blob: 8aa5f1188e9ba1eaae67835283933dced8fc6f20 [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)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * 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
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61#include <linux/config.h>
62#include <linux/mm.h>
63#include <linux/smp_lock.h>
64#include <linux/socket.h>
65#include <linux/file.h>
66#include <linux/net.h>
67#include <linux/interrupt.h>
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
71#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>
80#include <linux/divert.h>
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
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);
98static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
99 size_t size, loff_t pos);
100static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
101 size_t size, loff_t pos);
102static int sock_mmap(struct file *file, struct vm_area_struct * vma);
103
104static int sock_close(struct inode *inode, struct file *file);
105static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
107static long sock_ioctl(struct file *file,
108 unsigned int cmd, unsigned long arg);
109static int sock_fasync(int fd, struct file *filp, int on);
110static ssize_t sock_readv(struct file *file, const struct iovec *vector,
111 unsigned long count, loff_t *ppos);
112static ssize_t sock_writev(struct file *file, const struct iovec *vector,
113 unsigned long count, loff_t *ppos);
114static ssize_t sock_sendpage(struct file *file, struct page *page,
115 int offset, size_t size, loff_t *ppos, int more);
116
117
118/*
119 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
120 * in the operation structures but are done directly via the socketcall() multiplexor.
121 */
122
123static struct file_operations socket_file_ops = {
124 .owner = THIS_MODULE,
125 .llseek = no_llseek,
126 .aio_read = sock_aio_read,
127 .aio_write = sock_aio_write,
128 .poll = sock_poll,
129 .unlocked_ioctl = sock_ioctl,
130 .mmap = sock_mmap,
131 .open = sock_no_open, /* special open code to disallow open via /proc */
132 .release = sock_close,
133 .fasync = sock_fasync,
134 .readv = sock_readv,
135 .writev = sock_writev,
136 .sendpage = sock_sendpage
137};
138
139/*
140 * The protocol list. Each protocol is registered in here.
141 */
142
143static struct net_proto_family *net_families[NPROTO];
144
145#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
146static atomic_t net_family_lockct = ATOMIC_INIT(0);
147static DEFINE_SPINLOCK(net_family_lock);
148
149/* The strategy is: modifications net_family vector are short, do not
150 sleep and veeery rare, but read access should be free of any exclusive
151 locks.
152 */
153
154static void net_family_write_lock(void)
155{
156 spin_lock(&net_family_lock);
157 while (atomic_read(&net_family_lockct) != 0) {
158 spin_unlock(&net_family_lock);
159
160 yield();
161
162 spin_lock(&net_family_lock);
163 }
164}
165
166static __inline__ void net_family_write_unlock(void)
167{
168 spin_unlock(&net_family_lock);
169}
170
171static __inline__ void net_family_read_lock(void)
172{
173 atomic_inc(&net_family_lockct);
174 spin_unlock_wait(&net_family_lock);
175}
176
177static __inline__ void net_family_read_unlock(void)
178{
179 atomic_dec(&net_family_lockct);
180}
181
182#else
183#define net_family_write_lock() do { } while(0)
184#define net_family_write_unlock() do { } while(0)
185#define net_family_read_lock() do { } while(0)
186#define net_family_read_unlock() do { } while(0)
187#endif
188
189
190/*
191 * Statistics counters of the socket lists
192 */
193
194static DEFINE_PER_CPU(int, sockets_in_use) = 0;
195
196/*
197 * Support routines. Move socket addresses back and forth across the kernel/user
198 * divide and look after the messy bits.
199 */
200
201#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
202 16 for IP, 16 for IPX,
203 24 for IPv6,
204 about 80 for AX.25
205 must be at least one bigger than
206 the AF_UNIX size (see net/unix/af_unix.c
207 :unix_mkname()).
208 */
209
210/**
211 * move_addr_to_kernel - copy a socket address into kernel space
212 * @uaddr: Address in user space
213 * @kaddr: Address in kernel space
214 * @ulen: Length in user space
215 *
216 * The address is copied into kernel space. If the provided address is
217 * too long an error code of -EINVAL is returned. If the copy gives
218 * invalid addresses -EFAULT is returned. On a success 0 is returned.
219 */
220
221int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
222{
223 if(ulen<0||ulen>MAX_SOCK_ADDR)
224 return -EINVAL;
225 if(ulen==0)
226 return 0;
227 if(copy_from_user(kaddr,uaddr,ulen))
228 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100229 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230}
231
232/**
233 * move_addr_to_user - copy an address to user space
234 * @kaddr: kernel space address
235 * @klen: length of address in kernel
236 * @uaddr: user space address
237 * @ulen: pointer to user length field
238 *
239 * The value pointed to by ulen on entry is the buffer length available.
240 * This is overwritten with the buffer space used. -EINVAL is returned
241 * if an overlong buffer is specified or a negative buffer size. -EFAULT
242 * is returned if either the buffer or the length field are not
243 * accessible.
244 * After copying the data up to the limit the user specifies, the true
245 * length of the data is written over the length limit the user
246 * specified. Zero is returned for a success.
247 */
248
249int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
250{
251 int err;
252 int len;
253
254 if((err=get_user(len, ulen)))
255 return err;
256 if(len>klen)
257 len=klen;
258 if(len<0 || len> MAX_SOCK_ADDR)
259 return -EINVAL;
260 if(len)
261 {
262 if(copy_to_user(uaddr,kaddr,len))
263 return -EFAULT;
264 }
265 /*
266 * "fromlen shall refer to the value before truncation.."
267 * 1003.1g
268 */
269 return __put_user(klen, ulen);
270}
271
272#define SOCKFS_MAGIC 0x534F434B
273
Eric Dumazetba899662005-08-26 12:05:31 -0700274static kmem_cache_t * sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275
276static struct inode *sock_alloc_inode(struct super_block *sb)
277{
278 struct socket_alloc *ei;
279 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
280 if (!ei)
281 return NULL;
282 init_waitqueue_head(&ei->socket.wait);
283
284 ei->socket.fasync_list = NULL;
285 ei->socket.state = SS_UNCONNECTED;
286 ei->socket.flags = 0;
287 ei->socket.ops = NULL;
288 ei->socket.sk = NULL;
289 ei->socket.file = NULL;
290 ei->socket.flags = 0;
291
292 return &ei->vfs_inode;
293}
294
295static void sock_destroy_inode(struct inode *inode)
296{
297 kmem_cache_free(sock_inode_cachep,
298 container_of(inode, struct socket_alloc, vfs_inode));
299}
300
301static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
302{
303 struct socket_alloc *ei = (struct socket_alloc *) foo;
304
305 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
306 SLAB_CTOR_CONSTRUCTOR)
307 inode_init_once(&ei->vfs_inode);
308}
309
310static int init_inodecache(void)
311{
312 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
313 sizeof(struct socket_alloc),
314 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
315 init_once, NULL);
316 if (sock_inode_cachep == NULL)
317 return -ENOMEM;
318 return 0;
319}
320
321static struct super_operations sockfs_ops = {
322 .alloc_inode = sock_alloc_inode,
323 .destroy_inode =sock_destroy_inode,
324 .statfs = simple_statfs,
325};
326
327static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
328 int flags, const char *dev_name, void *data)
329{
330 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
331}
332
Eric Dumazetba899662005-08-26 12:05:31 -0700333static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334
335static struct file_system_type sock_fs_type = {
336 .name = "sockfs",
337 .get_sb = sockfs_get_sb,
338 .kill_sb = kill_anon_super,
339};
340static int sockfs_delete_dentry(struct dentry *dentry)
341{
342 return 1;
343}
344static struct dentry_operations sockfs_dentry_operations = {
345 .d_delete = sockfs_delete_dentry,
346};
347
348/*
349 * Obtains the first available file descriptor and sets it up for use.
