blob: 24a77400b65e4e8fd63f825144bdc4891b5fced9 [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
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *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
stephen hemminger11165f12010-10-18 14:27:29 +0000212static int move_addr_to_user(struct sockaddr *kaddr, int klen,
213 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 Dumazeteaefd1102011-02-18 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 Dumazeteaefd1102011-02-18 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 Dumazeteaefd1102011-02-18 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 Dumazeteaefd1102011-02-18 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 Dumazeteaefd1102011-02-18 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 Dumazeteaefd1102011-02-18 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;
Patrick Ohly20d49472009-02-12 05:03:38 +0000541 return 0;
542}
543EXPORT_SYMBOL(sock_tx_timestamp);
544
Anton Blanchard228e5482011-05-02 20:21:35 +0000545static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
546 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547{
548 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
Herbert Xuf8451722010-05-24 00:12:34 -0700550 sock_update_classid(sock->sk);
551
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 si->sock = sock;
553 si->scm = NULL;
554 si->msg = msg;
555 si->size = size;
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 return sock->ops->sendmsg(iocb, sock, msg, size);
558}
559
Anton Blanchard228e5482011-05-02 20:21:35 +0000560static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
561 struct msghdr *msg, size_t size)
562{
563 int err = security_socket_sendmsg(sock, msg, size);
564
565 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
566}
567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
569{
570 struct kiocb iocb;
571 struct sock_iocb siocb;
572 int ret;
573
574 init_sync_kiocb(&iocb, NULL);
575 iocb.private = &siocb;
576 ret = __sock_sendmsg(&iocb, sock, msg, size);
577 if (-EIOCBQUEUED == ret)
578 ret = wait_on_sync_kiocb(&iocb);
579 return ret;
580}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700581EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
Eric Dumazet894dc242011-07-26 02:39:41 +0000583static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000584{
585 struct kiocb iocb;
586 struct sock_iocb siocb;
587 int ret;
588
589 init_sync_kiocb(&iocb, NULL);
590 iocb.private = &siocb;
591 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
592 if (-EIOCBQUEUED == ret)
593 ret = wait_on_sync_kiocb(&iocb);
594 return ret;
595}
596
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
598 struct kvec *vec, size_t num, size_t size)
599{
600 mm_segment_t oldfs = get_fs();
601 int result;
602
603 set_fs(KERNEL_DS);
604 /*
605 * the following is safe, since for compiler definitions of kvec and
606 * iovec are identical, yielding the same in-core layout and alignment
607 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700608 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 msg->msg_iovlen = num;
610 result = sock_sendmsg(sock, msg, size);
611 set_fs(oldfs);
612 return result;
613}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700614EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615
Patrick Ohly20d49472009-02-12 05:03:38 +0000616static int ktime2ts(ktime_t kt, struct timespec *ts)
617{
618 if (kt.tv64) {
619 *ts = ktime_to_timespec(kt);
620 return 1;
621 } else {
622 return 0;
623 }
624}
625
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700626/*
627 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
628 */
629void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
630 struct sk_buff *skb)
631{
Patrick Ohly20d49472009-02-12 05:03:38 +0000632 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
633 struct timespec ts[3];
634 int empty = 1;
635 struct skb_shared_hwtstamps *shhwtstamps =
636 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700637
Patrick Ohly20d49472009-02-12 05:03:38 +0000638 /* Race occurred between timestamp enabling and packet
639 receiving. Fill in the current time for now. */
640 if (need_software_tstamp && skb->tstamp.tv64 == 0)
641 __net_timestamp(skb);
642
643 if (need_software_tstamp) {
644 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
645 struct timeval tv;
646 skb_get_timestamp(skb, &tv);
647 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
648 sizeof(tv), &tv);
649 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000650 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000651 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000652 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000653 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700654 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000655
656
657 memset(ts, 0, sizeof(ts));
658 if (skb->tstamp.tv64 &&
659 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
660 skb_get_timestampns(skb, ts + 0);
661 empty = 0;
662 }
663 if (shhwtstamps) {
664 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
665 ktime2ts(shhwtstamps->syststamp, ts + 1))
666 empty = 0;
667 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
668 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
669 empty = 0;
670 }
671 if (!empty)
672 put_cmsg(msg, SOL_SOCKET,
673 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700674}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300675EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
676
stephen hemminger11165f12010-10-18 14:27:29 +0000677static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
678 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700679{
680 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
681 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
682 sizeof(__u32), &skb->dropcount);
683}
684
Eric Dumazet767dd032010-04-28 19:14:43 +0000685void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700686 struct sk_buff *skb)
687{
688 sock_recv_timestamp(msg, sk, skb);
689 sock_recv_drops(msg, sk, skb);
690}
Eric Dumazet767dd032010-04-28 19:14:43 +0000691EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700692
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700693static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
694 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 struct sock_iocb *si = kiocb_to_siocb(iocb);
697
Herbert Xuf8451722010-05-24 00:12:34 -0700698 sock_update_classid(sock->sk);
699
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700 si->sock = sock;
701 si->scm = NULL;
702 si->msg = msg;
703 si->size = size;
704 si->flags = flags;
705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
707}
708
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700709static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
710 struct msghdr *msg, size_t size, int flags)
711{
712 int err = security_socket_recvmsg(sock, msg, size, flags);
713
714 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
715}
716
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700717int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 size_t size, int flags)
719{
720 struct kiocb iocb;
721 struct sock_iocb siocb;
722 int ret;
723
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700724 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725 iocb.private = &siocb;
726 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
727 if (-EIOCBQUEUED == ret)
728 ret = wait_on_sync_kiocb(&iocb);
729 return ret;
730}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700731EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700733static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
734 size_t size, int flags)
735{
736 struct kiocb iocb;
737 struct sock_iocb siocb;
738 int ret;
739
740 init_sync_kiocb(&iocb, NULL);
741 iocb.private = &siocb;
742 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
743 if (-EIOCBQUEUED == ret)
744 ret = wait_on_sync_kiocb(&iocb);
745 return ret;
746}
747
Martin Lucinac1249c02010-12-10 00:04:05 +0000748/**
749 * kernel_recvmsg - Receive a message from a socket (kernel space)
750 * @sock: The socket to receive the message from
751 * @msg: Received message
752 * @vec: Input s/g array for message data
753 * @num: Size of input s/g array
754 * @size: Number of bytes to read
755 * @flags: Message flags (MSG_DONTWAIT, etc...)
756 *
757 * On return the msg structure contains the scatter/gather array passed in the
758 * vec argument. The array is modified so that it consists of the unfilled
759 * portion of the original array.
760 *
761 * The returned value is the total number of bytes received, or an error.
762 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700763int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
764 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765{
766 mm_segment_t oldfs = get_fs();
767 int result;
768
769 set_fs(KERNEL_DS);
770 /*
771 * the following is safe, since for compiler definitions of kvec and
772 * iovec are identical, yielding the same in-core layout and alignment
773 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700774 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 result = sock_recvmsg(sock, msg, size, flags);
776 set_fs(oldfs);
777 return result;
778}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700779EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780
781static void sock_aio_dtor(struct kiocb *iocb)
782{
783 kfree(iocb->private);
784}
785
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300786static ssize_t sock_sendpage(struct file *file, struct page *page,
787 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788{
789 struct socket *sock;
790 int flags;
791
Eric Dumazetb69aee02005-09-06 14:42:45 -0700792 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
795 if (more)
796 flags |= MSG_MORE;
797
Linus Torvaldse6949582009-08-13 08:28:36 -0700798 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
800
Jens Axboe9c55e012007-11-06 23:30:13 -0800801static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700802 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800803 unsigned int flags)
804{
805 struct socket *sock = file->private_data;
806
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800807 if (unlikely(!sock->ops->splice_read))
808 return -EINVAL;
809
Herbert Xuf8451722010-05-24 00:12:34 -0700810 sock_update_classid(sock->sk);
811
Jens Axboe9c55e012007-11-06 23:30:13 -0800812 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
813}
814
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800815static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700816 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800817{
818 if (!is_sync_kiocb(iocb)) {
819 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
820 if (!siocb)
821 return NULL;
822 iocb->ki_dtor = sock_aio_dtor;
823 }
824
825 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800826 iocb->private = siocb;
827 return siocb;
828}
829
830static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700831 struct file *file, const struct iovec *iov,
832 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800833{
834 struct socket *sock = file->private_data;
835 size_t size = 0;
836 int i;
837
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700838 for (i = 0; i < nr_segs; i++)
839 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800840
841 msg->msg_name = NULL;
842 msg->msg_namelen = 0;
843 msg->msg_control = NULL;
844 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700845 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800846 msg->msg_iovlen = nr_segs;
847 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
848
849 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
850}
851
Badari Pulavarty027445c2006-09-30 23:28:46 -0700852static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
853 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800854{
855 struct sock_iocb siocb, *x;
856
857 if (pos != 0)
858 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700859
860 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800861 return 0;
862
Badari Pulavarty027445c2006-09-30 23:28:46 -0700863
864 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800865 if (!x)
866 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700867 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800868}
869
870static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700871 struct file *file, const struct iovec *iov,
872 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800873{
874 struct socket *sock = file->private_data;
875 size_t size = 0;
876 int i;
877
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700878 for (i = 0; i < nr_segs; i++)
879 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800880
881 msg->msg_name = NULL;
882 msg->msg_namelen = 0;
883 msg->msg_control = NULL;
884 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700885 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800886 msg->msg_iovlen = nr_segs;
887 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
888 if (sock->type == SOCK_SEQPACKET)
889 msg->msg_flags |= MSG_EOR;
890
891 return __sock_sendmsg(iocb, sock, msg, size);
892}
893
Badari Pulavarty027445c2006-09-30 23:28:46 -0700894static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
895 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800897 struct sock_iocb siocb, *x;
898
899 if (pos != 0)
900 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700901
Badari Pulavarty027445c2006-09-30 23:28:46 -0700902 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800903 if (!x)
904 return -ENOMEM;
905
Badari Pulavarty027445c2006-09-30 23:28:46 -0700906 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907}
908
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909/*
910 * Atomic setting of ioctl hooks to avoid race
911 * with module unload.
