blob: 7651de0085028dc4d02b15e5844e239dd938d01e [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>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070066#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/netdevice.h>
68#include <linux/proc_fs.h>
69#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080070#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/wanrouter.h>
72#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030073#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <linux/init.h>
76#include <linux/poll.h>
77#include <linux/cache.h>
78#include <linux/module.h>
79#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#include <linux/mount.h>
81#include <linux/security.h>
82#include <linux/syscalls.h>
83#include <linux/compat.h>
84#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010085#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010086#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070087#include <linux/nsproxy.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070088
89#include <asm/uaccess.h>
90#include <asm/unistd.h>
91
92#include <net/compat.h>
93
94#include <net/sock.h>
95#include <linux/netfilter.h>
96
97static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -070098static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
99 unsigned long nr_segs, loff_t pos);
100static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
101 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700102static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
104static int sock_close(struct inode *inode, struct file *file);
105static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700107static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800108#ifdef CONFIG_COMPAT
109static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700110 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800111#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113static ssize_t sock_sendpage(struct file *file, struct page *page,
114 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800115static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
116 struct pipe_inode_info *pipe, size_t len,
117 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119/*
120 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
121 * in the operation structures but are done directly via the socketcall() multiplexor.
122 */
123
Arjan van de Venda7071d2007-02-12 00:55:36 -0800124static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125 .owner = THIS_MODULE,
126 .llseek = no_llseek,
127 .aio_read = sock_aio_read,
128 .aio_write = sock_aio_write,
129 .poll = sock_poll,
130 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800131#ifdef CONFIG_COMPAT
132 .compat_ioctl = compat_sock_ioctl,
133#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 .mmap = sock_mmap,
135 .open = sock_no_open, /* special open code to disallow open via /proc */
136 .release = sock_close,
137 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200138 .sendpage = sock_sendpage,
139 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800140 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141};
142
143/*
144 * The protocol list. Each protocol is registered in here.
145 */
146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700148static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150/*
151 * Statistics counters of the socket lists
152 */
153
154static DEFINE_PER_CPU(int, sockets_in_use) = 0;
155
156/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700157 * Support routines.
158 * Move socket addresses back and forth across the kernel/user
159 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 */
161
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700162#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 16 for IP, 16 for IPX,
164 24 for IPv6,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700165 about 80 for AX.25
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 must be at least one bigger than
167 the AF_UNIX size (see net/unix/af_unix.c
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700168 :unix_mkname()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700170
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171/**
172 * move_addr_to_kernel - copy a socket address into kernel space
173 * @uaddr: Address in user space
174 * @kaddr: Address in kernel space
175 * @ulen: Length in user space
176 *
177 * The address is copied into kernel space. If the provided address is
178 * too long an error code of -EINVAL is returned. If the copy gives
179 * invalid addresses -EFAULT is returned. On a success 0 is returned.
180 */
181
182int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
183{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700184 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700186 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700188 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100190 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191}
192
193/**
194 * move_addr_to_user - copy an address to user space
195 * @kaddr: kernel space address
196 * @klen: length of address in kernel
197 * @uaddr: user space address
198 * @ulen: pointer to user length field
199 *
200 * The value pointed to by ulen on entry is the buffer length available.
201 * This is overwritten with the buffer space used. -EINVAL is returned
202 * if an overlong buffer is specified or a negative buffer size. -EFAULT
203 * is returned if either the buffer or the length field are not
204 * accessible.
205 * After copying the data up to the limit the user specifies, the true
206 * length of the data is written over the length limit the user
207 * specified. Zero is returned for a success.
208 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700209
210int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
211 int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700212{
213 int err;
214 int len;
215
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700216 err = get_user(len, ulen);
217 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700219 if (len > klen)
220 len = klen;
221 if (len < 0 || len > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700223 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500224 if (audit_sockaddr(klen, kaddr))
225 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700226 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 return -EFAULT;
228 }
229 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700230 * "fromlen shall refer to the value before truncation.."
231 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 */
233 return __put_user(klen, ulen);
234}
235
236#define SOCKFS_MAGIC 0x534F434B
237
Christoph Lametere18b8902006-12-06 20:33:20 -0800238static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700243
Christoph Lametere94b1762006-12-06 20:33:17 -0800244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 if (!ei)
246 return NULL;
247 init_waitqueue_head(&ei->socket.wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 ei->socket.fasync_list = NULL;
250 ei->socket.state = SS_UNCONNECTED;
251 ei->socket.flags = 0;
252 ei->socket.ops = NULL;
253 ei->socket.sk = NULL;
254 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
256 return &ei->vfs_inode;
257}
258
259static void sock_destroy_inode(struct inode *inode)
260{
261 kmem_cache_free(sock_inode_cachep,
262 container_of(inode, struct socket_alloc, vfs_inode));
263}
264
Christoph Lameter4ba9b9d2007-10-16 23:25:51 -0700265static void init_once(struct kmem_cache *cachep, void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700267 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Christoph Lametera35afb82007-05-16 22:10:57 -0700269 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272static int init_inodecache(void)
273{
274 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700275 sizeof(struct socket_alloc),
276 0,
277 (SLAB_HWCACHE_ALIGN |
278 SLAB_RECLAIM_ACCOUNT |
279 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900280 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 if (sock_inode_cachep == NULL)
282 return -ENOMEM;
283 return 0;
284}
285
286static struct super_operations sockfs_ops = {
287 .alloc_inode = sock_alloc_inode,
288 .destroy_inode =sock_destroy_inode,
289 .statfs = simple_statfs,
290};
291
David Howells454e2392006-06-23 02:02:57 -0700292static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700293 int flags, const char *dev_name, void *data,
294 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295{
David Howells454e2392006-06-23 02:02:57 -0700296 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
297 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298}
299
Eric Dumazetba899662005-08-26 12:05:31 -0700300static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301
302static struct file_system_type sock_fs_type = {
303 .name = "sockfs",
304 .get_sb = sockfs_get_sb,
305 .kill_sb = kill_anon_super,
306};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700307
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308static int sockfs_delete_dentry(struct dentry *dentry)
309{
Eric Dumazet304e61e2006-12-06 20:38:49 -0800310 /*
311 * At creation time, we pretended this dentry was hashed
312 * (by clearing DCACHE_UNHASHED bit in d_flags)
313 * At delete time, we restore the truth : not hashed.
314 * (so that dput() can proceed correctly)
315 */
316 dentry->d_flags |= DCACHE_UNHASHED;
317 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318}
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700319
320/*
321 * sockfs_dname() is called from d_path().
322 */
323static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
324{
325 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
326 dentry->d_inode->i_ino);
327}
328
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329static struct dentry_operations sockfs_dentry_operations = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700330 .d_delete = sockfs_delete_dentry,
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700331 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332};
333
334/*
335 * Obtains the first available file descriptor and sets it up for use.
