[NEIGH]: Move netlink neighbour table bits to linux/neighbour.h
[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
1da177e4
LT
61#include <linux/mm.h>
62#include <linux/smp_lock.h>
63#include <linux/socket.h>
64#include <linux/file.h>
65#include <linux/net.h>
66#include <linux/interrupt.h>
67#include <linux/netdevice.h>
68#include <linux/proc_fs.h>
69#include <linux/seq_file.h>
4a3e2f71 70#include <linux/mutex.h>
1da177e4
LT
71#include <linux/wanrouter.h>
72#include <linux/if_bridge.h>
20380731
ACM
73#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
1da177e4
LT
75#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>
3ec3b2fb 86#include <linux/audit.h>
d86b5e0e 87#include <linux/wireless.h>
1da177e4
LT
88
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);
89bbfc95
SP
109#ifdef CONFIG_COMPAT
110static long compat_sock_ioctl(struct file *file,
111 unsigned int cmd, unsigned long arg);
112#endif
1da177e4
LT
113static int sock_fasync(int fd, struct file *filp, int on);
114static ssize_t sock_readv(struct file *file, const struct iovec *vector,
115 unsigned long count, loff_t *ppos);
116static ssize_t sock_writev(struct file *file, const struct iovec *vector,
117 unsigned long count, loff_t *ppos);
118static ssize_t sock_sendpage(struct file *file, struct page *page,
119 int offset, size_t size, loff_t *ppos, int more);
120
1da177e4
LT
121/*
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
124 */
125
126static struct file_operations socket_file_ops = {
127 .owner = THIS_MODULE,
128 .llseek = no_llseek,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
131 .poll = sock_poll,
132 .unlocked_ioctl = sock_ioctl,
89bbfc95
SP
133#ifdef CONFIG_COMPAT
134 .compat_ioctl = compat_sock_ioctl,
135#endif
1da177e4
LT
136 .mmap = sock_mmap,
137 .open = sock_no_open, /* special open code to disallow open via /proc */
138 .release = sock_close,
139 .fasync = sock_fasync,
140 .readv = sock_readv,
141 .writev = sock_writev,
5274f052
JA
142 .sendpage = sock_sendpage,
143 .splice_write = generic_splice_sendpage,
1da177e4
LT
144};
145
146/*
147 * The protocol list. Each protocol is registered in here.
148 */
149
150static struct net_proto_family *net_families[NPROTO];
151
152#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
153static atomic_t net_family_lockct = ATOMIC_INIT(0);
154static DEFINE_SPINLOCK(net_family_lock);
155
156/* The strategy is: modifications net_family vector are short, do not
157 sleep and veeery rare, but read access should be free of any exclusive
158 locks.
159 */
160
161static void net_family_write_lock(void)
162{
163 spin_lock(&net_family_lock);
164 while (atomic_read(&net_family_lockct) != 0) {
165 spin_unlock(&net_family_lock);
166
167 yield();
168
169 spin_lock(&net_family_lock);
170 }
171}
172
173static __inline__ void net_family_write_unlock(void)
174{
175 spin_unlock(&net_family_lock);
176}
177
178static __inline__ void net_family_read_lock(void)
179{
180 atomic_inc(&net_family_lockct);
181 spin_unlock_wait(&net_family_lock);
182}
183
184static __inline__ void net_family_read_unlock(void)
185{
186 atomic_dec(&net_family_lockct);
187}
188
189#else
190#define net_family_write_lock() do { } while(0)
191#define net_family_write_unlock() do { } while(0)
192#define net_family_read_lock() do { } while(0)
193#define net_family_read_unlock() do { } while(0)
194#endif
195
196
197/*
198 * Statistics counters of the socket lists
199 */
200
201static DEFINE_PER_CPU(int, sockets_in_use) = 0;
202
203/*
204 * Support routines. Move socket addresses back and forth across the kernel/user
205 * divide and look after the messy bits.
206 */
207
208#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
209 16 for IP, 16 for IPX,
210 24 for IPv6,
211 about 80 for AX.25
212 must be at least one bigger than
213 the AF_UNIX size (see net/unix/af_unix.c
214 :unix_mkname()).
215 */
216
217/**
218 * move_addr_to_kernel - copy a socket address into kernel space
219 * @uaddr: Address in user space
220 * @kaddr: Address in kernel space
221 * @ulen: Length in user space
222 *
223 * The address is copied into kernel space. If the provided address is
224 * too long an error code of -EINVAL is returned. If the copy gives
225 * invalid addresses -EFAULT is returned. On a success 0 is returned.
226 */
227
228int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
229{
230 if(ulen<0||ulen>MAX_SOCK_ADDR)
231 return -EINVAL;
232 if(ulen==0)
233 return 0;
234 if(copy_from_user(kaddr,uaddr,ulen))
235 return -EFAULT;
3ec3b2fb 236 return audit_sockaddr(ulen, kaddr);
1da177e4
LT
237}
238
239/**
240 * move_addr_to_user - copy an address to user space
241 * @kaddr: kernel space address
242 * @klen: length of address in kernel
243 * @uaddr: user space address
244 * @ulen: pointer to user length field
245 *
246 * The value pointed to by ulen on entry is the buffer length available.
247 * This is overwritten with the buffer space used. -EINVAL is returned
248 * if an overlong buffer is specified or a negative buffer size. -EFAULT
249 * is returned if either the buffer or the length field are not
250 * accessible.
251 * After copying the data up to the limit the user specifies, the true
252 * length of the data is written over the length limit the user
253 * specified. Zero is returned for a success.
254 */
255
256int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
257{
258 int err;
259 int len;
260
261 if((err=get_user(len, ulen)))
262 return err;
263 if(len>klen)
264 len=klen;
265 if(len<0 || len> MAX_SOCK_ADDR)
266 return -EINVAL;
267 if(len)
268 {
d6fe3945
SG
269 if (audit_sockaddr(klen, kaddr))
270 return -ENOMEM;
1da177e4
LT
271 if(copy_to_user(uaddr,kaddr,len))
272 return -EFAULT;
273 }
274 /*
275 * "fromlen shall refer to the value before truncation.."
276 * 1003.1g
277 */
278 return __put_user(klen, ulen);
279}
280
281#define SOCKFS_MAGIC 0x534F434B
282
ba89966c 283static kmem_cache_t * sock_inode_cachep __read_mostly;
1da177e4
LT
284
285static struct inode *sock_alloc_inode(struct super_block *sb)
286{
287 struct socket_alloc *ei;
288 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
289 if (!ei)
290 return NULL;
291 init_waitqueue_head(&ei->socket.wait);
292
293 ei->socket.fasync_list = NULL;
294 ei->socket.state = SS_UNCONNECTED;
295 ei->socket.flags = 0;
296 ei->socket.ops = NULL;
297 ei->socket.sk = NULL;
298 ei->socket.file = NULL;
299 ei->socket.flags = 0;
300
301 return &ei->vfs_inode;
302}
303
304static void sock_destroy_inode(struct inode *inode)
305{
306 kmem_cache_free(sock_inode_cachep,
307 container_of(inode, struct socket_alloc, vfs_inode));
308}
309
310static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
311{
312 struct socket_alloc *ei = (struct socket_alloc *) foo;
313
314 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
315 SLAB_CTOR_CONSTRUCTOR)
316 inode_init_once(&ei->vfs_inode);
317}
318
319static int init_inodecache(void)
320{
321 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
322 sizeof(struct socket_alloc),
fffb60f9
PJ
323 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
324 SLAB_MEM_SPREAD),
1da177e4
LT
325 init_once, NULL);
326 if (sock_inode_cachep == NULL)
327 return -ENOMEM;
328 return 0;
329}
330
331static struct super_operations sockfs_ops = {
332 .alloc_inode = sock_alloc_inode,
333 .destroy_inode =sock_destroy_inode,
334 .statfs = simple_statfs,
335};
336
454e2398
DH
337static int sockfs_get_sb(struct file_system_type *fs_type,
338 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
1da177e4 339{
454e2398
DH
340 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
341 mnt);
1da177e4
LT
342}
343
ba89966c 344static struct vfsmount *sock_mnt __read_mostly;
1da177e4
LT
345
346static struct file_system_type sock_fs_type = {
347 .name = "sockfs",
348 .get_sb = sockfs_get_sb,
349 .kill_sb = kill_anon_super,
350};
351static int sockfs_delete_dentry(struct dentry *dentry)
352{
353 return 1;
354}
355static struct dentry_operations sockfs_dentry_operations = {
356 .d_delete = sockfs_delete_dentry,
357};
358
359/*
360 * Obtains the first available file descriptor and sets it up for use.
