blob: bbc80780f45afa84b073aff901640337b4dcfe95 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
Jesper Juhl02c30a82005-05-05 16:16:16 -07007 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61#include <linux/config.h>
62#include <linux/mm.h>
63#include <linux/smp_lock.h>
64#include <linux/socket.h>
65#include <linux/file.h>
66#include <linux/net.h>
67#include <linux/interrupt.h>
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080071#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030074#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#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>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010087#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010088#include <linux/wireless.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
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);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800110#ifdef CONFIG_COMPAT
111static long compat_sock_ioctl(struct file *file,
112 unsigned int cmd, unsigned long arg);
113#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114static 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
Jens Axboe5274f052006-03-30 15:15:30 +0200122extern ssize_t generic_splice_sendpage(struct inode *inode, struct file *out,
123 size_t len, unsigned int flags);
124
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125
126/*
127 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
128 * in the operation structures but are done directly via the socketcall() multiplexor.
129 */
130
131static struct file_operations socket_file_ops = {
132 .owner = THIS_MODULE,
133 .llseek = no_llseek,
134 .aio_read = sock_aio_read,
135 .aio_write = sock_aio_write,
136 .poll = sock_poll,
137 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800138#ifdef CONFIG_COMPAT
139 .compat_ioctl = compat_sock_ioctl,
140#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 .mmap = sock_mmap,
142 .open = sock_no_open, /* special open code to disallow open via /proc */
143 .release = sock_close,
144 .fasync = sock_fasync,
145 .readv = sock_readv,
146 .writev = sock_writev,
Jens Axboe5274f052006-03-30 15:15:30 +0200147 .sendpage = sock_sendpage,
148 .splice_write = generic_splice_sendpage,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149};
150
151/*
152 * The protocol list. Each protocol is registered in here.
153 */
154
155static struct net_proto_family *net_families[NPROTO];
156
157#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
158static atomic_t net_family_lockct = ATOMIC_INIT(0);
159static DEFINE_SPINLOCK(net_family_lock);
160
161/* The strategy is: modifications net_family vector are short, do not
162 sleep and veeery rare, but read access should be free of any exclusive
163 locks.
164 */
165
166static void net_family_write_lock(void)
167{
168 spin_lock(&net_family_lock);
169 while (atomic_read(&net_family_lockct) != 0) {
170 spin_unlock(&net_family_lock);
171
172 yield();
173
174 spin_lock(&net_family_lock);
175 }
176}
177
178static __inline__ void net_family_write_unlock(void)
179{
180 spin_unlock(&net_family_lock);
181}
182
183static __inline__ void net_family_read_lock(void)
184{
185 atomic_inc(&net_family_lockct);
186 spin_unlock_wait(&net_family_lock);
187}
188
189static __inline__ void net_family_read_unlock(void)
190{
191 atomic_dec(&net_family_lockct);
192}
193
194#else
195#define net_family_write_lock() do { } while(0)
196#define net_family_write_unlock() do { } while(0)
197#define net_family_read_lock() do { } while(0)
198#define net_family_read_unlock() do { } while(0)
199#endif
200
201
202/*
203 * Statistics counters of the socket lists
204 */
205
206static DEFINE_PER_CPU(int, sockets_in_use) = 0;
207
208/*
209 * Support routines. Move socket addresses back and forth across the kernel/user
210 * divide and look after the messy bits.
211 */
212
213#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
214 16 for IP, 16 for IPX,
215 24 for IPv6,
216 about 80 for AX.25
217 must be at least one bigger than
218 the AF_UNIX size (see net/unix/af_unix.c
219 :unix_mkname()).
220 */
221
222/**
223 * move_addr_to_kernel - copy a socket address into kernel space
224 * @uaddr: Address in user space
225 * @kaddr: Address in kernel space
226 * @ulen: Length in user space
227 *
228 * The address is copied into kernel space. If the provided address is
229 * too long an error code of -EINVAL is returned. If the copy gives
230 * invalid addresses -EFAULT is returned. On a success 0 is returned.
231 */
232
233int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
234{
235 if(ulen<0||ulen>MAX_SOCK_ADDR)
236 return -EINVAL;
237 if(ulen==0)
238 return 0;
239 if(copy_from_user(kaddr,uaddr,ulen))
240 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100241 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242}
243
244/**
245 * move_addr_to_user - copy an address to user space
246 * @kaddr: kernel space address
247 * @klen: length of address in kernel
248 * @uaddr: user space address
249 * @ulen: pointer to user length field
250 *
251 * The value pointed to by ulen on entry is the buffer length available.
252 * This is overwritten with the buffer space used. -EINVAL is returned
253 * if an overlong buffer is specified or a negative buffer size. -EFAULT
254 * is returned if either the buffer or the length field are not
255 * accessible.
256 * After copying the data up to the limit the user specifies, the true
257 * length of the data is written over the length limit the user
258 * specified. Zero is returned for a success.
259 */
260
261int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
262{
263 int err;
264 int len;
265
266 if((err=get_user(len, ulen)))
267 return err;
268 if(len>klen)
269 len=klen;
270 if(len<0 || len> MAX_SOCK_ADDR)
271 return -EINVAL;
272 if(len)
273 {
274 if(copy_to_user(uaddr,kaddr,len))
275 return -EFAULT;
276 }
277 /*
278 * "fromlen shall refer to the value before truncation.."
279 * 1003.1g
280 */
281 return __put_user(klen, ulen);
282}
283
284#define SOCKFS_MAGIC 0x534F434B
285
Eric Dumazetba899662005-08-26 12:05:31 -0700286static kmem_cache_t * sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
288static struct inode *sock_alloc_inode(struct super_block *sb)
289{
290 struct socket_alloc *ei;
291 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
292 if (!ei)
293 return NULL;
294 init_waitqueue_head(&ei->socket.wait);
295
296 ei->socket.fasync_list = NULL;
297 ei->socket.state = SS_UNCONNECTED;
298 ei->socket.flags = 0;
299 ei->socket.ops = NULL;
300 ei->socket.sk = NULL;
301 ei->socket.file = NULL;
302 ei->socket.flags = 0;
303
304 return &ei->vfs_inode;
305}
306
307static void sock_destroy_inode(struct inode *inode)
308{
309 kmem_cache_free(sock_inode_cachep,
310 container_of(inode, struct socket_alloc, vfs_inode));
311}
312
313static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
314{
315 struct socket_alloc *ei = (struct socket_alloc *) foo;
316
317 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
318 SLAB_CTOR_CONSTRUCTOR)
319 inode_init_once(&ei->vfs_inode);
320}
321
322static int init_inodecache(void)
323{
324 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
325 sizeof(struct socket_alloc),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800326 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
327 SLAB_MEM_SPREAD),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328 init_once, NULL);
329 if (sock_inode_cachep == NULL)
330 return -ENOMEM;
331 return 0;
332}
333
334static struct super_operations sockfs_ops = {
335 .alloc_inode = sock_alloc_inode,
336 .destroy_inode =sock_destroy_inode,
337 .statfs = simple_statfs,
338};
339
340static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
341 int flags, const char *dev_name, void *data)
342{
343 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
344}
345
Eric Dumazetba899662005-08-26 12:05:31 -0700346static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
348static struct file_system_type sock_fs_type = {
349 .name = "sockfs",
350 .get_sb = sockfs_get_sb,
351 .kill_sb = kill_anon_super,
352};
353static int sockfs_delete_dentry(struct dentry *dentry)
354{
355 return 1;
356}
357static struct dentry_operations sockfs_dentry_operations = {
358 .d_delete = sockfs_delete_dentry,
359};
360
361/*
362 * Obtains the first available file descriptor and sets it up for use.
