blob: fcd77eac0ccff184294eab4888e3ff61ad35d168 [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
122
123/*
124 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
125 * in the operation structures but are done directly via the socketcall() multiplexor.
126 */
127
128static struct file_operations socket_file_ops = {
129 .owner = THIS_MODULE,
130 .llseek = no_llseek,
131 .aio_read = sock_aio_read,
132 .aio_write = sock_aio_write,
133 .poll = sock_poll,
134 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800135#ifdef CONFIG_COMPAT
136 .compat_ioctl = compat_sock_ioctl,
137#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138 .mmap = sock_mmap,
139 .open = sock_no_open, /* special open code to disallow open via /proc */
140 .release = sock_close,
141 .fasync = sock_fasync,
142 .readv = sock_readv,
143 .writev = sock_writev,
144 .sendpage = sock_sendpage
145};
146
147/*
148 * The protocol list. Each protocol is registered in here.
149 */
150
151static struct net_proto_family *net_families[NPROTO];
152
153#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
154static atomic_t net_family_lockct = ATOMIC_INIT(0);
155static DEFINE_SPINLOCK(net_family_lock);
156
157/* The strategy is: modifications net_family vector are short, do not
158 sleep and veeery rare, but read access should be free of any exclusive
159 locks.
160 */
161
162static void net_family_write_lock(void)
163{
164 spin_lock(&net_family_lock);
165 while (atomic_read(&net_family_lockct) != 0) {
166 spin_unlock(&net_family_lock);
167
168 yield();
169
170 spin_lock(&net_family_lock);
171 }
172}
173
174static __inline__ void net_family_write_unlock(void)
175{
176 spin_unlock(&net_family_lock);
177}
178
179static __inline__ void net_family_read_lock(void)
180{
181 atomic_inc(&net_family_lockct);
182 spin_unlock_wait(&net_family_lock);
183}
184
185static __inline__ void net_family_read_unlock(void)
186{
187 atomic_dec(&net_family_lockct);
188}
189
190#else
191#define net_family_write_lock() do { } while(0)
192#define net_family_write_unlock() do { } while(0)
193#define net_family_read_lock() do { } while(0)
194#define net_family_read_unlock() do { } while(0)
195#endif
196
197
198/*
199 * Statistics counters of the socket lists
200 */
201
202static DEFINE_PER_CPU(int, sockets_in_use) = 0;
203
204/*
205 * Support routines. Move socket addresses back and forth across the kernel/user
206 * divide and look after the messy bits.
207 */
208
209#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
210 16 for IP, 16 for IPX,
211 24 for IPv6,
212 about 80 for AX.25
213 must be at least one bigger than
214 the AF_UNIX size (see net/unix/af_unix.c
215 :unix_mkname()).
216 */
217
218/**
219 * move_addr_to_kernel - copy a socket address into kernel space
220 * @uaddr: Address in user space
221 * @kaddr: Address in kernel space
222 * @ulen: Length in user space
223 *
224 * The address is copied into kernel space. If the provided address is
225 * too long an error code of -EINVAL is returned. If the copy gives
226 * invalid addresses -EFAULT is returned. On a success 0 is returned.
227 */
228
229int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
230{
231 if(ulen<0||ulen>MAX_SOCK_ADDR)
232 return -EINVAL;
233 if(ulen==0)
234 return 0;
235 if(copy_from_user(kaddr,uaddr,ulen))
236 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100237 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238}
239
240/**
241 * move_addr_to_user - copy an address to user space
242 * @kaddr: kernel space address
243 * @klen: length of address in kernel
244 * @uaddr: user space address
245 * @ulen: pointer to user length field
246 *
247 * The value pointed to by ulen on entry is the buffer length available.
248 * This is overwritten with the buffer space used. -EINVAL is returned
249 * if an overlong buffer is specified or a negative buffer size. -EFAULT
250 * is returned if either the buffer or the length field are not
251 * accessible.
252 * After copying the data up to the limit the user specifies, the true
253 * length of the data is written over the length limit the user
254 * specified. Zero is returned for a success.
255 */
256
257int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
258{
259 int err;
260 int len;
261
262 if((err=get_user(len, ulen)))
263 return err;
264 if(len>klen)
265 len=klen;
266 if(len<0 || len> MAX_SOCK_ADDR)
267 return -EINVAL;
268 if(len)
269 {
270 if(copy_to_user(uaddr,kaddr,len))
271 return -EFAULT;
272 }
273 /*
274 * "fromlen shall refer to the value before truncation.."
275 * 1003.1g
276 */
277 return __put_user(klen, ulen);
278}
279
280#define SOCKFS_MAGIC 0x534F434B
281
Eric Dumazetba899662005-08-26 12:05:31 -0700282static kmem_cache_t * sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283
284static struct inode *sock_alloc_inode(struct super_block *sb)
285{
286 struct socket_alloc *ei;
287 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
288 if (!ei)
289 return NULL;
290 init_waitqueue_head(&ei->socket.wait);
291
292 ei->socket.fasync_list = NULL;
293 ei->socket.state = SS_UNCONNECTED;
294 ei->socket.flags = 0;
295 ei->socket.ops = NULL;
296 ei->socket.sk = NULL;
297 ei->socket.file = NULL;
298 ei->socket.flags = 0;
299
300 return &ei->vfs_inode;
301}
302
303static void sock_destroy_inode(struct inode *inode)
304{
305 kmem_cache_free(sock_inode_cachep,
306 container_of(inode, struct socket_alloc, vfs_inode));
307}
308
309static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
310{
311 struct socket_alloc *ei = (struct socket_alloc *) foo;
312
313 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
314 SLAB_CTOR_CONSTRUCTOR)
315 inode_init_once(&ei->vfs_inode);
316}
317
318static int init_inodecache(void)
319{
320 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
321 sizeof(struct socket_alloc),
Paul Jacksonfffb60f2006-03-24 03:16:06 -0800322 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
323 SLAB_MEM_SPREAD),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 init_once, NULL);
325 if (sock_inode_cachep == NULL)
326 return -ENOMEM;
327 return 0;
328}
329
330static struct super_operations sockfs_ops = {
331 .alloc_inode = sock_alloc_inode,
332 .destroy_inode =sock_destroy_inode,
333 .statfs = simple_statfs,
334};
335
336static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
337 int flags, const char *dev_name, void *data)
338{
339 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
340}
341
Eric Dumazetba899662005-08-26 12:05:31 -0700342static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
344static struct file_system_type sock_fs_type = {
345 .name = "sockfs",
346 .get_sb = sockfs_get_sb,
347 .kill_sb = kill_anon_super,
348};
349static int sockfs_delete_dentry(struct dentry *dentry)
350{
351 return 1;
352}
353static struct dentry_operations sockfs_dentry_operations = {
354 .d_delete = sockfs_delete_dentry,
355};
356
357/*
358 * Obtains the first available file descriptor and sets it up for use.