361 *
39d8c1b6
DM
362 * These functions create file structures and maps them to fd space
363 * of the current process. On success it returns file descriptor
1da177e4
LT
364 * and file struct implicitly stored in sock->file.
365 * Note that another thread may close file descriptor before we return
366 * from this function. We use the fact that now we do not refer
367 * to socket after mapping. If one day we will need it, this
368 * function will increment ref. count on file by 1.
369 *
370 * In any case returned fd MAY BE not valid!
371 * This race condition is unavoidable
372 * with shared fd spaces, we cannot solve it inside kernel,
373 * but we take care of internal coherence yet.
374 */
375
39d8c1b6 376static int sock_alloc_fd(struct file **filep)
1da177e4
LT
377{
378 int fd;
1da177e4
LT
379
380 fd = get_unused_fd();
39d8c1b6 381 if (likely(fd >= 0)) {
1da177e4
LT
382 struct file *file = get_empty_filp();
383
39d8c1b6
DM
384 *filep = file;
385 if (unlikely(!file)) {
1da177e4 386 put_unused_fd(fd);
39d8c1b6 387 return -ENFILE;
1da177e4 388 }
39d8c1b6
DM
389 } else
390 *filep = NULL;
391 return fd;
392}
1da177e4 393
39d8c1b6
DM
394static int sock_attach_fd(struct socket *sock, struct file *file)
395{
396 struct qstr this;
397 char name[32];
398
399 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
400 this.name = name;
401 this.hash = SOCK_INODE(sock)->i_ino;
402
403 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
404 if (unlikely(!file->f_dentry))
405 return -ENOMEM;
406
407 file->f_dentry->d_op = &sockfs_dentry_operations;
408 d_add(file->f_dentry, SOCK_INODE(sock));
409 file->f_vfsmnt = mntget(sock_mnt);
410 file->f_mapping = file->f_dentry->d_inode->i_mapping;
411
412 sock->file = file;
413 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
414 file->f_mode = FMODE_READ | FMODE_WRITE;
415 file->f_flags = O_RDWR;
416 file->f_pos = 0;
417 file->private_data = sock;
1da177e4 418
39d8c1b6
DM
419 return 0;
420}
421
422int sock_map_fd(struct socket *sock)
423{
424 struct file *newfile;
425 int fd = sock_alloc_fd(&newfile);
426
427 if (likely(fd >= 0)) {
428 int err = sock_attach_fd(sock, newfile);
429
430 if (unlikely(err < 0)) {
431 put_filp(newfile);
1da177e4 432 put_unused_fd(fd);
39d8c1b6 433 return err;
1da177e4 434 }
39d8c1b6 435 fd_install(fd, newfile);
1da177e4 436 }
1da177e4
LT
437 return fd;
438}
439
6cb153ca
BL
440static struct socket *sock_from_file(struct file *file, int *err)
441{
442 struct inode *inode;
443 struct socket *sock;
444
445 if (file->f_op == &socket_file_ops)
446 return file->private_data; /* set in sock_map_fd */
447
448 inode = file->f_dentry->d_inode;
449 if (!S_ISSOCK(inode->i_mode)) {
450 *err = -ENOTSOCK;
451 return NULL;
452 }
453
454 sock = SOCKET_I(inode);
455 if (sock->file != file) {
456 printk(KERN_ERR "socki_lookup: socket file changed!\n");
457 sock->file = file;
458 }
459 return sock;
460}
461
1da177e4
LT
462/**
463 * sockfd_lookup - Go from a file number to its socket slot
464 * @fd: file handle
465 * @err: pointer to an error code return
466 *
467 * The file handle passed in is locked and the socket it is bound
468 * too is returned. If an error occurs the err pointer is overwritten
469 * with a negative errno code and NULL is returned. The function checks
470 * for both invalid handles and passing a handle which is not a socket.
471 *
472 * On a success the socket object pointer is returned.
473 */
474
475struct socket *sockfd_lookup(int fd, int *err)
476{
477 struct file *file;
1da177e4
LT
478 struct socket *sock;
479
6cb153ca 480 if (!(file = fget(fd))) {
1da177e4
LT
481 *err = -EBADF;
482 return NULL;
483 }
6cb153ca
BL
484 sock = sock_from_file(file, err);
485 if (!sock)
1da177e4 486 fput(file);
6cb153ca
BL
487 return sock;
488}
1da177e4 489
6cb153ca
BL
490static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
491{
492 struct file *file;
493 struct socket *sock;
494
3672558c 495 *err = -EBADF;
6cb153ca
BL
496 file = fget_light(fd, fput_needed);
497 if (file) {
498 sock = sock_from_file(file, err);
499 if (sock)
500 return sock;
501 fput_light(file, *fput_needed);
1da177e4 502 }
6cb153ca 503 return NULL;
1da177e4
LT
504}
505
506/**
507 * sock_alloc - allocate a socket
508 *
509 * Allocate a new inode and socket object. The two are bound together
510 * and initialised. The socket is then returned. If we are out of inodes
511 * NULL is returned.
512 */
513
514static struct socket *sock_alloc(void)
515{
516 struct inode * inode;
517 struct socket * sock;
518
519 inode = new_inode(sock_mnt->mnt_sb);
520 if (!inode)
521 return NULL;
522
523 sock = SOCKET_I(inode);
524
525 inode->i_mode = S_IFSOCK|S_IRWXUGO;
526 inode->i_uid = current->fsuid;
527 inode->i_gid = current->fsgid;
528
529 get_cpu_var(sockets_in_use)++;
530 put_cpu_var(sockets_in_use);
531 return sock;
532}
533
534/*
535 * In theory you can't get an open on this inode, but /proc provides
536 * a back door. Remember to keep it shut otherwise you'll let the
537 * creepy crawlies in.
538 */
539
540static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
541{
542 return -ENXIO;
543}
544
4b6f5d20 545const struct file_operations bad_sock_fops = {
1da177e4
LT
546 .owner = THIS_MODULE,
547 .open = sock_no_open,
548};
549
550/**
551 * sock_release - close a socket
552 * @sock: socket to close
553 *
554 * The socket is released from the protocol stack if it has a release
555 * callback, and the inode is then released if the socket is bound to
556 * an inode not a file.