363 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800364 * These functions create file structures and maps them to fd space
365 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 * and file struct implicitly stored in sock->file.
367 * Note that another thread may close file descriptor before we return
368 * from this function. We use the fact that now we do not refer
369 * to socket after mapping. If one day we will need it, this
370 * function will increment ref. count on file by 1.
371 *
372 * In any case returned fd MAY BE not valid!
373 * This race condition is unavoidable
374 * with shared fd spaces, we cannot solve it inside kernel,
375 * but we take care of internal coherence yet.
376 */
377
David S. Miller39d8c1b2006-03-20 17:13:49 -0800378static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379{
380 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800381
382 fd = get_unused_fd();
383 if (likely(fd >= 0)) {
384 struct file *file = get_empty_filp();
385
386 *filep = file;
387 if (unlikely(!file)) {
388 put_unused_fd(fd);
389 return -ENFILE;
390 }
391 } else
392 *filep = NULL;
393 return fd;
394}
395
396static int sock_attach_fd(struct socket *sock, struct file *file)
397{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 struct qstr this;
399 char name[32];
400
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
402 this.name = name;
403 this.hash = SOCK_INODE(sock)->i_ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
David S. Miller39d8c1b2006-03-20 17:13:49 -0800405 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
406 if (unlikely(!file->f_dentry))
407 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
David S. Miller39d8c1b2006-03-20 17:13:49 -0800409 file->f_dentry->d_op = &sockfs_dentry_operations;
410 d_add(file->f_dentry, SOCK_INODE(sock));
411 file->f_vfsmnt = mntget(sock_mnt);
412 file->f_mapping = file->f_dentry->d_inode->i_mapping;
413
414 sock->file = file;
415 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
416 file->f_mode = FMODE_READ | FMODE_WRITE;
417 file->f_flags = O_RDWR;
418 file->f_pos = 0;
419 file->private_data = sock;
420
421 return 0;
422}
423
424int sock_map_fd(struct socket *sock)
425{
426 struct file *newfile;
427 int fd = sock_alloc_fd(&newfile);
428
429 if (likely(fd >= 0)) {
430 int err = sock_attach_fd(sock, newfile);
431
432 if (unlikely(err < 0)) {
433 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800435 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800437 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 return fd;
440}
441
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800442static struct socket *sock_from_file(struct file *file, int *err)
443{
444 struct inode *inode;
445 struct socket *sock;
446
447 if (file->f_op == &socket_file_ops)
448 return file->private_data; /* set in sock_map_fd */
449
450 inode = file->f_dentry->d_inode;
451 if (!S_ISSOCK(inode->i_mode)) {
452 *err = -ENOTSOCK;
453 return NULL;
454 }
455
456 sock = SOCKET_I(inode);
457 if (sock->file != file) {
458 printk(KERN_ERR "socki_lookup: socket file changed!\n");
459 sock->file = file;
460 }
461 return sock;
462}
463
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464/**
465 * sockfd_lookup - Go from a file number to its socket slot
466 * @fd: file handle
467 * @err: pointer to an error code return
468 *
469 * The file handle passed in is locked and the socket it is bound
470 * too is returned. If an error occurs the err pointer is overwritten
471 * with a negative errno code and NULL is returned. The function checks
472 * for both invalid handles and passing a handle which is not a socket.
473 *
474 * On a success the socket object pointer is returned.
475 */
476
477struct socket *sockfd_lookup(int fd, int *err)
478{
479 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 struct socket *sock;
481
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800482 if (!(file = fget(fd))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 *err = -EBADF;
484 return NULL;
485 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800486 sock = sock_from_file(file, err);
487 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 return sock;
490}
491
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800492static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
493{
494 struct file *file;
495 struct socket *sock;
496
497 file = fget_light(fd, fput_needed);
498 if (file) {
499 sock = sock_from_file(file, err);
500 if (sock)
501 return sock;
502 fput_light(file, *fput_needed);
503 }
504 return NULL;
505}
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507/**
508 * sock_alloc - allocate a socket
509 *
510 * Allocate a new inode and socket object. The two are bound together
511 * and initialised. The socket is then returned. If we are out of inodes
512 * NULL is returned.
513 */
514
515static struct socket *sock_alloc(void)
516{
517 struct inode * inode;
518 struct socket * sock;
519
520 inode = new_inode(sock_mnt->mnt_sb);
521 if (!inode)
522 return NULL;
523
524 sock = SOCKET_I(inode);
525
526 inode->i_mode = S_IFSOCK|S_IRWXUGO;
527 inode->i_uid = current->fsuid;
528 inode->i_gid = current->fsgid;
529
530 get_cpu_var(sockets_in_use)++;
531 put_cpu_var(sockets_in_use);
532 return sock;
533}
534
535/*
536 * In theory you can't get an open on this inode, but /proc provides
537 * a back door. Remember to keep it shut otherwise you'll let the
538 * creepy crawlies in.
539 */
540
541static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
542{
543 return -ENXIO;
544}
545
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800546const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 .owner = THIS_MODULE,
548 .open = sock_no_open,
549};
550
551/**
552 * sock_release - close a socket
553 * @sock: socket to close
554 *
555 * The socket is released from the protocol stack if it has a release
556 * callback, and the inode is then released if the socket is bound to
557 * an inode not a file.