359 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800360 * These functions create file structures and maps them to fd space
361 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 * and file struct implicitly stored in sock->file.
363 * Note that another thread may close file descriptor before we return
364 * from this function. We use the fact that now we do not refer
365 * to socket after mapping. If one day we will need it, this
366 * function will increment ref. count on file by 1.
367 *
368 * In any case returned fd MAY BE not valid!
369 * This race condition is unavoidable
370 * with shared fd spaces, we cannot solve it inside kernel,
371 * but we take care of internal coherence yet.
372 */
373
David S. Miller39d8c1b2006-03-20 17:13:49 -0800374static int sock_alloc_fd(struct file **filep)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375{
376 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800377
378 fd = get_unused_fd();
379 if (likely(fd >= 0)) {
380 struct file *file = get_empty_filp();
381
382 *filep = file;
383 if (unlikely(!file)) {
384 put_unused_fd(fd);
385 return -ENFILE;
386 }
387 } else
388 *filep = NULL;
389 return fd;
390}
391
392static int sock_attach_fd(struct socket *sock, struct file *file)
393{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 struct qstr this;
395 char name[32];
396
David S. Miller39d8c1b2006-03-20 17:13:49 -0800397 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
398 this.name = name;
399 this.hash = SOCK_INODE(sock)->i_ino;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
402 if (unlikely(!file->f_dentry))
403 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
David S. Miller39d8c1b2006-03-20 17:13:49 -0800405 file->f_dentry->d_op = &sockfs_dentry_operations;
406 d_add(file->f_dentry, SOCK_INODE(sock));
407 file->f_vfsmnt = mntget(sock_mnt);
408 file->f_mapping = file->f_dentry->d_inode->i_mapping;
409
410 sock->file = file;
411 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
412 file->f_mode = FMODE_READ | FMODE_WRITE;
413 file->f_flags = O_RDWR;
414 file->f_pos = 0;
415 file->private_data = sock;
416
417 return 0;
418}
419
420int sock_map_fd(struct socket *sock)
421{
422 struct file *newfile;
423 int fd = sock_alloc_fd(&newfile);
424
425 if (likely(fd >= 0)) {
426 int err = sock_attach_fd(sock, newfile);
427
428 if (unlikely(err < 0)) {
429 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800431 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800433 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 return fd;
436}
437
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800438static struct socket *sock_from_file(struct file *file, int *err)
439{
440 struct inode *inode;
441 struct socket *sock;
442
443 if (file->f_op == &socket_file_ops)
444 return file->private_data; /* set in sock_map_fd */
445
446 inode = file->f_dentry->d_inode;
447 if (!S_ISSOCK(inode->i_mode)) {
448 *err = -ENOTSOCK;
449 return NULL;
450 }
451
452 sock = SOCKET_I(inode);
453 if (sock->file != file) {
454 printk(KERN_ERR "socki_lookup: socket file changed!\n");
455 sock->file = file;
456 }
457 return sock;
458}
459
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460/**
461 * sockfd_lookup - Go from a file number to its socket slot
462 * @fd: file handle
463 * @err: pointer to an error code return
464 *
465 * The file handle passed in is locked and the socket it is bound
466 * too is returned. If an error occurs the err pointer is overwritten
467 * with a negative errno code and NULL is returned. The function checks
468 * for both invalid handles and passing a handle which is not a socket.
469 *
470 * On a success the socket object pointer is returned.
471 */
472
473struct socket *sockfd_lookup(int fd, int *err)
474{
475 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 struct socket *sock;
477
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800478 if (!(file = fget(fd))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 *err = -EBADF;
480 return NULL;
481 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800482 sock = sock_from_file(file, err);
483 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 return sock;
486}
487
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800488static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
489{
490 struct file *file;
491 struct socket *sock;
492
493 file = fget_light(fd, fput_needed);
494 if (file) {
495 sock = sock_from_file(file, err);
496 if (sock)
497 return sock;
498 fput_light(file, *fput_needed);
499 }
500 return NULL;
501}
502
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503/**
504 * sock_alloc - allocate a socket
505 *
506 * Allocate a new inode and socket object. The two are bound together
507 * and initialised. The socket is then returned. If we are out of inodes
508 * NULL is returned.
509 */
510
511static struct socket *sock_alloc(void)
512{
513 struct inode * inode;
514 struct socket * sock;
515
516 inode = new_inode(sock_mnt->mnt_sb);
517 if (!inode)
518 return NULL;
519
520 sock = SOCKET_I(inode);
521
522 inode->i_mode = S_IFSOCK|S_IRWXUGO;
523 inode->i_uid = current->fsuid;
524 inode->i_gid = current->fsgid;
525
526 get_cpu_var(sockets_in_use)++;
527 put_cpu_var(sockets_in_use);
528 return sock;
529}
530
531/*
532 * In theory you can't get an open on this inode, but /proc provides
533 * a back door. Remember to keep it shut otherwise you'll let the
534 * creepy crawlies in.
535 */
536
537static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
538{
539 return -ENXIO;
540}
541
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800542const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 .owner = THIS_MODULE,
544 .open = sock_no_open,
545};
546
547/**
548 * sock_release - close a socket
549 * @sock: socket to close
550 *
551 * The socket is released from the protocol stack if it has a release
552 * callback, and the inode is then released if the socket is bound to
553 * an inode not a file.
