blob: e3c21d5ec2885aa4fea17eea23621580c07a6fb7 [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);
110static int sock_fasync(int fd, struct file *filp, int on);
111static ssize_t sock_readv(struct file *file, const struct iovec *vector,
112 unsigned long count, loff_t *ppos);
113static ssize_t sock_writev(struct file *file, const struct iovec *vector,
114 unsigned long count, loff_t *ppos);
115static ssize_t sock_sendpage(struct file *file, struct page *page,
116 int offset, size_t size, loff_t *ppos, int more);
117
118
119/*
120 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
121 * in the operation structures but are done directly via the socketcall() multiplexor.
122 */
123
124static struct file_operations socket_file_ops = {
125 .owner = THIS_MODULE,
126 .llseek = no_llseek,
127 .aio_read = sock_aio_read,
128 .aio_write = sock_aio_write,
129 .poll = sock_poll,
130 .unlocked_ioctl = sock_ioctl,
131 .mmap = sock_mmap,
132 .open = sock_no_open, /* special open code to disallow open via /proc */
133 .release = sock_close,
134 .fasync = sock_fasync,
135 .readv = sock_readv,
136 .writev = sock_writev,
137 .sendpage = sock_sendpage
138};
139
140/*
141 * The protocol list. Each protocol is registered in here.
142 */
143
144static struct net_proto_family *net_families[NPROTO];
145
146#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
147static atomic_t net_family_lockct = ATOMIC_INIT(0);
148static DEFINE_SPINLOCK(net_family_lock);
149
150/* The strategy is: modifications net_family vector are short, do not
151 sleep and veeery rare, but read access should be free of any exclusive
152 locks.
153 */
154
155static void net_family_write_lock(void)
156{
157 spin_lock(&net_family_lock);
158 while (atomic_read(&net_family_lockct) != 0) {
159 spin_unlock(&net_family_lock);
160
161 yield();
162
163 spin_lock(&net_family_lock);
164 }
165}
166
167static __inline__ void net_family_write_unlock(void)
168{
169 spin_unlock(&net_family_lock);
170}
171
172static __inline__ void net_family_read_lock(void)
173{
174 atomic_inc(&net_family_lockct);
175 spin_unlock_wait(&net_family_lock);
176}
177
178static __inline__ void net_family_read_unlock(void)
179{
180 atomic_dec(&net_family_lockct);
181}
182
183#else
184#define net_family_write_lock() do { } while(0)
185#define net_family_write_unlock() do { } while(0)
186#define net_family_read_lock() do { } while(0)
187#define net_family_read_unlock() do { } while(0)
188#endif
189
190
191/*
192 * Statistics counters of the socket lists
193 */
194
195static DEFINE_PER_CPU(int, sockets_in_use) = 0;
196
197/*
198 * Support routines. Move socket addresses back and forth across the kernel/user
199 * divide and look after the messy bits.
200 */
201
202#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
203 16 for IP, 16 for IPX,
204 24 for IPv6,
205 about 80 for AX.25
206 must be at least one bigger than
207 the AF_UNIX size (see net/unix/af_unix.c
208 :unix_mkname()).
209 */
210
211/**
212 * move_addr_to_kernel - copy a socket address into kernel space
213 * @uaddr: Address in user space
214 * @kaddr: Address in kernel space
215 * @ulen: Length in user space
216 *
217 * The address is copied into kernel space. If the provided address is
218 * too long an error code of -EINVAL is returned. If the copy gives
219 * invalid addresses -EFAULT is returned. On a success 0 is returned.
220 */
221
222int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
223{
224 if(ulen<0||ulen>MAX_SOCK_ADDR)
225 return -EINVAL;
226 if(ulen==0)
227 return 0;
228 if(copy_from_user(kaddr,uaddr,ulen))
229 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100230 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231}
232
233/**
234 * move_addr_to_user - copy an address to user space
235 * @kaddr: kernel space address
236 * @klen: length of address in kernel
237 * @uaddr: user space address
238 * @ulen: pointer to user length field
239 *
240 * The value pointed to by ulen on entry is the buffer length available.
241 * This is overwritten with the buffer space used. -EINVAL is returned
242 * if an overlong buffer is specified or a negative buffer size. -EFAULT
243 * is returned if either the buffer or the length field are not
244 * accessible.
245 * After copying the data up to the limit the user specifies, the true
246 * length of the data is written over the length limit the user
247 * specified. Zero is returned for a success.
248 */
249
250int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
251{
252 int err;
253 int len;
254
255 if((err=get_user(len, ulen)))
256 return err;
257 if(len>klen)
258 len=klen;
259 if(len<0 || len> MAX_SOCK_ADDR)
260 return -EINVAL;
261 if(len)
262 {
263 if(copy_to_user(uaddr,kaddr,len))
264 return -EFAULT;
265 }
266 /*
267 * "fromlen shall refer to the value before truncation.."
268 * 1003.1g
269 */
270 return __put_user(klen, ulen);
271}
272
273#define SOCKFS_MAGIC 0x534F434B
274
Eric Dumazetba899662005-08-26 12:05:31 -0700275static kmem_cache_t * sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276
277static struct inode *sock_alloc_inode(struct super_block *sb)
278{
279 struct socket_alloc *ei;
280 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
281 if (!ei)
282 return NULL;
283 init_waitqueue_head(&ei->socket.wait);
284
285 ei->socket.fasync_list = NULL;
286 ei->socket.state = SS_UNCONNECTED;
287 ei->socket.flags = 0;
288 ei->socket.ops = NULL;
289 ei->socket.sk = NULL;
290 ei->socket.file = NULL;
291 ei->socket.flags = 0;
292
293 return &ei->vfs_inode;
294}
295
296static void sock_destroy_inode(struct inode *inode)
297{
298 kmem_cache_free(sock_inode_cachep,
299 container_of(inode, struct socket_alloc, vfs_inode));
300}
301
302static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
303{
304 struct socket_alloc *ei = (struct socket_alloc *) foo;
305
306 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
307 SLAB_CTOR_CONSTRUCTOR)
308 inode_init_once(&ei->vfs_inode);
309}
310
311static int init_inodecache(void)
312{
313 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
314 sizeof(struct socket_alloc),
315 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
316 init_once, NULL);
317 if (sock_inode_cachep == NULL)
318 return -ENOMEM;
319 return 0;
320}
321
322static struct super_operations sockfs_ops = {
323 .alloc_inode = sock_alloc_inode,
324 .destroy_inode =sock_destroy_inode,
325 .statfs = simple_statfs,
326};
327
328static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
329 int flags, const char *dev_name, void *data)
330{
331 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
332}
333
Eric Dumazetba899662005-08-26 12:05:31 -0700334static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
336static struct file_system_type sock_fs_type = {
337 .name = "sockfs",
338 .get_sb = sockfs_get_sb,
339 .kill_sb = kill_anon_super,
340};
341static int sockfs_delete_dentry(struct dentry *dentry)
342{
343 return 1;
344}
345static struct dentry_operations sockfs_dentry_operations = {
346 .d_delete = sockfs_delete_dentry,
347};
348
349/*
350 * Obtains the first available file descriptor and sets it up for use.