350 *
351 * This function creates file structure and maps it to fd space
352 * of current process. On success it returns file descriptor
353 * and file struct implicitly stored in sock->file.
354 * Note that another thread may close file descriptor before we return
355 * from this function. We use the fact that now we do not refer
356 * to socket after mapping. If one day we will need it, this
357 * function will increment ref. count on file by 1.
358 *
359 * In any case returned fd MAY BE not valid!
360 * This race condition is unavoidable
361 * with shared fd spaces, we cannot solve it inside kernel,
362 * but we take care of internal coherence yet.
363 */
364
365int sock_map_fd(struct socket *sock)
366{
367 int fd;
368 struct qstr this;
369 char name[32];
370
371 /*
372 * Find a file descriptor suitable for return to the user.
373 */
374
375 fd = get_unused_fd();
376 if (fd >= 0) {
377 struct file *file = get_empty_filp();
378
379 if (!file) {
380 put_unused_fd(fd);
381 fd = -ENFILE;
382 goto out;
383 }
384
Eric Dumazetf31f5f02005-06-22 14:32:51 -0700385 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 this.name = name;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 this.hash = SOCK_INODE(sock)->i_ino;
388
389 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
390 if (!file->f_dentry) {
391 put_filp(file);
392 put_unused_fd(fd);
393 fd = -ENOMEM;
394 goto out;
395 }
396 file->f_dentry->d_op = &sockfs_dentry_operations;
397 d_add(file->f_dentry, SOCK_INODE(sock));
398 file->f_vfsmnt = mntget(sock_mnt);
399 file->f_mapping = file->f_dentry->d_inode->i_mapping;
400
401 sock->file = file;
402 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
403 file->f_mode = FMODE_READ | FMODE_WRITE;
404 file->f_flags = O_RDWR;
405 file->f_pos = 0;
Benjamin LaHaise07dc3f02005-08-10 14:16:04 -0700406 file->private_data = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 fd_install(fd, file);
408 }
409
410out:
411 return fd;
412}
413
414/**
415 * sockfd_lookup - Go from a file number to its socket slot
416 * @fd: file handle
417 * @err: pointer to an error code return
418 *
419 * The file handle passed in is locked and the socket it is bound
420 * too is returned. If an error occurs the err pointer is overwritten
421 * with a negative errno code and NULL is returned. The function checks
422 * for both invalid handles and passing a handle which is not a socket.
423 *
424 * On a success the socket object pointer is returned.
425 */
426
427struct socket *sockfd_lookup(int fd, int *err)
428{
429 struct file *file;
430 struct inode *inode;
431 struct socket *sock;
432
433 if (!(file = fget(fd)))
434 {
435 *err = -EBADF;
436 return NULL;
437 }
438
Benjamin LaHaise07dc3f02005-08-10 14:16:04 -0700439 if (file->f_op == &socket_file_ops)
440 return file->private_data; /* set in sock_map_fd */
441
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 inode = file->f_dentry->d_inode;
443 if (!S_ISSOCK(inode->i_mode)) {
444 *err = -ENOTSOCK;
445 fput(file);
446 return NULL;
447 }
448
449 sock = SOCKET_I(inode);
450 if (sock->file != file) {
451 printk(KERN_ERR "socki_lookup: socket file changed!\n");
452 sock->file = file;
453 }
454 return sock;
455}
456
457/**
458 * sock_alloc - allocate a socket
459 *
460 * Allocate a new inode and socket object. The two are bound together
461 * and initialised. The socket is then returned. If we are out of inodes
462 * NULL is returned.
463 */
464
465static struct socket *sock_alloc(void)
466{
467 struct inode * inode;
468 struct socket * sock;
469
470 inode = new_inode(sock_mnt->mnt_sb);
471 if (!inode)
472 return NULL;
473
474 sock = SOCKET_I(inode);
475
476 inode->i_mode = S_IFSOCK|S_IRWXUGO;
477 inode->i_uid = current->fsuid;
478 inode->i_gid = current->fsgid;
479
480 get_cpu_var(sockets_in_use)++;
481 put_cpu_var(sockets_in_use);
482 return sock;
483}
484
485/*
486 * In theory you can't get an open on this inode, but /proc provides
487 * a back door. Remember to keep it shut otherwise you'll let the
488 * creepy crawlies in.
489 */
490
491static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
492{
493 return -ENXIO;
494}
495
496struct file_operations bad_sock_fops = {
497 .owner = THIS_MODULE,
498 .open = sock_no_open,
499};
500
501/**
502 * sock_release - close a socket
503 * @sock: socket to close
504 *
505 * The socket is released from the protocol stack if it has a release
506 * callback, and the inode is then released if the socket is bound to
507 * an inode not a file.
