blob: 2a022c00d85c509e2346a8afac33eaa63154439a [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
Eric Dumazet29a020d2009-09-15 02:39:20 -0700492 kmemcheck_annotate_bitfield(sock, type);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700493 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100494 inode->i_uid = current_fsuid();
495 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
Eric Dumazet4e694892009-04-04 16:41:09 -0700497 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 return sock;
499}
500
501/*
502 * In theory you can't get an open on this inode, but /proc provides
503 * a back door. Remember to keep it shut otherwise you'll let the
504 * creepy crawlies in.
505 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
508{
509 return -ENXIO;
510}
511
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800512const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 .owner = THIS_MODULE,
514 .open = sock_no_open,
515};
516
517/**
518 * sock_release - close a socket
519 * @sock: socket to close
520 *
521 * The socket is released from the protocol stack if it has a release
522 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700523 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700525
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526void sock_release(struct socket *sock)
527{
528 if (sock->ops) {
529 struct module *owner = sock->ops->owner;
530
531 sock->ops->release(sock);
532 sock->ops = NULL;
533 module_put(owner);
534 }
535
536 if (sock->fasync_list)
537 printk(KERN_ERR "sock_release: fasync list not empty!\n");
538
Eric Dumazet4e694892009-04-04 16:41:09 -0700539 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 if (!sock->file) {
541 iput(SOCK_INODE(sock));
542 return;
543 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700544 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545}
546
Patrick Ohly20d49472009-02-12 05:03:38 +0000547int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
548 union skb_shared_tx *shtx)
549{
550 shtx->flags = 0;
551 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
552 shtx->hardware = 1;
553 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
554 shtx->software = 1;
555 return 0;
556}
557EXPORT_SYMBOL(sock_tx_timestamp);
558
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700559static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 struct msghdr *msg, size_t size)
561{
562 struct sock_iocb *si = kiocb_to_siocb(iocb);
563 int err;
564
565 si->sock = sock;
566 si->scm = NULL;
567 si->msg = msg;
568 si->size = size;
569
570 err = security_socket_sendmsg(sock, msg, size);
571 if (err)
572 return err;
573
574 return sock->ops->sendmsg(iocb, sock, msg, size);
575}
576
577int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
578{
579 struct kiocb iocb;
580 struct sock_iocb siocb;
581 int ret;
582
583 init_sync_kiocb(&iocb, NULL);
584 iocb.private = &siocb;
585 ret = __sock_sendmsg(&iocb, sock, msg, size);
586 if (-EIOCBQUEUED == ret)
587 ret = wait_on_sync_kiocb(&iocb);
588 return ret;
589}
590
591int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
592 struct kvec *vec, size_t num, size_t size)
593{
594 mm_segment_t oldfs = get_fs();
595 int result;
596
597 set_fs(KERNEL_DS);
598 /*
599 * the following is safe, since for compiler definitions of kvec and
600 * iovec are identical, yielding the same in-core layout and alignment
601 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700602 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 msg->msg_iovlen = num;
604 result = sock_sendmsg(sock, msg, size);
605 set_fs(oldfs);
606 return result;
607}
608
Patrick Ohly20d49472009-02-12 05:03:38 +0000609static int ktime2ts(ktime_t kt, struct timespec *ts)
610{
611 if (kt.tv64) {
612 *ts = ktime_to_timespec(kt);
613 return 1;
614 } else {
615 return 0;
616 }
617}
618
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700619/*
620 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
621 */
622void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
623 struct sk_buff *skb)
624{
Patrick Ohly20d49472009-02-12 05:03:38 +0000625 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
626 struct timespec ts[3];
627 int empty = 1;
628 struct skb_shared_hwtstamps *shhwtstamps =
629 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700630
Patrick Ohly20d49472009-02-12 05:03:38 +0000631 /* Race occurred between timestamp enabling and packet
632 receiving. Fill in the current time for now. */
633 if (need_software_tstamp && skb->tstamp.tv64 == 0)
634 __net_timestamp(skb);
635
636 if (need_software_tstamp) {
637 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
638 struct timeval tv;
639 skb_get_timestamp(skb, &tv);
640 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
641 sizeof(tv), &tv);
642 } else {
643 struct timespec ts;
644 skb_get_timestampns(skb, &ts);
645 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
646 sizeof(ts), &ts);
647 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700648 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000649
650
651 memset(ts, 0, sizeof(ts));
652 if (skb->tstamp.tv64 &&
653 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
654 skb_get_timestampns(skb, ts + 0);
655 empty = 0;
656 }
657 if (shhwtstamps) {
658 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
659 ktime2ts(shhwtstamps->syststamp, ts + 1))
660 empty = 0;
661 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
662 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
663 empty = 0;
664 }
665 if (!empty)
666 put_cmsg(msg, SOL_SOCKET,
667 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700668}
669
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300670EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
671
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700672static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 struct msghdr *msg, size_t size, int flags)
674{
675 int err;
676 struct sock_iocb *si = kiocb_to_siocb(iocb);
677
678 si->sock = sock;
679 si->scm = NULL;
680 si->msg = msg;
681 si->size = size;
682 si->flags = flags;
683
684 err = security_socket_recvmsg(sock, msg, size, flags);
685 if (err)
686 return err;
687
688 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
689}
690
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700691int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 size_t size, int flags)
693{
694 struct kiocb iocb;
695 struct sock_iocb siocb;
696 int ret;
697
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700698 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 iocb.private = &siocb;
700 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
701 if (-EIOCBQUEUED == ret)
702 ret = wait_on_sync_kiocb(&iocb);
703 return ret;
704}
705
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700706int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
707 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708{
709 mm_segment_t oldfs = get_fs();
710 int result;
711
712 set_fs(KERNEL_DS);
713 /*
714 * the following is safe, since for compiler definitions of kvec and
715 * iovec are identical, yielding the same in-core layout and alignment
716 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700717 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 result = sock_recvmsg(sock, msg, size, flags);
719 set_fs(oldfs);
720 return result;
721}
722
723static void sock_aio_dtor(struct kiocb *iocb)
724{
725 kfree(iocb->private);
726}
727
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300728static ssize_t sock_sendpage(struct file *file, struct page *page,
729 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730{
731 struct socket *sock;
732 int flags;
733
Eric Dumazetb69aee02005-09-06 14:42:45 -0700734 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735
736 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
737 if (more)
738 flags |= MSG_MORE;
739
Linus Torvaldse6949582009-08-13 08:28:36 -0700740 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741}