336 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800337 * These functions create file structures and maps them to fd space
338 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 * and file struct implicitly stored in sock->file.
340 * Note that another thread may close file descriptor before we return
341 * from this function. We use the fact that now we do not refer
342 * to socket after mapping. If one day we will need it, this
343 * function will increment ref. count on file by 1.
344 *
345 * In any case returned fd MAY BE not valid!
346 * This race condition is unavoidable
347 * with shared fd spaces, we cannot solve it inside kernel,
348 * but we take care of internal coherence yet.
349 */
350
David S. Miller39d8c1b2006-03-20 17:13:49 -0800351static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352{
353 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800354
355 fd = get_unused_fd();
356 if (likely(fd >= 0)) {
357 struct file *file = get_empty_filp();
358
359 *filep = file;
360 if (unlikely(!file)) {
361 put_unused_fd(fd);
362 return -ENFILE;
363 }
364 } else
365 *filep = NULL;
366 return fd;
367}
368
369static int sock_attach_fd(struct socket *sock, struct file *file)
370{
Dave Hansence8d2cd2007-10-16 23:31:13 -0700371 struct dentry *dentry;
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700372 struct qstr name = { .name = "" };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373
Dave Hansence8d2cd2007-10-16 23:31:13 -0700374 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
375 if (unlikely(!dentry))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800376 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377
Dave Hansence8d2cd2007-10-16 23:31:13 -0700378 dentry->d_op = &sockfs_dentry_operations;
Eric Dumazet304e61e2006-12-06 20:38:49 -0800379 /*
380 * We dont want to push this dentry into global dentry hash table.
381 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
382 * This permits a working /proc/$pid/fd/XXX on sockets
383 */
Dave Hansence8d2cd2007-10-16 23:31:13 -0700384 dentry->d_flags &= ~DCACHE_UNHASHED;
385 d_instantiate(dentry, SOCK_INODE(sock));
David S. Miller39d8c1b2006-03-20 17:13:49 -0800386
387 sock->file = file;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700388 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
389 &socket_file_ops);
390 SOCK_INODE(sock)->i_fop = &socket_file_ops;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800391 file->f_flags = O_RDWR;
392 file->f_pos = 0;
393 file->private_data = sock;
394
395 return 0;
396}
397
398int sock_map_fd(struct socket *sock)
399{
400 struct file *newfile;
401 int fd = sock_alloc_fd(&newfile);
402
403 if (likely(fd >= 0)) {
404 int err = sock_attach_fd(sock, newfile);
405
406 if (unlikely(err < 0)) {
407 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800409 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800411 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 return fd;
414}
415
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800416static struct socket *sock_from_file(struct file *file, int *err)
417{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800418 if (file->f_op == &socket_file_ops)
419 return file->private_data; /* set in sock_map_fd */
420
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800421 *err = -ENOTSOCK;
422 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800423}
424
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425/**
426 * sockfd_lookup - Go from a file number to its socket slot
427 * @fd: file handle
428 * @err: pointer to an error code return
429 *
430 * The file handle passed in is locked and the socket it is bound
431 * too is returned. If an error occurs the err pointer is overwritten
432 * with a negative errno code and NULL is returned. The function checks
433 * for both invalid handles and passing a handle which is not a socket.
434 *
435 * On a success the socket object pointer is returned.
436 */
437
438struct socket *sockfd_lookup(int fd, int *err)
439{
440 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 struct socket *sock;
442
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700443 file = fget(fd);
444 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 *err = -EBADF;
446 return NULL;
447 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700448
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800449 sock = sock_from_file(file, err);
450 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 return sock;
453}
454
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800455static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
456{
457 struct file *file;
458 struct socket *sock;
459
Hua Zhong36725582006-04-19 15:25:02 -0700460 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800461 file = fget_light(fd, fput_needed);
462 if (file) {
463 sock = sock_from_file(file, err);
464 if (sock)
465 return sock;
466 fput_light(file, *fput_needed);
467 }
468 return NULL;
469}
470
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471/**
472 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700473 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 * Allocate a new inode and socket object. The two are bound together
475 * and initialised. The socket is then returned. If we are out of inodes
476 * NULL is returned.
477 */
478
479static struct socket *sock_alloc(void)
480{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700481 struct inode *inode;
482 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
484 inode = new_inode(sock_mnt->mnt_sb);
485 if (!inode)
486 return NULL;
487
488 sock = SOCKET_I(inode);
489
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700490 inode->i_mode = S_IFSOCK | S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 inode->i_uid = current->fsuid;
492 inode->i_gid = current->fsgid;
493
494 get_cpu_var(sockets_in_use)++;
495 put_cpu_var(sockets_in_use);
496 return sock;
497}
498
499/*
500 * In theory you can't get an open on this inode, but /proc provides
501 * a back door. Remember to keep it shut otherwise you'll let the
502 * creepy crawlies in.
