blob: 791d71a36a93dfec5166fe05e2e0cb394cfa904b [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
90#include <asm/uaccess.h>
91#include <asm/unistd.h>
92
93#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070094#include <net/wext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96#include <net/sock.h>
97#include <linux/netfilter.h>
98
99static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700100static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
101 unsigned long nr_segs, loff_t pos);
102static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
103 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700104static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
106static int sock_close(struct inode *inode, struct file *file);
107static unsigned int sock_poll(struct file *file,
108 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700109static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800110#ifdef CONFIG_COMPAT
111static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700112 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800113#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115static ssize_t sock_sendpage(struct file *file, struct page *page,
116 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800117static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
118 struct pipe_inode_info *pipe, size_t len,
119 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121/*
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
124 */
125
Arjan van de Venda7071d2007-02-12 00:55:36 -0800126static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 .owner = THIS_MODULE,
128 .llseek = no_llseek,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
131 .poll = sock_poll,
132 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800133#ifdef CONFIG_COMPAT
134 .compat_ioctl = compat_sock_ioctl,
135#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 .mmap = sock_mmap,
137 .open = sock_no_open, /* special open code to disallow open via /proc */
138 .release = sock_close,
139 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200140 .sendpage = sock_sendpage,
141 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800142 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143};
144
145/*
146 * The protocol list. Each protocol is registered in here.
147 */
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700150static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152/*
153 * Statistics counters of the socket lists
154 */
155
156static DEFINE_PER_CPU(int, sockets_in_use) = 0;
157
158/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700159 * Support routines.
160 * Move socket addresses back and forth across the kernel/user
161 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 */
163
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700164#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 16 for IP, 16 for IPX,
166 24 for IPv6,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700167 about 80 for AX.25
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 must be at least one bigger than
169 the AF_UNIX size (see net/unix/af_unix.c
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700170 :unix_mkname()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700172
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
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700212int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700213 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
238#define SOCKFS_MAGIC 0x534F434B
239
Christoph Lametere18b8902006-12-06 20:33:20 -0800240static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241
242static struct inode *sock_alloc_inode(struct super_block *sb)
243{
244 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700245
Christoph Lametere94b1762006-12-06 20:33:17 -0800246 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 if (!ei)
248 return NULL;
249 init_waitqueue_head(&ei->socket.wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700250
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 ei->socket.fasync_list = NULL;
252 ei->socket.state = SS_UNCONNECTED;
253 ei->socket.flags = 0;
254 ei->socket.ops = NULL;
255 ei->socket.sk = NULL;
256 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257
258 return &ei->vfs_inode;
259}
260
261static void sock_destroy_inode(struct inode *inode)
262{
263 kmem_cache_free(sock_inode_cachep,
264 container_of(inode, struct socket_alloc, vfs_inode));
265}
266
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700267static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700269 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Christoph Lametera35afb82007-05-16 22:10:57 -0700271 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274static int init_inodecache(void)
275{
276 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700277 sizeof(struct socket_alloc),
278 0,
279 (SLAB_HWCACHE_ALIGN |
280 SLAB_RECLAIM_ACCOUNT |
281 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900282 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 if (sock_inode_cachep == NULL)
284 return -ENOMEM;
285 return 0;
286}
287
288static struct super_operations sockfs_ops = {
289 .alloc_inode = sock_alloc_inode,
290 .destroy_inode =sock_destroy_inode,
291 .statfs = simple_statfs,
292};
293
David Howells454e2392006-06-23 02:02:57 -0700294static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700295 int flags, const char *dev_name, void *data,
296 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297{
David Howells454e2392006-06-23 02:02:57 -0700298 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
299 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300}
301
Eric Dumazetba899662005-08-26 12:05:31 -0700302static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
304static struct file_system_type sock_fs_type = {
305 .name = "sockfs",
306 .get_sb = sockfs_get_sb,
307 .kill_sb = kill_anon_super,
308};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310static int sockfs_delete_dentry(struct dentry *dentry)
311{
Eric Dumazet304e61e2006-12-06 20:38:49 -0800312 /*
313 * At creation time, we pretended this dentry was hashed
314 * (by clearing DCACHE_UNHASHED bit in d_flags)
315 * At delete time, we restore the truth : not hashed.
316 * (so that dput() can proceed correctly)
317 */
318 dentry->d_flags |= DCACHE_UNHASHED;
319 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320}
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700321
322/*
323 * sockfs_dname() is called from d_path().
324 */
325static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
326{
327 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
328 dentry->d_inode->i_ino);
329}
330
Al Viro3ba13d12009-02-20 06:02:22 +0000331static const struct dentry_operations sockfs_dentry_operations = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700332 .d_delete = sockfs_delete_dentry,
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700333 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334};
335
336/*
337 * Obtains the first available file descriptor and sets it up for use.
338 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800339 * These functions create file structures and maps them to fd space
340 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 * and file struct implicitly stored in sock->file.
342 * Note that another thread may close file descriptor before we return
343 * from this function. We use the fact that now we do not refer
344 * to socket after mapping. If one day we will need it, this
345 * function will increment ref. count on file by 1.
346 *
347 * In any case returned fd MAY BE not valid!
348 * This race condition is unavoidable
349 * with shared fd spaces, we cannot solve it inside kernel,
350 * but we take care of internal coherence yet.
351 */
352
Ulrich Dreppera677a032008-07-23 21:29:17 -0700353static int sock_alloc_fd(struct file **filep, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354{
355 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800356
Ulrich Dreppera677a032008-07-23 21:29:17 -0700357 fd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800358 if (likely(fd >= 0)) {
359 struct file *file = get_empty_filp();
360
361 *filep = file;
362 if (unlikely(!file)) {
363 put_unused_fd(fd);
364 return -ENFILE;
365 }
366 } else
367 *filep = NULL;
368 return fd;
369}
370
Ulrich Drepper77d27202008-07-23 21:29:35 -0700371static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800372{
Dave Hansence8d2cd2007-10-16 23:31:13 -0700373 struct dentry *dentry;
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700374 struct qstr name = { .name = "" };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
Dave Hansence8d2cd2007-10-16 23:31:13 -0700376 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
377 if (unlikely(!dentry))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800378 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379
Dave Hansence8d2cd2007-10-16 23:31:13 -0700380 dentry->d_op = &sockfs_dentry_operations;
Eric Dumazet304e61e2006-12-06 20:38:49 -0800381 /*
382 * We dont want to push this dentry into global dentry hash table.