554 */
555
556void sock_release(struct socket *sock)
557{
558 if (sock->ops) {
559 struct module *owner = sock->ops->owner;
560
561 sock->ops->release(sock);
562 sock->ops = NULL;
563 module_put(owner);
564 }
565
566 if (sock->fasync_list)
567 printk(KERN_ERR "sock_release: fasync list not empty!\n");
568
569 get_cpu_var(sockets_in_use)--;
570 put_cpu_var(sockets_in_use);
571 if (!sock->file) {
572 iput(SOCK_INODE(sock));
573 return;
574 }
575 sock->file=NULL;
576}
577
578static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
579 struct msghdr *msg, size_t size)
580{
581 struct sock_iocb *si = kiocb_to_siocb(iocb);
582 int err;
583
584 si->sock = sock;
585 si->scm = NULL;
586 si->msg = msg;
587 si->size = size;
588
589 err = security_socket_sendmsg(sock, msg, size);
590 if (err)
591 return err;
592
593 return sock->ops->sendmsg(iocb, sock, msg, size);
594}
595
596int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
597{
598 struct kiocb iocb;
599 struct sock_iocb siocb;
600 int ret;
601
602 init_sync_kiocb(&iocb, NULL);
603 iocb.private = &siocb;
604 ret = __sock_sendmsg(&iocb, sock, msg, size);
605 if (-EIOCBQUEUED == ret)
606 ret = wait_on_sync_kiocb(&iocb);
607 return ret;
608}
609
610int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
611 struct kvec *vec, size_t num, size_t size)
612{
613 mm_segment_t oldfs = get_fs();
614 int result;
615
616 set_fs(KERNEL_DS);
617 /*
618 * the following is safe, since for compiler definitions of kvec and
619 * iovec are identical, yielding the same in-core layout and alignment
620 */
621 msg->msg_iov = (struct iovec *)vec,
622 msg->msg_iovlen = num;
623 result = sock_sendmsg(sock, msg, size);
624 set_fs(oldfs);
625 return result;
626}
627
628static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
629 struct msghdr *msg, size_t size, int flags)
630{
631 int err;
632 struct sock_iocb *si = kiocb_to_siocb(iocb);
633
634 si->sock = sock;
635 si->scm = NULL;
636 si->msg = msg;
637 si->size = size;
638 si->flags = flags;
639
640 err = security_socket_recvmsg(sock, msg, size, flags);
641 if (err)
642 return err;
643
644 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
645}
646
647int sock_recvmsg(struct socket *sock, struct msghdr *msg,
648 size_t size, int flags)
649{
650 struct kiocb iocb;
651 struct sock_iocb siocb;
652 int ret;
653
654 init_sync_kiocb(&iocb, NULL);
655 iocb.private = &siocb;
656 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
657 if (-EIOCBQUEUED == ret)
658 ret = wait_on_sync_kiocb(&iocb);
659 return ret;
660}
661
662int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
663 struct kvec *vec, size_t num,
664 size_t size, int flags)
665{
666 mm_segment_t oldfs = get_fs();
667 int result;
668
669 set_fs(KERNEL_DS);
670 /*
671 * the following is safe, since for compiler definitions of kvec and
672 * iovec are identical, yielding the same in-core layout and alignment
673 */
674 msg->msg_iov = (struct iovec *)vec,
675 msg->msg_iovlen = num;
676 result = sock_recvmsg(sock, msg, size, flags);
677 set_fs(oldfs);
678 return result;
679}
680
681static void sock_aio_dtor(struct kiocb *iocb)
682{
683 kfree(iocb->private);
684}
685
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300686static ssize_t sock_sendpage(struct file *file, struct page *page,
687 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688{
689 struct socket *sock;
690 int flags;
691
Eric Dumazetb69aee02005-09-06 14:42:45 -0700692 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
694 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
695 if (more)
696 flags |= MSG_MORE;
697
698 return sock->ops->sendpage(sock, page, offset, size, flags);
699}
700
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800701static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
702 char __user *ubuf, size_t size, struct sock_iocb *siocb)
703{
704 if (!is_sync_kiocb(iocb)) {
705 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
706 if (!siocb)
707 return NULL;
708 iocb->ki_dtor = sock_aio_dtor;
709 }
710
711 siocb->kiocb = iocb;
712 siocb->async_iov.iov_base = ubuf;
713 siocb->async_iov.iov_len = size;
714
715 iocb->private = siocb;
716 return siocb;
717}
718
719static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
720 struct file *file, struct iovec *iov, unsigned long nr_segs)
721{
722 struct socket *sock = file->private_data;
723 size_t size = 0;
724 int i;
725
726 for (i = 0 ; i < nr_segs ; i++)
727 size += iov[i].iov_len;
728
729 msg->msg_name = NULL;
730 msg->msg_namelen = 0;
731 msg->msg_control = NULL;
732 msg->msg_controllen = 0;
733 msg->msg_iov = (struct iovec *) iov;
734 msg->msg_iovlen = nr_segs;
735 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
736
737 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
738}
739
740static ssize_t sock_readv(struct file *file, const struct iovec *iov,
741 unsigned long nr_segs, loff_t *ppos)
742{
743 struct kiocb iocb;
744 struct sock_iocb siocb;
745 struct msghdr msg;
746 int ret;
747
748 init_sync_kiocb(&iocb, NULL);
749 iocb.private = &siocb;
750
751 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
752 if (-EIOCBQUEUED == ret)
753 ret = wait_on_sync_kiocb(&iocb);
754 return ret;
755}
756
757static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
758 size_t count, loff_t pos)
759{
760 struct sock_iocb siocb, *x;
761
762 if (pos != 0)
763 return -ESPIPE;
764 if (count == 0) /* Match SYS5 behaviour */
765 return 0;
766
767 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
768 if (!x)
769 return -ENOMEM;
770 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
771 &x->async_iov, 1);
772}
773
774static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
775 struct file *file, struct iovec *iov, unsigned long nr_segs)
776{
777 struct socket *sock = file->private_data;
778 size_t size = 0;
779 int i;
780
781 for (i = 0 ; i < nr_segs ; i++)
782 size += iov[i].iov_len;
783
784 msg->msg_name = NULL;
785 msg->msg_namelen = 0;
786 msg->msg_control = NULL;
787 msg->msg_controllen = 0;
788 msg->msg_iov = (struct iovec *) iov;
789 msg->msg_iovlen = nr_segs;
790 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
791 if (sock->type == SOCK_SEQPACKET)
792 msg->msg_flags |= MSG_EOR;
793
794 return __sock_sendmsg(iocb, sock, msg, size);
795}
796
797static ssize_t sock_writev(struct file *file, const struct iovec *iov,
798 unsigned long nr_segs, loff_t *ppos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799{
800 struct msghdr msg;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800801 struct kiocb iocb;
802 struct sock_iocb siocb;
803 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800805 init_sync_kiocb(&iocb, NULL);
806 iocb.private = &siocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800808 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
809 if (-EIOCBQUEUED == ret)
810 ret = wait_on_sync_kiocb(&iocb);
811 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812}
813
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800814static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
815 size_t count, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800817 struct sock_iocb siocb, *x;
818
819 if (pos != 0)
820 return -ESPIPE;
821 if (count == 0) /* Match SYS5 behaviour */
822 return 0;
823
824 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
825 if (!x)
826 return -ENOMEM;
827
828 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
829 &x->async_iov, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830}
831
832
833/*
834 * Atomic setting of ioctl hooks to avoid race
835 * with module unload.