558 */
559
560void sock_release(struct socket *sock)
561{
562 if (sock->ops) {
563 struct module *owner = sock->ops->owner;
564
565 sock->ops->release(sock);
566 sock->ops = NULL;
567 module_put(owner);
568 }
569
570 if (sock->fasync_list)
571 printk(KERN_ERR "sock_release: fasync list not empty!\n");
572
573 get_cpu_var(sockets_in_use)--;
574 put_cpu_var(sockets_in_use);
575 if (!sock->file) {
576 iput(SOCK_INODE(sock));
577 return;
578 }
579 sock->file=NULL;
580}
581
582static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
583 struct msghdr *msg, size_t size)
584{
585 struct sock_iocb *si = kiocb_to_siocb(iocb);
586 int err;
587
588 si->sock = sock;
589 si->scm = NULL;
590 si->msg = msg;
591 si->size = size;
592
593 err = security_socket_sendmsg(sock, msg, size);
594 if (err)
595 return err;
596
597 return sock->ops->sendmsg(iocb, sock, msg, size);
598}
599
600int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
601{
602 struct kiocb iocb;
603 struct sock_iocb siocb;
604 int ret;
605
606 init_sync_kiocb(&iocb, NULL);
607 iocb.private = &siocb;
608 ret = __sock_sendmsg(&iocb, sock, msg, size);
609 if (-EIOCBQUEUED == ret)
610 ret = wait_on_sync_kiocb(&iocb);
611 return ret;
612}
613
614int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
615 struct kvec *vec, size_t num, size_t size)
616{
617 mm_segment_t oldfs = get_fs();
618 int result;
619
620 set_fs(KERNEL_DS);
621 /*
622 * the following is safe, since for compiler definitions of kvec and
623 * iovec are identical, yielding the same in-core layout and alignment
624 */
625 msg->msg_iov = (struct iovec *)vec,
626 msg->msg_iovlen = num;
627 result = sock_sendmsg(sock, msg, size);
628 set_fs(oldfs);
629 return result;
630}
631
632static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
633 struct msghdr *msg, size_t size, int flags)
634{
635 int err;
636 struct sock_iocb *si = kiocb_to_siocb(iocb);
637
638 si->sock = sock;
639 si->scm = NULL;
640 si->msg = msg;
641 si->size = size;
642 si->flags = flags;
643
644 err = security_socket_recvmsg(sock, msg, size, flags);
645 if (err)
646 return err;
647
648 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
649}
650
651int sock_recvmsg(struct socket *sock, struct msghdr *msg,
652 size_t size, int flags)
653{
654 struct kiocb iocb;
655 struct sock_iocb siocb;
656 int ret;
657
658 init_sync_kiocb(&iocb, NULL);
659 iocb.private = &siocb;
660 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
661 if (-EIOCBQUEUED == ret)
662 ret = wait_on_sync_kiocb(&iocb);
663 return ret;
664}
665
666int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
667 struct kvec *vec, size_t num,
668 size_t size, int flags)
669{
670 mm_segment_t oldfs = get_fs();
671 int result;
672
673 set_fs(KERNEL_DS);
674 /*
675 * the following is safe, since for compiler definitions of kvec and
676 * iovec are identical, yielding the same in-core layout and alignment
677 */
678 msg->msg_iov = (struct iovec *)vec,
679 msg->msg_iovlen = num;
680 result = sock_recvmsg(sock, msg, size, flags);
681 set_fs(oldfs);
682 return result;
683}
684
685static void sock_aio_dtor(struct kiocb *iocb)
686{
687 kfree(iocb->private);
688}
689
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300690static ssize_t sock_sendpage(struct file *file, struct page *page,
691 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692{
693 struct socket *sock;
694 int flags;
695
Eric Dumazetb69aee02005-09-06 14:42:45 -0700696 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
698 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
699 if (more)
700 flags |= MSG_MORE;
701
702 return sock->ops->sendpage(sock, page, offset, size, flags);
703}
704
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800705static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
706 char __user *ubuf, size_t size, struct sock_iocb *siocb)
707{
708 if (!is_sync_kiocb(iocb)) {
709 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
710 if (!siocb)
711 return NULL;
712 iocb->ki_dtor = sock_aio_dtor;
713 }
714
715 siocb->kiocb = iocb;
716 siocb->async_iov.iov_base = ubuf;
717 siocb->async_iov.iov_len = size;
718
719 iocb->private = siocb;
720 return siocb;
721}
722
723static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
724 struct file *file, struct iovec *iov, unsigned long nr_segs)
725{
726 struct socket *sock = file->private_data;
727 size_t size = 0;
728 int i;
729
730 for (i = 0 ; i < nr_segs ; i++)
731 size += iov[i].iov_len;
732
733 msg->msg_name = NULL;
734 msg->msg_namelen = 0;
735 msg->msg_control = NULL;
736 msg->msg_controllen = 0;
737 msg->msg_iov = (struct iovec *) iov;
738 msg->msg_iovlen = nr_segs;
739 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
740
741 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
742}
743
744static ssize_t sock_readv(struct file *file, const struct iovec *iov,
745 unsigned long nr_segs, loff_t *ppos)
746{
747 struct kiocb iocb;
748 struct sock_iocb siocb;
749 struct msghdr msg;
750 int ret;
751
752 init_sync_kiocb(&iocb, NULL);
753 iocb.private = &siocb;
754
755 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
756 if (-EIOCBQUEUED == ret)
757 ret = wait_on_sync_kiocb(&iocb);
758 return ret;
759}
760
761static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
762 size_t count, loff_t pos)
763{
764 struct sock_iocb siocb, *x;
765
766 if (pos != 0)
767 return -ESPIPE;
768 if (count == 0) /* Match SYS5 behaviour */
769 return 0;
770
771 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
772 if (!x)
773 return -ENOMEM;
774 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
775 &x->async_iov, 1);
776}
777
778static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
779 struct file *file, struct iovec *iov, unsigned long nr_segs)
780{
781 struct socket *sock = file->private_data;
782 size_t size = 0;
783 int i;
784
785 for (i = 0 ; i < nr_segs ; i++)
786 size += iov[i].iov_len;
787
788 msg->msg_name = NULL;
789 msg->msg_namelen = 0;
790 msg->msg_control = NULL;
791 msg->msg_controllen = 0;
792 msg->msg_iov = (struct iovec *) iov;
793 msg->msg_iovlen = nr_segs;
794 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
795 if (sock->type == SOCK_SEQPACKET)
796 msg->msg_flags |= MSG_EOR;
797
798 return __sock_sendmsg(iocb, sock, msg, size);
799}
800
801static ssize_t sock_writev(struct file *file, const struct iovec *iov,
802 unsigned long nr_segs, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803{
804 struct msghdr msg;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800805 struct kiocb iocb;
806 struct sock_iocb siocb;
807 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800809 init_sync_kiocb(&iocb, NULL);
810 iocb.private = &siocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800812 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
813 if (-EIOCBQUEUED == ret)
814 ret = wait_on_sync_kiocb(&iocb);
815 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816}
817
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800818static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
819 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800821 struct sock_iocb siocb, *x;
822
823 if (pos != 0)
824 return -ESPIPE;
825 if (count == 0) /* Match SYS5 behaviour */
826 return 0;
827
828 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
829 if (!x)
830 return -ENOMEM;
831
832 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
833 &x->async_iov, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834}
835
836
837/*
838 * Atomic setting of ioctl hooks to avoid race
839 * with module unload.