383 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
384 * This permits a working /proc/$pid/fd/XXX on sockets
385 */
Dave Hansence8d2cd2007-10-16 23:31:13 -0700386 dentry->d_flags &= ~DCACHE_UNHASHED;
387 d_instantiate(dentry, SOCK_INODE(sock));
David S. Miller39d8c1b2006-03-20 17:13:49 -0800388
389 sock->file = file;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700390 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
391 &socket_file_ops);
392 SOCK_INODE(sock)->i_fop = &socket_file_ops;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700393 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800394 file->f_pos = 0;
395 file->private_data = sock;
396
397 return 0;
398}
399
Ulrich Dreppera677a032008-07-23 21:29:17 -0700400int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401{
402 struct file *newfile;
Ulrich Dreppera677a032008-07-23 21:29:17 -0700403 int fd = sock_alloc_fd(&newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800404
405 if (likely(fd >= 0)) {
Ulrich Drepper77d27202008-07-23 21:29:35 -0700406 int err = sock_attach_fd(sock, newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800407
408 if (unlikely(err < 0)) {
409 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800411 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800413 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 return fd;
416}
417
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800418static struct socket *sock_from_file(struct file *file, int *err)
419{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800420 if (file->f_op == &socket_file_ops)
421 return file->private_data; /* set in sock_map_fd */
422
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800423 *err = -ENOTSOCK;
424 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800425}
426
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427/**
428 * sockfd_lookup - Go from a file number to its socket slot
429 * @fd: file handle
430 * @err: pointer to an error code return
431 *
432 * The file handle passed in is locked and the socket it is bound
433 * too is returned. If an error occurs the err pointer is overwritten
434 * with a negative errno code and NULL is returned. The function checks
435 * for both invalid handles and passing a handle which is not a socket.
436 *
437 * On a success the socket object pointer is returned.
438 */
439
440struct socket *sockfd_lookup(int fd, int *err)
441{
442 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 struct socket *sock;
444
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700445 file = fget(fd);
446 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 *err = -EBADF;
448 return NULL;
449 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700450
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451 sock = sock_from_file(file, err);
452 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 return sock;
455}
456
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800457static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
458{
459 struct file *file;
460 struct socket *sock;
461
Hua Zhong36725582006-04-19 15:25:02 -0700462 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800463 file = fget_light(fd, fput_needed);
464 if (file) {
465 sock = sock_from_file(file, err);
466 if (sock)
467 return sock;
468 fput_light(file, *fput_needed);
469 }
470 return NULL;
471}
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473/**
474 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700475 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 * Allocate a new inode and socket object. The two are bound together
477 * and initialised. The socket is then returned. If we are out of inodes
478 * NULL is returned.
479 */
480
481static struct socket *sock_alloc(void)
482{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700483 struct inode *inode;
484 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
486 inode = new_inode(sock_mnt->mnt_sb);
487 if (!inode)
488 return NULL;
489
490 sock = SOCKET_I(inode);
491
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700492 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100493 inode->i_uid = current_fsuid();
494 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Eric Dumazet4e694892009-04-04 16:41:09 -0700496 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 return sock;
498}
499
500/*
501 * In theory you can't get an open on this inode, but /proc provides
502 * a back door. Remember to keep it shut otherwise you'll let the
503 * creepy crawlies in.
504 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700505
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
507{
508 return -ENXIO;
509}
510
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800511const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 .owner = THIS_MODULE,
513 .open = sock_no_open,
514};
515
516/**
517 * sock_release - close a socket
518 * @sock: socket to close
519 *
520 * The socket is released from the protocol stack if it has a release
521 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700522 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700524
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525void sock_release(struct socket *sock)
526{
527 if (sock->ops) {
528 struct module *owner = sock->ops->owner;
529
530 sock->ops->release(sock);
531 sock->ops = NULL;
532 module_put(owner);
533 }
534
535 if (sock->fasync_list)
536 printk(KERN_ERR "sock_release: fasync list not empty!\n");
537
Eric Dumazet4e694892009-04-04 16:41:09 -0700538 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 if (!sock->file) {
540 iput(SOCK_INODE(sock));
541 return;
542 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700543 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544}
545
Patrick Ohly20d49472009-02-12 05:03:38 +0000546int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
547 union skb_shared_tx *shtx)
548{
549 shtx->flags = 0;
550 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
551 shtx->hardware = 1;
552 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
553 shtx->software = 1;
554 return 0;
555}
556EXPORT_SYMBOL(sock_tx_timestamp);
557
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700558static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 struct msghdr *msg, size_t size)
560{
561 struct sock_iocb *si = kiocb_to_siocb(iocb);
562 int err;
563
564 si->sock = sock;
565 si->scm = NULL;
566 si->msg = msg;
567 si->size = size;
568
569 err = security_socket_sendmsg(sock, msg, size);
570 if (err)
571 return err;
572
573 return sock->ops->sendmsg(iocb, sock, msg, size);
574}
575
576int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
577{
578 struct kiocb iocb;
579 struct sock_iocb siocb;
580 int ret;
581
582 init_sync_kiocb(&iocb, NULL);
583 iocb.private = &siocb;
584 ret = __sock_sendmsg(&iocb, sock, msg, size);
585 if (-EIOCBQUEUED == ret)
586 ret = wait_on_sync_kiocb(&iocb);
587 return ret;
588}
589
590int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
591 struct kvec *vec, size_t num, size_t size)
592{
593 mm_segment_t oldfs = get_fs();
594 int result;
595
596 set_fs(KERNEL_DS);
597 /*
598 * the following is safe, since for compiler definitions of kvec and
599 * iovec are identical, yielding the same in-core layout and alignment
600 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700601 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 msg->msg_iovlen = num;
603 result = sock_sendmsg(sock, msg, size);
604 set_fs(oldfs);
605 return result;
606}
607
Patrick Ohly20d49472009-02-12 05:03:38 +0000608static int ktime2ts(ktime_t kt, struct timespec *ts)
609{
610 if (kt.tv64) {
611 *ts = ktime_to_timespec(kt);
612 return 1;
613 } else {
614 return 0;
615 }
616}
617
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700618/*
619 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
620 */
621void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
622 struct sk_buff *skb)
623{
Patrick Ohly20d49472009-02-12 05:03:38 +0000624 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
625 struct timespec ts[3];
626 int empty = 1;
627 struct skb_shared_hwtstamps *shhwtstamps =
628 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700629
Patrick Ohly20d49472009-02-12 05:03:38 +0000630 /* Race occurred between timestamp enabling and packet
631 receiving. Fill in the current time for now. */
632 if (need_software_tstamp && skb->tstamp.tv64 == 0)
633 __net_timestamp(skb);
634
635 if (need_software_tstamp) {
636 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
637 struct timeval tv;
638 skb_get_timestamp(skb, &tv);
639 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
640 sizeof(tv), &tv);
641 } else {
642 struct timespec ts;
643 skb_get_timestampns(skb, &ts);
644 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
645 sizeof(ts), &ts);
646 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700647 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000648
649
650 memset(ts, 0, sizeof(ts));
651 if (skb->tstamp.tv64 &&
652 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
653 skb_get_timestampns(skb, ts + 0);
654 empty = 0;
655 }
656 if (shhwtstamps) {
657 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
658 ktime2ts(shhwtstamps->syststamp, ts + 1))
659 empty = 0;
660 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
661 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
662 empty = 0;
663 }
664 if (!empty)
665 put_cmsg(msg, SOL_SOCKET,
666 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700667}
668
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300669EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
670
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700671static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 struct msghdr *msg, size_t size, int flags)
673{
674 int err;
675 struct sock_iocb *si = kiocb_to_siocb(iocb);
676
677 si->sock = sock;
678 si->scm = NULL;
679 si->msg = msg;