351 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800352 * These functions create file structures and maps them to fd space
353 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 * and file struct implicitly stored in sock->file.
355 * Note that another thread may close file descriptor before we return
356 * from this function. We use the fact that now we do not refer
357 * to socket after mapping. If one day we will need it, this
358 * function will increment ref. count on file by 1.
359 *
360 * In any case returned fd MAY BE not valid!
361 * This race condition is unavoidable
362 * with shared fd spaces, we cannot solve it inside kernel,
363 * but we take care of internal coherence yet.
364 */
365
David S. Miller39d8c1b2006-03-20 17:13:49 -0800366static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367{
368 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800369
370 fd = get_unused_fd();
371 if (likely(fd >= 0)) {
372 struct file *file = get_empty_filp();
373
374 *filep = file;
375 if (unlikely(!file)) {
376 put_unused_fd(fd);
377 return -ENFILE;
378 }
379 } else
380 *filep = NULL;
381 return fd;
382}
383
384static int sock_attach_fd(struct socket *sock, struct file *file)
385{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 struct qstr this;
387 char name[32];
388
David S. Miller39d8c1b2006-03-20 17:13:49 -0800389 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
390 this.name = name;
391 this.hash = SOCK_INODE(sock)->i_ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
394 if (unlikely(!file->f_dentry))
395 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396
David S. Miller39d8c1b2006-03-20 17:13:49 -0800397 file->f_dentry->d_op = &sockfs_dentry_operations;
398 d_add(file->f_dentry, SOCK_INODE(sock));
399 file->f_vfsmnt = mntget(sock_mnt);
400 file->f_mapping = file->f_dentry->d_inode->i_mapping;
401
402 sock->file = file;
403 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
404 file->f_mode = FMODE_READ | FMODE_WRITE;
405 file->f_flags = O_RDWR;
406 file->f_pos = 0;
407 file->private_data = sock;
408
409 return 0;
410}
411
412int sock_map_fd(struct socket *sock)
413{
414 struct file *newfile;
415 int fd = sock_alloc_fd(&newfile);
416
417 if (likely(fd >= 0)) {
418 int err = sock_attach_fd(sock, newfile);
419
420 if (unlikely(err < 0)) {
421 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800423 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800425 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 return fd;
428}
429
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800430static struct socket *sock_from_file(struct file *file, int *err)
431{
432 struct inode *inode;
433 struct socket *sock;
434
435 if (file->f_op == &socket_file_ops)
436 return file->private_data; /* set in sock_map_fd */
437
438 inode = file->f_dentry->d_inode;
439 if (!S_ISSOCK(inode->i_mode)) {
440 *err = -ENOTSOCK;
441 return NULL;
442 }
443
444 sock = SOCKET_I(inode);
445 if (sock->file != file) {
446 printk(KERN_ERR "socki_lookup: socket file changed!\n");
447 sock->file = file;
448 }
449 return sock;
450}
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452/**
453 * sockfd_lookup - Go from a file number to its socket slot
454 * @fd: file handle
455 * @err: pointer to an error code return
456 *
457 * The file handle passed in is locked and the socket it is bound
458 * too is returned. If an error occurs the err pointer is overwritten
459 * with a negative errno code and NULL is returned. The function checks
460 * for both invalid handles and passing a handle which is not a socket.
461 *
462 * On a success the socket object pointer is returned.
463 */
464
465struct socket *sockfd_lookup(int fd, int *err)
466{
467 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 struct socket *sock;
469
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800470 if (!(file = fget(fd))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 *err = -EBADF;
472 return NULL;
473 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800474 sock = sock_from_file(file, err);
475 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 return sock;
478}
479
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800480static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
481{
482 struct file *file;
483 struct socket *sock;
484
485 file = fget_light(fd, fput_needed);
486 if (file) {
487 sock = sock_from_file(file, err);
488 if (sock)
489 return sock;
490 fput_light(file, *fput_needed);
491 }
492 return NULL;
493}
494
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495/**
496 * sock_alloc - allocate a socket
497 *
498 * Allocate a new inode and socket object. The two are bound together
499 * and initialised. The socket is then returned. If we are out of inodes
500 * NULL is returned.
501 */
502
503static struct socket *sock_alloc(void)
504{
505 struct inode * inode;
506 struct socket * sock;
507
508 inode = new_inode(sock_mnt->mnt_sb);
509 if (!inode)
510 return NULL;
511
512 sock = SOCKET_I(inode);
513
514 inode->i_mode = S_IFSOCK|S_IRWXUGO;
515 inode->i_uid = current->fsuid;
516 inode->i_gid = current->fsgid;
517
518 get_cpu_var(sockets_in_use)++;
519 put_cpu_var(sockets_in_use);
520 return sock;
521}
522
523/*
524 * In theory you can't get an open on this inode, but /proc provides
525 * a back door. Remember to keep it shut otherwise you'll let the
526 * creepy crawlies in.
527 */
528
529static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
530{
531 return -ENXIO;
532}
533
534struct file_operations bad_sock_fops = {
535 .owner = THIS_MODULE,
536 .open = sock_no_open,
537};
538
539/**
540 * sock_release - close a socket
541 * @sock: socket to close
542 *
543 * The socket is released from the protocol stack if it has a release
544 * callback, and the inode is then released if the socket is bound to
545 * an inode not a file.
