blob: a0ce8ad722524b8290ffc6f046e66c9ab2365728 [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
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700267static void init_once(struct kmem_cache *cachep, 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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331static 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
371static int sock_attach_fd(struct socket *sock, struct file *file)
372{
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;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393 file->f_flags = O_RDWR;
394 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)) {
406 int err = sock_attach_fd(sock, newfile);
407
408 if (unlikely(err < 0)) {
409 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800411 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800413 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 return fd;
416}
417
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800418static struct socket *sock_from_file(struct file *file, int *err)
419{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800420 if (file->f_op == &socket_file_ops)
421 return file->private_data; /* set in sock_map_fd */
422
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800423 *err = -ENOTSOCK;
424 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800425}
426
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427/**
428 * sockfd_lookup - Go from a file number to its socket slot
429 * @fd: file handle
430 * @err: pointer to an error code return
431 *
432 * The file handle passed in is locked and the socket it is bound
433 * too is returned. If an error occurs the err pointer is overwritten
434 * with a negative errno code and NULL is returned. The function checks
435 * for both invalid handles and passing a handle which is not a socket.
436 *
437 * On a success the socket object pointer is returned.
438 */
439
440struct socket *sockfd_lookup(int fd, int *err)
441{
442 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 struct socket *sock;
444
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700445 file = fget(fd);
446 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 *err = -EBADF;
448 return NULL;
449 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700450
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451 sock = sock_from_file(file, err);
452 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 return sock;
455}
456
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800457static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
458{
459 struct file *file;
460 struct socket *sock;
461
Hua Zhong36725582006-04-19 15:25:02 -0700462 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800463 file = fget_light(fd, fput_needed);
464 if (file) {
465 sock = sock_from_file(file, err);
466 if (sock)
467 return sock;
468 fput_light(file, *fput_needed);
469 }
470 return NULL;
471}
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473/**
474 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700475 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 * Allocate a new inode and socket object. The two are bound together
477 * and initialised. The socket is then returned. If we are out of inodes
478 * NULL is returned.
479 */
480
481static struct socket *sock_alloc(void)
482{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700483 struct inode *inode;
484 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
486 inode = new_inode(sock_mnt->mnt_sb);
487 if (!inode)
488 return NULL;
489
490 sock = SOCKET_I(inode);
491
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700492 inode->i_mode = S_IFSOCK | S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 inode->i_uid = current->fsuid;
494 inode->i_gid = current->fsgid;
495
496 get_cpu_var(sockets_in_use)++;
497 put_cpu_var(sockets_in_use);
498 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
539 get_cpu_var(sockets_in_use)--;
540 put_cpu_var(sockets_in_use);
541 if (!sock->file) {
542 iput(SOCK_INODE(sock));
543 return;
544 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700545 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546}
547
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700548static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 struct msghdr *msg, size_t size)
550{
551 struct sock_iocb *si = kiocb_to_siocb(iocb);
552 int err;
553
554 si->sock = sock;
555 si->scm = NULL;
556 si->msg = msg;
557 si->size = size;
558
559 err = security_socket_sendmsg(sock, msg, size);
560 if (err)
561 return err;
562
563 return sock->ops->sendmsg(iocb, sock, msg, size);
564}
565
566int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
567{
568 struct kiocb iocb;
569 struct sock_iocb siocb;
570 int ret;
571
572 init_sync_kiocb(&iocb, NULL);
573 iocb.private = &siocb;
574 ret = __sock_sendmsg(&iocb, sock, msg, size);
575 if (-EIOCBQUEUED == ret)
576 ret = wait_on_sync_kiocb(&iocb);
577 return ret;
578}
579
580int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
581 struct kvec *vec, size_t num, size_t size)
582{
583 mm_segment_t oldfs = get_fs();
584 int result;
585
586 set_fs(KERNEL_DS);
587 /*
588 * the following is safe, since for compiler definitions of kvec and
589 * iovec are identical, yielding the same in-core layout and alignment
590 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700591 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 msg->msg_iovlen = num;
593 result = sock_sendmsg(sock, msg, size);
594 set_fs(oldfs);
595 return result;
596}
597
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700598/*
599 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
600 */
601void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
602 struct sk_buff *skb)
603{
604 ktime_t kt = skb->tstamp;
605
606 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
607 struct timeval tv;
608 /* Race occurred between timestamp enabling and packet
609 receiving. Fill in the current time for now. */
610 if (kt.tv64 == 0)
611 kt = ktime_get_real();
612 skb->tstamp = kt;
613 tv = ktime_to_timeval(kt);
614 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
615 } else {
616 struct timespec ts;
617 /* Race occurred between timestamp enabling and packet
618 receiving. Fill in the current time for now. */
619 if (kt.tv64 == 0)
620 kt = ktime_get_real();
621 skb->tstamp = kt;
622 ts = ktime_to_timespec(kt);
623 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
624 }
625}
626
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300627EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
628
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700629static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 struct msghdr *msg, size_t size, int flags)
631{
632 int err;
633 struct sock_iocb *si = kiocb_to_siocb(iocb);
634
635 si->sock = sock;
636 si->scm = NULL;
637 si->msg = msg;
638 si->size = size;
639 si->flags = flags;
640
641 err = security_socket_recvmsg(sock, msg, size, flags);
642 if (err)
643 return err;
644
645 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
646}
647
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700648int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 size_t size, int flags)
650{
651 struct kiocb iocb;
652 struct sock_iocb siocb;
653 int ret;
654
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700655 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 iocb.private = &siocb;
657 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
658 if (-EIOCBQUEUED == ret)
659 ret = wait_on_sync_kiocb(&iocb);
660 return ret;
661}
662
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700663int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
664 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665{
666 mm_segment_t oldfs = get_fs();
667 int result;
668
669 set_fs(KERNEL_DS);
670 /*
671 * the following is safe, since for compiler definitions of kvec and
672 * iovec are identical, yielding the same in-core layout and alignment
673 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700674 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 result = sock_recvmsg(sock, msg, size, flags);
676 set_fs(oldfs);
677 return result;
678}
679
680static void sock_aio_dtor(struct kiocb *iocb)
681{
682 kfree(iocb->private);
683}
684
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300685static ssize_t sock_sendpage(struct file *file, struct page *page,
