 sock->sk->sk_err = -err;
2308 }
2309
2310 return datagrams;
2311 }
2312
2313 return err;
2314}
2315
2316SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2317 unsigned int, vlen, unsigned int, flags,
2318 struct timespec __user *, timeout)
2319{
2320 int datagrams;
2321 struct timespec timeout_sys;
2322
2323 if (!timeout)
2324 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2325
2326 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2327 return -EFAULT;
2328
2329 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2330
2331 if (datagrams > 0 &&
2332 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2333 datagrams = -EFAULT;
2334
2335 return datagrams;
2336}
2337
2338#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339/* Argument list sizes for sys_socketcall */
2340#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002341static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002342 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2343 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2344 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002345 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002346};
2347
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348#undef AL
2349
2350/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002351 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 *
2353 * Argument checking cleaned up. Saved 20% in size.
2354 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002355 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356 */
2357
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002358SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359{
2360 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002361 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002363 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364
Anton Blanchard228e5482011-05-02 20:21:35 +00002365 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366 return -EINVAL;
2367
Arjan van de Ven47379052009-09-28 12:57:44 -07002368 len = nargs[call];
2369 if (len > sizeof(a))
2370 return -EINVAL;
2371
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002373 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002375
Al Virof3298dc2008-12-10 03:16:51 -05002376 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002377
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002378 a0 = a[0];
2379 a1 = a[1];
2380
2381 switch (call) {
2382 case SYS_SOCKET:
2383 err = sys_socket(a0, a1, a[2]);
2384 break;
2385 case SYS_BIND:
2386 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2387 break;
2388 case SYS_CONNECT:
2389 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2390 break;
2391 case SYS_LISTEN:
2392 err = sys_listen(a0, a1);
2393 break;
2394 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002395 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2396 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002397 break;
2398 case SYS_GETSOCKNAME:
2399 err =
2400 sys_getsockname(a0, (struct sockaddr __user *)a1,
2401 (int __user *)a[2]);
2402 break;
2403 case SYS_GETPEERNAME:
2404 err =
2405 sys_getpeername(a0, (struct sockaddr __user *)a1,
2406 (int __user *)a[2]);
2407 break;
2408 case SYS_SOCKETPAIR:
2409 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2410 break;
2411 case SYS_SEND:
2412 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2413 break;
2414 case SYS_SENDTO:
2415 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2416 (struct sockaddr __user *)a[4], a[5]);
2417 break;
2418 case SYS_RECV:
2419 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2420 break;
2421 case SYS_RECVFROM:
2422 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2423 (struct sockaddr __user *)a[4],
2424 (int __user *)a[5]);
2425 break;
2426 case SYS_SHUTDOWN:
2427 err = sys_shutdown(a0, a1);
2428 break;
2429 case SYS_SETSOCKOPT:
2430 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2431 break;
2432 case SYS_GETSOCKOPT:
2433 err =
2434 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2435 (int __user *)a[4]);
2436 break;
2437 case SYS_SENDMSG:
2438 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2439 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002440 case SYS_SENDMMSG:
2441 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2442 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002443 case SYS_RECVMSG:
2444 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2445 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002446 case SYS_RECVMMSG:
2447 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2448 (struct timespec __user *)a[4]);
2449 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002450 case SYS_ACCEPT4:
2451 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2452 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002453 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002454 default:
2455 err = -EINVAL;
2456 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457 }
2458 return err;
2459}
2460
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002461#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002463/**
2464 * sock_register - add a socket protocol handler
2465 * @ops: description of protocol
2466 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467 * This function is called by a protocol handler that wants to
2468 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002469 * socket interface. The value ops->family coresponds to the
2470 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002472int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473{
2474 int err;
2475
2476 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002477 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2478 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479 return -ENOBUFS;
2480 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002481
2482 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002483 if (rcu_dereference_protected(net_families[ops->family],
2484 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002485 err = -EEXIST;
2486 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002487 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002488 err = 0;
2489 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002490 spin_unlock(&net_family_lock);
2491
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002492 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493 return err;
2494}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002495EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002497/**
2498 * sock_unregister - remove a protocol handler
2499 * @family: protocol family to remove
2500 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002501 * This function is called by a protocol handler that wants to
2502 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002503 * new socket creation.