508 */
509
510void sock_release(struct socket *sock)
511{
512 if (sock->ops) {
513 struct module *owner = sock->ops->owner;
514
515 sock->ops->release(sock);
516 sock->ops = NULL;
517 module_put(owner);
518 }
519
520 if (sock->fasync_list)
521 printk(KERN_ERR "sock_release: fasync list not empty!\n");
522
523 get_cpu_var(sockets_in_use)--;
524 put_cpu_var(sockets_in_use);
525 if (!sock->file) {
526 iput(SOCK_INODE(sock));
527 return;
528 }
529 sock->file=NULL;
530}
531
532static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
533 struct msghdr *msg, size_t size)
534{
535 struct sock_iocb *si = kiocb_to_siocb(iocb);
536 int err;
537
538 si->sock = sock;
539 si->scm = NULL;
540 si->msg = msg;
541 si->size = size;
542
543 err = security_socket_sendmsg(sock, msg, size);
544 if (err)
545 return err;
546
547 return sock->ops->sendmsg(iocb, sock, msg, size);
548}
549
550int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
551{
552 struct kiocb iocb;
553 struct sock_iocb siocb;
554 int ret;
555
556 init_sync_kiocb(&iocb, NULL);
557 iocb.private = &siocb;
558 ret = __sock_sendmsg(&iocb, sock, msg, size);
559 if (-EIOCBQUEUED == ret)
560 ret = wait_on_sync_kiocb(&iocb);
561 return ret;
562}
563
564int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
565 struct kvec *vec, size_t num, size_t size)
566{
567 mm_segment_t oldfs = get_fs();
568 int result;
569
570 set_fs(KERNEL_DS);
571 /*
572 * the following is safe, since for compiler definitions of kvec and
573 * iovec are identical, yielding the same in-core layout and alignment
574 */
575 msg->msg_iov = (struct iovec *)vec,
576 msg->msg_iovlen = num;
577 result = sock_sendmsg(sock, msg, size);
578 set_fs(oldfs);
579 return result;
580}
581
582static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
583 struct msghdr *msg, size_t size, int flags)
584{
585 int err;
586 struct sock_iocb *si = kiocb_to_siocb(iocb);
587
588 si->sock = sock;
589 si->scm = NULL;
590 si->msg = msg;
591 si->size = size;
592 si->flags = flags;
593
594 err = security_socket_recvmsg(sock, msg, size, flags);
595 if (err)
596 return err;
597
598 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
599}
600
601int sock_recvmsg(struct socket *sock, struct msghdr *msg,
602 size_t size, int flags)
603{
604 struct kiocb iocb;
605 struct sock_iocb siocb;
606 int ret;
607
608 init_sync_kiocb(&iocb, NULL);
609 iocb.private = &siocb;
610 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
611 if (-EIOCBQUEUED == ret)
612 ret = wait_on_sync_kiocb(&iocb);
613 return ret;
614}
615
616int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
617 struct kvec *vec, size_t num,
618 size_t size, int flags)
619{
620 mm_segment_t oldfs = get_fs();
621 int result;
622
623 set_fs(KERNEL_DS);
624 /*
625 * the following is safe, since for compiler definitions of kvec and
626 * iovec are identical, yielding the same in-core layout and alignment
627 */
628 msg->msg_iov = (struct iovec *)vec,
629 msg->msg_iovlen = num;
630 result = sock_recvmsg(sock, msg, size, flags);
631 set_fs(oldfs);
632 return result;
633}
634
635static void sock_aio_dtor(struct kiocb *iocb)
636{
637 kfree(iocb->private);
638}
639
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300640static ssize_t sock_sendpage(struct file *file, struct page *page,
641 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642{
643 struct socket *sock;
644 int flags;
645
Eric Dumazetb69aee02005-09-06 14:42:45 -0700646 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647
648 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
649 if (more)
650 flags |= MSG_MORE;
651
652 return sock->ops->sendpage(sock, page, offset, size, flags);
653}
654
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800655static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
656 char __user *ubuf, size_t size, struct sock_iocb *siocb)
657{
658 if (!is_sync_kiocb(iocb)) {
659 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
660 if (!siocb)
661 return NULL;
662 iocb->ki_dtor = sock_aio_dtor;
663 }
664
665 siocb->kiocb = iocb;
666 siocb->async_iov.iov_base = ubuf;
667 siocb->async_iov.iov_len = size;
668
669 iocb->private = siocb;
670 return siocb;
671}
672
673static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
674 struct file *file, struct iovec *iov, unsigned long nr_segs)
675{
676 struct socket *sock = file->private_data;
677 size_t size = 0;
678 int i;
679
680 for (i = 0 ; i < nr_segs ; i++)
681 size += iov[i].iov_len;
682
683 msg->msg_name = NULL;
684 msg->msg_namelen = 0;
685 msg->msg_control = NULL;
686 msg->msg_controllen = 0;
687 msg->msg_iov = (struct iovec *) iov;
688 msg->msg_iovlen = nr_segs;
689 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
690
691 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
692}
693
694static ssize_t sock_readv(struct file *file, const struct iovec *iov,
695 unsigned long nr_segs, loff_t *ppos)
696{
697 struct kiocb iocb;
698 struct sock_iocb siocb;
699 struct msghdr msg;
700 int ret;
701
702 init_sync_kiocb(&iocb, NULL);
703 iocb.private = &siocb;
704
705 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
706 if (-EIOCBQUEUED == ret)
707 ret = wait_on_sync_kiocb(&iocb);
708 return ret;
709}
710
711static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
712 size_t count, loff_t pos)
713{
714 struct sock_iocb siocb, *x;
715
716 if (pos != 0)
717 return -ESPIPE;
718 if (count == 0) /* Match SYS5 behaviour */
719 return 0;
720
721 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
722 if (!x)
723 return -ENOMEM;
724 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
725 &x->async_iov, 1);
726}
727
728static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
729 struct file *file, struct iovec *iov, unsigned long nr_segs)
730{
731 struct socket *sock = file->private_data;
732 size_t size = 0;
733 int i;
734
735 for (i = 0 ; i < nr_segs ; i++)
736 size += iov[i].iov_len;
737
738 msg->msg_name = NULL;
739 msg->msg_namelen = 0;
740 msg->msg_control = NULL;
741 msg->msg_controllen = 0;
742 msg->msg_iov = (struct iovec *) iov;
743 msg->msg_iovlen = nr_segs;
744 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
745 if (sock->type == SOCK_SEQPACKET)
746 msg->msg_flags |= MSG_EOR;
747
748 return __sock_sendmsg(iocb, sock, msg, size);
749}
750
751static ssize_t sock_writev(struct file *file, const struct iovec *iov,
752 unsigned long nr_segs, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753{
754 struct msghdr msg;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800755 struct kiocb iocb;
756 struct sock_iocb siocb;
757 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800759 init_sync_kiocb(&iocb, NULL);
760 iocb.private = &siocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800762 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
763 if (-EIOCBQUEUED == ret)
764 ret = wait_on_sync_kiocb(&iocb);
765 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766}
767
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800768static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
769 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800771 struct sock_iocb siocb, *x;
772
773 if (pos != 0)
774 return -ESPIPE;
775 if (count == 0) /* Match SYS5 behaviour */
776 return 0;
777
778 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
779 if (!x)
780 return -ENOMEM;
781
782 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
783 &x->async_iov, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784}
785
786
787/*
788 * Atomic setting of ioctl hooks to avoid race
789 * with module unload.
790 */
791
792static DECLARE_MUTEX(br_ioctl_mutex);
793static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
794
795void brioctl_set(int (*hook)(unsigned int, void __user *))
796{
797 down(&br_ioctl_mutex);
798 br_ioctl_hook = hook;
799 up(&br_ioctl_mutex);
800}
801EXPORT_SYMBOL(brioctl_set);
802
803static DECLARE_MUTEX(vlan_ioctl_mutex);
804static int (*vlan_ioctl_hook)(void __user *arg);
805
806void vlan_ioctl_set(int (*hook)(void __user *))
807{
808 down(&vlan_ioctl_mutex);
809 vlan_ioctl_hook = hook;
810 up(&vlan_ioctl_mutex);
811}
812EXPORT_SYMBOL(vlan_ioctl_set);
813
814static DECLARE_MUTEX(dlci_ioctl_mutex);
815static int (*dlci_ioctl_hook)(unsigned int, void __user *);
816
817void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
818{
819 down(&dlci_ioctl_mutex);
820 dlci_ioctl_hook = hook;
821 up(&dlci_ioctl_mutex);
822}
823EXPORT_SYMBOL(dlci_ioctl_set);
824
825/*
826 * With an ioctl, arg may well be a user mode pointer, but we don't know
827 * what to do with it - that's up to the protocol still.