503 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
506{
507 return -ENXIO;
508}
509
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800510const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 .owner = THIS_MODULE,
512 .open = sock_no_open,
513};
514
515/**
516 * sock_release - close a socket
517 * @sock: socket to close
518 *
519 * The socket is released from the protocol stack if it has a release
520 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700521 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700523
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524void sock_release(struct socket *sock)
525{
526 if (sock->ops) {
527 struct module *owner = sock->ops->owner;
528
529 sock->ops->release(sock);
530 sock->ops = NULL;
531 module_put(owner);
532 }
533
534 if (sock->fasync_list)
535 printk(KERN_ERR "sock_release: fasync list not empty!\n");
536
537 get_cpu_var(sockets_in_use)--;
538 put_cpu_var(sockets_in_use);
539 if (!sock->file) {
540 iput(SOCK_INODE(sock));
541 return;
542 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700543 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544}
545
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700546static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 struct msghdr *msg, size_t size)
548{
549 struct sock_iocb *si = kiocb_to_siocb(iocb);
550 int err;
551
552 si->sock = sock;
553 si->scm = NULL;
554 si->msg = msg;
555 si->size = size;
556
557 err = security_socket_sendmsg(sock, msg, size);
558 if (err)
559 return err;
560
561 return sock->ops->sendmsg(iocb, sock, msg, size);
562}
563
564int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
565{
566 struct kiocb iocb;
567 struct sock_iocb siocb;
568 int ret;
569
570 init_sync_kiocb(&iocb, NULL);
571 iocb.private = &siocb;
572 ret = __sock_sendmsg(&iocb, sock, msg, size);
573 if (-EIOCBQUEUED == ret)
574 ret = wait_on_sync_kiocb(&iocb);
575 return ret;
576}
577
578int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
579 struct kvec *vec, size_t num, size_t size)
580{
581 mm_segment_t oldfs = get_fs();
582 int result;
583
584 set_fs(KERNEL_DS);
585 /*
586 * the following is safe, since for compiler definitions of kvec and
587 * iovec are identical, yielding the same in-core layout and alignment
588 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700589 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 msg->msg_iovlen = num;
591 result = sock_sendmsg(sock, msg, size);
592 set_fs(oldfs);
593 return result;
594}
595
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700596/*
597 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
598 */
599void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
600 struct sk_buff *skb)
601{
602 ktime_t kt = skb->tstamp;
603
604 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
605 struct timeval tv;
606 /* Race occurred between timestamp enabling and packet
607 receiving. Fill in the current time for now. */
608 if (kt.tv64 == 0)
609 kt = ktime_get_real();
610 skb->tstamp = kt;
611 tv = ktime_to_timeval(kt);
612 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
613 } else {
614 struct timespec ts;
615 /* Race occurred between timestamp enabling and packet
616 receiving. Fill in the current time for now. */
617 if (kt.tv64 == 0)
618 kt = ktime_get_real();
619 skb->tstamp = kt;
620 ts = ktime_to_timespec(kt);
621 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
622 }
623}
624
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300625EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
626
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700627static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 struct msghdr *msg, size_t size, int flags)
629{
630 int err;
631 struct sock_iocb *si = kiocb_to_siocb(iocb);
632
633 si->sock = sock;
634 si->scm = NULL;
635 si->msg = msg;
636 si->size = size;
637 si->flags = flags;
638
639 err = security_socket_recvmsg(sock, msg, size, flags);
640 if (err)
641 return err;
642
643 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
644}
645
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700646int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 size_t size, int flags)
648{
649 struct kiocb iocb;
650 struct sock_iocb siocb;
651 int ret;
652
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700653 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 iocb.private = &siocb;
655 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
656 if (-EIOCBQUEUED == ret)
657 ret = wait_on_sync_kiocb(&iocb);
658 return ret;
659}
660
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700661int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
662 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663{
664 mm_segment_t oldfs = get_fs();
665 int result;
666
667 set_fs(KERNEL_DS);
668 /*
669 * the following is safe, since for compiler definitions of kvec and
670 * iovec are identical, yielding the same in-core layout and alignment
671 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700672 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 result = sock_recvmsg(sock, msg, size, flags);
674 set_fs(oldfs);
675 return result;
676}
677
678static void sock_aio_dtor(struct kiocb *iocb)
679{
680 kfree(iocb->private);
681}
682
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300683static ssize_t sock_sendpage(struct file *file, struct page *page,
684 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685{
686 struct socket *sock;
687 int flags;
688
Eric Dumazetb69aee02005-09-06 14:42:45 -0700689 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
691 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
692 if (more)
693 flags |= MSG_MORE;
694
695 return sock->ops->sendpage(sock, page, offset, size, flags);
696}
697
Jens Axboe9c55e012007-11-06 23:30:13 -0800698static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
699 struct pipe_inode_info *pipe, size_t len,
700 unsigned int flags)
701{
702 struct socket *sock = file->private_data;
703
704 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
705}
706
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800707static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700708 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800709{
710 if (!is_sync_kiocb(iocb)) {
711 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
712 if (!siocb)
713 return NULL;
714 iocb->ki_dtor = sock_aio_dtor;
715 }
716
717 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800718 iocb->private = siocb;
719 return siocb;
720}
721
722static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700723 struct file *file, const struct iovec *iov,
724 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800725{
726 struct socket *sock = file->private_data;
727 size_t size = 0;
728 int i;
729
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700730 for (i = 0; i < nr_segs; i++)
731 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800732
733 msg->msg_name = NULL;
734 msg->msg_namelen = 0;
735 msg->msg_control = NULL;
736 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700737 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800738 msg->msg_iovlen = nr_segs;
739 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
740
741 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
742}
743
Badari Pulavarty027445c2006-09-30 23:28:46 -0700744static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
745 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800746{
747 struct sock_iocb siocb, *x;
748
749 if (pos != 0)
750 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700751
752 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800753 return 0;
754
Badari Pulavarty027445c2006-09-30 23:28:46 -0700755
756 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800757 if (!x)
758 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700759 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800760}
761
762static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700763 struct file *file, const struct iovec *iov,
764 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800765{
766 struct socket *sock = file->private_data;
767 size_t size = 0;
768 int i;
769
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700770 for (i = 0; i < nr_segs; i++)
771 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800772
773 msg->msg_name = NULL;
774 msg->msg_namelen = 0;
775 msg->msg_control = NULL;
776 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700777 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800778 msg->msg_iovlen = nr_segs;
779 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
780 if (sock->type == SOCK_SEQPACKET)
781 msg->msg_flags |= MSG_EOR;
782
783 return __sock_sendmsg(iocb, sock, msg, size);
784}
785
Badari Pulavarty027445c2006-09-30 23:28:46 -0700786static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
787 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800789 struct sock_iocb siocb, *x;
790
791 if (pos != 0)
792 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700793
Badari Pulavarty027445c2006-09-30 23:28:46 -0700794 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800795 if (!x)
796 return -ENOMEM;
797
Badari Pulavarty027445c2006-09-30 23:28:46 -0700798 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
800
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801/*
802 * Atomic setting of ioctl hooks to avoid race
803 * with module unload.
804 */
805
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800806static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700807static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Eric W. Biederman881d9662007-09-17 11:56:21 -0700809void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800811 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800813 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700815
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816EXPORT_SYMBOL(brioctl_set);
817
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800818static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700819static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
Eric W. Biederman881d9662007-09-17 11:56:21 -0700821void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800823 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800825 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700827
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828EXPORT_SYMBOL(vlan_ioctl_set);
829
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800830static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700831static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700833void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800835 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800837 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700839
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840EXPORT_SYMBOL(dlci_ioctl_set);
841
842/*
843 * With an ioctl, arg may well be a user mode pointer, but we don't know
844 * what to do with it - that's up to the protocol still.
845 */
846
847static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
848{
849 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700850 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 void __user *argp = (void __user *)arg;
852 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700853 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854
Eric Dumazetb69aee02005-09-06 14:42:45 -0700855 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700856 sk = sock->sk;
857 net = sk->sk_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700859 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100861#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700863 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700865#endif /* CONFIG_WIRELESS_EXT */
866 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 case FIOSETOWN:
868 case SIOCSPGRP:
869 err = -EFAULT;
870 if (get_user(pid, (int __user *)argp))
871 break;
872 err = f_setown(sock->file, pid, 1);
873 break;
874 case FIOGETOWN:
875 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700876 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700877 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 break;
879 case SIOCGIFBR:
880 case SIOCSIFBR:
881 case SIOCBRADDBR:
882 case SIOCBRDELBR:
883 err = -ENOPKG;
884 if (!br_ioctl_hook)
885 request_module("bridge");
886
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800887 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700888 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700889 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800890 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 break;
892 case SIOCGIFVLAN:
893 case SIOCSIFVLAN:
894 err = -ENOPKG;
895 if (!vlan_ioctl_hook)
896 request_module("8021q");
897
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800898 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700900 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800901 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 case SIOCADDDLCI:
904 case SIOCDELDLCI:
905 err = -ENOPKG;
906 if (!dlci_ioctl_hook)
907 request_module("dlci");
908
909 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800910 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800912 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 }
914 break;
915 default:
916 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800917
918 /*
919 * If this ioctl is unknown try to hand it down
920 * to the NIC driver.
