blob: 74283610db159f41bfd4f7b8d20c7b0cbfff075d [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 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800351 * These functions create file structures and maps them to fd space
352 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 * 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
David S. Miller39d8c1b2006-03-20 17:13:49 -0800365static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366{
367 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800368
369 fd = get_unused_fd();
370 if (likely(fd >= 0)) {
371 struct file *file = get_empty_filp();
372
373 *filep = file;
374 if (unlikely(!file)) {
375 put_unused_fd(fd);
376 return -ENFILE;
377 }
378 } else
379 *filep = NULL;
380 return fd;
381}
382
383static int sock_attach_fd(struct socket *sock, struct file *file)
384{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700385 struct qstr this;
386 char name[32];
387
David S. Miller39d8c1b2006-03-20 17:13:49 -0800388 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
389 this.name = name;
390 this.hash = SOCK_INODE(sock)->i_ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
393 if (unlikely(!file->f_dentry))
394 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395
David S. Miller39d8c1b2006-03-20 17:13:49 -0800396 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;
406 file->private_data = sock;
407
408 return 0;
409}
410
411int sock_map_fd(struct socket *sock)
412{
413 struct file *newfile;
414 int fd = sock_alloc_fd(&newfile);
415
416 if (likely(fd >= 0)) {
417 int err = sock_attach_fd(sock, newfile);
418
419 if (unlikely(err < 0)) {
420 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800422 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800424 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 return fd;
427}
428
429/**
430 * sockfd_lookup - Go from a file number to its socket slot
431 * @fd: file handle
432 * @err: pointer to an error code return
433 *
434 * The file handle passed in is locked and the socket it is bound
435 * too is returned. If an error occurs the err pointer is overwritten
436 * with a negative errno code and NULL is returned. The function checks
437 * for both invalid handles and passing a handle which is not a socket.
438 *
439 * On a success the socket object pointer is returned.
440 */
441
442struct socket *sockfd_lookup(int fd, int *err)
443{
444 struct file *file;
445 struct inode *inode;
446 struct socket *sock;
447
448 if (!(file = fget(fd)))
449 {
450 *err = -EBADF;
451 return NULL;
452 }
453
Benjamin LaHaise07dc3f02005-08-10 14:16:04 -0700454 if (file->f_op == &socket_file_ops)
455 return file->private_data; /* set in sock_map_fd */
456
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 inode = file->f_dentry->d_inode;
458 if (!S_ISSOCK(inode->i_mode)) {
459 *err = -ENOTSOCK;
460 fput(file);
461 return NULL;
462 }
463
464 sock = SOCKET_I(inode);
465 if (sock->file != file) {
466 printk(KERN_ERR "socki_lookup: socket file changed!\n");
467 sock->file = file;
468 }
469 return sock;
470}
471
472/**
473 * sock_alloc - allocate a socket
474 *
475 * Allocate a new inode and socket object. The two are bound together
476 * and initialised. The socket is then returned. If we are out of inodes
477 * NULL is returned.
478 */
479
480static struct socket *sock_alloc(void)
481{
482 struct inode * inode;
483 struct socket * sock;
484
485 inode = new_inode(sock_mnt->mnt_sb);
486 if (!inode)
487 return NULL;
488
489 sock = SOCKET_I(inode);
490
491 inode->i_mode = S_IFSOCK|S_IRWXUGO;
492 inode->i_uid = current->fsuid;
493 inode->i_gid = current->fsgid;
494
495 get_cpu_var(sockets_in_use)++;
496 put_cpu_var(sockets_in_use);
497 return sock;
498}
499
500/*
501 * In theory you can't get an open on this inode, but /proc provides
502 * a back door. Remember to keep it shut otherwise you'll let the
503 * creepy crawlies in.
504 */
505
506static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
507{
508 return -ENXIO;
509}
510
511struct file_operations bad_sock_fops = {
512 .owner = THIS_MODULE,
513 .open = sock_no_open,
514};
515
516/**
517 * sock_release - close a socket
518 * @sock: socket to close
519 *
520 * The socket is released from the protocol stack if it has a release
521 * callback, and the inode is then released if the socket is bound to
522 * an inode not a file.
