[PATCH] deal with deadlocks in audit_free()
[linux-2.6-block.git] / net / socket.c
CommitLineData
1da177e4
LT
1/*
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>
02c30a84 7 * Ross Biro
1da177e4
LT
8 * 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>
4a3e2f71 71#include <linux/mutex.h>
1da177e4
LT
72#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
20380731
ACM
74#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
1da177e4
LT
76#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
81#include <linux/divert.h>
82#include <linux/mount.h>
83#include <linux/security.h>
84#include <linux/syscalls.h>
85#include <linux/compat.h>
86#include <linux/kmod.h>
3ec3b2fb 87#include <linux/audit.h>
d86b5e0e 88#include <linux/wireless.h>
1da177e4
LT
89
90#include <asm/uaccess.h>
91#include <asm/unistd.h>
92
93#include <net/compat.h>
94
95#include <net/sock.h>
96#include <linux/netfilter.h>
97
98static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
99static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
100 size_t size, loff_t pos);
101static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
102 size_t size, loff_t pos);
103static int sock_mmap(struct file *file, struct vm_area_struct * vma);
104
105static int sock_close(struct inode *inode, struct file *file);
106static unsigned int sock_poll(struct file *file,
107 struct poll_table_struct *wait);
108static long sock_ioctl(struct file *file,
109 unsigned int cmd, unsigned long arg);
89bbfc95
SP
110#ifdef CONFIG_COMPAT
111static long compat_sock_ioctl(struct file *file,
112 unsigned int cmd, unsigned long arg);
113#endif
1da177e4
LT
114static int sock_fasync(int fd, struct file *filp, int on);
115static ssize_t sock_readv(struct file *file, const struct iovec *vector,
116 unsigned long count, loff_t *ppos);
117static ssize_t sock_writev(struct file *file, const struct iovec *vector,
118 unsigned long count, loff_t *ppos);
119static ssize_t sock_sendpage(struct file *file, struct page *page,
120 int offset, size_t size, loff_t *ppos, int more);
121
1da177e4
LT
122/*
123 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
124 * in the operation structures but are done directly via the socketcall() multiplexor.
125 */
126
127static struct file_operations socket_file_ops = {
128 .owner = THIS_MODULE,
129 .llseek = no_llseek,
130 .aio_read = sock_aio_read,
131 .aio_write = sock_aio_write,
132 .poll = sock_poll,
133 .unlocked_ioctl = sock_ioctl,
89bbfc95
SP
134#ifdef CONFIG_COMPAT
135 .compat_ioctl = compat_sock_ioctl,
136#endif
1da177e4
LT
137 .mmap = sock_mmap,
138 .open = sock_no_open, /* special open code to disallow open via /proc */
139 .release = sock_close,
140 .fasync = sock_fasync,
141 .readv = sock_readv,
142 .writev = sock_writev,
5274f052
JA
143 .sendpage = sock_sendpage,
144 .splice_write = generic_splice_sendpage,
1da177e4
LT
145};
146
147/*
148 * The protocol list. Each protocol is registered in here.
149 */
150
151static struct net_proto_family *net_families[NPROTO];
152
153#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
154static atomic_t net_family_lockct = ATOMIC_INIT(0);
155static DEFINE_SPINLOCK(net_family_lock);
156
157/* The strategy is: modifications net_family vector are short, do not
158 sleep and veeery rare, but read access should be free of any exclusive
159 locks.
160 */
161
162static void net_family_write_lock(void)
163{
164 spin_lock(&net_family_lock);
165 while (atomic_read(&net_family_lockct) != 0) {
166 spin_unlock(&net_family_lock);
167
168 yield();
169
170 spin_lock(&net_family_lock);
171 }
172}
173
174static __inline__ void net_family_write_unlock(void)
175{
176 spin_unlock(&net_family_lock);
177}
178
179static __inline__ void net_family_read_lock(void)
180{
181 atomic_inc(&net_family_lockct);
182 spin_unlock_wait(&net_family_lock);
183}
184
185static __inline__ void net_family_read_unlock(void)
186{
187 atomic_dec(&net_family_lockct);
188}
189
190#else
191#define net_family_write_lock() do { } while(0)
192#define net_family_write_unlock() do { } while(0)
193#define net_family_read_lock() do { } while(0)
194#define net_family_read_unlock() do { } while(0)
195#endif
196
197
198/*
199 * Statistics counters of the socket lists
200 */
201
202static DEFINE_PER_CPU(int, sockets_in_use) = 0;
203
204/*
205 * Support routines. Move socket addresses back and forth across the kernel/user
206 * divide and look after the messy bits.
207 */
208
209#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
210 16 for IP, 16 for IPX,
211 24 for IPv6,
212 about 80 for AX.25
213 must be at least one bigger than
214 the AF_UNIX size (see net/unix/af_unix.c
215 :unix_mkname()).
216 */
217
218/**
219 * move_addr_to_kernel - copy a socket address into kernel space
220 * @uaddr: Address in user space
221 * @kaddr: Address in kernel space
222 * @ulen: Length in user space
223 *
224 * The address is copied into kernel space. If the provided address is
225 * too long an error code of -EINVAL is returned. If the copy gives
226 * invalid addresses -EFAULT is returned. On a success 0 is returned.
227 */
228
229int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
230{
231 if(ulen<0||ulen>MAX_SOCK_ADDR)
232 return -EINVAL;
233 if(ulen==0)
234 return 0;
235 if(copy_from_user(kaddr,uaddr,ulen))
236 return -EFAULT;
3ec3b2fb 237 return audit_sockaddr(ulen, kaddr);
1da177e4
LT
238}
239
240/**
241 * move_addr_to_user - copy an address to user space
242 * @kaddr: kernel space address
243 * @klen: length of address in kernel
244 * @uaddr: user space address
245 * @ulen: pointer to user length field
246 *
247 * The value pointed to by ulen on entry is the buffer length available.
248 * This is overwritten with the buffer space used. -EINVAL is returned
249 * if an overlong buffer is specified or a negative buffer size. -EFAULT
250 * is returned if either the buffer or the length field are not
251 * accessible.
252 * After copying the data up to the limit the user specifies, the true
253 * length of the data is written over the length limit the user
254 * specified. Zero is returned for a success.
255 */
256
257int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
258{
259 int err;
260 int len;
261
262 if((err=get_user(len, ulen)))
263 return err;
264 if(len>klen)
265 len=klen;
266 if(len<0 || len> MAX_SOCK_ADDR)
267 return -EINVAL;
268 if(len)
269 {
270 if(copy_to_user(uaddr,kaddr,len))
271 return -EFAULT;
272 }
273 /*
274 * "fromlen shall refer to the value before truncation.."
275 * 1003.1g
276 */
277 return __put_user(klen, ulen);
278}
279
280#define SOCKFS_MAGIC 0x534F434B
281
ba89966c 282static kmem_cache_t * sock_inode_cachep __read_mostly;
1da177e4
LT
283
284static struct inode *sock_alloc_inode(struct super_block *sb)
285{
286 struct socket_alloc *ei;
287 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
288 if (!ei)
289 return NULL;
290 init_waitqueue_head(&ei->socket.wait);
291
292 ei->socket.fasync_list = NULL;
293 ei->socket.state = SS_UNCONNECTED;
294 ei->socket.flags = 0;
295 ei->socket.ops = NULL;
296 ei->socket.sk = NULL;
297 ei->socket.file = NULL;
298 ei->socket.flags = 0;
299
300 return &ei->vfs_inode;
301}
302
303static void sock_destroy_inode(struct inode *inode)
304{
305 kmem_cache_free(sock_inode_cachep,
306 container_of(inode, struct socket_alloc, vfs_inode));
307}
308
309static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
310{
311 struct socket_alloc *ei = (struct socket_alloc *) foo;
312
313 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
314 SLAB_CTOR_CONSTRUCTOR)
315 inode_init_once(&ei->vfs_inode);
316}
317
318static int init_inodecache(void)
319{
320 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
321 sizeof(struct socket_alloc),
fffb60f9
PJ
322 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
323 SLAB_MEM_SPREAD),
1da177e4
LT
324 init_once, NULL);
325 if (sock_inode_cachep == NULL)
326 return -ENOMEM;
327 return 0;
328}
329
330static struct super_operations sockfs_ops = {
331 .alloc_inode = sock_alloc_inode,
332 .destroy_inode =sock_destroy_inode,
333 .statfs = simple_statfs,
334};
335
336static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
337 int flags, const char *dev_name, void *data)
338{
339 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
340}
341
ba89966c 342static struct vfsmount *sock_mnt __read_mostly;
1da177e4
LT
343
344static struct file_system_type sock_fs_type = {
345 .name = "sockfs",
346 .get_sb = sockfs_get_sb,
347 .kill_sb = kill_anon_super,
348};
349static int sockfs_delete_dentry(struct dentry *dentry)
350{
351 return 1;
352}
353static struct dentry_operations sockfs_dentry_operations = {
354 .d_delete = sockfs_delete_dentry,
355};
356
357/*
358 * Obtains the first available file descriptor and sets it up for use.