921 */
922 if (err == -ENOIOCTLCMD)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700923 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700925 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 return err;
927}
928
929int sock_create_lite(int family, int type, int protocol, struct socket **res)
930{
931 int err;
932 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700933
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 err = security_socket_create(family, type, protocol, 1);
935 if (err)
936 goto out;
937
938 sock = sock_alloc();
939 if (!sock) {
940 err = -ENOMEM;
941 goto out;
942 }
943
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700945 err = security_socket_post_create(sock, family, type, protocol, 1);
946 if (err)
947 goto out_release;
948
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949out:
950 *res = sock;
951 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700952out_release:
953 sock_release(sock);
954 sock = NULL;
955 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956}
957
958/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700959static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960{
961 struct socket *sock;
962
963 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700964 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700966 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 return sock->ops->poll(file, sock, wait);
968}
969
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700970static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700972 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973
974 return sock->ops->mmap(file, sock, vma);
975}
976
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300977static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978{
979 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700980 * It was possible the inode is NULL we were
981 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 */
983
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700984 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 printk(KERN_DEBUG "sock_close: NULL inode\n");
986 return 0;
987 }
988 sock_fasync(-1, filp, 0);
989 sock_release(SOCKET_I(inode));
990 return 0;
991}
992
993/*
994 * Update the socket async list
995 *
996 * Fasync_list locking strategy.
997 *
998 * 1. fasync_list is modified only under process context socket lock
999 * i.e. under semaphore.
1000 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1001 * or under socket lock.
1002 * 3. fasync_list can be used from softirq context, so that
1003 * modification under socket lock have to be enhanced with
1004 * write_lock_bh(&sk->sk_callback_lock).
1005 * --ANK (990710)
1006 */
1007
1008static int sock_fasync(int fd, struct file *filp, int on)
1009{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001010 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 struct socket *sock;
1012 struct sock *sk;
1013
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001014 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001015 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001016 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 return -ENOMEM;
1018 }
1019
Eric Dumazetb69aee02005-09-06 14:42:45 -07001020 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001022 sk = sock->sk;
1023 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 kfree(fna);
1025 return -EINVAL;
1026 }
1027
1028 lock_sock(sk);
1029
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001030 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001032 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1033 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 break;
1035
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001036 if (on) {
1037 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001039 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 write_unlock_bh(&sk->sk_callback_lock);
1041
1042 kfree(fna);
1043 goto out;
1044 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001045 fna->fa_file = filp;
1046 fna->fa_fd = fd;
1047 fna->magic = FASYNC_MAGIC;
1048 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001050 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001052 } else {
1053 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001055 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 write_unlock_bh(&sk->sk_callback_lock);
1057 kfree(fa);
1058 }
1059 }
1060
1061out:
1062 release_sock(sock->sk);
1063 return 0;
1064}
1065
1066/* This function may be called only under socket lock or callback_lock */
1067
1068int sock_wake_async(struct socket *sock, int how, int band)
1069{
1070 if (!sock || !sock->fasync_list)
1071 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001072 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001073 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1075 break;
1076 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001077 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1079 break;
1080 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001081 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001082call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 __kill_fasync(sock->fasync_list, SIGIO, band);
1084 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001085 case SOCK_WAKE_URG:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 __kill_fasync(sock->fasync_list, SIGURG, band);
1087 }
1088 return 0;
1089}
1090
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001091static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001092 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093{
1094 int err;
1095 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001096 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
1098 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001099 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 */
1101 if (family < 0 || family >= NPROTO)
1102 return -EAFNOSUPPORT;
1103 if (type < 0 || type >= SOCK_MAX)
1104 return -EINVAL;
1105
1106 /* Compatibility.
1107
1108 This uglymoron is moved from INET layer to here to avoid
1109 deadlock in module load.
1110 */
1111 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001112 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 if (!warned) {
1114 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001115 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1116 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 }
1118 family = PF_PACKET;
1119 }
1120
1121 err = security_socket_create(family, type, protocol, kern);
1122 if (err)
1123 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001124
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001125 /*
1126 * Allocate the socket and allow the family to set things up. if
1127 * the protocol is 0, the family is instructed to select an appropriate
1128 * default.
1129 */
1130 sock = sock_alloc();
1131 if (!sock) {
1132 if (net_ratelimit())
1133 printk(KERN_WARNING "socket: no more sockets\n");
1134 return -ENFILE; /* Not exactly a match, but its the
1135 closest posix thing */
1136 }
1137
1138 sock->type = type;
1139
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140#if defined(CONFIG_KMOD)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001141 /* Attempt to load a protocol module if the find failed.
1142 *
1143 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 * requested real, full-featured networking support upon configuration.
1145 * Otherwise module support will break!
1146 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001147 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001148 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149#endif
1150
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001151 rcu_read_lock();
1152 pf = rcu_dereference(net_families[family]);
1153 err = -EAFNOSUPPORT;
1154 if (!pf)
1155 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
1157 /*
1158 * We will call the ->create function, that possibly is in a loadable
1159 * module, so we have to bump that loadable module refcnt first.
1160 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001161 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 goto out_release;
1163
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001164 /* Now protected by module ref count */
1165 rcu_read_unlock();
1166
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001167 err = pf->create(net, sock, protocol);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001168 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001170
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 /*
1172 * Now to bump the refcnt of the [loadable] module that owns this
1173 * socket at sock_release time we decrement its refcnt.