523 */
524
525void sock_release(struct socket *sock)
526{
527 if (sock->ops) {
528 struct module *owner = sock->ops->owner;
529
530 sock->ops->release(sock);
531 sock->ops = NULL;
532 module_put(owner);
533 }
534
535 if (sock->fasync_list)
536 printk(KERN_ERR "sock_release: fasync list not empty!\n");
537
538 get_cpu_var(sockets_in_use)--;
539 put_cpu_var(sockets_in_use);
540 if (!sock->file) {
541 iput(SOCK_INODE(sock));
542 return;
543 }
544 sock->file=NULL;
545}
546
547static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
548 struct msghdr *msg, size_t size)
549{
550 struct sock_iocb *si = kiocb_to_siocb(iocb);
551 int err;
552
553 si->sock = sock;
554 si->scm = NULL;
555 si->msg = msg;
556 si->size = size;
557
558 err = security_socket_sendmsg(sock, msg, size);
559 if (err)
560 return err;
561
562 return sock->ops->sendmsg(iocb, sock, msg, size);
563}
564
565int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
566{
567 struct kiocb iocb;
568 struct sock_iocb siocb;
569 int ret;
570
571 init_sync_kiocb(&iocb, NULL);
572 iocb.private = &siocb;
573 ret = __sock_sendmsg(&iocb, sock, msg, size);
574 if (-EIOCBQUEUED == ret)
575 ret = wait_on_sync_kiocb(&iocb);
576 return ret;
577}
578
579int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
580 struct kvec *vec, size_t num, size_t size)
581{
582 mm_segment_t oldfs = get_fs();
583 int result;
584
585 set_fs(KERNEL_DS);
586 /*
587 * the following is safe, since for compiler definitions of kvec and
588 * iovec are identical, yielding the same in-core layout and alignment
589 */
590 msg->msg_iov = (struct iovec *)vec,
591 msg->msg_iovlen = num;
592 result = sock_sendmsg(sock, msg, size);
593 set_fs(oldfs);
594 return result;
595}
596
597static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
598 struct msghdr *msg, size_t size, int flags)
599{
600 int err;
601 struct sock_iocb *si = kiocb_to_siocb(iocb);
602
603 si->sock = sock;
604 si->scm = NULL;
605 si->msg = msg;
606 si->size = size;
607 si->flags = flags;
608
609 err = security_socket_recvmsg(sock, msg, size, flags);
610 if (err)
611 return err;
612
613 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
614}
615
616int sock_recvmsg(struct socket *sock, struct msghdr *msg,
617 size_t size, int flags)
618{
619 struct kiocb iocb;
620 struct sock_iocb siocb;
621 int ret;
622
623 init_sync_kiocb(&iocb, NULL);
624 iocb.private = &siocb;
625 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
626 if (-EIOCBQUEUED == ret)
627 ret = wait_on_sync_kiocb(&iocb);
628 return ret;
629}
630
631int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
632 struct kvec *vec, size_t num,
633 size_t size, int flags)
634{
635 mm_segment_t oldfs = get_fs();
636 int result;
637
638 set_fs(KERNEL_DS);
639 /*
640 * the following is safe, since for compiler definitions of kvec and
641 * iovec are identical, yielding the same in-core layout and alignment
642 */
643 msg->msg_iov = (struct iovec *)vec,
644 msg->msg_iovlen = num;
645 result = sock_recvmsg(sock, msg, size, flags);
646 set_fs(oldfs);
647 return result;
648}
649
650static void sock_aio_dtor(struct kiocb *iocb)
651{
652 kfree(iocb->private);
653}
654
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300655static ssize_t sock_sendpage(struct file *file, struct page *page,
656 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657{
658 struct socket *sock;
659 int flags;
660
Eric Dumazetb69aee02005-09-06 14:42:45 -0700661 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662
663 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
664 if (more)
665 flags |= MSG_MORE;
666
667 return sock->ops->sendpage(sock, page, offset, size, flags);
668}
669
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800670static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
671 char __user *ubuf, size_t size, struct sock_iocb *siocb)
672{
673 if (!is_sync_kiocb(iocb)) {
674 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
675 if (!siocb)
676 return NULL;
677 iocb->ki_dtor = sock_aio_dtor;
678 }
679
680 siocb->kiocb = iocb;
681 siocb->async_iov.iov_base = ubuf;
682 siocb->async_iov.iov_len = size;
683
684 iocb->private = siocb;
685 return siocb;
686}
687
688static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
689 struct file *file, struct iovec *iov, unsigned long nr_segs)
690{
691 struct socket *sock = file->private_data;
692 size_t size = 0;
693 int i;
694
695 for (i = 0 ; i < nr_segs ; i++)
696 size += iov[i].iov_len;
697
698 msg->msg_name = NULL;
699 msg->msg_namelen = 0;
700 msg->msg_control = NULL;
701 msg->msg_controllen = 0;
702 msg->msg_iov = (struct iovec *) iov;
703 msg->msg_iovlen = nr_segs;
704 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
705
706 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
707}
708
709static ssize_t sock_readv(struct file *file, const struct iovec *iov,
710 unsigned long nr_segs, loff_t *ppos)
711{
712 struct kiocb iocb;
713 struct sock_iocb siocb;
714 struct msghdr msg;
715 int ret;
716
717 init_sync_kiocb(&iocb, NULL);
718 iocb.private = &siocb;
719
720 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
721 if (-EIOCBQUEUED == ret)
722 ret = wait_on_sync_kiocb(&iocb);
723 return ret;
724}
725
726static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
727 size_t count, loff_t pos)
728{
729 struct sock_iocb siocb, *x;
730
731 if (pos != 0)
732 return -ESPIPE;
733 if (count == 0) /* Match SYS5 behaviour */
734 return 0;
735
736 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
737 if (!x)
738 return -ENOMEM;
739 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
740 &x->async_iov, 1);
741}
742
743static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
744 struct file *file, struct iovec *iov, unsigned long nr_segs)
745{
746 struct socket *sock = file->private_data;
747 size_t size = 0;
748 int i;
749
750 for (i = 0 ; i < nr_segs ; i++)
751 size += iov[i].iov_len;
752
753 msg->msg_name = NULL;
754 msg->msg_namelen = 0;
755 msg->msg_control = NULL;
756 msg->msg_controllen = 0;
757 msg->msg_iov = (struct iovec *) iov;
758 msg->msg_iovlen = nr_segs;
759 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
760 if (sock->type == SOCK_SEQPACKET)
761 msg->msg_flags |= MSG_EOR;
762
763 return __sock_sendmsg(iocb, sock, msg, size);
764}
765
766static ssize_t sock_writev(struct file *file, const struct iovec *iov,
767 unsigned long nr_segs, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768{
769 struct msghdr msg;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800770 struct kiocb iocb;
771 struct sock_iocb siocb;
772 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800774 init_sync_kiocb(&iocb, NULL);
775 iocb.private = &siocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800777 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
778 if (-EIOCBQUEUED == ret)
779 ret = wait_on_sync_kiocb(&iocb);
780 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781}
782
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800783static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
784 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800786 struct sock_iocb siocb, *x;
787
788 if (pos != 0)
789 return -ESPIPE;
790 if (count == 0) /* Match SYS5 behaviour */
791 return 0;
792
793 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
794 if (!x)
795 return -ENOMEM;
796
797 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
798 &x->async_iov, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799}
800
801
802/*
803 * Atomic setting of ioctl hooks to avoid race
804 * with module unload.