359 *
39d8c1b6
DM
360 * These functions create file structures and maps them to fd space
361 * of the current process. On success it returns file descriptor
1da177e4
LT
362 * and file struct implicitly stored in sock->file.
363 * Note that another thread may close file descriptor before we return
364 * from this function. We use the fact that now we do not refer
365 * to socket after mapping. If one day we will need it, this
366 * function will increment ref. count on file by 1.
367 *
368 * In any case returned fd MAY BE not valid!
369 * This race condition is unavoidable
370 * with shared fd spaces, we cannot solve it inside kernel,
371 * but we take care of internal coherence yet.
372 */
373
39d8c1b6 374static int sock_alloc_fd(struct file **filep)
1da177e4
LT
375{
376 int fd;
1da177e4
LT
377
378 fd = get_unused_fd();
39d8c1b6 379 if (likely(fd >= 0)) {
1da177e4
LT
380 struct file *file = get_empty_filp();
381
39d8c1b6
DM
382 *filep = file;
383 if (unlikely(!file)) {
1da177e4 384 put_unused_fd(fd);
39d8c1b6 385 return -ENFILE;
1da177e4 386 }
39d8c1b6
DM
387 } else
388 *filep = NULL;
389 return fd;
390}
1da177e4 391
39d8c1b6
DM
392static int sock_attach_fd(struct socket *sock, struct file *file)
393{
394 struct qstr this;
395 char name[32];
396
397 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
398 this.name = name;
399 this.hash = SOCK_INODE(sock)->i_ino;
400
401 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
402 if (unlikely(!file->f_dentry))
403 return -ENOMEM;
404
405 file->f_dentry->d_op = &sockfs_dentry_operations;
406 d_add(file->f_dentry, SOCK_INODE(sock));
407 file->f_vfsmnt = mntget(sock_mnt);
408 file->f_mapping = file->f_dentry->d_inode->i_mapping;
409
410 sock->file = file;
411 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
412 file->f_mode = FMODE_READ | FMODE_WRITE;
413 file->f_flags = O_RDWR;
414 file->f_pos = 0;
415 file->private_data = sock;
1da177e4 416
39d8c1b6
DM
417 return 0;
418}
419
420int sock_map_fd(struct socket *sock)
421{
422 struct file *newfile;
423 int fd = sock_alloc_fd(&newfile);
424
425 if (likely(fd >= 0)) {
426 int err = sock_attach_fd(sock, newfile);
427
428 if (unlikely(err < 0)) {
429 put_filp(newfile);
1da177e4 430 put_unused_fd(fd);
39d8c1b6 431 return err;
1da177e4 432 }
39d8c1b6 433 fd_install(fd, newfile);
1da177e4 434 }
1da177e4
LT
435 return fd;
436}
437
6cb153ca
BL
438static struct socket *sock_from_file(struct file *file, int *err)
439{
440 struct inode *inode;
441 struct socket *sock;
442
443 if (file->f_op == &socket_file_ops)
444 return file->private_data; /* set in sock_map_fd */
445
446 inode = file->f_dentry->d_inode;
447 if (!S_ISSOCK(inode->i_mode)) {
448 *err = -ENOTSOCK;
449 return NULL;
450 }
451
452 sock = SOCKET_I(inode);
453 if (sock->file != file) {
454 printk(KERN_ERR "socki_lookup: socket file changed!\n");
455 sock->file = file;
456 }
457 return sock;
458}
459
1da177e4
LT
460/**
461 * sockfd_lookup - Go from a file number to its socket slot
462 * @fd: file handle
463 * @err: pointer to an error code return
464 *
465 * The file handle passed in is locked and the socket it is bound
466 * too is returned. If an error occurs the err pointer is overwritten
467 * with a negative errno code and NULL is returned. The function checks
468 * for both invalid handles and passing a handle which is not a socket.
469 *
470 * On a success the socket object pointer is returned.
471 */
472
473struct socket *sockfd_lookup(int fd, int *err)
474{
475 struct file *file;
1da177e4
LT
476 struct socket *sock;
477
6cb153ca 478 if (!(file = fget(fd))) {
1da177e4
LT
479 *err = -EBADF;
480 return NULL;
481 }
6cb153ca
BL
482 sock = sock_from_file(file, err);
483 if (!sock)
1da177e4 484 fput(file);
6cb153ca
BL
485 return sock;
486}
1da177e4 487
6cb153ca
BL
488static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
489{
490 struct file *file;
491 struct socket *sock;
492
3672558c 493 *err = -EBADF;
6cb153ca
BL
494 file = fget_light(fd, fput_needed);
495 if (file) {
496 sock = sock_from_file(file, err);
497 if (sock)
498 return sock;
499 fput_light(file, *fput_needed);
1da177e4 500 }
6cb153ca 501 return NULL;
1da177e4
LT
502}
503
504/**
505 * sock_alloc - allocate a socket
506 *
507 * Allocate a new inode and socket object. The two are bound together
508 * and initialised. The socket is then returned. If we are out of inodes
509 * NULL is returned.
510 */
511
512static struct socket *sock_alloc(void)
513{
514 struct inode * inode;
515 struct socket * sock;
516
517 inode = new_inode(sock_mnt->mnt_sb);
518 if (!inode)
519 return NULL;
520
521 sock = SOCKET_I(inode);
522
523 inode->i_mode = S_IFSOCK|S_IRWXUGO;
524 inode->i_uid = current->fsuid;
525 inode->i_gid = current->fsgid;
526
527 get_cpu_var(sockets_in_use)++;
528 put_cpu_var(sockets_in_use);
529 return sock;
530}
531
532/*
533 * In theory you can't get an open on this inode, but /proc provides
534 * a back door. Remember to keep it shut otherwise you'll let the
535 * creepy crawlies in.
536 */
537
538static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
539{
540 return -ENXIO;
541}
542
4b6f5d20 543const struct file_operations bad_sock_fops = {
1da177e4
LT
544 .owner = THIS_MODULE,
545 .open = sock_no_open,
546};
547
548/**
549 * sock_release - close a socket
550 * @sock: socket to close
551 *
552 * The socket is released from the protocol stack if it has a release
553 * callback, and the inode is then released if the socket is bound to
554 * an inode not a file.