912 */
913
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800914static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700915static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916
Eric W. Biederman881d9662007-09-17 11:56:21 -0700917void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800919 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800921 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922}
923EXPORT_SYMBOL(brioctl_set);
924
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800925static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700926static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Eric W. Biederman881d9662007-09-17 11:56:21 -0700928void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800930 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800932 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700933}
934EXPORT_SYMBOL(vlan_ioctl_set);
935
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800936static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700937static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700939void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800941 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800943 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944}
945EXPORT_SYMBOL(dlci_ioctl_set);
946
Arnd Bergmann6b960182009-11-06 23:10:54 -0800947static long sock_do_ioctl(struct net *net, struct socket *sock,
948 unsigned int cmd, unsigned long arg)
949{
950 int err;
951 void __user *argp = (void __user *)arg;
952
953 err = sock->ops->ioctl(sock, cmd, arg);
954
955 /*
956 * If this ioctl is unknown try to hand it down
957 * to the NIC driver.
958 */
959 if (err == -ENOIOCTLCMD)
960 err = dev_ioctl(net, cmd, argp);
961
962 return err;
963}
964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965/*
966 * With an ioctl, arg may well be a user mode pointer, but we don't know
967 * what to do with it - that's up to the protocol still.
968 */
969
970static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
971{
972 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700973 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974 void __user *argp = (void __user *)arg;
975 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700976 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
Eric Dumazetb69aee02005-09-06 14:42:45 -0700978 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700979 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900980 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700982 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200984#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700986 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200988#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700989 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 case FIOSETOWN:
991 case SIOCSPGRP:
992 err = -EFAULT;
993 if (get_user(pid, (int __user *)argp))
994 break;
995 err = f_setown(sock->file, pid, 1);
996 break;
997 case FIOGETOWN:
998 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700999 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001000 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 break;
1002 case SIOCGIFBR:
1003 case SIOCSIFBR:
1004 case SIOCBRADDBR:
1005 case SIOCBRDELBR:
1006 err = -ENOPKG;
1007 if (!br_ioctl_hook)
1008 request_module("bridge");
1009
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001010 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001011 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001012 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001013 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 break;
1015 case SIOCGIFVLAN:
1016 case SIOCSIFVLAN:
1017 err = -ENOPKG;
1018 if (!vlan_ioctl_hook)
1019 request_module("8021q");
1020
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001021 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001023 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001024 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 case SIOCADDDLCI:
1027 case SIOCDELDLCI:
1028 err = -ENOPKG;
1029 if (!dlci_ioctl_hook)
1030 request_module("dlci");
1031
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001032 mutex_lock(&dlci_ioctl_mutex);
1033 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001035 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 break;
1037 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001038 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001040 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 return err;
1042}
1043
1044int sock_create_lite(int family, int type, int protocol, struct socket **res)
1045{
1046 int err;
1047 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001048
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 err = security_socket_create(family, type, protocol, 1);
1050 if (err)
1051 goto out;
1052
1053 sock = sock_alloc();
1054 if (!sock) {
1055 err = -ENOMEM;
1056 goto out;
1057 }
1058
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001060 err = security_socket_post_create(sock, family, type, protocol, 1);
1061 if (err)
1062 goto out_release;
1063
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064out:
1065 *res = sock;
1066 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001067out_release:
1068 sock_release(sock);
1069 sock = NULL;
1070 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001072EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073
1074/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001075static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076{
1077 struct socket *sock;
1078
1079 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001080 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001082 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 return sock->ops->poll(file, sock, wait);
1084}
1085
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001086static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001088 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
1090 return sock->ops->mmap(file, sock, vma);
1091}
1092
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001093static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094{
1095 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001096 * It was possible the inode is NULL we were
1097 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 */
1099
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001100 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 printk(KERN_DEBUG "sock_close: NULL inode\n");
1102 return 0;
1103 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 sock_release(SOCKET_I(inode));
1105 return 0;
1106}
1107
1108/*
1109 * Update the socket async list
1110 *
1111 * Fasync_list locking strategy.
1112 *
1113 * 1. fasync_list is modified only under process context socket lock
1114 * i.e. under semaphore.
1115 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001116 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 */
1118
1119static int sock_fasync(int fd, struct file *filp, int on)
1120{
Eric Dumazet989a2972010-04-14 09:55:35 +00001121 struct socket *sock = filp->private_data;
1122 struct sock *sk = sock->sk;
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001123 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
Eric Dumazet989a2972010-04-14 09:55:35 +00001125 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127
1128 lock_sock(sk);
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001129 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1130 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001132 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001133 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001134 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001135 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
Eric Dumazet989a2972010-04-14 09:55:35 +00001137 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 return 0;
1139}
1140
Eric Dumazet43815482010-04-29 11:01:49 +00001141/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142
1143int sock_wake_async(struct socket *sock, int how, int band)
1144{
Eric Dumazet43815482010-04-29 11:01:49 +00001145 struct socket_wq *wq;
1146
1147 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001149 rcu_read_lock();
1150 wq = rcu_dereference(sock->wq);
1151 if (!wq || !wq->fasync_list) {
1152 rcu_read_unlock();
1153 return -1;
1154 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001155 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001156 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1158 break;
1159 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001160 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1162 break;
1163 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001164 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001165call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001166 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001168 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001169 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 }
Eric Dumazet43815482010-04-29 11:01:49 +00001171 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 return 0;
1173}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001174EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
Pavel Emelyanov721db932010-09-29 16:06:32 +04001176int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001177 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178{
1179 int err;
1180 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001181 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
1183 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001184 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 */
1186 if (family < 0 || family >= NPROTO)
1187 return -EAFNOSUPPORT;
1188 if (type < 0 || type >= SOCK_MAX)
1189 return -EINVAL;
1190
1191 /* Compatibility.
1192
1193 This uglymoron is moved from INET layer to here to avoid
1194 deadlock in module load.
1195 */
1196 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001197 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 if (!warned) {
1199 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001200 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1201 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 }
1203 family = PF_PACKET;
1204 }
1205
1206 err = security_socket_create(family, type, protocol, kern);
1207 if (err)
1208 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001209
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001210 /*
1211 * Allocate the socket and allow the family to set things up. if
1212 * the protocol is 0, the family is instructed to select an appropriate
1213 * default.
1214 */
1215 sock = sock_alloc();
1216 if (!sock) {
1217 if (net_ratelimit())
1218 printk(KERN_WARNING "socket: no more sockets\n");
1219 return -ENFILE; /* Not exactly a match, but its the
1220 closest posix thing */
1221 }
1222
1223 sock->type = type;
1224
Johannes Berg95a5afc2008-10-16 15:24:51 -07001225#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001226 /* Attempt to load a protocol module if the find failed.
1227 *
1228 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 * requested real, full-featured networking support upon configuration.