557 */
558
559void sock_release(struct socket *sock)
560{
561 if (sock->ops) {
562 struct module *owner = sock->ops->owner;
563
564 sock->ops->release(sock);
565 sock->ops = NULL;
566 module_put(owner);
567 }
568
569 if (sock->fasync_list)
570 printk(KERN_ERR "sock_release: fasync list not empty!\n");
571
572 get_cpu_var(sockets_in_use)--;
573 put_cpu_var(sockets_in_use);
574 if (!sock->file) {
575 iput(SOCK_INODE(sock));
576 return;
577 }
578 sock->file=NULL;
579}
580
581static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
582 struct msghdr *msg, size_t size)
583{
584 struct sock_iocb *si = kiocb_to_siocb(iocb);
585 int err;
586
587 si->sock = sock;
588 si->scm = NULL;
589 si->msg = msg;
590 si->size = size;
591
592 err = security_socket_sendmsg(sock, msg, size);
593 if (err)
594 return err;
595
596 return sock->ops->sendmsg(iocb, sock, msg, size);
597}
598
599int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
600{
601 struct kiocb iocb;
602 struct sock_iocb siocb;
603 int ret;
604
605 init_sync_kiocb(&iocb, NULL);
606 iocb.private = &siocb;
607 ret = __sock_sendmsg(&iocb, sock, msg, size);
608 if (-EIOCBQUEUED == ret)
609 ret = wait_on_sync_kiocb(&iocb);
610 return ret;
611}
612
613int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
614 struct kvec *vec, size_t num, size_t size)
615{
616 mm_segment_t oldfs = get_fs();
617 int result;
618
619 set_fs(KERNEL_DS);
620 /*
621 * the following is safe, since for compiler definitions of kvec and
622 * iovec are identical, yielding the same in-core layout and alignment
623 */
624 msg->msg_iov = (struct iovec *)vec,
625 msg->msg_iovlen = num;
626 result = sock_sendmsg(sock, msg, size);
627 set_fs(oldfs);
628 return result;
629}
630
631static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
632 struct msghdr *msg, size_t size, int flags)
633{
634 int err;
635 struct sock_iocb *si = kiocb_to_siocb(iocb);
636
637 si->sock = sock;
638 si->scm = NULL;
639 si->msg = msg;
640 si->size = size;
641 si->flags = flags;
642
643 err = security_socket_recvmsg(sock, msg, size, flags);
644 if (err)
645 return err;
646
647 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
648}
649
650int sock_recvmsg(struct socket *sock, struct msghdr *msg,
651 size_t size, int flags)
652{
653 struct kiocb iocb;
654 struct sock_iocb siocb;
655 int ret;
656
657 init_sync_kiocb(&iocb, NULL);
658 iocb.private = &siocb;
659 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
660 if (-EIOCBQUEUED == ret)
661 ret = wait_on_sync_kiocb(&iocb);
662 return ret;
663}
664
665int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
666 struct kvec *vec, size_t num,
667 size_t size, int flags)
668{
669 mm_segment_t oldfs = get_fs();
670 int result;
671
672 set_fs(KERNEL_DS);
673 /*
674 * the following is safe, since for compiler definitions of kvec and
675 * iovec are identical, yielding the same in-core layout and alignment
676 */
677 msg->msg_iov = (struct iovec *)vec,
678 msg->msg_iovlen = num;
679 result = sock_recvmsg(sock, msg, size, flags);
680 set_fs(oldfs);
681 return result;
682}
683
684static void sock_aio_dtor(struct kiocb *iocb)
685{
686 kfree(iocb->private);
687}
688
ce1d4d3e
CH
689static ssize_t sock_sendpage(struct file *file, struct page *page,
690 int offset, size_t size, loff_t *ppos, int more)
1da177e4 691{
1da177e4
LT
692 struct socket *sock;
693 int flags;
694
ce1d4d3e
CH
695 sock = file->private_data;
696
697 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
698 if (more)
699 flags |= MSG_MORE;
700
701 return sock->ops->sendpage(sock, page, offset, size, flags);
702}
1da177e4 703
ce1d4d3e
CH
704static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
705 char __user *ubuf, size_t size, struct sock_iocb *siocb)
706{
707 if (!is_sync_kiocb(iocb)) {
708 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
709 if (!siocb)
710 return NULL;
1da177e4
LT
711 iocb->ki_dtor = sock_aio_dtor;
712 }
1da177e4 713
ce1d4d3e
CH
714 siocb->kiocb = iocb;
715 siocb->async_iov.iov_base = ubuf;
716 siocb->async_iov.iov_len = size;
1da177e4 717
ce1d4d3e
CH
718 iocb->private = siocb;
719 return siocb;
1da177e4
LT
720}
721
ce1d4d3e
CH
722static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
723 struct file *file, struct iovec *iov, unsigned long nr_segs)
724{
725 struct socket *sock = file->private_data;
726 size_t size = 0;
727 int i;
1da177e4 728
ce1d4d3e
CH
729 for (i = 0 ; i < nr_segs ; i++)
730 size += iov[i].iov_len;
1da177e4 731
ce1d4d3e
CH
732 msg->msg_name = NULL;
733 msg->msg_namelen = 0;
734 msg->msg_control = NULL;
735 msg->msg_controllen = 0;
736 msg->msg_iov = (struct iovec *) iov;
737 msg->msg_iovlen = nr_segs;
738 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
739
740 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
741}
742
743static ssize_t sock_readv(struct file *file, const struct iovec *iov,
744 unsigned long nr_segs, loff_t *ppos)
1da177e4 745{
ce1d4d3e
CH
746 struct kiocb iocb;
747 struct sock_iocb siocb;
748 struct msghdr msg;
749 int ret;
750
751 init_sync_kiocb(&iocb, NULL);
752 iocb.private = &siocb;
753
754 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
755 if (-EIOCBQUEUED == ret)
756 ret = wait_on_sync_kiocb(&iocb);
757 return ret;
758}
759
760static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
761 size_t count, loff_t pos)
762{
763 struct sock_iocb siocb, *x;
764
1da177e4
LT
765 if (pos != 0)
766 return -ESPIPE;
ce1d4d3e 767 if (count == 0) /* Match SYS5 behaviour */
1da177e4
LT
768 return 0;
769
ce1d4d3e
CH
770 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
771 if (!x)
772 return -ENOMEM;
773 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
774 &x->async_iov, 1);
1da177e4
LT
775}
776
ce1d4d3e
CH
777static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
778 struct file *file, struct iovec *iov, unsigned long nr_segs)
1da177e4 779{
ce1d4d3e
CH
780 struct socket *sock = file->private_data;
781 size_t size = 0;
782 int i;
1da177e4 783
ce1d4d3e
CH
784 for (i = 0 ; i < nr_segs ; i++)
785 size += iov[i].iov_len;
1da177e4 786
ce1d4d3e
CH
787 msg->msg_name = NULL;
788 msg->msg_namelen = 0;
789 msg->msg_control = NULL;
790 msg->msg_controllen = 0;
791 msg->msg_iov = (struct iovec *) iov;
792 msg->msg_iovlen = nr_segs;
793 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
794 if (sock->type == SOCK_SEQPACKET)
795 msg->msg_flags |= MSG_EOR;
1da177e4 796
ce1d4d3e 797 return __sock_sendmsg(iocb, sock, msg, size);
1da177e4
LT
798}
799
ce1d4d3e
CH
800static ssize_t sock_writev(struct file *file, const struct iovec *iov,
801 unsigned long nr_segs, loff_t *ppos)
1da177e4
LT
802{
803 struct msghdr msg;
ce1d4d3e
CH
804 struct kiocb iocb;
805 struct sock_iocb siocb;
806 int ret;
1da177e4 807
ce1d4d3e
CH
808 init_sync_kiocb(&iocb, NULL);
809 iocb.private = &siocb;
1da177e4 810
ce1d4d3e
CH
811 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
812 if (-EIOCBQUEUED == ret)
813 ret = wait_on_sync_kiocb(&iocb);
814 return ret;
815}
1da177e4 816
ce1d4d3e
CH
817static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
818 size_t count, loff_t pos)
819{
820 struct sock_iocb siocb, *x;
1da177e4 821
ce1d4d3e
CH
822 if (pos != 0)
823 return -ESPIPE;
824 if (count == 0) /* Match SYS5 behaviour */
825 return 0;
1da177e4 826
ce1d4d3e
CH
827 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
828 if (!x)
829 return -ENOMEM;
1da177e4 830
ce1d4d3e
CH
831 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
832 &x->async_iov, 1);
1da177e4
LT
833}
834
835
836/*
837 * Atomic setting of ioctl hooks to avoid race
838 * with module unload.