555 */
556
557void sock_release(struct socket *sock)
558{
559 if (sock->ops) {
560 struct module *owner = sock->ops->owner;
561
562 sock->ops->release(sock);
563 sock->ops = NULL;
564 module_put(owner);
565 }
566
567 if (sock->fasync_list)
568 printk(KERN_ERR "sock_release: fasync list not empty!\n");
569
570 get_cpu_var(sockets_in_use)--;
571 put_cpu_var(sockets_in_use);
572 if (!sock->file) {
573 iput(SOCK_INODE(sock));
574 return;
575 }
576 sock->file=NULL;
577}
578
579static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
580 struct msghdr *msg, size_t size)
581{
582 struct sock_iocb *si = kiocb_to_siocb(iocb);
583 int err;
584
585 si->sock = sock;
586 si->scm = NULL;
587 si->msg = msg;
588 si->size = size;
589
590 err = security_socket_sendmsg(sock, msg, size);
591 if (err)
592 return err;
593
594 return sock->ops->sendmsg(iocb, sock, msg, size);
595}
596
597int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
598{
599 struct kiocb iocb;
600 struct sock_iocb siocb;
601 int ret;
602
603 init_sync_kiocb(&iocb, NULL);
604 iocb.private = &siocb;
605 ret = __sock_sendmsg(&iocb, sock, msg, size);
606 if (-EIOCBQUEUED == ret)
607 ret = wait_on_sync_kiocb(&iocb);
608 return ret;
609}
610
611int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
612 struct kvec *vec, size_t num, size_t size)
613{
614 mm_segment_t oldfs = get_fs();
615 int result;
616
617 set_fs(KERNEL_DS);
618 /*
619 * the following is safe, since for compiler definitions of kvec and
620 * iovec are identical, yielding the same in-core layout and alignment
621 */
622 msg->msg_iov = (struct iovec *)vec,
623 msg->msg_iovlen = num;
624 result = sock_sendmsg(sock, msg, size);
625 set_fs(oldfs);
626 return result;
627}
628
629static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
630 struct msghdr *msg, size_t size, int flags)
631{
632 int err;
633 struct sock_iocb *si = kiocb_to_siocb(iocb);
634
635 si->sock = sock;
636 si->scm = NULL;
637 si->msg = msg;
638 si->size = size;
639 si->flags = flags;
640
641 err = security_socket_recvmsg(sock, msg, size, flags);
642 if (err)
643 return err;
644
645 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
646}
647
648int sock_recvmsg(struct socket *sock, struct msghdr *msg,
649 size_t size, int flags)
650{
651 struct kiocb iocb;
652 struct sock_iocb siocb;
653 int ret;
654
655 init_sync_kiocb(&iocb, NULL);
656 iocb.private = &siocb;
657 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
658 if (-EIOCBQUEUED == ret)
659 ret = wait_on_sync_kiocb(&iocb);
660 return ret;
661}
662
663int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
664 struct kvec *vec, size_t num,
665 size_t size, int flags)
666{
667 mm_segment_t oldfs = get_fs();
668 int result;
669
670 set_fs(KERNEL_DS);
671 /*
672 * the following is safe, since for compiler definitions of kvec and
673 * iovec are identical, yielding the same in-core layout and alignment
674 */
675 msg->msg_iov = (struct iovec *)vec,
676 msg->msg_iovlen = num;
677 result = sock_recvmsg(sock, msg, size, flags);
678 set_fs(oldfs);
679 return result;
680}
681
682static void sock_aio_dtor(struct kiocb *iocb)
683{
684 kfree(iocb->private);
685}
686
ce1d4d3e
CH
687static ssize_t sock_sendpage(struct file *file, struct page *page,
688 int offset, size_t size, loff_t *ppos, int more)
1da177e4 689{
1da177e4
LT
690 struct socket *sock;
691 int flags;
692
ce1d4d3e
CH
693 sock = file->private_data;
694
695 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
696 if (more)
697 flags |= MSG_MORE;
698
699 return sock->ops->sendpage(sock, page, offset, size, flags);
700}
1da177e4 701
ce1d4d3e
CH
702static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
703 char __user *ubuf, size_t size, struct sock_iocb *siocb)
704{
705 if (!is_sync_kiocb(iocb)) {
706 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
707 if (!siocb)
708 return NULL;
1da177e4
LT
709 iocb->ki_dtor = sock_aio_dtor;
710 }
1da177e4 711
ce1d4d3e
CH
712 siocb->kiocb = iocb;
713 siocb->async_iov.iov_base = ubuf;
714 siocb->async_iov.iov_len = size;
1da177e4 715
ce1d4d3e
CH
716 iocb->private = siocb;
717 return siocb;
1da177e4
LT
718}
719
ce1d4d3e
CH
720static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
721 struct file *file, struct iovec *iov, unsigned long nr_segs)
722{
723 struct socket *sock = file->private_data;
724 size_t size = 0;
725 int i;
1da177e4 726
ce1d4d3e
CH
727 for (i = 0 ; i < nr_segs ; i++)
728 size += iov[i].iov_len;
1da177e4 729
ce1d4d3e
CH
730 msg->msg_name = NULL;
731 msg->msg_namelen = 0;
732 msg->msg_control = NULL;
733 msg->msg_controllen = 0;
734 msg->msg_iov = (struct iovec *) iov;
735 msg->msg_iovlen = nr_segs;
736 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
737
738 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
739}
740
741static ssize_t sock_readv(struct file *file, const struct iovec *iov,
742 unsigned long nr_segs, loff_t *ppos)
1da177e4 743{
ce1d4d3e
CH
744 struct kiocb iocb;
745 struct sock_iocb siocb;
746 struct msghdr msg;
747 int ret;
748
749 init_sync_kiocb(&iocb, NULL);
750 iocb.private = &siocb;
751
752 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
753 if (-EIOCBQUEUED == ret)
754 ret = wait_on_sync_kiocb(&iocb);
755 return ret;
756}
757
758static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
759 size_t count, loff_t pos)
760{
761 struct sock_iocb siocb, *x;
762
1da177e4
LT
763 if (pos != 0)
764 return -ESPIPE;
ce1d4d3e 765 if (count == 0) /* Match SYS5 behaviour */
1da177e4
LT
766 return 0;
767
ce1d4d3e
CH
768 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
769 if (!x)
770 return -ENOMEM;
771 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
772 &x->async_iov, 1);
1da177e4
LT
773}
774
ce1d4d3e
CH
775static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
776 struct file *file, struct iovec *iov, unsigned long nr_segs)
1da177e4 777{
ce1d4d3e
CH
778 struct socket *sock = file->private_data;
779 size_t size = 0;
780 int i;
1da177e4 781
ce1d4d3e
CH
782 for (i = 0 ; i < nr_segs ; i++)
783 size += iov[i].iov_len;
1da177e4 784
ce1d4d3e
CH
785 msg->msg_name = NULL;
786 msg->msg_namelen = 0;
787 msg->msg_control = NULL;
788 msg->msg_controllen = 0;
789 msg->msg_iov = (struct iovec *) iov;
790 msg->msg_iovlen = nr_segs;
791 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
792 if (sock->type == SOCK_SEQPACKET)
793 msg->msg_flags |= MSG_EOR;
1da177e4 794
ce1d4d3e 795 return __sock_sendmsg(iocb, sock, msg, size);
1da177e4
LT
796}
797
ce1d4d3e
CH
798static ssize_t sock_writev(struct file *file, const struct iovec *iov,
799 unsigned long nr_segs, loff_t *ppos)
1da177e4
LT
800{
801 struct msghdr msg;
ce1d4d3e
CH
802 struct kiocb iocb;
803 struct sock_iocb siocb;
804 int ret;
1da177e4 805
ce1d4d3e
CH
806 init_sync_kiocb(&iocb, NULL);
807 iocb.private = &siocb;
1da177e4 808
ce1d4d3e
CH
809 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
810 if (-EIOCBQUEUED == ret)
811 ret = wait_on_sync_kiocb(&iocb);
812 return ret;
813}
1da177e4 814
ce1d4d3e
CH
815static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
816 size_t count, loff_t pos)
817{
818 struct sock_iocb siocb, *x;
1da177e4 819
ce1d4d3e
CH
820 if (pos != 0)
821 return -ESPIPE;
822 if (count == 0) /* Match SYS5 behaviour */
823 return 0;
1da177e4 824
ce1d4d3e
CH
825 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
826 if (!x)
827 return -ENOMEM;
1da177e4 828
ce1d4d3e
CH
829 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
830 &x->async_iov, 1);
1da177e4
LT
831}
832
833
834/*
835 * Atomic setting of ioctl hooks to avoid race
836 * with module unload.