1230 * Otherwise module support will break!
1231 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001232 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001233 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234#endif
1235
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001236 rcu_read_lock();
1237 pf = rcu_dereference(net_families[family]);
1238 err = -EAFNOSUPPORT;
1239 if (!pf)
1240 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
1242 /*
1243 * We will call the ->create function, that possibly is in a loadable
1244 * module, so we have to bump that loadable module refcnt first.
1245 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001246 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 goto out_release;
1248
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001249 /* Now protected by module ref count */
1250 rcu_read_unlock();
1251
Eric Paris3f378b682009-11-05 22:18:14 -08001252 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001253 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001255
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 /*
1257 * Now to bump the refcnt of the [loadable] module that owns this
1258 * socket at sock_release time we decrement its refcnt.
1259 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001260 if (!try_module_get(sock->ops->owner))
1261 goto out_module_busy;
1262
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 /*
1264 * Now that we're done with the ->create function, the [loadable]
1265 * module can have its refcnt decremented
1266 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001267 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001268 err = security_socket_post_create(sock, family, type, protocol, kern);
1269 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001270 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001271 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001273 return 0;
1274
1275out_module_busy:
1276 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001278 sock->ops = NULL;
1279 module_put(pf->owner);
1280out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001282 return err;
1283
1284out_release:
1285 rcu_read_unlock();
1286 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001288EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
1290int sock_create(int family, int type, int protocol, struct socket **res)
1291{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001292 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001294EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
1296int sock_create_kern(int family, int type, int protocol, struct socket **res)
1297{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001298 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001300EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001302SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303{
1304 int retval;
1305 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001306 int flags;
1307
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001308 /* Check the SOCK_* constants for consistency. */
1309 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1310 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1311 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1312 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1313
Ulrich Dreppera677a032008-07-23 21:29:17 -07001314 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001315 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001316 return -EINVAL;
1317 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001319 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1320 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1321
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 retval = sock_create(family, type, protocol, &sock);
1323 if (retval < 0)
1324 goto out;
1325
Ulrich Drepper77d27202008-07-23 21:29:35 -07001326 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 if (retval < 0)
1328 goto out_release;
1329
1330out:
1331 /* It may be already another descriptor 8) Not kernel problem. */
1332 return retval;
1333
1334out_release:
1335 sock_release(sock);
1336 return retval;
1337}
1338
1339/*
1340 * Create a pair of connected sockets.
1341 */
1342
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001343SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1344 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345{
1346 struct socket *sock1, *sock2;
1347 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001348 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001349 int flags;
1350
1351 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001352 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001353 return -EINVAL;
1354 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001356 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1357 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1358
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 /*
1360 * Obtain the first socket and check if the underlying protocol
1361 * supports the socketpair call.
1362 */
1363
1364 err = sock_create(family, type, protocol, &sock1);
1365 if (err < 0)
1366 goto out;
1367
1368 err = sock_create(family, type, protocol, &sock2);
1369 if (err < 0)
1370 goto out_release_1;
1371
1372 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001373 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 goto out_release_both;
1375
Al Viro7cbe66b2009-08-05 19:59:08 +04001376 fd1 = sock_alloc_file(sock1, &newfile1, flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001377 if (unlikely(fd1 < 0)) {
1378 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001380 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381
Al Viro7cbe66b2009-08-05 19:59:08 +04001382 fd2 = sock_alloc_file(sock2, &newfile2, flags);
Al Viro198de4d2009-08-05 19:29:23 +04001383 if (unlikely(fd2 < 0)) {
1384 err = fd2;
1385 fput(newfile1);
1386 put_unused_fd(fd1);
1387 sock_release(sock2);
1388 goto out;
Al Virodb349502007-02-07 01:48:00 -05001389 }
1390
Al Viro157cf642008-12-14 04:57:47 -05001391 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001392 fd_install(fd1, newfile1);
1393 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 /* fd1 and fd2 may be already another descriptors.
1395 * Not kernel problem.
1396 */
1397
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001398 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 if (!err)
1400 err = put_user(fd2, &usockvec[1]);
1401 if (!err)
1402 return 0;
1403
1404 sys_close(fd2);
1405 sys_close(fd1);
1406 return err;
1407
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001409 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001411 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412out:
1413 return err;
1414}
1415
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416/*
1417 * Bind a name to a socket. Nothing much to do here since it's
1418 * the protocol's responsibility to handle the local address.
1419 *
1420 * We move the socket address to kernel space before we call
1421 * the protocol layer (having also checked the address is ok).
1422 */
1423
Heiko Carstens20f37032009-01-14 14:14:23 +01001424SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425{
1426 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001427 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001428 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001430 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001431 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001432 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001433 if (err >= 0) {
1434 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001435 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001436 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001437 if (!err)
1438 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001439 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001440 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001442 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001443 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 return err;
1445}
1446
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447/*
1448 * Perform a listen. Basically, we allow the protocol to do anything
1449 * necessary for a listen, and if that works, we mark the socket as
1450 * ready for listening.
1451 */
1452
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001453SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454{
1455 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001456 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001457 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001458
1459 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1460 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001461 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001462 if ((unsigned)backlog > somaxconn)
1463 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
1465 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001466 if (!err)
1467 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001469 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 }
1471 return err;
1472}
1473
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474/*
1475 * For accept, we attempt to create a new socket, set up the link
1476 * with the client, wake up the client, then return the new
1477 * connected fd. We collect the address of the connector in kernel
1478 * space and move it to user at the very end. This is unclean because
1479 * we open the socket then return an error.
1480 *
1481 * 1003.1g adds the ability to recvmsg() to query connection pending
1482 * status to recvmsg. We need to add that support in a way thats
1483 * clean when we restucture accept also.
1484 */
1485
Heiko Carstens20f37032009-01-14 14:14:23 +01001486SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1487 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488{
1489 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001490 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001491 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001492 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493
Ulrich Drepper77d27202008-07-23 21:29:35 -07001494 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001495 return -EINVAL;
1496
1497 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1498 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1499
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001500 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 if (!sock)
1502 goto out;
1503
1504 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001505 newsock = sock_alloc();
1506 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 goto out_put;
1508
1509 newsock->type = sock->type;
1510 newsock->ops = sock->ops;
1511
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 /*
1513 * We don't need try_module_get here, as the listening socket (sock)
1514 * has the protocol module (sock->ops->owner) held.
1515 */
1516 __module_get(newsock->ops->owner);
1517
Al Viro7cbe66b2009-08-05 19:59:08 +04001518 newfd = sock_alloc_file(newsock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001519 if (unlikely(newfd < 0)) {
1520 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001521 sock_release(newsock);
1522 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001523 }
1524
Frank Filza79af592005-09-27 15:23:38 -07001525 err = security_socket_accept(sock, newsock);
1526 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001527 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001528
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1530 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001531 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
1533 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001534 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001535 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001537 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001539 err = move_addr_to_user((struct sockaddr *)&address,
1540 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001542 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 }
1544
1545 /* File flags are not inherited via accept() unlike another OSes. */
1546
David S. Miller39d8c1b2006-03-20 17:13:49 -08001547 fd_install(newfd, newfile);
1548 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001551 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552out:
1553 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001554out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001555 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001556 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 goto out_put;
1558}
1559
Heiko Carstens20f37032009-01-14 14:14:23 +01001560SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1561 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001562{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001563 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001564}
1565
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566/*
1567 * Attempt to connect to a socket with the server address. The address
1568 * is in user space so we verify it is OK and move it to kernel space.
1569 *
1570 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1571 * break bindings
1572 *
1573 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1574 * other SEQPACKET protocols that take time to connect() as it doesn't
1575 * include the -EINPROGRESS status for such sockets.
1576 */
1577
Heiko Carstens20f37032009-01-14 14:14:23 +01001578SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1579 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580{
1581 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001582 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001583 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001585 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 if (!sock)
1587 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001588 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 if (err < 0)
1590 goto out_put;
1591
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001592 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001593 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 if (err)
1595 goto out_put;
1596
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001597 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 sock->file->f_flags);
1599out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001600 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601out:
1602 return err;
1603}
1604
1605/*
1606 * Get the local address ('name') of a socket object. Move the obtained
1607 * name to user space.
