blob: 379b3a3907551f50d42776664d0bc66aafcfa671 [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);
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116/*
117 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
118 * in the operation structures but are done directly via the socketcall() multiplexor.
119 */
120
Arjan van de Venda7071d2007-02-12 00:55:36 -0800121static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 .owner = THIS_MODULE,
123 .llseek = no_llseek,
124 .aio_read = sock_aio_read,
125 .aio_write = sock_aio_write,
126 .poll = sock_poll,
127 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800128#ifdef CONFIG_COMPAT
129 .compat_ioctl = compat_sock_ioctl,
130#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 .mmap = sock_mmap,
132 .open = sock_no_open, /* special open code to disallow open via /proc */
133 .release = sock_close,
134 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200135 .sendpage = sock_sendpage,
136 .splice_write = generic_splice_sendpage,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137};
138
139/*
140 * The protocol list. Each protocol is registered in here.
141 */
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700144static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146/*
147 * Statistics counters of the socket lists
148 */
149
150static DEFINE_PER_CPU(int, sockets_in_use) = 0;
151
152/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700153 * Support routines.
154 * Move socket addresses back and forth across the kernel/user
155 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 */
157
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700158#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 16 for IP, 16 for IPX,
160 24 for IPv6,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700161 about 80 for AX.25
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 must be at least one bigger than
163 the AF_UNIX size (see net/unix/af_unix.c
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700164 :unix_mkname()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167/**
168 * move_addr_to_kernel - copy a socket address into kernel space
169 * @uaddr: Address in user space
170 * @kaddr: Address in kernel space
171 * @ulen: Length in user space
172 *
173 * The address is copied into kernel space. If the provided address is
174 * too long an error code of -EINVAL is returned. If the copy gives
175 * invalid addresses -EFAULT is returned. On a success 0 is returned.
176 */
177
178int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
179{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700180 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700182 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700184 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100186 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187}
188
189/**
190 * move_addr_to_user - copy an address to user space
191 * @kaddr: kernel space address
192 * @klen: length of address in kernel
193 * @uaddr: user space address
194 * @ulen: pointer to user length field
195 *
196 * The value pointed to by ulen on entry is the buffer length available.
197 * This is overwritten with the buffer space used. -EINVAL is returned
198 * if an overlong buffer is specified or a negative buffer size. -EFAULT
199 * is returned if either the buffer or the length field are not
200 * accessible.
201 * After copying the data up to the limit the user specifies, the true
202 * length of the data is written over the length limit the user
203 * specified. Zero is returned for a success.
204 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700205
206int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
207 int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208{
209 int err;
210 int len;
211
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700212 err = get_user(len, ulen);
213 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700215 if (len > klen)
216 len = klen;
217 if (len < 0 || len > MAX_SOCK_ADDR)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700219 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500220 if (audit_sockaddr(klen, kaddr))
221 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700222 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223 return -EFAULT;
224 }
225 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700226 * "fromlen shall refer to the value before truncation.."
227 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 */
229 return __put_user(klen, ulen);
230}
231
232#define SOCKFS_MAGIC 0x534F434B
233
Christoph Lametere18b8902006-12-06 20:33:20 -0800234static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235
236static struct inode *sock_alloc_inode(struct super_block *sb)
237{
238 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700239
Christoph Lametere94b1762006-12-06 20:33:17 -0800240 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 if (!ei)
242 return NULL;
243 init_waitqueue_head(&ei->socket.wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700244
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 ei->socket.fasync_list = NULL;
246 ei->socket.state = SS_UNCONNECTED;
247 ei->socket.flags = 0;
248 ei->socket.ops = NULL;
249 ei->socket.sk = NULL;
250 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
252 return &ei->vfs_inode;
253}
254
255static void sock_destroy_inode(struct inode *inode)
256{
257 kmem_cache_free(sock_inode_cachep,
258 container_of(inode, struct socket_alloc, vfs_inode));
259}
260
Christoph Lametere18b8902006-12-06 20:33:20 -0800261static void init_once(void *foo, struct kmem_cache *cachep, unsigned long flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700263 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264
Christoph Lametera35afb82007-05-16 22:10:57 -0700265 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700267
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268static int init_inodecache(void)
269{
270 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700271 sizeof(struct socket_alloc),
272 0,
273 (SLAB_HWCACHE_ALIGN |
274 SLAB_RECLAIM_ACCOUNT |
275 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900276 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277 if (sock_inode_cachep == NULL)
278 return -ENOMEM;
279 return 0;
280}
281
282static struct super_operations sockfs_ops = {
283 .alloc_inode = sock_alloc_inode,
284 .destroy_inode =sock_destroy_inode,
285 .statfs = simple_statfs,
286};
287
David Howells454e2392006-06-23 02:02:57 -0700288static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700289 int flags, const char *dev_name, void *data,
290 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291{
David Howells454e2392006-06-23 02:02:57 -0700292 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
293 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294}
295
Eric Dumazetba899662005-08-26 12:05:31 -0700296static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297
298static struct file_system_type sock_fs_type = {
299 .name = "sockfs",
300 .get_sb = sockfs_get_sb,
301 .kill_sb = kill_anon_super,
302};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700303
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304static int sockfs_delete_dentry(struct dentry *dentry)
305{
Eric Dumazet304e61e2006-12-06 20:38:49 -0800306 /*
307 * At creation time, we pretended this dentry was hashed
308 * (by clearing DCACHE_UNHASHED bit in d_flags)
309 * At delete time, we restore the truth : not hashed.
310 * (so that dput() can proceed correctly)
311 */
312 dentry->d_flags |= DCACHE_UNHASHED;
313 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314}
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700315
316/*
317 * sockfs_dname() is called from d_path().
318 */
319static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
320{
321 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
322 dentry->d_inode->i_ino);
323}
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325static struct dentry_operations sockfs_dentry_operations = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700326 .d_delete = sockfs_delete_dentry,
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700327 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328};
329
330/*
331 * Obtains the first available file descriptor and sets it up for use.
332 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800333 * These functions create file structures and maps them to fd space
334 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 * and file struct implicitly stored in sock->file.
336 * Note that another thread may close file descriptor before we return
337 * from this function. We use the fact that now we do not refer
338 * to socket after mapping. If one day we will need it, this
339 * function will increment ref. count on file by 1.
340 *
341 * In any case returned fd MAY BE not valid!
342 * This race condition is unavoidable
343 * with shared fd spaces, we cannot solve it inside kernel,
344 * but we take care of internal coherence yet.