837 */
838
4a3e2f71 839static DEFINE_MUTEX(br_ioctl_mutex);
1da177e4
LT
840static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
841
842void brioctl_set(int (*hook)(unsigned int, void __user *))
843{
4a3e2f71 844 mutex_lock(&br_ioctl_mutex);
1da177e4 845 br_ioctl_hook = hook;
4a3e2f71 846 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
847}
848EXPORT_SYMBOL(brioctl_set);
849
4a3e2f71 850static DEFINE_MUTEX(vlan_ioctl_mutex);
1da177e4
LT
851static int (*vlan_ioctl_hook)(void __user *arg);
852
853void vlan_ioctl_set(int (*hook)(void __user *))
854{
4a3e2f71 855 mutex_lock(&vlan_ioctl_mutex);
1da177e4 856 vlan_ioctl_hook = hook;
4a3e2f71 857 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
858}
859EXPORT_SYMBOL(vlan_ioctl_set);
860
4a3e2f71 861static DEFINE_MUTEX(dlci_ioctl_mutex);
1da177e4
LT
862static int (*dlci_ioctl_hook)(unsigned int, void __user *);
863
864void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
865{
4a3e2f71 866 mutex_lock(&dlci_ioctl_mutex);
1da177e4 867 dlci_ioctl_hook = hook;
4a3e2f71 868 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
869}
870EXPORT_SYMBOL(dlci_ioctl_set);
871
872/*
873 * With an ioctl, arg may well be a user mode pointer, but we don't know
874 * what to do with it - that's up to the protocol still.
875 */
876
877static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
878{
879 struct socket *sock;
880 void __user *argp = (void __user *)arg;
881 int pid, err;
882
b69aee04 883 sock = file->private_data;
1da177e4
LT
884 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
885 err = dev_ioctl(cmd, argp);
886 } else
d86b5e0e 887#ifdef CONFIG_WIRELESS_EXT
1da177e4
LT
888 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
889 err = dev_ioctl(cmd, argp);
890 } else
d86b5e0e 891#endif /* CONFIG_WIRELESS_EXT */
1da177e4
LT
892 switch (cmd) {
893 case FIOSETOWN:
894 case SIOCSPGRP:
895 err = -EFAULT;
896 if (get_user(pid, (int __user *)argp))
897 break;
898 err = f_setown(sock->file, pid, 1);
899 break;
900 case FIOGETOWN:
901 case SIOCGPGRP:
902 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
903 break;
904 case SIOCGIFBR:
905 case SIOCSIFBR:
906 case SIOCBRADDBR:
907 case SIOCBRDELBR:
908 err = -ENOPKG;
909 if (!br_ioctl_hook)
910 request_module("bridge");
911
4a3e2f71 912 mutex_lock(&br_ioctl_mutex);
1da177e4
LT
913 if (br_ioctl_hook)
914 err = br_ioctl_hook(cmd, argp);
4a3e2f71 915 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
916 break;
917 case SIOCGIFVLAN:
918 case SIOCSIFVLAN:
919 err = -ENOPKG;
920 if (!vlan_ioctl_hook)
921 request_module("8021q");
922
4a3e2f71 923 mutex_lock(&vlan_ioctl_mutex);
1da177e4
LT
924 if (vlan_ioctl_hook)
925 err = vlan_ioctl_hook(argp);
4a3e2f71 926 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
927 break;
928 case SIOCGIFDIVERT:
929 case SIOCSIFDIVERT:
930 /* Convert this to call through a hook */
931 err = divert_ioctl(cmd, argp);
932 break;
933 case SIOCADDDLCI:
934 case SIOCDELDLCI:
935 err = -ENOPKG;
936 if (!dlci_ioctl_hook)
937 request_module("dlci");
938
939 if (dlci_ioctl_hook) {
4a3e2f71 940 mutex_lock(&dlci_ioctl_mutex);
1da177e4 941 err = dlci_ioctl_hook(cmd, argp);
4a3e2f71 942 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
943 }
944 break;
945 default:
946 err = sock->ops->ioctl(sock, cmd, arg);
b5e5fa5e
CH
947
948 /*
949 * If this ioctl is unknown try to hand it down
950 * to the NIC driver.
951 */
952 if (err == -ENOIOCTLCMD)
953 err = dev_ioctl(cmd, argp);
1da177e4
LT
954 break;
955 }
956 return err;
957}
958
959int sock_create_lite(int family, int type, int protocol, struct socket **res)
960{
961 int err;
962 struct socket *sock = NULL;
963
964 err = security_socket_create(family, type, protocol, 1);
965 if (err)
966 goto out;
967
968 sock = sock_alloc();
969 if (!sock) {
970 err = -ENOMEM;
971 goto out;
972 }
973
974 security_socket_post_create(sock, family, type, protocol, 1);
975 sock->type = type;
976out:
977 *res = sock;
978 return err;
979}
980
981/* No kernel lock held - perfect */
982static unsigned int sock_poll(struct file *file, poll_table * wait)
983{
984 struct socket *sock;
985
986 /*
987 * We can't return errors to poll, so it's either yes or no.
988 */
b69aee04 989 sock = file->private_data;
1da177e4
LT
990 return sock->ops->poll(file, sock, wait);
991}
992
993static int sock_mmap(struct file * file, struct vm_area_struct * vma)
994{
b69aee04 995 struct socket *sock = file->private_data;
1da177e4
LT
996
997 return sock->ops->mmap(file, sock, vma);
998}
999
20380731 1000static int sock_close(struct inode *inode, struct file *filp)
1da177e4
LT
1001{
1002 /*
1003 * It was possible the inode is NULL we were
1004 * closing an unfinished socket.
1005 */
1006
1007 if (!inode)
1008 {
1009 printk(KERN_DEBUG "sock_close: NULL inode\n");
1010 return 0;
1011 }
1012 sock_fasync(-1, filp, 0);
1013 sock_release(SOCKET_I(inode));
1014 return 0;
1015}
1016
1017/*
1018 * Update the socket async list
1019 *
1020 * Fasync_list locking strategy.
1021 *
1022 * 1. fasync_list is modified only under process context socket lock
1023 * i.e. under semaphore.
1024 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1025 * or under socket lock.
1026 * 3. fasync_list can be used from softirq context, so that
1027 * modification under socket lock have to be enhanced with
1028 * write_lock_bh(&sk->sk_callback_lock).
1029 * --ANK (990710)
1030 */
1031
1032static int sock_fasync(int fd, struct file *filp, int on)
1033{
1034 struct fasync_struct *fa, *fna=NULL, **prev;
1035 struct socket *sock;
1036 struct sock *sk;
1037
1038 if (on)
1039 {
8b3a7005 1040 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1da177e4
LT
1041 if(fna==NULL)
1042 return -ENOMEM;
1043 }
1044
b69aee04 1045 sock = filp->private_data;
1da177e4
LT
1046
1047 if ((sk=sock->sk) == NULL) {
1048 kfree(fna);
1049 return -EINVAL;
1050 }
1051
1052 lock_sock(sk);
1053
1054 prev=&(sock->fasync_list);
1055
1056 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1057 if (fa->fa_file==filp)
1058 break;
1059
1060 if(on)
1061 {
1062 if(fa!=NULL)
1063 {
1064 write_lock_bh(&sk->sk_callback_lock);
1065 fa->fa_fd=fd;
1066 write_unlock_bh(&sk->sk_callback_lock);
1067
1068 kfree(fna);
1069 goto out;
1070 }
1071 fna->fa_file=filp;
1072 fna->fa_fd=fd;
1073 fna->magic=FASYNC_MAGIC;
1074 fna->fa_next=sock->fasync_list;
1075 write_lock_bh(&sk->sk_callback_lock);
1076 sock->fasync_list=fna;
1077 write_unlock_bh(&sk->sk_callback_lock);
1078 }
1079 else
1080 {
1081 if (fa!=NULL)
1082 {
1083 write_lock_bh(&sk->sk_callback_lock);
1084 *prev=fa->fa_next;
1085 write_unlock_bh(&sk->sk_callback_lock);
1086 kfree(fa);
1087 }
1088 }
1089
1090out:
1091 release_sock(sock->sk);
1092 return 0;
1093}
1094
1095/* This function may be called only under socket lock or callback_lock */
1096
1097int sock_wake_async(struct socket *sock, int how, int band)
1098{
1099 if (!sock || !sock->fasync_list)
1100 return -1;
1101 switch (how)
1102 {
1103 case 1:
1104
1105 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1106 break;
1107 goto call_kill;
1108 case 2:
1109 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1110 break;
1111 /* fall through */
1112 case 0:
1113 call_kill:
1114 __kill_fasync(sock->fasync_list, SIGIO, band);
1115 break;
1116 case 3:
1117 __kill_fasync(sock->fasync_list, SIGURG, band);
1118 }
1119 return 0;
1120}
1121
1122static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1123{
1124 int err;
1125 struct socket *sock;
1126
1127 /*
1128 * Check protocol is in range
1129 */
1130 if (family < 0 || family >= NPROTO)
1131 return -EAFNOSUPPORT;
1132 if (type < 0 || type >= SOCK_MAX)
1133 return -EINVAL;
1134
1135 /* Compatibility.