840 */
841
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800842static DEFINE_MUTEX(br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
844
845void brioctl_set(int (*hook)(unsigned int, void __user *))
846{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800847 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800849 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850}
851EXPORT_SYMBOL(brioctl_set);
852
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800853static DEFINE_MUTEX(vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854static int (*vlan_ioctl_hook)(void __user *arg);
855
856void vlan_ioctl_set(int (*hook)(void __user *))
857{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800858 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800860 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861}
862EXPORT_SYMBOL(vlan_ioctl_set);
863
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800864static DEFINE_MUTEX(dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865static int (*dlci_ioctl_hook)(unsigned int, void __user *);
866
867void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
868{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800869 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800871 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872}
873EXPORT_SYMBOL(dlci_ioctl_set);
874
875/*
876 * With an ioctl, arg may well be a user mode pointer, but we don't know
877 * what to do with it - that's up to the protocol still.
878 */
879
880static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
881{
882 struct socket *sock;
883 void __user *argp = (void __user *)arg;
884 int pid, err;
885
Eric Dumazetb69aee02005-09-06 14:42:45 -0700886 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
888 err = dev_ioctl(cmd, argp);
889 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100890#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
892 err = dev_ioctl(cmd, argp);
893 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100894#endif /* CONFIG_WIRELESS_EXT */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 switch (cmd) {
896 case FIOSETOWN:
897 case SIOCSPGRP:
898 err = -EFAULT;
899 if (get_user(pid, (int __user *)argp))
900 break;
901 err = f_setown(sock->file, pid, 1);
902 break;
903 case FIOGETOWN:
904 case SIOCGPGRP:
905 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
906 break;
907 case SIOCGIFBR:
908 case SIOCSIFBR:
909 case SIOCBRADDBR:
910 case SIOCBRDELBR:
911 err = -ENOPKG;
912 if (!br_ioctl_hook)
913 request_module("bridge");
914
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800915 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 if (br_ioctl_hook)
917 err = br_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800918 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 break;
920 case SIOCGIFVLAN:
921 case SIOCSIFVLAN:
922 err = -ENOPKG;
923 if (!vlan_ioctl_hook)
924 request_module("8021q");
925
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800926 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 if (vlan_ioctl_hook)
928 err = vlan_ioctl_hook(argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800929 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 break;
931 case SIOCGIFDIVERT:
932 case SIOCSIFDIVERT:
933 /* Convert this to call through a hook */
934 err = divert_ioctl(cmd, argp);
935 break;
936 case SIOCADDDLCI:
937 case SIOCDELDLCI:
938 err = -ENOPKG;
939 if (!dlci_ioctl_hook)
940 request_module("dlci");
941
942 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800943 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800945 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 }
947 break;
948 default:
949 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800950
951 /*
952 * If this ioctl is unknown try to hand it down
953 * to the NIC driver.
954 */
955 if (err == -ENOIOCTLCMD)
956 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 break;
958 }
959 return err;
960}
961
962int sock_create_lite(int family, int type, int protocol, struct socket **res)
963{
964 int err;
965 struct socket *sock = NULL;
966
967 err = security_socket_create(family, type, protocol, 1);
968 if (err)
969 goto out;
970
971 sock = sock_alloc();
972 if (!sock) {
973 err = -ENOMEM;
974 goto out;
975 }
976
977 security_socket_post_create(sock, family, type, protocol, 1);
978 sock->type = type;
979out:
980 *res = sock;
981 return err;
982}
983
984/* No kernel lock held - perfect */
985static unsigned int sock_poll(struct file *file, poll_table * wait)
986{
987 struct socket *sock;
988
989 /*
990 * We can't return errors to poll, so it's either yes or no.
991 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700992 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 return sock->ops->poll(file, sock, wait);
994}
995
996static int sock_mmap(struct file * file, struct vm_area_struct * vma)
997{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700998 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000 return sock->ops->mmap(file, sock, vma);
1001}
1002
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001003static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004{
1005 /*
1006 * It was possible the inode is NULL we were
1007 * closing an unfinished socket.
1008 */
1009
1010 if (!inode)
1011 {
1012 printk(KERN_DEBUG "sock_close: NULL inode\n");
1013 return 0;
1014 }
1015 sock_fasync(-1, filp, 0);
1016 sock_release(SOCKET_I(inode));
1017 return 0;
1018}
1019
1020/*
1021 * Update the socket async list
1022 *
1023 * Fasync_list locking strategy.
1024 *
1025 * 1. fasync_list is modified only under process context socket lock
1026 * i.e. under semaphore.
1027 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1028 * or under socket lock.
1029 * 3. fasync_list can be used from softirq context, so that
1030 * modification under socket lock have to be enhanced with
1031 * write_lock_bh(&sk->sk_callback_lock).
1032 * --ANK (990710)
1033 */
1034
1035static int sock_fasync(int fd, struct file *filp, int on)
1036{
1037 struct fasync_struct *fa, *fna=NULL, **prev;
1038 struct socket *sock;
1039 struct sock *sk;
1040
1041 if (on)
1042 {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001043 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 if(fna==NULL)
1045 return -ENOMEM;
1046 }
1047
Eric Dumazetb69aee02005-09-06 14:42:45 -07001048 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049
1050 if ((sk=sock->sk) == NULL) {
1051 kfree(fna);
1052 return -EINVAL;
1053 }
1054
1055 lock_sock(sk);
1056
1057 prev=&(sock->fasync_list);
1058
1059 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1060 if (fa->fa_file==filp)
1061 break;
1062
1063 if(on)
1064 {
1065 if(fa!=NULL)
1066 {
1067 write_lock_bh(&sk->sk_callback_lock);
1068 fa->fa_fd=fd;
1069 write_unlock_bh(&sk->sk_callback_lock);
1070
1071 kfree(fna);
1072 goto out;
1073 }
1074 fna->fa_file=filp;
1075 fna->fa_fd=fd;
1076 fna->magic=FASYNC_MAGIC;
1077 fna->fa_next=sock->fasync_list;
1078 write_lock_bh(&sk->sk_callback_lock);
1079 sock->fasync_list=fna;
1080 write_unlock_bh(&sk->sk_callback_lock);
1081 }
1082 else
1083 {
1084 if (fa!=NULL)
1085 {
1086 write_lock_bh(&sk->sk_callback_lock);
1087 *prev=fa->fa_next;
1088 write_unlock_bh(&sk->sk_callback_lock);
1089 kfree(fa);
1090 }
1091 }
1092
1093out:
1094 release_sock(sock->sk);
1095 return 0;
1096}
1097
1098/* This function may be called only under socket lock or callback_lock */
1099
1100int sock_wake_async(struct socket *sock, int how, int band)
1101{
1102 if (!sock || !sock->fasync_list)
1103 return -1;
1104 switch (how)
1105 {
1106 case 1:
1107
1108 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1109 break;
1110 goto call_kill;
1111 case 2:
1112 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1113 break;
1114 /* fall through */
1115 case 0:
1116 call_kill:
1117 __kill_fasync(sock->fasync_list, SIGIO, band);
1118 break;
1119 case 3:
1120 __kill_fasync(sock->fasync_list, SIGURG, band);
1121 }
1122 return 0;
1123}
1124
1125static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1126{
1127 int err;
1128 struct socket *sock;
1129
1130 /*
1131 * Check protocol is in range
1132 */
1133 if (family < 0 || family >= NPROTO)
1134 return -EAFNOSUPPORT;
1135 if (type < 0 || type >= SOCK_MAX)
1136 return -EINVAL;
1137
1138 /* Compatibility.