1174 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001175 if (!try_module_get(sock->ops->owner))
1176 goto out_module_busy;
1177
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 /*
1179 * Now that we're done with the ->create function, the [loadable]
1180 * module can have its refcnt decremented
1181 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001182 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001183 err = security_socket_post_create(sock, family, type, protocol, kern);
1184 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001185 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001186 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001188 return 0;
1189
1190out_module_busy:
1191 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001193 sock->ops = NULL;
1194 module_put(pf->owner);
1195out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001197 return err;
1198
1199out_release:
1200 rcu_read_unlock();
1201 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202}
1203
1204int sock_create(int family, int type, int protocol, struct socket **res)
1205{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001206 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207}
1208
1209int sock_create_kern(int family, int type, int protocol, struct socket **res)
1210{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001211 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212}
1213
1214asmlinkage long sys_socket(int family, int type, int protocol)
1215{
1216 int retval;
1217 struct socket *sock;
1218
1219 retval = sock_create(family, type, protocol, &sock);
1220 if (retval < 0)
1221 goto out;
1222
1223 retval = sock_map_fd(sock);
1224 if (retval < 0)
1225 goto out_release;
1226
1227out:
1228 /* It may be already another descriptor 8) Not kernel problem. */
1229 return retval;
1230
1231out_release:
1232 sock_release(sock);
1233 return retval;
1234}
1235
1236/*
1237 * Create a pair of connected sockets.
1238 */
1239
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001240asmlinkage long sys_socketpair(int family, int type, int protocol,
1241 int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242{
1243 struct socket *sock1, *sock2;
1244 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001245 struct file *newfile1, *newfile2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
1247 /*
1248 * Obtain the first socket and check if the underlying protocol
1249 * supports the socketpair call.
1250 */
1251
1252 err = sock_create(family, type, protocol, &sock1);
1253 if (err < 0)
1254 goto out;
1255
1256 err = sock_create(family, type, protocol, &sock2);
1257 if (err < 0)
1258 goto out_release_1;
1259
1260 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001261 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 goto out_release_both;
1263
Al Virodb349502007-02-07 01:48:00 -05001264 fd1 = sock_alloc_fd(&newfile1);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001265 if (unlikely(fd1 < 0)) {
1266 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001268 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269
Al Virodb349502007-02-07 01:48:00 -05001270 fd2 = sock_alloc_fd(&newfile2);
1271 if (unlikely(fd2 < 0)) {
David S. Millerbf3c23d2007-10-29 21:54:02 -07001272 err = fd2;
Al Virodb349502007-02-07 01:48:00 -05001273 put_filp(newfile1);
1274 put_unused_fd(fd1);
1275 goto out_release_both;
1276 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
Al Virodb349502007-02-07 01:48:00 -05001278 err = sock_attach_fd(sock1, newfile1);
1279 if (unlikely(err < 0)) {
1280 goto out_fd2;
1281 }
1282
1283 err = sock_attach_fd(sock2, newfile2);
1284 if (unlikely(err < 0)) {
1285 fput(newfile1);
1286 goto out_fd1;
1287 }
1288
1289 err = audit_fd_pair(fd1, fd2);
1290 if (err < 0) {
1291 fput(newfile1);
1292 fput(newfile2);
1293 goto out_fd;
1294 }
1295
1296 fd_install(fd1, newfile1);
1297 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 /* fd1 and fd2 may be already another descriptors.
1299 * Not kernel problem.
1300 */
1301
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001302 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 if (!err)
1304 err = put_user(fd2, &usockvec[1]);
1305 if (!err)
1306 return 0;
1307
1308 sys_close(fd2);
1309 sys_close(fd1);
1310 return err;
1311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001313 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001315 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316out:
1317 return err;
Al Virodb349502007-02-07 01:48:00 -05001318
1319out_fd2:
1320 put_filp(newfile1);
1321 sock_release(sock1);
1322out_fd1:
1323 put_filp(newfile2);
1324 sock_release(sock2);
1325out_fd:
1326 put_unused_fd(fd1);
1327 put_unused_fd(fd2);
1328 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329}
1330
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331/*
1332 * Bind a name to a socket. Nothing much to do here since it's
1333 * the protocol's responsibility to handle the local address.
1334 *
1335 * We move the socket address to kernel space before we call
1336 * the protocol layer (having also checked the address is ok).
1337 */
1338
1339asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1340{
1341 struct socket *sock;
1342 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001343 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001345 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001346 if (sock) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001347 err = move_addr_to_kernel(umyaddr, addrlen, address);
1348 if (err >= 0) {
1349 err = security_socket_bind(sock,
1350 (struct sockaddr *)address,
1351 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001352 if (!err)
1353 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001354 (struct sockaddr *)
1355 address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001357 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001358 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 return err;
1360}
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362/*
1363 * Perform a listen. Basically, we allow the protocol to do anything
1364 * necessary for a listen, and if that works, we mark the socket as
1365 * ready for listening.
1366 */
1367
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368asmlinkage long sys_listen(int fd, int backlog)
1369{
1370 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001371 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001372 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001373
1374 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1375 if (sock) {
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001376 somaxconn = sock->sk->sk_net->sysctl_somaxconn;
1377 if ((unsigned)backlog > somaxconn)
1378 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379
1380 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001381 if (!err)
1382 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001384 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 }
1386 return err;
1387}
1388
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389/*
1390 * For accept, we attempt to create a new socket, set up the link
1391 * with the client, wake up the client, then return the new
1392 * connected fd. We collect the address of the connector in kernel
1393 * space and move it to user at the very end. This is unclean because
1394 * we open the socket then return an error.
1395 *
1396 * 1003.1g adds the ability to recvmsg() to query connection pending
1397 * status to recvmsg. We need to add that support in a way thats
1398 * clean when we restucture accept also.
1399 */
1400
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001401asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1402 int __user *upeer_addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403{
1404 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001405 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001406 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 char address[MAX_SOCK_ADDR];
1408
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001409 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 if (!sock)
1411 goto out;
1412
1413 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001414 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 goto out_put;
1416
1417 newsock->type = sock->type;
1418 newsock->ops = sock->ops;
1419
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 /*
1421 * We don't need try_module_get here, as the listening socket (sock)
1422 * has the protocol module (sock->ops->owner) held.
1423 */
1424 __module_get(newsock->ops->owner);
1425
David S. Miller39d8c1b2006-03-20 17:13:49 -08001426 newfd = sock_alloc_fd(&newfile);
1427 if (unlikely(newfd < 0)) {
1428 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001429 sock_release(newsock);
1430 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001431 }
1432
1433 err = sock_attach_fd(newsock, newfile);
1434 if (err < 0)
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001435 goto out_fd_simple;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001436
Frank Filza79af592005-09-27 15:23:38 -07001437 err = security_socket_accept(sock, newsock);
1438 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001439 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001440
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1442 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001443 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
1445 if (upeer_sockaddr) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001446 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1447 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001449 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001451 err = move_addr_to_user(address, len, upeer_sockaddr,
1452 upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001454 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 }
1456
1457 /* File flags are not inherited via accept() unlike another OSes. */
1458
David S. Miller39d8c1b2006-03-20 17:13:49 -08001459 fd_install(newfd, newfile);
1460 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461
1462 security_socket_post_accept(sock, newsock);
1463
1464out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001465 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466out:
1467 return err;
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001468out_fd_simple:
1469 sock_release(newsock);
1470 put_filp(newfile);
1471 put_unused_fd(newfd);
1472 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001473out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001474 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001475 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 goto out_put;
1477}
1478
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479/*
1480 * Attempt to connect to a socket with the server address. The address
1481 * is in user space so we verify it is OK and move it to kernel space.