839 */
840
4a3e2f71 841static DEFINE_MUTEX(br_ioctl_mutex);
1da177e4
LT
842static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
843
844void brioctl_set(int (*hook)(unsigned int, void __user *))
845{
4a3e2f71 846 mutex_lock(&br_ioctl_mutex);
1da177e4 847 br_ioctl_hook = hook;
4a3e2f71 848 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
849}
850EXPORT_SYMBOL(brioctl_set);
851
4a3e2f71 852static DEFINE_MUTEX(vlan_ioctl_mutex);
1da177e4
LT
853static int (*vlan_ioctl_hook)(void __user *arg);
854
855void vlan_ioctl_set(int (*hook)(void __user *))
856{
4a3e2f71 857 mutex_lock(&vlan_ioctl_mutex);
1da177e4 858 vlan_ioctl_hook = hook;
4a3e2f71 859 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
860}
861EXPORT_SYMBOL(vlan_ioctl_set);
862
4a3e2f71 863static DEFINE_MUTEX(dlci_ioctl_mutex);
1da177e4
LT
864static int (*dlci_ioctl_hook)(unsigned int, void __user *);
865
866void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
867{
4a3e2f71 868 mutex_lock(&dlci_ioctl_mutex);
1da177e4 869 dlci_ioctl_hook = hook;
4a3e2f71 870 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
871}
872EXPORT_SYMBOL(dlci_ioctl_set);
873
874/*
875 * With an ioctl, arg may well be a user mode pointer, but we don't know
876 * what to do with it - that's up to the protocol still.
877 */
878
879static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
880{
881 struct socket *sock;
882 void __user *argp = (void __user *)arg;
883 int pid, err;
884
b69aee04 885 sock = file->private_data;
1da177e4
LT
886 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
887 err = dev_ioctl(cmd, argp);
888 } else
d86b5e0e 889#ifdef CONFIG_WIRELESS_EXT
1da177e4
LT
890 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
891 err = dev_ioctl(cmd, argp);
892 } else
d86b5e0e 893#endif /* CONFIG_WIRELESS_EXT */
1da177e4
LT
894 switch (cmd) {
895 case FIOSETOWN:
896 case SIOCSPGRP:
897 err = -EFAULT;
898 if (get_user(pid, (int __user *)argp))
899 break;
900 err = f_setown(sock->file, pid, 1);
901 break;
902 case FIOGETOWN:
903 case SIOCGPGRP:
904 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
905 break;
906 case SIOCGIFBR:
907 case SIOCSIFBR:
908 case SIOCBRADDBR:
909 case SIOCBRDELBR:
910 err = -ENOPKG;
911 if (!br_ioctl_hook)
912 request_module("bridge");
913
4a3e2f71 914 mutex_lock(&br_ioctl_mutex);
1da177e4
LT
915 if (br_ioctl_hook)
916 err = br_ioctl_hook(cmd, argp);
4a3e2f71 917 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
918 break;
919 case SIOCGIFVLAN:
920 case SIOCSIFVLAN:
921 err = -ENOPKG;
922 if (!vlan_ioctl_hook)
923 request_module("8021q");
924
4a3e2f71 925 mutex_lock(&vlan_ioctl_mutex);
1da177e4
LT
926 if (vlan_ioctl_hook)
927 err = vlan_ioctl_hook(argp);
4a3e2f71 928 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
929 break;
930 case SIOCGIFDIVERT:
931 case SIOCSIFDIVERT:
932 /* Convert this to call through a hook */
933 err = divert_ioctl(cmd, argp);
934 break;
935 case SIOCADDDLCI:
936 case SIOCDELDLCI:
937 err = -ENOPKG;
938 if (!dlci_ioctl_hook)
939 request_module("dlci");
940
941 if (dlci_ioctl_hook) {
4a3e2f71 942 mutex_lock(&dlci_ioctl_mutex);
1da177e4 943 err = dlci_ioctl_hook(cmd, argp);
4a3e2f71 944 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
945 }
946 break;
947 default:
948 err = sock->ops->ioctl(sock, cmd, arg);
b5e5fa5e
CH
949
950 /*
951 * If this ioctl is unknown try to hand it down
952 * to the NIC driver.
953 */
954 if (err == -ENOIOCTLCMD)
955 err = dev_ioctl(cmd, argp);
1da177e4
LT
956 break;
957 }
958 return err;
959}
960
961int sock_create_lite(int family, int type, int protocol, struct socket **res)
962{
963 int err;
964 struct socket *sock = NULL;
965
966 err = security_socket_create(family, type, protocol, 1);
967 if (err)
968 goto out;
969
970 sock = sock_alloc();
971 if (!sock) {
972 err = -ENOMEM;
973 goto out;
974 }
975
1da177e4 976 sock->type = type;
7420ed23
VY
977 err = security_socket_post_create(sock, family, type, protocol, 1);
978 if (err)
979 goto out_release;
980
1da177e4
LT
981out:
982 *res = sock;
983 return err;
7420ed23
VY
984out_release:
985 sock_release(sock);
986 sock = NULL;
987 goto out;
1da177e4
LT
988}
989
990/* No kernel lock held - perfect */
991static unsigned int sock_poll(struct file *file, poll_table * wait)
992{
993 struct socket *sock;
994
995 /*
996 * We can't return errors to poll, so it's either yes or no.
997 */
b69aee04 998 sock = file->private_data;
1da177e4
LT
999 return sock->ops->poll(file, sock, wait);
1000}
1001
1002static int sock_mmap(struct file * file, struct vm_area_struct * vma)
1003{
b69aee04 1004 struct socket *sock = file->private_data;
1da177e4
LT
1005
1006 return sock->ops->mmap(file, sock, vma);
1007}
1008
20380731 1009static int sock_close(struct inode *inode, struct file *filp)
1da177e4
LT
1010{
1011 /*
1012 * It was possible the inode is NULL we were
1013 * closing an unfinished socket.
1014 */
1015
1016 if (!inode)
1017 {
1018 printk(KERN_DEBUG "sock_close: NULL inode\n");
1019 return 0;
1020 }
1021 sock_fasync(-1, filp, 0);
1022 sock_release(SOCKET_I(inode));
1023 return 0;
1024}
1025
1026/*
1027 * Update the socket async list
1028 *
1029 * Fasync_list locking strategy.
1030 *
1031 * 1. fasync_list is modified only under process context socket lock
1032 * i.e. under semaphore.
1033 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1034 * or under socket lock.
1035 * 3. fasync_list can be used from softirq context, so that
1036 * modification under socket lock have to be enhanced with
1037 * write_lock_bh(&sk->sk_callback_lock).
1038 * --ANK (990710)
1039 */
1040
1041static int sock_fasync(int fd, struct file *filp, int on)
1042{
1043 struct fasync_struct *fa, *fna=NULL, **prev;
1044 struct socket *sock;
1045 struct sock *sk;
1046
1047 if (on)
1048 {
8b3a7005 1049 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1da177e4
LT
1050 if(fna==NULL)
1051 return -ENOMEM;
1052 }
1053
b69aee04 1054 sock = filp->private_data;
1da177e4
LT
1055
1056 if ((sk=sock->sk) == NULL) {
1057 kfree(fna);
1058 return -EINVAL;
1059 }
1060
1061 lock_sock(sk);
1062
1063 prev=&(sock->fasync_list);
1064
1065 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1066 if (fa->fa_file==filp)
1067 break;
1068
1069 if(on)
1070 {
1071 if(fa!=NULL)
1072 {
1073 write_lock_bh(&sk->sk_callback_lock);
1074 fa->fa_fd=fd;
1075 write_unlock_bh(&sk->sk_callback_lock);
1076
1077 kfree(fna);
1078 goto out;
1079 }
1080 fna->fa_file=filp;
1081 fna->fa_fd=fd;
1082 fna->magic=FASYNC_MAGIC;
1083 fna->fa_next=sock->fasync_list;
1084 write_lock_bh(&sk->sk_callback_lock);
1085 sock->fasync_list=fna;
1086 write_unlock_bh(&sk->sk_callback_lock);
1087 }
1088 else
1089 {
1090 if (fa!=NULL)
1091 {
1092 write_lock_bh(&sk->sk_callback_lock);
1093 *prev=fa->fa_next;
1094 write_unlock_bh(&sk->sk_callback_lock);
1095 kfree(fa);
1096 }
1097 }
1098
1099out:
1100 release_sock(sock->sk);
1101 return 0;
1102}
1103
1104/* This function may be called only under socket lock or callback_lock */
1105
1106int sock_wake_async(struct socket *sock, int how, int band)
1107{
1108 if (!sock || !sock->fasync_list)
1109 return -1;
1110 switch (how)
1111 {
1112 case 1:
1113
1114 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1115 break;
1116 goto call_kill;
1117 case 2:
1118 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1119 break;
1120 /* fall through */
1121 case 0:
1122 call_kill:
1123 __kill_fasync(sock->fasync_list, SIGIO, band);
1124 break;
1125 case 3:
1126 __kill_fasync(sock->fasync_list, SIGURG, band);
1127 }
1128 return 0;
1129}
1130
1131static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1132{
1133 int err;
1134 struct socket *sock;
1135
1136 /*
1137 * Check protocol is in range
1138 */
1139 if (family < 0 || family >= NPROTO)
1140 return -EAFNOSUPPORT;
1141 if (type < 0 || type >= SOCK_MAX)
1142 return -EINVAL;
1143
1144 /* Compatibility.