828 */
829
830static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
831{
832 struct socket *sock;
833 void __user *argp = (void __user *)arg;
834 int pid, err;
835
Eric Dumazetb69aee02005-09-06 14:42:45 -0700836 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
838 err = dev_ioctl(cmd, argp);
839 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100840#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
842 err = dev_ioctl(cmd, argp);
843 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100844#endif /* CONFIG_WIRELESS_EXT */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 switch (cmd) {
846 case FIOSETOWN:
847 case SIOCSPGRP:
848 err = -EFAULT;
849 if (get_user(pid, (int __user *)argp))
850 break;
851 err = f_setown(sock->file, pid, 1);
852 break;
853 case FIOGETOWN:
854 case SIOCGPGRP:
855 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
856 break;
857 case SIOCGIFBR:
858 case SIOCSIFBR:
859 case SIOCBRADDBR:
860 case SIOCBRDELBR:
861 err = -ENOPKG;
862 if (!br_ioctl_hook)
863 request_module("bridge");
864
865 down(&br_ioctl_mutex);
866 if (br_ioctl_hook)
867 err = br_ioctl_hook(cmd, argp);
868 up(&br_ioctl_mutex);
869 break;
870 case SIOCGIFVLAN:
871 case SIOCSIFVLAN:
872 err = -ENOPKG;
873 if (!vlan_ioctl_hook)
874 request_module("8021q");
875
876 down(&vlan_ioctl_mutex);
877 if (vlan_ioctl_hook)
878 err = vlan_ioctl_hook(argp);
879 up(&vlan_ioctl_mutex);
880 break;
881 case SIOCGIFDIVERT:
882 case SIOCSIFDIVERT:
883 /* Convert this to call through a hook */
884 err = divert_ioctl(cmd, argp);
885 break;
886 case SIOCADDDLCI:
887 case SIOCDELDLCI:
888 err = -ENOPKG;
889 if (!dlci_ioctl_hook)
890 request_module("dlci");
891
892 if (dlci_ioctl_hook) {
893 down(&dlci_ioctl_mutex);
894 err = dlci_ioctl_hook(cmd, argp);
895 up(&dlci_ioctl_mutex);
896 }
897 break;
898 default:
899 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800900
901 /*
902 * If this ioctl is unknown try to hand it down
903 * to the NIC driver.
904 */
905 if (err == -ENOIOCTLCMD)
906 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 break;
908 }
909 return err;
910}
911
912int sock_create_lite(int family, int type, int protocol, struct socket **res)
913{
914 int err;
915 struct socket *sock = NULL;
916
917 err = security_socket_create(family, type, protocol, 1);
918 if (err)
919 goto out;
920
921 sock = sock_alloc();
922 if (!sock) {
923 err = -ENOMEM;
924 goto out;
925 }
926
927 security_socket_post_create(sock, family, type, protocol, 1);
928 sock->type = type;
929out:
930 *res = sock;
931 return err;
932}
933
934/* No kernel lock held - perfect */
935static unsigned int sock_poll(struct file *file, poll_table * wait)
936{
937 struct socket *sock;
938
939 /*
940 * We can't return errors to poll, so it's either yes or no.
941 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700942 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 return sock->ops->poll(file, sock, wait);
944}
945
946static int sock_mmap(struct file * file, struct vm_area_struct * vma)
947{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700948 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
950 return sock->ops->mmap(file, sock, vma);
951}
952
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300953static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954{
955 /*
956 * It was possible the inode is NULL we were
957 * closing an unfinished socket.
958 */
959
960 if (!inode)
961 {
962 printk(KERN_DEBUG "sock_close: NULL inode\n");
963 return 0;
964 }
965 sock_fasync(-1, filp, 0);
966 sock_release(SOCKET_I(inode));
967 return 0;
968}
969
970/*
971 * Update the socket async list
972 *
973 * Fasync_list locking strategy.
974 *
975 * 1. fasync_list is modified only under process context socket lock
976 * i.e. under semaphore.
977 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
978 * or under socket lock.
979 * 3. fasync_list can be used from softirq context, so that
980 * modification under socket lock have to be enhanced with
981 * write_lock_bh(&sk->sk_callback_lock).
982 * --ANK (990710)
983 */
984
985static int sock_fasync(int fd, struct file *filp, int on)
986{
987 struct fasync_struct *fa, *fna=NULL, **prev;
988 struct socket *sock;
989 struct sock *sk;
990
991 if (on)
992 {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -0800993 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 if(fna==NULL)
995 return -ENOMEM;
996 }
997
Eric Dumazetb69aee02005-09-06 14:42:45 -0700998 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000 if ((sk=sock->sk) == NULL) {
1001 kfree(fna);
1002 return -EINVAL;
1003 }
1004
1005 lock_sock(sk);
1006
1007 prev=&(sock->fasync_list);
1008
1009 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1010 if (fa->fa_file==filp)
1011 break;
1012
1013 if(on)
1014 {
1015 if(fa!=NULL)
1016 {
1017 write_lock_bh(&sk->sk_callback_lock);
1018 fa->fa_fd=fd;
1019 write_unlock_bh(&sk->sk_callback_lock);
1020
1021 kfree(fna);
1022 goto out;
1023 }
1024 fna->fa_file=filp;
1025 fna->fa_fd=fd;
1026 fna->magic=FASYNC_MAGIC;
1027 fna->fa_next=sock->fasync_list;
1028 write_lock_bh(&sk->sk_callback_lock);
1029 sock->fasync_list=fna;
1030 write_unlock_bh(&sk->sk_callback_lock);
1031 }
1032 else
1033 {
1034 if (fa!=NULL)
1035 {
1036 write_lock_bh(&sk->sk_callback_lock);
1037 *prev=fa->fa_next;
1038 write_unlock_bh(&sk->sk_callback_lock);
1039 kfree(fa);
1040 }
1041 }
1042
1043out:
1044 release_sock(sock->sk);
1045 return 0;
1046}
1047
1048/* This function may be called only under socket lock or callback_lock */
1049
1050int sock_wake_async(struct socket *sock, int how, int band)
1051{
1052 if (!sock || !sock->fasync_list)
1053 return -1;
1054 switch (how)
1055 {
1056 case 1:
1057
1058 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1059 break;
1060 goto call_kill;
1061 case 2:
1062 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1063 break;
1064 /* fall through */
1065 case 0:
1066 call_kill:
1067 __kill_fasync(sock->fasync_list, SIGIO, band);
1068 break;
1069 case 3:
1070 __kill_fasync(sock->fasync_list, SIGURG, band);
1071 }
1072 return 0;
1073}
1074
1075static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1076{
1077 int err;
1078 struct socket *sock;
1079
1080 /*
1081 * Check protocol is in range
1082 */
1083 if (family < 0 || family >= NPROTO)
1084 return -EAFNOSUPPORT;
1085 if (type < 0 || type >= SOCK_MAX)
1086 return -EINVAL;
1087
1088 /* Compatibility.