686 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687{
688 struct socket *sock;
689 int flags;
690
Eric Dumazetb69aee02005-09-06 14:42:45 -0700691 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
693 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
694 if (more)
695 flags |= MSG_MORE;
696
697 return sock->ops->sendpage(sock, page, offset, size, flags);
698}
699
Jens Axboe9c55e012007-11-06 23:30:13 -0800700static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
701 struct pipe_inode_info *pipe, size_t len,
702 unsigned int flags)
703{
704 struct socket *sock = file->private_data;
705
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800706 if (unlikely(!sock->ops->splice_read))
707 return -EINVAL;
708
Jens Axboe9c55e012007-11-06 23:30:13 -0800709 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
710}
711
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800712static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700713 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800714{
715 if (!is_sync_kiocb(iocb)) {
716 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
717 if (!siocb)
718 return NULL;
719 iocb->ki_dtor = sock_aio_dtor;
720 }
721
722 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800723 iocb->private = siocb;
724 return siocb;
725}
726
727static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700728 struct file *file, const struct iovec *iov,
729 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800730{
731 struct socket *sock = file->private_data;
732 size_t size = 0;
733 int i;
734
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700735 for (i = 0; i < nr_segs; i++)
736 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800737
738 msg->msg_name = NULL;
739 msg->msg_namelen = 0;
740 msg->msg_control = NULL;
741 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700742 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800743 msg->msg_iovlen = nr_segs;
744 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
745
746 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
747}
748
Badari Pulavarty027445c2006-09-30 23:28:46 -0700749static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
750 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800751{
752 struct sock_iocb siocb, *x;
753
754 if (pos != 0)
755 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700756
757 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800758 return 0;
759
Badari Pulavarty027445c2006-09-30 23:28:46 -0700760
761 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800762 if (!x)
763 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700764 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800765}
766
767static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700768 struct file *file, const struct iovec *iov,
769 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800770{
771 struct socket *sock = file->private_data;
772 size_t size = 0;
773 int i;
774
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700775 for (i = 0; i < nr_segs; i++)
776 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800777
778 msg->msg_name = NULL;
779 msg->msg_namelen = 0;
780 msg->msg_control = NULL;
781 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700782 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800783 msg->msg_iovlen = nr_segs;
784 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
785 if (sock->type == SOCK_SEQPACKET)
786 msg->msg_flags |= MSG_EOR;
787
788 return __sock_sendmsg(iocb, sock, msg, size);
789}
790
Badari Pulavarty027445c2006-09-30 23:28:46 -0700791static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
792 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800794 struct sock_iocb siocb, *x;
795
796 if (pos != 0)
797 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700798
Badari Pulavarty027445c2006-09-30 23:28:46 -0700799 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800800 if (!x)
801 return -ENOMEM;
802
Badari Pulavarty027445c2006-09-30 23:28:46 -0700803 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804}
805
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806/*
807 * Atomic setting of ioctl hooks to avoid race
808 * with module unload.
809 */
810
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800811static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700812static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813
Eric W. Biederman881d9662007-09-17 11:56:21 -0700814void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800816 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800818 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700820
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821EXPORT_SYMBOL(brioctl_set);
822
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800823static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700824static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
Eric W. Biederman881d9662007-09-17 11:56:21 -0700826void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800828 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800830 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700832
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833EXPORT_SYMBOL(vlan_ioctl_set);
834
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800835static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700836static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700838void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800840 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800842 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700844
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845EXPORT_SYMBOL(dlci_ioctl_set);
846
847/*
848 * With an ioctl, arg may well be a user mode pointer, but we don't know
849 * what to do with it - that's up to the protocol still.
850 */
851
852static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
853{
854 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700855 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 void __user *argp = (void __user *)arg;
857 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700858 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
Eric Dumazetb69aee02005-09-06 14:42:45 -0700860 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700861 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900862 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700864 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100866#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700868 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700870#endif /* CONFIG_WIRELESS_EXT */
871 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 case FIOSETOWN:
873 case SIOCSPGRP:
874 err = -EFAULT;
875 if (get_user(pid, (int __user *)argp))
876 break;
877 err = f_setown(sock->file, pid, 1);
878 break;
879 case FIOGETOWN:
880 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700881 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700882 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 break;
884 case SIOCGIFBR:
885 case SIOCSIFBR:
886 case SIOCBRADDBR:
887 case SIOCBRDELBR:
888 err = -ENOPKG;
889 if (!br_ioctl_hook)
890 request_module("bridge");
891
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800892 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700893 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700894 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800895 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 break;
897 case SIOCGIFVLAN:
898 case SIOCSIFVLAN:
899 err = -ENOPKG;
900 if (!vlan_ioctl_hook)
901 request_module("8021q");
902
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800903 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700905 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800906 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 case SIOCADDDLCI:
909 case SIOCDELDLCI:
910 err = -ENOPKG;
911 if (!dlci_ioctl_hook)
912 request_module("dlci");
913
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700914 mutex_lock(&dlci_ioctl_mutex);
915 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700917 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 break;
919 default:
920 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800921
922 /*
923 * If this ioctl is unknown try to hand it down
924 * to the NIC driver.