680 si->size = size;
681 si->flags = flags;
682
683 err = security_socket_recvmsg(sock, msg, size, flags);
684 if (err)
685 return err;
686
687 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
688}
689
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700690int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691 size_t size, int flags)
692{
693 struct kiocb iocb;
694 struct sock_iocb siocb;
695 int ret;
696
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700697 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 iocb.private = &siocb;
699 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
700 if (-EIOCBQUEUED == ret)
701 ret = wait_on_sync_kiocb(&iocb);
702 return ret;
703}
704
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700705int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
706 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707{
708 mm_segment_t oldfs = get_fs();
709 int result;
710
711 set_fs(KERNEL_DS);
712 /*
713 * the following is safe, since for compiler definitions of kvec and
714 * iovec are identical, yielding the same in-core layout and alignment
715 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700716 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 result = sock_recvmsg(sock, msg, size, flags);
718 set_fs(oldfs);
719 return result;
720}
721
722static void sock_aio_dtor(struct kiocb *iocb)
723{
724 kfree(iocb->private);
725}
726
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300727static ssize_t sock_sendpage(struct file *file, struct page *page,
728 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729{
730 struct socket *sock;
731 int flags;
732
Eric Dumazetb69aee02005-09-06 14:42:45 -0700733 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734
735 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
736 if (more)
737 flags |= MSG_MORE;
738
739 return sock->ops->sendpage(sock, page, offset, size, flags);
740}
741
Jens Axboe9c55e012007-11-06 23:30:13 -0800742static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
743 struct pipe_inode_info *pipe, size_t len,
744 unsigned int flags)
745{
746 struct socket *sock = file->private_data;
747
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800748 if (unlikely(!sock->ops->splice_read))
749 return -EINVAL;
750
Jens Axboe9c55e012007-11-06 23:30:13 -0800751 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
752}
753
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800754static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700755 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800756{
757 if (!is_sync_kiocb(iocb)) {
758 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
759 if (!siocb)
760 return NULL;
761 iocb->ki_dtor = sock_aio_dtor;
762 }
763
764 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800765 iocb->private = siocb;
766 return siocb;
767}
768
769static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700770 struct file *file, const struct iovec *iov,
771 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800772{
773 struct socket *sock = file->private_data;
774 size_t size = 0;
775 int i;
776
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700777 for (i = 0; i < nr_segs; i++)
778 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800779
780 msg->msg_name = NULL;
781 msg->msg_namelen = 0;
782 msg->msg_control = NULL;
783 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700784 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800785 msg->msg_iovlen = nr_segs;
786 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
787
788 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
789}
790
Badari Pulavarty027445c2006-09-30 23:28:46 -0700791static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
792 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800793{
794 struct sock_iocb siocb, *x;
795
796 if (pos != 0)
797 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700798
799 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800800 return 0;
801
Badari Pulavarty027445c2006-09-30 23:28:46 -0700802
803 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800804 if (!x)
805 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700806 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800807}
808
809static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700810 struct file *file, const struct iovec *iov,
811 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800812{
813 struct socket *sock = file->private_data;
814 size_t size = 0;
815 int i;
816
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700817 for (i = 0; i < nr_segs; i++)
818 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800819
820 msg->msg_name = NULL;
821 msg->msg_namelen = 0;
822 msg->msg_control = NULL;
823 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700824 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800825 msg->msg_iovlen = nr_segs;
826 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
827 if (sock->type == SOCK_SEQPACKET)
828 msg->msg_flags |= MSG_EOR;
829
830 return __sock_sendmsg(iocb, sock, msg, size);
831}
832
Badari Pulavarty027445c2006-09-30 23:28:46 -0700833static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
834 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800836 struct sock_iocb siocb, *x;
837
838 if (pos != 0)
839 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700840
Badari Pulavarty027445c2006-09-30 23:28:46 -0700841 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800842 if (!x)
843 return -ENOMEM;
844
Badari Pulavarty027445c2006-09-30 23:28:46 -0700845 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846}
847
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848/*
849 * Atomic setting of ioctl hooks to avoid race
850 * with module unload.
851 */
852
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800853static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700854static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
Eric W. Biederman881d9662007-09-17 11:56:21 -0700856void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800858 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800860 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700862
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863EXPORT_SYMBOL(brioctl_set);
864
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800865static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700866static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
Eric W. Biederman881d9662007-09-17 11:56:21 -0700868void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800870 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800872 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700874
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875EXPORT_SYMBOL(vlan_ioctl_set);
876
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800877static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700878static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700880void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800882 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800884 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700886
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887EXPORT_SYMBOL(dlci_ioctl_set);
888
889/*
890 * With an ioctl, arg may well be a user mode pointer, but we don't know
891 * what to do with it - that's up to the protocol still.
892 */
893
894static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
895{
896 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700897 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 void __user *argp = (void __user *)arg;
899 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700900 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901
Eric Dumazetb69aee02005-09-06 14:42:45 -0700902 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700903 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900904 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700906 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100908#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700910 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700912#endif /* CONFIG_WIRELESS_EXT */
913 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 case FIOSETOWN:
915 case SIOCSPGRP:
916 err = -EFAULT;
917 if (get_user(pid, (int __user *)argp))
918 break;
919 err = f_setown(sock->file, pid, 1);
920 break;
921 case FIOGETOWN:
922 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700923 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700924 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 break;
926 case SIOCGIFBR:
927 case SIOCSIFBR:
928 case SIOCBRADDBR:
929 case SIOCBRDELBR:
930 err = -ENOPKG;
931 if (!br_ioctl_hook)
932 request_module("bridge");
933
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800934 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700935 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700936 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800937 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 break;
939 case SIOCGIFVLAN:
940 case SIOCSIFVLAN:
941 err = -ENOPKG;
942 if (!vlan_ioctl_hook)
943 request_module("8021q");
944
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800945 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700947 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800948 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 case SIOCADDDLCI:
951 case SIOCDELDLCI:
952 err = -ENOPKG;
953 if (!dlci_ioctl_hook)
954 request_module("dlci");
955
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700956 mutex_lock(&dlci_ioctl_mutex);
957 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700959 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 break;
961 default:
962 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800963
964 /*
965 * If this ioctl is unknown try to hand it down
966 * to the NIC driver.