1089
1090 This uglymoron is moved from INET layer to here to avoid
1091 deadlock in module load.
1092 */
1093 if (family == PF_INET && type == SOCK_PACKET) {
1094 static int warned;
1095 if (!warned) {
1096 warned = 1;
1097 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1098 }
1099 family = PF_PACKET;
1100 }
1101
1102 err = security_socket_create(family, type, protocol, kern);
1103 if (err)
1104 return err;
1105
1106#if defined(CONFIG_KMOD)
1107 /* Attempt to load a protocol module if the find failed.
1108 *
1109 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1110 * requested real, full-featured networking support upon configuration.
1111 * Otherwise module support will break!
1112 */
1113 if (net_families[family]==NULL)
1114 {
1115 request_module("net-pf-%d",family);
1116 }
1117#endif
1118
1119 net_family_read_lock();
1120 if (net_families[family] == NULL) {
1121 err = -EAFNOSUPPORT;
1122 goto out;
1123 }
1124
1125/*
1126 * Allocate the socket and allow the family to set things up. if
1127 * the protocol is 0, the family is instructed to select an appropriate
1128 * default.
1129 */
1130
1131 if (!(sock = sock_alloc())) {
1132 printk(KERN_WARNING "socket: no more sockets\n");
1133 err = -ENFILE; /* Not exactly a match, but its the
1134 closest posix thing */
1135 goto out;
1136 }
1137
1138 sock->type = type;
1139
1140 /*
1141 * We will call the ->create function, that possibly is in a loadable
1142 * module, so we have to bump that loadable module refcnt first.
1143 */
1144 err = -EAFNOSUPPORT;
1145 if (!try_module_get(net_families[family]->owner))
1146 goto out_release;
1147
Frank Filza79af592005-09-27 15:23:38 -07001148 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1149 sock->ops = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001151 }
1152
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 /*
1154 * Now to bump the refcnt of the [loadable] module that owns this
1155 * socket at sock_release time we decrement its refcnt.
1156 */
1157 if (!try_module_get(sock->ops->owner)) {
1158 sock->ops = NULL;
1159 goto out_module_put;
1160 }
1161 /*
1162 * Now that we're done with the ->create function, the [loadable]
1163 * module can have its refcnt decremented
1164 */
1165 module_put(net_families[family]->owner);
1166 *res = sock;
1167 security_socket_post_create(sock, family, type, protocol, kern);
1168
1169out:
1170 net_family_read_unlock();
1171 return err;
1172out_module_put:
1173 module_put(net_families[family]->owner);
1174out_release:
1175 sock_release(sock);
1176 goto out;
1177}
1178
1179int sock_create(int family, int type, int protocol, struct socket **res)
1180{
1181 return __sock_create(family, type, protocol, res, 0);
1182}
1183
1184int sock_create_kern(int family, int type, int protocol, struct socket **res)
1185{
1186 return __sock_create(family, type, protocol, res, 1);
1187}
1188
1189asmlinkage long sys_socket(int family, int type, int protocol)
1190{
1191 int retval;
1192 struct socket *sock;
1193
1194 retval = sock_create(family, type, protocol, &sock);
1195 if (retval < 0)
1196 goto out;
1197
1198 retval = sock_map_fd(sock);
1199 if (retval < 0)
1200 goto out_release;
1201
1202out:
1203 /* It may be already another descriptor 8) Not kernel problem. */
1204 return retval;
1205
1206out_release:
1207 sock_release(sock);
1208 return retval;
1209}
1210
1211/*
1212 * Create a pair of connected sockets.
1213 */
1214
1215asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1216{
1217 struct socket *sock1, *sock2;
1218 int fd1, fd2, err;
1219
1220 /*
1221 * Obtain the first socket and check if the underlying protocol
1222 * supports the socketpair call.
1223 */
1224
1225 err = sock_create(family, type, protocol, &sock1);
1226 if (err < 0)
1227 goto out;
1228
1229 err = sock_create(family, type, protocol, &sock2);
1230 if (err < 0)
1231 goto out_release_1;
1232
1233 err = sock1->ops->socketpair(sock1, sock2);
1234 if (err < 0)
1235 goto out_release_both;
1236
1237 fd1 = fd2 = -1;
1238
1239 err = sock_map_fd(sock1);
1240 if (err < 0)
1241 goto out_release_both;
1242 fd1 = err;
1243
1244 err = sock_map_fd(sock2);
1245 if (err < 0)
1246 goto out_close_1;
1247 fd2 = err;
1248
1249 /* fd1 and fd2 may be already another descriptors.
1250 * Not kernel problem.
1251 */
1252
1253 err = put_user(fd1, &usockvec[0]);
1254 if (!err)
1255 err = put_user(fd2, &usockvec[1]);
1256 if (!err)
1257 return 0;
1258
1259 sys_close(fd2);
1260 sys_close(fd1);
1261 return err;
1262
1263out_close_1:
1264 sock_release(sock2);
1265 sys_close(fd1);
1266 return err;
1267
1268out_release_both:
1269 sock_release(sock2);
1270out_release_1:
1271 sock_release(sock1);
1272out:
1273 return err;
1274}
1275
1276
1277/*
1278 * Bind a name to a socket. Nothing much to do here since it's
1279 * the protocol's responsibility to handle the local address.
1280 *
1281 * We move the socket address to kernel space before we call
1282 * the protocol layer (having also checked the address is ok).
1283 */
1284
1285asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1286{
1287 struct socket *sock;
1288 char address[MAX_SOCK_ADDR];
1289 int err;
1290
1291 if((sock = sockfd_lookup(fd,&err))!=NULL)
1292 {
1293 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1294 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1295 if (err) {
1296 sockfd_put(sock);
1297 return err;
1298 }
1299 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1300 }
1301 sockfd_put(sock);
1302 }
1303 return err;
1304}
1305
1306
1307/*
1308 * Perform a listen. Basically, we allow the protocol to do anything
1309 * necessary for a listen, and if that works, we mark the socket as
1310 * ready for listening.