546 */
547
548void sock_release(struct socket *sock)
549{
550 if (sock->ops) {
551 struct module *owner = sock->ops->owner;
552
553 sock->ops->release(sock);
554 sock->ops = NULL;
555 module_put(owner);
556 }
557
558 if (sock->fasync_list)
559 printk(KERN_ERR "sock_release: fasync list not empty!\n");
560
561 get_cpu_var(sockets_in_use)--;
562 put_cpu_var(sockets_in_use);
563 if (!sock->file) {
564 iput(SOCK_INODE(sock));
565 return;
566 }
567 sock->file=NULL;
568}
569
570static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
571 struct msghdr *msg, size_t size)
572{
573 struct sock_iocb *si = kiocb_to_siocb(iocb);
574 int err;
575
576 si->sock = sock;
577 si->scm = NULL;
578 si->msg = msg;
579 si->size = size;
580
581 err = security_socket_sendmsg(sock, msg, size);
582 if (err)
583 return err;
584
585 return sock->ops->sendmsg(iocb, sock, msg, size);
586}
587
588int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
589{
590 struct kiocb iocb;
591 struct sock_iocb siocb;
592 int ret;
593
594 init_sync_kiocb(&iocb, NULL);
595 iocb.private = &siocb;
596 ret = __sock_sendmsg(&iocb, sock, msg, size);
597 if (-EIOCBQUEUED == ret)
598 ret = wait_on_sync_kiocb(&iocb);
599 return ret;
600}
601
602int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
603 struct kvec *vec, size_t num, size_t size)
604{
605 mm_segment_t oldfs = get_fs();
606 int result;
607
608 set_fs(KERNEL_DS);
609 /*
610 * the following is safe, since for compiler definitions of kvec and
611 * iovec are identical, yielding the same in-core layout and alignment
612 */
613 msg->msg_iov = (struct iovec *)vec,
614 msg->msg_iovlen = num;
615 result = sock_sendmsg(sock, msg, size);
616 set_fs(oldfs);
617 return result;
618}
619
620static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
621 struct msghdr *msg, size_t size, int flags)
622{
623 int err;
624 struct sock_iocb *si = kiocb_to_siocb(iocb);
625
626 si->sock = sock;
627 si->scm = NULL;
628 si->msg = msg;
629 si->size = size;
630 si->flags = flags;
631
632 err = security_socket_recvmsg(sock, msg, size, flags);
633 if (err)
634 return err;
635
636 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
637}
638
639int sock_recvmsg(struct socket *sock, struct msghdr *msg,
640 size_t size, int flags)
641{
642 struct kiocb iocb;
643 struct sock_iocb siocb;
644 int ret;
645
646 init_sync_kiocb(&iocb, NULL);
647 iocb.private = &siocb;
648 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
649 if (-EIOCBQUEUED == ret)
650 ret = wait_on_sync_kiocb(&iocb);
651 return ret;
652}
653
654int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
655 struct kvec *vec, size_t num,
656 size_t size, int flags)
657{
658 mm_segment_t oldfs = get_fs();
659 int result;
660
661 set_fs(KERNEL_DS);
662 /*
663 * the following is safe, since for compiler definitions of kvec and
664 * iovec are identical, yielding the same in-core layout and alignment
665 */
666 msg->msg_iov = (struct iovec *)vec,
667 msg->msg_iovlen = num;
668 result = sock_recvmsg(sock, msg, size, flags);
669 set_fs(oldfs);
670 return result;
671}
672
673static void sock_aio_dtor(struct kiocb *iocb)
674{
675 kfree(iocb->private);
676}
677
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300678static ssize_t sock_sendpage(struct file *file, struct page *page,
679 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680{
681 struct socket *sock;
682 int flags;
683
Eric Dumazetb69aee02005-09-06 14:42:45 -0700684 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685
686 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
687 if (more)
688 flags |= MSG_MORE;
689
690 return sock->ops->sendpage(sock, page, offset, size, flags);
691}
692
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800693static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
694 char __user *ubuf, size_t size, struct sock_iocb *siocb)
695{
696 if (!is_sync_kiocb(iocb)) {
697 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
698 if (!siocb)
699 return NULL;
700 iocb->ki_dtor = sock_aio_dtor;
701 }
702
703 siocb->kiocb = iocb;
704 siocb->async_iov.iov_base = ubuf;
705 siocb->async_iov.iov_len = size;
706
707 iocb->private = siocb;
708 return siocb;
709}
710
711static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
712 struct file *file, struct iovec *iov, unsigned long nr_segs)
713{
714 struct socket *sock = file->private_data;
715 size_t size = 0;
716 int i;
717
718 for (i = 0 ; i < nr_segs ; i++)
719 size += iov[i].iov_len;
720
721 msg->msg_name = NULL;
722 msg->msg_namelen = 0;
723 msg->msg_control = NULL;
724 msg->msg_controllen = 0;
725 msg->msg_iov = (struct iovec *) iov;
726 msg->msg_iovlen = nr_segs;
727 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
728
729 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
730}
731
732static ssize_t sock_readv(struct file *file, const struct iovec *iov,
733 unsigned long nr_segs, loff_t *ppos)
734{
735 struct kiocb iocb;
736 struct sock_iocb siocb;
737 struct msghdr msg;
738 int ret;
739
740 init_sync_kiocb(&iocb, NULL);
741 iocb.private = &siocb;
742
743 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
744 if (-EIOCBQUEUED == ret)
745 ret = wait_on_sync_kiocb(&iocb);
746 return ret;
747}
748
749static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
750 size_t count, loff_t pos)
751{
752 struct sock_iocb siocb, *x;
753
754 if (pos != 0)
755 return -ESPIPE;
756 if (count == 0) /* Match SYS5 behaviour */
757 return 0;
758
759 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
760 if (!x)
761 return -ENOMEM;
762 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
763 &x->async_iov, 1);
764}
765
766static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
767 struct file *file, struct iovec *iov, unsigned long nr_segs)
768{
769 struct socket *sock = file->private_data;
770 size_t size = 0;
771 int i;
772
773 for (i = 0 ; i < nr_segs ; i++)
774 size += iov[i].iov_len;
775
776 msg->msg_name = NULL;
777 msg->msg_namelen = 0;
778 msg->msg_control = NULL;
779 msg->msg_controllen = 0;
780 msg->msg_iov = (struct iovec *) iov;
781 msg->msg_iovlen = nr_segs;
782 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
783 if (sock->type == SOCK_SEQPACKET)
784 msg->msg_flags |= MSG_EOR;
785
786 return __sock_sendmsg(iocb, sock, msg, size);
787}
788
789static ssize_t sock_writev(struct file *file, const struct iovec *iov,
790 unsigned long nr_segs, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791{
792 struct msghdr msg;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800793 struct kiocb iocb;
794 struct sock_iocb siocb;
795 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800797 init_sync_kiocb(&iocb, NULL);
798 iocb.private = &siocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800800 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
801 if (-EIOCBQUEUED == ret)
802 ret = wait_on_sync_kiocb(&iocb);
803 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804}
805
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800806static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
807 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800809 struct sock_iocb siocb, *x;
810
811 if (pos != 0)
812 return -ESPIPE;
813 if (count == 0) /* Match SYS5 behaviour */
814 return 0;
815
816 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
817 if (!x)
818 return -ENOMEM;
819
820 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
821 &x->async_iov, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822}
823
824
825/*
826 * Atomic setting of ioctl hooks to avoid race
827 * with module unload.