1139
1140 This uglymoron is moved from INET layer to here to avoid
1141 deadlock in module load.
1142 */
1143 if (family == PF_INET && type == SOCK_PACKET) {
1144 static int warned;
1145 if (!warned) {
1146 warned = 1;
1147 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1148 }
1149 family = PF_PACKET;
1150 }
1151
1152 err = security_socket_create(family, type, protocol, kern);
1153 if (err)
1154 return err;
1155
1156#if defined(CONFIG_KMOD)
1157 /* Attempt to load a protocol module if the find failed.
1158 *
1159 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1160 * requested real, full-featured networking support upon configuration.
1161 * Otherwise module support will break!
1162 */
1163 if (net_families[family]==NULL)
1164 {
1165 request_module("net-pf-%d",family);
1166 }
1167#endif
1168
1169 net_family_read_lock();
1170 if (net_families[family] == NULL) {
1171 err = -EAFNOSUPPORT;
1172 goto out;
1173 }
1174
1175/*
1176 * Allocate the socket and allow the family to set things up. if
1177 * the protocol is 0, the family is instructed to select an appropriate
1178 * default.
1179 */
1180
1181 if (!(sock = sock_alloc())) {
1182 printk(KERN_WARNING "socket: no more sockets\n");
1183 err = -ENFILE; /* Not exactly a match, but its the
1184 closest posix thing */
1185 goto out;
1186 }
1187
1188 sock->type = type;
1189
1190 /*
1191 * We will call the ->create function, that possibly is in a loadable
1192 * module, so we have to bump that loadable module refcnt first.
1193 */
1194 err = -EAFNOSUPPORT;
1195 if (!try_module_get(net_families[family]->owner))
1196 goto out_release;
1197
Frank Filza79af592005-09-27 15:23:38 -07001198 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1199 sock->ops = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001201 }
1202
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 /*
1204 * Now to bump the refcnt of the [loadable] module that owns this
1205 * socket at sock_release time we decrement its refcnt.
1206 */
1207 if (!try_module_get(sock->ops->owner)) {
1208 sock->ops = NULL;
1209 goto out_module_put;
1210 }
1211 /*
1212 * Now that we're done with the ->create function, the [loadable]
1213 * module can have its refcnt decremented
1214 */
1215 module_put(net_families[family]->owner);
1216 *res = sock;
1217 security_socket_post_create(sock, family, type, protocol, kern);
1218
1219out:
1220 net_family_read_unlock();
1221 return err;
1222out_module_put:
1223 module_put(net_families[family]->owner);
1224out_release:
1225 sock_release(sock);
1226 goto out;
1227}
1228
1229int sock_create(int family, int type, int protocol, struct socket **res)
1230{
1231 return __sock_create(family, type, protocol, res, 0);
1232}
1233
1234int sock_create_kern(int family, int type, int protocol, struct socket **res)
1235{
1236 return __sock_create(family, type, protocol, res, 1);
1237}
1238
1239asmlinkage long sys_socket(int family, int type, int protocol)
1240{
1241 int retval;
1242 struct socket *sock;
1243
1244 retval = sock_create(family, type, protocol, &sock);
1245 if (retval < 0)
1246 goto out;
1247
1248 retval = sock_map_fd(sock);
1249 if (retval < 0)
1250 goto out_release;
1251
1252out:
1253 /* It may be already another descriptor 8) Not kernel problem. */
1254 return retval;
1255
1256out_release:
1257 sock_release(sock);
1258 return retval;
1259}
1260
1261/*
1262 * Create a pair of connected sockets.
1263 */
1264
1265asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1266{
1267 struct socket *sock1, *sock2;
1268 int fd1, fd2, err;
1269
1270 /*
1271 * Obtain the first socket and check if the underlying protocol
1272 * supports the socketpair call.
1273 */
1274
1275 err = sock_create(family, type, protocol, &sock1);
1276 if (err < 0)
1277 goto out;
1278
1279 err = sock_create(family, type, protocol, &sock2);
1280 if (err < 0)
1281 goto out_release_1;
1282
1283 err = sock1->ops->socketpair(sock1, sock2);
1284 if (err < 0)
1285 goto out_release_both;
1286
1287 fd1 = fd2 = -1;
1288
1289 err = sock_map_fd(sock1);
1290 if (err < 0)
1291 goto out_release_both;
1292 fd1 = err;
1293
1294 err = sock_map_fd(sock2);
1295 if (err < 0)
1296 goto out_close_1;
1297 fd2 = err;
1298
1299 /* fd1 and fd2 may be already another descriptors.
1300 * Not kernel problem.
1301 */
1302
1303 err = put_user(fd1, &usockvec[0]);
1304 if (!err)
1305 err = put_user(fd2, &usockvec[1]);
1306 if (!err)
1307 return 0;
1308
1309 sys_close(fd2);
1310 sys_close(fd1);
1311 return err;
1312
1313out_close_1:
1314 sock_release(sock2);
1315 sys_close(fd1);
1316 return err;
1317
1318out_release_both:
1319 sock_release(sock2);
1320out_release_1:
1321 sock_release(sock1);
1322out:
1323 return err;
1324}
1325
1326
1327/*
1328 * Bind a name to a socket. Nothing much to do here since it's
1329 * the protocol's responsibility to handle the local address.
1330 *
1331 * We move the socket address to kernel space before we call
1332 * the protocol layer (having also checked the address is ok).
1333 */
1334
1335asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1336{
1337 struct socket *sock;
1338 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001339 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001341 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 {
1343 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1344 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001345 if (!err)
1346 err = sock->ops->bind(sock,
1347 (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001349 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 }
1351 return err;
1352}
1353
1354
1355/*
1356 * Perform a listen. Basically, we allow the protocol to do anything
1357 * necessary for a listen, and if that works, we mark the socket as
1358 * ready for listening.