1311 */
1312
1313int sysctl_somaxconn = SOMAXCONN;
1314
1315asmlinkage long sys_listen(int fd, int backlog)
1316{
1317 struct socket *sock;
1318 int err;
1319
1320 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1321 if ((unsigned) backlog > sysctl_somaxconn)
1322 backlog = sysctl_somaxconn;
1323
1324 err = security_socket_listen(sock, backlog);
1325 if (err) {
1326 sockfd_put(sock);
1327 return err;
1328 }
1329
1330 err=sock->ops->listen(sock, backlog);
1331 sockfd_put(sock);
1332 }
1333 return err;
1334}
1335
1336
1337/*
1338 * For accept, we attempt to create a new socket, set up the link
1339 * with the client, wake up the client, then return the new
1340 * connected fd. We collect the address of the connector in kernel
1341 * space and move it to user at the very end. This is unclean because
1342 * we open the socket then return an error.
1343 *
1344 * 1003.1g adds the ability to recvmsg() to query connection pending
1345 * status to recvmsg. We need to add that support in a way thats
1346 * clean when we restucture accept also.
1347 */
1348
1349asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1350{
1351 struct socket *sock, *newsock;
1352 int err, len;
1353 char address[MAX_SOCK_ADDR];
1354
1355 sock = sockfd_lookup(fd, &err);
1356 if (!sock)
1357 goto out;
1358
1359 err = -ENFILE;
1360 if (!(newsock = sock_alloc()))
1361 goto out_put;
1362
1363 newsock->type = sock->type;
1364 newsock->ops = sock->ops;
1365
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 /*
1367 * We don't need try_module_get here, as the listening socket (sock)
1368 * has the protocol module (sock->ops->owner) held.
1369 */
1370 __module_get(newsock->ops->owner);
1371
Frank Filza79af592005-09-27 15:23:38 -07001372 err = security_socket_accept(sock, newsock);
1373 if (err)
1374 goto out_release;
1375
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1377 if (err < 0)
1378 goto out_release;
1379
1380 if (upeer_sockaddr) {
1381 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1382 err = -ECONNABORTED;
1383 goto out_release;
1384 }
1385 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1386 if (err < 0)
1387 goto out_release;
1388 }
1389
1390 /* File flags are not inherited via accept() unlike another OSes. */
1391
1392 if ((err = sock_map_fd(newsock)) < 0)
1393 goto out_release;
1394
1395 security_socket_post_accept(sock, newsock);
1396
1397out_put:
1398 sockfd_put(sock);
1399out:
1400 return err;
1401out_release:
1402 sock_release(newsock);
1403 goto out_put;
1404}
1405
1406
1407/*
1408 * Attempt to connect to a socket with the server address. The address
1409 * is in user space so we verify it is OK and move it to kernel space.
1410 *
1411 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1412 * break bindings
1413 *
1414 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1415 * other SEQPACKET protocols that take time to connect() as it doesn't
1416 * include the -EINPROGRESS status for such sockets.
1417 */
1418
1419asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1420{
1421 struct socket *sock;
1422 char address[MAX_SOCK_ADDR];
1423 int err;
1424
1425 sock = sockfd_lookup(fd, &err);
1426 if (!sock)
1427 goto out;
1428 err = move_addr_to_kernel(uservaddr, addrlen, address);
1429 if (err < 0)
1430 goto out_put;
1431
1432 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1433 if (err)
1434 goto out_put;
1435
1436 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1437 sock->file->f_flags);
1438out_put:
1439 sockfd_put(sock);
1440out:
1441 return err;
1442}
1443
1444/*
1445 * Get the local address ('name') of a socket object. Move the obtained
1446 * name to user space.
1447 */
1448
1449asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1450{
1451 struct socket *sock;
1452 char address[MAX_SOCK_ADDR];
1453 int len, err;
1454
1455 sock = sockfd_lookup(fd, &err);
1456 if (!sock)
1457 goto out;
1458
1459 err = security_socket_getsockname(sock);
1460 if (err)
1461 goto out_put;
1462
1463 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1464 if (err)
1465 goto out_put;
1466 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1467
1468out_put:
1469 sockfd_put(sock);
1470out:
1471 return err;
1472}
1473
1474/*
1475 * Get the remote address ('name') of a socket object. Move the obtained
1476 * name to user space.
1477 */
1478
1479asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1480{
1481 struct socket *sock;
1482 char address[MAX_SOCK_ADDR];
1483 int len, err;
1484
1485 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1486 {
1487 err = security_socket_getpeername(sock);
1488 if (err) {
1489 sockfd_put(sock);
1490 return err;
1491 }
1492
1493 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1494 if (!err)
1495 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1496 sockfd_put(sock);
1497 }
1498 return err;
1499}
1500
1501/*
1502 * Send a datagram to a given address. We move the address into kernel
1503 * space and check the user space data area is readable before invoking
1504 * the protocol.
1505 */
1506
1507asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1508 struct sockaddr __user *addr, int addr_len)
1509{
1510 struct socket *sock;
1511 char address[MAX_SOCK_ADDR];
1512 int err;
1513 struct msghdr msg;
1514 struct iovec iov;
1515
1516 sock = sockfd_lookup(fd, &err);
1517 if (!sock)
1518 goto out;
1519 iov.iov_base=buff;
1520 iov.iov_len=len;
1521 msg.msg_name=NULL;
1522 msg.msg_iov=&iov;
1523 msg.msg_iovlen=1;
1524 msg.msg_control=NULL;
1525 msg.msg_controllen=0;
1526 msg.msg_namelen=0;
1527 if(addr)
1528 {
1529 err = move_addr_to_kernel(addr, addr_len, address);
1530 if (err < 0)
1531 goto out_put;
1532 msg.msg_name=address;
1533 msg.msg_namelen=addr_len;
1534 }
1535 if (sock->file->f_flags & O_NONBLOCK)
1536 flags |= MSG_DONTWAIT;
1537 msg.msg_flags = flags;
1538 err = sock_sendmsg(sock, &msg, len);
1539
1540out_put:
1541 sockfd_put(sock);
1542out:
1543 return err;
1544}
1545
1546/*
1547 * Send a datagram down a socket.
1548 */
1549
1550asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1551{
1552 return sys_sendto(fd, buff, len, flags, NULL, 0);
1553}
1554
1555/*
1556 * Receive a frame from the socket and optionally record the address of the
1557 * sender. We verify the buffers are writable and if needed move the
1558 * sender address from kernel to user space.
1559 */
1560
1561asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1562 struct sockaddr __user *addr, int __user *addr_len)
1563{
1564 struct socket *sock;
1565 struct iovec iov;
1566 struct msghdr msg;
1567 char address[MAX_SOCK_ADDR];
1568 int err,err2;
1569
1570 sock = sockfd_lookup(fd, &err);
1571 if (!sock)
1572 goto out;
1573
1574 msg.msg_control=NULL;
1575 msg.msg_controllen=0;
1576 msg.msg_iovlen=1;
1577 msg.msg_iov=&iov;
1578 iov.iov_len=size;
1579 iov.iov_base=ubuf;
1580 msg.msg_name=address;
1581 msg.msg_namelen=MAX_SOCK_ADDR;
1582 if (sock->file->f_flags & O_NONBLOCK)
1583 flags |= MSG_DONTWAIT;
1584 err=sock_recvmsg(sock, &msg, size, flags);
1585
1586 if(err >= 0 && addr != NULL)
1587 {
1588 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1589 if(err2<0)
1590 err=err2;
1591 }
1592 sockfd_put(sock);
1593out:
1594 return err;
1595}
1596
1597/*
1598 * Receive a datagram from a socket.