742
Jens Axboe9c55e012007-11-06 23:30:13 -0800743static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
744 struct pipe_inode_info *pipe, size_t len,
745 unsigned int flags)
746{
747 struct socket *sock = file->private_data;
748
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800749 if (unlikely(!sock->ops->splice_read))
750 return -EINVAL;
751
Jens Axboe9c55e012007-11-06 23:30:13 -0800752 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
753}
754
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800755static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700756 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800757{
758 if (!is_sync_kiocb(iocb)) {
759 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
760 if (!siocb)
761 return NULL;
762 iocb->ki_dtor = sock_aio_dtor;
763 }
764
765 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800766 iocb->private = siocb;
767 return siocb;
768}
769
770static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700771 struct file *file, const struct iovec *iov,
772 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800773{
774 struct socket *sock = file->private_data;
775 size_t size = 0;
776 int i;
777
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700778 for (i = 0; i < nr_segs; i++)
779 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800780
781 msg->msg_name = NULL;
782 msg->msg_namelen = 0;
783 msg->msg_control = NULL;
784 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700785 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800786 msg->msg_iovlen = nr_segs;
787 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
788
789 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
790}
791
Badari Pulavarty027445c2006-09-30 23:28:46 -0700792static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
793 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800794{
795 struct sock_iocb siocb, *x;
796
797 if (pos != 0)
798 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700799
800 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800801 return 0;
802
Badari Pulavarty027445c2006-09-30 23:28:46 -0700803
804 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800805 if (!x)
806 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700807 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800808}
809
810static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700811 struct file *file, const struct iovec *iov,
812 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800813{
814 struct socket *sock = file->private_data;
815 size_t size = 0;
816 int i;
817
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700818 for (i = 0; i < nr_segs; i++)
819 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800820
821 msg->msg_name = NULL;
822 msg->msg_namelen = 0;
823 msg->msg_control = NULL;
824 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700825 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800826 msg->msg_iovlen = nr_segs;
827 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
828 if (sock->type == SOCK_SEQPACKET)
829 msg->msg_flags |= MSG_EOR;
830
831 return __sock_sendmsg(iocb, sock, msg, size);
832}
833
Badari Pulavarty027445c2006-09-30 23:28:46 -0700834static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
835 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800837 struct sock_iocb siocb, *x;
838
839 if (pos != 0)
840 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700841
Badari Pulavarty027445c2006-09-30 23:28:46 -0700842 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800843 if (!x)
844 return -ENOMEM;
845
Badari Pulavarty027445c2006-09-30 23:28:46 -0700846 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847}
848
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849/*
850 * Atomic setting of ioctl hooks to avoid race
851 * with module unload.
852 */
853
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800854static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700855static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
Eric W. Biederman881d9662007-09-17 11:56:21 -0700857void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800859 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800861 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700863
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864EXPORT_SYMBOL(brioctl_set);
865
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800866static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700867static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
Eric W. Biederman881d9662007-09-17 11:56:21 -0700869void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800871 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800873 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700875
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876EXPORT_SYMBOL(vlan_ioctl_set);
877
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800878static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700879static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700881void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800883 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800885 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888EXPORT_SYMBOL(dlci_ioctl_set);
889
890/*
891 * With an ioctl, arg may well be a user mode pointer, but we don't know
892 * what to do with it - that's up to the protocol still.
893 */
894
895static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
896{
897 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700898 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 void __user *argp = (void __user *)arg;
900 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700901 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
Eric Dumazetb69aee02005-09-06 14:42:45 -0700903 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700904 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900905 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700907 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100909#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700911 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700913#endif /* CONFIG_WIRELESS_EXT */
914 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 case FIOSETOWN:
916 case SIOCSPGRP:
917 err = -EFAULT;
918 if (get_user(pid, (int __user *)argp))
919 break;
920 err = f_setown(sock->file, pid, 1);
921 break;
922 case FIOGETOWN:
923 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700924 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700925 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 break;
927 case SIOCGIFBR:
928 case SIOCSIFBR:
929 case SIOCBRADDBR:
930 case SIOCBRDELBR:
931 err = -ENOPKG;
932 if (!br_ioctl_hook)
933 request_module("bridge");
934
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800935 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700936 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700937 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800938 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 break;
940 case SIOCGIFVLAN:
941 case SIOCSIFVLAN:
942 err = -ENOPKG;
943 if (!vlan_ioctl_hook)
944 request_module("8021q");
945
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800946 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700948 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800949 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 case SIOCADDDLCI:
952 case SIOCDELDLCI:
953 err = -ENOPKG;
954 if (!dlci_ioctl_hook)
955 request_module("dlci");
956
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700957 mutex_lock(&dlci_ioctl_mutex);
958 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700960 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 break;
962 default:
963 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800964
965 /*
966 * If this ioctl is unknown try to hand it down
967 * to the NIC driver.