1145
1146 This uglymoron is moved from INET layer to here to avoid
1147 deadlock in module load.
1148 */
1149 if (family == PF_INET && type == SOCK_PACKET) {
1150 static int warned;
1151 if (!warned) {
1152 warned = 1;
1153 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1154 }
1155 family = PF_PACKET;
1156 }
1157
1158 err = security_socket_create(family, type, protocol, kern);
1159 if (err)
1160 return err;
1161
1162#if defined(CONFIG_KMOD)
1163 /* Attempt to load a protocol module if the find failed.
1164 *
1165 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1166 * requested real, full-featured networking support upon configuration.
1167 * Otherwise module support will break!
1168 */
1169 if (net_families[family]==NULL)
1170 {
1171 request_module("net-pf-%d",family);
1172 }
1173#endif
1174
1175 net_family_read_lock();
1176 if (net_families[family] == NULL) {
1177 err = -EAFNOSUPPORT;
1178 goto out;
1179 }
1180
1181/*
1182 * Allocate the socket and allow the family to set things up. if
1183 * the protocol is 0, the family is instructed to select an appropriate
1184 * default.
1185 */
1186
1187 if (!(sock = sock_alloc())) {
5991c844
AM
1188 if (net_ratelimit())
1189 printk(KERN_WARNING "socket: no more sockets\n");
1da177e4
LT
1190 err = -ENFILE; /* Not exactly a match, but its the
1191 closest posix thing */
1192 goto out;
1193 }
1194
1195 sock->type = type;
1196
1197 /*
1198 * We will call the ->create function, that possibly is in a loadable
1199 * module, so we have to bump that loadable module refcnt first.
1200 */
1201 err = -EAFNOSUPPORT;
1202 if (!try_module_get(net_families[family]->owner))
1203 goto out_release;
1204
a79af59e
FF
1205 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1206 sock->ops = NULL;
1da177e4 1207 goto out_module_put;
a79af59e
FF
1208 }
1209
1da177e4
LT
1210 /*
1211 * Now to bump the refcnt of the [loadable] module that owns this
1212 * socket at sock_release time we decrement its refcnt.
1213 */
1214 if (!try_module_get(sock->ops->owner)) {
1215 sock->ops = NULL;
1216 goto out_module_put;
1217 }
1218 /*
1219 * Now that we're done with the ->create function, the [loadable]
1220 * module can have its refcnt decremented
1221 */
1222 module_put(net_families[family]->owner);
1223 *res = sock;
7420ed23
VY
1224 err = security_socket_post_create(sock, family, type, protocol, kern);
1225 if (err)
1226 goto out_release;
1da177e4
LT
1227
1228out:
1229 net_family_read_unlock();
1230 return err;
1231out_module_put:
1232 module_put(net_families[family]->owner);
1233out_release:
1234 sock_release(sock);
1235 goto out;
1236}
1237
1238int sock_create(int family, int type, int protocol, struct socket **res)
1239{
1240 return __sock_create(family, type, protocol, res, 0);
1241}
1242
1243int sock_create_kern(int family, int type, int protocol, struct socket **res)
1244{
1245 return __sock_create(family, type, protocol, res, 1);
1246}
1247
1248asmlinkage long sys_socket(int family, int type, int protocol)
1249{
1250 int retval;
1251 struct socket *sock;
1252
1253 retval = sock_create(family, type, protocol, &sock);
1254 if (retval < 0)
1255 goto out;
1256
1257 retval = sock_map_fd(sock);
1258 if (retval < 0)
1259 goto out_release;
1260
1261out:
1262 /* It may be already another descriptor 8) Not kernel problem. */
1263 return retval;
1264
1265out_release:
1266 sock_release(sock);
1267 return retval;
1268}
1269
1270/*
1271 * Create a pair of connected sockets.
1272 */
1273
1274asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1275{
1276 struct socket *sock1, *sock2;
1277 int fd1, fd2, err;
1278
1279 /*
1280 * Obtain the first socket and check if the underlying protocol
1281 * supports the socketpair call.
1282 */
1283
1284 err = sock_create(family, type, protocol, &sock1);
1285 if (err < 0)
1286 goto out;
1287
1288 err = sock_create(family, type, protocol, &sock2);
1289 if (err < 0)
1290 goto out_release_1;
1291
1292 err = sock1->ops->socketpair(sock1, sock2);
1293 if (err < 0)
1294 goto out_release_both;
1295
1296 fd1 = fd2 = -1;
1297
1298 err = sock_map_fd(sock1);
1299 if (err < 0)
1300 goto out_release_both;
1301 fd1 = err;
1302
1303 err = sock_map_fd(sock2);
1304 if (err < 0)
1305 goto out_close_1;
1306 fd2 = err;
1307
1308 /* fd1 and fd2 may be already another descriptors.
1309 * Not kernel problem.
1310 */
1311
1312 err = put_user(fd1, &usockvec[0]);
1313 if (!err)
1314 err = put_user(fd2, &usockvec[1]);
1315 if (!err)
1316 return 0;
1317
1318 sys_close(fd2);
1319 sys_close(fd1);
1320 return err;
1321
1322out_close_1:
1323 sock_release(sock2);
1324 sys_close(fd1);
1325 return err;
1326
1327out_release_both:
1328 sock_release(sock2);
1329out_release_1:
1330 sock_release(sock1);
1331out:
1332 return err;
1333}
1334
1335
1336/*
1337 * Bind a name to a socket. Nothing much to do here since it's
1338 * the protocol's responsibility to handle the local address.
1339 *
1340 * We move the socket address to kernel space before we call
1341 * the protocol layer (having also checked the address is ok).
1342 */
1343
1344asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1345{
1346 struct socket *sock;
1347 char address[MAX_SOCK_ADDR];
6cb153ca 1348 int err, fput_needed;
1da177e4 1349
6cb153ca 1350 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1da177e4
LT
1351 {
1352 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1353 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
6cb153ca
BL
1354 if (!err)
1355 err = sock->ops->bind(sock,
1356 (struct sockaddr *)address, addrlen);
1da177e4 1357 }
6cb153ca 1358 fput_light(sock->file, fput_needed);
1da177e4
LT
1359 }
1360 return err;
1361}
1362
1363
1364/*
1365 * Perform a listen. Basically, we allow the protocol to do anything
1366 * necessary for a listen, and if that works, we mark the socket as
1367 * ready for listening.
1368 */
1369
1370int sysctl_somaxconn = SOMAXCONN;
1371
1372asmlinkage long sys_listen(int fd, int backlog)
1373{
1374 struct socket *sock;
6cb153ca 1375 int err, fput_needed;
1da177e4 1376
6cb153ca 1377 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1da177e4
LT
1378 if ((unsigned) backlog > sysctl_somaxconn)
1379 backlog = sysctl_somaxconn;
1380
1381 err = security_socket_listen(sock, backlog);
6cb153ca
BL
1382 if (!err)
1383 err = sock->ops->listen(sock, backlog);
1da177e4 1384
6cb153ca 1385 fput_light(sock->file, fput_needed);
1da177e4
LT
1386 }
1387 return err;
1388}
1389
1390
1391/*
1392 * For accept, we attempt to create a new socket, set up the link
1393 * with the client, wake up the client, then return the new
1394 * connected fd. We collect the address of the connector in kernel
1395 * space and move it to user at the very end. This is unclean because
1396 * we open the socket then return an error.