1608 */
1609
Heiko Carstens20f37032009-01-14 14:14:23 +01001610SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1611 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612{
1613 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001614 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001615 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001616
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 if (!sock)
1619 goto out;
1620
1621 err = security_socket_getsockname(sock);
1622 if (err)
1623 goto out_put;
1624
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001625 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 if (err)
1627 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001628 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629
1630out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001631 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632out:
1633 return err;
1634}
1635
1636/*
1637 * Get the remote address ('name') of a socket object. Move the obtained
1638 * name to user space.
1639 */
1640
Heiko Carstens20f37032009-01-14 14:14:23 +01001641SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1642 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643{
1644 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001645 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001646 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001648 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1649 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 err = security_socket_getpeername(sock);
1651 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001652 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 return err;
1654 }
1655
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001656 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001657 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001658 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001660 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001661 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001662 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 }
1664 return err;
1665}
1666
1667/*
1668 * Send a datagram to a given address. We move the address into kernel
1669 * space and check the user space data area is readable before invoking
1670 * the protocol.
1671 */
1672
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001673SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1674 unsigned, flags, struct sockaddr __user *, addr,
1675 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676{
1677 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001678 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679 int err;
1680 struct msghdr msg;
1681 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001682 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001683
Linus Torvalds253eacc2010-10-30 16:43:10 -07001684 if (len > INT_MAX)
1685 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001686 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1687 if (!sock)
David S. Miller4387ff752007-02-08 15:06:08 -08001688 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001689
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001690 iov.iov_base = buff;
1691 iov.iov_len = len;
1692 msg.msg_name = NULL;
1693 msg.msg_iov = &iov;
1694 msg.msg_iovlen = 1;
1695 msg.msg_control = NULL;
1696 msg.msg_controllen = 0;
1697 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001698 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001699 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 if (err < 0)
1701 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001702 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001703 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 }
1705 if (sock->file->f_flags & O_NONBLOCK)
1706 flags |= MSG_DONTWAIT;
1707 msg.msg_flags = flags;
1708 err = sock_sendmsg(sock, &msg, len);
1709
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001710out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001711 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08001712out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 return err;
1714}
1715
1716/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001717 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 */
1719
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001720SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1721 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722{
1723 return sys_sendto(fd, buff, len, flags, NULL, 0);
1724}
1725
1726/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001727 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 * sender. We verify the buffers are writable and if needed move the
1729 * sender address from kernel to user space.
1730 */
1731
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001732SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1733 unsigned, flags, struct sockaddr __user *, addr,
1734 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735{
1736 struct socket *sock;
1737 struct iovec iov;
1738 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001739 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001740 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001741 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742
Linus Torvalds253eacc2010-10-30 16:43:10 -07001743 if (size > INT_MAX)
1744 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001745 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001747 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001749 msg.msg_control = NULL;
1750 msg.msg_controllen = 0;
1751 msg.msg_iovlen = 1;
1752 msg.msg_iov = &iov;
1753 iov.iov_len = size;
1754 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001755 msg.msg_name = (struct sockaddr *)&address;
1756 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 if (sock->file->f_flags & O_NONBLOCK)
1758 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001759 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001761 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001762 err2 = move_addr_to_user((struct sockaddr *)&address,
1763 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001764 if (err2 < 0)
1765 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001767
1768 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08001769out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 return err;
1771}
1772
1773/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001774 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 */
1776
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001777asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1778 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779{
1780 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1781}
1782
1783/*
1784 * Set a socket option. Because we don't know the option lengths we have
1785 * to pass the user mode parameter for the protocols to sort out.
1786 */
1787
Heiko Carstens20f37032009-01-14 14:14:23 +01001788SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1789 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001791 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 struct socket *sock;
1793
1794 if (optlen < 0)
1795 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001796
1797 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1798 if (sock != NULL) {
1799 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001800 if (err)
1801 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802
1803 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001804 err =
1805 sock_setsockopt(sock, level, optname, optval,
1806 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001808 err =
1809 sock->ops->setsockopt(sock, level, optname, optval,
1810 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001811out_put:
1812 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813 }
1814 return err;
1815}
1816
1817/*
1818 * Get a socket option. Because we don't know the option lengths we have
1819 * to pass a user mode parameter for the protocols to sort out.
1820 */
1821
Heiko Carstens20f37032009-01-14 14:14:23 +01001822SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1823 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001825 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 struct socket *sock;
1827
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1829 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001830 err = security_socket_getsockopt(sock, level, optname);
1831 if (err)
1832 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833
1834 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001835 err =
1836 sock_getsockopt(sock, level, optname, optval,
1837 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001839 err =
1840 sock->ops->getsockopt(sock, level, optname, optval,
1841 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001842out_put:
1843 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 }
1845 return err;
1846}
1847
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848/*
1849 * Shutdown a socket.
1850 */
1851
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001852SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001854 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 struct socket *sock;
1856
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001857 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1858 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001860 if (!err)
1861 err = sock->ops->shutdown(sock, how);
1862 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 }
1864 return err;
1865}
1866
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001867/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 * fields which are the same type (int / unsigned) on our platforms.
1869 */
1870#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1871#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1872#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1873
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001874struct used_address {
1875 struct sockaddr_storage name;
1876 unsigned int name_len;
1877};
1878
Anton Blanchard228e5482011-05-02 20:21:35 +00001879static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001880 struct msghdr *msg_sys, unsigned flags,
1881 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001883 struct compat_msghdr __user *msg_compat =
1884 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001885 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001887 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001888 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1889 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890 unsigned char *ctl_buf = ctl;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 int err, ctl_len, iov_size, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001892
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 err = -EFAULT;
1894 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001895 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001897 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 return -EFAULT;
1899
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 /* do not move before msg_sys is valid */
1901 err = -EMSGSIZE;
Anton Blanchard228e5482011-05-02 20:21:35 +00001902 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1903 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001905 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 err = -ENOMEM;
Anton Blanchard228e5482011-05-02 20:21:35 +00001907 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1908 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1910 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00001911 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 }
1913
1914 /* This will also move the address data into kernel space */
1915 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001916 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001917 (struct sockaddr *)&address,
1918 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 } else
Anton Blanchard228e5482011-05-02 20:21:35 +00001920 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001921 (struct sockaddr *)&address,
1922 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001923 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924 goto out_freeiov;
1925 total_len = err;
1926
1927 err = -ENOBUFS;
1928
Anton Blanchard228e5482011-05-02 20:21:35 +00001929 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001931 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001933 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001934 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001935 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 if (err)
1937 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001938 ctl_buf = msg_sys->msg_control;
1939 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001941 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001943 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 goto out_freeiov;
1945 }
1946 err = -EFAULT;
1947 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00001948 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1950 * checking falls down on this.
1951 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001952 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00001953 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001954 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00001956 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 }
Anton Blanchard228e5482011-05-02 20:21:35 +00001958 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959
1960 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00001961 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001962 /*
1963 * If this is sendmmsg() and current destination address is same as
1964 * previously succeeded address, omit asking LSM's decision.
1965 * used_address->name_len is initialized to UINT_MAX so that the first
1966 * destination address never matches.
1967 */
1968 if (used_address && used_address->name_len == msg_sys->msg_namelen &&
1969 !memcmp(&used_address->name, msg->msg_name,
1970 used_address->name_len)) {
1971 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
1972 goto out_freectl;
1973 }
1974 err = sock_sendmsg(sock, msg_sys, total_len);
1975 /*
1976 * If this is sendmmsg() and sending to current destination address was
1977 * successful, remember it.