967 */
968 if (err == -ENOIOCTLCMD)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700969 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700971 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 return err;
973}
974
975int sock_create_lite(int family, int type, int protocol, struct socket **res)
976{
977 int err;
978 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 err = security_socket_create(family, type, protocol, 1);
981 if (err)
982 goto out;
983
984 sock = sock_alloc();
985 if (!sock) {
986 err = -ENOMEM;
987 goto out;
988 }
989
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700991 err = security_socket_post_create(sock, family, type, protocol, 1);
992 if (err)
993 goto out_release;
994
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995out:
996 *res = sock;
997 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700998out_release:
999 sock_release(sock);
1000 sock = NULL;
1001 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002}
1003
1004/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001005static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006{
1007 struct socket *sock;
1008
1009 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001010 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001012 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 return sock->ops->poll(file, sock, wait);
1014}
1015
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001016static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001018 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
1020 return sock->ops->mmap(file, sock, vma);
1021}
1022
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001023static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024{
1025 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001026 * It was possible the inode is NULL we were
1027 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 */
1029
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001030 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 printk(KERN_DEBUG "sock_close: NULL inode\n");
1032 return 0;
1033 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 sock_release(SOCKET_I(inode));
1035 return 0;
1036}
1037
1038/*
1039 * Update the socket async list
1040 *
1041 * Fasync_list locking strategy.
1042 *
1043 * 1. fasync_list is modified only under process context socket lock
1044 * i.e. under semaphore.
1045 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1046 * or under socket lock.
1047 * 3. fasync_list can be used from softirq context, so that
1048 * modification under socket lock have to be enhanced with
1049 * write_lock_bh(&sk->sk_callback_lock).
1050 * --ANK (990710)
1051 */
1052
1053static int sock_fasync(int fd, struct file *filp, int on)
1054{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001055 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 struct socket *sock;
1057 struct sock *sk;
1058
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001059 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001060 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001061 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 return -ENOMEM;
1063 }
1064
Eric Dumazetb69aee02005-09-06 14:42:45 -07001065 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001067 sk = sock->sk;
1068 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 kfree(fna);
1070 return -EINVAL;
1071 }
1072
1073 lock_sock(sk);
1074
Jonathan Corbet76398422009-02-01 14:26:59 -07001075 spin_lock(&filp->f_lock);
1076 if (on)
1077 filp->f_flags |= FASYNC;
1078 else
1079 filp->f_flags &= ~FASYNC;
1080 spin_unlock(&filp->f_lock);
1081
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001082 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001084 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1085 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 break;
1087
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001088 if (on) {
1089 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001091 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 write_unlock_bh(&sk->sk_callback_lock);
1093
1094 kfree(fna);
1095 goto out;
1096 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001097 fna->fa_file = filp;
1098 fna->fa_fd = fd;
1099 fna->magic = FASYNC_MAGIC;
1100 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001102 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001104 } else {
1105 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001107 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 write_unlock_bh(&sk->sk_callback_lock);
1109 kfree(fa);
1110 }
1111 }
1112
1113out:
1114 release_sock(sock->sk);
1115 return 0;
1116}
1117
1118/* This function may be called only under socket lock or callback_lock */
1119
1120int sock_wake_async(struct socket *sock, int how, int band)
1121{
1122 if (!sock || !sock->fasync_list)
1123 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001124 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001125 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1127 break;
1128 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001129 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1131 break;
1132 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001133 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001134call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 __kill_fasync(sock->fasync_list, SIGIO, band);
1136 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001137 case SOCK_WAKE_URG:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 __kill_fasync(sock->fasync_list, SIGURG, band);
1139 }
1140 return 0;
1141}
1142
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001143static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001144 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145{
1146 int err;
1147 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001148 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
1150 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001151 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 */
1153 if (family < 0 || family >= NPROTO)
1154 return -EAFNOSUPPORT;
1155 if (type < 0 || type >= SOCK_MAX)
1156 return -EINVAL;
1157
1158 /* Compatibility.
1159
1160 This uglymoron is moved from INET layer to here to avoid
1161 deadlock in module load.
1162 */
1163 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001164 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 if (!warned) {
1166 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001167 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1168 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 }
1170 family = PF_PACKET;
1171 }
1172
1173 err = security_socket_create(family, type, protocol, kern);
1174 if (err)
1175 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001176
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001177 /*
1178 * Allocate the socket and allow the family to set things up. if
1179 * the protocol is 0, the family is instructed to select an appropriate
1180 * default.
1181 */
1182 sock = sock_alloc();
1183 if (!sock) {
1184 if (net_ratelimit())
1185 printk(KERN_WARNING "socket: no more sockets\n");
1186 return -ENFILE; /* Not exactly a match, but its the
1187 closest posix thing */
1188 }
1189
1190 sock->type = type;
1191
Johannes Berg95a5afc2008-10-16 15:24:51 -07001192#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001193 /* Attempt to load a protocol module if the find failed.
1194 *
1195 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 * requested real, full-featured networking support upon configuration.
1197 * Otherwise module support will break!
1198 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001199 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001200 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201#endif
1202
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001203 rcu_read_lock();
1204 pf = rcu_dereference(net_families[family]);
1205 err = -EAFNOSUPPORT;
1206 if (!pf)
1207 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208
1209 /*
1210 * We will call the ->create function, that possibly is in a loadable
1211 * module, so we have to bump that loadable module refcnt first.
1212 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001213 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 goto out_release;
1215
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001216 /* Now protected by module ref count */
1217 rcu_read_unlock();
1218
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001219 err = pf->create(net, sock, protocol);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001220 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001222
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 /*
1224 * Now to bump the refcnt of the [loadable] module that owns this
1225 * socket at sock_release time we decrement its refcnt.