1482 *
1483 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1484 * break bindings
1485 *
1486 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1487 * other SEQPACKET protocols that take time to connect() as it doesn't
1488 * include the -EINPROGRESS status for such sockets.
1489 */
1490
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001491asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1492 int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493{
1494 struct socket *sock;
1495 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001496 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001498 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 if (!sock)
1500 goto out;
1501 err = move_addr_to_kernel(uservaddr, addrlen, address);
1502 if (err < 0)
1503 goto out_put;
1504
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001505 err =
1506 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 if (err)
1508 goto out_put;
1509
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001510 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 sock->file->f_flags);
1512out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001513 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514out:
1515 return err;
1516}
1517
1518/*
1519 * Get the local address ('name') of a socket object. Move the obtained
1520 * name to user space.
1521 */
1522
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001523asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1524 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525{
1526 struct socket *sock;
1527 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001528 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001529
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001530 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 if (!sock)
1532 goto out;
1533
1534 err = security_socket_getsockname(sock);
1535 if (err)
1536 goto out_put;
1537
1538 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1539 if (err)
1540 goto out_put;
1541 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1542
1543out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001544 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545out:
1546 return err;
1547}
1548
1549/*
1550 * Get the remote address ('name') of a socket object. Move the obtained
1551 * name to user space.
1552 */
1553
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001554asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1555 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556{
1557 struct socket *sock;
1558 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001559 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001561 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1562 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 err = security_socket_getpeername(sock);
1564 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001565 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 return err;
1567 }
1568
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001569 err =
1570 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1571 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572 if (!err)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001573 err = move_addr_to_user(address, len, usockaddr,
1574 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001575 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 }
1577 return err;
1578}
1579
1580/*
1581 * Send a datagram to a given address. We move the address into kernel
1582 * space and check the user space data area is readable before invoking
1583 * the protocol.
1584 */
1585
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001586asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1587 unsigned flags, struct sockaddr __user *addr,
1588 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589{
1590 struct socket *sock;
1591 char address[MAX_SOCK_ADDR];
1592 int err;
1593 struct msghdr msg;
1594 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001595 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001596
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001597 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1598 if (!sock)
David S. Miller4387ff752007-02-08 15:06:08 -08001599 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001600
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001601 iov.iov_base = buff;
1602 iov.iov_len = len;
1603 msg.msg_name = NULL;
1604 msg.msg_iov = &iov;
1605 msg.msg_iovlen = 1;
1606 msg.msg_control = NULL;
1607 msg.msg_controllen = 0;
1608 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001609 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 err = move_addr_to_kernel(addr, addr_len, address);
1611 if (err < 0)
1612 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001613 msg.msg_name = address;
1614 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615 }
1616 if (sock->file->f_flags & O_NONBLOCK)
1617 flags |= MSG_DONTWAIT;
1618 msg.msg_flags = flags;
1619 err = sock_sendmsg(sock, &msg, len);
1620
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001621out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001622 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08001623out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 return err;
1625}
1626
1627/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001628 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 */
1630
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001631asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632{
1633 return sys_sendto(fd, buff, len, flags, NULL, 0);
1634}
1635
1636/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001637 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 * sender. We verify the buffers are writable and if needed move the
1639 * sender address from kernel to user space.
1640 */
1641
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001642asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1643 unsigned flags, struct sockaddr __user *addr,
1644 int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645{
1646 struct socket *sock;
1647 struct iovec iov;
1648 struct msghdr msg;
1649 char address[MAX_SOCK_ADDR];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001650 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001651 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001653 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001655 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001657 msg.msg_control = NULL;
1658 msg.msg_controllen = 0;
1659 msg.msg_iovlen = 1;
1660 msg.msg_iov = &iov;
1661 iov.iov_len = size;
1662 iov.iov_base = ubuf;
1663 msg.msg_name = address;
1664 msg.msg_namelen = MAX_SOCK_ADDR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 if (sock->file->f_flags & O_NONBLOCK)
1666 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001667 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001669 if (err >= 0 && addr != NULL) {
1670 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1671 if (err2 < 0)
1672 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001674
1675 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08001676out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 return err;
1678}
1679
1680/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001681 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 */
1683
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001684asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1685 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686{
1687 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1688}
1689
1690/*
1691 * Set a socket option. Because we don't know the option lengths we have
1692 * to pass the user mode parameter for the protocols to sort out.
1693 */
1694
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001695asmlinkage long sys_setsockopt(int fd, int level, int optname,
1696 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001698 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 struct socket *sock;
1700
1701 if (optlen < 0)
1702 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001703
1704 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1705 if (sock != NULL) {
1706 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001707 if (err)
1708 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709
1710 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001711 err =
1712 sock_setsockopt(sock, level, optname, optval,
1713 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001715 err =
1716 sock->ops->setsockopt(sock, level, optname, optval,
1717 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001718out_put:
1719 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 }
1721 return err;
1722}
1723
1724/*
1725 * Get a socket option. Because we don't know the option lengths we have
1726 * to pass a user mode parameter for the protocols to sort out.
1727 */
1728
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001729asmlinkage long sys_getsockopt(int fd, int level, int optname,
1730 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001732 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 struct socket *sock;
1734
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001735 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1736 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001737 err = security_socket_getsockopt(sock, level, optname);
1738 if (err)
1739 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
1741 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001742 err =
1743 sock_getsockopt(sock, level, optname, optval,
1744 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001746 err =
1747 sock->ops->getsockopt(sock, level, optname, optval,
1748 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001749out_put:
1750 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 }
1752 return err;
1753}
1754
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755/*
1756 * Shutdown a socket.
1757 */
1758
1759asmlinkage long sys_shutdown(int fd, int how)
1760{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001761 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 struct socket *sock;
1763
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001764 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1765 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001767 if (!err)
1768 err = sock->ops->shutdown(sock, how);
1769 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 }
1771 return err;
1772}
1773
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001774/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 * fields which are the same type (int / unsigned) on our platforms.