828 */
829
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800830static DEFINE_MUTEX(br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
832
833void brioctl_set(int (*hook)(unsigned int, void __user *))
834{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800835 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800837 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
839EXPORT_SYMBOL(brioctl_set);
840
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800841static DEFINE_MUTEX(vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842static int (*vlan_ioctl_hook)(void __user *arg);
843
844void vlan_ioctl_set(int (*hook)(void __user *))
845{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800846 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800848 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849}
850EXPORT_SYMBOL(vlan_ioctl_set);
851
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800852static DEFINE_MUTEX(dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853static int (*dlci_ioctl_hook)(unsigned int, void __user *);
854
855void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
856{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800857 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800859 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860}
861EXPORT_SYMBOL(dlci_ioctl_set);
862
863/*
864 * With an ioctl, arg may well be a user mode pointer, but we don't know
865 * what to do with it - that's up to the protocol still.
866 */
867
868static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
869{
870 struct socket *sock;
871 void __user *argp = (void __user *)arg;
872 int pid, err;
873
Eric Dumazetb69aee02005-09-06 14:42:45 -0700874 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
876 err = dev_ioctl(cmd, argp);
877 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100878#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
880 err = dev_ioctl(cmd, argp);
881 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100882#endif /* CONFIG_WIRELESS_EXT */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 switch (cmd) {
884 case FIOSETOWN:
885 case SIOCSPGRP:
886 err = -EFAULT;
887 if (get_user(pid, (int __user *)argp))
888 break;
889 err = f_setown(sock->file, pid, 1);
890 break;
891 case FIOGETOWN:
892 case SIOCGPGRP:
893 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
894 break;
895 case SIOCGIFBR:
896 case SIOCSIFBR:
897 case SIOCBRADDBR:
898 case SIOCBRDELBR:
899 err = -ENOPKG;
900 if (!br_ioctl_hook)
901 request_module("bridge");
902
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800903 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 if (br_ioctl_hook)
905 err = br_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800906 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 break;
908 case SIOCGIFVLAN:
909 case SIOCSIFVLAN:
910 err = -ENOPKG;
911 if (!vlan_ioctl_hook)
912 request_module("8021q");
913
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800914 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 if (vlan_ioctl_hook)
916 err = vlan_ioctl_hook(argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800917 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 break;
919 case SIOCGIFDIVERT:
920 case SIOCSIFDIVERT:
921 /* Convert this to call through a hook */
922 err = divert_ioctl(cmd, argp);
923 break;
924 case SIOCADDDLCI:
925 case SIOCDELDLCI:
926 err = -ENOPKG;
927 if (!dlci_ioctl_hook)
928 request_module("dlci");
929
930 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800931 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800933 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 }
935 break;
936 default:
937 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800938
939 /*
940 * If this ioctl is unknown try to hand it down
941 * to the NIC driver.
942 */
943 if (err == -ENOIOCTLCMD)
944 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 break;
946 }
947 return err;
948}
949
950int sock_create_lite(int family, int type, int protocol, struct socket **res)
951{
952 int err;
953 struct socket *sock = NULL;
954
955 err = security_socket_create(family, type, protocol, 1);
956 if (err)
957 goto out;
958
959 sock = sock_alloc();
960 if (!sock) {
961 err = -ENOMEM;
962 goto out;
963 }
964
965 security_socket_post_create(sock, family, type, protocol, 1);
966 sock->type = type;
967out:
968 *res = sock;
969 return err;
970}
971
972/* No kernel lock held - perfect */
973static unsigned int sock_poll(struct file *file, poll_table * wait)
974{
975 struct socket *sock;
976
977 /*
978 * We can't return errors to poll, so it's either yes or no.
979 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700980 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 return sock->ops->poll(file, sock, wait);
982}
983
984static int sock_mmap(struct file * file, struct vm_area_struct * vma)
985{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700986 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
988 return sock->ops->mmap(file, sock, vma);
989}
990
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300991static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992{
993 /*
994 * It was possible the inode is NULL we were
995 * closing an unfinished socket.
996 */
997
998 if (!inode)
999 {
1000 printk(KERN_DEBUG "sock_close: NULL inode\n");
1001 return 0;
1002 }
1003 sock_fasync(-1, filp, 0);
1004 sock_release(SOCKET_I(inode));
1005 return 0;
1006}
1007
1008/*
1009 * Update the socket async list
1010 *
1011 * Fasync_list locking strategy.
1012 *
1013 * 1. fasync_list is modified only under process context socket lock
1014 * i.e. under semaphore.
1015 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1016 * or under socket lock.
1017 * 3. fasync_list can be used from softirq context, so that
1018 * modification under socket lock have to be enhanced with
1019 * write_lock_bh(&sk->sk_callback_lock).
1020 * --ANK (990710)
1021 */
1022
1023static int sock_fasync(int fd, struct file *filp, int on)
1024{
1025 struct fasync_struct *fa, *fna=NULL, **prev;
1026 struct socket *sock;
1027 struct sock *sk;
1028
1029 if (on)
1030 {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001031 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 if(fna==NULL)
1033 return -ENOMEM;
1034 }
1035
Eric Dumazetb69aee02005-09-06 14:42:45 -07001036 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037
1038 if ((sk=sock->sk) == NULL) {
1039 kfree(fna);
1040 return -EINVAL;
1041 }
1042
1043 lock_sock(sk);
1044
1045 prev=&(sock->fasync_list);
1046
1047 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1048 if (fa->fa_file==filp)
1049 break;
1050
1051 if(on)
1052 {
1053 if(fa!=NULL)
1054 {
1055 write_lock_bh(&sk->sk_callback_lock);
1056 fa->fa_fd=fd;
1057 write_unlock_bh(&sk->sk_callback_lock);
1058
1059 kfree(fna);
1060 goto out;
1061 }
1062 fna->fa_file=filp;
1063 fna->fa_fd=fd;
1064 fna->magic=FASYNC_MAGIC;
1065 fna->fa_next=sock->fasync_list;
1066 write_lock_bh(&sk->sk_callback_lock);
1067 sock->fasync_list=fna;
1068 write_unlock_bh(&sk->sk_callback_lock);
1069 }
1070 else
1071 {
1072 if (fa!=NULL)
1073 {
1074 write_lock_bh(&sk->sk_callback_lock);
1075 *prev=fa->fa_next;
1076 write_unlock_bh(&sk->sk_callback_lock);
1077 kfree(fa);
1078 }
1079 }
1080
1081out:
1082 release_sock(sock->sk);
1083 return 0;
1084}
1085
1086/* This function may be called only under socket lock or callback_lock */
1087
1088int sock_wake_async(struct socket *sock, int how, int band)
1089{
1090 if (!sock || !sock->fasync_list)
1091 return -1;
1092 switch (how)
1093 {
1094 case 1:
1095
1096 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1097 break;
1098 goto call_kill;
1099 case 2:
1100 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1101 break;
1102 /* fall through */
1103 case 0:
1104 call_kill:
1105 __kill_fasync(sock->fasync_list, SIGIO, band);
1106 break;
1107 case 3:
1108 __kill_fasync(sock->fasync_list, SIGURG, band);
1109 }
1110 return 0;
1111}
1112
1113static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1114{
1115 int err;
1116 struct socket *sock;
1117
1118 /*
1119 * Check protocol is in range
1120 */
1121 if (family < 0 || family >= NPROTO)
1122 return -EAFNOSUPPORT;
1123 if (type < 0 || type >= SOCK_MAX)
1124 return -EINVAL;
1125
1126 /* Compatibility.