836 */
837
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800838static DEFINE_MUTEX(br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
840
841void brioctl_set(int (*hook)(unsigned int, void __user *))
842{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800843 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800845 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846}
847EXPORT_SYMBOL(brioctl_set);
848
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800849static DEFINE_MUTEX(vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850static int (*vlan_ioctl_hook)(void __user *arg);
851
852void vlan_ioctl_set(int (*hook)(void __user *))
853{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800854 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800856 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857}
858EXPORT_SYMBOL(vlan_ioctl_set);
859
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800860static DEFINE_MUTEX(dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861static int (*dlci_ioctl_hook)(unsigned int, void __user *);
862
863void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
864{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800865 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800867 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868}
869EXPORT_SYMBOL(dlci_ioctl_set);
870
871/*
872 * With an ioctl, arg may well be a user mode pointer, but we don't know
873 * what to do with it - that's up to the protocol still.
874 */
875
876static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
877{
878 struct socket *sock;
879 void __user *argp = (void __user *)arg;
880 int pid, err;
881
Eric Dumazetb69aee02005-09-06 14:42:45 -0700882 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
884 err = dev_ioctl(cmd, argp);
885 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100886#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
888 err = dev_ioctl(cmd, argp);
889 } else
Adrian Bunkd86b5e02006-01-21 00:46:55 +0100890#endif /* CONFIG_WIRELESS_EXT */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 switch (cmd) {
892 case FIOSETOWN:
893 case SIOCSPGRP:
894 err = -EFAULT;
895 if (get_user(pid, (int __user *)argp))
896 break;
897 err = f_setown(sock->file, pid, 1);
898 break;
899 case FIOGETOWN:
900 case SIOCGPGRP:
901 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
902 break;
903 case SIOCGIFBR:
904 case SIOCSIFBR:
905 case SIOCBRADDBR:
906 case SIOCBRDELBR:
907 err = -ENOPKG;
908 if (!br_ioctl_hook)
909 request_module("bridge");
910
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800911 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 if (br_ioctl_hook)
913 err = br_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800914 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 break;
916 case SIOCGIFVLAN:
917 case SIOCSIFVLAN:
918 err = -ENOPKG;
919 if (!vlan_ioctl_hook)
920 request_module("8021q");
921
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800922 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 if (vlan_ioctl_hook)
924 err = vlan_ioctl_hook(argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800925 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 break;
927 case SIOCGIFDIVERT:
928 case SIOCSIFDIVERT:
929 /* Convert this to call through a hook */
930 err = divert_ioctl(cmd, argp);
931 break;
932 case SIOCADDDLCI:
933 case SIOCDELDLCI:
934 err = -ENOPKG;
935 if (!dlci_ioctl_hook)
936 request_module("dlci");
937
938 if (dlci_ioctl_hook) {
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800939 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 err = dlci_ioctl_hook(cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800941 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 }
943 break;
944 default:
945 err = sock->ops->ioctl(sock, cmd, arg);
Christoph Hellwigb5e5fa52006-01-03 14:18:33 -0800946
947 /*
948 * If this ioctl is unknown try to hand it down
949 * to the NIC driver.
950 */
951 if (err == -ENOIOCTLCMD)
952 err = dev_ioctl(cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 break;
954 }
955 return err;
956}
957
958int sock_create_lite(int family, int type, int protocol, struct socket **res)
959{
960 int err;
961 struct socket *sock = NULL;
962
963 err = security_socket_create(family, type, protocol, 1);
964 if (err)
965 goto out;
966
967 sock = sock_alloc();
968 if (!sock) {
969 err = -ENOMEM;
970 goto out;
971 }
972
973 security_socket_post_create(sock, family, type, protocol, 1);
974 sock->type = type;
975out:
976 *res = sock;
977 return err;
978}
979
980/* No kernel lock held - perfect */
981static unsigned int sock_poll(struct file *file, poll_table * wait)
982{
983 struct socket *sock;
984
985 /*
986 * We can't return errors to poll, so it's either yes or no.
987 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700988 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 return sock->ops->poll(file, sock, wait);
990}
991
992static int sock_mmap(struct file * file, struct vm_area_struct * vma)
993{
Eric Dumazetb69aee02005-09-06 14:42:45 -0700994 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
996 return sock->ops->mmap(file, sock, vma);
997}
998
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300999static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000{
1001 /*
1002 * It was possible the inode is NULL we were
1003 * closing an unfinished socket.
1004 */
1005
1006 if (!inode)
1007 {
1008 printk(KERN_DEBUG "sock_close: NULL inode\n");
1009 return 0;
1010 }
1011 sock_fasync(-1, filp, 0);
1012 sock_release(SOCKET_I(inode));
1013 return 0;
1014}
1015
1016/*
1017 * Update the socket async list
1018 *
1019 * Fasync_list locking strategy.
1020 *
1021 * 1. fasync_list is modified only under process context socket lock
1022 * i.e. under semaphore.
1023 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1024 * or under socket lock.
1025 * 3. fasync_list can be used from softirq context, so that
1026 * modification under socket lock have to be enhanced with
1027 * write_lock_bh(&sk->sk_callback_lock).
1028 * --ANK (990710)
1029 */
1030
1031static int sock_fasync(int fd, struct file *filp, int on)
1032{
1033 struct fasync_struct *fa, *fna=NULL, **prev;
1034 struct socket *sock;
1035 struct sock *sk;
1036
1037 if (on)
1038 {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001039 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 if(fna==NULL)
1041 return -ENOMEM;
1042 }
1043
Eric Dumazetb69aee02005-09-06 14:42:45 -07001044 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045
1046 if ((sk=sock->sk) == NULL) {
1047 kfree(fna);
1048 return -EINVAL;
1049 }
1050
1051 lock_sock(sk);
1052
1053 prev=&(sock->fasync_list);
1054
1055 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1056 if (fa->fa_file==filp)
1057 break;
1058
1059 if(on)
1060 {
1061 if(fa!=NULL)
1062 {
1063 write_lock_bh(&sk->sk_callback_lock);
1064 fa->fa_fd=fd;
1065 write_unlock_bh(&sk->sk_callback_lock);
1066
1067 kfree(fna);
1068 goto out;
1069 }
1070 fna->fa_file=filp;
1071 fna->fa_fd=fd;
1072 fna->magic=FASYNC_MAGIC;
1073 fna->fa_next=sock->fasync_list;
1074 write_lock_bh(&sk->sk_callback_lock);
1075 sock->fasync_list=fna;
1076 write_unlock_bh(&sk->sk_callback_lock);
1077 }
1078 else
1079 {
1080 if (fa!=NULL)
1081 {
1082 write_lock_bh(&sk->sk_callback_lock);
1083 *prev=fa->fa_next;
1084 write_unlock_bh(&sk->sk_callback_lock);
1085 kfree(fa);
1086 }
1087 }
1088
1089out:
1090 release_sock(sock->sk);
1091 return 0;
1092}
1093
1094/* This function may be called only under socket lock or callback_lock */
1095
1096int sock_wake_async(struct socket *sock, int how, int band)
1097{
1098 if (!sock || !sock->fasync_list)
1099 return -1;
1100 switch (how)
1101 {
1102 case 1:
1103
1104 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1105 break;
1106 goto call_kill;
1107 case 2:
1108 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1109 break;
1110 /* fall through */
1111 case 0:
1112 call_kill:
1113 __kill_fasync(sock->fasync_list, SIGIO, band);
1114 break;
1115 case 3:
1116 __kill_fasync(sock->fasync_list, SIGURG, band);
1117 }
1118 return 0;
1119}
1120
1121static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1122{
1123 int err;
1124 struct socket *sock;
1125
1126 /*
1127 * Check protocol is in range
1128 */
1129 if (family < 0 || family >= NPROTO)
1130 return -EAFNOSUPPORT;
1131 if (type < 0 || type >= SOCK_MAX)
1132 return -EINVAL;
1133
1134 /* Compatibility.