1776 */
1777#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1778#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1779#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1780
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781/*
1782 * BSD sendmsg interface
1783 */
1784
1785asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1786{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001787 struct compat_msghdr __user *msg_compat =
1788 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789 struct socket *sock;
1790 char address[MAX_SOCK_ADDR];
1791 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001792 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001793 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1794 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 unsigned char *ctl_buf = ctl;
1796 struct msghdr msg_sys;
1797 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001798 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001799
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 err = -EFAULT;
1801 if (MSG_CMSG_COMPAT & flags) {
1802 if (get_compat_msghdr(&msg_sys, msg_compat))
1803 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001804 }
1805 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 return -EFAULT;
1807
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001808 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001809 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 goto out;
1811
1812 /* do not move before msg_sys is valid */
1813 err = -EMSGSIZE;
1814 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1815 goto out_put;
1816
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001817 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 err = -ENOMEM;
1819 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1820 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1821 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1822 if (!iov)
1823 goto out_put;
1824 }
1825
1826 /* This will also move the address data into kernel space */
1827 if (MSG_CMSG_COMPAT & flags) {
1828 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1829 } else
1830 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001831 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 goto out_freeiov;
1833 total_len = err;
1834
1835 err = -ENOBUFS;
1836
1837 if (msg_sys.msg_controllen > INT_MAX)
1838 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001839 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001841 err =
1842 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1843 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 if (err)
1845 goto out_freeiov;
1846 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001847 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001849 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001851 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 goto out_freeiov;
1853 }
1854 err = -EFAULT;
1855 /*
1856 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1857 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1858 * checking falls down on this.
1859 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001860 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1861 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862 goto out_freectl;
1863 msg_sys.msg_control = ctl_buf;
1864 }
1865 msg_sys.msg_flags = flags;
1866
1867 if (sock->file->f_flags & O_NONBLOCK)
1868 msg_sys.msg_flags |= MSG_DONTWAIT;
1869 err = sock_sendmsg(sock, &msg_sys, total_len);
1870
1871out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001872 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1874out_freeiov:
1875 if (iov != iovstack)
1876 sock_kfree_s(sock->sk, iov, iov_size);
1877out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001878 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001879out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 return err;
1881}
1882
1883/*
1884 * BSD recvmsg interface
1885 */
1886
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001887asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1888 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890 struct compat_msghdr __user *msg_compat =
1891 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 struct socket *sock;
1893 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001894 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 struct msghdr msg_sys;
1896 unsigned long cmsg_ptr;
1897 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001898 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899
1900 /* kernel mode address */
1901 char addr[MAX_SOCK_ADDR];
1902
1903 /* user mode address pointers */
1904 struct sockaddr __user *uaddr;
1905 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001906
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 if (MSG_CMSG_COMPAT & flags) {
1908 if (get_compat_msghdr(&msg_sys, msg_compat))
1909 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910 }
1911 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1912 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001914 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 if (!sock)
1916 goto out;
1917
1918 err = -EMSGSIZE;
1919 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1920 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001921
1922 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 err = -ENOMEM;
1924 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1925 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1926 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1927 if (!iov)
1928 goto out_put;
1929 }
1930
1931 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001932 * Save the user-mode address (verify_iovec will change the
1933 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001935
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -07001936 uaddr = (__force void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 uaddr_len = COMPAT_NAMELEN(msg);
1938 if (MSG_CMSG_COMPAT & flags) {
1939 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1940 } else
1941 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1942 if (err < 0)
1943 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001944 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945
1946 cmsg_ptr = (unsigned long)msg_sys.msg_control;
Ulrich Drepper4a195422007-07-15 23:40:34 -07001947 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001948
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 if (sock->file->f_flags & O_NONBLOCK)
1950 flags |= MSG_DONTWAIT;
1951 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1952 if (err < 0)
1953 goto out_freeiov;
1954 len = err;
1955
1956 if (uaddr != NULL) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001957 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1958 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 if (err < 0)
1960 goto out_freeiov;
1961 }
David S. Miller37f7f422005-09-16 16:51:01 -07001962 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1963 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 if (err)
1965 goto out_freeiov;
1966 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001967 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 &msg_compat->msg_controllen);
1969 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001970 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 &msg->msg_controllen);
1972 if (err)
1973 goto out_freeiov;
1974 err = len;
1975
1976out_freeiov:
1977 if (iov != iovstack)
1978 sock_kfree_s(sock->sk, iov, iov_size);
1979out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001980 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981out:
1982 return err;
1983}
1984
1985#ifdef __ARCH_WANT_SYS_SOCKETCALL
1986
1987/* Argument list sizes for sys_socketcall */
1988#define AL(x) ((x) * sizeof(unsigned long))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001989static const unsigned char nargs[18]={
1990 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1991 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1992 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1993};
1994
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995#undef AL
1996
1997/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001998 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 *
2000 * Argument checking cleaned up. Saved 20% in size.
2001 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002002 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 */
2004
2005asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2006{
2007 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002008 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 int err;
2010
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002011 if (call < 1 || call > SYS_RECVMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012 return -EINVAL;
2013
2014 /* copy_from_user should be SMP safe. */
2015 if (copy_from_user(a, args, nargs[call]))
2016 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002017
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002018 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002019 if (err)
2020 return err;
2021
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002022 a0 = a[0];
2023 a1 = a[1];
2024
2025 switch (call) {
2026 case SYS_SOCKET:
2027 err = sys_socket(a0, a1, a[2]);
2028 break;
2029 case SYS_BIND:
2030 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2031 break;
2032 case SYS_CONNECT:
2033 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2034 break;
2035 case SYS_LISTEN:
2036 err = sys_listen(a0, a1);
2037 break;
2038 case SYS_ACCEPT:
2039 err =
2040 sys_accept(a0, (struct sockaddr __user *)a1,
2041 (int __user *)a[2]);
2042 break;
2043 case SYS_GETSOCKNAME:
2044 err =
2045 sys_getsockname(a0, (struct sockaddr __user *)a1,
2046 (int __user *)a[2]);
2047 break;
2048 case SYS_GETPEERNAME:
2049 err =
2050 sys_getpeername(a0, (struct sockaddr __user *)a1,
2051 (int __user *)a[2]);
2052 break;
2053 case SYS_SOCKETPAIR:
2054 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2055 break;
2056 case SYS_SEND:
2057 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2058 break;
2059 case SYS_SENDTO:
2060 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2061 (struct sockaddr __user *)a[4], a[5]);
2062 break;
2063 case SYS_RECV:
2064 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2065 break;
2066 case SYS_RECVFROM:
2067 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2068 (struct sockaddr __user *)a[4],
2069 (int __user *)a[5]);
2070 break;
2071 case SYS_SHUTDOWN:
2072 err = sys_shutdown(a0, a1);
2073 break;
2074 case SYS_SETSOCKOPT:
2075 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2076 break;
2077 case SYS_GETSOCKOPT:
2078 err =
2079 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2080 (int __user *)a[4]);
2081 break;
2082 case SYS_SENDMSG:
2083 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2084 break;
2085 case SYS_RECVMSG:
2086 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2087 break;
2088 default:
2089 err = -EINVAL;
2090 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 }
2092 return err;
2093}
2094
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002095#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002097/**
2098 * sock_register - add a socket protocol handler
2099 * @ops: description of protocol
2100 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 * This function is called by a protocol handler that wants to
2102 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002103 * socket interface. The value ops->family coresponds to the
2104 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002106int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107{
2108 int err;
2109
2110 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002111 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2112 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 return -ENOBUFS;
2114 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002115
2116 spin_lock(&net_family_lock);
2117 if (net_families[ops->family])
2118 err = -EEXIST;
2119 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002120 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 err = 0;
2122 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002123 spin_unlock(&net_family_lock);
2124
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002125 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 return err;
2127}
2128
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002129/**
2130 * sock_unregister - remove a protocol handler
2131 * @family: protocol family to remove
2132 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 * This function is called by a protocol handler that wants to
2134 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002135 * new socket creation.