1359 */
1360
1361int sysctl_somaxconn = SOMAXCONN;
1362
1363asmlinkage long sys_listen(int fd, int backlog)
1364{
1365 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001366 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001368 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 if ((unsigned) backlog > sysctl_somaxconn)
1370 backlog = sysctl_somaxconn;
1371
1372 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001373 if (!err)
1374 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001376 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 }
1378 return err;
1379}
1380
1381
1382/*
1383 * For accept, we attempt to create a new socket, set up the link
1384 * with the client, wake up the client, then return the new
1385 * connected fd. We collect the address of the connector in kernel
1386 * space and move it to user at the very end. This is unclean because
1387 * we open the socket then return an error.
1388 *
1389 * 1003.1g adds the ability to recvmsg() to query connection pending
1390 * status to recvmsg. We need to add that support in a way thats
1391 * clean when we restucture accept also.
1392 */
1393
1394asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1395{
1396 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001397 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001398 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 char address[MAX_SOCK_ADDR];
1400
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001401 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 if (!sock)
1403 goto out;
1404
1405 err = -ENFILE;
1406 if (!(newsock = sock_alloc()))
1407 goto out_put;
1408
1409 newsock->type = sock->type;
1410 newsock->ops = sock->ops;
1411
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 /*
1413 * We don't need try_module_get here, as the listening socket (sock)
1414 * has the protocol module (sock->ops->owner) held.
1415 */
1416 __module_get(newsock->ops->owner);
1417
David S. Miller39d8c1b2006-03-20 17:13:49 -08001418 newfd = sock_alloc_fd(&newfile);
1419 if (unlikely(newfd < 0)) {
1420 err = newfd;
1421 goto out_release;
1422 }
1423
1424 err = sock_attach_fd(newsock, newfile);
1425 if (err < 0)
1426 goto out_fd;
1427
Frank Filza79af592005-09-27 15:23:38 -07001428 err = security_socket_accept(sock, newsock);
1429 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001430 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001431
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1433 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001434 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
1436 if (upeer_sockaddr) {
1437 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1438 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001439 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 }
1441 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1442 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001443 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 }
1445
1446 /* File flags are not inherited via accept() unlike another OSes. */
1447
David S. Miller39d8c1b2006-03-20 17:13:49 -08001448 fd_install(newfd, newfile);
1449 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450
1451 security_socket_post_accept(sock, newsock);
1452
1453out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001454 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455out:
1456 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001457out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001458 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001459 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460out_release:
1461 sock_release(newsock);
1462 goto out_put;
1463}
1464
1465
1466/*
1467 * Attempt to connect to a socket with the server address. The address
1468 * is in user space so we verify it is OK and move it to kernel space.
1469 *
1470 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1471 * break bindings
1472 *
1473 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1474 * other SEQPACKET protocols that take time to connect() as it doesn't
1475 * include the -EINPROGRESS status for such sockets.
1476 */
1477
1478asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1479{
1480 struct socket *sock;
1481 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001482 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 if (!sock)
1486 goto out;
1487 err = move_addr_to_kernel(uservaddr, addrlen, address);
1488 if (err < 0)
1489 goto out_put;
1490
1491 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1492 if (err)
1493 goto out_put;
1494
1495 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1496 sock->file->f_flags);
1497out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001498 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499out:
1500 return err;
1501}
1502
1503/*
1504 * Get the local address ('name') of a socket object. Move the obtained
1505 * name to user space.
1506 */
1507
1508asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1509{
1510 struct socket *sock;
1511 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001512 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001514 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 if (!sock)
1516 goto out;
1517
1518 err = security_socket_getsockname(sock);
1519 if (err)
1520 goto out_put;
1521
1522 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1523 if (err)
1524 goto out_put;
1525 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1526
1527out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001528 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529out:
1530 return err;
1531}
1532
1533/*
1534 * Get the remote address ('name') of a socket object. Move the obtained
1535 * name to user space.
1536 */
1537
1538asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1539{
1540 struct socket *sock;
1541 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001542 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001544 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 err = security_socket_getpeername(sock);
1546 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001547 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 return err;
1549 }
1550
1551 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1552 if (!err)
1553 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001554 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 }
1556 return err;
1557}
1558
1559/*
1560 * Send a datagram to a given address. We move the address into kernel
1561 * space and check the user space data area is readable before invoking
1562 * the protocol.
1563 */
1564
1565asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1566 struct sockaddr __user *addr, int addr_len)
1567{
1568 struct socket *sock;
1569 char address[MAX_SOCK_ADDR];
1570 int err;
1571 struct msghdr msg;
1572 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001573 int fput_needed;
1574 struct file *sock_file;
1575
1576 sock_file = fget_light(fd, &fput_needed);
1577 if (!sock_file)
1578 return -EBADF;
1579
1580 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001582 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 iov.iov_base=buff;
1584 iov.iov_len=len;
1585 msg.msg_name=NULL;
1586 msg.msg_iov=&iov;
1587 msg.msg_iovlen=1;
1588 msg.msg_control=NULL;
1589 msg.msg_controllen=0;
1590 msg.msg_namelen=0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001591 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 err = move_addr_to_kernel(addr, addr_len, address);
1593 if (err < 0)
1594 goto out_put;
1595 msg.msg_name=address;
1596 msg.msg_namelen=addr_len;
1597 }
1598 if (sock->file->f_flags & O_NONBLOCK)
1599 flags |= MSG_DONTWAIT;
1600 msg.msg_flags = flags;
1601 err = sock_sendmsg(sock, &msg, len);
1602
1603out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001604 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 return err;
1606}
1607
1608/*
1609 * Send a datagram down a socket.
1610 */
1611
1612asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1613{
1614 return sys_sendto(fd, buff, len, flags, NULL, 0);
1615}
1616
1617/*
1618 * Receive a frame from the socket and optionally record the address of the
1619 * sender. We verify the buffers are writable and if needed move the
1620 * sender address from kernel to user space.
1621 */
1622
1623asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1624 struct sockaddr __user *addr, int __user *addr_len)
1625{
1626 struct socket *sock;
1627 struct iovec iov;
1628 struct msghdr msg;
1629 char address[MAX_SOCK_ADDR];
1630 int err,err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001631 struct file *sock_file;
1632 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001634 sock_file = fget_light(fd, &fput_needed);
1635 if (!sock_file)
1636 return -EBADF;
1637
1638 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 if (!sock)
1640 goto out;
1641
1642 msg.msg_control=NULL;
1643 msg.msg_controllen=0;
1644 msg.msg_iovlen=1;
1645 msg.msg_iov=&iov;
1646 iov.iov_len=size;
1647 iov.iov_base=ubuf;
1648 msg.msg_name=address;
1649 msg.msg_namelen=MAX_SOCK_ADDR;
1650 if (sock->file->f_flags & O_NONBLOCK)
1651 flags |= MSG_DONTWAIT;
1652 err=sock_recvmsg(sock, &msg, size, flags);
1653
1654 if(err >= 0 && addr != NULL)
1655 {
1656 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1657 if(err2<0)
1658 err=err2;
1659 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660out:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001661 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 return err;
1663}
1664
1665/*
1666 * Receive a datagram from a socket.