1127
1128 This uglymoron is moved from INET layer to here to avoid
1129 deadlock in module load.
1130 */
1131 if (family == PF_INET && type == SOCK_PACKET) {
1132 static int warned;
1133 if (!warned) {
1134 warned = 1;
1135 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1136 }
1137 family = PF_PACKET;
1138 }
1139
1140 err = security_socket_create(family, type, protocol, kern);
1141 if (err)
1142 return err;
1143
1144#if defined(CONFIG_KMOD)
1145 /* Attempt to load a protocol module if the find failed.
1146 *
1147 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1148 * requested real, full-featured networking support upon configuration.
1149 * Otherwise module support will break!
1150 */
1151 if (net_families[family]==NULL)
1152 {
1153 request_module("net-pf-%d",family);
1154 }
1155#endif
1156
1157 net_family_read_lock();
1158 if (net_families[family] == NULL) {
1159 err = -EAFNOSUPPORT;
1160 goto out;
1161 }
1162
1163/*
1164 * Allocate the socket and allow the family to set things up. if
1165 * the protocol is 0, the family is instructed to select an appropriate
1166 * default.
1167 */
1168
1169 if (!(sock = sock_alloc())) {
1170 printk(KERN_WARNING "socket: no more sockets\n");
1171 err = -ENFILE; /* Not exactly a match, but its the
1172 closest posix thing */
1173 goto out;
1174 }
1175
1176 sock->type = type;
1177
1178 /*
1179 * We will call the ->create function, that possibly is in a loadable
1180 * module, so we have to bump that loadable module refcnt first.
1181 */
1182 err = -EAFNOSUPPORT;
1183 if (!try_module_get(net_families[family]->owner))
1184 goto out_release;
1185
Frank Filza79af592005-09-27 15:23:38 -07001186 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1187 sock->ops = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001189 }
1190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 /*
1192 * Now to bump the refcnt of the [loadable] module that owns this
1193 * socket at sock_release time we decrement its refcnt.
1194 */
1195 if (!try_module_get(sock->ops->owner)) {
1196 sock->ops = NULL;
1197 goto out_module_put;
1198 }
1199 /*
1200 * Now that we're done with the ->create function, the [loadable]
1201 * module can have its refcnt decremented
1202 */
1203 module_put(net_families[family]->owner);
1204 *res = sock;
1205 security_socket_post_create(sock, family, type, protocol, kern);
1206
1207out:
1208 net_family_read_unlock();
1209 return err;
1210out_module_put:
1211 module_put(net_families[family]->owner);
1212out_release:
1213 sock_release(sock);
1214 goto out;
1215}
1216
1217int sock_create(int family, int type, int protocol, struct socket **res)
1218{
1219 return __sock_create(family, type, protocol, res, 0);
1220}
1221
1222int sock_create_kern(int family, int type, int protocol, struct socket **res)
1223{
1224 return __sock_create(family, type, protocol, res, 1);
1225}
1226
1227asmlinkage long sys_socket(int family, int type, int protocol)
1228{
1229 int retval;
1230 struct socket *sock;
1231
1232 retval = sock_create(family, type, protocol, &sock);
1233 if (retval < 0)
1234 goto out;
1235
1236 retval = sock_map_fd(sock);
1237 if (retval < 0)
1238 goto out_release;
1239
1240out:
1241 /* It may be already another descriptor 8) Not kernel problem. */
1242 return retval;
1243
1244out_release:
1245 sock_release(sock);
1246 return retval;
1247}
1248
1249/*
1250 * Create a pair of connected sockets.
1251 */
1252
1253asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1254{
1255 struct socket *sock1, *sock2;
1256 int fd1, fd2, err;
1257
1258 /*
1259 * Obtain the first socket and check if the underlying protocol
1260 * supports the socketpair call.
1261 */
1262
1263 err = sock_create(family, type, protocol, &sock1);
1264 if (err < 0)
1265 goto out;
1266
1267 err = sock_create(family, type, protocol, &sock2);
1268 if (err < 0)
1269 goto out_release_1;
1270
1271 err = sock1->ops->socketpair(sock1, sock2);
1272 if (err < 0)
1273 goto out_release_both;
1274
1275 fd1 = fd2 = -1;
1276
1277 err = sock_map_fd(sock1);
1278 if (err < 0)
1279 goto out_release_both;
1280 fd1 = err;
1281
1282 err = sock_map_fd(sock2);
1283 if (err < 0)
1284 goto out_close_1;
1285 fd2 = err;
1286
1287 /* fd1 and fd2 may be already another descriptors.
1288 * Not kernel problem.
1289 */
1290
1291 err = put_user(fd1, &usockvec[0]);
1292 if (!err)
1293 err = put_user(fd2, &usockvec[1]);
1294 if (!err)
1295 return 0;
1296
1297 sys_close(fd2);
1298 sys_close(fd1);
1299 return err;
1300
1301out_close_1:
1302 sock_release(sock2);
1303 sys_close(fd1);
1304 return err;
1305
1306out_release_both:
1307 sock_release(sock2);
1308out_release_1:
1309 sock_release(sock1);
1310out:
1311 return err;
1312}
1313
1314
1315/*
1316 * Bind a name to a socket. Nothing much to do here since it's
1317 * the protocol's responsibility to handle the local address.
1318 *
1319 * We move the socket address to kernel space before we call
1320 * the protocol layer (having also checked the address is ok).
1321 */
1322
1323asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1324{
1325 struct socket *sock;
1326 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001327 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001329 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 {
1331 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1332 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001333 if (!err)
1334 err = sock->ops->bind(sock,
1335 (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001337 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 }
1339 return err;
1340}
1341
1342
1343/*
1344 * Perform a listen. Basically, we allow the protocol to do anything
1345 * necessary for a listen, and if that works, we mark the socket as
1346 * ready for listening.