925 */
926 if (err == -ENOIOCTLCMD)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700927 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700929 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 return err;
931}
932
933int sock_create_lite(int family, int type, int protocol, struct socket **res)
934{
935 int err;
936 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700937
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 err = security_socket_create(family, type, protocol, 1);
939 if (err)
940 goto out;
941
942 sock = sock_alloc();
943 if (!sock) {
944 err = -ENOMEM;
945 goto out;
946 }
947
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700949 err = security_socket_post_create(sock, family, type, protocol, 1);
950 if (err)
951 goto out_release;
952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953out:
954 *res = sock;
955 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700956out_release:
957 sock_release(sock);
958 sock = NULL;
959 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960}
961
962/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700963static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964{
965 struct socket *sock;
966
967 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700968 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700970 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 return sock->ops->poll(file, sock, wait);
972}
973
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700974static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700976 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
978 return sock->ops->mmap(file, sock, vma);
979}
980
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300981static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982{
983 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700984 * It was possible the inode is NULL we were
985 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 */
987
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700988 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 printk(KERN_DEBUG "sock_close: NULL inode\n");
990 return 0;
991 }
992 sock_fasync(-1, filp, 0);
993 sock_release(SOCKET_I(inode));
994 return 0;
995}
996
997/*
998 * Update the socket async list
999 *
1000 * Fasync_list locking strategy.
1001 *
1002 * 1. fasync_list is modified only under process context socket lock
1003 * i.e. under semaphore.
1004 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1005 * or under socket lock.
1006 * 3. fasync_list can be used from softirq context, so that
1007 * modification under socket lock have to be enhanced with
1008 * write_lock_bh(&sk->sk_callback_lock).
1009 * --ANK (990710)
1010 */
1011
1012static int sock_fasync(int fd, struct file *filp, int on)
1013{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001014 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 struct socket *sock;
1016 struct sock *sk;
1017
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001018 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001019 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001020 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 return -ENOMEM;
1022 }
1023
Eric Dumazetb69aee02005-09-06 14:42:45 -07001024 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001026 sk = sock->sk;
1027 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 kfree(fna);
1029 return -EINVAL;
1030 }
1031
1032 lock_sock(sk);
1033
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001034 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001036 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1037 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 break;
1039
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001040 if (on) {
1041 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001043 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 write_unlock_bh(&sk->sk_callback_lock);
1045
1046 kfree(fna);
1047 goto out;
1048 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001049 fna->fa_file = filp;
1050 fna->fa_fd = fd;
1051 fna->magic = FASYNC_MAGIC;
1052 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001054 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001056 } else {
1057 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001059 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 write_unlock_bh(&sk->sk_callback_lock);
1061 kfree(fa);
1062 }
1063 }
1064
1065out:
1066 release_sock(sock->sk);
1067 return 0;
1068}
1069
1070/* This function may be called only under socket lock or callback_lock */
1071
1072int sock_wake_async(struct socket *sock, int how, int band)
1073{
1074 if (!sock || !sock->fasync_list)
1075 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001076 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001077 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1079 break;
1080 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001081 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1083 break;
1084 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001085 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001086call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 __kill_fasync(sock->fasync_list, SIGIO, band);
1088 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001089 case SOCK_WAKE_URG:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 __kill_fasync(sock->fasync_list, SIGURG, band);
1091 }
1092 return 0;
1093}
1094
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001095static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001096 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097{
1098 int err;
1099 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001100 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101
1102 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001103 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 */
1105 if (family < 0 || family >= NPROTO)
1106 return -EAFNOSUPPORT;
1107 if (type < 0 || type >= SOCK_MAX)
1108 return -EINVAL;
1109
1110 /* Compatibility.
1111
1112 This uglymoron is moved from INET layer to here to avoid
1113 deadlock in module load.
1114 */
1115 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001116 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 if (!warned) {
1118 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001119 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1120 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 }
1122 family = PF_PACKET;
1123 }
1124
1125 err = security_socket_create(family, type, protocol, kern);
1126 if (err)
1127 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001128
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001129 /*
1130 * Allocate the socket and allow the family to set things up. if
1131 * the protocol is 0, the family is instructed to select an appropriate
1132 * default.
1133 */
1134 sock = sock_alloc();
1135 if (!sock) {
1136 if (net_ratelimit())
1137 printk(KERN_WARNING "socket: no more sockets\n");
1138 return -ENFILE; /* Not exactly a match, but its the
1139 closest posix thing */
1140 }
1141
1142 sock->type = type;
1143
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144#if defined(CONFIG_KMOD)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001145 /* Attempt to load a protocol module if the find failed.
1146 *
1147 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 * requested real, full-featured networking support upon configuration.
1149 * Otherwise module support will break!
1150 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001151 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001152 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153#endif
1154
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001155 rcu_read_lock();
1156 pf = rcu_dereference(net_families[family]);
1157 err = -EAFNOSUPPORT;
1158 if (!pf)
1159 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160
1161 /*
1162 * We will call the ->create function, that possibly is in a loadable
1163 * module, so we have to bump that loadable module refcnt first.
1164 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001165 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 goto out_release;
1167
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001168 /* Now protected by module ref count */
1169 rcu_read_unlock();
1170
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001171 err = pf->create(net, sock, protocol);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001172 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001174
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 /*
1176 * Now to bump the refcnt of the [loadable] module that owns this
1177 * socket at sock_release time we decrement its refcnt.