1667 */
1668
1669asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1670{
1671 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1672}
1673
1674/*
1675 * Set a socket option. Because we don't know the option lengths we have
1676 * to pass the user mode parameter for the protocols to sort out.
1677 */
1678
1679asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1680{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001681 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 struct socket *sock;
1683
1684 if (optlen < 0)
1685 return -EINVAL;
1686
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001687 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 {
1689 err = security_socket_setsockopt(sock,level,optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001690 if (err)
1691 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692
1693 if (level == SOL_SOCKET)
1694 err=sock_setsockopt(sock,level,optname,optval,optlen);
1695 else
1696 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001697out_put:
1698 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 }
1700 return err;
1701}
1702
1703/*
1704 * Get a socket option. Because we don't know the option lengths we have
1705 * to pass a user mode parameter for the protocols to sort out.
1706 */
1707
1708asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1709{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001710 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 struct socket *sock;
1712
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001713 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1714 err = security_socket_getsockopt(sock, level, optname);
1715 if (err)
1716 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717
1718 if (level == SOL_SOCKET)
1719 err=sock_getsockopt(sock,level,optname,optval,optlen);
1720 else
1721 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001722out_put:
1723 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 }
1725 return err;
1726}
1727
1728
1729/*
1730 * Shutdown a socket.
1731 */
1732
1733asmlinkage long sys_shutdown(int fd, int how)
1734{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001735 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 struct socket *sock;
1737
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001738 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 {
1740 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001741 if (!err)
1742 err = sock->ops->shutdown(sock, how);
1743 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 }
1745 return err;
1746}
1747
1748/* A couple of helpful macros for getting the address of the 32/64 bit
1749 * fields which are the same type (int / unsigned) on our platforms.
1750 */
1751#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1752#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1753#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1754
1755
1756/*
1757 * BSD sendmsg interface
1758 */
1759
1760asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1761{
1762 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1763 struct socket *sock;
1764 char address[MAX_SOCK_ADDR];
1765 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001766 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1767 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1768 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 unsigned char *ctl_buf = ctl;
1770 struct msghdr msg_sys;
1771 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001772 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
1774 err = -EFAULT;
1775 if (MSG_CMSG_COMPAT & flags) {
1776 if (get_compat_msghdr(&msg_sys, msg_compat))
1777 return -EFAULT;
1778 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1779 return -EFAULT;
1780
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001781 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 if (!sock)
1783 goto out;
1784
1785 /* do not move before msg_sys is valid */
1786 err = -EMSGSIZE;
1787 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1788 goto out_put;
1789
1790 /* Check whether to allocate the iovec area*/
1791 err = -ENOMEM;
1792 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1793 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1794 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1795 if (!iov)
1796 goto out_put;
1797 }
1798
1799 /* This will also move the address data into kernel space */
1800 if (MSG_CMSG_COMPAT & flags) {
1801 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1802 } else
1803 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1804 if (err < 0)
1805 goto out_freeiov;
1806 total_len = err;
1807
1808 err = -ENOBUFS;
1809
1810 if (msg_sys.msg_controllen > INT_MAX)
1811 goto out_freeiov;
1812 ctl_len = msg_sys.msg_controllen;
1813 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Al Viro8920e8f2005-09-07 18:28:51 -07001814 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 if (err)
1816 goto out_freeiov;
1817 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001818 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 } else if (ctl_len) {
1820 if (ctl_len > sizeof(ctl))
1821 {
1822 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1823 if (ctl_buf == NULL)
1824 goto out_freeiov;
1825 }
1826 err = -EFAULT;
1827 /*
1828 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1829 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1830 * checking falls down on this.
1831 */
1832 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1833 goto out_freectl;
1834 msg_sys.msg_control = ctl_buf;
1835 }
1836 msg_sys.msg_flags = flags;
1837
1838 if (sock->file->f_flags & O_NONBLOCK)
1839 msg_sys.msg_flags |= MSG_DONTWAIT;
1840 err = sock_sendmsg(sock, &msg_sys, total_len);
1841
1842out_freectl:
1843 if (ctl_buf != ctl)
1844 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1845out_freeiov:
1846 if (iov != iovstack)
1847 sock_kfree_s(sock->sk, iov, iov_size);
1848out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001849 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850out:
1851 return err;
1852}
1853
1854/*
1855 * BSD recvmsg interface
1856 */
1857
1858asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1859{
1860 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1861 struct socket *sock;
1862 struct iovec iovstack[UIO_FASTIOV];
1863 struct iovec *iov=iovstack;
1864 struct msghdr msg_sys;
1865 unsigned long cmsg_ptr;
1866 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001867 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868
1869 /* kernel mode address */
1870 char addr[MAX_SOCK_ADDR];
1871
1872 /* user mode address pointers */
1873 struct sockaddr __user *uaddr;
1874 int __user *uaddr_len;
1875
1876 if (MSG_CMSG_COMPAT & flags) {
1877 if (get_compat_msghdr(&msg_sys, msg_compat))
1878 return -EFAULT;
1879 } else
1880 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1881 return -EFAULT;
1882
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001883 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 if (!sock)
1885 goto out;
1886
1887 err = -EMSGSIZE;
1888 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1889 goto out_put;
1890
1891 /* Check whether to allocate the iovec area*/
1892 err = -ENOMEM;
1893 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1894 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1895 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1896 if (!iov)
1897 goto out_put;
1898 }
1899
1900 /*
1901 * Save the user-mode address (verify_iovec will change the
1902 * kernel msghdr to use the kernel address space)
1903 */
1904
1905 uaddr = (void __user *) msg_sys.msg_name;
1906 uaddr_len = COMPAT_NAMELEN(msg);
1907 if (MSG_CMSG_COMPAT & flags) {
1908 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1909 } else
1910 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1911 if (err < 0)
1912 goto out_freeiov;
1913 total_len=err;
1914
1915 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1916 msg_sys.msg_flags = 0;
1917 if (MSG_CMSG_COMPAT & flags)
1918 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1919
1920 if (sock->file->f_flags & O_NONBLOCK)
1921 flags |= MSG_DONTWAIT;
1922 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1923 if (err < 0)
1924 goto out_freeiov;
1925 len = err;
1926
1927 if (uaddr != NULL) {
1928 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1929 if (err < 0)
1930 goto out_freeiov;
1931 }
David S. Miller37f7f422005-09-16 16:51:01 -07001932 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1933 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 if (err)
1935 goto out_freeiov;
1936 if (MSG_CMSG_COMPAT & flags)
1937 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1938 &msg_compat->msg_controllen);
1939 else
1940 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1941 &msg->msg_controllen);
1942 if (err)
1943 goto out_freeiov;
1944 err = len;
1945
1946out_freeiov:
1947 if (iov != iovstack)
1948 sock_kfree_s(sock->sk, iov, iov_size);
1949out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001950 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951out:
1952 return err;
1953}
1954
1955#ifdef __ARCH_WANT_SYS_SOCKETCALL
1956
1957/* Argument list sizes for sys_socketcall */
1958#define AL(x) ((x) * sizeof(unsigned long))
1959static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1960 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1961 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1962#undef AL
1963
1964/*
1965 * System call vectors.