1136
1137 This uglymoron is moved from INET layer to here to avoid
1138 deadlock in module load.
1139 */
1140 if (family == PF_INET && type == SOCK_PACKET) {
1141 static int warned;
1142 if (!warned) {
1143 warned = 1;
1144 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1145 }
1146 family = PF_PACKET;
1147 }
1148
1149 err = security_socket_create(family, type, protocol, kern);
1150 if (err)
1151 return err;
1152
1153#if defined(CONFIG_KMOD)
1154 /* Attempt to load a protocol module if the find failed.
1155 *
1156 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1157 * requested real, full-featured networking support upon configuration.
1158 * Otherwise module support will break!
1159 */
1160 if (net_families[family]==NULL)
1161 {
1162 request_module("net-pf-%d",family);
1163 }
1164#endif
1165
1166 net_family_read_lock();
1167 if (net_families[family] == NULL) {
1168 err = -EAFNOSUPPORT;
1169 goto out;
1170 }
1171
1172/*
1173 * Allocate the socket and allow the family to set things up. if
1174 * the protocol is 0, the family is instructed to select an appropriate
1175 * default.
1176 */
1177
1178 if (!(sock = sock_alloc())) {
1179 printk(KERN_WARNING "socket: no more sockets\n");
1180 err = -ENFILE; /* Not exactly a match, but its the
1181 closest posix thing */
1182 goto out;
1183 }
1184
1185 sock->type = type;
1186
1187 /*
1188 * We will call the ->create function, that possibly is in a loadable
1189 * module, so we have to bump that loadable module refcnt first.
1190 */
1191 err = -EAFNOSUPPORT;
1192 if (!try_module_get(net_families[family]->owner))
1193 goto out_release;
1194
a79af59e
FF
1195 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1196 sock->ops = NULL;
1da177e4 1197 goto out_module_put;
a79af59e
FF
1198 }
1199
1da177e4
LT
1200 /*
1201 * Now to bump the refcnt of the [loadable] module that owns this
1202 * socket at sock_release time we decrement its refcnt.
1203 */
1204 if (!try_module_get(sock->ops->owner)) {
1205 sock->ops = NULL;
1206 goto out_module_put;
1207 }
1208 /*
1209 * Now that we're done with the ->create function, the [loadable]
1210 * module can have its refcnt decremented
1211 */
1212 module_put(net_families[family]->owner);
1213 *res = sock;
1214 security_socket_post_create(sock, family, type, protocol, kern);
1215
1216out:
1217 net_family_read_unlock();
1218 return err;
1219out_module_put:
1220 module_put(net_families[family]->owner);
1221out_release:
1222 sock_release(sock);
1223 goto out;
1224}
1225
1226int sock_create(int family, int type, int protocol, struct socket **res)
1227{
1228 return __sock_create(family, type, protocol, res, 0);
1229}
1230
1231int sock_create_kern(int family, int type, int protocol, struct socket **res)
1232{
1233 return __sock_create(family, type, protocol, res, 1);
1234}
1235
1236asmlinkage long sys_socket(int family, int type, int protocol)
1237{
1238 int retval;
1239 struct socket *sock;
1240
1241 retval = sock_create(family, type, protocol, &sock);
1242 if (retval < 0)
1243 goto out;
1244
1245 retval = sock_map_fd(sock);
1246 if (retval < 0)
1247 goto out_release;
1248
1249out:
1250 /* It may be already another descriptor 8) Not kernel problem. */
1251 return retval;
1252
1253out_release:
1254 sock_release(sock);
1255 return retval;
1256}
1257
1258/*
1259 * Create a pair of connected sockets.
1260 */
1261
1262asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1263{
1264 struct socket *sock1, *sock2;
1265 int fd1, fd2, err;
1266
1267 /*
1268 * Obtain the first socket and check if the underlying protocol
1269 * supports the socketpair call.
1270 */
1271
1272 err = sock_create(family, type, protocol, &sock1);
1273 if (err < 0)
1274 goto out;
1275
1276 err = sock_create(family, type, protocol, &sock2);
1277 if (err < 0)
1278 goto out_release_1;
1279
1280 err = sock1->ops->socketpair(sock1, sock2);
1281 if (err < 0)
1282 goto out_release_both;
1283
1284 fd1 = fd2 = -1;
1285
1286 err = sock_map_fd(sock1);
1287 if (err < 0)
1288 goto out_release_both;
1289 fd1 = err;
1290
1291 err = sock_map_fd(sock2);
1292 if (err < 0)
1293 goto out_close_1;
1294 fd2 = err;
1295
1296 /* fd1 and fd2 may be already another descriptors.
1297 * Not kernel problem.
1298 */
1299
1300 err = put_user(fd1, &usockvec[0]);
1301 if (!err)
1302 err = put_user(fd2, &usockvec[1]);
1303 if (!err)
1304 return 0;
1305
1306 sys_close(fd2);
1307 sys_close(fd1);
1308 return err;
1309
1310out_close_1:
1311 sock_release(sock2);
1312 sys_close(fd1);
1313 return err;
1314
1315out_release_both:
1316 sock_release(sock2);
1317out_release_1:
1318 sock_release(sock1);
1319out:
1320 return err;
1321}
1322
1323
1324/*
1325 * Bind a name to a socket. Nothing much to do here since it's
1326 * the protocol's responsibility to handle the local address.
1327 *
1328 * We move the socket address to kernel space before we call
1329 * the protocol layer (having also checked the address is ok).
1330 */
1331
1332asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1333{
1334 struct socket *sock;
1335 char address[MAX_SOCK_ADDR];
6cb153ca 1336 int err, fput_needed;
1da177e4 1337
6cb153ca 1338 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1da177e4
LT
1339 {
1340 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1341 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
6cb153ca
BL
1342 if (!err)
1343 err = sock->ops->bind(sock,
1344 (struct sockaddr *)address, addrlen);
1da177e4 1345 }
6cb153ca 1346 fput_light(sock->file, fput_needed);
1da177e4
LT
1347 }
1348 return err;
1349}
1350
1351
1352/*
1353 * Perform a listen. Basically, we allow the protocol to do anything
1354 * necessary for a listen, and if that works, we mark the socket as
1355 * ready for listening.
1356 */
1357
1358int sysctl_somaxconn = SOMAXCONN;
1359
1360asmlinkage long sys_listen(int fd, int backlog)
1361{
1362 struct socket *sock;
6cb153ca 1363 int err, fput_needed;
1da177e4 1364
6cb153ca 1365 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1da177e4
LT
1366 if ((unsigned) backlog > sysctl_somaxconn)
1367 backlog = sysctl_somaxconn;
1368
1369 err = security_socket_listen(sock, backlog);
6cb153ca
BL
1370 if (!err)
1371 err = sock->ops->listen(sock, backlog);
1da177e4 1372
6cb153ca 1373 fput_light(sock->file, fput_needed);
1da177e4
LT
1374 }
1375 return err;
1376}
1377
1378
1379/*
1380 * For accept, we attempt to create a new socket, set up the link
1381 * with the client, wake up the client, then return the new
1382 * connected fd. We collect the address of the connector in kernel
1383 * space and move it to user at the very end. This is unclean because
1384 * we open the socket then return an error.