1178 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001179 if (!try_module_get(sock->ops->owner))
1180 goto out_module_busy;
1181
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 /*
1183 * Now that we're done with the ->create function, the [loadable]
1184 * module can have its refcnt decremented
1185 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001186 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001187 err = security_socket_post_create(sock, family, type, protocol, kern);
1188 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001189 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001190 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001192 return 0;
1193
1194out_module_busy:
1195 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001197 sock->ops = NULL;
1198 module_put(pf->owner);
1199out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001201 return err;
1202
1203out_release:
1204 rcu_read_unlock();
1205 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206}
1207
1208int sock_create(int family, int type, int protocol, struct socket **res)
1209{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001210 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211}
1212
1213int sock_create_kern(int family, int type, int protocol, struct socket **res)
1214{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001215 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216}
1217
1218asmlinkage long sys_socket(int family, int type, int protocol)
1219{
1220 int retval;
1221 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001222 int flags;
1223
1224 flags = type & ~SOCK_TYPE_MASK;
1225 if (flags & ~SOCK_CLOEXEC)
1226 return -EINVAL;
1227 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001229 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1230 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1231
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 retval = sock_create(family, type, protocol, &sock);
1233 if (retval < 0)
1234 goto out;
1235
Ulrich Dreppera677a032008-07-23 21:29:17 -07001236 retval = sock_map_fd(sock, flags & O_CLOEXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 if (retval < 0)
1238 goto out_release;
1239
1240out:
1241 /* It may be already another descriptor 8) Not kernel problem. */
1242 return retval;
1243
1244out_release:
1245 sock_release(sock);
1246 return retval;
1247}
1248
1249/*
1250 * Create a pair of connected sockets.
1251 */
1252
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001253asmlinkage long sys_socketpair(int family, int type, int protocol,
1254 int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255{
1256 struct socket *sock1, *sock2;
1257 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001258 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001259 int flags;
1260
1261 flags = type & ~SOCK_TYPE_MASK;
1262 if (flags & ~SOCK_CLOEXEC)
1263 return -EINVAL;
1264 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001266 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1267 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1268
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 /*
1270 * Obtain the first socket and check if the underlying protocol
1271 * supports the socketpair call.
1272 */
1273
1274 err = sock_create(family, type, protocol, &sock1);
1275 if (err < 0)
1276 goto out;
1277
1278 err = sock_create(family, type, protocol, &sock2);
1279 if (err < 0)
1280 goto out_release_1;
1281
1282 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001283 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 goto out_release_both;
1285
Ulrich Dreppera677a032008-07-23 21:29:17 -07001286 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001287 if (unlikely(fd1 < 0)) {
1288 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001290 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291
Ulrich Dreppera677a032008-07-23 21:29:17 -07001292 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
Al Virodb349502007-02-07 01:48:00 -05001293 if (unlikely(fd2 < 0)) {
David S. Millerbf3c23d2007-10-29 21:54:02 -07001294 err = fd2;
Al Virodb349502007-02-07 01:48:00 -05001295 put_filp(newfile1);
1296 put_unused_fd(fd1);
1297 goto out_release_both;
1298 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
Al Virodb349502007-02-07 01:48:00 -05001300 err = sock_attach_fd(sock1, newfile1);
1301 if (unlikely(err < 0)) {
1302 goto out_fd2;
1303 }
1304
1305 err = sock_attach_fd(sock2, newfile2);
1306 if (unlikely(err < 0)) {
1307 fput(newfile1);
1308 goto out_fd1;
1309 }
1310
1311 err = audit_fd_pair(fd1, fd2);
1312 if (err < 0) {
1313 fput(newfile1);
1314 fput(newfile2);
1315 goto out_fd;
1316 }
1317
1318 fd_install(fd1, newfile1);
1319 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 /* fd1 and fd2 may be already another descriptors.
1321 * Not kernel problem.
1322 */
1323
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001324 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 if (!err)
1326 err = put_user(fd2, &usockvec[1]);
1327 if (!err)
1328 return 0;
1329
1330 sys_close(fd2);
1331 sys_close(fd1);
1332 return err;
1333
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001335 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001337 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338out:
1339 return err;
Al Virodb349502007-02-07 01:48:00 -05001340
1341out_fd2:
1342 put_filp(newfile1);
1343 sock_release(sock1);
1344out_fd1:
1345 put_filp(newfile2);
1346 sock_release(sock2);
1347out_fd:
1348 put_unused_fd(fd1);
1349 put_unused_fd(fd2);
1350 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351}
1352
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353/*
1354 * Bind a name to a socket. Nothing much to do here since it's
1355 * the protocol's responsibility to handle the local address.
1356 *
1357 * We move the socket address to kernel space before we call
1358 * the protocol layer (having also checked the address is ok).
1359 */
1360
1361asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1362{
1363 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001364 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001365 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001367 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001368 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001369 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001370 if (err >= 0) {
1371 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001372 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001373 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001374 if (!err)
1375 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001376 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001377 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001379 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001380 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 return err;
1382}
1383
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384/*
1385 * Perform a listen. Basically, we allow the protocol to do anything
1386 * necessary for a listen, and if that works, we mark the socket as
1387 * ready for listening.
1388 */
1389
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390asmlinkage long sys_listen(int fd, int backlog)
1391{
1392 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001393 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001394 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001395
1396 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1397 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001398 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001399 if ((unsigned)backlog > somaxconn)
1400 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401
1402 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001403 if (!err)
1404 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001406 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 }
1408 return err;
1409}
1410
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411/*
1412 * For accept, we attempt to create a new socket, set up the link
1413 * with the client, wake up the client, then return the new
1414 * connected fd. We collect the address of the connector in kernel
1415 * space and move it to user at the very end. This is unclean because
1416 * we open the socket then return an error.
1417 *
1418 * 1003.1g adds the ability to recvmsg() to query connection pending
1419 * status to recvmsg. We need to add that support in a way thats
1420 * clean when we restucture accept also.