1135
1136 This uglymoron is moved from INET layer to here to avoid
1137 deadlock in module load.
1138 */
1139 if (family == PF_INET && type == SOCK_PACKET) {
1140 static int warned;
1141 if (!warned) {
1142 warned = 1;
1143 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1144 }
1145 family = PF_PACKET;
1146 }
1147
1148 err = security_socket_create(family, type, protocol, kern);
1149 if (err)
1150 return err;
1151
1152#if defined(CONFIG_KMOD)
1153 /* Attempt to load a protocol module if the find failed.
1154 *
1155 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1156 * requested real, full-featured networking support upon configuration.
1157 * Otherwise module support will break!
1158 */
1159 if (net_families[family]==NULL)
1160 {
1161 request_module("net-pf-%d",family);
1162 }
1163#endif
1164
1165 net_family_read_lock();
1166 if (net_families[family] == NULL) {
1167 err = -EAFNOSUPPORT;
1168 goto out;
1169 }
1170
1171/*
1172 * Allocate the socket and allow the family to set things up. if
1173 * the protocol is 0, the family is instructed to select an appropriate
1174 * default.
1175 */
1176
1177 if (!(sock = sock_alloc())) {
1178 printk(KERN_WARNING "socket: no more sockets\n");
1179 err = -ENFILE; /* Not exactly a match, but its the
1180 closest posix thing */
1181 goto out;
1182 }
1183
1184 sock->type = type;
1185
1186 /*
1187 * We will call the ->create function, that possibly is in a loadable
1188 * module, so we have to bump that loadable module refcnt first.
1189 */
1190 err = -EAFNOSUPPORT;
1191 if (!try_module_get(net_families[family]->owner))
1192 goto out_release;
1193
Frank Filza79af592005-09-27 15:23:38 -07001194 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1195 sock->ops = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001197 }
1198
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 /*
1200 * Now to bump the refcnt of the [loadable] module that owns this
1201 * socket at sock_release time we decrement its refcnt.
1202 */
1203 if (!try_module_get(sock->ops->owner)) {
1204 sock->ops = NULL;
1205 goto out_module_put;
1206 }
1207 /*
1208 * Now that we're done with the ->create function, the [loadable]
1209 * module can have its refcnt decremented
1210 */
1211 module_put(net_families[family]->owner);
1212 *res = sock;
1213 security_socket_post_create(sock, family, type, protocol, kern);
1214
1215out:
1216 net_family_read_unlock();
1217 return err;
1218out_module_put:
1219 module_put(net_families[family]->owner);
1220out_release:
1221 sock_release(sock);
1222 goto out;
1223}
1224
1225int sock_create(int family, int type, int protocol, struct socket **res)
1226{
1227 return __sock_create(family, type, protocol, res, 0);
1228}
1229
1230int sock_create_kern(int family, int type, int protocol, struct socket **res)
1231{
1232 return __sock_create(family, type, protocol, res, 1);
1233}
1234
1235asmlinkage long sys_socket(int family, int type, int protocol)
1236{
1237 int retval;
1238 struct socket *sock;
1239
1240 retval = sock_create(family, type, protocol, &sock);
1241 if (retval < 0)
1242 goto out;
1243
1244 retval = sock_map_fd(sock);
1245 if (retval < 0)
1246 goto out_release;
1247
1248out:
1249 /* It may be already another descriptor 8) Not kernel problem. */
1250 return retval;
1251
1252out_release:
1253 sock_release(sock);
1254 return retval;
1255}
1256
1257/*
1258 * Create a pair of connected sockets.
1259 */
1260
1261asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1262{
1263 struct socket *sock1, *sock2;
1264 int fd1, fd2, err;
1265
1266 /*
1267 * Obtain the first socket and check if the underlying protocol
1268 * supports the socketpair call.
1269 */
1270
1271 err = sock_create(family, type, protocol, &sock1);
1272 if (err < 0)
1273 goto out;
1274
1275 err = sock_create(family, type, protocol, &sock2);
1276 if (err < 0)
1277 goto out_release_1;
1278
1279 err = sock1->ops->socketpair(sock1, sock2);
1280 if (err < 0)
1281 goto out_release_both;
1282
1283 fd1 = fd2 = -1;
1284
1285 err = sock_map_fd(sock1);
1286 if (err < 0)
1287 goto out_release_both;
1288 fd1 = err;
1289
1290 err = sock_map_fd(sock2);
1291 if (err < 0)
1292 goto out_close_1;
1293 fd2 = err;
1294
1295 /* fd1 and fd2 may be already another descriptors.
1296 * Not kernel problem.
1297 */
1298
1299 err = put_user(fd1, &usockvec[0]);
1300 if (!err)
1301 err = put_user(fd2, &usockvec[1]);
1302 if (!err)
1303 return 0;
1304
1305 sys_close(fd2);
1306 sys_close(fd1);
1307 return err;
1308
1309out_close_1:
1310 sock_release(sock2);
1311 sys_close(fd1);
1312 return err;
1313
1314out_release_both:
1315 sock_release(sock2);
1316out_release_1:
1317 sock_release(sock1);
1318out:
1319 return err;
1320}
1321
1322
1323/*
1324 * Bind a name to a socket. Nothing much to do here since it's
1325 * the protocol's responsibility to handle the local address.
1326 *
1327 * We move the socket address to kernel space before we call
1328 * the protocol layer (having also checked the address is ok).
1329 */
1330
1331asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1332{
1333 struct socket *sock;
1334 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001335 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001337 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 {
1339 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1340 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001341 if (!err)
1342 err = sock->ops->bind(sock,
1343 (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001345 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 }
1347 return err;
1348}
1349
1350
1351/*
1352 * Perform a listen. Basically, we allow the protocol to do anything
1353 * necessary for a listen, and if that works, we mark the socket as
1354 * ready for listening.