805 */
806
807static DECLARE_MUTEX(br_ioctl_mutex);
808static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
809
810void brioctl_set(int (*hook)(unsigned int, void __user *))
811{
812 down(&br_ioctl_mutex);
813 br_ioctl_hook = hook;
814 up(&br_ioctl_mutex);
815}
816EXPORT_SYMBOL(brioctl_set);
817
818static DECLARE_MUTEX(vlan_ioctl_mutex);
819static int (*vlan_ioctl_hook)(void __user *arg);
820
821void vlan_ioctl_set(int (*hook)(void __user *))
822{
823 down(&vlan_ioctl_mutex);
824 vlan_ioctl_hook = hook;
825 up(&vlan_ioctl_mutex);
826}
827EXPORT_SYMBOL(vlan_ioctl_set);
828
829static DECLARE_MUTEX(dlci_ioctl_mutex);
830static int (*dlci_ioctl_hook)(unsigned int, void __user *);
831
832void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
833{
834 down(&dlci_ioctl_mutex);
835 dlci_ioctl_hook = hook;
836 up(&dlci_ioctl_mutex);
837}
838EXPORT_SYMBOL(dlci_ioctl_set);
839
840/*
841 * With an ioctl, arg may well be a user mode pointer, but we don't know
842 * what to do with it - that's up to the protocol still.
843 */
844
845static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
846{
847 struct socket *sock;
848 void __user *argp = (void __user *)arg;
849 int pid, err;
850
Eric Dumazetb69aee02005-09-06 14:42:45 -0700851 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
853 err = dev_ioctl(cmd, argp);
854 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100855#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
857 err = dev_ioctl(cmd, argp);
858 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100859#endif /* CONFIG_WIRELESS_EXT */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 switch (cmd) {
861 case FIOSETOWN:
862 case SIOCSPGRP:
863 err = -EFAULT;
864 if (get_user(pid, (int __user *)argp))
865 break;
866 err = f_setown(sock->file, pid, 1);
867 break;
868 case FIOGETOWN:
869 case SIOCGPGRP:
870 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
871 break;
872 case SIOCGIFBR:
873 case SIOCSIFBR:
874 case SIOCBRADDBR:
875 case SIOCBRDELBR:
876 err = -ENOPKG;
877 if (!br_ioctl_hook)
878 request_module("bridge");
879
880 down(&br_ioctl_mutex);
881 if (br_ioctl_hook)
882 err = br_ioctl_hook(cmd, argp);
883 up(&br_ioctl_mutex);
884 break;
885 case SIOCGIFVLAN:
886 case SIOCSIFVLAN:
887 err = -ENOPKG;
888 if (!vlan_ioctl_hook)
889 request_module("8021q");
890
891 down(&vlan_ioctl_mutex);
892 if (vlan_ioctl_hook)
893 err = vlan_ioctl_hook(argp);
894 up(&vlan_ioctl_mutex);
895 break;
896 case SIOCGIFDIVERT:
897 case SIOCSIFDIVERT:
898 /* Convert this to call through a hook */
899 err = divert_ioctl(cmd, argp);
900 break;
901 case SIOCADDDLCI:
902 case SIOCDELDLCI:
903 err = -ENOPKG;
904 if (!dlci_ioctl_hook)
905 request_module("dlci");
906
907 if (dlci_ioctl_hook) {
908 down(&dlci_ioctl_mutex);
909 err = dlci_ioctl_hook(cmd, argp);
910 up(&dlci_ioctl_mutex);
911 }
912 break;
913 default:
914 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800915
916 /*
917 * If this ioctl is unknown try to hand it down
918 * to the NIC driver.
919 */
920 if (err == -ENOIOCTLCMD)
921 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 break;
923 }
924 return err;
925}
926
927int sock_create_lite(int family, int type, int protocol, struct socket **res)
928{
929 int err;
930 struct socket *sock = NULL;
931
932 err = security_socket_create(family, type, protocol, 1);
933 if (err)
934 goto out;
935
936 sock = sock_alloc();
937 if (!sock) {
938 err = -ENOMEM;
939 goto out;
940 }
941
942 security_socket_post_create(sock, family, type, protocol, 1);
943 sock->type = type;
944out:
945 *res = sock;
946 return err;
947}
948
949/* No kernel lock held - perfect */
950static unsigned int sock_poll(struct file *file, poll_table * wait)
951{
952 struct socket *sock;
953
954 /*
955 * We can't return errors to poll, so it's either yes or no.
956 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700957 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 return sock->ops->poll(file, sock, wait);
959}
960
961static int sock_mmap(struct file * file, struct vm_area_struct * vma)
962{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700963 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964
965 return sock->ops->mmap(file, sock, vma);
966}
967
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300968static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969{
970 /*
971 * It was possible the inode is NULL we were
972 * closing an unfinished socket.
973 */
974
975 if (!inode)
976 {
977 printk(KERN_DEBUG "sock_close: NULL inode\n");
978 return 0;
979 }
980 sock_fasync(-1, filp, 0);
981 sock_release(SOCKET_I(inode));
982 return 0;
983}
984
985/*
986 * Update the socket async list
987 *
988 * Fasync_list locking strategy.
989 *
990 * 1. fasync_list is modified only under process context socket lock
991 * i.e. under semaphore.
992 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
993 * or under socket lock.
994 * 3. fasync_list can be used from softirq context, so that
995 * modification under socket lock have to be enhanced with
996 * write_lock_bh(&sk->sk_callback_lock).