968 */
969 if (err == -ENOIOCTLCMD)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700970 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700972 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 return err;
974}
975
976int sock_create_lite(int family, int type, int protocol, struct socket **res)
977{
978 int err;
979 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700980
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 err = security_socket_create(family, type, protocol, 1);
982 if (err)
983 goto out;
984
985 sock = sock_alloc();
986 if (!sock) {
987 err = -ENOMEM;
988 goto out;
989 }
990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700992 err = security_socket_post_create(sock, family, type, protocol, 1);
993 if (err)
994 goto out_release;
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996out:
997 *res = sock;
998 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700999out_release:
1000 sock_release(sock);
1001 sock = NULL;
1002 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003}
1004
1005/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001006static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007{
1008 struct socket *sock;
1009
1010 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001011 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001013 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 return sock->ops->poll(file, sock, wait);
1015}
1016
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001017static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001019 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020
1021 return sock->ops->mmap(file, sock, vma);
1022}
1023
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001024static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025{
1026 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001027 * It was possible the inode is NULL we were
1028 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 */
1030
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001031 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 printk(KERN_DEBUG "sock_close: NULL inode\n");
1033 return 0;
1034 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 sock_release(SOCKET_I(inode));
1036 return 0;
1037}
1038
1039/*
1040 * Update the socket async list
1041 *
1042 * Fasync_list locking strategy.
1043 *
1044 * 1. fasync_list is modified only under process context socket lock
1045 * i.e. under semaphore.
1046 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1047 * or under socket lock.
1048 * 3. fasync_list can be used from softirq context, so that
1049 * modification under socket lock have to be enhanced with
1050 * write_lock_bh(&sk->sk_callback_lock).
1051 * --ANK (990710)
1052 */
1053
1054static int sock_fasync(int fd, struct file *filp, int on)
1055{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001056 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 struct socket *sock;
1058 struct sock *sk;
1059
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001060 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001061 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001062 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 return -ENOMEM;
1064 }
1065
Eric Dumazetb69aee02005-09-06 14:42:45 -07001066 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001068 sk = sock->sk;
1069 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 kfree(fna);
1071 return -EINVAL;
1072 }
1073
1074 lock_sock(sk);
1075
Jonathan Corbet76398422009-02-01 14:26:59 -07001076 spin_lock(&filp->f_lock);
1077 if (on)
1078 filp->f_flags |= FASYNC;
1079 else
1080 filp->f_flags &= ~FASYNC;
1081 spin_unlock(&filp->f_lock);
1082
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001083 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001085 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1086 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 break;
1088
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001089 if (on) {
1090 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001092 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 write_unlock_bh(&sk->sk_callback_lock);
1094
1095 kfree(fna);
1096 goto out;
1097 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001098 fna->fa_file = filp;
1099 fna->fa_fd = fd;
1100 fna->magic = FASYNC_MAGIC;
1101 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001103 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001105 } else {
1106 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001108 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 write_unlock_bh(&sk->sk_callback_lock);
1110 kfree(fa);
1111 }
1112 }
1113
1114out:
1115 release_sock(sock->sk);
1116 return 0;
1117}
1118
1119/* This function may be called only under socket lock or callback_lock */
1120
1121int sock_wake_async(struct socket *sock, int how, int band)
1122{
1123 if (!sock || !sock->fasync_list)
1124 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001125 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001126 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1128 break;
1129 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001130 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1132 break;
1133 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001134 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001135call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 __kill_fasync(sock->fasync_list, SIGIO, band);
1137 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001138 case SOCK_WAKE_URG:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 __kill_fasync(sock->fasync_list, SIGURG, band);
1140 }
1141 return 0;
1142}
1143
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001144static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001145 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146{
1147 int err;
1148 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001149 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
1151 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001152 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 */
1154 if (family < 0 || family >= NPROTO)
1155 return -EAFNOSUPPORT;
1156 if (type < 0 || type >= SOCK_MAX)
1157 return -EINVAL;
1158
1159 /* Compatibility.
1160
1161 This uglymoron is moved from INET layer to here to avoid
1162 deadlock in module load.
1163 */
1164 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001165 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 if (!warned) {
1167 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001168 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1169 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 }
1171 family = PF_PACKET;
1172 }
1173
1174 err = security_socket_create(family, type, protocol, kern);
1175 if (err)
1176 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001177
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001178 /*
1179 * Allocate the socket and allow the family to set things up. if
1180 * the protocol is 0, the family is instructed to select an appropriate
1181 * default.
1182 */
1183 sock = sock_alloc();
1184 if (!sock) {
1185 if (net_ratelimit())
1186 printk(KERN_WARNING "socket: no more sockets\n");
1187 return -ENFILE; /* Not exactly a match, but its the
1188 closest posix thing */
1189 }
1190
1191 sock->type = type;
1192
Johannes Berg95a5afc2008-10-16 15:24:51 -07001193#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001194 /* Attempt to load a protocol module if the find failed.
1195 *
1196 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 * requested real, full-featured networking support upon configuration.
1198 * Otherwise module support will break!
1199 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001200 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001201 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202#endif
1203
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001204 rcu_read_lock();
1205 pf = rcu_dereference(net_families[family]);
1206 err = -EAFNOSUPPORT;
1207 if (!pf)
1208 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
1210 /*
1211 * We will call the ->create function, that possibly is in a loadable
1212 * module, so we have to bump that loadable module refcnt first.