1347 */
1348
1349int sysctl_somaxconn = SOMAXCONN;
1350
1351asmlinkage long sys_listen(int fd, int backlog)
1352{
1353 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001354 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001356 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 if ((unsigned) backlog > sysctl_somaxconn)
1358 backlog = sysctl_somaxconn;
1359
1360 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001361 if (!err)
1362 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001364 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 }
1366 return err;
1367}
1368
1369
1370/*
1371 * For accept, we attempt to create a new socket, set up the link
1372 * with the client, wake up the client, then return the new
1373 * connected fd. We collect the address of the connector in kernel
1374 * space and move it to user at the very end. This is unclean because
1375 * we open the socket then return an error.
1376 *
1377 * 1003.1g adds the ability to recvmsg() to query connection pending
1378 * status to recvmsg. We need to add that support in a way thats
1379 * clean when we restucture accept also.
1380 */
1381
1382asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1383{
1384 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001385 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001386 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 char address[MAX_SOCK_ADDR];
1388
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001389 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 if (!sock)
1391 goto out;
1392
1393 err = -ENFILE;
1394 if (!(newsock = sock_alloc()))
1395 goto out_put;
1396
1397 newsock->type = sock->type;
1398 newsock->ops = sock->ops;
1399
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 /*
1401 * We don't need try_module_get here, as the listening socket (sock)
1402 * has the protocol module (sock->ops->owner) held.
1403 */
1404 __module_get(newsock->ops->owner);
1405
David S. Miller39d8c1b2006-03-20 17:13:49 -08001406 newfd = sock_alloc_fd(&newfile);
1407 if (unlikely(newfd < 0)) {
1408 err = newfd;
1409 goto out_release;
1410 }
1411
1412 err = sock_attach_fd(newsock, newfile);
1413 if (err < 0)
1414 goto out_fd;
1415
Frank Filza79af592005-09-27 15:23:38 -07001416 err = security_socket_accept(sock, newsock);
1417 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001418 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001419
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1421 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001422 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
1424 if (upeer_sockaddr) {
1425 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1426 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001427 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 }
1429 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1430 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001431 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 }
1433
1434 /* File flags are not inherited via accept() unlike another OSes. */
1435
David S. Miller39d8c1b2006-03-20 17:13:49 -08001436 fd_install(newfd, newfile);
1437 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
1439 security_socket_post_accept(sock, newsock);
1440
1441out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001442 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443out:
1444 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001445out_fd:
1446 put_filp(newfile);
1447 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448out_release:
1449 sock_release(newsock);
1450 goto out_put;
1451}
1452
1453
1454/*
1455 * Attempt to connect to a socket with the server address. The address
1456 * is in user space so we verify it is OK and move it to kernel space.
1457 *
1458 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1459 * break bindings
1460 *
1461 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1462 * other SEQPACKET protocols that take time to connect() as it doesn't
1463 * include the -EINPROGRESS status for such sockets.
1464 */
1465
1466asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1467{
1468 struct socket *sock;
1469 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001470 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001472 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 if (!sock)
1474 goto out;
1475 err = move_addr_to_kernel(uservaddr, addrlen, address);
1476 if (err < 0)
1477 goto out_put;
1478
1479 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1480 if (err)
1481 goto out_put;
1482
1483 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1484 sock->file->f_flags);
1485out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001486 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487out:
1488 return err;
1489}
1490
1491/*
1492 * Get the local address ('name') of a socket object. Move the obtained
1493 * name to user space.
1494 */
1495
1496asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1497{
1498 struct socket *sock;
1499 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001500 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001502 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 if (!sock)
1504 goto out;
1505
1506 err = security_socket_getsockname(sock);
1507 if (err)
1508 goto out_put;
1509
1510 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1511 if (err)
1512 goto out_put;
1513 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1514
1515out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001516 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517out:
1518 return err;
1519}
1520
1521/*
1522 * Get the remote address ('name') of a socket object. Move the obtained
1523 * name to user space.
1524 */
1525
1526asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1527{
1528 struct socket *sock;
1529 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001530 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001532 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 err = security_socket_getpeername(sock);
1534 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001535 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 return err;
1537 }
1538
1539 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1540 if (!err)
1541 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001542 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 }
1544 return err;
1545}
1546
1547/*
1548 * Send a datagram to a given address. We move the address into kernel
1549 * space and check the user space data area is readable before invoking
1550 * the protocol.
1551 */
1552
1553asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1554 struct sockaddr __user *addr, int addr_len)
1555{
1556 struct socket *sock;
1557 char address[MAX_SOCK_ADDR];
1558 int err;
1559 struct msghdr msg;
1560 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001561 int fput_needed;
1562 struct file *sock_file;
1563
1564 sock_file = fget_light(fd, &fput_needed);
1565 if (!sock_file)
1566 return -EBADF;
1567
1568 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001570 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 iov.iov_base=buff;
1572 iov.iov_len=len;
1573 msg.msg_name=NULL;
1574 msg.msg_iov=&iov;
1575 msg.msg_iovlen=1;
1576 msg.msg_control=NULL;
1577 msg.msg_controllen=0;
1578 msg.msg_namelen=0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001579 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 err = move_addr_to_kernel(addr, addr_len, address);
1581 if (err < 0)
1582 goto out_put;
1583 msg.msg_name=address;
1584 msg.msg_namelen=addr_len;
1585 }
1586 if (sock->file->f_flags & O_NONBLOCK)
1587 flags |= MSG_DONTWAIT;
1588 msg.msg_flags = flags;
1589 err = sock_sendmsg(sock, &msg, len);
1590
1591out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001592 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 return err;
1594}
1595
1596/*
1597 * Send a datagram down a socket.
1598 */
1599
1600asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1601{
1602 return sys_sendto(fd, buff, len, flags, NULL, 0);
1603}
1604
1605/*
1606 * Receive a frame from the socket and optionally record the address of the
1607 * sender. We verify the buffers are writable and if needed move the
1608 * sender address from kernel to user space.
1609 */
1610
1611asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1612 struct sockaddr __user *addr, int __user *addr_len)
1613{
1614 struct socket *sock;
1615 struct iovec iov;
1616 struct msghdr msg;
1617 char address[MAX_SOCK_ADDR];
1618 int err,err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001619 struct file *sock_file;
1620 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001622 sock_file = fget_light(fd, &fput_needed);
1623 if (!sock_file)
1624 return -EBADF;
1625
1626 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 if (!sock)
1628 goto out;
1629
1630 msg.msg_control=NULL;
1631 msg.msg_controllen=0;
1632 msg.msg_iovlen=1;
1633 msg.msg_iov=&iov;
1634 iov.iov_len=size;
1635 iov.iov_base=ubuf;
1636 msg.msg_name=address;
1637 msg.msg_namelen=MAX_SOCK_ADDR;
1638 if (sock->file->f_flags & O_NONBLOCK)
1639 flags |= MSG_DONTWAIT;
1640 err=sock_recvmsg(sock, &msg, size, flags);
1641
1642 if(err >= 0 && addr != NULL)
1643 {
1644 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1645 if(err2<0)
1646 err=err2;
1647 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648out:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001649 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 return err;
1651}
1652
1653/*
1654 * Receive a datagram from a socket.