1421 */
1422
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001423long do_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1424 int __user *upeer_addrlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425{
1426 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001427 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001428 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001429 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001431 if (flags & ~SOCK_CLOEXEC)
1432 return -EINVAL;
1433
1434 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1435 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1436
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001437 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 if (!sock)
1439 goto out;
1440
1441 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001442 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 goto out_put;
1444
1445 newsock->type = sock->type;
1446 newsock->ops = sock->ops;
1447
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 /*
1449 * We don't need try_module_get here, as the listening socket (sock)
1450 * has the protocol module (sock->ops->owner) held.
1451 */
1452 __module_get(newsock->ops->owner);
1453
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001454 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001455 if (unlikely(newfd < 0)) {
1456 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001457 sock_release(newsock);
1458 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001459 }
1460
1461 err = sock_attach_fd(newsock, newfile);
1462 if (err < 0)
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001463 goto out_fd_simple;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001464
Frank Filza79af592005-09-27 15:23:38 -07001465 err = security_socket_accept(sock, newsock);
1466 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001467 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001468
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1470 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001471 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
1473 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001474 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001475 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001477 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001479 err = move_addr_to_user((struct sockaddr *)&address,
1480 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001482 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 }
1484
1485 /* File flags are not inherited via accept() unlike another OSes. */
1486
David S. Miller39d8c1b2006-03-20 17:13:49 -08001487 fd_install(newfd, newfile);
1488 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489
1490 security_socket_post_accept(sock, newsock);
1491
1492out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001493 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494out:
1495 return err;
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001496out_fd_simple:
1497 sock_release(newsock);
1498 put_filp(newfile);
1499 put_unused_fd(newfd);
1500 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001501out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001502 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001503 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 goto out_put;
1505}
1506
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001507asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1508 int __user *upeer_addrlen,
1509 const sigset_t __user *sigmask,
1510 size_t sigsetsize, int flags)
1511{
1512 sigset_t ksigmask, sigsaved;
1513 int ret;
1514
1515 if (sigmask) {
1516 /* XXX: Don't preclude handling different sized sigset_t's. */
1517 if (sigsetsize != sizeof(sigset_t))
1518 return -EINVAL;
1519 if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
1520 return -EFAULT;
1521
1522 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1523 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1524 }
1525
1526 ret = do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1527
1528 if (ret < 0 && signal_pending(current)) {
1529 /*
1530 * Don't restore the signal mask yet. Let do_signal() deliver
1531 * the signal on the way back to userspace, before the signal
1532 * mask is restored.
1533 */
1534 if (sigmask) {
1535 memcpy(&current->saved_sigmask, &sigsaved,
1536 sizeof(sigsaved));
1537 set_restore_sigmask();
1538 }
1539 } else if (sigmask)
1540 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1541
1542 return ret;
1543}
1544
1545asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1546 int __user *upeer_addrlen)
1547{
1548 return do_accept(fd, upeer_sockaddr, upeer_addrlen, 0);
1549}
1550
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551/*
1552 * Attempt to connect to a socket with the server address. The address
1553 * is in user space so we verify it is OK and move it to kernel space.
1554 *
1555 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1556 * break bindings
1557 *
1558 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1559 * other SEQPACKET protocols that take time to connect() as it doesn't
1560 * include the -EINPROGRESS status for such sockets.
1561 */
1562
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001563asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1564 int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565{
1566 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001567 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001568 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001570 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 if (!sock)
1572 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001573 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 if (err < 0)
1575 goto out_put;
1576
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001577 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001578 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 if (err)
1580 goto out_put;
1581
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001582 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 sock->file->f_flags);
1584out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001585 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586out:
1587 return err;
1588}
1589
1590/*
1591 * Get the local address ('name') of a socket object. Move the obtained
1592 * name to user space.
1593 */
1594
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001595asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1596 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597{
1598 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001599 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001600 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001601
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001602 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 if (!sock)
1604 goto out;
1605
1606 err = security_socket_getsockname(sock);
1607 if (err)
1608 goto out_put;
1609
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001610 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 if (err)
1612 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001613 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614
1615out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001616 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617out:
1618 return err;
1619}
1620
1621/*
1622 * Get the remote address ('name') of a socket object. Move the obtained
1623 * name to user space.
1624 */
1625
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001626asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1627 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628{
1629 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001630 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001631 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001633 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1634 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 err = security_socket_getpeername(sock);
1636 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001637 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 return err;
1639 }
1640
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001641 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001642 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001643 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001645 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001646 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001647 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 }
1649 return err;
1650}
1651
1652/*
1653 * Send a datagram to a given address. We move the address into kernel
1654 * space and check the user space data area is readable before invoking
1655 * the protocol.
1656 */
1657
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001658asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1659 unsigned flags, struct sockaddr __user *addr,
1660 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661{
1662 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001663 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 int err;
1665 struct msghdr msg;
1666 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001667 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001668
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001669 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1670 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001671 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001672
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001673 iov.iov_base = buff;
1674 iov.iov_len = len;
1675 msg.msg_name = NULL;
1676 msg.msg_iov = &iov;
1677 msg.msg_iovlen = 1;
1678 msg.msg_control = NULL;
1679 msg.msg_controllen = 0;
1680 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001681 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001682 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (err < 0)
1684 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001685 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001686 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 }
1688 if (sock->file->f_flags & O_NONBLOCK)
1689 flags |= MSG_DONTWAIT;
1690 msg.msg_flags = flags;
1691 err = sock_sendmsg(sock, &msg, len);
1692
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001693out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001694 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001695out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 return err;
1697}
1698
1699/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001700 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 */
1702
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001703asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704{
1705 return sys_sendto(fd, buff, len, flags, NULL, 0);
1706}
1707
1708/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001709 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 * sender. We verify the buffers are writable and if needed move the
1711 * sender address from kernel to user space.