1226 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001227 if (!try_module_get(sock->ops->owner))
1228 goto out_module_busy;
1229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 /*
1231 * Now that we're done with the ->create function, the [loadable]
1232 * module can have its refcnt decremented
1233 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001234 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001235 err = security_socket_post_create(sock, family, type, protocol, kern);
1236 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001237 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001238 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001240 return 0;
1241
1242out_module_busy:
1243 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001245 sock->ops = NULL;
1246 module_put(pf->owner);
1247out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001249 return err;
1250
1251out_release:
1252 rcu_read_unlock();
1253 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254}
1255
1256int sock_create(int family, int type, int protocol, struct socket **res)
1257{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001258 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259}
1260
1261int sock_create_kern(int family, int type, int protocol, struct socket **res)
1262{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001263 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264}
1265
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001266SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267{
1268 int retval;
1269 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001270 int flags;
1271
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001272 /* Check the SOCK_* constants for consistency. */
1273 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1274 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1275 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1276 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1277
Ulrich Dreppera677a032008-07-23 21:29:17 -07001278 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001279 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001280 return -EINVAL;
1281 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001283 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1284 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1285
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 retval = sock_create(family, type, protocol, &sock);
1287 if (retval < 0)
1288 goto out;
1289
Ulrich Drepper77d27202008-07-23 21:29:35 -07001290 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 if (retval < 0)
1292 goto out_release;
1293
1294out:
1295 /* It may be already another descriptor 8) Not kernel problem. */
1296 return retval;
1297
1298out_release:
1299 sock_release(sock);
1300 return retval;
1301}
1302
1303/*
1304 * Create a pair of connected sockets.
1305 */
1306
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001307SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1308 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309{
1310 struct socket *sock1, *sock2;
1311 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001312 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001313 int flags;
1314
1315 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001316 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001317 return -EINVAL;
1318 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001320 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1321 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1322
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 /*
1324 * Obtain the first socket and check if the underlying protocol
1325 * supports the socketpair call.
1326 */
1327
1328 err = sock_create(family, type, protocol, &sock1);
1329 if (err < 0)
1330 goto out;
1331
1332 err = sock_create(family, type, protocol, &sock2);
1333 if (err < 0)
1334 goto out_release_1;
1335
1336 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001337 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 goto out_release_both;
1339
Ulrich Dreppera677a032008-07-23 21:29:17 -07001340 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001341 if (unlikely(fd1 < 0)) {
1342 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001344 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
Ulrich Dreppera677a032008-07-23 21:29:17 -07001346 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
Al Virodb349502007-02-07 01:48:00 -05001347 if (unlikely(fd2 < 0)) {
David S. Millerbf3c23d2007-10-29 21:54:02 -07001348 err = fd2;
Al Virodb349502007-02-07 01:48:00 -05001349 put_filp(newfile1);
1350 put_unused_fd(fd1);
1351 goto out_release_both;
1352 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353
Ulrich Drepper77d27202008-07-23 21:29:35 -07001354 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
Al Virodb349502007-02-07 01:48:00 -05001355 if (unlikely(err < 0)) {
1356 goto out_fd2;
1357 }
1358
Ulrich Drepper77d27202008-07-23 21:29:35 -07001359 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
Al Virodb349502007-02-07 01:48:00 -05001360 if (unlikely(err < 0)) {
1361 fput(newfile1);
1362 goto out_fd1;
1363 }
1364
Al Viro157cf642008-12-14 04:57:47 -05001365 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001366 fd_install(fd1, newfile1);
1367 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 /* fd1 and fd2 may be already another descriptors.
1369 * Not kernel problem.
1370 */
1371
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001372 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 if (!err)
1374 err = put_user(fd2, &usockvec[1]);
1375 if (!err)
1376 return 0;
1377
1378 sys_close(fd2);
1379 sys_close(fd1);
1380 return err;
1381
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001383 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001385 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386out:
1387 return err;
Al Virodb349502007-02-07 01:48:00 -05001388
1389out_fd2:
1390 put_filp(newfile1);
1391 sock_release(sock1);
1392out_fd1:
1393 put_filp(newfile2);
1394 sock_release(sock2);
Al Virodb349502007-02-07 01:48:00 -05001395 put_unused_fd(fd1);
1396 put_unused_fd(fd2);
1397 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398}
1399
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400/*
1401 * Bind a name to a socket. Nothing much to do here since it's
1402 * the protocol's responsibility to handle the local address.
1403 *
1404 * We move the socket address to kernel space before we call
1405 * the protocol layer (having also checked the address is ok).
1406 */
1407
Heiko Carstens20f37032009-01-14 14:14:23 +01001408SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409{
1410 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001411 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001412 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001414 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001415 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001416 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001417 if (err >= 0) {
1418 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001419 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001420 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001421 if (!err)
1422 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001423 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001424 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001426 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001427 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 return err;
1429}
1430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431/*
1432 * Perform a listen. Basically, we allow the protocol to do anything
1433 * necessary for a listen, and if that works, we mark the socket as
1434 * ready for listening.
1435 */
1436
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001437SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438{
1439 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001440 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001441 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001442
1443 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1444 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001445 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001446 if ((unsigned)backlog > somaxconn)
1447 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448
1449 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001450 if (!err)
1451 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001453 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 }
1455 return err;
1456}
1457
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458/*
1459 * For accept, we attempt to create a new socket, set up the link
1460 * with the client, wake up the client, then return the new
1461 * connected fd. We collect the address of the connector in kernel
1462 * space and move it to user at the very end. This is unclean because
1463 * we open the socket then return an error.
1464 *
1465 * 1003.1g adds the ability to recvmsg() to query connection pending
1466 * status to recvmsg. We need to add that support in a way thats
1467 * clean when we restucture accept also.
1468 */
1469
Heiko Carstens20f37032009-01-14 14:14:23 +01001470SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1471 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472{
1473 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001474 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001475 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001476 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
Ulrich Drepper77d27202008-07-23 21:29:35 -07001478 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001479 return -EINVAL;
1480
1481 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1482 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1483
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 if (!sock)
1486 goto out;
1487
1488 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001489 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 goto out_put;
1491
1492 newsock->type = sock->type;
1493 newsock->ops = sock->ops;
1494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 /*
1496 * We don't need try_module_get here, as the listening socket (sock)
1497 * has the protocol module (sock->ops->owner) held.
1498 */
1499 __module_get(newsock->ops->owner);
1500
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001501 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001502 if (unlikely(newfd < 0)) {
1503 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001504 sock_release(newsock);
1505 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001506 }
1507
Ulrich Drepper77d27202008-07-23 21:29:35 -07001508 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001509 if (err < 0)
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001510 goto out_fd_simple;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001511
Frank Filza79af592005-09-27 15:23:38 -07001512 err = security_socket_accept(sock, newsock);
1513 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001514 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1517 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001518 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
1520 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001521 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001522 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001524 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001526 err = move_addr_to_user((struct sockaddr *)&address,
1527 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001529 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 }
1531
1532 /* File flags are not inherited via accept() unlike another OSes. */
1533
David S. Miller39d8c1b2006-03-20 17:13:49 -08001534 fd_install(newfd, newfile);
1535 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001538 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539out:
1540 return err;
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001541out_fd_simple:
1542 sock_release(newsock);
1543 put_filp(newfile);
1544 put_unused_fd(newfd);
1545 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001546out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001547 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001548 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 goto out_put;
1550}
1551
Heiko Carstens20f37032009-01-14 14:14:23 +01001552SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1553 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001554{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001555 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001556}
1557
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558/*
1559 * Attempt to connect to a socket with the server address. The address
1560 * is in user space so we verify it is OK and move it to kernel space.