2504 *
2505 * If protocol handler is a module, then it can use module reference
2506 * counts to protect against new references. If protocol handler is not
2507 * a module then it needs to provide its own protection in
2508 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002510void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002512 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002514 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002515 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002516 spin_unlock(&net_family_lock);
2517
2518 synchronize_rcu();
2519
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002520 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002521}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002522EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523
Andi Kleen77d76ea2005-12-22 12:43:42 -08002524static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002525{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002526 int err;
2527
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002529 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002531
Linus Torvalds1da177e2005-04-16 15:20:36 -07002532 sk_init();
2533
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002535 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536 */
2537 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
2539 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002540 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541 */
2542
2543 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002544
2545 err = register_filesystem(&sock_fs_type);
2546 if (err)
2547 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002549 if (IS_ERR(sock_mnt)) {
2550 err = PTR_ERR(sock_mnt);
2551 goto out_mount;
2552 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002553
2554 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555 */
2556
2557#ifdef CONFIG_NETFILTER
2558 netfilter_init();
2559#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002560
Richard Cochranc1f19b52010-07-17 08:49:36 +00002561#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2562 skb_timestamping_init();
2563#endif
2564
Nick Pigginb3e19d92011-01-07 17:50:11 +11002565out:
2566 return err;
2567
2568out_mount:
2569 unregister_filesystem(&sock_fs_type);
2570out_fs:
2571 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572}
2573
Andi Kleen77d76ea2005-12-22 12:43:42 -08002574core_initcall(sock_init); /* early initcall */
2575
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576#ifdef CONFIG_PROC_FS
2577void socket_seq_show(struct seq_file *seq)
2578{
2579 int cpu;
2580 int counter = 0;
2581
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002582 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002583 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584
2585 /* It can be negative, by the way. 8) */
2586 if (counter < 0)
2587 counter = 0;
2588
2589 seq_printf(seq, "sockets: used %d\n", counter);
2590}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002591#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002593#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002594static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002595 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002596{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002597 mm_segment_t old_fs = get_fs();
2598 struct timeval ktv;
2599 int err;
2600
2601 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002602 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002603 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002604 if (!err)
2605 err = compat_put_timeval(up, &ktv);
2606
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002607 return err;
2608}
2609
Arnd Bergmann6b960182009-11-06 23:10:54 -08002610static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002611 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002612{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002613 mm_segment_t old_fs = get_fs();
2614 struct timespec kts;
2615 int err;
2616
2617 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002618 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002619 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002620 if (!err)
2621 err = compat_put_timespec(up, &kts);
2622
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002623 return err;
2624}
2625
Arnd Bergmann6b960182009-11-06 23:10:54 -08002626static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002627{
2628 struct ifreq __user *uifr;
2629 int err;
2630
2631 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002632 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002633 return -EFAULT;
2634
Arnd Bergmann6b960182009-11-06 23:10:54 -08002635 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002636 if (err)
2637 return err;
2638
Arnd Bergmann6b960182009-11-06 23:10:54 -08002639 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002640 return -EFAULT;
2641
2642 return 0;
2643}
2644
Arnd Bergmann6b960182009-11-06 23:10:54 -08002645static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002646{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002647 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002648 struct ifconf ifc;
2649 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002650 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002651 struct ifreq __user *ifr;
2652 unsigned int i, j;
2653 int err;
2654
Arnd Bergmann6b960182009-11-06 23:10:54 -08002655 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002656 return -EFAULT;
2657
2658 if (ifc32.ifcbuf == 0) {
2659 ifc32.ifc_len = 0;
2660 ifc.ifc_len = 0;
2661 ifc.ifc_req = NULL;
2662 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2663 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002664 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2665 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002666 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2667 ifc.ifc_len = len;
2668 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2669 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002670 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002671 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002672 return -EFAULT;
2673 ifr++;
2674 ifr32++;
2675 }
2676 }
2677 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2678 return -EFAULT;
2679
Arnd Bergmann6b960182009-11-06 23:10:54 -08002680 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002681 if (err)
2682 return err;
2683
2684 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2685 return -EFAULT;
2686
2687 ifr = ifc.ifc_req;
2688 ifr32 = compat_ptr(ifc32.ifcbuf);
2689 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002690 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2691 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2692 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002693 return -EFAULT;
2694 ifr32++;
2695 ifr++;
2696 }
2697
2698 if (ifc32.ifcbuf == 0) {
2699 /* Translate from 64-bit structure multiple to
2700 * a 32-bit one.