1978 */
1979 if (used_address && err >= 0) {
1980 used_address->name_len = msg_sys->msg_namelen;
1981 memcpy(&used_address->name, msg->msg_name,
1982 used_address->name_len);
1983 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984
1985out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001986 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1988out_freeiov:
1989 if (iov != iovstack)
1990 sock_kfree_s(sock->sk, iov, iov_size);
Anton Blanchard228e5482011-05-02 20:21:35 +00001991out:
1992 return err;
1993}
1994
1995/*
1996 * BSD sendmsg interface
1997 */
1998
1999SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
2000{
2001 int fput_needed, err;
2002 struct msghdr msg_sys;
2003 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2004
2005 if (!sock)
2006 goto out;
2007
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002008 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002009
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002010 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002011out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012 return err;
2013}
2014
Anton Blanchard228e5482011-05-02 20:21:35 +00002015/*
2016 * Linux sendmmsg interface
2017 */
2018
2019int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2020 unsigned int flags)
2021{
2022 int fput_needed, err, datagrams;
2023 struct socket *sock;
2024 struct mmsghdr __user *entry;
2025 struct compat_mmsghdr __user *compat_entry;
2026 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002027 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002028
Anton Blanchard98382f42011-08-04 14:07:39 +00002029 if (vlen > UIO_MAXIOV)
2030 vlen = UIO_MAXIOV;
2031
Anton Blanchard228e5482011-05-02 20:21:35 +00002032 datagrams = 0;
2033
2034 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2035 if (!sock)
2036 return err;
2037
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002038 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002039 entry = mmsg;
2040 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002041 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002042
2043 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002044 if (MSG_CMSG_COMPAT & flags) {
2045 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002046 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002047 if (err < 0)
2048 break;
2049 err = __put_user(err, &compat_entry->msg_len);
2050 ++compat_entry;
2051 } else {
2052 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002053 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002054 if (err < 0)
2055 break;
2056 err = put_user(err, &entry->msg_len);
2057 ++entry;
2058 }
2059
2060 if (err)
2061 break;
2062 ++datagrams;
2063 }
2064
Anton Blanchard228e5482011-05-02 20:21:35 +00002065 fput_light(sock->file, fput_needed);
2066
Anton Blanchard728ffb82011-08-04 14:07:38 +00002067 /* We only return an error if no datagrams were able to be sent */
2068 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002069 return datagrams;
2070
Anton Blanchard228e5482011-05-02 20:21:35 +00002071 return err;
2072}
2073
2074SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2075 unsigned int, vlen, unsigned int, flags)
2076{
2077 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2078}
2079
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002080static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2081 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002083 struct compat_msghdr __user *msg_compat =
2084 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002086 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087 unsigned long cmsg_ptr;
2088 int err, iov_size, total_len, len;
2089
2090 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002091 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
2093 /* user mode address pointers */
2094 struct sockaddr __user *uaddr;
2095 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002096
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002098 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002100 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002101 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002104 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2105 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002106
2107 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002109 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2110 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2112 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002113 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 }
2115
2116 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002117 * Save the user-mode address (verify_iovec will change the
2118 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002120
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002121 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 uaddr_len = COMPAT_NAMELEN(msg);
2123 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002124 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002125 (struct sockaddr *)&addr,
2126 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 } else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002128 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002129 (struct sockaddr *)&addr,
2130 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131 if (err < 0)
2132 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002133 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002135 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2136 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002137
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138 if (sock->file->f_flags & O_NONBLOCK)
2139 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002140 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2141 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 if (err < 0)
2143 goto out_freeiov;
2144 len = err;
2145
2146 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002147 err = move_addr_to_user((struct sockaddr *)&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002148 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002149 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150 if (err < 0)
2151 goto out_freeiov;
2152 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002153 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002154 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 if (err)
2156 goto out_freeiov;
2157 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002158 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159 &msg_compat->msg_controllen);
2160 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002161 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162 &msg->msg_controllen);
2163 if (err)
2164 goto out_freeiov;
2165 err = len;
2166
2167out_freeiov:
2168 if (iov != iovstack)
2169 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002170out:
2171 return err;
2172}
2173
2174/*
2175 * BSD recvmsg interface
2176 */
2177
2178SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2179 unsigned int, flags)
2180{
2181 int fput_needed, err;
2182 struct msghdr msg_sys;
2183 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2184
2185 if (!sock)
2186 goto out;
2187
2188 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2189
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002190 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191out:
2192 return err;
2193}
2194
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002195/*
2196 * Linux recvmmsg interface
2197 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002199int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2200 unsigned int flags, struct timespec *timeout)
2201{
2202 int fput_needed, err, datagrams;
2203 struct socket *sock;
2204 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002205 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002206 struct msghdr msg_sys;
2207 struct timespec end_time;
2208
2209 if (timeout &&
2210 poll_select_set_timeout(&end_time, timeout->tv_sec,
2211 timeout->tv_nsec))
2212 return -EINVAL;
2213
2214 datagrams = 0;
2215
2216 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2217 if (!sock)
2218 return err;
2219
2220 err = sock_error(sock->sk);
2221 if (err)
2222 goto out_put;
2223
2224 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002225 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002226
2227 while (datagrams < vlen) {
2228 /*
2229 * No need to ask LSM for more than the first datagram.
2230 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002231 if (MSG_CMSG_COMPAT & flags) {
2232 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002233 &msg_sys, flags & ~MSG_WAITFORONE,
2234 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002235 if (err < 0)
2236 break;
2237 err = __put_user(err, &compat_entry->msg_len);
2238 ++compat_entry;
2239 } else {
2240 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002241 &msg_sys, flags & ~MSG_WAITFORONE,
2242 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002243 if (err < 0)
2244 break;
2245 err = put_user(err, &entry->msg_len);
2246 ++entry;
2247 }
2248
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002249 if (err)
2250 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002251 ++datagrams;
2252
Brandon L Black71c5c1592010-03-26 16:18:03 +00002253 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2254 if (flags & MSG_WAITFORONE)
2255 flags |= MSG_DONTWAIT;
2256
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002257 if (timeout) {
2258 ktime_get_ts(timeout);
2259 *timeout = timespec_sub(end_time, *timeout);
2260 if (timeout->tv_sec < 0) {
2261 timeout->tv_sec = timeout->tv_nsec = 0;
2262 break;
2263 }
2264
2265 /* Timeout, return less than vlen datagrams */
2266 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2267 break;
2268 }
2269
2270 /* Out of band data, return right away */
2271 if (msg_sys.msg_flags & MSG_OOB)
2272 break;
2273 }
2274
2275out_put:
2276 fput_light(sock->file, fput_needed);
2277
2278 if (err == 0)
2279 return datagrams;
2280
2281 if (datagrams != 0) {
2282 /*
2283 * We may return less entries than requested (vlen) if the
2284 * sock is non block and there aren't enough datagrams...
2285 */
2286 if (err != -EAGAIN) {
2287 /*
2288 * ... or if recvmsg returns an error after we
2289 * received some datagrams, where we record the
2290 * error to return on the next call or if the
2291 * app asks about it using getsockopt(SO_ERROR).
2292 */
2293 sock->sk->sk_err = -err;
2294 }
2295
2296 return datagrams;
2297 }
2298
2299 return err;
2300}
2301
2302SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2303 unsigned int, vlen, unsigned int, flags,
2304 struct timespec __user *, timeout)
2305{
2306 int datagrams;
2307 struct timespec timeout_sys;
2308
2309 if (!timeout)
2310 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2311
2312 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2313 return -EFAULT;
2314
2315 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2316
2317 if (datagrams > 0 &&
2318 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2319 datagrams = -EFAULT;
2320
2321 return datagrams;
2322}
2323
2324#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325/* Argument list sizes for sys_socketcall */
2326#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002327static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002328 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2329 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2330 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002331 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002332};
2333
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334#undef AL
2335
2336/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002337 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 *
2339 * Argument checking cleaned up. Saved 20% in size.