1213 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001214 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 goto out_release;
1216
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001217 /* Now protected by module ref count */
1218 rcu_read_unlock();
1219
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001220 err = pf->create(net, sock, protocol);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001221 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001223
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 /*
1225 * Now to bump the refcnt of the [loadable] module that owns this
1226 * socket at sock_release time we decrement its refcnt.
1227 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001228 if (!try_module_get(sock->ops->owner))
1229 goto out_module_busy;
1230
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 /*
1232 * Now that we're done with the ->create function, the [loadable]
1233 * module can have its refcnt decremented
1234 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001235 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001236 err = security_socket_post_create(sock, family, type, protocol, kern);
1237 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001238 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001239 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001241 return 0;
1242
1243out_module_busy:
1244 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001246 sock->ops = NULL;
1247 module_put(pf->owner);
1248out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001250 return err;
1251
1252out_release:
1253 rcu_read_unlock();
1254 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255}
1256
1257int sock_create(int family, int type, int protocol, struct socket **res)
1258{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001259 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260}
1261
1262int sock_create_kern(int family, int type, int protocol, struct socket **res)
1263{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001264 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265}
1266
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001267SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268{
1269 int retval;
1270 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001271 int flags;
1272
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001273 /* Check the SOCK_* constants for consistency. */
1274 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1275 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1276 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1277 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1278
Ulrich Dreppera677a032008-07-23 21:29:17 -07001279 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001280 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001281 return -EINVAL;
1282 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001284 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1285 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1286
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 retval = sock_create(family, type, protocol, &sock);
1288 if (retval < 0)
1289 goto out;
1290
Ulrich Drepper77d27202008-07-23 21:29:35 -07001291 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 if (retval < 0)
1293 goto out_release;
1294
1295out:
1296 /* It may be already another descriptor 8) Not kernel problem. */
1297 return retval;
1298
1299out_release:
1300 sock_release(sock);
1301 return retval;
1302}
1303
1304/*
1305 * Create a pair of connected sockets.
1306 */
1307
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001308SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1309 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310{
1311 struct socket *sock1, *sock2;
1312 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001313 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001314 int flags;
1315
1316 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001317 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001318 return -EINVAL;
1319 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001321 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1322 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1323
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 /*
1325 * Obtain the first socket and check if the underlying protocol
1326 * supports the socketpair call.
1327 */
1328
1329 err = sock_create(family, type, protocol, &sock1);
1330 if (err < 0)
1331 goto out;
1332
1333 err = sock_create(family, type, protocol, &sock2);
1334 if (err < 0)
1335 goto out_release_1;
1336
1337 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001338 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 goto out_release_both;
1340
Ulrich Dreppera677a032008-07-23 21:29:17 -07001341 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001342 if (unlikely(fd1 < 0)) {
1343 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
Ulrich Dreppera677a032008-07-23 21:29:17 -07001347 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
Al Virodb349502007-02-07 01:48:00 -05001348 if (unlikely(fd2 < 0)) {
David S. Millerbf3c23d2007-10-29 21:54:02 -07001349 err = fd2;
Al Virodb349502007-02-07 01:48:00 -05001350 put_filp(newfile1);
1351 put_unused_fd(fd1);
1352 goto out_release_both;
1353 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
Ulrich Drepper77d27202008-07-23 21:29:35 -07001355 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
Al Virodb349502007-02-07 01:48:00 -05001356 if (unlikely(err < 0)) {
1357 goto out_fd2;
1358 }
1359
Ulrich Drepper77d27202008-07-23 21:29:35 -07001360 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
Al Virodb349502007-02-07 01:48:00 -05001361 if (unlikely(err < 0)) {
1362 fput(newfile1);
1363 goto out_fd1;
1364 }
1365
Al Viro157cf642008-12-14 04:57:47 -05001366 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001367 fd_install(fd1, newfile1);
1368 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 /* fd1 and fd2 may be already another descriptors.
1370 * Not kernel problem.
1371 */
1372
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001373 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 if (!err)
1375 err = put_user(fd2, &usockvec[1]);
1376 if (!err)
1377 return 0;
1378
1379 sys_close(fd2);
1380 sys_close(fd1);
1381 return err;
1382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001384 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001386 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387out:
1388 return err;
Al Virodb349502007-02-07 01:48:00 -05001389
1390out_fd2:
1391 put_filp(newfile1);
1392 sock_release(sock1);
1393out_fd1:
1394 put_filp(newfile2);
1395 sock_release(sock2);
Al Virodb349502007-02-07 01:48:00 -05001396 put_unused_fd(fd1);
1397 put_unused_fd(fd2);
1398 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399}
1400
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401/*
1402 * Bind a name to a socket. Nothing much to do here since it's
1403 * the protocol's responsibility to handle the local address.
1404 *
1405 * We move the socket address to kernel space before we call
1406 * the protocol layer (having also checked the address is ok).
1407 */
1408
Heiko Carstens20f37032009-01-14 14:14:23 +01001409SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410{
1411 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001412 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001413 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001415 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001416 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001417 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001418 if (err >= 0) {
1419 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001420 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001421 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001422 if (!err)
1423 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001424 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001425 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001427 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001428 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 return err;
1430}
1431
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432/*
1433 * Perform a listen. Basically, we allow the protocol to do anything
1434 * necessary for a listen, and if that works, we mark the socket as
1435 * ready for listening.
1436 */
1437
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001438SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439{
1440 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001441 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001442 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001443
1444 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1445 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001446 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001447 if ((unsigned)backlog > somaxconn)
1448 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449
1450 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001451 if (!err)
1452 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001454 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 }
1456 return err;
1457}
1458
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459/*
1460 * For accept, we attempt to create a new socket, set up the link
1461 * with the client, wake up the client, then return the new
1462 * connected fd. We collect the address of the connector in kernel
1463 * space and move it to user at the very end. This is unclean because
1464 * we open the socket then return an error.