2701 */
2702 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002703 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002704 ifc32.ifc_len = i;
2705 } else {
2706 ifc32.ifc_len = i;
2707 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002708 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002709 return -EFAULT;
2710
2711 return 0;
2712}
2713
Arnd Bergmann6b960182009-11-06 23:10:54 -08002714static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002715{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002716 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2717 bool convert_in = false, convert_out = false;
2718 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2719 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002720 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002721 u32 rule_cnt = 0, actual_rule_cnt;
2722 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002723 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002724 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002725
2726 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2727 return -EFAULT;
2728
Ben Hutchings3a7da392011-03-17 07:34:32 +00002729 compat_rxnfc = compat_ptr(data);
2730
2731 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002732 return -EFAULT;
2733
Ben Hutchings3a7da392011-03-17 07:34:32 +00002734 /* Most ethtool structures are defined without padding.
2735 * Unfortunately struct ethtool_rxnfc is an exception.
2736 */
2737 switch (ethcmd) {
2738 default:
2739 break;
2740 case ETHTOOL_GRXCLSRLALL:
2741 /* Buffer size is variable */
2742 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2743 return -EFAULT;
2744 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2745 return -ENOMEM;
2746 buf_size += rule_cnt * sizeof(u32);
2747 /* fall through */
2748 case ETHTOOL_GRXRINGS:
2749 case ETHTOOL_GRXCLSRLCNT:
2750 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002751 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002752 convert_out = true;
2753 /* fall through */
2754 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002755 buf_size += sizeof(struct ethtool_rxnfc);
2756 convert_in = true;
2757 break;
2758 }
2759
2760 ifr = compat_alloc_user_space(buf_size);
2761 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2762
2763 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2764 return -EFAULT;
2765
2766 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2767 &ifr->ifr_ifru.ifru_data))
2768 return -EFAULT;
2769
2770 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002771 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002772 * fs.ring_cookie and at the end of fs, but nowhere else.
2773 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002774 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2775 sizeof(compat_rxnfc->fs.m_ext) !=
2776 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2777 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002778 BUILD_BUG_ON(
2779 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2780 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2781 offsetof(struct ethtool_rxnfc, fs.location) -
2782 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2783
2784 if (copy_in_user(rxnfc, compat_rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002785 (void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002786 (void *)rxnfc) ||
2787 copy_in_user(&rxnfc->fs.ring_cookie,
2788 &compat_rxnfc->fs.ring_cookie,
2789 (void *)(&rxnfc->fs.location + 1) -
2790 (void *)&rxnfc->fs.ring_cookie) ||
2791 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2792 sizeof(rxnfc->rule_cnt)))
2793 return -EFAULT;
2794 }
2795
2796 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2797 if (ret)
2798 return ret;
2799
2800 if (convert_out) {
2801 if (copy_in_user(compat_rxnfc, rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002802 (const void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002803 (const void *)rxnfc) ||
2804 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2805 &rxnfc->fs.ring_cookie,
2806 (const void *)(&rxnfc->fs.location + 1) -
2807 (const void *)&rxnfc->fs.ring_cookie) ||
2808 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2809 sizeof(rxnfc->rule_cnt)))
2810 return -EFAULT;
2811
2812 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2813 /* As an optimisation, we only copy the actual
2814 * number of rules that the underlying
2815 * function returned. Since Mallory might
2816 * change the rule count in user memory, we
2817 * check that it is less than the rule count
2818 * originally given (as the user buffer size),
2819 * which has been range-checked.