345 */
346
David S. Miller39d8c1b2006-03-20 17:13:49 -0800347static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348{
349 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800350
351 fd = get_unused_fd();
352 if (likely(fd >= 0)) {
353 struct file *file = get_empty_filp();
354
355 *filep = file;
356 if (unlikely(!file)) {
357 put_unused_fd(fd);
358 return -ENFILE;
359 }
360 } else
361 *filep = NULL;
362 return fd;
363}
364
365static int sock_attach_fd(struct socket *sock, struct file *file)
366{
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700367 struct qstr name = { .name = "" };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700369 file->f_path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
Josef Sipek3126a422006-12-08 02:37:23 -0800370 if (unlikely(!file->f_path.dentry))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800371 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
Josef Sipek3126a422006-12-08 02:37:23 -0800373 file->f_path.dentry->d_op = &sockfs_dentry_operations;
Eric Dumazet304e61e2006-12-06 20:38:49 -0800374 /*
375 * We dont want to push this dentry into global dentry hash table.
376 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
377 * This permits a working /proc/$pid/fd/XXX on sockets
378 */
Josef Sipek3126a422006-12-08 02:37:23 -0800379 file->f_path.dentry->d_flags &= ~DCACHE_UNHASHED;
380 d_instantiate(file->f_path.dentry, SOCK_INODE(sock));
381 file->f_path.mnt = mntget(sock_mnt);
382 file->f_mapping = file->f_path.dentry->d_inode->i_mapping;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800383
384 sock->file = file;
385 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
386 file->f_mode = FMODE_READ | FMODE_WRITE;
387 file->f_flags = O_RDWR;
388 file->f_pos = 0;
389 file->private_data = sock;
390
391 return 0;
392}
393
394int sock_map_fd(struct socket *sock)
395{
396 struct file *newfile;
397 int fd = sock_alloc_fd(&newfile);
398
399 if (likely(fd >= 0)) {
400 int err = sock_attach_fd(sock, newfile);
401
402 if (unlikely(err < 0)) {
403 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800405 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800407 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 return fd;
410}
411
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800412static struct socket *sock_from_file(struct file *file, int *err)
413{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800414 if (file->f_op == &socket_file_ops)
415 return file->private_data; /* set in sock_map_fd */
416
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800417 *err = -ENOTSOCK;
418 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800419}
420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421/**
422 * sockfd_lookup - Go from a file number to its socket slot
423 * @fd: file handle
424 * @err: pointer to an error code return
425 *
426 * The file handle passed in is locked and the socket it is bound
427 * too is returned. If an error occurs the err pointer is overwritten
428 * with a negative errno code and NULL is returned. The function checks
429 * for both invalid handles and passing a handle which is not a socket.
430 *
431 * On a success the socket object pointer is returned.
432 */
433
434struct socket *sockfd_lookup(int fd, int *err)
435{
436 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 struct socket *sock;
438
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700439 file = fget(fd);
440 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 *err = -EBADF;
442 return NULL;
443 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700444
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800445 sock = sock_from_file(file, err);
446 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 return sock;
449}
450
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452{
453 struct file *file;
454 struct socket *sock;
455
Hua Zhong36725582006-04-19 15:25:02 -0700456 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800457 file = fget_light(fd, fput_needed);
458 if (file) {
459 sock = sock_from_file(file, err);
460 if (sock)
461 return sock;
462 fput_light(file, *fput_needed);
463 }
464 return NULL;
465}
466
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467/**
468 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700469 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 * Allocate a new inode and socket object. The two are bound together
471 * and initialised. The socket is then returned. If we are out of inodes
472 * NULL is returned.
473 */
474
475static struct socket *sock_alloc(void)
476{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700477 struct inode *inode;
478 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479
480 inode = new_inode(sock_mnt->mnt_sb);
481 if (!inode)
482 return NULL;
483
484 sock = SOCKET_I(inode);
485
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700486 inode->i_mode = S_IFSOCK | S_IRWXUGO;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 inode->i_uid = current->fsuid;
488 inode->i_gid = current->fsgid;
489
490 get_cpu_var(sockets_in_use)++;
491 put_cpu_var(sockets_in_use);
492 return sock;
493}
494
495/*
496 * In theory you can't get an open on this inode, but /proc provides
497 * a back door. Remember to keep it shut otherwise you'll let the
498 * creepy crawlies in.
499 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700500
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
502{
503 return -ENXIO;
504}
505
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800506const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 .owner = THIS_MODULE,
508 .open = sock_no_open,
509};
510
511/**
512 * sock_release - close a socket
513 * @sock: socket to close
514 *
515 * The socket is released from the protocol stack if it has a release
516 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700517 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700519
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520void sock_release(struct socket *sock)
521{
522 if (sock->ops) {
523 struct module *owner = sock->ops->owner;
524
525 sock->ops->release(sock);
526 sock->ops = NULL;
527 module_put(owner);
528 }
529
530 if (sock->fasync_list)
531 printk(KERN_ERR "sock_release: fasync list not empty!\n");
532
533 get_cpu_var(sockets_in_use)--;
534 put_cpu_var(sockets_in_use);
535 if (!sock->file) {
536 iput(SOCK_INODE(sock));
537 return;
538 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700539 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540}
541
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700542static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 struct msghdr *msg, size_t size)
544{
545 struct sock_iocb *si = kiocb_to_siocb(iocb);
546 int err;
547
548 si->sock = sock;
549 si->scm = NULL;
550 si->msg = msg;
551 si->size = size;
552
553 err = security_socket_sendmsg(sock, msg, size);
554 if (err)
555 return err;
556
557 return sock->ops->sendmsg(iocb, sock, msg, size);
558}
559
560int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
561{
562 struct kiocb iocb;
563 struct sock_iocb siocb;
564 int ret;
565
566 init_sync_kiocb(&iocb, NULL);
567 iocb.private = &siocb;
568 ret = __sock_sendmsg(&iocb, sock, msg, size);
569 if (-EIOCBQUEUED == ret)
570 ret = wait_on_sync_kiocb(&iocb);
571 return ret;
572}
573
574int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
575 struct kvec *vec, size_t num, size_t size)
576{
577 mm_segment_t oldfs = get_fs();
578 int result;
579
580 set_fs(KERNEL_DS);
581 /*
582 * the following is safe, since for compiler definitions of kvec and
583 * iovec are identical, yielding the same in-core layout and alignment
584 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700585 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 msg->msg_iovlen = num;
587 result = sock_sendmsg(sock, msg, size);
588 set_fs(oldfs);
589 return result;
590}
591
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700592/*
593 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
594 */
595void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
596 struct sk_buff *skb)
597{
598 ktime_t kt = skb->tstamp;
599
600 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
601 struct timeval tv;
602 /* Race occurred between timestamp enabling and packet
603 receiving. Fill in the current time for now. */
604 if (kt.tv64 == 0)
605 kt = ktime_get_real();
606 skb->tstamp = kt;
607 tv = ktime_to_timeval(kt);
608 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
609 } else {
610 struct timespec ts;
611 /* Race occurred between timestamp enabling and packet
612 receiving. Fill in the current time for now. */
613 if (kt.tv64 == 0)
614 kt = ktime_get_real();
615 skb->tstamp = kt;
616 ts = ktime_to_timespec(kt);
617 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
618 }
619}
620
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300621EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
622
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700623static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 struct msghdr *msg, size_t size, int flags)
625{
626 int err;
627 struct sock_iocb *si = kiocb_to_siocb(iocb);
628
629 si->sock = sock;
630 si->scm = NULL;
631 si->msg = msg;
632 si->size = size;
633 si->flags = flags;
634
635 err = security_socket_recvmsg(sock, msg, size, flags);
636 if (err)