997 * --ANK (990710)
998 */
999
1000static int sock_fasync(int fd, struct file *filp, int on)
1001{
1002 struct fasync_struct *fa, *fna=NULL, **prev;
1003 struct socket *sock;
1004 struct sock *sk;
1005
1006 if (on)
1007 {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001008 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 if(fna==NULL)
1010 return -ENOMEM;
1011 }
1012
Eric Dumazetb69aee02005-09-06 14:42:45 -07001013 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
1015 if ((sk=sock->sk) == NULL) {
1016 kfree(fna);
1017 return -EINVAL;
1018 }
1019
1020 lock_sock(sk);
1021
1022 prev=&(sock->fasync_list);
1023
1024 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1025 if (fa->fa_file==filp)
1026 break;
1027
1028 if(on)
1029 {
1030 if(fa!=NULL)
1031 {
1032 write_lock_bh(&sk->sk_callback_lock);
1033 fa->fa_fd=fd;
1034 write_unlock_bh(&sk->sk_callback_lock);
1035
1036 kfree(fna);
1037 goto out;
1038 }
1039 fna->fa_file=filp;
1040 fna->fa_fd=fd;
1041 fna->magic=FASYNC_MAGIC;
1042 fna->fa_next=sock->fasync_list;
1043 write_lock_bh(&sk->sk_callback_lock);
1044 sock->fasync_list=fna;
1045 write_unlock_bh(&sk->sk_callback_lock);
1046 }
1047 else
1048 {
1049 if (fa!=NULL)
1050 {
1051 write_lock_bh(&sk->sk_callback_lock);
1052 *prev=fa->fa_next;
1053 write_unlock_bh(&sk->sk_callback_lock);
1054 kfree(fa);
1055 }
1056 }
1057
1058out:
1059 release_sock(sock->sk);
1060 return 0;
1061}
1062
1063/* This function may be called only under socket lock or callback_lock */
1064
1065int sock_wake_async(struct socket *sock, int how, int band)
1066{
1067 if (!sock || !sock->fasync_list)
1068 return -1;
1069 switch (how)
1070 {
1071 case 1:
1072
1073 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1074 break;
1075 goto call_kill;
1076 case 2:
1077 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1078 break;
1079 /* fall through */
1080 case 0:
1081 call_kill:
1082 __kill_fasync(sock->fasync_list, SIGIO, band);
1083 break;
1084 case 3:
1085 __kill_fasync(sock->fasync_list, SIGURG, band);
1086 }
1087 return 0;
1088}
1089
1090static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1091{
1092 int err;
1093 struct socket *sock;
1094
1095 /*
1096 * Check protocol is in range
1097 */
1098 if (family < 0 || family >= NPROTO)
1099 return -EAFNOSUPPORT;
1100 if (type < 0 || type >= SOCK_MAX)
1101 return -EINVAL;
1102
1103 /* Compatibility.
1104
1105 This uglymoron is moved from INET layer to here to avoid
1106 deadlock in module load.
1107 */
1108 if (family == PF_INET && type == SOCK_PACKET) {
1109 static int warned;
1110 if (!warned) {
1111 warned = 1;
1112 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1113 }
1114 family = PF_PACKET;
1115 }
1116
1117 err = security_socket_create(family, type, protocol, kern);
1118 if (err)
1119 return err;
1120
1121#if defined(CONFIG_KMOD)
1122 /* Attempt to load a protocol module if the find failed.
1123 *
1124 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1125 * requested real, full-featured networking support upon configuration.
1126 * Otherwise module support will break!
1127 */
1128 if (net_families[family]==NULL)
1129 {
1130 request_module("net-pf-%d",family);
1131 }
1132#endif
1133
1134 net_family_read_lock();
1135 if (net_families[family] == NULL) {
1136 err = -EAFNOSUPPORT;
1137 goto out;
1138 }
1139
1140/*
1141 * Allocate the socket and allow the family to set things up. if
1142 * the protocol is 0, the family is instructed to select an appropriate
1143 * default.
1144 */
1145
1146 if (!(sock = sock_alloc())) {
1147 printk(KERN_WARNING "socket: no more sockets\n");
1148 err = -ENFILE; /* Not exactly a match, but its the
1149 closest posix thing */
1150 goto out;
1151 }
1152
1153 sock->type = type;
1154
1155 /*
1156 * We will call the ->create function, that possibly is in a loadable
1157 * module, so we have to bump that loadable module refcnt first.
1158 */
1159 err = -EAFNOSUPPORT;
1160 if (!try_module_get(net_families[family]->owner))
1161 goto out_release;
1162
Frank Filza79af592005-09-27 15:23:38 -07001163 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1164 sock->ops = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001166 }
1167
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 /*
1169 * Now to bump the refcnt of the [loadable] module that owns this
1170 * socket at sock_release time we decrement its refcnt.
1171 */
1172 if (!try_module_get(sock->ops->owner)) {
1173 sock->ops = NULL;
1174 goto out_module_put;
1175 }
1176 /*
1177 * Now that we're done with the ->create function, the [loadable]
1178 * module can have its refcnt decremented
1179 */
1180 module_put(net_families[family]->owner);
1181 *res = sock;
1182 security_socket_post_create(sock, family, type, protocol, kern);
1183
1184out:
1185 net_family_read_unlock();
1186 return err;
1187out_module_put:
1188 module_put(net_families[family]->owner);
1189out_release:
1190 sock_release(sock);
1191 goto out;
1192}
1193
1194int sock_create(int family, int type, int protocol, struct socket **res)
1195{
1196 return __sock_create(family, type, protocol, res, 0);
1197}
1198
1199int sock_create_kern(int family, int type, int protocol, struct socket **res)
1200{
1201 return __sock_create(family, type, protocol, res, 1);
1202}
1203
1204asmlinkage long sys_socket(int family, int type, int protocol)
1205{
1206 int retval;
1207 struct socket *sock;
1208
1209 retval = sock_create(family, type, protocol, &sock);
1210 if (retval < 0)
1211 goto out;
1212
1213 retval = sock_map_fd(sock);
1214 if (retval < 0)
1215 goto out_release;
1216
1217out:
1218 /* It may be already another descriptor 8) Not kernel problem. */
1219 return retval;
1220
1221out_release:
1222 sock_release(sock);
1223 return retval;
1224}
1225
1226/*
1227 * Create a pair of connected sockets.