2340 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002341 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342 */
2343
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002344SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345{
2346 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002347 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002349 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350
Anton Blanchard228e5482011-05-02 20:21:35 +00002351 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 return -EINVAL;
2353
Arjan van de Ven47379052009-09-28 12:57:44 -07002354 len = nargs[call];
2355 if (len > sizeof(a))
2356 return -EINVAL;
2357
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002359 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002361
Al Virof3298dc2008-12-10 03:16:51 -05002362 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002363
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002364 a0 = a[0];
2365 a1 = a[1];
2366
2367 switch (call) {
2368 case SYS_SOCKET:
2369 err = sys_socket(a0, a1, a[2]);
2370 break;
2371 case SYS_BIND:
2372 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2373 break;
2374 case SYS_CONNECT:
2375 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2376 break;
2377 case SYS_LISTEN:
2378 err = sys_listen(a0, a1);
2379 break;
2380 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002381 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2382 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002383 break;
2384 case SYS_GETSOCKNAME:
2385 err =
2386 sys_getsockname(a0, (struct sockaddr __user *)a1,
2387 (int __user *)a[2]);
2388 break;
2389 case SYS_GETPEERNAME:
2390 err =
2391 sys_getpeername(a0, (struct sockaddr __user *)a1,
2392 (int __user *)a[2]);
2393 break;
2394 case SYS_SOCKETPAIR:
2395 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2396 break;
2397 case SYS_SEND:
2398 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2399 break;
2400 case SYS_SENDTO:
2401 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2402 (struct sockaddr __user *)a[4], a[5]);
2403 break;
2404 case SYS_RECV:
2405 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2406 break;
2407 case SYS_RECVFROM:
2408 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2409 (struct sockaddr __user *)a[4],
2410 (int __user *)a[5]);
2411 break;
2412 case SYS_SHUTDOWN:
2413 err = sys_shutdown(a0, a1);
2414 break;
2415 case SYS_SETSOCKOPT:
2416 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2417 break;
2418 case SYS_GETSOCKOPT:
2419 err =
2420 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2421 (int __user *)a[4]);
2422 break;
2423 case SYS_SENDMSG:
2424 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2425 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002426 case SYS_SENDMMSG:
2427 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2428 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002429 case SYS_RECVMSG:
2430 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2431 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002432 case SYS_RECVMMSG:
2433 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2434 (struct timespec __user *)a[4]);
2435 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002436 case SYS_ACCEPT4:
2437 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2438 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002439 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002440 default:
2441 err = -EINVAL;
2442 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443 }
2444 return err;
2445}
2446
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002447#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002449/**
2450 * sock_register - add a socket protocol handler
2451 * @ops: description of protocol
2452 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 * This function is called by a protocol handler that wants to
2454 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002455 * socket interface. The value ops->family coresponds to the
2456 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002458int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459{
2460 int err;
2461
2462 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002463 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2464 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465 return -ENOBUFS;
2466 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002467
2468 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002469 if (rcu_dereference_protected(net_families[ops->family],
2470 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002471 err = -EEXIST;
2472 else {
Eric Dumazet190683a2010-11-10 10:50:44 +00002473 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474 err = 0;
2475 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002476 spin_unlock(&net_family_lock);
2477
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002478 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479 return err;
2480}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002481EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002483/**
2484 * sock_unregister - remove a protocol handler
2485 * @family: protocol family to remove
2486 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487 * This function is called by a protocol handler that wants to
2488 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002489 * new socket creation.
2490 *
2491 * If protocol handler is a module, then it can use module reference
2492 * counts to protect against new references. If protocol handler is not
2493 * a module then it needs to provide its own protection in
2494 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002496void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002498 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002499
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002500 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002501 rcu_assign_pointer(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002502 spin_unlock(&net_family_lock);
2503
2504 synchronize_rcu();
2505
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002506 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002507}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002508EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509
Andi Kleen77d76ea2005-12-22 12:43:42 -08002510static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002512 int err;
2513
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002515 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002517
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518 sk_init();
2519
Linus Torvalds1da177e2005-04-16 15:20:36 -07002520 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002521 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522 */
2523 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524
2525 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002526 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002527 */
2528
2529 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002530
2531 err = register_filesystem(&sock_fs_type);
2532 if (err)
2533 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002535 if (IS_ERR(sock_mnt)) {
2536 err = PTR_ERR(sock_mnt);
2537 goto out_mount;
2538 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002539
2540 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541 */
2542
2543#ifdef CONFIG_NETFILTER
2544 netfilter_init();
2545#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002546
Richard Cochranc1f19b52010-07-17 08:49:36 +00002547#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2548 skb_timestamping_init();
2549#endif
2550
Nick Pigginb3e19d92011-01-07 17:50:11 +11002551out:
2552 return err;
2553
2554out_mount:
2555 unregister_filesystem(&sock_fs_type);
2556out_fs:
2557 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558}
2559
Andi Kleen77d76ea2005-12-22 12:43:42 -08002560core_initcall(sock_init); /* early initcall */
2561
Linus Torvalds1da177e2005-04-16 15:20:36 -07002562#ifdef CONFIG_PROC_FS
2563void socket_seq_show(struct seq_file *seq)
2564{
2565 int cpu;
2566 int counter = 0;
2567
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002568 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002569 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570
2571 /* It can be negative, by the way. 8) */
2572 if (counter < 0)
2573 counter = 0;
2574
2575 seq_printf(seq, "sockets: used %d\n", counter);
2576}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002577#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002579#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002580static int do_siocgstamp(struct net *net, struct socket *sock,
2581 unsigned int cmd, struct compat_timeval __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002582{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002583 mm_segment_t old_fs = get_fs();
2584 struct timeval ktv;
2585 int err;
2586
2587 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002588 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002589 set_fs(old_fs);
2590 if (!err) {
2591 err = put_user(ktv.tv_sec, &up->tv_sec);
2592 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2593 }
2594 return err;
2595}
2596
Arnd Bergmann6b960182009-11-06 23:10:54 -08002597static int do_siocgstampns(struct net *net, struct socket *sock,
2598 unsigned int cmd, struct compat_timespec __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002599{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002600 mm_segment_t old_fs = get_fs();
2601 struct timespec kts;
2602 int err;
2603
2604 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002605 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002606 set_fs(old_fs);
2607 if (!err) {
2608 err = put_user(kts.tv_sec, &up->tv_sec);
2609 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2610 }
2611 return err;
2612}
2613
Arnd Bergmann6b960182009-11-06 23:10:54 -08002614static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002615{
2616 struct ifreq __user *uifr;
2617 int err;
2618
2619 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002620 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002621 return -EFAULT;
2622
Arnd Bergmann6b960182009-11-06 23:10:54 -08002623 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002624 if (err)
2625 return err;
2626
Arnd Bergmann6b960182009-11-06 23:10:54 -08002627 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002628 return -EFAULT;
2629
2630 return 0;
2631}
2632
Arnd Bergmann6b960182009-11-06 23:10:54 -08002633static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002634{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002635 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002636 struct ifconf ifc;
2637 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002638 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002639 struct ifreq __user *ifr;
2640 unsigned int i, j;
2641 int err;
2642
Arnd Bergmann6b960182009-11-06 23:10:54 -08002643 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002644 return -EFAULT;
2645
2646 if (ifc32.ifcbuf == 0) {
2647 ifc32.ifc_len = 0;
2648 ifc.ifc_len = 0;
2649 ifc.ifc_req = NULL;
2650 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2651 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002652 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2653 sizeof(struct ifreq);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002654 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2655 ifc.ifc_len = len;
2656 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2657 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002658 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002659 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002660 return -EFAULT;
2661 ifr++;
2662 ifr32++;
2663 }
2664 }
2665 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2666 return -EFAULT;
2667
Arnd Bergmann6b960182009-11-06 23:10:54 -08002668 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002669 if (err)
2670 return err;
2671
2672 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2673 return -EFAULT;
2674
2675 ifr = ifc.ifc_req;
2676 ifr32 = compat_ptr(ifc32.ifcbuf);
2677 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002678 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2679 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2680 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002681 return -EFAULT;
2682 ifr32++;
2683 ifr++;
2684 }
2685
2686 if (ifc32.ifcbuf == 0) {
2687 /* Translate from 64-bit structure multiple to
2688 * a 32-bit one.
2689 */
2690 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002691 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002692 ifc32.ifc_len = i;
2693 } else {
2694 ifc32.ifc_len = i;
2695 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002696 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002697 return -EFAULT;
2698
2699 return 0;
2700}
2701
Arnd Bergmann6b960182009-11-06 23:10:54 -08002702static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002703{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002704 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2705 bool convert_in = false, convert_out = false;
2706 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2707 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002708 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002709 u32 rule_cnt = 0, actual_rule_cnt;
2710 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002711 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002712 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002713
2714 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2715 return -EFAULT;
2716
Ben Hutchings3a7da392011-03-17 07:34:32 +00002717 compat_rxnfc = compat_ptr(data);
2718
2719 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002720 return -EFAULT;
2721
Ben Hutchings3a7da392011-03-17 07:34:32 +00002722 /* Most ethtool structures are defined without padding.
2723 * Unfortunately struct ethtool_rxnfc is an exception.
2724 */
2725 switch (ethcmd) {
2726 default:
2727 break;
2728 case ETHTOOL_GRXCLSRLALL:
2729 /* Buffer size is variable */
2730 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2731 return -EFAULT;
2732 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2733 return -ENOMEM;
2734 buf_size += rule_cnt * sizeof(u32);
2735 /* fall through */
2736 case ETHTOOL_GRXRINGS:
2737 case ETHTOOL_GRXCLSRLCNT:
2738 case ETHTOOL_GRXCLSRULE:
2739 convert_out = true;
2740 /* fall through */
2741 case ETHTOOL_SRXCLSRLDEL:
2742 case ETHTOOL_SRXCLSRLINS:
2743 buf_size += sizeof(struct ethtool_rxnfc);
2744 convert_in = true;
2745 break;
2746 }
2747
2748 ifr = compat_alloc_user_space(buf_size);
2749 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2750
2751 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2752 return -EFAULT;
2753
2754 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2755 &ifr->ifr_ifru.ifru_data))
2756 return -EFAULT;
2757
2758 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002759 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002760 * fs.ring_cookie and at the end of fs, but nowhere else.