637 return err;
638
639 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
640}
641
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700642int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 size_t size, int flags)
644{
645 struct kiocb iocb;
646 struct sock_iocb siocb;
647 int ret;
648
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700649 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 iocb.private = &siocb;
651 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
652 if (-EIOCBQUEUED == ret)
653 ret = wait_on_sync_kiocb(&iocb);
654 return ret;
655}
656
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700657int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
658 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659{
660 mm_segment_t oldfs = get_fs();
661 int result;
662
663 set_fs(KERNEL_DS);
664 /*
665 * the following is safe, since for compiler definitions of kvec and
666 * iovec are identical, yielding the same in-core layout and alignment
667 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700668 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 result = sock_recvmsg(sock, msg, size, flags);
670 set_fs(oldfs);
671 return result;
672}
673
674static void sock_aio_dtor(struct kiocb *iocb)
675{
676 kfree(iocb->private);
677}
678
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300679static ssize_t sock_sendpage(struct file *file, struct page *page,
680 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681{
682 struct socket *sock;
683 int flags;
684
Eric Dumazetb69aee02005-09-06 14:42:45 -0700685 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
687 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
688 if (more)
689 flags |= MSG_MORE;
690
691 return sock->ops->sendpage(sock, page, offset, size, flags);
692}
693
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800694static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700695 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800696{
697 if (!is_sync_kiocb(iocb)) {
698 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
699 if (!siocb)
700 return NULL;
701 iocb->ki_dtor = sock_aio_dtor;
702 }
703
704 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800705 iocb->private = siocb;
706 return siocb;
707}
708
709static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700710 struct file *file, const struct iovec *iov,
711 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800712{
713 struct socket *sock = file->private_data;
714 size_t size = 0;
715 int i;
716
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700717 for (i = 0; i < nr_segs; i++)
718 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800719
720 msg->msg_name = NULL;
721 msg->msg_namelen = 0;
722 msg->msg_control = NULL;
723 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700724 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800725 msg->msg_iovlen = nr_segs;
726 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
727
728 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
729}
730
Badari Pulavarty027445c2006-09-30 23:28:46 -0700731static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
732 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800733{
734 struct sock_iocb siocb, *x;
735
736 if (pos != 0)
737 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700738
739 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800740 return 0;
741
Badari Pulavarty027445c2006-09-30 23:28:46 -0700742
743 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800744 if (!x)
745 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700746 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800747}
748
749static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700750 struct file *file, const struct iovec *iov,
751 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800752{
753 struct socket *sock = file->private_data;
754 size_t size = 0;
755 int i;
756
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700757 for (i = 0; i < nr_segs; i++)
758 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800759
760 msg->msg_name = NULL;
761 msg->msg_namelen = 0;
762 msg->msg_control = NULL;
763 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700764 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800765 msg->msg_iovlen = nr_segs;
766 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
767 if (sock->type == SOCK_SEQPACKET)
768 msg->msg_flags |= MSG_EOR;
769
770 return __sock_sendmsg(iocb, sock, msg, size);
771}
772
Badari Pulavarty027445c2006-09-30 23:28:46 -0700773static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
774 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800776 struct sock_iocb siocb, *x;
777
778 if (pos != 0)
779 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700780
Badari Pulavarty027445c2006-09-30 23:28:46 -0700781 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800782 if (!x)
783 return -ENOMEM;
784
Badari Pulavarty027445c2006-09-30 23:28:46 -0700785 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786}
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788/*
789 * Atomic setting of ioctl hooks to avoid race
790 * with module unload.
791 */
792
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800793static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700794static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Eric W. Biederman881d9662007-09-17 11:56:21 -0700796void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800798 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800800 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700802
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803EXPORT_SYMBOL(brioctl_set);
804
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800805static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700806static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
Eric W. Biederman881d9662007-09-17 11:56:21 -0700808void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800810 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800812 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700814
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815EXPORT_SYMBOL(vlan_ioctl_set);
816
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800817static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700818static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700820void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800822 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800824 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700826
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827EXPORT_SYMBOL(dlci_ioctl_set);
828
829/*
830 * With an ioctl, arg may well be a user mode pointer, but we don't know
831 * what to do with it - that's up to the protocol still.
832 */
833
834static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
835{
836 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700837 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 void __user *argp = (void __user *)arg;
839 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700840 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
Eric Dumazetb69aee02005-09-06 14:42:45 -0700842 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700843 sk = sock->sk;
844 net = sk->sk_net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700846 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100848#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700850 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 } else
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700852#endif /* CONFIG_WIRELESS_EXT */
853 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 case FIOSETOWN:
855 case SIOCSPGRP:
856 err = -EFAULT;
857 if (get_user(pid, (int __user *)argp))
858 break;
859 err = f_setown(sock->file, pid, 1);
860 break;
861 case FIOGETOWN:
862 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700863 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700864 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 break;
866 case SIOCGIFBR:
867 case SIOCSIFBR:
868 case SIOCBRADDBR:
869 case SIOCBRDELBR:
870 err = -ENOPKG;
871 if (!br_ioctl_hook)
872 request_module("bridge");
873
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800874 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700875 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700876 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800877 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 break;
879 case SIOCGIFVLAN:
880 case SIOCSIFVLAN:
881 err = -ENOPKG;
882 if (!vlan_ioctl_hook)
883 request_module("8021q");
884
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800885 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700887 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800888 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 case SIOCADDDLCI:
891 case SIOCDELDLCI:
892 err = -ENOPKG;
893 if (!dlci_ioctl_hook)
894 request_module("dlci");
895
896 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800897 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800899 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 }
901 break;
902 default:
903 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800904
905 /*
906 * If this ioctl is unknown try to hand it down
907 * to the NIC driver.