1465 *
1466 * 1003.1g adds the ability to recvmsg() to query connection pending
1467 * status to recvmsg. We need to add that support in a way thats
1468 * clean when we restucture accept also.
1469 */
1470
Heiko Carstens20f37032009-01-14 14:14:23 +01001471SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1472 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473{
1474 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001475 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001476 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001477 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478
Ulrich Drepper77d27202008-07-23 21:29:35 -07001479 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001480 return -EINVAL;
1481
1482 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1483 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1484
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001485 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 if (!sock)
1487 goto out;
1488
1489 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001490 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 goto out_put;
1492
1493 newsock->type = sock->type;
1494 newsock->ops = sock->ops;
1495
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 /*
1497 * We don't need try_module_get here, as the listening socket (sock)
1498 * has the protocol module (sock->ops->owner) held.
1499 */
1500 __module_get(newsock->ops->owner);
1501
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001502 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001503 if (unlikely(newfd < 0)) {
1504 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001505 sock_release(newsock);
1506 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001507 }
1508
Ulrich Drepper77d27202008-07-23 21:29:35 -07001509 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001510 if (err < 0)
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001511 goto out_fd_simple;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001512
Frank Filza79af592005-09-27 15:23:38 -07001513 err = security_socket_accept(sock, newsock);
1514 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001515 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001516
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1518 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001519 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520
1521 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001522 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001523 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001525 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001527 err = move_addr_to_user((struct sockaddr *)&address,
1528 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001530 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 }
1532
1533 /* File flags are not inherited via accept() unlike another OSes. */
1534
David S. Miller39d8c1b2006-03-20 17:13:49 -08001535 fd_install(newfd, newfile);
1536 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001539 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540out:
1541 return err;
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001542out_fd_simple:
1543 sock_release(newsock);
1544 put_filp(newfile);
1545 put_unused_fd(newfd);
1546 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001547out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001548 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001549 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 goto out_put;
1551}
1552
Heiko Carstens20f37032009-01-14 14:14:23 +01001553SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1554 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001555{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001556 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001557}
1558
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559/*
1560 * Attempt to connect to a socket with the server address. The address
1561 * is in user space so we verify it is OK and move it to kernel space.
1562 *
1563 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1564 * break bindings
1565 *
1566 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1567 * other SEQPACKET protocols that take time to connect() as it doesn't
1568 * include the -EINPROGRESS status for such sockets.
1569 */
1570
Heiko Carstens20f37032009-01-14 14:14:23 +01001571SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1572 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573{
1574 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001575 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001576 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001578 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 if (!sock)
1580 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001581 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 if (err < 0)
1583 goto out_put;
1584
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001585 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001586 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 if (err)
1588 goto out_put;
1589
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001590 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 sock->file->f_flags);
1592out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001593 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594out:
1595 return err;
1596}
1597
1598/*
1599 * Get the local address ('name') of a socket object. Move the obtained
1600 * name to user space.
1601 */
1602
Heiko Carstens20f37032009-01-14 14:14:23 +01001603SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1604 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605{
1606 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001607 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001608 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001609
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001610 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 if (!sock)
1612 goto out;
1613
1614 err = security_socket_getsockname(sock);
1615 if (err)
1616 goto out_put;
1617
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001618 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 if (err)
1620 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001621 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
1623out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001624 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625out:
1626 return err;
1627}
1628
1629/*
1630 * Get the remote address ('name') of a socket object. Move the obtained
1631 * name to user space.
1632 */
1633
Heiko Carstens20f37032009-01-14 14:14:23 +01001634SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1635 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636{
1637 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001638 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001639 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001641 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1642 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 err = security_socket_getpeername(sock);
1644 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001645 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 return err;
1647 }
1648
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001649 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001650 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001651 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001653 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001654 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001655 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 }
1657 return err;
1658}
1659
1660/*
1661 * Send a datagram to a given address. We move the address into kernel
1662 * space and check the user space data area is readable before invoking
1663 * the protocol.
1664 */
1665
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001666SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1667 unsigned, flags, struct sockaddr __user *, addr,
1668 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669{
1670 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001671 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 int err;
1673 struct msghdr msg;
1674 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001675 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001676
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001677 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1678 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001679 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001680
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001681 iov.iov_base = buff;
1682 iov.iov_len = len;
1683 msg.msg_name = NULL;
1684 msg.msg_iov = &iov;
1685 msg.msg_iovlen = 1;
1686 msg.msg_control = NULL;
1687 msg.msg_controllen = 0;
1688 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001689 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001690 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 if (err < 0)
1692 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001693 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001694 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 }
1696 if (sock->file->f_flags & O_NONBLOCK)
1697 flags |= MSG_DONTWAIT;
1698 msg.msg_flags = flags;
1699 err = sock_sendmsg(sock, &msg, len);
1700
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001701out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001702 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001703out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 return err;
1705}
1706
1707/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001708 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 */
1710
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001711SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1712 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713{
1714 return sys_sendto(fd, buff, len, flags, NULL, 0);
1715}
1716
1717/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001718 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 * sender. We verify the buffers are writable and if needed move the
1720 * sender address from kernel to user space.