1228 */
1229
1230asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1231{
1232 struct socket *sock1, *sock2;
1233 int fd1, fd2, err;
1234
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);
1249 if (err < 0)
1250 goto out_release_both;
1251
1252 fd1 = fd2 = -1;
1253
1254 err = sock_map_fd(sock1);
1255 if (err < 0)
1256 goto out_release_both;
1257 fd1 = err;
1258
1259 err = sock_map_fd(sock2);
1260 if (err < 0)
1261 goto out_close_1;
1262 fd2 = err;
1263
1264 /* fd1 and fd2 may be already another descriptors.
1265 * Not kernel problem.
1266 */
1267
1268 err = put_user(fd1, &usockvec[0]);
1269 if (!err)
1270 err = put_user(fd2, &usockvec[1]);
1271 if (!err)
1272 return 0;
1273
1274 sys_close(fd2);
1275 sys_close(fd1);
1276 return err;
1277
1278out_close_1:
1279 sock_release(sock2);
1280 sys_close(fd1);
1281 return err;
1282
1283out_release_both:
1284 sock_release(sock2);
1285out_release_1:
1286 sock_release(sock1);
1287out:
1288 return err;
1289}
1290
1291
1292/*
1293 * Bind a name to a socket. Nothing much to do here since it's
1294 * the protocol's responsibility to handle the local address.
1295 *
1296 * We move the socket address to kernel space before we call
1297 * the protocol layer (having also checked the address is ok).
1298 */
1299
1300asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1301{
1302 struct socket *sock;
1303 char address[MAX_SOCK_ADDR];
1304 int err;
1305
1306 if((sock = sockfd_lookup(fd,&err))!=NULL)
1307 {
1308 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1309 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1310 if (err) {
1311 sockfd_put(sock);
1312 return err;
1313 }
1314 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1315 }
1316 sockfd_put(sock);
1317 }
1318 return err;
1319}
1320
1321
1322/*
1323 * Perform a listen. Basically, we allow the protocol to do anything
1324 * necessary for a listen, and if that works, we mark the socket as
1325 * ready for listening.
1326 */
1327
1328int sysctl_somaxconn = SOMAXCONN;
1329
1330asmlinkage long sys_listen(int fd, int backlog)
1331{
1332 struct socket *sock;
1333 int err;
1334
1335 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1336 if ((unsigned) backlog > sysctl_somaxconn)
1337 backlog = sysctl_somaxconn;
1338
1339 err = security_socket_listen(sock, backlog);
1340 if (err) {
1341 sockfd_put(sock);
1342 return err;
1343 }
1344
1345 err=sock->ops->listen(sock, backlog);
1346 sockfd_put(sock);
1347 }
1348 return err;
1349}
1350
1351
1352/*
1353 * For accept, we attempt to create a new socket, set up the link
1354 * with the client, wake up the client, then return the new
1355 * connected fd. We collect the address of the connector in kernel
1356 * space and move it to user at the very end. This is unclean because
1357 * we open the socket then return an error.
1358 *
1359 * 1003.1g adds the ability to recvmsg() to query connection pending
1360 * status to recvmsg. We need to add that support in a way thats
1361 * clean when we restucture accept also.
1362 */
1363
1364asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1365{
1366 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001367 struct file *newfile;
1368 int err, len, newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 char address[MAX_SOCK_ADDR];
1370
1371 sock = sockfd_lookup(fd, &err);
1372 if (!sock)
1373 goto out;
1374
1375 err = -ENFILE;
1376 if (!(newsock = sock_alloc()))
1377 goto out_put;
1378
1379 newsock->type = sock->type;
1380 newsock->ops = sock->ops;
1381
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 /*
1383 * We don't need try_module_get here, as the listening socket (sock)
1384 * has the protocol module (sock->ops->owner) held.
1385 */
1386 __module_get(newsock->ops->owner);
1387
David S. Miller39d8c1b2006-03-20 17:13:49 -08001388 newfd = sock_alloc_fd(&newfile);
1389 if (unlikely(newfd < 0)) {
1390 err = newfd;
1391 goto out_release;
1392 }
1393
1394 err = sock_attach_fd(newsock, newfile);
1395 if (err < 0)
1396 goto out_fd;
1397
Frank Filza79af592005-09-27 15:23:38 -07001398 err = security_socket_accept(sock, newsock);
1399 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001400 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001401
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1403 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001404 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405
1406 if (upeer_sockaddr) {
1407 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1408 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001409 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 }
1411 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1412 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001413 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 }
1415
1416 /* File flags are not inherited via accept() unlike another OSes. */
1417
David S. Miller39d8c1b2006-03-20 17:13:49 -08001418 fd_install(newfd, newfile);
1419 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
1421 security_socket_post_accept(sock, newsock);
1422
1423out_put:
1424 sockfd_put(sock);
1425out:
1426 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001427out_fd:
1428 put_filp(newfile);
1429 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430out_release:
1431 sock_release(newsock);
1432 goto out_put;
1433}
1434
1435
1436/*
1437 * Attempt to connect to a socket with the server address. The address
1438 * is in user space so we verify it is OK and move it to kernel space.
1439 *
1440 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1441 * break bindings
1442 *
1443 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1444 * other SEQPACKET protocols that take time to connect() as it doesn't
1445 * include the -EINPROGRESS status for such sockets.
1446 */
1447
1448asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1449{
1450 struct socket *sock;
1451 char address[MAX_SOCK_ADDR];
1452 int err;
1453
1454 sock = sockfd_lookup(fd, &err);
1455 if (!sock)
1456 goto out;
1457 err = move_addr_to_kernel(uservaddr, addrlen, address);
1458 if (err < 0)
1459 goto out_put;
1460
1461 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1462 if (err)
1463 goto out_put;
1464
1465 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1466 sock->file->f_flags);
1467out_put:
1468 sockfd_put(sock);
1469out:
1470 return err;
1471}
1472
1473/*
1474 * Get the local address ('name') of a socket object. Move the obtained
1475 * name to user space.