2136 *
2137 * If protocol handler is a module, then it can use module reference
2138 * counts to protect against new references. If protocol handler is not
2139 * a module then it needs to provide its own protection in
2140 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002142void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002144 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002146 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002147 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002148 spin_unlock(&net_family_lock);
2149
2150 synchronize_rcu();
2151
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002152 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153}
2154
Andi Kleen77d76ea2005-12-22 12:43:42 -08002155static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156{
2157 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002158 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002160
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161 sk_init();
2162
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002164 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165 */
2166 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167
2168 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002169 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170 */
2171
2172 init_inodecache();
2173 register_filesystem(&sock_fs_type);
2174 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002175
2176 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 */
2178
2179#ifdef CONFIG_NETFILTER
2180 netfilter_init();
2181#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002182
2183 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184}
2185
Andi Kleen77d76ea2005-12-22 12:43:42 -08002186core_initcall(sock_init); /* early initcall */
2187
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188#ifdef CONFIG_PROC_FS
2189void socket_seq_show(struct seq_file *seq)
2190{
2191 int cpu;
2192 int counter = 0;
2193
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002194 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002195 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196
2197 /* It can be negative, by the way. 8) */
2198 if (counter < 0)
2199 counter = 0;
2200
2201 seq_printf(seq, "sockets: used %d\n", counter);
2202}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002203#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002205#ifdef CONFIG_COMPAT
2206static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002207 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002208{
2209 struct socket *sock = file->private_data;
2210 int ret = -ENOIOCTLCMD;
2211
2212 if (sock->ops->compat_ioctl)
2213 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2214
2215 return ret;
2216}
2217#endif
2218
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002219int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2220{
2221 return sock->ops->bind(sock, addr, addrlen);
2222}
2223
2224int kernel_listen(struct socket *sock, int backlog)
2225{
2226 return sock->ops->listen(sock, backlog);
2227}
2228
2229int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2230{
2231 struct sock *sk = sock->sk;
2232 int err;
2233
2234 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2235 newsock);
2236 if (err < 0)
2237 goto done;
2238
2239 err = sock->ops->accept(sock, *newsock, flags);
2240 if (err < 0) {
2241 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07002242 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002243 goto done;
2244 }
2245
2246 (*newsock)->ops = sock->ops;
2247
2248done:
2249 return err;
2250}
2251
2252int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09002253 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002254{
2255 return sock->ops->connect(sock, addr, addrlen, flags);
2256}
2257
2258int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2259 int *addrlen)
2260{
2261 return sock->ops->getname(sock, addr, addrlen, 0);
2262}
2263
2264int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2265 int *addrlen)
2266{
2267 return sock->ops->getname(sock, addr, addrlen, 1);
2268}
2269
2270int kernel_getsockopt(struct socket *sock, int level, int optname,
2271 char *optval, int *optlen)
2272{
2273 mm_segment_t oldfs = get_fs();
2274 int err;
2275
2276 set_fs(KERNEL_DS);
2277 if (level == SOL_SOCKET)
2278 err = sock_getsockopt(sock, level, optname, optval, optlen);
2279 else
2280 err = sock->ops->getsockopt(sock, level, optname, optval,
2281 optlen);
2282 set_fs(oldfs);
2283 return err;
2284}
2285
2286int kernel_setsockopt(struct socket *sock, int level, int optname,
2287 char *optval, int optlen)
2288{
2289 mm_segment_t oldfs = get_fs();
2290 int err;
2291
2292 set_fs(KERNEL_DS);
2293 if (level == SOL_SOCKET)
2294 err = sock_setsockopt(sock, level, optname, optval, optlen);
2295 else
2296 err = sock->ops->setsockopt(sock, level, optname, optval,
2297 optlen);
2298 set_fs(oldfs);
2299 return err;
2300}
2301
2302int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2303 size_t size, int flags)
2304{
2305 if (sock->ops->sendpage)
2306 return sock->ops->sendpage(sock, page, offset, size, flags);
2307
2308 return sock_no_sendpage(sock, page, offset, size, flags);
2309}
2310
2311int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2312{
2313 mm_segment_t oldfs = get_fs();
2314 int err;
2315
2316 set_fs(KERNEL_DS);
2317 err = sock->ops->ioctl(sock, cmd, arg);
2318 set_fs(oldfs);
2319
2320 return err;
2321}
2322
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002323int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2324{
2325 return sock->ops->shutdown(sock, how);
2326}
2327
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328/* ABI emulation layers need these two */
2329EXPORT_SYMBOL(move_addr_to_kernel);
2330EXPORT_SYMBOL(move_addr_to_user);
2331EXPORT_SYMBOL(sock_create);
2332EXPORT_SYMBOL(sock_create_kern);
2333EXPORT_SYMBOL(sock_create_lite);
2334EXPORT_SYMBOL(sock_map_fd);
2335EXPORT_SYMBOL(sock_recvmsg);
2336EXPORT_SYMBOL(sock_register);
2337EXPORT_SYMBOL(sock_release);
2338EXPORT_SYMBOL(sock_sendmsg);
2339EXPORT_SYMBOL(sock_unregister);
2340EXPORT_SYMBOL(sock_wake_async);
2341EXPORT_SYMBOL(sockfd_lookup);
2342EXPORT_SYMBOL(kernel_sendmsg);
2343EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002344EXPORT_SYMBOL(kernel_bind);
2345EXPORT_SYMBOL(kernel_listen);
2346EXPORT_SYMBOL(kernel_accept);
2347EXPORT_SYMBOL(kernel_connect);
2348EXPORT_SYMBOL(kernel_getsockname);
2349EXPORT_SYMBOL(kernel_getpeername);
2350EXPORT_SYMBOL(kernel_getsockopt);
2351EXPORT_SYMBOL(kernel_setsockopt);
2352EXPORT_SYMBOL(kernel_sendpage);
2353EXPORT_SYMBOL(kernel_sock_ioctl);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08002354EXPORT_SYMBOL(kernel_sock_shutdown);