1599 */
1600
1601asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1602{
1603 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1604}
1605
1606/*
1607 * Set a socket option. Because we don't know the option lengths we have
1608 * to pass the user mode parameter for the protocols to sort out.
1609 */
1610
1611asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1612{
1613 int err;
1614 struct socket *sock;
1615
1616 if (optlen < 0)
1617 return -EINVAL;
1618
1619 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1620 {
1621 err = security_socket_setsockopt(sock,level,optname);
1622 if (err) {
1623 sockfd_put(sock);
1624 return err;
1625 }
1626
1627 if (level == SOL_SOCKET)
1628 err=sock_setsockopt(sock,level,optname,optval,optlen);
1629 else
1630 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1631 sockfd_put(sock);
1632 }
1633 return err;
1634}
1635
1636/*
1637 * Get a socket option. Because we don't know the option lengths we have
1638 * to pass a user mode parameter for the protocols to sort out.
1639 */
1640
1641asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1642{
1643 int err;
1644 struct socket *sock;
1645
1646 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1647 {
1648 err = security_socket_getsockopt(sock, level,
1649 optname);
1650 if (err) {
1651 sockfd_put(sock);
1652 return err;
1653 }
1654
1655 if (level == SOL_SOCKET)
1656 err=sock_getsockopt(sock,level,optname,optval,optlen);
1657 else
1658 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1659 sockfd_put(sock);
1660 }
1661 return err;
1662}
1663
1664
1665/*
1666 * Shutdown a socket.
1667 */
1668
1669asmlinkage long sys_shutdown(int fd, int how)
1670{
1671 int err;
1672 struct socket *sock;
1673
1674 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1675 {
1676 err = security_socket_shutdown(sock, how);
1677 if (err) {
1678 sockfd_put(sock);
1679 return err;
1680 }
1681
1682 err=sock->ops->shutdown(sock, how);
1683 sockfd_put(sock);
1684 }
1685 return err;
1686}
1687
1688/* A couple of helpful macros for getting the address of the 32/64 bit
1689 * fields which are the same type (int / unsigned) on our platforms.
1690 */
1691#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1692#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1693#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1694
1695
1696/*
1697 * BSD sendmsg interface
1698 */
1699
1700asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1701{
1702 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1703 struct socket *sock;
1704 char address[MAX_SOCK_ADDR];
1705 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001706 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1707 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1708 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 unsigned char *ctl_buf = ctl;
1710 struct msghdr msg_sys;
1711 int err, ctl_len, iov_size, total_len;
1712
1713 err = -EFAULT;
1714 if (MSG_CMSG_COMPAT & flags) {
1715 if (get_compat_msghdr(&msg_sys, msg_compat))
1716 return -EFAULT;
1717 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1718 return -EFAULT;
1719
1720 sock = sockfd_lookup(fd, &err);
1721 if (!sock)
1722 goto out;
1723
1724 /* do not move before msg_sys is valid */
1725 err = -EMSGSIZE;
1726 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1727 goto out_put;
1728
1729 /* Check whether to allocate the iovec area*/
1730 err = -ENOMEM;
1731 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1732 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1733 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1734 if (!iov)
1735 goto out_put;
1736 }
1737
1738 /* This will also move the address data into kernel space */
1739 if (MSG_CMSG_COMPAT & flags) {
1740 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1741 } else
1742 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1743 if (err < 0)
1744 goto out_freeiov;
1745 total_len = err;
1746
1747 err = -ENOBUFS;
1748
1749 if (msg_sys.msg_controllen > INT_MAX)
1750 goto out_freeiov;
1751 ctl_len = msg_sys.msg_controllen;
1752 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Al Viro8920e8f2005-09-07 18:28:51 -07001753 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 if (err)
1755 goto out_freeiov;
1756 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001757 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 } else if (ctl_len) {
1759 if (ctl_len > sizeof(ctl))
1760 {
1761 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1762 if (ctl_buf == NULL)
1763 goto out_freeiov;
1764 }
1765 err = -EFAULT;
1766 /*
1767 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1768 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1769 * checking falls down on this.
1770 */
1771 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1772 goto out_freectl;
1773 msg_sys.msg_control = ctl_buf;
1774 }
1775 msg_sys.msg_flags = flags;
1776
1777 if (sock->file->f_flags & O_NONBLOCK)
1778 msg_sys.msg_flags |= MSG_DONTWAIT;
1779 err = sock_sendmsg(sock, &msg_sys, total_len);
1780
1781out_freectl:
1782 if (ctl_buf != ctl)
1783 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1784out_freeiov:
1785 if (iov != iovstack)
1786 sock_kfree_s(sock->sk, iov, iov_size);
1787out_put:
1788 sockfd_put(sock);
1789out:
1790 return err;
1791}
1792
1793/*
1794 * BSD recvmsg interface
1795 */
1796
1797asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1798{
1799 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1800 struct socket *sock;
1801 struct iovec iovstack[UIO_FASTIOV];
1802 struct iovec *iov=iovstack;
1803 struct msghdr msg_sys;
1804 unsigned long cmsg_ptr;
1805 int err, iov_size, total_len, len;
1806
1807 /* kernel mode address */
1808 char addr[MAX_SOCK_ADDR];
1809
1810 /* user mode address pointers */
1811 struct sockaddr __user *uaddr;
1812 int __user *uaddr_len;
1813
1814 if (MSG_CMSG_COMPAT & flags) {
1815 if (get_compat_msghdr(&msg_sys, msg_compat))
1816 return -EFAULT;
1817 } else
1818 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1819 return -EFAULT;
1820
1821 sock = sockfd_lookup(fd, &err);
1822 if (!sock)
1823 goto out;
1824
1825 err = -EMSGSIZE;
1826 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1827 goto out_put;
1828
1829 /* Check whether to allocate the iovec area*/
1830 err = -ENOMEM;
1831 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1832 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1833 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1834 if (!iov)
1835 goto out_put;
1836 }
1837
1838 /*
1839 * Save the user-mode address (verify_iovec will change the
1840 * kernel msghdr to use the kernel address space)
1841 */
1842
1843 uaddr = (void __user *) msg_sys.msg_name;
1844 uaddr_len = COMPAT_NAMELEN(msg);
1845 if (MSG_CMSG_COMPAT & flags) {
1846 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1847 } else
1848 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1849 if (err < 0)
1850 goto out_freeiov;
1851 total_len=err;
1852
1853 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1854 msg_sys.msg_flags = 0;
1855 if (MSG_CMSG_COMPAT & flags)
1856 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1857
1858 if (sock->file->f_flags & O_NONBLOCK)
1859 flags |= MSG_DONTWAIT;
1860 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1861 if (err < 0)
1862 goto out_freeiov;
1863 len = err;
1864
1865 if (uaddr != NULL) {
1866 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1867 if (err < 0)
1868 goto out_freeiov;
1869 }
David S. Miller37f7f422005-09-16 16:51:01 -07001870 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1871 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 if (err)
1873 goto out_freeiov;
1874 if (MSG_CMSG_COMPAT & flags)
1875 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1876 &msg_compat->msg_controllen);
1877 else
1878 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1879 &msg->msg_controllen);
1880 if (err)
1881 goto out_freeiov;
1882 err = len;
1883
1884out_freeiov:
1885 if (iov != iovstack)
1886 sock_kfree_s(sock->sk, iov, iov_size);
1887out_put:
1888 sockfd_put(sock);
1889out:
1890 return err;
1891}
1892
1893#ifdef __ARCH_WANT_SYS_SOCKETCALL
1894
1895/* Argument list sizes for sys_socketcall */
1896#define AL(x) ((x) * sizeof(unsigned long))
1897static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1898 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1899 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1900#undef AL
1901
1902/*
1903 * System call vectors.