1355 */
1356
1357int sysctl_somaxconn = SOMAXCONN;
1358
1359asmlinkage long sys_listen(int fd, int backlog)
1360{
1361 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001362 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001364 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 if ((unsigned) backlog > sysctl_somaxconn)
1366 backlog = sysctl_somaxconn;
1367
1368 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001369 if (!err)
1370 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001372 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 }
1374 return err;
1375}
1376
1377
1378/*
1379 * For accept, we attempt to create a new socket, set up the link
1380 * with the client, wake up the client, then return the new
1381 * connected fd. We collect the address of the connector in kernel
1382 * space and move it to user at the very end. This is unclean because
1383 * we open the socket then return an error.
1384 *
1385 * 1003.1g adds the ability to recvmsg() to query connection pending
1386 * status to recvmsg. We need to add that support in a way thats
1387 * clean when we restucture accept also.
1388 */
1389
1390asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1391{
1392 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001393 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001394 int err, len, newfd, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 char address[MAX_SOCK_ADDR];
1396
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001397 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 if (!sock)
1399 goto out;
1400
1401 err = -ENFILE;
1402 if (!(newsock = sock_alloc()))
1403 goto out_put;
1404
1405 newsock->type = sock->type;
1406 newsock->ops = sock->ops;
1407
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 /*
1409 * We don't need try_module_get here, as the listening socket (sock)
1410 * has the protocol module (sock->ops->owner) held.
1411 */
1412 __module_get(newsock->ops->owner);
1413
David S. Miller39d8c1b2006-03-20 17:13:49 -08001414 newfd = sock_alloc_fd(&newfile);
1415 if (unlikely(newfd < 0)) {
1416 err = newfd;
1417 goto out_release;
1418 }
1419
1420 err = sock_attach_fd(newsock, newfile);
1421 if (err < 0)
1422 goto out_fd;
1423
Frank Filza79af592005-09-27 15:23:38 -07001424 err = security_socket_accept(sock, newsock);
1425 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001426 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1429 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001430 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
1432 if (upeer_sockaddr) {
1433 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1434 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001435 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 }
1437 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1438 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001439 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 }
1441
1442 /* File flags are not inherited via accept() unlike another OSes. */
1443
David S. Miller39d8c1b2006-03-20 17:13:49 -08001444 fd_install(newfd, newfile);
1445 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446
1447 security_socket_post_accept(sock, newsock);
1448
1449out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001450 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451out:
1452 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001453out_fd:
1454 put_filp(newfile);
1455 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456out_release:
1457 sock_release(newsock);
1458 goto out_put;
1459}
1460
1461
1462/*
1463 * Attempt to connect to a socket with the server address. The address
1464 * is in user space so we verify it is OK and move it to kernel space.
1465 *
1466 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1467 * break bindings
1468 *
1469 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1470 * other SEQPACKET protocols that take time to connect() as it doesn't
1471 * include the -EINPROGRESS status for such sockets.
1472 */
1473
1474asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1475{
1476 struct socket *sock;
1477 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001478 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001480 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 if (!sock)
1482 goto out;
1483 err = move_addr_to_kernel(uservaddr, addrlen, address);
1484 if (err < 0)
1485 goto out_put;
1486
1487 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1488 if (err)
1489 goto out_put;
1490
1491 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1492 sock->file->f_flags);
1493out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001494 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495out:
1496 return err;
1497}
1498
1499/*
1500 * Get the local address ('name') of a socket object. Move the obtained
1501 * name to user space.
1502 */
1503
1504asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1505{
1506 struct socket *sock;
1507 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001508 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001510 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 if (!sock)
1512 goto out;
1513
1514 err = security_socket_getsockname(sock);
1515 if (err)
1516 goto out_put;
1517
1518 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1519 if (err)
1520 goto out_put;
1521 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1522
1523out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001524 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525out:
1526 return err;
1527}
1528
1529/*
1530 * Get the remote address ('name') of a socket object. Move the obtained
1531 * name to user space.
1532 */
1533
1534asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1535{
1536 struct socket *sock;
1537 char address[MAX_SOCK_ADDR];
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001538 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001540 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 err = security_socket_getpeername(sock);
1542 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001543 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 return err;
1545 }
1546
1547 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1548 if (!err)
1549 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001550 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 }
1552 return err;
1553}
1554
1555/*
1556 * Send a datagram to a given address. We move the address into kernel
1557 * space and check the user space data area is readable before invoking
1558 * the protocol.
1559 */
1560
1561asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1562 struct sockaddr __user *addr, int addr_len)
1563{
1564 struct socket *sock;
1565 char address[MAX_SOCK_ADDR];
1566 int err;
1567 struct msghdr msg;
1568 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001569 int fput_needed;
1570 struct file *sock_file;
1571
1572 sock_file = fget_light(fd, &fput_needed);
1573 if (!sock_file)
1574 return -EBADF;
1575
1576 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 if (!sock)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001578 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 iov.iov_base=buff;
1580 iov.iov_len=len;
1581 msg.msg_name=NULL;
1582 msg.msg_iov=&iov;
1583 msg.msg_iovlen=1;
1584 msg.msg_control=NULL;
1585 msg.msg_controllen=0;
1586 msg.msg_namelen=0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001587 if (addr) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 err = move_addr_to_kernel(addr, addr_len, address);
1589 if (err < 0)
1590 goto out_put;
1591 msg.msg_name=address;
1592 msg.msg_namelen=addr_len;
1593 }
1594 if (sock->file->f_flags & O_NONBLOCK)
1595 flags |= MSG_DONTWAIT;
1596 msg.msg_flags = flags;
1597 err = sock_sendmsg(sock, &msg, len);
1598
1599out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001600 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 return err;
1602}
1603
1604/*
1605 * Send a datagram down a socket.
1606 */
1607
1608asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1609{
1610 return sys_sendto(fd, buff, len, flags, NULL, 0);
1611}
1612
1613/*
1614 * Receive a frame from the socket and optionally record the address of the
1615 * sender. We verify the buffers are writable and if needed move the
1616 * sender address from kernel to user space.