1476 */
1477
1478asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1479{
1480 struct socket *sock;
1481 char address[MAX_SOCK_ADDR];
1482 int len, err;
1483
1484 sock = sockfd_lookup(fd, &err);
1485 if (!sock)
1486 goto out;
1487
1488 err = security_socket_getsockname(sock);
1489 if (err)
1490 goto out_put;
1491
1492 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1493 if (err)
1494 goto out_put;
1495 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1496
1497out_put:
1498 sockfd_put(sock);
1499out:
1500 return err;
1501}
1502
1503/*
1504 * Get the remote address ('name') of a socket object. Move the obtained
1505 * name to user space.
1506 */
1507
1508asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1509{
1510 struct socket *sock;
1511 char address[MAX_SOCK_ADDR];
1512 int len, err;
1513
1514 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1515 {
1516 err = security_socket_getpeername(sock);
1517 if (err) {
1518 sockfd_put(sock);
1519 return err;
1520 }
1521
1522 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1523 if (!err)
1524 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1525 sockfd_put(sock);
1526 }
1527 return err;
1528}
1529
1530/*
1531 * Send a datagram to a given address. We move the address into kernel
1532 * space and check the user space data area is readable before invoking
1533 * the protocol.
1534 */
1535
1536asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1537 struct sockaddr __user *addr, int addr_len)
1538{
1539 struct socket *sock;
1540 char address[MAX_SOCK_ADDR];
1541 int err;
1542 struct msghdr msg;
1543 struct iovec iov;
1544
1545 sock = sockfd_lookup(fd, &err);
1546 if (!sock)
1547 goto out;
1548 iov.iov_base=buff;
1549 iov.iov_len=len;
1550 msg.msg_name=NULL;
1551 msg.msg_iov=&iov;
1552 msg.msg_iovlen=1;
1553 msg.msg_control=NULL;
1554 msg.msg_controllen=0;
1555 msg.msg_namelen=0;
1556 if(addr)
1557 {
1558 err = move_addr_to_kernel(addr, addr_len, address);
1559 if (err < 0)
1560 goto out_put;
1561 msg.msg_name=address;
1562 msg.msg_namelen=addr_len;
1563 }
1564 if (sock->file->f_flags & O_NONBLOCK)
1565 flags |= MSG_DONTWAIT;
1566 msg.msg_flags = flags;
1567 err = sock_sendmsg(sock, &msg, len);
1568
1569out_put:
1570 sockfd_put(sock);
1571out:
1572 return err;
1573}
1574
1575/*
1576 * Send a datagram down a socket.
1577 */
1578
1579asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1580{
1581 return sys_sendto(fd, buff, len, flags, NULL, 0);
1582}
1583
1584/*
1585 * Receive a frame from the socket and optionally record the address of the
1586 * sender. We verify the buffers are writable and if needed move the
1587 * sender address from kernel to user space.
1588 */
1589
1590asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1591 struct sockaddr __user *addr, int __user *addr_len)
1592{
1593 struct socket *sock;
1594 struct iovec iov;
1595 struct msghdr msg;
1596 char address[MAX_SOCK_ADDR];
1597 int err,err2;
1598
1599 sock = sockfd_lookup(fd, &err);
1600 if (!sock)
1601 goto out;
1602
1603 msg.msg_control=NULL;
1604 msg.msg_controllen=0;
1605 msg.msg_iovlen=1;
1606 msg.msg_iov=&iov;
1607 iov.iov_len=size;
1608 iov.iov_base=ubuf;
1609 msg.msg_name=address;
1610 msg.msg_namelen=MAX_SOCK_ADDR;
1611 if (sock->file->f_flags & O_NONBLOCK)
1612 flags |= MSG_DONTWAIT;
1613 err=sock_recvmsg(sock, &msg, size, flags);
1614
1615 if(err >= 0 && addr != NULL)
1616 {
1617 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1618 if(err2<0)
1619 err=err2;
1620 }
1621 sockfd_put(sock);
1622out:
1623 return err;
1624}
1625
1626/*
1627 * Receive a datagram from a socket.
1628 */
1629
1630asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1631{
1632 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1633}
1634
1635/*
1636 * Set a socket option. Because we don't know the option lengths we have
1637 * to pass the user mode parameter for the protocols to sort out.
1638 */
1639
1640asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1641{
1642 int err;
1643 struct socket *sock;
1644
1645 if (optlen < 0)
1646 return -EINVAL;
1647
1648 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1649 {
1650 err = security_socket_setsockopt(sock,level,optname);
1651 if (err) {
1652 sockfd_put(sock);
1653 return err;
1654 }
1655
1656 if (level == SOL_SOCKET)
1657 err=sock_setsockopt(sock,level,optname,optval,optlen);
1658 else
1659 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1660 sockfd_put(sock);
1661 }
1662 return err;
1663}
1664
1665/*
1666 * Get a socket option. Because we don't know the option lengths we have
1667 * to pass a user mode parameter for the protocols to sort out.
1668 */
1669
1670asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1671{
1672 int err;
1673 struct socket *sock;
1674
1675 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1676 {
1677 err = security_socket_getsockopt(sock, level,
1678 optname);
1679 if (err) {
1680 sockfd_put(sock);
1681 return err;
1682 }
1683
1684 if (level == SOL_SOCKET)
1685 err=sock_getsockopt(sock,level,optname,optval,optlen);
1686 else
1687 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1688 sockfd_put(sock);
1689 }
1690 return err;
1691}
1692
1693
1694/*
1695 * Shutdown a socket.