1655 */
1656
1657asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1658{
1659 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1660}
1661
1662/*
1663 * Set a socket option. Because we don't know the option lengths we have
1664 * to pass the user mode parameter for the protocols to sort out.
1665 */
1666
1667asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1668{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001669 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 struct socket *sock;
1671
1672 if (optlen < 0)
1673 return -EINVAL;
1674
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001675 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 {
1677 err = security_socket_setsockopt(sock,level,optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001678 if (err)
1679 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680
1681 if (level == SOL_SOCKET)
1682 err=sock_setsockopt(sock,level,optname,optval,optlen);
1683 else
1684 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001685out_put:
1686 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 }
1688 return err;
1689}
1690
1691/*
1692 * Get a socket option. Because we don't know the option lengths we have
1693 * to pass a user mode parameter for the protocols to sort out.
1694 */
1695
1696asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1697{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001698 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 struct socket *sock;
1700
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001701 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1702 err = security_socket_getsockopt(sock, level, optname);
1703 if (err)
1704 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
1706 if (level == SOL_SOCKET)
1707 err=sock_getsockopt(sock,level,optname,optval,optlen);
1708 else
1709 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001710out_put:
1711 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 }
1713 return err;
1714}
1715
1716
1717/*
1718 * Shutdown a socket.
1719 */
1720
1721asmlinkage long sys_shutdown(int fd, int how)
1722{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001723 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 struct socket *sock;
1725
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001726 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 {
1728 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001729 if (!err)
1730 err = sock->ops->shutdown(sock, how);
1731 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 }
1733 return err;
1734}
1735
1736/* A couple of helpful macros for getting the address of the 32/64 bit
1737 * fields which are the same type (int / unsigned) on our platforms.
1738 */
1739#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1740#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1741#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1742
1743
1744/*
1745 * BSD sendmsg interface
1746 */
1747
1748asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1749{
1750 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1751 struct socket *sock;
1752 char address[MAX_SOCK_ADDR];
1753 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001754 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1755 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1756 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 unsigned char *ctl_buf = ctl;
1758 struct msghdr msg_sys;
1759 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001760 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
1762 err = -EFAULT;
1763 if (MSG_CMSG_COMPAT & flags) {
1764 if (get_compat_msghdr(&msg_sys, msg_compat))
1765 return -EFAULT;
1766 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1767 return -EFAULT;
1768
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001769 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 if (!sock)
1771 goto out;
1772
1773 /* do not move before msg_sys is valid */
1774 err = -EMSGSIZE;
1775 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1776 goto out_put;
1777
1778 /* Check whether to allocate the iovec area*/
1779 err = -ENOMEM;
1780 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1781 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1782 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1783 if (!iov)
1784 goto out_put;
1785 }
1786
1787 /* This will also move the address data into kernel space */
1788 if (MSG_CMSG_COMPAT & flags) {
1789 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1790 } else
1791 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1792 if (err < 0)
1793 goto out_freeiov;
1794 total_len = err;
1795
1796 err = -ENOBUFS;
1797
1798 if (msg_sys.msg_controllen > INT_MAX)
1799 goto out_freeiov;
1800 ctl_len = msg_sys.msg_controllen;
1801 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Al Viro8920e8f2005-09-07 18:28:51 -07001802 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 if (err)
1804 goto out_freeiov;
1805 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001806 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 } else if (ctl_len) {
1808 if (ctl_len > sizeof(ctl))
1809 {
1810 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1811 if (ctl_buf == NULL)
1812 goto out_freeiov;
1813 }
1814 err = -EFAULT;
1815 /*
1816 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1817 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1818 * checking falls down on this.
1819 */
1820 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1821 goto out_freectl;
1822 msg_sys.msg_control = ctl_buf;
1823 }
1824 msg_sys.msg_flags = flags;
1825
1826 if (sock->file->f_flags & O_NONBLOCK)
1827 msg_sys.msg_flags |= MSG_DONTWAIT;
1828 err = sock_sendmsg(sock, &msg_sys, total_len);
1829
1830out_freectl:
1831 if (ctl_buf != ctl)
1832 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1833out_freeiov:
1834 if (iov != iovstack)
1835 sock_kfree_s(sock->sk, iov, iov_size);
1836out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001837 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838out:
1839 return err;
1840}
1841
1842/*
1843 * BSD recvmsg interface
1844 */
1845
1846asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1847{
1848 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1849 struct socket *sock;
1850 struct iovec iovstack[UIO_FASTIOV];
1851 struct iovec *iov=iovstack;
1852 struct msghdr msg_sys;
1853 unsigned long cmsg_ptr;
1854 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001855 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
1857 /* kernel mode address */
1858 char addr[MAX_SOCK_ADDR];
1859
1860 /* user mode address pointers */
1861 struct sockaddr __user *uaddr;
1862 int __user *uaddr_len;
1863
1864 if (MSG_CMSG_COMPAT & flags) {
1865 if (get_compat_msghdr(&msg_sys, msg_compat))
1866 return -EFAULT;
1867 } else
1868 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1869 return -EFAULT;
1870
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001871 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 if (!sock)
1873 goto out;
1874
1875 err = -EMSGSIZE;
1876 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1877 goto out_put;
1878
1879 /* Check whether to allocate the iovec area*/
1880 err = -ENOMEM;
1881 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1882 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1883 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1884 if (!iov)
1885 goto out_put;
1886 }
1887
1888 /*
1889 * Save the user-mode address (verify_iovec will change the
1890 * kernel msghdr to use the kernel address space)
1891 */
1892
1893 uaddr = (void __user *) msg_sys.msg_name;
1894 uaddr_len = COMPAT_NAMELEN(msg);
1895 if (MSG_CMSG_COMPAT & flags) {
1896 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1897 } else
1898 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1899 if (err < 0)
1900 goto out_freeiov;
1901 total_len=err;
1902
1903 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1904 msg_sys.msg_flags = 0;
1905 if (MSG_CMSG_COMPAT & flags)
1906 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1907
1908 if (sock->file->f_flags & O_NONBLOCK)
1909 flags |= MSG_DONTWAIT;
1910 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1911 if (err < 0)
1912 goto out_freeiov;
1913 len = err;
1914
1915 if (uaddr != NULL) {
1916 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1917 if (err < 0)
1918 goto out_freeiov;
1919 }
David S. Miller37f7f422005-09-16 16:51:01 -07001920 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1921 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 if (err)
1923 goto out_freeiov;
1924 if (MSG_CMSG_COMPAT & flags)
1925 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1926 &msg_compat->msg_controllen);
1927 else
1928 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1929 &msg->msg_controllen);
1930 if (err)
1931 goto out_freeiov;
1932 err = len;
1933
1934out_freeiov:
1935 if (iov != iovstack)
1936 sock_kfree_s(sock->sk, iov, iov_size);
1937out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001938 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939out:
1940 return err;
1941}
1942
1943#ifdef __ARCH_WANT_SYS_SOCKETCALL
1944
1945/* Argument list sizes for sys_socketcall */
1946#define AL(x) ((x) * sizeof(unsigned long))
1947static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1948 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1949 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1950#undef AL
1951
1952/*
1953 * System call vectors.