1385 *
1386 * 1003.1g adds the ability to recvmsg() to query connection pending
1387 * status to recvmsg. We need to add that support in a way thats
1388 * clean when we restucture accept also.
1389 */
1390
1391asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1392{
1393 struct socket *sock, *newsock;
39d8c1b6 1394 struct file *newfile;
6cb153ca 1395 int err, len, newfd, fput_needed;
1da177e4
LT
1396 char address[MAX_SOCK_ADDR];
1397
6cb153ca 1398 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1399 if (!sock)
1400 goto out;
1401
1402 err = -ENFILE;
1403 if (!(newsock = sock_alloc()))
1404 goto out_put;
1405
1406 newsock->type = sock->type;
1407 newsock->ops = sock->ops;
1408
1da177e4
LT
1409 /*
1410 * We don't need try_module_get here, as the listening socket (sock)
1411 * has the protocol module (sock->ops->owner) held.
1412 */
1413 __module_get(newsock->ops->owner);
1414
39d8c1b6
DM
1415 newfd = sock_alloc_fd(&newfile);
1416 if (unlikely(newfd < 0)) {
1417 err = newfd;
9a1875e6
DM
1418 sock_release(newsock);
1419 goto out_put;
39d8c1b6
DM
1420 }
1421
1422 err = sock_attach_fd(newsock, newfile);
1423 if (err < 0)
1424 goto out_fd;
1425
a79af59e
FF
1426 err = security_socket_accept(sock, newsock);
1427 if (err)
39d8c1b6 1428 goto out_fd;
a79af59e 1429
1da177e4
LT
1430 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1431 if (err < 0)
39d8c1b6 1432 goto out_fd;
1da177e4
LT
1433
1434 if (upeer_sockaddr) {
1435 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1436 err = -ECONNABORTED;
39d8c1b6 1437 goto out_fd;
1da177e4
LT
1438 }
1439 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1440 if (err < 0)
39d8c1b6 1441 goto out_fd;
1da177e4
LT
1442 }
1443
1444 /* File flags are not inherited via accept() unlike another OSes. */
1445
39d8c1b6
DM
1446 fd_install(newfd, newfile);
1447 err = newfd;
1da177e4
LT
1448
1449 security_socket_post_accept(sock, newsock);
1450
1451out_put:
6cb153ca 1452 fput_light(sock->file, fput_needed);
1da177e4
LT
1453out:
1454 return err;
39d8c1b6 1455out_fd:
9606a216 1456 fput(newfile);
39d8c1b6 1457 put_unused_fd(newfd);
1da177e4
LT
1458 goto out_put;
1459}
1460
1461
1462/*
1463 * Attempt to connect to a socket with the server address. The address
1464 * is in user space so we verify it is OK and move it to kernel space.
1465 *
1466 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1467 * break bindings
1468 *
1469 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1470 * other SEQPACKET protocols that take time to connect() as it doesn't
1471 * include the -EINPROGRESS status for such sockets.
1472 */
1473
1474asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1475{
1476 struct socket *sock;
1477 char address[MAX_SOCK_ADDR];
6cb153ca 1478 int err, fput_needed;
1da177e4 1479
6cb153ca 1480 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1481 if (!sock)
1482 goto out;
1483 err = move_addr_to_kernel(uservaddr, addrlen, address);
1484 if (err < 0)
1485 goto out_put;
1486
1487 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1488 if (err)
1489 goto out_put;
1490
1491 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1492 sock->file->f_flags);
1493out_put:
6cb153ca 1494 fput_light(sock->file, fput_needed);
1da177e4
LT
1495out:
1496 return err;
1497}
1498
1499/*
1500 * Get the local address ('name') of a socket object. Move the obtained
1501 * name to user space.
1502 */
1503
1504asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1505{
1506 struct socket *sock;
1507 char address[MAX_SOCK_ADDR];
6cb153ca 1508 int len, err, fput_needed;
1da177e4 1509
6cb153ca 1510 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1511 if (!sock)
1512 goto out;
1513
1514 err = security_socket_getsockname(sock);
1515 if (err)
1516 goto out_put;
1517
1518 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1519 if (err)
1520 goto out_put;
1521 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1522
1523out_put:
6cb153ca 1524 fput_light(sock->file, fput_needed);
1da177e4
LT
1525out:
1526 return err;
1527}
1528
1529/*
1530 * Get the remote address ('name') of a socket object. Move the obtained
1531 * name to user space.
1532 */
1533
1534asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1535{
1536 struct socket *sock;
1537 char address[MAX_SOCK_ADDR];
6cb153ca 1538 int len, err, fput_needed;
1da177e4 1539
6cb153ca 1540 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1da177e4
LT
1541 err = security_socket_getpeername(sock);
1542 if (err) {
6cb153ca 1543 fput_light(sock->file, fput_needed);
1da177e4
LT
1544 return err;
1545 }
1546
1547 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1548 if (!err)
1549 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
6cb153ca 1550 fput_light(sock->file, fput_needed);
1da177e4
LT
1551 }
1552 return err;
1553}
1554
1555/*
1556 * Send a datagram to a given address. We move the address into kernel
1557 * space and check the user space data area is readable before invoking
1558 * the protocol.
1559 */
1560
1561asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1562 struct sockaddr __user *addr, int addr_len)
1563{
1564 struct socket *sock;
1565 char address[MAX_SOCK_ADDR];
1566 int err;
1567 struct msghdr msg;
1568 struct iovec iov;
6cb153ca
BL
1569 int fput_needed;
1570 struct file *sock_file;
1571
1572 sock_file = fget_light(fd, &fput_needed);
1573 if (!sock_file)
1574 return -EBADF;
1575
1576 sock = sock_from_file(sock_file, &err);
1da177e4 1577 if (!sock)
6cb153ca 1578 goto out_put;
1da177e4
LT
1579 iov.iov_base=buff;
1580 iov.iov_len=len;
1581 msg.msg_name=NULL;
1582 msg.msg_iov=&iov;
1583 msg.msg_iovlen=1;
1584 msg.msg_control=NULL;
1585 msg.msg_controllen=0;
1586 msg.msg_namelen=0;
6cb153ca 1587 if (addr) {
1da177e4
LT
1588 err = move_addr_to_kernel(addr, addr_len, address);
1589 if (err < 0)
1590 goto out_put;
1591 msg.msg_name=address;
1592 msg.msg_namelen=addr_len;
1593 }
1594 if (sock->file->f_flags & O_NONBLOCK)
1595 flags |= MSG_DONTWAIT;
1596 msg.msg_flags = flags;
1597 err = sock_sendmsg(sock, &msg, len);
1598
1599out_put:
6cb153ca 1600 fput_light(sock_file, fput_needed);
1da177e4
LT
1601 return err;
1602}
1603
1604/*
1605 * Send a datagram down a socket.
1606 */
1607
1608asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1609{
1610 return sys_sendto(fd, buff, len, flags, NULL, 0);
1611}
1612
1613/*
1614 * Receive a frame from the socket and optionally record the address of the
1615 * sender. We verify the buffers are writable and if needed move the
1616 * sender address from kernel to user space.