908 */
909 if (err == -ENOIOCTLCMD)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700910 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700912 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 return err;
914}
915
916int sock_create_lite(int family, int type, int protocol, struct socket **res)
917{
918 int err;
919 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700920
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 err = security_socket_create(family, type, protocol, 1);
922 if (err)
923 goto out;
924
925 sock = sock_alloc();
926 if (!sock) {
927 err = -ENOMEM;
928 goto out;
929 }
930
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700932 err = security_socket_post_create(sock, family, type, protocol, 1);
933 if (err)
934 goto out_release;
935
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936out:
937 *res = sock;
938 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700939out_release:
940 sock_release(sock);
941 sock = NULL;
942 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943}
944
945/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700946static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947{
948 struct socket *sock;
949
950 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700951 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700953 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 return sock->ops->poll(file, sock, wait);
955}
956
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700957static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700959 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960
961 return sock->ops->mmap(file, sock, vma);
962}
963
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300964static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965{
966 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700967 * It was possible the inode is NULL we were
968 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 */
970
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700971 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 printk(KERN_DEBUG "sock_close: NULL inode\n");
973 return 0;
974 }
975 sock_fasync(-1, filp, 0);
976 sock_release(SOCKET_I(inode));
977 return 0;
978}
979
980/*
981 * Update the socket async list
982 *
983 * Fasync_list locking strategy.
984 *
985 * 1. fasync_list is modified only under process context socket lock
986 * i.e. under semaphore.
987 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
988 * or under socket lock.
989 * 3. fasync_list can be used from softirq context, so that
990 * modification under socket lock have to be enhanced with
991 * write_lock_bh(&sk->sk_callback_lock).
992 * --ANK (990710)
993 */
994
995static int sock_fasync(int fd, struct file *filp, int on)
996{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700997 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 struct socket *sock;
999 struct sock *sk;
1000
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001001 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001002 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001003 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 return -ENOMEM;
1005 }
1006
Eric Dumazetb69aee02005-09-06 14:42:45 -07001007 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001009 sk = sock->sk;
1010 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 kfree(fna);
1012 return -EINVAL;
1013 }
1014
1015 lock_sock(sk);
1016
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001017 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001019 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1020 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 break;
1022
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001023 if (on) {
1024 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001026 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 write_unlock_bh(&sk->sk_callback_lock);
1028
1029 kfree(fna);
1030 goto out;
1031 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001032 fna->fa_file = filp;
1033 fna->fa_fd = fd;
1034 fna->magic = FASYNC_MAGIC;
1035 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001037 sock->fasync_list = fna;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001039 } else {
1040 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001042 *prev = fa->fa_next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 write_unlock_bh(&sk->sk_callback_lock);
1044 kfree(fa);
1045 }
1046 }
1047
1048out:
1049 release_sock(sock->sk);
1050 return 0;
1051}
1052
1053/* This function may be called only under socket lock or callback_lock */
1054
1055int sock_wake_async(struct socket *sock, int how, int band)
1056{
1057 if (!sock || !sock->fasync_list)
1058 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001059 switch (how) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 case 1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001061
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1063 break;
1064 goto call_kill;
1065 case 2:
1066 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1067 break;
1068 /* fall through */
1069 case 0:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001070call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 __kill_fasync(sock->fasync_list, SIGIO, band);
1072 break;
1073 case 3:
1074 __kill_fasync(sock->fasync_list, SIGURG, band);
1075 }
1076 return 0;
1077}
1078
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001079static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001080 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081{
1082 int err;
1083 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001084 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085
1086 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001087 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 */
1089 if (family < 0 || family >= NPROTO)
1090 return -EAFNOSUPPORT;
1091 if (type < 0 || type >= SOCK_MAX)
1092 return -EINVAL;
1093
1094 /* Compatibility.
1095
1096 This uglymoron is moved from INET layer to here to avoid
1097 deadlock in module load.
1098 */
1099 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001100 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 if (!warned) {
1102 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001103 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1104 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 }
1106 family = PF_PACKET;
1107 }
1108
1109 err = security_socket_create(family, type, protocol, kern);
1110 if (err)
1111 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001112
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001113 /*
1114 * Allocate the socket and allow the family to set things up. if
1115 * the protocol is 0, the family is instructed to select an appropriate
1116 * default.
1117 */
1118 sock = sock_alloc();
1119 if (!sock) {
1120 if (net_ratelimit())
1121 printk(KERN_WARNING "socket: no more sockets\n");
1122 return -ENFILE; /* Not exactly a match, but its the
1123 closest posix thing */
1124 }
1125
1126 sock->type = type;
1127
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128#if defined(CONFIG_KMOD)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001129 /* Attempt to load a protocol module if the find failed.
1130 *
1131 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 * requested real, full-featured networking support upon configuration.
1133 * Otherwise module support will break!
1134 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001135 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001136 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137#endif
1138
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001139 rcu_read_lock();
1140 pf = rcu_dereference(net_families[family]);
1141 err = -EAFNOSUPPORT;
1142 if (!pf)
1143 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
1145 /*
1146 * We will call the ->create function, that possibly is in a loadable
1147 * module, so we have to bump that loadable module refcnt first.
1148 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001149 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 goto out_release;
1151
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001152 /* Now protected by module ref count */
1153 rcu_read_unlock();
1154
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001155 err = pf->create(net, sock, protocol);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001156 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001158
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 /*
1160 * Now to bump the refcnt of the [loadable] module that owns this
1161 * socket at sock_release time we decrement its refcnt.
1162 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001163 if (!try_module_get(sock->ops->owner))
1164 goto out_module_busy;
1165
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 /*
1167 * Now that we're done with the ->create function, the [loadable]
1168 * module can have its refcnt decremented
1169 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001170 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001171 err = security_socket_post_create(sock, family, type, protocol, kern);
1172 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001173 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001174 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001176 return 0;
1177
1178out_module_busy:
1179 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001181 sock->ops = NULL;
1182 module_put(pf->owner);
1183out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001185 return err;
1186
1187out_release:
1188 rcu_read_unlock();
1189 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190}
1191
1192int sock_create(int family, int type, int protocol, struct socket **res)
1193{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001194 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195}
1196
1197int sock_create_kern(int family, int type, int protocol, struct socket **res)
1198{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001199 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200}
1201
1202asmlinkage long sys_socket(int family, int type, int protocol)
1203{
1204 int retval;
1205 struct socket *sock;
1206
1207 retval = sock_create(family, type, protocol, &sock);
1208 if (retval < 0)
1209 goto out;
1210
1211 retval = sock_map_fd(sock);
1212 if (retval < 0)
1213 goto out_release;
1214
1215out:
1216 /* It may be already another descriptor 8) Not kernel problem. */
1217 return retval;
1218
1219out_release:
1220 sock_release(sock);
1221 return retval;
1222}
1223
1224/*
1225 * Create a pair of connected sockets.