1617 */
1618
1619asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1620 struct sockaddr __user *addr, int __user *addr_len)
1621{
1622 struct socket *sock;
1623 struct iovec iov;
1624 struct msghdr msg;
1625 char address[MAX_SOCK_ADDR];
1626 int err,err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001627 struct file *sock_file;
1628 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001630 sock_file = fget_light(fd, &fput_needed);
1631 if (!sock_file)
1632 return -EBADF;
1633
1634 sock = sock_from_file(sock_file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 if (!sock)
1636 goto out;
1637
1638 msg.msg_control=NULL;
1639 msg.msg_controllen=0;
1640 msg.msg_iovlen=1;
1641 msg.msg_iov=&iov;
1642 iov.iov_len=size;
1643 iov.iov_base=ubuf;
1644 msg.msg_name=address;
1645 msg.msg_namelen=MAX_SOCK_ADDR;
1646 if (sock->file->f_flags & O_NONBLOCK)
1647 flags |= MSG_DONTWAIT;
1648 err=sock_recvmsg(sock, &msg, size, flags);
1649
1650 if(err >= 0 && addr != NULL)
1651 {
1652 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1653 if(err2<0)
1654 err=err2;
1655 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656out:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001657 fput_light(sock_file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 return err;
1659}
1660
1661/*
1662 * Receive a datagram from a socket.
1663 */
1664
1665asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1666{
1667 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1668}
1669
1670/*
1671 * Set a socket option. Because we don't know the option lengths we have
1672 * to pass the user mode parameter for the protocols to sort out.
1673 */
1674
1675asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1676{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001677 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 struct socket *sock;
1679
1680 if (optlen < 0)
1681 return -EINVAL;
1682
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001683 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 {
1685 err = security_socket_setsockopt(sock,level,optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001686 if (err)
1687 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688
1689 if (level == SOL_SOCKET)
1690 err=sock_setsockopt(sock,level,optname,optval,optlen);
1691 else
1692 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001693out_put:
1694 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 }
1696 return err;
1697}
1698
1699/*
1700 * Get a socket option. Because we don't know the option lengths we have
1701 * to pass a user mode parameter for the protocols to sort out.
1702 */
1703
1704asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1705{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001706 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 struct socket *sock;
1708
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001709 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1710 err = security_socket_getsockopt(sock, level, optname);
1711 if (err)
1712 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
1714 if (level == SOL_SOCKET)
1715 err=sock_getsockopt(sock,level,optname,optval,optlen);
1716 else
1717 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001718out_put:
1719 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 }
1721 return err;
1722}
1723
1724
1725/*
1726 * Shutdown a socket.
1727 */
1728
1729asmlinkage long sys_shutdown(int fd, int how)
1730{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001731 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 struct socket *sock;
1733
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001734 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 {
1736 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001737 if (!err)
1738 err = sock->ops->shutdown(sock, how);
1739 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 }
1741 return err;
1742}
1743
1744/* A couple of helpful macros for getting the address of the 32/64 bit
1745 * fields which are the same type (int / unsigned) on our platforms.
1746 */
1747#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1748#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1749#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1750
1751
1752/*
1753 * BSD sendmsg interface
1754 */
1755
1756asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1757{
1758 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1759 struct socket *sock;
1760 char address[MAX_SOCK_ADDR];
1761 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001762 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1763 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1764 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 unsigned char *ctl_buf = ctl;
1766 struct msghdr msg_sys;
1767 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001768 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769
1770 err = -EFAULT;
1771 if (MSG_CMSG_COMPAT & flags) {
1772 if (get_compat_msghdr(&msg_sys, msg_compat))
1773 return -EFAULT;
1774 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1775 return -EFAULT;
1776
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001777 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 if (!sock)
1779 goto out;
1780
1781 /* do not move before msg_sys is valid */
1782 err = -EMSGSIZE;
1783 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1784 goto out_put;
1785
1786 /* Check whether to allocate the iovec area*/
1787 err = -ENOMEM;
1788 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1789 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1790 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1791 if (!iov)
1792 goto out_put;
1793 }
1794
1795 /* This will also move the address data into kernel space */
1796 if (MSG_CMSG_COMPAT & flags) {
1797 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1798 } else
1799 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1800 if (err < 0)
1801 goto out_freeiov;
1802 total_len = err;
1803
1804 err = -ENOBUFS;
1805
1806 if (msg_sys.msg_controllen > INT_MAX)
1807 goto out_freeiov;
1808 ctl_len = msg_sys.msg_controllen;
1809 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Al Viro8920e8f2005-09-07 18:28:51 -07001810 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 if (err)
1812 goto out_freeiov;
1813 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001814 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 } else if (ctl_len) {
1816 if (ctl_len > sizeof(ctl))
1817 {
1818 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1819 if (ctl_buf == NULL)
1820 goto out_freeiov;
1821 }
1822 err = -EFAULT;
1823 /*
1824 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1825 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1826 * checking falls down on this.
1827 */
1828 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1829 goto out_freectl;
1830 msg_sys.msg_control = ctl_buf;
1831 }
1832 msg_sys.msg_flags = flags;
1833
1834 if (sock->file->f_flags & O_NONBLOCK)
1835 msg_sys.msg_flags |= MSG_DONTWAIT;
1836 err = sock_sendmsg(sock, &msg_sys, total_len);
1837
1838out_freectl:
1839 if (ctl_buf != ctl)
1840 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1841out_freeiov:
1842 if (iov != iovstack)
1843 sock_kfree_s(sock->sk, iov, iov_size);
1844out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001845 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846out:
1847 return err;
1848}
1849
1850/*
1851 * BSD recvmsg interface
1852 */
1853
1854asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1855{
1856 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1857 struct socket *sock;
1858 struct iovec iovstack[UIO_FASTIOV];
1859 struct iovec *iov=iovstack;
1860 struct msghdr msg_sys;
1861 unsigned long cmsg_ptr;
1862 int err, iov_size, total_len, len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001863 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864
1865 /* kernel mode address */
1866 char addr[MAX_SOCK_ADDR];
1867
1868 /* user mode address pointers */
1869 struct sockaddr __user *uaddr;
1870 int __user *uaddr_len;
1871
1872 if (MSG_CMSG_COMPAT & flags) {
1873 if (get_compat_msghdr(&msg_sys, msg_compat))
1874 return -EFAULT;
1875 } else
1876 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1877 return -EFAULT;
1878
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001879 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 if (!sock)
1881 goto out;
1882
1883 err = -EMSGSIZE;
1884 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1885 goto out_put;
1886
1887 /* Check whether to allocate the iovec area*/
1888 err = -ENOMEM;
1889 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1890 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1891 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1892 if (!iov)
1893 goto out_put;
1894 }
1895
1896 /*
1897 * Save the user-mode address (verify_iovec will change the
1898 * kernel msghdr to use the kernel address space)
1899 */
1900
1901 uaddr = (void __user *) msg_sys.msg_name;
1902 uaddr_len = COMPAT_NAMELEN(msg);
1903 if (MSG_CMSG_COMPAT & flags) {
1904 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1905 } else
1906 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1907 if (err < 0)
1908 goto out_freeiov;
1909 total_len=err;
1910
1911 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1912 msg_sys.msg_flags = 0;
1913 if (MSG_CMSG_COMPAT & flags)
1914 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1915
1916 if (sock->file->f_flags & O_NONBLOCK)
1917 flags |= MSG_DONTWAIT;
1918 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1919 if (err < 0)
1920 goto out_freeiov;
1921 len = err;
1922
1923 if (uaddr != NULL) {
1924 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1925 if (err < 0)
1926 goto out_freeiov;
1927 }
David S. Miller37f7f422005-09-16 16:51:01 -07001928 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1929 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 if (err)
1931 goto out_freeiov;
1932 if (MSG_CMSG_COMPAT & flags)
1933 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1934 &msg_compat->msg_controllen);
1935 else
1936 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1937 &msg->msg_controllen);
1938 if (err)
1939 goto out_freeiov;
1940 err = len;
1941
1942out_freeiov:
1943 if (iov != iovstack)
1944 sock_kfree_s(sock->sk, iov, iov_size);
1945out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001946 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947out:
1948 return err;
1949}
1950
1951#ifdef __ARCH_WANT_SYS_SOCKETCALL
1952
1953/* Argument list sizes for sys_socketcall */
1954#define AL(x) ((x) * sizeof(unsigned long))
1955static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1956 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1957 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1958#undef AL
1959
1960/*
1961 * System call vectors.