1696 */
1697
1698asmlinkage long sys_shutdown(int fd, int how)
1699{
1700 int err;
1701 struct socket *sock;
1702
1703 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1704 {
1705 err = security_socket_shutdown(sock, how);
1706 if (err) {
1707 sockfd_put(sock);
1708 return err;
1709 }
1710
1711 err=sock->ops->shutdown(sock, how);
1712 sockfd_put(sock);
1713 }
1714 return err;
1715}
1716
1717/* A couple of helpful macros for getting the address of the 32/64 bit
1718 * fields which are the same type (int / unsigned) on our platforms.
1719 */
1720#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1721#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1722#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1723
1724
1725/*
1726 * BSD sendmsg interface
1727 */
1728
1729asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1730{
1731 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1732 struct socket *sock;
1733 char address[MAX_SOCK_ADDR];
1734 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001735 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1736 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1737 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 unsigned char *ctl_buf = ctl;
1739 struct msghdr msg_sys;
1740 int err, ctl_len, iov_size, total_len;
1741
1742 err = -EFAULT;
1743 if (MSG_CMSG_COMPAT & flags) {
1744 if (get_compat_msghdr(&msg_sys, msg_compat))
1745 return -EFAULT;
1746 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1747 return -EFAULT;
1748
1749 sock = sockfd_lookup(fd, &err);
1750 if (!sock)
1751 goto out;
1752
1753 /* do not move before msg_sys is valid */
1754 err = -EMSGSIZE;
1755 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1756 goto out_put;
1757
1758 /* Check whether to allocate the iovec area*/
1759 err = -ENOMEM;
1760 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1761 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1762 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1763 if (!iov)
1764 goto out_put;
1765 }
1766
1767 /* This will also move the address data into kernel space */
1768 if (MSG_CMSG_COMPAT & flags) {
1769 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1770 } else
1771 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1772 if (err < 0)
1773 goto out_freeiov;
1774 total_len = err;
1775
1776 err = -ENOBUFS;
1777
1778 if (msg_sys.msg_controllen > INT_MAX)
1779 goto out_freeiov;
1780 ctl_len = msg_sys.msg_controllen;
1781 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Al Viro8920e8f2005-09-07 18:28:51 -07001782 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 if (err)
1784 goto out_freeiov;
1785 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001786 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 } else if (ctl_len) {
1788 if (ctl_len > sizeof(ctl))
1789 {
1790 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1791 if (ctl_buf == NULL)
1792 goto out_freeiov;
1793 }
1794 err = -EFAULT;
1795 /*
1796 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1797 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1798 * checking falls down on this.
1799 */
1800 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1801 goto out_freectl;
1802 msg_sys.msg_control = ctl_buf;
1803 }
1804 msg_sys.msg_flags = flags;
1805
1806 if (sock->file->f_flags & O_NONBLOCK)
1807 msg_sys.msg_flags |= MSG_DONTWAIT;
1808 err = sock_sendmsg(sock, &msg_sys, total_len);
1809
1810out_freectl:
1811 if (ctl_buf != ctl)
1812 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1813out_freeiov:
1814 if (iov != iovstack)
1815 sock_kfree_s(sock->sk, iov, iov_size);
1816out_put:
1817 sockfd_put(sock);
1818out:
1819 return err;
1820}
1821
1822/*
1823 * BSD recvmsg interface
1824 */
1825
1826asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1827{
1828 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1829 struct socket *sock;
1830 struct iovec iovstack[UIO_FASTIOV];
1831 struct iovec *iov=iovstack;
1832 struct msghdr msg_sys;
1833 unsigned long cmsg_ptr;
1834 int err, iov_size, total_len, len;
1835
1836 /* kernel mode address */
1837 char addr[MAX_SOCK_ADDR];
1838
1839 /* user mode address pointers */
1840 struct sockaddr __user *uaddr;
1841 int __user *uaddr_len;
1842
1843 if (MSG_CMSG_COMPAT & flags) {
1844 if (get_compat_msghdr(&msg_sys, msg_compat))
1845 return -EFAULT;
1846 } else
1847 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1848 return -EFAULT;
1849
1850 sock = sockfd_lookup(fd, &err);
1851 if (!sock)
1852 goto out;
1853
1854 err = -EMSGSIZE;
1855 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1856 goto out_put;
1857
1858 /* Check whether to allocate the iovec area*/
1859 err = -ENOMEM;
1860 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1861 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1862 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1863 if (!iov)
1864 goto out_put;
1865 }
1866
1867 /*
1868 * Save the user-mode address (verify_iovec will change the
1869 * kernel msghdr to use the kernel address space)
1870 */
1871
1872 uaddr = (void __user *) msg_sys.msg_name;
1873 uaddr_len = COMPAT_NAMELEN(msg);
1874 if (MSG_CMSG_COMPAT & flags) {
1875 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1876 } else
1877 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1878 if (err < 0)
1879 goto out_freeiov;
1880 total_len=err;
1881
1882 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1883 msg_sys.msg_flags = 0;
1884 if (MSG_CMSG_COMPAT & flags)
1885 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1886
1887 if (sock->file->f_flags & O_NONBLOCK)
1888 flags |= MSG_DONTWAIT;
1889 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1890 if (err < 0)
1891 goto out_freeiov;
1892 len = err;
1893
1894 if (uaddr != NULL) {
1895 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1896 if (err < 0)
1897 goto out_freeiov;
1898 }
David S. Miller37f7f422005-09-16 16:51:01 -07001899 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1900 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 if (err)
1902 goto out_freeiov;
1903 if (MSG_CMSG_COMPAT & flags)
1904 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1905 &msg_compat->msg_controllen);
1906 else
1907 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1908 &msg->msg_controllen);
1909 if (err)
1910 goto out_freeiov;
1911 err = len;
1912
1913out_freeiov:
1914 if (iov != iovstack)
1915 sock_kfree_s(sock->sk, iov, iov_size);
1916out_put:
1917 sockfd_put(sock);
1918out:
1919 return err;
1920}
1921
1922#ifdef __ARCH_WANT_SYS_SOCKETCALL
1923
1924/* Argument list sizes for sys_socketcall */
1925#define AL(x) ((x) * sizeof(unsigned long))
1926static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1927 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1928 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1929#undef AL
1930
1931/*
1932 * System call vectors.