1721 */
1722
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001723SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1724 unsigned, flags, struct sockaddr __user *, addr,
1725 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726{
1727 struct socket *sock;
1728 struct iovec iov;
1729 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001730 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001731 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001732 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001734 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001736 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001738 msg.msg_control = NULL;
1739 msg.msg_controllen = 0;
1740 msg.msg_iovlen = 1;
1741 msg.msg_iov = &iov;
1742 iov.iov_len = size;
1743 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001744 msg.msg_name = (struct sockaddr *)&address;
1745 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (sock->file->f_flags & O_NONBLOCK)
1747 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001748 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001750 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001751 err2 = move_addr_to_user((struct sockaddr *)&address,
1752 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001753 if (err2 < 0)
1754 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001756
1757 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001758out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 return err;
1760}
1761
1762/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001763 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 */
1765
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001766asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1767 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768{
1769 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1770}
1771
1772/*
1773 * Set a socket option. Because we don't know the option lengths we have
1774 * to pass the user mode parameter for the protocols to sort out.
1775 */
1776
Heiko Carstens20f37032009-01-14 14:14:23 +01001777SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1778 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001780 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 struct socket *sock;
1782
1783 if (optlen < 0)
1784 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001785
1786 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1787 if (sock != NULL) {
1788 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001789 if (err)
1790 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
1792 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001793 err =
1794 sock_setsockopt(sock, level, optname, optval,
1795 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797 err =
1798 sock->ops->setsockopt(sock, level, optname, optval,
1799 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001800out_put:
1801 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 }
1803 return err;
1804}
1805
1806/*
1807 * Get a socket option. Because we don't know the option lengths we have
1808 * to pass a user mode parameter for the protocols to sort out.
1809 */
1810
Heiko Carstens20f37032009-01-14 14:14:23 +01001811SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1812 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001814 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 struct socket *sock;
1816
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1818 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001819 err = security_socket_getsockopt(sock, level, optname);
1820 if (err)
1821 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
1823 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001824 err =
1825 sock_getsockopt(sock, level, optname, optval,
1826 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 err =
1829 sock->ops->getsockopt(sock, level, optname, optval,
1830 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001831out_put:
1832 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 }
1834 return err;
1835}
1836
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837/*
1838 * Shutdown a socket.
1839 */
1840
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001841SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001843 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 struct socket *sock;
1845
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001846 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1847 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001849 if (!err)
1850 err = sock->ops->shutdown(sock, how);
1851 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 }
1853 return err;
1854}
1855
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001856/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 * fields which are the same type (int / unsigned) on our platforms.
1858 */
1859#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1860#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1861#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1862
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863/*
1864 * BSD sendmsg interface
1865 */
1866
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001867SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001869 struct compat_msghdr __user *msg_compat =
1870 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001872 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001874 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001875 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1876 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877 unsigned char *ctl_buf = ctl;
1878 struct msghdr msg_sys;
1879 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001880 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001881
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 err = -EFAULT;
1883 if (MSG_CMSG_COMPAT & flags) {
1884 if (get_compat_msghdr(&msg_sys, msg_compat))
1885 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001886 }
1887 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 return -EFAULT;
1889
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001890 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001891 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 goto out;
1893
1894 /* do not move before msg_sys is valid */
1895 err = -EMSGSIZE;
1896 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1897 goto out_put;
1898
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001899 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 err = -ENOMEM;
1901 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1902 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1903 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1904 if (!iov)
1905 goto out_put;
1906 }
1907
1908 /* This will also move the address data into kernel space */
1909 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001910 err = verify_compat_iovec(&msg_sys, iov,
1911 (struct sockaddr *)&address,
1912 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001914 err = verify_iovec(&msg_sys, iov,
1915 (struct sockaddr *)&address,
1916 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001917 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 goto out_freeiov;
1919 total_len = err;
1920
1921 err = -ENOBUFS;
1922
1923 if (msg_sys.msg_controllen > INT_MAX)
1924 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001925 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001927 err =
1928 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1929 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 if (err)
1931 goto out_freeiov;
1932 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001933 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001935 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001937 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 goto out_freeiov;
1939 }
1940 err = -EFAULT;
1941 /*
1942 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1943 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1944 * checking falls down on this.