2820 */
2821 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2822 return -EFAULT;
2823 if (actual_rule_cnt < rule_cnt)
2824 rule_cnt = actual_rule_cnt;
2825 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2826 &rxnfc->rule_locs[0],
2827 rule_cnt * sizeof(u32)))
2828 return -EFAULT;
2829 }
2830 }
2831
2832 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002833}
2834
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002835static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2836{
2837 void __user *uptr;
2838 compat_uptr_t uptr32;
2839 struct ifreq __user *uifr;
2840
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002841 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002842 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2843 return -EFAULT;
2844
2845 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2846 return -EFAULT;
2847
2848 uptr = compat_ptr(uptr32);
2849
2850 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2851 return -EFAULT;
2852
2853 return dev_ioctl(net, SIOCWANDEV, uifr);
2854}
2855
Arnd Bergmann6b960182009-11-06 23:10:54 -08002856static int bond_ioctl(struct net *net, unsigned int cmd,
2857 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002858{
2859 struct ifreq kifr;
2860 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002861 mm_segment_t old_fs;
2862 int err;
2863 u32 data;
2864 void __user *datap;
2865
2866 switch (cmd) {
2867 case SIOCBONDENSLAVE:
2868 case SIOCBONDRELEASE:
2869 case SIOCBONDSETHWADDR:
2870 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002871 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002872 return -EFAULT;
2873
2874 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002875 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002876 err = dev_ioctl(net, cmd,
2877 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002878 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002879
2880 return err;
2881 case SIOCBONDSLAVEINFOQUERY:
2882 case SIOCBONDINFOQUERY:
2883 uifr = compat_alloc_user_space(sizeof(*uifr));
2884 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2885 return -EFAULT;
2886
2887 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2888 return -EFAULT;
2889
2890 datap = compat_ptr(data);
2891 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2892 return -EFAULT;
2893
Arnd Bergmann6b960182009-11-06 23:10:54 -08002894 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002895 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002896 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002897 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002898}
2899
Arnd Bergmann6b960182009-11-06 23:10:54 -08002900static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2901 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002902{
2903 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002904 char tmp_buf[IFNAMSIZ];
2905 void __user *data64;
2906 u32 data32;
2907
2908 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2909 IFNAMSIZ))
2910 return -EFAULT;
2911 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2912 return -EFAULT;
2913 data64 = compat_ptr(data32);
2914
2915 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2916
2917 /* Don't check these user accesses, just let that get trapped
2918 * in the ioctl handler instead.
2919 */
2920 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2921 IFNAMSIZ))
2922 return -EFAULT;
2923 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2924 return -EFAULT;
2925
Arnd Bergmann6b960182009-11-06 23:10:54 -08002926 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002927}
2928
Arnd Bergmann6b960182009-11-06 23:10:54 -08002929static int dev_ifsioc(struct net *net, struct socket *sock,
2930 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002931{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002932 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002933 int err;
2934
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002935 uifr = compat_alloc_user_space(sizeof(*uifr));
2936 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2937 return -EFAULT;
2938
2939 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2940
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002941 if (!err) {
2942 switch (cmd) {
2943 case SIOCGIFFLAGS:
2944 case SIOCGIFMETRIC:
2945 case SIOCGIFMTU:
2946 case SIOCGIFMEM:
2947 case SIOCGIFHWADDR:
2948 case SIOCGIFINDEX:
2949 case SIOCGIFADDR:
2950 case SIOCGIFBRDADDR:
2951 case SIOCGIFDSTADDR:
2952 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002953 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002954 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002955 case SIOCGMIIPHY:
2956 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002957 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002958 err = -EFAULT;
2959 break;
2960 }
2961 }
2962 return err;
2963}
2964
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002965static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2966 struct compat_ifreq __user *uifr32)
2967{
2968 struct ifreq ifr;
2969 struct compat_ifmap __user *uifmap32;
2970 mm_segment_t old_fs;
2971 int err;
2972
2973 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2974 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2975 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2976 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2977 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2978 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2979 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2980 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2981 if (err)
2982 return -EFAULT;
2983
2984 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002985 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002986 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002987 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002988
2989 if (cmd == SIOCGIFMAP && !err) {
2990 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2991 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2992 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2993 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2994 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2995 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2996 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2997 if (err)
2998 err = -EFAULT;
2999 }
3000 return err;
3001}
3002
3003static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3004{
3005 void __user *uptr;
3006 compat_uptr_t uptr32;
3007 struct ifreq __user *uifr;
3008
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003009 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003010 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3011 return -EFAULT;
3012
3013 if (get_user(uptr32, &uifr32->ifr_data))