2761 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002762 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2763 sizeof(compat_rxnfc->fs.m_ext) !=
2764 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2765 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002766 BUILD_BUG_ON(
2767 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2768 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2769 offsetof(struct ethtool_rxnfc, fs.location) -
2770 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2771
2772 if (copy_in_user(rxnfc, compat_rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002773 (void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002774 (void *)rxnfc) ||
2775 copy_in_user(&rxnfc->fs.ring_cookie,
2776 &compat_rxnfc->fs.ring_cookie,
2777 (void *)(&rxnfc->fs.location + 1) -
2778 (void *)&rxnfc->fs.ring_cookie) ||
2779 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2780 sizeof(rxnfc->rule_cnt)))
2781 return -EFAULT;
2782 }
2783
2784 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2785 if (ret)
2786 return ret;
2787
2788 if (convert_out) {
2789 if (copy_in_user(compat_rxnfc, rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002790 (const void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002791 (const void *)rxnfc) ||
2792 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2793 &rxnfc->fs.ring_cookie,
2794 (const void *)(&rxnfc->fs.location + 1) -
2795 (const void *)&rxnfc->fs.ring_cookie) ||
2796 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2797 sizeof(rxnfc->rule_cnt)))
2798 return -EFAULT;
2799
2800 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2801 /* As an optimisation, we only copy the actual
2802 * number of rules that the underlying
2803 * function returned. Since Mallory might
2804 * change the rule count in user memory, we
2805 * check that it is less than the rule count
2806 * originally given (as the user buffer size),
2807 * which has been range-checked.
2808 */
2809 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2810 return -EFAULT;
2811 if (actual_rule_cnt < rule_cnt)
2812 rule_cnt = actual_rule_cnt;
2813 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2814 &rxnfc->rule_locs[0],
2815 rule_cnt * sizeof(u32)))
2816 return -EFAULT;
2817 }
2818 }
2819
2820 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002821}
2822
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002823static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2824{
2825 void __user *uptr;
2826 compat_uptr_t uptr32;
2827 struct ifreq __user *uifr;
2828
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002829 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002830 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2831 return -EFAULT;
2832
2833 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2834 return -EFAULT;
2835
2836 uptr = compat_ptr(uptr32);
2837
2838 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2839 return -EFAULT;
2840
2841 return dev_ioctl(net, SIOCWANDEV, uifr);
2842}
2843
Arnd Bergmann6b960182009-11-06 23:10:54 -08002844static int bond_ioctl(struct net *net, unsigned int cmd,
2845 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002846{
2847 struct ifreq kifr;
2848 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002849 mm_segment_t old_fs;
2850 int err;
2851 u32 data;
2852 void __user *datap;
2853
2854 switch (cmd) {
2855 case SIOCBONDENSLAVE:
2856 case SIOCBONDRELEASE:
2857 case SIOCBONDSETHWADDR:
2858 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002859 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002860 return -EFAULT;
2861
2862 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002863 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002864 err = dev_ioctl(net, cmd,
2865 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002866 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002867
2868 return err;
2869 case SIOCBONDSLAVEINFOQUERY:
2870 case SIOCBONDINFOQUERY:
2871 uifr = compat_alloc_user_space(sizeof(*uifr));
2872 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2873 return -EFAULT;
2874
2875 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2876 return -EFAULT;
2877
2878 datap = compat_ptr(data);
2879 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2880 return -EFAULT;
2881
Arnd Bergmann6b960182009-11-06 23:10:54 -08002882 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002883 default:
2884 return -EINVAL;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002885 }
Arnd Bergmann7a229382009-11-06 23:00:29 -08002886}
2887
Arnd Bergmann6b960182009-11-06 23:10:54 -08002888static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2889 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002890{
2891 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002892 char tmp_buf[IFNAMSIZ];
2893 void __user *data64;
2894 u32 data32;
2895
2896 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2897 IFNAMSIZ))
2898 return -EFAULT;
2899 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2900 return -EFAULT;
2901 data64 = compat_ptr(data32);
2902
2903 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2904
2905 /* Don't check these user accesses, just let that get trapped
2906 * in the ioctl handler instead.
2907 */
2908 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2909 IFNAMSIZ))
2910 return -EFAULT;
2911 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2912 return -EFAULT;
2913
Arnd Bergmann6b960182009-11-06 23:10:54 -08002914 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002915}
2916
Arnd Bergmann6b960182009-11-06 23:10:54 -08002917static int dev_ifsioc(struct net *net, struct socket *sock,
2918 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002919{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002920 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002921 int err;
2922
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002923 uifr = compat_alloc_user_space(sizeof(*uifr));
2924 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2925 return -EFAULT;
2926
2927 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2928
Arnd Bergmann7a229382009-11-06 23:00:29 -08002929 if (!err) {
2930 switch (cmd) {
2931 case SIOCGIFFLAGS:
2932 case SIOCGIFMETRIC:
2933 case SIOCGIFMTU:
2934 case SIOCGIFMEM:
2935 case SIOCGIFHWADDR:
2936 case SIOCGIFINDEX:
2937 case SIOCGIFADDR:
2938 case SIOCGIFBRDADDR:
2939 case SIOCGIFDSTADDR:
2940 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002941 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002942 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002943 case SIOCGMIIPHY:
2944 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002945 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002946 err = -EFAULT;
2947 break;
2948 }
2949 }
2950 return err;
2951}
2952
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002953static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2954 struct compat_ifreq __user *uifr32)
2955{
2956 struct ifreq ifr;
2957 struct compat_ifmap __user *uifmap32;
2958 mm_segment_t old_fs;
2959 int err;
2960
2961 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2962 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2963 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2964 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2965 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2966 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2967 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2968 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2969 if (err)
2970 return -EFAULT;
2971
2972 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002973 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002974 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002975 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002976
2977 if (cmd == SIOCGIFMAP && !err) {
2978 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2979 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2980 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2981 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2982 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2983 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2984 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2985 if (err)
2986 err = -EFAULT;
2987 }
2988 return err;
2989}
2990
2991static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2992{
2993 void __user *uptr;
2994 compat_uptr_t uptr32;
2995 struct ifreq __user *uifr;
2996
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002997 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002998 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2999 return -EFAULT;
3000
3001 if (get_user(uptr32, &uifr32->ifr_data))
3002 return -EFAULT;
3003
3004 uptr = compat_ptr(uptr32);
3005
3006 if (put_user(uptr, &uifr->ifr_data))
3007 return -EFAULT;
3008
3009 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3010}
3011
Arnd Bergmann7a229382009-11-06 23:00:29 -08003012struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003013 u32 rt_pad1;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003014 struct sockaddr rt_dst; /* target address */
3015 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3016 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003017 unsigned short rt_flags;
3018 short rt_pad2;
3019 u32 rt_pad3;
3020 unsigned char rt_tos;
3021 unsigned char rt_class;
3022 short rt_pad4;
3023 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003024 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003025 u32 rt_mtu; /* per route MTU/Window */
3026 u32 rt_window; /* Window clamping */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003027 unsigned short rt_irtt; /* Initial RTT */
3028};
3029
3030struct in6_rtmsg32 {
3031 struct in6_addr rtmsg_dst;
3032 struct in6_addr rtmsg_src;
3033 struct in6_addr rtmsg_gateway;
3034 u32 rtmsg_type;
3035 u16 rtmsg_dst_len;
3036 u16 rtmsg_src_len;
3037 u32 rtmsg_metric;
3038 u32 rtmsg_info;
3039 u32 rtmsg_flags;
3040 s32 rtmsg_ifindex;
3041};
3042
Arnd Bergmann6b960182009-11-06 23:10:54 -08003043static int routing_ioctl(struct net *net, struct socket *sock,
3044 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003045{
3046 int ret;
3047 void *r = NULL;
3048 struct in6_rtmsg r6;
3049 struct rtentry r4;
3050 char devname[16];
3051 u32 rtdev;
3052 mm_segment_t old_fs = get_fs();
3053
Arnd Bergmann6b960182009-11-06 23:10:54 -08003054 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3055 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003056 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003057 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003058 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3059 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3060 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3061 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3062 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3063 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3064 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003065
3066 r = (void *) &r6;