1561 *
1562 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1563 * break bindings
1564 *
1565 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1566 * other SEQPACKET protocols that take time to connect() as it doesn't
1567 * include the -EINPROGRESS status for such sockets.
1568 */
1569
Heiko Carstens20f37032009-01-14 14:14:23 +01001570SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1571 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572{
1573 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001574 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001575 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001577 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 if (!sock)
1579 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001580 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 if (err < 0)
1582 goto out_put;
1583
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001584 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001585 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 if (err)
1587 goto out_put;
1588
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001589 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 sock->file->f_flags);
1591out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001592 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593out:
1594 return err;
1595}
1596
1597/*
1598 * Get the local address ('name') of a socket object. Move the obtained
1599 * name to user space.
1600 */
1601
Heiko Carstens20f37032009-01-14 14:14:23 +01001602SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1603 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604{
1605 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001606 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001607 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001608
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001609 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 if (!sock)
1611 goto out;
1612
1613 err = security_socket_getsockname(sock);
1614 if (err)
1615 goto out_put;
1616
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001617 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 if (err)
1619 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001620 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621
1622out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001623 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624out:
1625 return err;
1626}
1627
1628/*
1629 * Get the remote address ('name') of a socket object. Move the obtained
1630 * name to user space.
1631 */
1632
Heiko Carstens20f37032009-01-14 14:14:23 +01001633SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1634 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635{
1636 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001637 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001638 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001640 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1641 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 err = security_socket_getpeername(sock);
1643 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001644 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 return err;
1646 }
1647
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001648 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001649 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001650 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001652 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001653 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001654 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 }
1656 return err;
1657}
1658
1659/*
1660 * Send a datagram to a given address. We move the address into kernel
1661 * space and check the user space data area is readable before invoking
1662 * the protocol.
1663 */
1664
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001665SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1666 unsigned, flags, struct sockaddr __user *, addr,
1667 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668{
1669 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001670 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 int err;
1672 struct msghdr msg;
1673 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001674 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001675
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001676 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1677 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001678 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001679
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001680 iov.iov_base = buff;
1681 iov.iov_len = len;
1682 msg.msg_name = NULL;
1683 msg.msg_iov = &iov;
1684 msg.msg_iovlen = 1;
1685 msg.msg_control = NULL;
1686 msg.msg_controllen = 0;
1687 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001688 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001689 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 if (err < 0)
1691 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001692 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001693 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 }
1695 if (sock->file->f_flags & O_NONBLOCK)
1696 flags |= MSG_DONTWAIT;
1697 msg.msg_flags = flags;
1698 err = sock_sendmsg(sock, &msg, len);
1699
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001700out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001701 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001702out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 return err;
1704}
1705
1706/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001707 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 */
1709
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001710SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1711 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712{
1713 return sys_sendto(fd, buff, len, flags, NULL, 0);
1714}
1715
1716/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001717 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 * sender. We verify the buffers are writable and if needed move the
1719 * sender address from kernel to user space.
1720 */
1721
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001722SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1723 unsigned, flags, struct sockaddr __user *, addr,
1724 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725{
1726 struct socket *sock;
1727 struct iovec iov;
1728 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001729 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001730 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001731 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001733 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001735 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001737 msg.msg_control = NULL;
1738 msg.msg_controllen = 0;
1739 msg.msg_iovlen = 1;
1740 msg.msg_iov = &iov;
1741 iov.iov_len = size;
1742 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001743 msg.msg_name = (struct sockaddr *)&address;
1744 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 if (sock->file->f_flags & O_NONBLOCK)
1746 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001747 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001749 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001750 err2 = move_addr_to_user((struct sockaddr *)&address,
1751 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001752 if (err2 < 0)
1753 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001755
1756 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001757out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 return err;
1759}
1760
1761/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001762 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 */
1764
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001765asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1766 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767{
1768 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1769}
1770
1771/*
1772 * Set a socket option. Because we don't know the option lengths we have
1773 * to pass the user mode parameter for the protocols to sort out.
1774 */
1775
Heiko Carstens20f37032009-01-14 14:14:23 +01001776SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1777 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001779 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 struct socket *sock;
1781
1782 if (optlen < 0)
1783 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001784
1785 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1786 if (sock != NULL) {
1787 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001788 if (err)
1789 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
1791 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001792 err =
1793 sock_setsockopt(sock, level, optname, optval,
1794 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001796 err =
1797 sock->ops->setsockopt(sock, level, optname, optval,
1798 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001799out_put:
1800 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 }
1802 return err;
1803}
1804
1805/*
1806 * Get a socket option. Because we don't know the option lengths we have
1807 * to pass a user mode parameter for the protocols to sort out.
1808 */
1809
Heiko Carstens20f37032009-01-14 14:14:23 +01001810SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1811 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001813 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 struct socket *sock;
1815
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001816 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1817 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001818 err = security_socket_getsockopt(sock, level, optname);
1819 if (err)
1820 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821
1822 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001823 err =
1824 sock_getsockopt(sock, level, optname, optval,
1825 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001827 err =
1828 sock->ops->getsockopt(sock, level, optname, optval,
1829 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001830out_put:
1831 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 }
1833 return err;
1834}
1835
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836/*
1837 * Shutdown a socket.
1838 */
1839
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001840SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001842 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 struct socket *sock;
1844
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001845 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1846 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001848 if (!err)
1849 err = sock->ops->shutdown(sock, how);
1850 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 }
1852 return err;
1853}
1854
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001855/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 * fields which are the same type (int / unsigned) on our platforms.
1857 */
1858#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1859#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1860#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1861
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862/*
1863 * BSD sendmsg interface
1864 */
1865
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001866SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001868 struct compat_msghdr __user *msg_compat =
1869 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001871 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001873 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001874 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1875 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876 unsigned char *ctl_buf = ctl;
1877 struct msghdr msg_sys;
1878 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001879 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001880
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 err = -EFAULT;
1882 if (MSG_CMSG_COMPAT & flags) {
1883 if (get_compat_msghdr(&msg_sys, msg_compat))
1884 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001885 }
1886 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 return -EFAULT;
1888
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001889 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 goto out;
1892
1893 /* do not move before msg_sys is valid */
1894 err = -EMSGSIZE;
1895 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1896 goto out_put;
1897
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001898 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 err = -ENOMEM;
1900 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1901 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1902 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1903 if (!iov)
1904 goto out_put;
1905 }
1906
1907 /* This will also move the address data into kernel space */
1908 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001909 err = verify_compat_iovec(&msg_sys, iov,
1910 (struct sockaddr *)&address,
1911 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001913 err = verify_iovec(&msg_sys, iov,
1914 (struct sockaddr *)&address,
1915 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001916 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 goto out_freeiov;
1918 total_len = err;
1919
1920 err = -ENOBUFS;
1921
1922 if (msg_sys.msg_controllen > INT_MAX)
1923 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001924 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001926 err =
1927 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1928 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 if (err)
1930 goto out_freeiov;
1931 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001932 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001934 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001936 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 goto out_freeiov;
1938 }
1939 err = -EFAULT;
1940 /*
1941 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1942 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1943 * checking falls down on this.