1904 *
1905 * Argument checking cleaned up. Saved 20% in size.
1906 * This function doesn't need to set the kernel lock because
1907 * it is set by the callees.
1908 */
1909
1910asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1911{
1912 unsigned long a[6];
1913 unsigned long a0,a1;
1914 int err;
1915
1916 if(call<1||call>SYS_RECVMSG)
1917 return -EINVAL;
1918
1919 /* copy_from_user should be SMP safe. */
1920 if (copy_from_user(a, args, nargs[call]))
1921 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001922
David Woodhouse4bcff1b2005-06-02 12:13:21 +01001923 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001924 if (err)
1925 return err;
1926
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 a0=a[0];
1928 a1=a[1];
1929
1930 switch(call)
1931 {
1932 case SYS_SOCKET:
1933 err = sys_socket(a0,a1,a[2]);
1934 break;
1935 case SYS_BIND:
1936 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1937 break;
1938 case SYS_CONNECT:
1939 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1940 break;
1941 case SYS_LISTEN:
1942 err = sys_listen(a0,a1);
1943 break;
1944 case SYS_ACCEPT:
1945 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1946 break;
1947 case SYS_GETSOCKNAME:
1948 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1949 break;
1950 case SYS_GETPEERNAME:
1951 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1952 break;
1953 case SYS_SOCKETPAIR:
1954 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1955 break;
1956 case SYS_SEND:
1957 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1958 break;
1959 case SYS_SENDTO:
1960 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
1961 (struct sockaddr __user *)a[4], a[5]);
1962 break;
1963 case SYS_RECV:
1964 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
1965 break;
1966 case SYS_RECVFROM:
1967 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
1968 (struct sockaddr __user *)a[4], (int __user *)a[5]);
1969 break;
1970 case SYS_SHUTDOWN:
1971 err = sys_shutdown(a0,a1);
1972 break;
1973 case SYS_SETSOCKOPT:
1974 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
1975 break;
1976 case SYS_GETSOCKOPT:
1977 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
1978 break;
1979 case SYS_SENDMSG:
1980 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
1981 break;
1982 case SYS_RECVMSG:
1983 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
1984 break;
1985 default:
1986 err = -EINVAL;
1987 break;
1988 }
1989 return err;
1990}
1991
1992#endif /* __ARCH_WANT_SYS_SOCKETCALL */
1993
1994/*
1995 * This function is called by a protocol handler that wants to
1996 * advertise its address family, and have it linked into the
1997 * SOCKET module.
1998 */
1999
2000int sock_register(struct net_proto_family *ops)
2001{
2002 int err;
2003
2004 if (ops->family >= NPROTO) {
2005 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2006 return -ENOBUFS;
2007 }
2008 net_family_write_lock();
2009 err = -EEXIST;
2010 if (net_families[ops->family] == NULL) {
2011 net_families[ops->family]=ops;
2012 err = 0;
2013 }
2014 net_family_write_unlock();
2015 printk(KERN_INFO "NET: Registered protocol family %d\n",
2016 ops->family);
2017 return err;
2018}
2019
2020/*
2021 * This function is called by a protocol handler that wants to
2022 * remove its address family, and have it unlinked from the
2023 * SOCKET module.
2024 */
2025
2026int sock_unregister(int family)
2027{
2028 if (family < 0 || family >= NPROTO)
2029 return -1;
2030
2031 net_family_write_lock();
2032 net_families[family]=NULL;
2033 net_family_write_unlock();
2034 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2035 family);
2036 return 0;
2037}
2038
Andi Kleen77d76ea2005-12-22 12:43:42 -08002039static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040{
2041 /*
2042 * Initialize sock SLAB cache.
2043 */
2044
2045 sk_init();
2046
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 /*
2048 * Initialize skbuff SLAB cache
2049 */
2050 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051
2052 /*
2053 * Initialize the protocols module.
2054 */
2055
2056 init_inodecache();
2057 register_filesystem(&sock_fs_type);
2058 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002059
2060 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 */
2062
2063#ifdef CONFIG_NETFILTER
2064 netfilter_init();
2065#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002066
2067 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068}
2069
Andi Kleen77d76ea2005-12-22 12:43:42 -08002070core_initcall(sock_init); /* early initcall */
2071
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072#ifdef CONFIG_PROC_FS
2073void socket_seq_show(struct seq_file *seq)
2074{
2075 int cpu;
2076 int counter = 0;
2077
2078 for (cpu = 0; cpu < NR_CPUS; cpu++)
2079 counter += per_cpu(sockets_in_use, cpu);
2080
2081 /* It can be negative, by the way. 8) */
2082 if (counter < 0)
2083 counter = 0;
2084
2085 seq_printf(seq, "sockets: used %d\n", counter);
2086}
2087#endif /* CONFIG_PROC_FS */
2088
2089/* ABI emulation layers need these two */
2090EXPORT_SYMBOL(move_addr_to_kernel);
2091EXPORT_SYMBOL(move_addr_to_user);
2092EXPORT_SYMBOL(sock_create);
2093EXPORT_SYMBOL(sock_create_kern);
2094EXPORT_SYMBOL(sock_create_lite);
2095EXPORT_SYMBOL(sock_map_fd);
2096EXPORT_SYMBOL(sock_recvmsg);
2097EXPORT_SYMBOL(sock_register);
2098EXPORT_SYMBOL(sock_release);
2099EXPORT_SYMBOL(sock_sendmsg);
2100EXPORT_SYMBOL(sock_unregister);
2101EXPORT_SYMBOL(sock_wake_async);
2102EXPORT_SYMBOL(sockfd_lookup);
2103EXPORT_SYMBOL(kernel_sendmsg);
2104EXPORT_SYMBOL(kernel_recvmsg);