1226 */
1227
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001228asmlinkage long sys_socketpair(int family, int type, int protocol,
1229 int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230{
1231 struct socket *sock1, *sock2;
1232 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001233 struct file *newfile1, *newfile2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
1235 /*
1236 * Obtain the first socket and check if the underlying protocol
1237 * supports the socketpair call.
1238 */
1239
1240 err = sock_create(family, type, protocol, &sock1);
1241 if (err < 0)
1242 goto out;
1243
1244 err = sock_create(family, type, protocol, &sock2);
1245 if (err < 0)
1246 goto out_release_1;
1247
1248 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001249 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 goto out_release_both;
1251
Al Virodb349502007-02-07 01:48:00 -05001252 fd1 = sock_alloc_fd(&newfile1);
1253 if (unlikely(fd1 < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 goto out_release_both;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255
Al Virodb349502007-02-07 01:48:00 -05001256 fd2 = sock_alloc_fd(&newfile2);
1257 if (unlikely(fd2 < 0)) {
1258 put_filp(newfile1);
1259 put_unused_fd(fd1);
1260 goto out_release_both;
1261 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262
Al Virodb349502007-02-07 01:48:00 -05001263 err = sock_attach_fd(sock1, newfile1);
1264 if (unlikely(err < 0)) {
1265 goto out_fd2;
1266 }
1267
1268 err = sock_attach_fd(sock2, newfile2);
1269 if (unlikely(err < 0)) {
1270 fput(newfile1);
1271 goto out_fd1;
1272 }
1273
1274 err = audit_fd_pair(fd1, fd2);
1275 if (err < 0) {
1276 fput(newfile1);
1277 fput(newfile2);
1278 goto out_fd;
1279 }
1280
1281 fd_install(fd1, newfile1);
1282 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 /* fd1 and fd2 may be already another descriptors.
1284 * Not kernel problem.
1285 */
1286
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001287 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 if (!err)
1289 err = put_user(fd2, &usockvec[1]);
1290 if (!err)
1291 return 0;
1292
1293 sys_close(fd2);
1294 sys_close(fd1);
1295 return err;
1296
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001298 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001300 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301out:
1302 return err;
Al Virodb349502007-02-07 01:48:00 -05001303
1304out_fd2:
1305 put_filp(newfile1);
1306 sock_release(sock1);
1307out_fd1:
1308 put_filp(newfile2);
1309 sock_release(sock2);
1310out_fd:
1311 put_unused_fd(fd1);
1312 put_unused_fd(fd2);
1313 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314}
1315
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316/*
1317 * Bind a name to a socket. Nothing much to do here since it's
1318 * the protocol's responsibility to handle the local address.
1319 *
1320 * We move the socket address to kernel space before we call
1321 * the protocol layer (having also checked the address is ok).
1322 */
1323
1324asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1325{
1326 struct socket *sock;
1327 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001328 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001330 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001331 if (sock) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001332 err = move_addr_to_kernel(umyaddr, addrlen, address);
1333 if (err >= 0) {
1334 err = security_socket_bind(sock,
1335 (struct sockaddr *)address,
1336 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001337 if (!err)
1338 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001339 (struct sockaddr *)
1340 address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001342 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001343 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 return err;
1345}
1346
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347/*
1348 * Perform a listen. Basically, we allow the protocol to do anything
1349 * necessary for a listen, and if that works, we mark the socket as
1350 * ready for listening.
1351 */
1352
Brian Haley7a42c212006-08-31 15:03:02 -07001353int sysctl_somaxconn __read_mostly = SOMAXCONN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
1355asmlinkage long sys_listen(int fd, int backlog)
1356{
1357 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001358 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001359
1360 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1361 if (sock) {
1362 if ((unsigned)backlog > sysctl_somaxconn)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 backlog = sysctl_somaxconn;
1364
1365 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001366 if (!err)
1367 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001369 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 }
1371 return err;
1372}
1373
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374/*
1375 * For accept, we attempt to create a new socket, set up the link
1376 * with the client, wake up the client, then return the new
1377 * connected fd. We collect the address of the connector in kernel
1378 * space and move it to user at the very end. This is unclean because
1379 * we open the socket then return an error.
1380 *
1381 * 1003.1g adds the ability to recvmsg() to query connection pending
1382 * status to recvmsg. We need to add that support in a way thats
1383 * clean when we restucture accept also.
1384 */
1385
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001386asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1387 int __user *upeer_addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388{
1389 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001390 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001391 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 char address[MAX_SOCK_ADDR];
1393
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001394 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 if (!sock)
1396 goto out;
1397
1398 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001399 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 goto out_put;
1401
1402 newsock->type = sock->type;
1403 newsock->ops = sock->ops;
1404
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 /*
1406 * We don't need try_module_get here, as the listening socket (sock)
1407 * has the protocol module (sock->ops->owner) held.
1408 */
1409 __module_get(newsock->ops->owner);
1410
David S. Miller39d8c1b2006-03-20 17:13:49 -08001411 newfd = sock_alloc_fd(&newfile);
1412 if (unlikely(newfd < 0)) {
1413 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001414 sock_release(newsock);
1415 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001416 }
1417
1418 err = sock_attach_fd(newsock, newfile);
1419 if (err < 0)
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001420 goto out_fd_simple;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001421
Frank Filza79af592005-09-27 15:23:38 -07001422 err = security_socket_accept(sock, newsock);
1423 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001424 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001425
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1427 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001428 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
1430 if (upeer_sockaddr) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001431 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1432 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001434 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001436 err = move_addr_to_user(address, len, upeer_sockaddr,
1437 upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001439 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 }
1441
1442 /* File flags are not inherited via accept() unlike another OSes. */
1443
David S. Miller39d8c1b2006-03-20 17:13:49 -08001444 fd_install(newfd, newfile);
1445 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446
1447 security_socket_post_accept(sock, newsock);
1448
1449out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001450 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451out:
1452 return err;
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001453out_fd_simple:
1454 sock_release(newsock);
1455 put_filp(newfile);
1456 put_unused_fd(newfd);
1457 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001458out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001459 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001460 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 goto out_put;
1462}
1463
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464/*
1465 * Attempt to connect to a socket with the server address. The address
1466 * is in user space so we verify it is OK and move it to kernel space.
1467 *
1468 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1469 * break bindings
1470 *
1471 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1472 * other SEQPACKET protocols that take time to connect() as it doesn't
1473 * include the -EINPROGRESS status for such sockets.