1617 */
1618
1619asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1620 struct sockaddr __user *addr, int __user *addr_len)
1621{
1622 struct socket *sock;
1623 struct iovec iov;
1624 struct msghdr msg;
1625 char address[MAX_SOCK_ADDR];
1626 int err,err2;
6cb153ca
BL
1627 struct file *sock_file;
1628 int fput_needed;
1629
1630 sock_file = fget_light(fd, &fput_needed);
1631 if (!sock_file)
1632 return -EBADF;
1da177e4 1633
6cb153ca 1634 sock = sock_from_file(sock_file, &err);
1da177e4
LT
1635 if (!sock)
1636 goto out;
1637
1638 msg.msg_control=NULL;
1639 msg.msg_controllen=0;
1640 msg.msg_iovlen=1;
1641 msg.msg_iov=&iov;
1642 iov.iov_len=size;
1643 iov.iov_base=ubuf;
1644 msg.msg_name=address;
1645 msg.msg_namelen=MAX_SOCK_ADDR;
1646 if (sock->file->f_flags & O_NONBLOCK)
1647 flags |= MSG_DONTWAIT;
1648 err=sock_recvmsg(sock, &msg, size, flags);
1649
1650 if(err >= 0 && addr != NULL)
1651 {
1652 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1653 if(err2<0)
1654 err=err2;
1655 }
1da177e4 1656out:
6cb153ca 1657 fput_light(sock_file, fput_needed);
1da177e4
LT
1658 return err;
1659}
1660
1661/*
1662 * Receive a datagram from a socket.
1663 */
1664
1665asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1666{
1667 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1668}
1669
1670/*
1671 * Set a socket option. Because we don't know the option lengths we have
1672 * to pass the user mode parameter for the protocols to sort out.
1673 */
1674
1675asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1676{
6cb153ca 1677 int err, fput_needed;
1da177e4
LT
1678 struct socket *sock;
1679
1680 if (optlen < 0)
1681 return -EINVAL;
1682
6cb153ca 1683 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
1da177e4
LT
1684 {
1685 err = security_socket_setsockopt(sock,level,optname);
6cb153ca
BL
1686 if (err)
1687 goto out_put;
1da177e4
LT
1688
1689 if (level == SOL_SOCKET)
1690 err=sock_setsockopt(sock,level,optname,optval,optlen);
1691 else
1692 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
6cb153ca
BL
1693out_put:
1694 fput_light(sock->file, fput_needed);
1da177e4
LT
1695 }
1696 return err;
1697}
1698
1699/*
1700 * Get a socket option. Because we don't know the option lengths we have
1701 * to pass a user mode parameter for the protocols to sort out.
1702 */
1703
1704asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1705{
6cb153ca 1706 int err, fput_needed;
1da177e4
LT
1707 struct socket *sock;
1708
6cb153ca
BL
1709 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1710 err = security_socket_getsockopt(sock, level, optname);
1711 if (err)
1712 goto out_put;
1da177e4
LT
1713
1714 if (level == SOL_SOCKET)
1715 err=sock_getsockopt(sock,level,optname,optval,optlen);
1716 else
1717 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
6cb153ca
BL
1718out_put:
1719 fput_light(sock->file, fput_needed);
1da177e4
LT
1720 }
1721 return err;
1722}
1723
1724
1725/*
1726 * Shutdown a socket.
1727 */
1728
1729asmlinkage long sys_shutdown(int fd, int how)
1730{
6cb153ca 1731 int err, fput_needed;
1da177e4
LT
1732 struct socket *sock;
1733
6cb153ca 1734 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1da177e4
LT
1735 {
1736 err = security_socket_shutdown(sock, how);
6cb153ca
BL
1737 if (!err)
1738 err = sock->ops->shutdown(sock, how);
1739 fput_light(sock->file, fput_needed);
1da177e4
LT
1740 }
1741 return err;
1742}
1743
1744/* A couple of helpful macros for getting the address of the 32/64 bit
1745 * fields which are the same type (int / unsigned) on our platforms.
1746 */
1747#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1748#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1749#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1750
1751
1752/*
1753 * BSD sendmsg interface
1754 */
1755
1756asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1757{
1758 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1759 struct socket *sock;
1760 char address[MAX_SOCK_ADDR];
1761 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
b9d717a7
AW
1762 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1763 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1764 /* 20 is size of ipv6_pktinfo */
1da177e4
LT
1765 unsigned char *ctl_buf = ctl;
1766 struct msghdr msg_sys;
1767 int err, ctl_len, iov_size, total_len;
6cb153ca 1768 int fput_needed;
1da177e4
LT
1769
1770 err = -EFAULT;
1771 if (MSG_CMSG_COMPAT & flags) {
1772 if (get_compat_msghdr(&msg_sys, msg_compat))
1773 return -EFAULT;
1774 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1775 return -EFAULT;
1776
6cb153ca 1777 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1778 if (!sock)
1779 goto out;
1780
1781 /* do not move before msg_sys is valid */
1782 err = -EMSGSIZE;
1783 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1784 goto out_put;
1785
1786 /* Check whether to allocate the iovec area*/
1787 err = -ENOMEM;
1788 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1789 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1790 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1791 if (!iov)
1792 goto out_put;
1793 }
1794
1795 /* This will also move the address data into kernel space */
1796 if (MSG_CMSG_COMPAT & flags) {
1797 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1798 } else
1799 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1800 if (err < 0)
1801 goto out_freeiov;
1802 total_len = err;
1803
1804 err = -ENOBUFS;
1805
1806 if (msg_sys.msg_controllen > INT_MAX)
1807 goto out_freeiov;
1808 ctl_len = msg_sys.msg_controllen;
1809 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
8920e8f9 1810 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
1da177e4
LT
1811 if (err)
1812 goto out_freeiov;
1813 ctl_buf = msg_sys.msg_control;
8920e8f9 1814 ctl_len = msg_sys.msg_controllen;
1da177e4
LT
1815 } else if (ctl_len) {
1816 if (ctl_len > sizeof(ctl))
1817 {
1818 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1819 if (ctl_buf == NULL)
1820 goto out_freeiov;
1821 }
1822 err = -EFAULT;
1823 /*
1824 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1825 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1826 * checking falls down on this.
1827 */
1828 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1829 goto out_freectl;
1830 msg_sys.msg_control = ctl_buf;
1831 }
1832 msg_sys.msg_flags = flags;
1833
1834 if (sock->file->f_flags & O_NONBLOCK)
1835 msg_sys.msg_flags |= MSG_DONTWAIT;
1836 err = sock_sendmsg(sock, &msg_sys, total_len);
1837
1838out_freectl:
1839 if (ctl_buf != ctl)
1840 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1841out_freeiov:
1842 if (iov != iovstack)
1843 sock_kfree_s(sock->sk, iov, iov_size);
1844out_put:
6cb153ca 1845 fput_light(sock->file, fput_needed);
1da177e4
LT
1846out:
1847 return err;
1848}
1849
1850/*
1851 * BSD recvmsg interface
1852 */
1853
1854asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1855{
1856 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1857 struct socket *sock;
1858 struct iovec iovstack[UIO_FASTIOV];
1859 struct iovec *iov=iovstack;
1860 struct msghdr msg_sys;
1861 unsigned long cmsg_ptr;
1862 int err, iov_size, total_len, len;
6cb153ca 1863 int fput_needed;
1da177e4
LT
1864
1865 /* kernel mode address */
1866 char addr[MAX_SOCK_ADDR];
1867
1868 /* user mode address pointers */
1869 struct sockaddr __user *uaddr;
1870 int __user *uaddr_len;
1871
1872 if (MSG_CMSG_COMPAT & flags) {
1873 if (get_compat_msghdr(&msg_sys, msg_compat))
1874 return -EFAULT;
1875 } else
1876 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1877 return -EFAULT;
1878
6cb153ca 1879 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1880 if (!sock)
1881 goto out;
1882
1883 err = -EMSGSIZE;
1884 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1885 goto out_put;
1886
1887 /* Check whether to allocate the iovec area*/
1888 err = -ENOMEM;
1889 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1890 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1891 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1892 if (!iov)
1893 goto out_put;
1894 }
1895
1896 /*
1897 * Save the user-mode address (verify_iovec will change the
1898 * kernel msghdr to use the kernel address space)
1899 */
1900
1901 uaddr = (void __user *) msg_sys.msg_name;
1902 uaddr_len = COMPAT_NAMELEN(msg);
1903 if (MSG_CMSG_COMPAT & flags) {
1904 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1905 } else
1906 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1907 if (err < 0)
1908 goto out_freeiov;
1909 total_len=err;
1910
1911 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1912 msg_sys.msg_flags = 0;
1913 if (MSG_CMSG_COMPAT & flags)
1914 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1915
1916 if (sock->file->f_flags & O_NONBLOCK)
1917 flags |= MSG_DONTWAIT;
1918 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1919 if (err < 0)
1920 goto out_freeiov;
1921 len = err;
1922
1923 if (uaddr != NULL) {
1924 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1925 if (err < 0)
1926 goto out_freeiov;
1927 }
37f7f421
DM
1928 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1929 COMPAT_FLAGS(msg));
1da177e4
LT
1930 if (err)
1931 goto out_freeiov;
1932 if (MSG_CMSG_COMPAT & flags)
1933 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1934 &msg_compat->msg_controllen);
1935 else
1936 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1937 &msg->msg_controllen);
1938 if (err)
1939 goto out_freeiov;
1940 err = len;
1941
1942out_freeiov:
1943 if (iov != iovstack)
1944 sock_kfree_s(sock->sk, iov, iov_size);
1945out_put:
6cb153ca 1946 fput_light(sock->file, fput_needed);
1da177e4
LT
1947out:
1948 return err;
1949}
1950
1951#ifdef __ARCH_WANT_SYS_SOCKETCALL
1952
1953/* Argument list sizes for sys_socketcall */
1954#define AL(x) ((x) * sizeof(unsigned long))
1955static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1956 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1957 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1958#undef AL
1959
1960/*
1961 * System call vectors.