1933 *
1934 * Argument checking cleaned up. Saved 20% in size.
1935 * This function doesn't need to set the kernel lock because
1936 * it is set by the callees.
1937 */
1938
1939asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1940{
1941 unsigned long a[6];
1942 unsigned long a0,a1;
1943 int err;
1944
1945 if(call<1||call>SYS_RECVMSG)
1946 return -EINVAL;
1947
1948 /* copy_from_user should be SMP safe. */
1949 if (copy_from_user(a, args, nargs[call]))
1950 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001951
David Woodhouse4bcff1b2005-06-02 12:13:21 +01001952 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001953 if (err)
1954 return err;
1955
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 a0=a[0];
1957 a1=a[1];
1958
1959 switch(call)
1960 {
1961 case SYS_SOCKET:
1962 err = sys_socket(a0,a1,a[2]);
1963 break;
1964 case SYS_BIND:
1965 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1966 break;
1967 case SYS_CONNECT:
1968 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1969 break;
1970 case SYS_LISTEN:
1971 err = sys_listen(a0,a1);
1972 break;
1973 case SYS_ACCEPT:
1974 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1975 break;
1976 case SYS_GETSOCKNAME:
1977 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1978 break;
1979 case SYS_GETPEERNAME:
1980 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1981 break;
1982 case SYS_SOCKETPAIR:
1983 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1984 break;
1985 case SYS_SEND:
1986 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1987 break;
1988 case SYS_SENDTO:
1989 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
1990 (struct sockaddr __user *)a[4], a[5]);
1991 break;
1992 case SYS_RECV:
1993 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
1994 break;
1995 case SYS_RECVFROM:
1996 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
1997 (struct sockaddr __user *)a[4], (int __user *)a[5]);
1998 break;
1999 case SYS_SHUTDOWN:
2000 err = sys_shutdown(a0,a1);
2001 break;
2002 case SYS_SETSOCKOPT:
2003 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2004 break;
2005 case SYS_GETSOCKOPT:
2006 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2007 break;
2008 case SYS_SENDMSG:
2009 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2010 break;
2011 case SYS_RECVMSG:
2012 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2013 break;
2014 default:
2015 err = -EINVAL;
2016 break;
2017 }
2018 return err;
2019}
2020
2021#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2022
2023/*
2024 * This function is called by a protocol handler that wants to
2025 * advertise its address family, and have it linked into the
2026 * SOCKET module.
2027 */
2028
2029int sock_register(struct net_proto_family *ops)
2030{
2031 int err;
2032
2033 if (ops->family >= NPROTO) {
2034 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2035 return -ENOBUFS;
2036 }
2037 net_family_write_lock();
2038 err = -EEXIST;
2039 if (net_families[ops->family] == NULL) {
2040 net_families[ops->family]=ops;
2041 err = 0;
2042 }
2043 net_family_write_unlock();
2044 printk(KERN_INFO "NET: Registered protocol family %d\n",
2045 ops->family);
2046 return err;
2047}
2048
2049/*
2050 * This function is called by a protocol handler that wants to
2051 * remove its address family, and have it unlinked from the
2052 * SOCKET module.
2053 */
2054
2055int sock_unregister(int family)
2056{
2057 if (family < 0 || family >= NPROTO)
2058 return -1;
2059
2060 net_family_write_lock();
2061 net_families[family]=NULL;
2062 net_family_write_unlock();
2063 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2064 family);
2065 return 0;
2066}
2067
Andi Kleen77d76ea2005-12-22 12:43:42 -08002068static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069{
2070 /*
2071 * Initialize sock SLAB cache.
2072 */
2073
2074 sk_init();
2075
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 /*
2077 * Initialize skbuff SLAB cache
2078 */
2079 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080
2081 /*
2082 * Initialize the protocols module.
2083 */
2084
2085 init_inodecache();
2086 register_filesystem(&sock_fs_type);
2087 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002088
2089 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 */
2091
2092#ifdef CONFIG_NETFILTER
2093 netfilter_init();
2094#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002095
2096 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097}
2098
Andi Kleen77d76ea2005-12-22 12:43:42 -08002099core_initcall(sock_init); /* early initcall */
2100
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101#ifdef CONFIG_PROC_FS
2102void socket_seq_show(struct seq_file *seq)
2103{
2104 int cpu;
2105 int counter = 0;
2106
Eric Dumazet88a2a4ac2006-02-04 23:27:36 -08002107 for_each_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 counter += per_cpu(sockets_in_use, cpu);
2109
2110 /* It can be negative, by the way. 8) */
2111 if (counter < 0)
2112 counter = 0;
2113
2114 seq_printf(seq, "sockets: used %d\n", counter);
2115}
2116#endif /* CONFIG_PROC_FS */
2117
2118/* ABI emulation layers need these two */
2119EXPORT_SYMBOL(move_addr_to_kernel);
2120EXPORT_SYMBOL(move_addr_to_user);
2121EXPORT_SYMBOL(sock_create);
2122EXPORT_SYMBOL(sock_create_kern);
2123EXPORT_SYMBOL(sock_create_lite);
2124EXPORT_SYMBOL(sock_map_fd);
2125EXPORT_SYMBOL(sock_recvmsg);
2126EXPORT_SYMBOL(sock_register);
2127EXPORT_SYMBOL(sock_release);
2128EXPORT_SYMBOL(sock_sendmsg);
2129EXPORT_SYMBOL(sock_unregister);
2130EXPORT_SYMBOL(sock_wake_async);
2131EXPORT_SYMBOL(sockfd_lookup);
2132EXPORT_SYMBOL(kernel_sendmsg);
2133EXPORT_SYMBOL(kernel_recvmsg);