1474 */
1475
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001476asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1477 int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478{
1479 struct socket *sock;
1480 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001481 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001483 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 if (!sock)
1485 goto out;
1486 err = move_addr_to_kernel(uservaddr, addrlen, address);
1487 if (err < 0)
1488 goto out_put;
1489
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001490 err =
1491 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 if (err)
1493 goto out_put;
1494
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001495 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 sock->file->f_flags);
1497out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001498 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499out:
1500 return err;
1501}
1502
1503/*
1504 * Get the local address ('name') of a socket object. Move the obtained
1505 * name to user space.
1506 */
1507
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001508asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1509 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510{
1511 struct socket *sock;
1512 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001513 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001514
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001515 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 if (!sock)
1517 goto out;
1518
1519 err = security_socket_getsockname(sock);
1520 if (err)
1521 goto out_put;
1522
1523 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1524 if (err)
1525 goto out_put;
1526 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1527
1528out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001529 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530out:
1531 return err;
1532}
1533
1534/*
1535 * Get the remote address ('name') of a socket object. Move the obtained
1536 * name to user space.
1537 */
1538
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001539asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1540 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541{
1542 struct socket *sock;
1543 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001544 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001546 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1547 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 err = security_socket_getpeername(sock);
1549 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001550 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 return err;
1552 }
1553
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001554 err =
1555 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1556 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 if (!err)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001558 err = move_addr_to_user(address, len, usockaddr,
1559 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001560 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 }
1562 return err;
1563}
1564
1565/*
1566 * Send a datagram to a given address. We move the address into kernel
1567 * space and check the user space data area is readable before invoking
1568 * the protocol.
1569 */
1570
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001571asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1572 unsigned flags, struct sockaddr __user *addr,
1573 int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574{
1575 struct socket *sock;
1576 char address[MAX_SOCK_ADDR];
1577 int err;
1578 struct msghdr msg;
1579 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001580 int fput_needed;
1581 struct file *sock_file;
1582
1583 sock_file = fget_light(fd, &fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001584 err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001585 if (!sock_file)
David S. Miller4387ff72007-02-08 15:06:08 -08001586 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001587
1588 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001590 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001591 iov.iov_base = buff;
1592 iov.iov_len = len;
1593 msg.msg_name = NULL;
1594 msg.msg_iov = &iov;
1595 msg.msg_iovlen = 1;
1596 msg.msg_control = NULL;
1597 msg.msg_controllen = 0;
1598 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001599 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 err = move_addr_to_kernel(addr, addr_len, address);
1601 if (err < 0)
1602 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001603 msg.msg_name = address;
1604 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 }
1606 if (sock->file->f_flags & O_NONBLOCK)
1607 flags |= MSG_DONTWAIT;
1608 msg.msg_flags = flags;
1609 err = sock_sendmsg(sock, &msg, len);
1610
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001611out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001612 fput_light(sock_file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001613out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 return err;
1615}
1616
1617/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001618 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 */
1620
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001621asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622{
1623 return sys_sendto(fd, buff, len, flags, NULL, 0);
1624}
1625
1626/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001627 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 * sender. We verify the buffers are writable and if needed move the
1629 * sender address from kernel to user space.
1630 */
1631
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001632asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1633 unsigned flags, struct sockaddr __user *addr,
1634 int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635{
1636 struct socket *sock;
1637 struct iovec iov;
1638 struct msghdr msg;
1639 char address[MAX_SOCK_ADDR];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001640 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001641 struct file *sock_file;
1642 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001644 sock_file = fget_light(fd, &fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001645 err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001646 if (!sock_file)
David S. Miller4387ff72007-02-08 15:06:08 -08001647 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001648
1649 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001651 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001653 msg.msg_control = NULL;
1654 msg.msg_controllen = 0;
1655 msg.msg_iovlen = 1;
1656 msg.msg_iov = &iov;
1657 iov.iov_len = size;
1658 iov.iov_base = ubuf;
1659 msg.msg_name = address;
1660 msg.msg_namelen = MAX_SOCK_ADDR;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 if (sock->file->f_flags & O_NONBLOCK)
1662 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001663 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001665 if (err >= 0 && addr != NULL) {
1666 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1667 if (err2 < 0)
1668 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 }
David S. Miller4387ff72007-02-08 15:06:08 -08001670out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001671 fput_light(sock_file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001672out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673 return err;
1674}
1675
1676/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001677 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 */
1679
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001680asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1681 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682{
1683 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1684}
1685
1686/*
1687 * Set a socket option. Because we don't know the option lengths we have
1688 * to pass the user mode parameter for the protocols to sort out.
1689 */
1690
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001691asmlinkage long sys_setsockopt(int fd, int level, int optname,
1692 char __user *optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001694 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 struct socket *sock;
1696
1697 if (optlen < 0)
1698 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001699
1700 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1701 if (sock != NULL) {
1702 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001703 if (err)
1704 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
1706 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001707 err =
1708 sock_setsockopt(sock, level, optname, optval,
1709 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001711 err =
1712 sock->ops->setsockopt(sock, level, optname, optval,
1713 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001714out_put:
1715 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 }
1717 return err;
1718}
1719
1720/*
1721 * Get a socket option. Because we don't know the option lengths we have
1722 * to pass a user mode parameter for the protocols to sort out.
1723 */
1724
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001725asmlinkage long sys_getsockopt(int fd, int level, int optname,
1726 char __user *optval, int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001728 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 struct socket *sock;
1730
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001731 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1732 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001733 err = security_socket_getsockopt(sock, level, optname);
1734 if (err)
1735 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
1737 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001738 err =
1739 sock_getsockopt(sock, level, optname, optval,
1740 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001742 err =
1743 sock->ops->getsockopt(sock, level, optname, optval,
1744 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001745out_put:
1746 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 }
1748 return err;
1749}
1750
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751/*
1752 * Shutdown a socket.
1753 */
1754
1755asmlinkage long sys_shutdown(int fd, int how)
1756{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001757 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 struct socket *sock;
1759
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001760 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1761 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001763 if (!err)
1764 err = sock->ops->shutdown(sock, how);
1765 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 }
1767 return err;
1768}
1769
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001770/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 * fields which are the same type (int / unsigned) on our platforms.