1712 */
1713
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001714asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1715 unsigned flags, struct sockaddr __user *addr,
1716 int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717{
1718 struct socket *sock;
1719 struct iovec iov;
1720 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001721 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001722 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001723 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001725 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001727 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001729 msg.msg_control = NULL;
1730 msg.msg_controllen = 0;
1731 msg.msg_iovlen = 1;
1732 msg.msg_iov = &iov;
1733 iov.iov_len = size;
1734 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001735 msg.msg_name = (struct sockaddr *)&address;
1736 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 if (sock->file->f_flags & O_NONBLOCK)
1738 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001739 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001741 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001742 err2 = move_addr_to_user((struct sockaddr *)&address,
1743 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001744 if (err2 < 0)
1745 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001747
1748 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001749out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 return err;
1751}
1752
1753/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001754 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 */
1756
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001757asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1758 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759{
1760 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1761}
1762
1763/*
1764 * Set a socket option. Because we don't know the option lengths we have
1765 * to pass the user mode parameter for the protocols to sort out.
1766 */
1767
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001768asmlinkage long sys_setsockopt(int fd, int level, int optname,
1769 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001771 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 struct socket *sock;
1773
1774 if (optlen < 0)
1775 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001776
1777 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1778 if (sock != NULL) {
1779 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001780 if (err)
1781 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782
1783 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001784 err =
1785 sock_setsockopt(sock, level, optname, optval,
1786 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001788 err =
1789 sock->ops->setsockopt(sock, level, optname, optval,
1790 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001791out_put:
1792 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 }
1794 return err;
1795}
1796
1797/*
1798 * Get a socket option. Because we don't know the option lengths we have
1799 * to pass a user mode parameter for the protocols to sort out.
1800 */
1801
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001802asmlinkage long sys_getsockopt(int fd, int level, int optname,
1803 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001805 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 struct socket *sock;
1807
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001808 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1809 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001810 err = security_socket_getsockopt(sock, level, optname);
1811 if (err)
1812 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813
1814 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001815 err =
1816 sock_getsockopt(sock, level, optname, optval,
1817 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001819 err =
1820 sock->ops->getsockopt(sock, level, optname, optval,
1821 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001822out_put:
1823 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 }
1825 return err;
1826}
1827
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828/*
1829 * Shutdown a socket.
1830 */
1831
1832asmlinkage long sys_shutdown(int fd, int how)
1833{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001834 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 struct socket *sock;
1836
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001837 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1838 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001840 if (!err)
1841 err = sock->ops->shutdown(sock, how);
1842 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 }
1844 return err;
1845}
1846
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001847/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 * fields which are the same type (int / unsigned) on our platforms.
1849 */
1850#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1851#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1852#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1853
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854/*
1855 * BSD sendmsg interface
1856 */
1857
1858asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1859{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001860 struct compat_msghdr __user *msg_compat =
1861 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001863 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001865 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001866 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1867 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 unsigned char *ctl_buf = ctl;
1869 struct msghdr msg_sys;
1870 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001871 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001872
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 err = -EFAULT;
1874 if (MSG_CMSG_COMPAT & flags) {
1875 if (get_compat_msghdr(&msg_sys, msg_compat))
1876 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001877 }
1878 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 return -EFAULT;
1880
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001881 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001882 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 goto out;
1884
1885 /* do not move before msg_sys is valid */
1886 err = -EMSGSIZE;
1887 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1888 goto out_put;
1889
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 err = -ENOMEM;
1892 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1893 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1894 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1895 if (!iov)
1896 goto out_put;
1897 }
1898
1899 /* This will also move the address data into kernel space */
1900 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001901 err = verify_compat_iovec(&msg_sys, iov,
1902 (struct sockaddr *)&address,
1903 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001905 err = verify_iovec(&msg_sys, iov,
1906 (struct sockaddr *)&address,
1907 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 goto out_freeiov;
1910 total_len = err;
1911
1912 err = -ENOBUFS;
1913
1914 if (msg_sys.msg_controllen > INT_MAX)
1915 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001916 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001918 err =
1919 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1920 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921 if (err)
1922 goto out_freeiov;
1923 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001924 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001926 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001928 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 goto out_freeiov;
1930 }
1931 err = -EFAULT;
1932 /*
1933 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1934 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1935 * checking falls down on this.