1945 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001946 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1947 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 goto out_freectl;
1949 msg_sys.msg_control = ctl_buf;
1950 }
1951 msg_sys.msg_flags = flags;
1952
1953 if (sock->file->f_flags & O_NONBLOCK)
1954 msg_sys.msg_flags |= MSG_DONTWAIT;
1955 err = sock_sendmsg(sock, &msg_sys, total_len);
1956
1957out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001958 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1960out_freeiov:
1961 if (iov != iovstack)
1962 sock_kfree_s(sock->sk, iov, iov_size);
1963out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001964 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001965out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 return err;
1967}
1968
1969/*
1970 * BSD recvmsg interface
1971 */
1972
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001973SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
1974 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001976 struct compat_msghdr __user *msg_compat =
1977 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 struct socket *sock;
1979 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001980 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 struct msghdr msg_sys;
1982 unsigned long cmsg_ptr;
1983 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001984 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985
1986 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001987 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988
1989 /* user mode address pointers */
1990 struct sockaddr __user *uaddr;
1991 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001992
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 if (MSG_CMSG_COMPAT & flags) {
1994 if (get_compat_msghdr(&msg_sys, msg_compat))
1995 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001996 }
1997 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1998 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002000 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 if (!sock)
2002 goto out;
2003
2004 err = -EMSGSIZE;
2005 if (msg_sys.msg_iovlen > UIO_MAXIOV)
2006 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002007
2008 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 err = -ENOMEM;
2010 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2011 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2012 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2013 if (!iov)
2014 goto out_put;
2015 }
2016
2017 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002018 * Save the user-mode address (verify_iovec will change the
2019 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002021
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -07002022 uaddr = (__force void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 uaddr_len = COMPAT_NAMELEN(msg);
2024 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002025 err = verify_compat_iovec(&msg_sys, iov,
2026 (struct sockaddr *)&addr,
2027 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002029 err = verify_iovec(&msg_sys, iov,
2030 (struct sockaddr *)&addr,
2031 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 if (err < 0)
2033 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002034 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035
2036 cmsg_ptr = (unsigned long)msg_sys.msg_control;
Ulrich Drepper4a195422007-07-15 23:40:34 -07002037 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002038
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 if (sock->file->f_flags & O_NONBLOCK)
2040 flags |= MSG_DONTWAIT;
2041 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2042 if (err < 0)
2043 goto out_freeiov;
2044 len = err;
2045
2046 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002047 err = move_addr_to_user((struct sockaddr *)&addr,
2048 msg_sys.msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002049 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 if (err < 0)
2051 goto out_freeiov;
2052 }
David S. Miller37f7f422005-09-16 16:51:01 -07002053 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2054 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 if (err)
2056 goto out_freeiov;
2057 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002058 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 &msg_compat->msg_controllen);
2060 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002061 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062 &msg->msg_controllen);
2063 if (err)
2064 goto out_freeiov;
2065 err = len;
2066
2067out_freeiov:
2068 if (iov != iovstack)
2069 sock_kfree_s(sock->sk, iov, iov_size);
2070out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002071 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072out:
2073 return err;
2074}
2075
2076#ifdef __ARCH_WANT_SYS_SOCKETCALL
2077
2078/* Argument list sizes for sys_socketcall */
2079#define AL(x) ((x) * sizeof(unsigned long))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002080static const unsigned char nargs[19]={
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002081 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2082 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002083 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
Ulrich Drepperde11def2008-11-19 15:36:14 -08002084 AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002085};
2086
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087#undef AL
2088
2089/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002090 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 *
2092 * Argument checking cleaned up. Saved 20% in size.
2093 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002094 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 */
2096
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002097SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098{
2099 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002100 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 int err;
2102
Ulrich Drepperde11def2008-11-19 15:36:14 -08002103 if (call < 1 || call > SYS_ACCEPT4)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 return -EINVAL;
2105
2106 /* copy_from_user should be SMP safe. */
2107 if (copy_from_user(a, args, nargs[call]))
2108 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002109
Al Virof3298dc2008-12-10 03:16:51 -05002110 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002111
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002112 a0 = a[0];
2113 a1 = a[1];
2114
2115 switch (call) {
2116 case SYS_SOCKET:
2117 err = sys_socket(a0, a1, a[2]);
2118 break;
2119 case SYS_BIND:
2120 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2121 break;
2122 case SYS_CONNECT:
2123 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2124 break;
2125 case SYS_LISTEN:
2126 err = sys_listen(a0, a1);
2127 break;
2128 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002129 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2130 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002131 break;
2132 case SYS_GETSOCKNAME:
2133 err =
2134 sys_getsockname(a0, (struct sockaddr __user *)a1,
2135 (int __user *)a[2]);
2136 break;
2137 case SYS_GETPEERNAME:
2138 err =
2139 sys_getpeername(a0, (struct sockaddr __user *)a1,
2140 (int __user *)a[2]);
2141 break;
2142 case SYS_SOCKETPAIR:
2143 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2144 break;
2145 case SYS_SEND:
2146 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2147 break;
2148 case SYS_SENDTO:
2149 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2150 (struct sockaddr __user *)a[4], a[5]);
2151 break;
2152 case SYS_RECV:
2153 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2154 break;
2155 case SYS_RECVFROM:
2156 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2157 (struct sockaddr __user *)a[4],
2158 (int __user *)a[5]);
2159 break;
2160 case SYS_SHUTDOWN:
2161 err = sys_shutdown(a0, a1);
2162 break;
2163 case SYS_SETSOCKOPT:
2164 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2165 break;
2166 case SYS_GETSOCKOPT:
2167 err =
2168 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2169 (int __user *)a[4]);
2170 break;
2171 case SYS_SENDMSG:
2172 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2173 break;
2174 case SYS_RECVMSG:
2175 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2176 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002177 case SYS_ACCEPT4:
2178 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2179 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002180 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002181 default:
2182 err = -EINVAL;
2183 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 }
2185 return err;
2186}
2187
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002188#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002190/**
2191 * sock_register - add a socket protocol handler
2192 * @ops: description of protocol
2193 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 * This function is called by a protocol handler that wants to
2195 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002196 * socket interface. The value ops->family coresponds to the
2197 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002199int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200{
2201 int err;
2202
2203 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002204 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2205 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206 return -ENOBUFS;
2207 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002208
2209 spin_lock(&net_family_lock);
2210 if (net_families[ops->family])
2211 err = -EEXIST;
2212 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002213 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 err = 0;
2215 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002216 spin_unlock(&net_family_lock);
2217
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002218 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 return err;
2220}
2221
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002222/**
2223 * sock_unregister - remove a protocol handler
2224 * @family: protocol family to remove
2225 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226 * This function is called by a protocol handler that wants to
2227 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002228 * new socket creation.