3014 return -EFAULT;
3015
3016 uptr = compat_ptr(uptr32);
3017
3018 if (put_user(uptr, &uifr->ifr_data))
3019 return -EFAULT;
3020
3021 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3022}
3023
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003024struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003025 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003026 struct sockaddr rt_dst; /* target address */
3027 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3028 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003029 unsigned short rt_flags;
3030 short rt_pad2;
3031 u32 rt_pad3;
3032 unsigned char rt_tos;
3033 unsigned char rt_class;
3034 short rt_pad4;
3035 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003036 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003037 u32 rt_mtu; /* per route MTU/Window */
3038 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003039 unsigned short rt_irtt; /* Initial RTT */
3040};
3041
3042struct in6_rtmsg32 {
3043 struct in6_addr rtmsg_dst;
3044 struct in6_addr rtmsg_src;
3045 struct in6_addr rtmsg_gateway;
3046 u32 rtmsg_type;
3047 u16 rtmsg_dst_len;
3048 u16 rtmsg_src_len;
3049 u32 rtmsg_metric;
3050 u32 rtmsg_info;
3051 u32 rtmsg_flags;
3052 s32 rtmsg_ifindex;
3053};
3054
Arnd Bergmann6b960182009-11-06 23:10:54 -08003055static int routing_ioctl(struct net *net, struct socket *sock,
3056 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003057{
3058 int ret;
3059 void *r = NULL;
3060 struct in6_rtmsg r6;
3061 struct rtentry r4;
3062 char devname[16];
3063 u32 rtdev;
3064 mm_segment_t old_fs = get_fs();
3065
Arnd Bergmann6b960182009-11-06 23:10:54 -08003066 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3067 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003068 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003069 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003070 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3071 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3072 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3073 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3074 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3075 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3076 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003077
3078 r = (void *) &r6;
3079 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003080 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003081 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003082 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003083 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3084 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3085 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3086 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3087 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3088 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003089 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003090 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003091 r4.rt_dev = (char __user __force *)devname;
3092 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003093 } else
3094 r4.rt_dev = NULL;
3095
3096 r = (void *) &r4;
3097 }
3098
3099 if (ret) {
3100 ret = -EFAULT;
3101 goto out;
3102 }
3103
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003104 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003105 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003106 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003107
3108out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003109 return ret;
3110}
3111
3112/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3113 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003114 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003115 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003116static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003117{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003118 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003119
Arnd Bergmann6b960182009-11-06 23:10:54 -08003120 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003121 return -EFAULT;
3122 if (tmp == BRCTL_GET_VERSION)
3123 return BRCTL_VERSION + 1;
3124 return -EINVAL;
3125}
3126
Arnd Bergmann6b960182009-11-06 23:10:54 -08003127static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3128 unsigned int cmd, unsigned long arg)
3129{
3130 void __user *argp = compat_ptr(arg);
3131 struct sock *sk = sock->sk;
3132 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003133
Arnd Bergmann6b960182009-11-06 23:10:54 -08003134 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3135 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003136
Arnd Bergmann6b960182009-11-06 23:10:54 -08003137 switch (cmd) {
3138 case SIOCSIFBR:
3139 case SIOCGIFBR:
3140 return old_bridge_ioctl(argp);
3141 case SIOCGIFNAME:
3142 return dev_ifname32(net, argp);
3143 case SIOCGIFCONF:
3144 return dev_ifconf(net, argp);
3145 case SIOCETHTOOL:
3146 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003147 case SIOCWANDEV:
3148 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003149 case SIOCGIFMAP:
3150 case SIOCSIFMAP:
3151 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003152 case SIOCBONDENSLAVE:
3153 case SIOCBONDRELEASE:
3154 case SIOCBONDSETHWADDR:
3155 case SIOCBONDSLAVEINFOQUERY:
3156 case SIOCBONDINFOQUERY:
3157 case SIOCBONDCHANGEACTIVE:
3158 return bond_ioctl(net, cmd, argp);
3159 case SIOCADDRT:
3160 case SIOCDELRT:
3161 return routing_ioctl(net, sock, cmd, argp);
3162 case SIOCGSTAMP:
3163 return do_siocgstamp(net, sock, cmd, argp);
3164 case SIOCGSTAMPNS:
3165 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003166 case SIOCSHWTSTAMP:
3167 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003168
Arnd Bergmann6b960182009-11-06 23:10:54 -08003169 case FIOSETOWN:
3170 case SIOCSPGRP:
3171 case FIOGETOWN:
3172 case SIOCGPGRP:
3173 case SIOCBRADDBR:
3174 case SIOCBRDELBR:
3175 case SIOCGIFVLAN:
3176 case SIOCSIFVLAN:
3177 case SIOCADDDLCI:
3178 case SIOCDELDLCI:
3179 return sock_ioctl(file, cmd, arg);
3180
3181 case SIOCGIFFLAGS:
3182 case SIOCSIFFLAGS:
3183 case SIOCGIFMETRIC:
3184 case SIOCSIFMETRIC:
3185 case SIOCGIFMTU:
3186 case SIOCSIFMTU:
3187 case SIOCGIFMEM:
3188 case SIOCSIFMEM:
3189 case SIOCGIFHWADDR:
3190 case SIOCSIFHWADDR:
3191 case SIOCADDMULTI:
3192 case SIOCDELMULTI:
3193 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003194 case SIOCGIFADDR:
3195 case SIOCSIFADDR:
3196 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003197 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003198 case SIOCGIFBRDADDR:
3199 case SIOCSIFBRDADDR:
3200 case SIOCGIFDSTADDR:
3201 case SIOCSIFDSTADDR:
3202 case SIOCGIFNETMASK:
3203 case SIOCSIFNETMASK:
3204 case SIOCSIFPFLAGS:
3205 case SIOCGIFPFLAGS:
3206 case SIOCGIFTXQLEN:
3207 case SIOCSIFTXQLEN:
3208 case SIOCBRADDIF:
3209 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003210 case SIOCSIFNAME:
3211 case SIOCGMIIPHY:
3212 case SIOCGMIIREG:
3213 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003214 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003215
Arnd Bergmann6b960182009-11-06 23:10:54 -08003216 case SIOCSARP:
3217 case SIOCGARP:
3218 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003219 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003220 return sock_do_ioctl(net, sock, cmd, arg);
3221 }
3222
Arnd Bergmann6b960182009-11-06 23:10:54 -08003223 return -ENOIOCTLCMD;
3224}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003225
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003226static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003227 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003228{
3229 struct socket *sock = file->private_data;
3230 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003231 struct sock *sk;
3232 struct net *net;
3233
3234 sk = sock->sk;
3235 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003236
3237 if (sock->ops->compat_ioctl)
3238 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3239
David S. Miller87de87d2008-06-03 09:14:03 -07003240 if (ret == -ENOIOCTLCMD &&
3241 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3242 ret = compat_wext_handle_ioctl(net, cmd, arg);
3243
Arnd Bergmann6b960182009-11-06 23:10:54 -08003244 if (ret == -ENOIOCTLCMD)
3245 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3246
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003247 return ret;
3248}
3249#endif
3250
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003251int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3252{
3253 return sock->ops->bind(sock, addr, addrlen);
3254}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003255EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003256
3257int kernel_listen(struct socket *sock, int backlog)
3258{
3259 return sock->ops->listen(sock, backlog);
3260}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003261EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003262
3263int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3264{
3265 struct sock *sk = sock->sk;
3266 int err;
3267
3268 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3269 newsock);
3270 if (err < 0)
3271 goto done;
3272
3273 err = sock->ops->accept(sock, *newsock, flags);
3274 if (err < 0) {
3275 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003276 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003277 goto done;
3278 }
3279
3280 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003281 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003282
3283done:
3284 return err;
3285}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003286EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003287
3288int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003289 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003290{
3291 return sock->ops->connect(sock, addr, addrlen, flags);
3292}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003293EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003294
3295int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3296 int *addrlen)
3297{
3298 return sock->ops->getname(sock, addr, addrlen, 0);
3299}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003300EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003301
3302int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3303 int *addrlen)
3304{
3305 return sock->ops->getname(sock, addr, addrlen, 1);
3306}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003307EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003308
3309int kernel_getsockopt(struct socket *sock, int level, int optname,
3310 char *optval, int *optlen)
3311{
3312 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003313 char __user *uoptval;
3314 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003315 int err;
3316
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003317 uoptval = (char __user __force *) optval;
3318 uoptlen = (int __user __force *) optlen;
3319
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003320 set_fs(KERNEL_DS);
3321 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003322 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003323 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003324 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3325 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003326 set_fs(oldfs);
3327 return err;
3328}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003329EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003330
3331int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003332 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003333{
3334 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003335 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003336 int err;
3337
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003338 uoptval = (char __user __force *) optval;
3339
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003340 set_fs(KERNEL_DS);
3341 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003342 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003343 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003344 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003345 optlen);
3346 set_fs(oldfs);
3347 return err;
3348}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003349EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003350
3351int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3352 size_t size, int flags)
3353{
Herbert Xuf8451722010-05-24 00:12:34 -07003354 sock_update_classid(sock->sk);
3355
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003356 if (sock->ops->sendpage)
3357 return sock->ops->sendpage(sock, page, offset, size, flags);
3358
3359 return sock_no_sendpage(sock, page, offset, size, flags);
3360}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003361EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003362
3363int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3364{
3365 mm_segment_t oldfs = get_fs();
3366 int err;
3367
3368 set_fs(KERNEL_DS);
3369 err = sock->ops->ioctl(sock, cmd, arg);
3370 set_fs(oldfs);
3371
3372 return err;
3373}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003374EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003375
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003376int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3377{
3378 return sock->ops->shutdown(sock, how);
3379}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003380EXPORT_SYMBOL(kernel_sock_shutdown);