1944 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001945 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1946 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947 goto out_freectl;
1948 msg_sys.msg_control = ctl_buf;
1949 }
1950 msg_sys.msg_flags = flags;
1951
1952 if (sock->file->f_flags & O_NONBLOCK)
1953 msg_sys.msg_flags |= MSG_DONTWAIT;
1954 err = sock_sendmsg(sock, &msg_sys, total_len);
1955
1956out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001957 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1959out_freeiov:
1960 if (iov != iovstack)
1961 sock_kfree_s(sock->sk, iov, iov_size);
1962out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001963 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001964out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 return err;
1966}
1967
1968/*
1969 * BSD recvmsg interface
1970 */
1971
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001972SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
1973 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001975 struct compat_msghdr __user *msg_compat =
1976 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 struct socket *sock;
1978 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001979 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 struct msghdr msg_sys;
1981 unsigned long cmsg_ptr;
1982 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001983 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984
1985 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001986 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987
1988 /* user mode address pointers */
1989 struct sockaddr __user *uaddr;
1990 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001991
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 if (MSG_CMSG_COMPAT & flags) {
1993 if (get_compat_msghdr(&msg_sys, msg_compat))
1994 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001995 }
1996 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1997 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001999 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 if (!sock)
2001 goto out;
2002
2003 err = -EMSGSIZE;
2004 if (msg_sys.msg_iovlen > UIO_MAXIOV)
2005 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002006
2007 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 err = -ENOMEM;
2009 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2010 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2011 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2012 if (!iov)
2013 goto out_put;
2014 }
2015
2016 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002017 * Save the user-mode address (verify_iovec will change the
2018 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -07002021 uaddr = (__force void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 uaddr_len = COMPAT_NAMELEN(msg);
2023 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002024 err = verify_compat_iovec(&msg_sys, iov,
2025 (struct sockaddr *)&addr,
2026 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002028 err = verify_iovec(&msg_sys, iov,
2029 (struct sockaddr *)&addr,
2030 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 if (err < 0)
2032 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002033 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034
2035 cmsg_ptr = (unsigned long)msg_sys.msg_control;
Ulrich Drepper4a195422007-07-15 23:40:34 -07002036 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002037
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 if (sock->file->f_flags & O_NONBLOCK)
2039 flags |= MSG_DONTWAIT;
2040 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2041 if (err < 0)
2042 goto out_freeiov;
2043 len = err;
2044
2045 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002046 err = move_addr_to_user((struct sockaddr *)&addr,
2047 msg_sys.msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002048 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 if (err < 0)
2050 goto out_freeiov;
2051 }
David S. Miller37f7f422005-09-16 16:51:01 -07002052 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2053 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 if (err)
2055 goto out_freeiov;
2056 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002057 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 &msg_compat->msg_controllen);
2059 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002060 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 &msg->msg_controllen);
2062 if (err)
2063 goto out_freeiov;
2064 err = len;
2065
2066out_freeiov:
2067 if (iov != iovstack)
2068 sock_kfree_s(sock->sk, iov, iov_size);
2069out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002070 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071out:
2072 return err;
2073}
2074
2075#ifdef __ARCH_WANT_SYS_SOCKETCALL
2076
2077/* Argument list sizes for sys_socketcall */
2078#define AL(x) ((x) * sizeof(unsigned long))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002079static const unsigned char nargs[19]={
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002080 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2081 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002082 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
Ulrich Drepperde11def2008-11-19 15:36:14 -08002083 AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002084};
2085
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086#undef AL
2087
2088/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002089 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 *
2091 * Argument checking cleaned up. Saved 20% in size.
2092 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002093 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 */
2095
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002096SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097{
2098 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002099 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 int err;
2101
Ulrich Drepperde11def2008-11-19 15:36:14 -08002102 if (call < 1 || call > SYS_ACCEPT4)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 return -EINVAL;
2104
2105 /* copy_from_user should be SMP safe. */
2106 if (copy_from_user(a, args, nargs[call]))
2107 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002108
Al Virof3298dc2008-12-10 03:16:51 -05002109 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002110
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002111 a0 = a[0];
2112 a1 = a[1];
2113
2114 switch (call) {
2115 case SYS_SOCKET:
2116 err = sys_socket(a0, a1, a[2]);
2117 break;
2118 case SYS_BIND:
2119 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2120 break;
2121 case SYS_CONNECT:
2122 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2123 break;
2124 case SYS_LISTEN:
2125 err = sys_listen(a0, a1);
2126 break;
2127 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002128 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2129 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002130 break;
2131 case SYS_GETSOCKNAME:
2132 err =
2133 sys_getsockname(a0, (struct sockaddr __user *)a1,
2134 (int __user *)a[2]);
2135 break;
2136 case SYS_GETPEERNAME:
2137 err =
2138 sys_getpeername(a0, (struct sockaddr __user *)a1,
2139 (int __user *)a[2]);
2140 break;
2141 case SYS_SOCKETPAIR:
2142 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2143 break;
2144 case SYS_SEND:
2145 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2146 break;
2147 case SYS_SENDTO:
2148 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2149 (struct sockaddr __user *)a[4], a[5]);
2150 break;
2151 case SYS_RECV:
2152 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2153 break;
2154 case SYS_RECVFROM:
2155 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2156 (struct sockaddr __user *)a[4],
2157 (int __user *)a[5]);
2158 break;
2159 case SYS_SHUTDOWN:
2160 err = sys_shutdown(a0, a1);
2161 break;
2162 case SYS_SETSOCKOPT:
2163 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2164 break;
2165 case SYS_GETSOCKOPT:
2166 err =
2167 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2168 (int __user *)a[4]);
2169 break;
2170 case SYS_SENDMSG:
2171 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2172 break;
2173 case SYS_RECVMSG:
2174 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2175 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002176 case SYS_ACCEPT4:
2177 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2178 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002179 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002180 default:
2181 err = -EINVAL;
2182 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 }
2184 return err;
2185}
2186
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002187#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002189/**
2190 * sock_register - add a socket protocol handler
2191 * @ops: description of protocol
2192 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 * This function is called by a protocol handler that wants to
2194 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002195 * socket interface. The value ops->family coresponds to the
2196 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002198int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199{
2200 int err;
2201
2202 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002203 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2204 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205 return -ENOBUFS;
2206 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002207
2208 spin_lock(&net_family_lock);
2209 if (net_families[ops->family])
2210 err = -EEXIST;
2211 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002212 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 err = 0;
2214 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002215 spin_unlock(&net_family_lock);
2216
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002217 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 return err;
2219}
2220
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002221/**
2222 * sock_unregister - remove a protocol handler
2223 * @family: protocol family to remove
2224 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225 * This function is called by a protocol handler that wants to
2226 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002227 * new socket creation.