1397 *
1398 * 1003.1g adds the ability to recvmsg() to query connection pending
1399 * status to recvmsg. We need to add that support in a way thats
1400 * clean when we restucture accept also.
1401 */
1402
1403asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1404{
1405 struct socket *sock, *newsock;
39d8c1b6 1406 struct file *newfile;
6cb153ca 1407 int err, len, newfd, fput_needed;
1da177e4
LT
1408 char address[MAX_SOCK_ADDR];
1409
6cb153ca 1410 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1411 if (!sock)
1412 goto out;
1413
1414 err = -ENFILE;
1415 if (!(newsock = sock_alloc()))
1416 goto out_put;
1417
1418 newsock->type = sock->type;
1419 newsock->ops = sock->ops;
1420
1da177e4
LT
1421 /*
1422 * We don't need try_module_get here, as the listening socket (sock)
1423 * has the protocol module (sock->ops->owner) held.
1424 */
1425 __module_get(newsock->ops->owner);
1426
39d8c1b6
DM
1427 newfd = sock_alloc_fd(&newfile);
1428 if (unlikely(newfd < 0)) {
1429 err = newfd;
9a1875e6
DM
1430 sock_release(newsock);
1431 goto out_put;
39d8c1b6
DM
1432 }
1433
1434 err = sock_attach_fd(newsock, newfile);
1435 if (err < 0)
1436 goto out_fd;
1437
a79af59e
FF
1438 err = security_socket_accept(sock, newsock);
1439 if (err)
39d8c1b6 1440 goto out_fd;
a79af59e 1441
1da177e4
LT
1442 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1443 if (err < 0)
39d8c1b6 1444 goto out_fd;
1da177e4
LT
1445
1446 if (upeer_sockaddr) {
1447 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1448 err = -ECONNABORTED;
39d8c1b6 1449 goto out_fd;
1da177e4
LT
1450 }
1451 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1452 if (err < 0)
39d8c1b6 1453 goto out_fd;
1da177e4
LT
1454 }
1455
1456 /* File flags are not inherited via accept() unlike another OSes. */
1457
39d8c1b6
DM
1458 fd_install(newfd, newfile);
1459 err = newfd;
1da177e4
LT
1460
1461 security_socket_post_accept(sock, newsock);
1462
1463out_put:
6cb153ca 1464 fput_light(sock->file, fput_needed);
1da177e4
LT
1465out:
1466 return err;
39d8c1b6 1467out_fd:
9606a216 1468 fput(newfile);
39d8c1b6 1469 put_unused_fd(newfd);
1da177e4
LT
1470 goto out_put;
1471}
1472
1473
1474/*
1475 * Attempt to connect to a socket with the server address. The address
1476 * is in user space so we verify it is OK and move it to kernel space.
1477 *
1478 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1479 * break bindings
1480 *
1481 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1482 * other SEQPACKET protocols that take time to connect() as it doesn't
1483 * include the -EINPROGRESS status for such sockets.
1484 */
1485
1486asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1487{
1488 struct socket *sock;
1489 char address[MAX_SOCK_ADDR];
6cb153ca 1490 int err, fput_needed;
1da177e4 1491
6cb153ca 1492 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1493 if (!sock)
1494 goto out;
1495 err = move_addr_to_kernel(uservaddr, addrlen, address);
1496 if (err < 0)
1497 goto out_put;
1498
1499 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1500 if (err)
1501 goto out_put;
1502
1503 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1504 sock->file->f_flags);
1505out_put:
6cb153ca 1506 fput_light(sock->file, fput_needed);
1da177e4
LT
1507out:
1508 return err;
1509}
1510
1511/*
1512 * Get the local address ('name') of a socket object. Move the obtained
1513 * name to user space.
1514 */
1515
1516asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1517{
1518 struct socket *sock;
1519 char address[MAX_SOCK_ADDR];
6cb153ca 1520 int len, err, fput_needed;
1da177e4 1521
6cb153ca 1522 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1523 if (!sock)
1524 goto out;
1525
1526 err = security_socket_getsockname(sock);
1527 if (err)
1528 goto out_put;
1529
1530 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1531 if (err)
1532 goto out_put;
1533 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1534
1535out_put:
6cb153ca 1536 fput_light(sock->file, fput_needed);
1da177e4
LT
1537out:
1538 return err;
1539}
1540
1541/*
1542 * Get the remote address ('name') of a socket object. Move the obtained
1543 * name to user space.
1544 */
1545
1546asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1547{
1548 struct socket *sock;
1549 char address[MAX_SOCK_ADDR];
6cb153ca 1550 int len, err, fput_needed;
1da177e4 1551
6cb153ca 1552 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1da177e4
LT
1553 err = security_socket_getpeername(sock);
1554 if (err) {
6cb153ca 1555 fput_light(sock->file, fput_needed);
1da177e4
LT
1556 return err;
1557 }
1558
1559 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1560 if (!err)
1561 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
6cb153ca 1562 fput_light(sock->file, fput_needed);
1da177e4
LT
1563 }
1564 return err;
1565}
1566
1567/*
1568 * Send a datagram to a given address. We move the address into kernel
1569 * space and check the user space data area is readable before invoking
1570 * the protocol.
1571 */
1572
1573asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1574 struct sockaddr __user *addr, int addr_len)
1575{
1576 struct socket *sock;
1577 char address[MAX_SOCK_ADDR];
1578 int err;
1579 struct msghdr msg;
1580 struct iovec iov;
6cb153ca
BL
1581 int fput_needed;
1582 struct file *sock_file;
1583
1584 sock_file = fget_light(fd, &fput_needed);
1585 if (!sock_file)
1586 return -EBADF;
1587
1588 sock = sock_from_file(sock_file, &err);
1da177e4 1589 if (!sock)
6cb153ca 1590 goto out_put;
1da177e4
LT
1591 iov.iov_base=buff;
1592 iov.iov_len=len;
1593 msg.msg_name=NULL;
1594 msg.msg_iov=&iov;
1595 msg.msg_iovlen=1;
1596 msg.msg_control=NULL;
1597 msg.msg_controllen=0;
1598 msg.msg_namelen=0;
6cb153ca 1599 if (addr) {
1da177e4
LT
1600 err = move_addr_to_kernel(addr, addr_len, address);
1601 if (err < 0)
1602 goto out_put;
1603 msg.msg_name=address;
1604 msg.msg_namelen=addr_len;
1605 }
1606 if (sock->file->f_flags & O_NONBLOCK)
1607 flags |= MSG_DONTWAIT;
1608 msg.msg_flags = flags;
1609 err = sock_sendmsg(sock, &msg, len);
1610
1611out_put:
6cb153ca 1612 fput_light(sock_file, fput_needed);
1da177e4
LT
1613 return err;
1614}
1615
1616/*
1617 * Send a datagram down a socket.
1618 */
1619
1620asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1621{
1622 return sys_sendto(fd, buff, len, flags, NULL, 0);
1623}
1624
1625/*
1626 * Receive a frame from the socket and optionally record the address of the
1627 * sender. We verify the buffers are writable and if needed move the
1628 * sender address from kernel to user space.