1772 */
1773#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1774#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1775#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1776
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777/*
1778 * BSD sendmsg interface
1779 */
1780
1781asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1782{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001783 struct compat_msghdr __user *msg_compat =
1784 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 struct socket *sock;
1786 char address[MAX_SOCK_ADDR];
1787 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001788 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001789 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1790 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 unsigned char *ctl_buf = ctl;
1792 struct msghdr msg_sys;
1793 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001794 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001795
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 err = -EFAULT;
1797 if (MSG_CMSG_COMPAT & flags) {
1798 if (get_compat_msghdr(&msg_sys, msg_compat))
1799 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001800 }
1801 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 return -EFAULT;
1803
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001804 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001805 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 goto out;
1807
1808 /* do not move before msg_sys is valid */
1809 err = -EMSGSIZE;
1810 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1811 goto out_put;
1812
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001813 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 err = -ENOMEM;
1815 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1816 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1817 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1818 if (!iov)
1819 goto out_put;
1820 }
1821
1822 /* This will also move the address data into kernel space */
1823 if (MSG_CMSG_COMPAT & flags) {
1824 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1825 } else
1826 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001827 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 goto out_freeiov;
1829 total_len = err;
1830
1831 err = -ENOBUFS;
1832
1833 if (msg_sys.msg_controllen > INT_MAX)
1834 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001835 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001837 err =
1838 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1839 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 if (err)
1841 goto out_freeiov;
1842 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001843 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001845 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001847 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 goto out_freeiov;
1849 }
1850 err = -EFAULT;
1851 /*
1852 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1853 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1854 * checking falls down on this.
1855 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001856 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1857 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 goto out_freectl;
1859 msg_sys.msg_control = ctl_buf;
1860 }
1861 msg_sys.msg_flags = flags;
1862
1863 if (sock->file->f_flags & O_NONBLOCK)
1864 msg_sys.msg_flags |= MSG_DONTWAIT;
1865 err = sock_sendmsg(sock, &msg_sys, total_len);
1866
1867out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001868 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1870out_freeiov:
1871 if (iov != iovstack)
1872 sock_kfree_s(sock->sk, iov, iov_size);
1873out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001874 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001875out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876 return err;
1877}
1878
1879/*
1880 * BSD recvmsg interface
1881 */
1882
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001883asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1884 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001886 struct compat_msghdr __user *msg_compat =
1887 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 struct socket *sock;
1889 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 struct msghdr msg_sys;
1892 unsigned long cmsg_ptr;
1893 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001894 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895
1896 /* kernel mode address */
1897 char addr[MAX_SOCK_ADDR];
1898
1899 /* user mode address pointers */
1900 struct sockaddr __user *uaddr;
1901 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001902
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 if (MSG_CMSG_COMPAT & flags) {
1904 if (get_compat_msghdr(&msg_sys, msg_compat))
1905 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001906 }
1907 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1908 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001910 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 if (!sock)
1912 goto out;
1913
1914 err = -EMSGSIZE;
1915 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1916 goto out_put;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001917
1918 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 err = -ENOMEM;
1920 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1921 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1922 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1923 if (!iov)
1924 goto out_put;
1925 }
1926
1927 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001928 * Save the user-mode address (verify_iovec will change the
1929 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001931
Stephen Hemmingercfcabdc2007-10-09 01:59:42 -07001932 uaddr = (__force void __user *)msg_sys.msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 uaddr_len = COMPAT_NAMELEN(msg);
1934 if (MSG_CMSG_COMPAT & flags) {
1935 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1936 } else
1937 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1938 if (err < 0)
1939 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001940 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941
1942 cmsg_ptr = (unsigned long)msg_sys.msg_control;
Ulrich Drepper4a195422007-07-15 23:40:34 -07001943 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001944
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 if (sock->file->f_flags & O_NONBLOCK)
1946 flags |= MSG_DONTWAIT;
1947 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1948 if (err < 0)
1949 goto out_freeiov;
1950 len = err;
1951
1952 if (uaddr != NULL) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001953 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1954 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 if (err < 0)
1956 goto out_freeiov;
1957 }
David S. Miller37f7f422005-09-16 16:51:01 -07001958 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1959 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 if (err)
1961 goto out_freeiov;
1962 if (MSG_CMSG_COMPAT & flags)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001963 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 &msg_compat->msg_controllen);
1965 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001966 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 &msg->msg_controllen);
1968 if (err)
1969 goto out_freeiov;
1970 err = len;
1971
1972out_freeiov:
1973 if (iov != iovstack)
1974 sock_kfree_s(sock->sk, iov, iov_size);
1975out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001976 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977out:
1978 return err;
1979}
1980
1981#ifdef __ARCH_WANT_SYS_SOCKETCALL
1982
1983/* Argument list sizes for sys_socketcall */
1984#define AL(x) ((x) * sizeof(unsigned long))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001985static const unsigned char nargs[18]={
1986 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1987 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1988 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1989};
1990
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991#undef AL
1992
1993/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001994 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 *
1996 * Argument checking cleaned up. Saved 20% in size.
1997 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001998 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 */
2000
2001asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2002{
2003 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002004 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 int err;
2006
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002007 if (call < 1 || call > SYS_RECVMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008 return -EINVAL;
2009
2010 /* copy_from_user should be SMP safe. */
2011 if (copy_from_user(a, args, nargs[call]))
2012 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002013
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002014 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002015 if (err)
2016 return err;
2017
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002018 a0 = a[0];
2019 a1 = a[1];
2020
2021 switch (call) {
2022 case SYS_SOCKET:
2023 err = sys_socket(a0, a1, a[2]);
2024 break;
2025 case SYS_BIND:
2026 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2027 break;
2028 case SYS_CONNECT:
2029 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2030 break;
2031 case SYS_LISTEN:
2032 err = sys_listen(a0, a1);
2033 break;
2034 case SYS_ACCEPT:
2035 err =
2036 sys_accept(a0, (struct sockaddr __user *)a1,
2037 (int __user *)a[2]);
2038 break;
2039 case SYS_GETSOCKNAME:
2040 err =
2041 sys_getsockname(a0, (struct sockaddr __user *)a1,
2042 (int __user *)a[2]);
2043 break;
2044 case SYS_GETPEERNAME:
2045 err =
2046 sys_getpeername(a0, (struct sockaddr __user *)a1,
2047 (int __user *)a[2]);
2048 break;
2049 case SYS_SOCKETPAIR:
2050 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2051 break;
2052 case SYS_SEND:
2053 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2054 break;
2055 case SYS_SENDTO:
2056 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2057 (struct sockaddr __user *)a[4], a[5]);
2058 break;
2059 case SYS_RECV:
2060 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2061 break;
2062 case SYS_RECVFROM:
2063 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2064 (struct sockaddr __user *)a[4],
2065 (int __user *)a[5]);
2066 break;
2067 case SYS_SHUTDOWN:
2068 err = sys_shutdown(a0, a1);
2069 break;
2070 case SYS_SETSOCKOPT:
2071 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2072 break;
2073 case SYS_GETSOCKOPT:
2074 err =
2075 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2076 (int __user *)a[4]);
2077 break;
2078 case SYS_SENDMSG:
2079 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2080 break;
2081 case SYS_RECVMSG:
2082 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2083 break;
2084 default:
2085 err = -EINVAL;
2086 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087 }
2088 return err;
2089}
2090
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002091#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002093/**
2094 * sock_register - add a socket protocol handler
2095 * @ops: description of protocol
2096 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 * This function is called by a protocol handler that wants to
2098 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002099 * socket interface. The value ops->family coresponds to the
2100 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002102int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103{
2104 int err;
2105
2106 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002107 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2108 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 return -ENOBUFS;
2110 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002111
2112 spin_lock(&net_family_lock);
2113 if (net_families[ops->family])
2114 err = -EEXIST;
2115 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002116 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 err = 0;
2118 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002119 spin_unlock(&net_family_lock);
2120
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002121 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 return err;
2123}
2124
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002125/**
2126 * sock_unregister - remove a protocol handler
2127 * @family: protocol family to remove
2128 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 * This function is called by a protocol handler that wants to
2130 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002131 * new socket creation.