1954 *
1955 * Argument checking cleaned up. Saved 20% in size.
1956 * This function doesn't need to set the kernel lock because
1957 * it is set by the callees.
1958 */
1959
1960asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1961{
1962 unsigned long a[6];
1963 unsigned long a0,a1;
1964 int err;
1965
1966 if(call<1||call>SYS_RECVMSG)
1967 return -EINVAL;
1968
1969 /* copy_from_user should be SMP safe. */
1970 if (copy_from_user(a, args, nargs[call]))
1971 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001972
David Woodhouse4bcff1b2005-06-02 12:13:21 +01001973 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001974 if (err)
1975 return err;
1976
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 a0=a[0];
1978 a1=a[1];
1979
1980 switch(call)
1981 {
1982 case SYS_SOCKET:
1983 err = sys_socket(a0,a1,a[2]);
1984 break;
1985 case SYS_BIND:
1986 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1987 break;
1988 case SYS_CONNECT:
1989 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1990 break;
1991 case SYS_LISTEN:
1992 err = sys_listen(a0,a1);
1993 break;
1994 case SYS_ACCEPT:
1995 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1996 break;
1997 case SYS_GETSOCKNAME:
1998 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1999 break;
2000 case SYS_GETPEERNAME:
2001 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2002 break;
2003 case SYS_SOCKETPAIR:
2004 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2005 break;
2006 case SYS_SEND:
2007 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2008 break;
2009 case SYS_SENDTO:
2010 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2011 (struct sockaddr __user *)a[4], a[5]);
2012 break;
2013 case SYS_RECV:
2014 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2015 break;
2016 case SYS_RECVFROM:
2017 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2018 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2019 break;
2020 case SYS_SHUTDOWN:
2021 err = sys_shutdown(a0,a1);
2022 break;
2023 case SYS_SETSOCKOPT:
2024 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2025 break;
2026 case SYS_GETSOCKOPT:
2027 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2028 break;
2029 case SYS_SENDMSG:
2030 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2031 break;
2032 case SYS_RECVMSG:
2033 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2034 break;
2035 default:
2036 err = -EINVAL;
2037 break;
2038 }
2039 return err;
2040}
2041
2042#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2043
2044/*
2045 * This function is called by a protocol handler that wants to
2046 * advertise its address family, and have it linked into the
2047 * SOCKET module.
2048 */
2049
2050int sock_register(struct net_proto_family *ops)
2051{
2052 int err;
2053
2054 if (ops->family >= NPROTO) {
2055 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2056 return -ENOBUFS;
2057 }
2058 net_family_write_lock();
2059 err = -EEXIST;
2060 if (net_families[ops->family] == NULL) {
2061 net_families[ops->family]=ops;
2062 err = 0;
2063 }
2064 net_family_write_unlock();
2065 printk(KERN_INFO "NET: Registered protocol family %d\n",
2066 ops->family);
2067 return err;
2068}
2069
2070/*
2071 * This function is called by a protocol handler that wants to
2072 * remove its address family, and have it unlinked from the
2073 * SOCKET module.
2074 */
2075
2076int sock_unregister(int family)
2077{
2078 if (family < 0 || family >= NPROTO)
2079 return -1;
2080
2081 net_family_write_lock();
2082 net_families[family]=NULL;
2083 net_family_write_unlock();
2084 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2085 family);
2086 return 0;
2087}
2088
Andi Kleen77d76ea2005-12-22 12:43:42 -08002089static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090{
2091 /*
2092 * Initialize sock SLAB cache.
2093 */
2094
2095 sk_init();
2096
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 /*
2098 * Initialize skbuff SLAB cache
2099 */
2100 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101
2102 /*
2103 * Initialize the protocols module.
2104 */
2105
2106 init_inodecache();
2107 register_filesystem(&sock_fs_type);
2108 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002109
2110 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 */
2112
2113#ifdef CONFIG_NETFILTER
2114 netfilter_init();
2115#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002116
2117 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118}
2119
Andi Kleen77d76ea2005-12-22 12:43:42 -08002120core_initcall(sock_init); /* early initcall */
2121
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122#ifdef CONFIG_PROC_FS
2123void socket_seq_show(struct seq_file *seq)
2124{
2125 int cpu;
2126 int counter = 0;
2127
Eric Dumazet88a2a4ac2006-02-04 23:27:36 -08002128 for_each_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 counter += per_cpu(sockets_in_use, cpu);
2130
2131 /* It can be negative, by the way. 8) */
2132 if (counter < 0)
2133 counter = 0;
2134
2135 seq_printf(seq, "sockets: used %d\n", counter);
2136}
2137#endif /* CONFIG_PROC_FS */
2138
2139/* ABI emulation layers need these two */
2140EXPORT_SYMBOL(move_addr_to_kernel);
2141EXPORT_SYMBOL(move_addr_to_user);
2142EXPORT_SYMBOL(sock_create);
2143EXPORT_SYMBOL(sock_create_kern);
2144EXPORT_SYMBOL(sock_create_lite);
2145EXPORT_SYMBOL(sock_map_fd);
2146EXPORT_SYMBOL(sock_recvmsg);
2147EXPORT_SYMBOL(sock_register);
2148EXPORT_SYMBOL(sock_release);
2149EXPORT_SYMBOL(sock_sendmsg);
2150EXPORT_SYMBOL(sock_unregister);
2151EXPORT_SYMBOL(sock_wake_async);
2152EXPORT_SYMBOL(sockfd_lookup);
2153EXPORT_SYMBOL(kernel_sendmsg);
2154EXPORT_SYMBOL(kernel_recvmsg);