1629 */
1630
1631asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1632 struct sockaddr __user *addr, int __user *addr_len)
1633{
1634 struct socket *sock;
1635 struct iovec iov;
1636 struct msghdr msg;
1637 char address[MAX_SOCK_ADDR];
1638 int err,err2;
6cb153ca
BL
1639 struct file *sock_file;
1640 int fput_needed;
1641
1642 sock_file = fget_light(fd, &fput_needed);
1643 if (!sock_file)
1644 return -EBADF;
1da177e4 1645
6cb153ca 1646 sock = sock_from_file(sock_file, &err);
1da177e4
LT
1647 if (!sock)
1648 goto out;
1649
1650 msg.msg_control=NULL;
1651 msg.msg_controllen=0;
1652 msg.msg_iovlen=1;
1653 msg.msg_iov=&iov;
1654 iov.iov_len=size;
1655 iov.iov_base=ubuf;
1656 msg.msg_name=address;
1657 msg.msg_namelen=MAX_SOCK_ADDR;
1658 if (sock->file->f_flags & O_NONBLOCK)
1659 flags |= MSG_DONTWAIT;
1660 err=sock_recvmsg(sock, &msg, size, flags);
1661
1662 if(err >= 0 && addr != NULL)
1663 {
1664 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1665 if(err2<0)
1666 err=err2;
1667 }
1da177e4 1668out:
6cb153ca 1669 fput_light(sock_file, fput_needed);
1da177e4
LT
1670 return err;
1671}
1672
1673/*
1674 * Receive a datagram from a socket.
1675 */
1676
1677asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1678{
1679 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1680}
1681
1682/*
1683 * Set a socket option. Because we don't know the option lengths we have
1684 * to pass the user mode parameter for the protocols to sort out.
1685 */
1686
1687asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1688{
6cb153ca 1689 int err, fput_needed;
1da177e4
LT
1690 struct socket *sock;
1691
1692 if (optlen < 0)
1693 return -EINVAL;
1694
6cb153ca 1695 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
1da177e4
LT
1696 {
1697 err = security_socket_setsockopt(sock,level,optname);
6cb153ca
BL
1698 if (err)
1699 goto out_put;
1da177e4
LT
1700
1701 if (level == SOL_SOCKET)
1702 err=sock_setsockopt(sock,level,optname,optval,optlen);
1703 else
1704 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
6cb153ca
BL
1705out_put:
1706 fput_light(sock->file, fput_needed);
1da177e4
LT
1707 }
1708 return err;
1709}
1710
1711/*
1712 * Get a socket option. Because we don't know the option lengths we have
1713 * to pass a user mode parameter for the protocols to sort out.
1714 */
1715
1716asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1717{
6cb153ca 1718 int err, fput_needed;
1da177e4
LT
1719 struct socket *sock;
1720
6cb153ca
BL
1721 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1722 err = security_socket_getsockopt(sock, level, optname);
1723 if (err)
1724 goto out_put;
1da177e4
LT
1725
1726 if (level == SOL_SOCKET)
1727 err=sock_getsockopt(sock,level,optname,optval,optlen);
1728 else
1729 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
6cb153ca
BL
1730out_put:
1731 fput_light(sock->file, fput_needed);
1da177e4
LT
1732 }
1733 return err;
1734}
1735
1736
1737/*
1738 * Shutdown a socket.
1739 */
1740
1741asmlinkage long sys_shutdown(int fd, int how)
1742{
6cb153ca 1743 int err, fput_needed;
1da177e4
LT
1744 struct socket *sock;
1745
6cb153ca 1746 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1da177e4
LT
1747 {
1748 err = security_socket_shutdown(sock, how);
6cb153ca
BL
1749 if (!err)
1750 err = sock->ops->shutdown(sock, how);
1751 fput_light(sock->file, fput_needed);
1da177e4
LT
1752 }
1753 return err;
1754}
1755
1756/* A couple of helpful macros for getting the address of the 32/64 bit
1757 * fields which are the same type (int / unsigned) on our platforms.
1758 */
1759#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1760#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1761#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1762
1763
1764/*
1765 * BSD sendmsg interface
1766 */
1767
1768asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1769{
1770 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1771 struct socket *sock;
1772 char address[MAX_SOCK_ADDR];
1773 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
b9d717a7
AW
1774 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1775 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1776 /* 20 is size of ipv6_pktinfo */
1da177e4
LT
1777 unsigned char *ctl_buf = ctl;
1778 struct msghdr msg_sys;
1779 int err, ctl_len, iov_size, total_len;
6cb153ca 1780 int fput_needed;
1da177e4
LT
1781
1782 err = -EFAULT;
1783 if (MSG_CMSG_COMPAT & flags) {
1784 if (get_compat_msghdr(&msg_sys, msg_compat))
1785 return -EFAULT;
1786 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1787 return -EFAULT;
1788
6cb153ca 1789 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1790 if (!sock)
1791 goto out;
1792
1793 /* do not move before msg_sys is valid */
1794 err = -EMSGSIZE;
1795 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1796 goto out_put;
1797
1798 /* Check whether to allocate the iovec area*/
1799 err = -ENOMEM;
1800 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1801 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1802 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1803 if (!iov)
1804 goto out_put;
1805 }
1806
1807 /* This will also move the address data into kernel space */
1808 if (MSG_CMSG_COMPAT & flags) {
1809 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1810 } else
1811 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1812 if (err < 0)
1813 goto out_freeiov;
1814 total_len = err;
1815
1816 err = -ENOBUFS;
1817
1818 if (msg_sys.msg_controllen > INT_MAX)
1819 goto out_freeiov;
1820 ctl_len = msg_sys.msg_controllen;
1821 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
8920e8f9 1822 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
1da177e4
LT
1823 if (err)
1824 goto out_freeiov;
1825 ctl_buf = msg_sys.msg_control;
8920e8f9 1826 ctl_len = msg_sys.msg_controllen;
1da177e4
LT
1827 } else if (ctl_len) {
1828 if (ctl_len > sizeof(ctl))
1829 {
1830 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1831 if (ctl_buf == NULL)
1832 goto out_freeiov;
1833 }
1834 err = -EFAULT;
1835 /*
1836 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1837 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1838 * checking falls down on this.
1839 */
1840 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1841 goto out_freectl;
1842 msg_sys.msg_control = ctl_buf;
1843 }
1844 msg_sys.msg_flags = flags;
1845
1846 if (sock->file->f_flags & O_NONBLOCK)
1847 msg_sys.msg_flags |= MSG_DONTWAIT;
1848 err = sock_sendmsg(sock, &msg_sys, total_len);
1849
1850out_freectl:
1851 if (ctl_buf != ctl)
1852 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1853out_freeiov:
1854 if (iov != iovstack)
1855 sock_kfree_s(sock->sk, iov, iov_size);
1856out_put:
6cb153ca 1857 fput_light(sock->file, fput_needed);
1da177e4
LT
1858out:
1859 return err;
1860}
1861
1862/*
1863 * BSD recvmsg interface
1864 */
1865
1866asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1867{
1868 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1869 struct socket *sock;
1870 struct iovec iovstack[UIO_FASTIOV];
1871 struct iovec *iov=iovstack;
1872 struct msghdr msg_sys;
1873 unsigned long cmsg_ptr;
1874 int err, iov_size, total_len, len;
6cb153ca 1875 int fput_needed;
1da177e4
LT
1876
1877 /* kernel mode address */
1878 char addr[MAX_SOCK_ADDR];
1879
1880 /* user mode address pointers */
1881 struct sockaddr __user *uaddr;
1882 int __user *uaddr_len;
1883
1884 if (MSG_CMSG_COMPAT & flags) {
1885 if (get_compat_msghdr(&msg_sys, msg_compat))
1886 return -EFAULT;
1887 } else
1888 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1889 return -EFAULT;
1890
6cb153ca 1891 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1892 if (!sock)
1893 goto out;
1894
1895 err = -EMSGSIZE;
1896 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1897 goto out_put;
1898
1899 /* Check whether to allocate the iovec area*/
1900 err = -ENOMEM;
1901 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1902 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1903 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1904 if (!iov)
1905 goto out_put;
1906 }
1907
1908 /*
1909 * Save the user-mode address (verify_iovec will change the
1910 * kernel msghdr to use the kernel address space)
1911 */
1912
1913 uaddr = (void __user *) msg_sys.msg_name;
1914 uaddr_len = COMPAT_NAMELEN(msg);
1915 if (MSG_CMSG_COMPAT & flags) {
1916 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1917 } else
1918 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1919 if (err < 0)
1920 goto out_freeiov;
1921 total_len=err;
1922
1923 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1924 msg_sys.msg_flags = 0;
1925 if (MSG_CMSG_COMPAT & flags)
1926 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1927
1928 if (sock->file->f_flags & O_NONBLOCK)
1929 flags |= MSG_DONTWAIT;
1930 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1931 if (err < 0)
1932 goto out_freeiov;
1933 len = err;
1934
1935 if (uaddr != NULL) {
1936 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1937 if (err < 0)
1938 goto out_freeiov;
1939 }
37f7f421
DM
1940 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1941 COMPAT_FLAGS(msg));
1da177e4
LT
1942 if (err)
1943 goto out_freeiov;
1944 if (MSG_CMSG_COMPAT & flags)
1945 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1946 &msg_compat->msg_controllen);
1947 else
1948 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1949 &msg->msg_controllen);
1950 if (err)
1951 goto out_freeiov;
1952 err = len;
1953
1954out_freeiov:
1955 if (iov != iovstack)
1956 sock_kfree_s(sock->sk, iov, iov_size);
1957out_put:
6cb153ca 1958 fput_light(sock->file, fput_needed);
1da177e4
LT
1959out:
1960 return err;
1961}
1962
1963#ifdef __ARCH_WANT_SYS_SOCKETCALL
1964
1965/* Argument list sizes for sys_socketcall */
1966#define AL(x) ((x) * sizeof(unsigned long))
1967static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1968 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1969 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1970#undef AL
1971
1972/*
1973 * System call vectors.