1962 *
1963 * Argument checking cleaned up. Saved 20% in size.
1964 * This function doesn't need to set the kernel lock because
1965 * it is set by the callees.
1966 */
1967
1968asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1969{
1970 unsigned long a[6];
1971 unsigned long a0,a1;
1972 int err;
1973
1974 if(call<1||call>SYS_RECVMSG)
1975 return -EINVAL;
1976
1977 /* copy_from_user should be SMP safe. */
1978 if (copy_from_user(a, args, nargs[call]))
1979 return -EFAULT;
3ec3b2fb 1980
4bcff1b3 1981 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
3ec3b2fb
DW
1982 if (err)
1983 return err;
1984
1da177e4
LT
1985 a0=a[0];
1986 a1=a[1];
1987
1988 switch(call)
1989 {
1990 case SYS_SOCKET:
1991 err = sys_socket(a0,a1,a[2]);
1992 break;
1993 case SYS_BIND:
1994 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1995 break;
1996 case SYS_CONNECT:
1997 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1998 break;
1999 case SYS_LISTEN:
2000 err = sys_listen(a0,a1);
2001 break;
2002 case SYS_ACCEPT:
2003 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2004 break;
2005 case SYS_GETSOCKNAME:
2006 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2007 break;
2008 case SYS_GETPEERNAME:
2009 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2010 break;
2011 case SYS_SOCKETPAIR:
2012 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2013 break;
2014 case SYS_SEND:
2015 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2016 break;
2017 case SYS_SENDTO:
2018 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2019 (struct sockaddr __user *)a[4], a[5]);
2020 break;
2021 case SYS_RECV:
2022 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2023 break;
2024 case SYS_RECVFROM:
2025 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2026 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2027 break;
2028 case SYS_SHUTDOWN:
2029 err = sys_shutdown(a0,a1);
2030 break;
2031 case SYS_SETSOCKOPT:
2032 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2033 break;
2034 case SYS_GETSOCKOPT:
2035 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2036 break;
2037 case SYS_SENDMSG:
2038 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2039 break;
2040 case SYS_RECVMSG:
2041 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2042 break;
2043 default:
2044 err = -EINVAL;
2045 break;
2046 }
2047 return err;
2048}
2049
2050#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2051
2052/*
2053 * This function is called by a protocol handler that wants to
2054 * advertise its address family, and have it linked into the
2055 * SOCKET module.
2056 */
2057
2058int sock_register(struct net_proto_family *ops)
2059{
2060 int err;
2061
2062 if (ops->family >= NPROTO) {
2063 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2064 return -ENOBUFS;
2065 }
2066 net_family_write_lock();
2067 err = -EEXIST;
2068 if (net_families[ops->family] == NULL) {
2069 net_families[ops->family]=ops;
2070 err = 0;
2071 }
2072 net_family_write_unlock();
2073 printk(KERN_INFO "NET: Registered protocol family %d\n",
2074 ops->family);
2075 return err;
2076}
2077
2078/*
2079 * This function is called by a protocol handler that wants to
2080 * remove its address family, and have it unlinked from the
2081 * SOCKET module.
2082 */
2083
2084int sock_unregister(int family)
2085{
2086 if (family < 0 || family >= NPROTO)
2087 return -1;
2088
2089 net_family_write_lock();
2090 net_families[family]=NULL;
2091 net_family_write_unlock();
2092 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2093 family);
2094 return 0;
2095}
2096
77d76ea3 2097static int __init sock_init(void)
1da177e4
LT
2098{
2099 /*
2100 * Initialize sock SLAB cache.
2101 */
2102
2103 sk_init();
2104
1da177e4
LT
2105 /*
2106 * Initialize skbuff SLAB cache
2107 */
2108 skb_init();
1da177e4
LT
2109
2110 /*
2111 * Initialize the protocols module.
2112 */
2113
2114 init_inodecache();
2115 register_filesystem(&sock_fs_type);
2116 sock_mnt = kern_mount(&sock_fs_type);
77d76ea3
AK
2117
2118 /* The real protocol initialization is performed in later initcalls.
1da177e4
LT
2119 */
2120
2121#ifdef CONFIG_NETFILTER
2122 netfilter_init();
2123#endif
cbeb321a
DM
2124
2125 return 0;
1da177e4
LT
2126}
2127
77d76ea3
AK
2128core_initcall(sock_init); /* early initcall */
2129
1da177e4
LT
2130#ifdef CONFIG_PROC_FS
2131void socket_seq_show(struct seq_file *seq)
2132{
2133 int cpu;
2134 int counter = 0;
2135
6f912042 2136 for_each_possible_cpu(cpu)
1da177e4
LT
2137 counter += per_cpu(sockets_in_use, cpu);
2138
2139 /* It can be negative, by the way. 8) */
2140 if (counter < 0)
2141 counter = 0;
2142
2143 seq_printf(seq, "sockets: used %d\n", counter);
2144}
2145#endif /* CONFIG_PROC_FS */
2146
89bbfc95
SP
2147#ifdef CONFIG_COMPAT
2148static long compat_sock_ioctl(struct file *file, unsigned cmd,
2149 unsigned long arg)
2150{
2151 struct socket *sock = file->private_data;
2152 int ret = -ENOIOCTLCMD;
2153
2154 if (sock->ops->compat_ioctl)
2155 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2156
2157 return ret;
2158}
2159#endif
2160
1da177e4
LT
2161/* ABI emulation layers need these two */
2162EXPORT_SYMBOL(move_addr_to_kernel);
2163EXPORT_SYMBOL(move_addr_to_user);
2164EXPORT_SYMBOL(sock_create);
2165EXPORT_SYMBOL(sock_create_kern);
2166EXPORT_SYMBOL(sock_create_lite);
2167EXPORT_SYMBOL(sock_map_fd);
2168EXPORT_SYMBOL(sock_recvmsg);
2169EXPORT_SYMBOL(sock_register);
2170EXPORT_SYMBOL(sock_release);
2171EXPORT_SYMBOL(sock_sendmsg);
2172EXPORT_SYMBOL(sock_unregister);
2173EXPORT_SYMBOL(sock_wake_async);
2174EXPORT_SYMBOL(sockfd_lookup);
2175EXPORT_SYMBOL(kernel_sendmsg);
2176EXPORT_SYMBOL(kernel_recvmsg);