1966 *
1967 * Argument checking cleaned up. Saved 20% in size.
1968 * This function doesn't need to set the kernel lock because
1969 * it is set by the callees.
1970 */
1971
1972asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1973{
1974 unsigned long a[6];
1975 unsigned long a0,a1;
1976 int err;
1977
1978 if(call<1||call>SYS_RECVMSG)
1979 return -EINVAL;
1980
1981 /* copy_from_user should be SMP safe. */
1982 if (copy_from_user(a, args, nargs[call]))
1983 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001984
David Woodhouse4bcff1b2005-06-02 12:13:21 +01001985 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001986 if (err)
1987 return err;
1988
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 a0=a[0];
1990 a1=a[1];
1991
1992 switch(call)
1993 {
1994 case SYS_SOCKET:
1995 err = sys_socket(a0,a1,a[2]);
1996 break;
1997 case SYS_BIND:
1998 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1999 break;
2000 case SYS_CONNECT:
2001 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2002 break;
2003 case SYS_LISTEN:
2004 err = sys_listen(a0,a1);
2005 break;
2006 case SYS_ACCEPT:
2007 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2008 break;
2009 case SYS_GETSOCKNAME:
2010 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2011 break;
2012 case SYS_GETPEERNAME:
2013 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2014 break;
2015 case SYS_SOCKETPAIR:
2016 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2017 break;
2018 case SYS_SEND:
2019 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2020 break;
2021 case SYS_SENDTO:
2022 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2023 (struct sockaddr __user *)a[4], a[5]);
2024 break;
2025 case SYS_RECV:
2026 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2027 break;
2028 case SYS_RECVFROM:
2029 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2030 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2031 break;
2032 case SYS_SHUTDOWN:
2033 err = sys_shutdown(a0,a1);
2034 break;
2035 case SYS_SETSOCKOPT:
2036 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2037 break;
2038 case SYS_GETSOCKOPT:
2039 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2040 break;
2041 case SYS_SENDMSG:
2042 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2043 break;
2044 case SYS_RECVMSG:
2045 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2046 break;
2047 default:
2048 err = -EINVAL;
2049 break;
2050 }
2051 return err;
2052}
2053
2054#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2055
2056/*
2057 * This function is called by a protocol handler that wants to
2058 * advertise its address family, and have it linked into the
2059 * SOCKET module.
2060 */
2061
2062int sock_register(struct net_proto_family *ops)
2063{
2064 int err;
2065
2066 if (ops->family >= NPROTO) {
2067 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2068 return -ENOBUFS;
2069 }
2070 net_family_write_lock();
2071 err = -EEXIST;
2072 if (net_families[ops->family] == NULL) {
2073 net_families[ops->family]=ops;
2074 err = 0;
2075 }
2076 net_family_write_unlock();
2077 printk(KERN_INFO "NET: Registered protocol family %d\n",
2078 ops->family);
2079 return err;
2080}
2081
2082/*
2083 * This function is called by a protocol handler that wants to
2084 * remove its address family, and have it unlinked from the
2085 * SOCKET module.
2086 */
2087
2088int sock_unregister(int family)
2089{
2090 if (family < 0 || family >= NPROTO)
2091 return -1;
2092
2093 net_family_write_lock();
2094 net_families[family]=NULL;
2095 net_family_write_unlock();
2096 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2097 family);
2098 return 0;
2099}
2100
Andi Kleen77d76ea2005-12-22 12:43:42 -08002101static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102{
2103 /*
2104 * Initialize sock SLAB cache.
2105 */
2106
2107 sk_init();
2108
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 /*
2110 * Initialize skbuff SLAB cache
2111 */
2112 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113
2114 /*
2115 * Initialize the protocols module.
2116 */
2117
2118 init_inodecache();
2119 register_filesystem(&sock_fs_type);
2120 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002121
2122 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 */
2124
2125#ifdef CONFIG_NETFILTER
2126 netfilter_init();
2127#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002128
2129 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130}
2131
Andi Kleen77d76ea2005-12-22 12:43:42 -08002132core_initcall(sock_init); /* early initcall */
2133
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134#ifdef CONFIG_PROC_FS
2135void socket_seq_show(struct seq_file *seq)
2136{
2137 int cpu;
2138 int counter = 0;
2139
Eric Dumazet88a2a4ac2006-02-04 23:27:36 -08002140 for_each_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 counter += per_cpu(sockets_in_use, cpu);
2142
2143 /* It can be negative, by the way. 8) */
2144 if (counter < 0)
2145 counter = 0;
2146
2147 seq_printf(seq, "sockets: used %d\n", counter);
2148}
2149#endif /* CONFIG_PROC_FS */
2150
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002151#ifdef CONFIG_COMPAT
2152static long compat_sock_ioctl(struct file *file, unsigned cmd,
2153 unsigned long arg)
2154{
2155 struct socket *sock = file->private_data;
2156 int ret = -ENOIOCTLCMD;
2157
2158 if (sock->ops->compat_ioctl)
2159 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2160
2161 return ret;
2162}
2163#endif
2164
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165/* ABI emulation layers need these two */
2166EXPORT_SYMBOL(move_addr_to_kernel);
2167EXPORT_SYMBOL(move_addr_to_user);
2168EXPORT_SYMBOL(sock_create);
2169EXPORT_SYMBOL(sock_create_kern);
2170EXPORT_SYMBOL(sock_create_lite);
2171EXPORT_SYMBOL(sock_map_fd);
2172EXPORT_SYMBOL(sock_recvmsg);
2173EXPORT_SYMBOL(sock_register);
2174EXPORT_SYMBOL(sock_release);
2175EXPORT_SYMBOL(sock_sendmsg);
2176EXPORT_SYMBOL(sock_unregister);
2177EXPORT_SYMBOL(sock_wake_async);
2178EXPORT_SYMBOL(sockfd_lookup);
2179EXPORT_SYMBOL(kernel_sendmsg);
2180EXPORT_SYMBOL(kernel_recvmsg);