2132 *
2133 * If protocol handler is a module, then it can use module reference
2134 * counts to protect against new references. If protocol handler is not
2135 * a module then it needs to provide its own protection in
2136 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002138void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002140 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002142 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002143 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002144 spin_unlock(&net_family_lock);
2145
2146 synchronize_rcu();
2147
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002148 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149}
2150
Andi Kleen77d76ea2005-12-22 12:43:42 -08002151static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152{
2153 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002154 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002156
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 sk_init();
2158
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002160 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161 */
2162 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163
2164 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002165 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002166 */
2167
2168 init_inodecache();
2169 register_filesystem(&sock_fs_type);
2170 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002171
2172 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173 */
2174
2175#ifdef CONFIG_NETFILTER
2176 netfilter_init();
2177#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002178
2179 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180}
2181
Andi Kleen77d76ea2005-12-22 12:43:42 -08002182core_initcall(sock_init); /* early initcall */
2183
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184#ifdef CONFIG_PROC_FS
2185void socket_seq_show(struct seq_file *seq)
2186{
2187 int cpu;
2188 int counter = 0;
2189
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002190 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002191 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192
2193 /* It can be negative, by the way. 8) */
2194 if (counter < 0)
2195 counter = 0;
2196
2197 seq_printf(seq, "sockets: used %d\n", counter);
2198}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002199#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002201#ifdef CONFIG_COMPAT
2202static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002203 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002204{
2205 struct socket *sock = file->private_data;
2206 int ret = -ENOIOCTLCMD;
2207
2208 if (sock->ops->compat_ioctl)
2209 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2210
2211 return ret;
2212}
2213#endif
2214
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002215int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2216{
2217 return sock->ops->bind(sock, addr, addrlen);
2218}
2219
2220int kernel_listen(struct socket *sock, int backlog)
2221{
2222 return sock->ops->listen(sock, backlog);
2223}
2224
2225int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2226{
2227 struct sock *sk = sock->sk;
2228 int err;
2229
2230 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2231 newsock);
2232 if (err < 0)
2233 goto done;
2234
2235 err = sock->ops->accept(sock, *newsock, flags);
2236 if (err < 0) {
2237 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07002238 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002239 goto done;
2240 }
2241
2242 (*newsock)->ops = sock->ops;
2243
2244done:
2245 return err;
2246}
2247
2248int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09002249 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002250{
2251 return sock->ops->connect(sock, addr, addrlen, flags);
2252}
2253
2254int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2255 int *addrlen)
2256{
2257 return sock->ops->getname(sock, addr, addrlen, 0);
2258}
2259
2260int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2261 int *addrlen)
2262{
2263 return sock->ops->getname(sock, addr, addrlen, 1);
2264}
2265
2266int kernel_getsockopt(struct socket *sock, int level, int optname,
2267 char *optval, int *optlen)
2268{
2269 mm_segment_t oldfs = get_fs();
2270 int err;
2271
2272 set_fs(KERNEL_DS);
2273 if (level == SOL_SOCKET)
2274 err = sock_getsockopt(sock, level, optname, optval, optlen);
2275 else
2276 err = sock->ops->getsockopt(sock, level, optname, optval,
2277 optlen);
2278 set_fs(oldfs);
2279 return err;
2280}
2281
2282int kernel_setsockopt(struct socket *sock, int level, int optname,
2283 char *optval, int optlen)
2284{
2285 mm_segment_t oldfs = get_fs();
2286 int err;
2287
2288 set_fs(KERNEL_DS);
2289 if (level == SOL_SOCKET)
2290 err = sock_setsockopt(sock, level, optname, optval, optlen);
2291 else
2292 err = sock->ops->setsockopt(sock, level, optname, optval,
2293 optlen);
2294 set_fs(oldfs);
2295 return err;
2296}
2297
2298int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2299 size_t size, int flags)
2300{
2301 if (sock->ops->sendpage)
2302 return sock->ops->sendpage(sock, page, offset, size, flags);
2303
2304 return sock_no_sendpage(sock, page, offset, size, flags);
2305}
2306
2307int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2308{
2309 mm_segment_t oldfs = get_fs();
2310 int err;
2311
2312 set_fs(KERNEL_DS);
2313 err = sock->ops->ioctl(sock, cmd, arg);
2314 set_fs(oldfs);
2315
2316 return err;
2317}
2318
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319/* ABI emulation layers need these two */
2320EXPORT_SYMBOL(move_addr_to_kernel);
2321EXPORT_SYMBOL(move_addr_to_user);
2322EXPORT_SYMBOL(sock_create);
2323EXPORT_SYMBOL(sock_create_kern);
2324EXPORT_SYMBOL(sock_create_lite);
2325EXPORT_SYMBOL(sock_map_fd);
2326EXPORT_SYMBOL(sock_recvmsg);
2327EXPORT_SYMBOL(sock_register);
2328EXPORT_SYMBOL(sock_release);
2329EXPORT_SYMBOL(sock_sendmsg);
2330EXPORT_SYMBOL(sock_unregister);
2331EXPORT_SYMBOL(sock_wake_async);
2332EXPORT_SYMBOL(sockfd_lookup);
2333EXPORT_SYMBOL(kernel_sendmsg);
2334EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002335EXPORT_SYMBOL(kernel_bind);
2336EXPORT_SYMBOL(kernel_listen);
2337EXPORT_SYMBOL(kernel_accept);
2338EXPORT_SYMBOL(kernel_connect);
2339EXPORT_SYMBOL(kernel_getsockname);
2340EXPORT_SYMBOL(kernel_getpeername);
2341EXPORT_SYMBOL(kernel_getsockopt);
2342EXPORT_SYMBOL(kernel_setsockopt);
2343EXPORT_SYMBOL(kernel_sendpage);
2344EXPORT_SYMBOL(kernel_sock_ioctl);