2229 *
2230 * If protocol handler is a module, then it can use module reference
2231 * counts to protect against new references. If protocol handler is not
2232 * a module then it needs to provide its own protection in
2233 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002235void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002237 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002239 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002240 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002241 spin_unlock(&net_family_lock);
2242
2243 synchronize_rcu();
2244
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002245 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246}
2247
Andi Kleen77d76ea2005-12-22 12:43:42 -08002248static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002249{
2250 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002251 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002253
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254 sk_init();
2255
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002257 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258 */
2259 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260
2261 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002262 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263 */
2264
2265 init_inodecache();
2266 register_filesystem(&sock_fs_type);
2267 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002268
2269 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270 */
2271
2272#ifdef CONFIG_NETFILTER
2273 netfilter_init();
2274#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002275
2276 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277}
2278
Andi Kleen77d76ea2005-12-22 12:43:42 -08002279core_initcall(sock_init); /* early initcall */
2280
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281#ifdef CONFIG_PROC_FS
2282void socket_seq_show(struct seq_file *seq)
2283{
2284 int cpu;
2285 int counter = 0;
2286
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002287 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002288 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289
2290 /* It can be negative, by the way. 8) */
2291 if (counter < 0)
2292 counter = 0;
2293
2294 seq_printf(seq, "sockets: used %d\n", counter);
2295}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002296#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002298#ifdef CONFIG_COMPAT
2299static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002300 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002301{
2302 struct socket *sock = file->private_data;
2303 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07002304 struct sock *sk;
2305 struct net *net;
2306
2307 sk = sock->sk;
2308 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002309
2310 if (sock->ops->compat_ioctl)
2311 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2312
David S. Miller87de87d2008-06-03 09:14:03 -07002313 if (ret == -ENOIOCTLCMD &&
2314 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2315 ret = compat_wext_handle_ioctl(net, cmd, arg);
2316
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002317 return ret;
2318}
2319#endif
2320
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002321int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2322{
2323 return sock->ops->bind(sock, addr, addrlen);
2324}
2325
2326int kernel_listen(struct socket *sock, int backlog)
2327{
2328 return sock->ops->listen(sock, backlog);
2329}
2330
2331int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2332{
2333 struct sock *sk = sock->sk;
2334 int err;
2335
2336 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2337 newsock);
2338 if (err < 0)
2339 goto done;
2340
2341 err = sock->ops->accept(sock, *newsock, flags);
2342 if (err < 0) {
2343 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07002344 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002345 goto done;
2346 }
2347
2348 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08002349 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002350
2351done:
2352 return err;
2353}
2354
2355int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09002356 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002357{
2358 return sock->ops->connect(sock, addr, addrlen, flags);
2359}
2360
2361int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2362 int *addrlen)
2363{
2364 return sock->ops->getname(sock, addr, addrlen, 0);
2365}
2366
2367int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2368 int *addrlen)
2369{
2370 return sock->ops->getname(sock, addr, addrlen, 1);
2371}
2372
2373int kernel_getsockopt(struct socket *sock, int level, int optname,
2374 char *optval, int *optlen)
2375{
2376 mm_segment_t oldfs = get_fs();
2377 int err;
2378
2379 set_fs(KERNEL_DS);
2380 if (level == SOL_SOCKET)
2381 err = sock_getsockopt(sock, level, optname, optval, optlen);
2382 else
2383 err = sock->ops->getsockopt(sock, level, optname, optval,
2384 optlen);
2385 set_fs(oldfs);
2386 return err;
2387}
2388
2389int kernel_setsockopt(struct socket *sock, int level, int optname,
2390 char *optval, int optlen)
2391{
2392 mm_segment_t oldfs = get_fs();
2393 int err;
2394
2395 set_fs(KERNEL_DS);
2396 if (level == SOL_SOCKET)
2397 err = sock_setsockopt(sock, level, optname, optval, optlen);
2398 else
2399 err = sock->ops->setsockopt(sock, level, optname, optval,
2400 optlen);
2401 set_fs(oldfs);
2402 return err;
2403}
2404
2405int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2406 size_t size, int flags)
2407{
2408 if (sock->ops->sendpage)
2409 return sock->ops->sendpage(sock, page, offset, size, flags);
2410
2411 return sock_no_sendpage(sock, page, offset, size, flags);
2412}
2413
2414int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2415{
2416 mm_segment_t oldfs = get_fs();
2417 int err;
2418
2419 set_fs(KERNEL_DS);
2420 err = sock->ops->ioctl(sock, cmd, arg);
2421 set_fs(oldfs);
2422
2423 return err;
2424}
2425
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002426int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2427{
2428 return sock->ops->shutdown(sock, how);
2429}
2430
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431EXPORT_SYMBOL(sock_create);
2432EXPORT_SYMBOL(sock_create_kern);
2433EXPORT_SYMBOL(sock_create_lite);
2434EXPORT_SYMBOL(sock_map_fd);
2435EXPORT_SYMBOL(sock_recvmsg);
2436EXPORT_SYMBOL(sock_register);
2437EXPORT_SYMBOL(sock_release);
2438EXPORT_SYMBOL(sock_sendmsg);
2439EXPORT_SYMBOL(sock_unregister);
2440EXPORT_SYMBOL(sock_wake_async);
2441EXPORT_SYMBOL(sockfd_lookup);
2442EXPORT_SYMBOL(kernel_sendmsg);
2443EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002444EXPORT_SYMBOL(kernel_bind);
2445EXPORT_SYMBOL(kernel_listen);
2446EXPORT_SYMBOL(kernel_accept);
2447EXPORT_SYMBOL(kernel_connect);
2448EXPORT_SYMBOL(kernel_getsockname);
2449EXPORT_SYMBOL(kernel_getpeername);
2450EXPORT_SYMBOL(kernel_getsockopt);
2451EXPORT_SYMBOL(kernel_setsockopt);
2452EXPORT_SYMBOL(kernel_sendpage);
2453EXPORT_SYMBOL(kernel_sock_ioctl);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002454EXPORT_SYMBOL(kernel_sock_shutdown);