1936 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001937 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1938 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 goto out_freectl;
1940 msg_sys.msg_control = ctl_buf;
1941 }
1942 msg_sys.msg_flags = flags;
1943
1944 if (sock->file->f_flags & O_NONBLOCK)
1945 msg_sys.msg_flags |= MSG_DONTWAIT;
1946 err = sock_sendmsg(sock, &msg_sys, total_len);
1947
1948out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001949 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1951out_freeiov:
1952 if (iov != iovstack)
1953 sock_kfree_s(sock->sk, iov, iov_size);
1954out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001955 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001956out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 return err;
1958}
1959
1960/*
1961 * BSD recvmsg interface
1962 */
1963
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001964asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1965 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001967 struct compat_msghdr __user *msg_compat =
1968 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969 struct socket *sock;
1970 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001971 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 struct msghdr msg_sys;
1973 unsigned long cmsg_ptr;
1974 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001975 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976
1977 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001978 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979
1980 /* user mode address pointers */
1981 struct sockaddr __user *uaddr;
1982 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001983
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 if (MSG_CMSG_COMPAT & flags) {
1985 if (get_compat_msghdr(&msg_sys, msg_compat))
1986 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001987 }
1988 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1989 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001991 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 if (!sock)
1993 goto out;
1994
1995 err = -EMSGSIZE;
1996 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1997 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001998
1999 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 err = -ENOMEM;
2001 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2002 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2003 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2004 if (!iov)
2005 goto out_put;
2006 }
2007
2008 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002009 * Save the user-mode address (verify_iovec will change the
2010 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002012
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -07002013 uaddr = (__force void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 uaddr_len = COMPAT_NAMELEN(msg);
2015 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002016 err = verify_compat_iovec(&msg_sys, iov,
2017 (struct sockaddr *)&addr,
2018 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002020 err = verify_iovec(&msg_sys, iov,
2021 (struct sockaddr *)&addr,
2022 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 if (err < 0)
2024 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002025 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
2027 cmsg_ptr = (unsigned long)msg_sys.msg_control;
Ulrich Drepper4a195422007-07-15 23:40:34 -07002028 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002029
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 if (sock->file->f_flags & O_NONBLOCK)
2031 flags |= MSG_DONTWAIT;
2032 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2033 if (err < 0)
2034 goto out_freeiov;
2035 len = err;
2036
2037 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002038 err = move_addr_to_user((struct sockaddr *)&addr,
2039 msg_sys.msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002040 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 if (err < 0)
2042 goto out_freeiov;
2043 }
David S. Miller37f7f422005-09-16 16:51:01 -07002044 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2045 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 if (err)
2047 goto out_freeiov;
2048 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002049 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 &msg_compat->msg_controllen);
2051 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002052 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 &msg->msg_controllen);
2054 if (err)
2055 goto out_freeiov;
2056 err = len;
2057
2058out_freeiov:
2059 if (iov != iovstack)
2060 sock_kfree_s(sock->sk, iov, iov_size);
2061out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002062 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063out:
2064 return err;
2065}
2066
2067#ifdef __ARCH_WANT_SYS_SOCKETCALL
2068
2069/* Argument list sizes for sys_socketcall */
2070#define AL(x) ((x) * sizeof(unsigned long))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002071static const unsigned char nargs[19]={
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002072 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2073 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002074 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2075 AL(6)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002076};
2077
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078#undef AL
2079
2080/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002081 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082 *
2083 * Argument checking cleaned up. Saved 20% in size.
2084 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002085 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 */
2087
2088asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2089{
2090 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002091 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 int err;
2093
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002094 if (call < 1 || call > SYS_PACCEPT)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 return -EINVAL;
2096
2097 /* copy_from_user should be SMP safe. */
2098 if (copy_from_user(a, args, nargs[call]))
2099 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002100
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002101 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002102 if (err)
2103 return err;
2104
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002105 a0 = a[0];
2106 a1 = a[1];
2107
2108 switch (call) {
2109 case SYS_SOCKET:
2110 err = sys_socket(a0, a1, a[2]);
2111 break;
2112 case SYS_BIND:
2113 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2114 break;
2115 case SYS_CONNECT:
2116 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2117 break;
2118 case SYS_LISTEN:
2119 err = sys_listen(a0, a1);
2120 break;
2121 case SYS_ACCEPT:
2122 err =
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002123 do_accept(a0, (struct sockaddr __user *)a1,
2124 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002125 break;
2126 case SYS_GETSOCKNAME:
2127 err =
2128 sys_getsockname(a0, (struct sockaddr __user *)a1,
2129 (int __user *)a[2]);
2130 break;
2131 case SYS_GETPEERNAME:
2132 err =
2133 sys_getpeername(a0, (struct sockaddr __user *)a1,
2134 (int __user *)a[2]);
2135 break;
2136 case SYS_SOCKETPAIR:
2137 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2138 break;
2139 case SYS_SEND:
2140 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2141 break;
2142 case SYS_SENDTO:
2143 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2144 (struct sockaddr __user *)a[4], a[5]);
2145 break;
2146 case SYS_RECV:
2147 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2148 break;
2149 case SYS_RECVFROM:
2150 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2151 (struct sockaddr __user *)a[4],
2152 (int __user *)a[5]);
2153 break;
2154 case SYS_SHUTDOWN:
2155 err = sys_shutdown(a0, a1);
2156 break;
2157 case SYS_SETSOCKOPT:
2158 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2159 break;
2160 case SYS_GETSOCKOPT:
2161 err =
2162 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2163 (int __user *)a[4]);
2164 break;
2165 case SYS_SENDMSG:
2166 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2167 break;
2168 case SYS_RECVMSG:
2169 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2170 break;
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002171 case SYS_PACCEPT:
2172 err =
2173 sys_paccept(a0, (struct sockaddr __user *)a1,
2174 (int __user *)a[2],
2175 (const sigset_t __user *) a[3],
2176 a[4], a[5]);
2177 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002178 default:
2179 err = -EINVAL;
2180 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 }
2182 return err;
2183}
2184
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002185#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002187/**
2188 * sock_register - add a socket protocol handler
2189 * @ops: description of protocol
2190 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191 * This function is called by a protocol handler that wants to
2192 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002193 * socket interface. The value ops->family coresponds to the
2194 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002196int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197{
2198 int err;
2199
2200 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002201 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2202 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203 return -ENOBUFS;
2204 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002205
2206 spin_lock(&net_family_lock);
2207 if (net_families[ops->family])
2208 err = -EEXIST;
2209 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002210 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 err = 0;
2212 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002213 spin_unlock(&net_family_lock);
2214
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002215 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216 return err;
2217}
2218
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002219/**
2220 * sock_unregister - remove a protocol handler
2221 * @family: protocol family to remove
2222 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 * This function is called by a protocol handler that wants to
2224 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002225 * new socket creation.