2228 *
2229 * If protocol handler is a module, then it can use module reference
2230 * counts to protect against new references. If protocol handler is not
2231 * a module then it needs to provide its own protection in
2232 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002234void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002236 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002238 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002239 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002240 spin_unlock(&net_family_lock);
2241
2242 synchronize_rcu();
2243
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002244 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245}
2246
Andi Kleen77d76ea2005-12-22 12:43:42 -08002247static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002248{
2249 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002250 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002252
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 sk_init();
2254
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002256 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257 */
2258 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259
2260 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002261 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262 */
2263
2264 init_inodecache();
2265 register_filesystem(&sock_fs_type);
2266 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002267
2268 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269 */
2270
2271#ifdef CONFIG_NETFILTER
2272 netfilter_init();
2273#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002274
2275 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276}
2277
Andi Kleen77d76ea2005-12-22 12:43:42 -08002278core_initcall(sock_init); /* early initcall */
2279
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280#ifdef CONFIG_PROC_FS
2281void socket_seq_show(struct seq_file *seq)
2282{
2283 int cpu;
2284 int counter = 0;
2285
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002286 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002287 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288
2289 /* It can be negative, by the way. 8) */
2290 if (counter < 0)
2291 counter = 0;
2292
2293 seq_printf(seq, "sockets: used %d\n", counter);
2294}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002295#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002297#ifdef CONFIG_COMPAT
2298static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002299 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002300{
2301 struct socket *sock = file->private_data;
2302 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07002303 struct sock *sk;
2304 struct net *net;
2305
2306 sk = sock->sk;
2307 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002308
2309 if (sock->ops->compat_ioctl)
2310 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2311
David S. Miller87de87d2008-06-03 09:14:03 -07002312 if (ret == -ENOIOCTLCMD &&
2313 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2314 ret = compat_wext_handle_ioctl(net, cmd, arg);
2315
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002316 return ret;
2317}
2318#endif
2319
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002320int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2321{
2322 return sock->ops->bind(sock, addr, addrlen);
2323}
2324
2325int kernel_listen(struct socket *sock, int backlog)
2326{
2327 return sock->ops->listen(sock, backlog);
2328}
2329
2330int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2331{
2332 struct sock *sk = sock->sk;
2333 int err;
2334
2335 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2336 newsock);
2337 if (err < 0)
2338 goto done;
2339
2340 err = sock->ops->accept(sock, *newsock, flags);
2341 if (err < 0) {
2342 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07002343 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002344 goto done;
2345 }
2346
2347 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08002348 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002349
2350done:
2351 return err;
2352}
2353
2354int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09002355 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002356{
2357 return sock->ops->connect(sock, addr, addrlen, flags);
2358}
2359
2360int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2361 int *addrlen)
2362{
2363 return sock->ops->getname(sock, addr, addrlen, 0);
2364}
2365
2366int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2367 int *addrlen)
2368{
2369 return sock->ops->getname(sock, addr, addrlen, 1);
2370}
2371
2372int kernel_getsockopt(struct socket *sock, int level, int optname,
2373 char *optval, int *optlen)
2374{
2375 mm_segment_t oldfs = get_fs();
2376 int err;
2377
2378 set_fs(KERNEL_DS);
2379 if (level == SOL_SOCKET)
2380 err = sock_getsockopt(sock, level, optname, optval, optlen);
2381 else
2382 err = sock->ops->getsockopt(sock, level, optname, optval,
2383 optlen);
2384 set_fs(oldfs);
2385 return err;
2386}
2387
2388int kernel_setsockopt(struct socket *sock, int level, int optname,
2389 char *optval, int optlen)
2390{
2391 mm_segment_t oldfs = get_fs();
2392 int err;
2393
2394 set_fs(KERNEL_DS);
2395 if (level == SOL_SOCKET)
2396 err = sock_setsockopt(sock, level, optname, optval, optlen);
2397 else
2398 err = sock->ops->setsockopt(sock, level, optname, optval,
2399 optlen);
2400 set_fs(oldfs);
2401 return err;
2402}
2403
2404int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2405 size_t size, int flags)
2406{
2407 if (sock->ops->sendpage)
2408 return sock->ops->sendpage(sock, page, offset, size, flags);
2409
2410 return sock_no_sendpage(sock, page, offset, size, flags);
2411}
2412
2413int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2414{
2415 mm_segment_t oldfs = get_fs();
2416 int err;
2417
2418 set_fs(KERNEL_DS);
2419 err = sock->ops->ioctl(sock, cmd, arg);
2420 set_fs(oldfs);
2421
2422 return err;
2423}
2424
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002425int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2426{
2427 return sock->ops->shutdown(sock, how);
2428}
2429
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430EXPORT_SYMBOL(sock_create);
2431EXPORT_SYMBOL(sock_create_kern);
2432EXPORT_SYMBOL(sock_create_lite);
2433EXPORT_SYMBOL(sock_map_fd);
2434EXPORT_SYMBOL(sock_recvmsg);
2435EXPORT_SYMBOL(sock_register);
2436EXPORT_SYMBOL(sock_release);
2437EXPORT_SYMBOL(sock_sendmsg);
2438EXPORT_SYMBOL(sock_unregister);
2439EXPORT_SYMBOL(sock_wake_async);
2440EXPORT_SYMBOL(sockfd_lookup);
2441EXPORT_SYMBOL(kernel_sendmsg);
2442EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002443EXPORT_SYMBOL(kernel_bind);
2444EXPORT_SYMBOL(kernel_listen);
2445EXPORT_SYMBOL(kernel_accept);
2446EXPORT_SYMBOL(kernel_connect);
2447EXPORT_SYMBOL(kernel_getsockname);
2448EXPORT_SYMBOL(kernel_getpeername);
2449EXPORT_SYMBOL(kernel_getsockopt);
2450EXPORT_SYMBOL(kernel_setsockopt);
2451EXPORT_SYMBOL(kernel_sendpage);
2452EXPORT_SYMBOL(kernel_sock_ioctl);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002453EXPORT_SYMBOL(kernel_sock_shutdown);