1974 *
1975 * Argument checking cleaned up. Saved 20% in size.
1976 * This function doesn't need to set the kernel lock because
1977 * it is set by the callees.
1978 */
1979
1980asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1981{
1982 unsigned long a[6];
1983 unsigned long a0,a1;
1984 int err;
1985
1986 if(call<1||call>SYS_RECVMSG)
1987 return -EINVAL;
1988
1989 /* copy_from_user should be SMP safe. */
1990 if (copy_from_user(a, args, nargs[call]))
1991 return -EFAULT;
3ec3b2fb 1992
4bcff1b3 1993 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
3ec3b2fb
DW
1994 if (err)
1995 return err;
1996
1da177e4
LT
1997 a0=a[0];
1998 a1=a[1];
1999
2000 switch(call)
2001 {
2002 case SYS_SOCKET:
2003 err = sys_socket(a0,a1,a[2]);
2004 break;
2005 case SYS_BIND:
2006 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
2007 break;
2008 case SYS_CONNECT:
2009 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2010 break;
2011 case SYS_LISTEN:
2012 err = sys_listen(a0,a1);
2013 break;
2014 case SYS_ACCEPT:
2015 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2016 break;
2017 case SYS_GETSOCKNAME:
2018 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2019 break;
2020 case SYS_GETPEERNAME:
2021 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2022 break;
2023 case SYS_SOCKETPAIR:
2024 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2025 break;
2026 case SYS_SEND:
2027 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2028 break;
2029 case SYS_SENDTO:
2030 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2031 (struct sockaddr __user *)a[4], a[5]);
2032 break;
2033 case SYS_RECV:
2034 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2035 break;
2036 case SYS_RECVFROM:
2037 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2038 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2039 break;
2040 case SYS_SHUTDOWN:
2041 err = sys_shutdown(a0,a1);
2042 break;
2043 case SYS_SETSOCKOPT:
2044 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2045 break;
2046 case SYS_GETSOCKOPT:
2047 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2048 break;
2049 case SYS_SENDMSG:
2050 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2051 break;
2052 case SYS_RECVMSG:
2053 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2054 break;
2055 default:
2056 err = -EINVAL;
2057 break;
2058 }
2059 return err;
2060}
2061
2062#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2063
2064/*
2065 * This function is called by a protocol handler that wants to
2066 * advertise its address family, and have it linked into the
2067 * SOCKET module.
2068 */
2069
2070int sock_register(struct net_proto_family *ops)
2071{
2072 int err;
2073
2074 if (ops->family >= NPROTO) {
2075 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2076 return -ENOBUFS;
2077 }
2078 net_family_write_lock();
2079 err = -EEXIST;
2080 if (net_families[ops->family] == NULL) {
2081 net_families[ops->family]=ops;
2082 err = 0;
2083 }
2084 net_family_write_unlock();
2085 printk(KERN_INFO "NET: Registered protocol family %d\n",
2086 ops->family);
2087 return err;
2088}
2089
2090/*
2091 * This function is called by a protocol handler that wants to
2092 * remove its address family, and have it unlinked from the
2093 * SOCKET module.
2094 */
2095
2096int sock_unregister(int family)
2097{
2098 if (family < 0 || family >= NPROTO)
2099 return -1;
2100
2101 net_family_write_lock();
2102 net_families[family]=NULL;
2103 net_family_write_unlock();
2104 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2105 family);
2106 return 0;
2107}
2108
77d76ea3 2109static int __init sock_init(void)
1da177e4
LT
2110{
2111 /*
2112 * Initialize sock SLAB cache.
2113 */
2114
2115 sk_init();
2116
1da177e4
LT
2117 /*
2118 * Initialize skbuff SLAB cache
2119 */
2120 skb_init();
1da177e4
LT
2121
2122 /*
2123 * Initialize the protocols module.
2124 */
2125
2126 init_inodecache();
2127 register_filesystem(&sock_fs_type);
2128 sock_mnt = kern_mount(&sock_fs_type);
77d76ea3
AK
2129
2130 /* The real protocol initialization is performed in later initcalls.
1da177e4
LT
2131 */
2132
2133#ifdef CONFIG_NETFILTER
2134 netfilter_init();
2135#endif
cbeb321a
DM
2136
2137 return 0;
1da177e4
LT
2138}
2139
77d76ea3
AK
2140core_initcall(sock_init); /* early initcall */
2141
1da177e4
LT
2142#ifdef CONFIG_PROC_FS
2143void socket_seq_show(struct seq_file *seq)
2144{
2145 int cpu;
2146 int counter = 0;
2147
6f912042 2148 for_each_possible_cpu(cpu)
1da177e4
LT
2149 counter += per_cpu(sockets_in_use, cpu);
2150
2151 /* It can be negative, by the way. 8) */
2152 if (counter < 0)
2153 counter = 0;
2154
2155 seq_printf(seq, "sockets: used %d\n", counter);
2156}
2157#endif /* CONFIG_PROC_FS */
2158
89bbfc95
SP
2159#ifdef CONFIG_COMPAT
2160static long compat_sock_ioctl(struct file *file, unsigned cmd,
2161 unsigned long arg)
2162{
2163 struct socket *sock = file->private_data;
2164 int ret = -ENOIOCTLCMD;
2165
2166 if (sock->ops->compat_ioctl)
2167 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2168
2169 return ret;
2170}
2171#endif
2172
1da177e4
LT
2173/* ABI emulation layers need these two */
2174EXPORT_SYMBOL(move_addr_to_kernel);
2175EXPORT_SYMBOL(move_addr_to_user);
2176EXPORT_SYMBOL(sock_create);
2177EXPORT_SYMBOL(sock_create_kern);
2178EXPORT_SYMBOL(sock_create_lite);
2179EXPORT_SYMBOL(sock_map_fd);
2180EXPORT_SYMBOL(sock_recvmsg);
2181EXPORT_SYMBOL(sock_register);
2182EXPORT_SYMBOL(sock_release);
2183EXPORT_SYMBOL(sock_sendmsg);
2184EXPORT_SYMBOL(sock_unregister);
2185EXPORT_SYMBOL(sock_wake_async);
2186EXPORT_SYMBOL(sockfd_lookup);
2187EXPORT_SYMBOL(kernel_sendmsg);
2188EXPORT_SYMBOL(kernel_recvmsg);