2226 *
2227 * If protocol handler is a module, then it can use module reference
2228 * counts to protect against new references. If protocol handler is not
2229 * a module then it needs to provide its own protection in
2230 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002232void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002234 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002236 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002237 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002238 spin_unlock(&net_family_lock);
2239
2240 synchronize_rcu();
2241
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002242 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243}
2244
Andi Kleen77d76ea2005-12-22 12:43:42 -08002245static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246{
2247 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002248 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002249 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002250
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251 sk_init();
2252
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002254 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255 */
2256 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257
2258 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002259 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260 */
2261
2262 init_inodecache();
2263 register_filesystem(&sock_fs_type);
2264 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002265
2266 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267 */
2268
2269#ifdef CONFIG_NETFILTER
2270 netfilter_init();
2271#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002272
2273 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274}
2275
Andi Kleen77d76ea2005-12-22 12:43:42 -08002276core_initcall(sock_init); /* early initcall */
2277
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278#ifdef CONFIG_PROC_FS
2279void socket_seq_show(struct seq_file *seq)
2280{
2281 int cpu;
2282 int counter = 0;
2283
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002284 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002285 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
2287 /* It can be negative, by the way. 8) */
2288 if (counter < 0)
2289 counter = 0;
2290
2291 seq_printf(seq, "sockets: used %d\n", counter);
2292}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002293#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002295#ifdef CONFIG_COMPAT
2296static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002297 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002298{
2299 struct socket *sock = file->private_data;
2300 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07002301 struct sock *sk;
2302 struct net *net;
2303
2304 sk = sock->sk;
2305 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002306
2307 if (sock->ops->compat_ioctl)
2308 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2309
David S. Miller87de87d2008-06-03 09:14:03 -07002310 if (ret == -ENOIOCTLCMD &&
2311 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2312 ret = compat_wext_handle_ioctl(net, cmd, arg);
2313
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002314 return ret;
2315}
2316#endif
2317
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002318int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2319{
2320 return sock->ops->bind(sock, addr, addrlen);
2321}
2322
2323int kernel_listen(struct socket *sock, int backlog)
2324{
2325 return sock->ops->listen(sock, backlog);
2326}
2327
2328int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2329{
2330 struct sock *sk = sock->sk;
2331 int err;
2332
2333 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2334 newsock);
2335 if (err < 0)
2336 goto done;
2337
2338 err = sock->ops->accept(sock, *newsock, flags);
2339 if (err < 0) {
2340 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07002341 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002342 goto done;
2343 }
2344
2345 (*newsock)->ops = sock->ops;
2346
2347done:
2348 return err;
2349}
2350
2351int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09002352 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002353{
2354 return sock->ops->connect(sock, addr, addrlen, flags);
2355}
2356
2357int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2358 int *addrlen)
2359{
2360 return sock->ops->getname(sock, addr, addrlen, 0);
2361}
2362
2363int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2364 int *addrlen)
2365{
2366 return sock->ops->getname(sock, addr, addrlen, 1);
2367}
2368
2369int kernel_getsockopt(struct socket *sock, int level, int optname,
2370 char *optval, int *optlen)
2371{
2372 mm_segment_t oldfs = get_fs();
2373 int err;
2374
2375 set_fs(KERNEL_DS);
2376 if (level == SOL_SOCKET)
2377 err = sock_getsockopt(sock, level, optname, optval, optlen);
2378 else
2379 err = sock->ops->getsockopt(sock, level, optname, optval,
2380 optlen);
2381 set_fs(oldfs);
2382 return err;
2383}
2384
2385int kernel_setsockopt(struct socket *sock, int level, int optname,
2386 char *optval, int optlen)
2387{
2388 mm_segment_t oldfs = get_fs();
2389 int err;
2390
2391 set_fs(KERNEL_DS);
2392 if (level == SOL_SOCKET)
2393 err = sock_setsockopt(sock, level, optname, optval, optlen);
2394 else
2395 err = sock->ops->setsockopt(sock, level, optname, optval,
2396 optlen);
2397 set_fs(oldfs);
2398 return err;
2399}
2400
2401int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2402 size_t size, int flags)
2403{
2404 if (sock->ops->sendpage)
2405 return sock->ops->sendpage(sock, page, offset, size, flags);
2406
2407 return sock_no_sendpage(sock, page, offset, size, flags);
2408}
2409
2410int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2411{
2412 mm_segment_t oldfs = get_fs();
2413 int err;
2414
2415 set_fs(KERNEL_DS);
2416 err = sock->ops->ioctl(sock, cmd, arg);
2417 set_fs(oldfs);
2418
2419 return err;
2420}
2421
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002422int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2423{
2424 return sock->ops->shutdown(sock, how);
2425}
2426
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427EXPORT_SYMBOL(sock_create);
2428EXPORT_SYMBOL(sock_create_kern);
2429EXPORT_SYMBOL(sock_create_lite);
2430EXPORT_SYMBOL(sock_map_fd);
2431EXPORT_SYMBOL(sock_recvmsg);
2432EXPORT_SYMBOL(sock_register);
2433EXPORT_SYMBOL(sock_release);
2434EXPORT_SYMBOL(sock_sendmsg);
2435EXPORT_SYMBOL(sock_unregister);
2436EXPORT_SYMBOL(sock_wake_async);
2437EXPORT_SYMBOL(sockfd_lookup);
2438EXPORT_SYMBOL(kernel_sendmsg);
2439EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002440EXPORT_SYMBOL(kernel_bind);
2441EXPORT_SYMBOL(kernel_listen);
2442EXPORT_SYMBOL(kernel_accept);
2443EXPORT_SYMBOL(kernel_connect);
2444EXPORT_SYMBOL(kernel_getsockname);
2445EXPORT_SYMBOL(kernel_getpeername);
2446EXPORT_SYMBOL(kernel_getsockopt);
2447EXPORT_SYMBOL(kernel_setsockopt);
2448EXPORT_SYMBOL(kernel_sendpage);
2449EXPORT_SYMBOL(kernel_sock_ioctl);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002450EXPORT_SYMBOL(kernel_sock_shutdown);