3067 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003068 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003069 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003070 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003071 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3072 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3073 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3074 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3075 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3076 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003077 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003078 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003079 r4.rt_dev = (char __user __force *)devname;
3080 devname[15] = 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003081 } else
3082 r4.rt_dev = NULL;
3083
3084 r = (void *) &r4;
3085 }
3086
3087 if (ret) {
3088 ret = -EFAULT;
3089 goto out;
3090 }
3091
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003092 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003093 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003094 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003095
3096out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003097 return ret;
3098}
3099
3100/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3101 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003102 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003103 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003104static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003105{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003106 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003107
Arnd Bergmann6b960182009-11-06 23:10:54 -08003108 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003109 return -EFAULT;
3110 if (tmp == BRCTL_GET_VERSION)
3111 return BRCTL_VERSION + 1;
3112 return -EINVAL;
3113}
3114
Arnd Bergmann6b960182009-11-06 23:10:54 -08003115static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3116 unsigned int cmd, unsigned long arg)
3117{
3118 void __user *argp = compat_ptr(arg);
3119 struct sock *sk = sock->sk;
3120 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003121
Arnd Bergmann6b960182009-11-06 23:10:54 -08003122 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3123 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003124
Arnd Bergmann6b960182009-11-06 23:10:54 -08003125 switch (cmd) {
3126 case SIOCSIFBR:
3127 case SIOCGIFBR:
3128 return old_bridge_ioctl(argp);
3129 case SIOCGIFNAME:
3130 return dev_ifname32(net, argp);
3131 case SIOCGIFCONF:
3132 return dev_ifconf(net, argp);
3133 case SIOCETHTOOL:
3134 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003135 case SIOCWANDEV:
3136 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003137 case SIOCGIFMAP:
3138 case SIOCSIFMAP:
3139 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003140 case SIOCBONDENSLAVE:
3141 case SIOCBONDRELEASE:
3142 case SIOCBONDSETHWADDR:
3143 case SIOCBONDSLAVEINFOQUERY:
3144 case SIOCBONDINFOQUERY:
3145 case SIOCBONDCHANGEACTIVE:
3146 return bond_ioctl(net, cmd, argp);
3147 case SIOCADDRT:
3148 case SIOCDELRT:
3149 return routing_ioctl(net, sock, cmd, argp);
3150 case SIOCGSTAMP:
3151 return do_siocgstamp(net, sock, cmd, argp);
3152 case SIOCGSTAMPNS:
3153 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003154 case SIOCSHWTSTAMP:
3155 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003156
Arnd Bergmann6b960182009-11-06 23:10:54 -08003157 case FIOSETOWN:
3158 case SIOCSPGRP:
3159 case FIOGETOWN:
3160 case SIOCGPGRP:
3161 case SIOCBRADDBR:
3162 case SIOCBRDELBR:
3163 case SIOCGIFVLAN:
3164 case SIOCSIFVLAN:
3165 case SIOCADDDLCI:
3166 case SIOCDELDLCI:
3167 return sock_ioctl(file, cmd, arg);
3168
3169 case SIOCGIFFLAGS:
3170 case SIOCSIFFLAGS:
3171 case SIOCGIFMETRIC:
3172 case SIOCSIFMETRIC:
3173 case SIOCGIFMTU:
3174 case SIOCSIFMTU:
3175 case SIOCGIFMEM:
3176 case SIOCSIFMEM:
3177 case SIOCGIFHWADDR:
3178 case SIOCSIFHWADDR:
3179 case SIOCADDMULTI:
3180 case SIOCDELMULTI:
3181 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003182 case SIOCGIFADDR:
3183 case SIOCSIFADDR:
3184 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003185 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003186 case SIOCGIFBRDADDR:
3187 case SIOCSIFBRDADDR:
3188 case SIOCGIFDSTADDR:
3189 case SIOCSIFDSTADDR:
3190 case SIOCGIFNETMASK:
3191 case SIOCSIFNETMASK:
3192 case SIOCSIFPFLAGS:
3193 case SIOCGIFPFLAGS:
3194 case SIOCGIFTXQLEN:
3195 case SIOCSIFTXQLEN:
3196 case SIOCBRADDIF:
3197 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003198 case SIOCSIFNAME:
3199 case SIOCGMIIPHY:
3200 case SIOCGMIIREG:
3201 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003202 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003203
Arnd Bergmann6b960182009-11-06 23:10:54 -08003204 case SIOCSARP:
3205 case SIOCGARP:
3206 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003207 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003208 return sock_do_ioctl(net, sock, cmd, arg);
3209 }
3210
3211 /* Prevent warning from compat_sys_ioctl, these always
3212 * result in -EINVAL in the native case anyway. */
3213 switch (cmd) {
3214 case SIOCRTMSG:
3215 case SIOCGIFCOUNT:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003216 case SIOCSRARP:
3217 case SIOCGRARP:
3218 case SIOCDRARP:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003219 case SIOCSIFLINK:
3220 case SIOCGIFSLAVE:
3221 case SIOCSIFSLAVE:
3222 return -EINVAL;
Arnd Bergmann6b960182009-11-06 23:10:54 -08003223 }
3224
3225 return -ENOIOCTLCMD;
3226}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003227
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003228static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003229 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003230{
3231 struct socket *sock = file->private_data;
3232 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003233 struct sock *sk;
3234 struct net *net;
3235
3236 sk = sock->sk;
3237 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003238
3239 if (sock->ops->compat_ioctl)
3240 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3241
David S. Miller87de87d2008-06-03 09:14:03 -07003242 if (ret == -ENOIOCTLCMD &&
3243 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3244 ret = compat_wext_handle_ioctl(net, cmd, arg);
3245
Arnd Bergmann6b960182009-11-06 23:10:54 -08003246 if (ret == -ENOIOCTLCMD)
3247 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3248
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003249 return ret;
3250}
3251#endif
3252
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003253int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3254{
3255 return sock->ops->bind(sock, addr, addrlen);
3256}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003257EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003258
3259int kernel_listen(struct socket *sock, int backlog)
3260{
3261 return sock->ops->listen(sock, backlog);
3262}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003263EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003264
3265int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3266{
3267 struct sock *sk = sock->sk;
3268 int err;
3269
3270 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3271 newsock);
3272 if (err < 0)
3273 goto done;
3274
3275 err = sock->ops->accept(sock, *newsock, flags);
3276 if (err < 0) {
3277 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003278 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003279 goto done;
3280 }
3281
3282 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003283 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003284
3285done:
3286 return err;
3287}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003288EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003289
3290int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003291 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003292{
3293 return sock->ops->connect(sock, addr, addrlen, flags);
3294}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003295EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003296
3297int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3298 int *addrlen)
3299{
3300 return sock->ops->getname(sock, addr, addrlen, 0);
3301}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003302EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003303
3304int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3305 int *addrlen)
3306{
3307 return sock->ops->getname(sock, addr, addrlen, 1);
3308}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003309EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003310
3311int kernel_getsockopt(struct socket *sock, int level, int optname,
3312 char *optval, int *optlen)
3313{
3314 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003315 char __user *uoptval;
3316 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003317 int err;
3318
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003319 uoptval = (char __user __force *) optval;
3320 uoptlen = (int __user __force *) optlen;
3321
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003322 set_fs(KERNEL_DS);
3323 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003324 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003325 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003326 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3327 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003328 set_fs(oldfs);
3329 return err;
3330}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003331EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003332
3333int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003334 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003335{
3336 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003337 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003338 int err;
3339
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003340 uoptval = (char __user __force *) optval;
3341
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003342 set_fs(KERNEL_DS);
3343 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003344 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003345 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003346 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003347 optlen);
3348 set_fs(oldfs);
3349 return err;
3350}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003351EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003352
3353int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3354 size_t size, int flags)
3355{
Herbert Xuf8451722010-05-24 00:12:34 -07003356 sock_update_classid(sock->sk);
3357
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003358 if (sock->ops->sendpage)
3359 return sock->ops->sendpage(sock, page, offset, size, flags);
3360
3361 return sock_no_sendpage(sock, page, offset, size, flags);
3362}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003363EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003364
3365int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3366{
3367 mm_segment_t oldfs = get_fs();
3368 int err;
3369
3370 set_fs(KERNEL_DS);
3371 err = sock->ops->ioctl(sock, cmd, arg);
3372 set_fs(oldfs);
3373
3374 return err;
3375}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003376EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003377
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003378int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3379{
3380 return sock->ops->shutdown(sock, how);
3381}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003382EXPORT_SYMBOL(kernel_sock_shutdown);