1962 *
1963 * Argument checking cleaned up. Saved 20% in size.
1964 * This function doesn't need to set the kernel lock because
1965 * it is set by the callees.
1966 */
1967
1968asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1969{
1970 unsigned long a[6];
1971 unsigned long a0,a1;
1972 int err;
1973
1974 if(call<1||call>SYS_RECVMSG)
1975 return -EINVAL;
1976
1977 /* copy_from_user should be SMP safe. */
1978 if (copy_from_user(a, args, nargs[call]))
1979 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001980
David Woodhouse4bcff1b2005-06-02 12:13:21 +01001981 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01001982 if (err)
1983 return err;
1984
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 a0=a[0];
1986 a1=a[1];
1987
1988 switch(call)
1989 {
1990 case SYS_SOCKET:
1991 err = sys_socket(a0,a1,a[2]);
1992 break;
1993 case SYS_BIND:
1994 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1995 break;
1996 case SYS_CONNECT:
1997 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1998 break;
1999 case SYS_LISTEN:
2000 err = sys_listen(a0,a1);
2001 break;
2002 case SYS_ACCEPT:
2003 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2004 break;
2005 case SYS_GETSOCKNAME:
2006 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2007 break;
2008 case SYS_GETPEERNAME:
2009 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2010 break;
2011 case SYS_SOCKETPAIR:
2012 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2013 break;
2014 case SYS_SEND:
2015 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2016 break;
2017 case SYS_SENDTO:
2018 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2019 (struct sockaddr __user *)a[4], a[5]);
2020 break;
2021 case SYS_RECV:
2022 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2023 break;
2024 case SYS_RECVFROM:
2025 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2026 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2027 break;
2028 case SYS_SHUTDOWN:
2029 err = sys_shutdown(a0,a1);
2030 break;
2031 case SYS_SETSOCKOPT:
2032 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2033 break;
2034 case SYS_GETSOCKOPT:
2035 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2036 break;
2037 case SYS_SENDMSG:
2038 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2039 break;
2040 case SYS_RECVMSG:
2041 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2042 break;
2043 default:
2044 err = -EINVAL;
2045 break;
2046 }
2047 return err;
2048}
2049
2050#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2051
2052/*
2053 * This function is called by a protocol handler that wants to
2054 * advertise its address family, and have it linked into the
2055 * SOCKET module.
2056 */
2057
2058int sock_register(struct net_proto_family *ops)
2059{
2060 int err;
2061
2062 if (ops->family >= NPROTO) {
2063 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2064 return -ENOBUFS;
2065 }
2066 net_family_write_lock();
2067 err = -EEXIST;
2068 if (net_families[ops->family] == NULL) {
2069 net_families[ops->family]=ops;
2070 err = 0;
2071 }
2072 net_family_write_unlock();
2073 printk(KERN_INFO "NET: Registered protocol family %d\n",
2074 ops->family);
2075 return err;
2076}
2077
2078/*
2079 * This function is called by a protocol handler that wants to
2080 * remove its address family, and have it unlinked from the
2081 * SOCKET module.
2082 */
2083
2084int sock_unregister(int family)
2085{
2086 if (family < 0 || family >= NPROTO)
2087 return -1;
2088
2089 net_family_write_lock();
2090 net_families[family]=NULL;
2091 net_family_write_unlock();
2092 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2093 family);
2094 return 0;
2095}
2096
Andi Kleen77d76ea2005-12-22 12:43:42 -08002097static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098{
2099 /*
2100 * Initialize sock SLAB cache.
2101 */
2102
2103 sk_init();
2104
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 /*
2106 * Initialize skbuff SLAB cache
2107 */
2108 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109
2110 /*
2111 * Initialize the protocols module.
2112 */
2113
2114 init_inodecache();
2115 register_filesystem(&sock_fs_type);
2116 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002117
2118 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 */
2120
2121#ifdef CONFIG_NETFILTER
2122 netfilter_init();
2123#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002124
2125 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126}
2127
Andi Kleen77d76ea2005-12-22 12:43:42 -08002128core_initcall(sock_init); /* early initcall */
2129
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130#ifdef CONFIG_PROC_FS
2131void socket_seq_show(struct seq_file *seq)
2132{
2133 int cpu;
2134 int counter = 0;
2135
Eric Dumazet88a2a4ac2006-02-04 23:27:36 -08002136 for_each_cpu(cpu)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 counter += per_cpu(sockets_in_use, cpu);
2138
2139 /* It can be negative, by the way. 8) */
2140 if (counter < 0)
2141 counter = 0;
2142
2143 seq_printf(seq, "sockets: used %d\n", counter);
2144}
2145#endif /* CONFIG_PROC_FS */
2146
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002147#ifdef CONFIG_COMPAT
2148static long compat_sock_ioctl(struct file *file, unsigned cmd,
2149 unsigned long arg)
2150{
2151 struct socket *sock = file->private_data;
2152 int ret = -ENOIOCTLCMD;
2153
2154 if (sock->ops->compat_ioctl)
2155 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2156
2157 return ret;
2158}
2159#endif
2160
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161/* ABI emulation layers need these two */
2162EXPORT_SYMBOL(move_addr_to_kernel);
2163EXPORT_SYMBOL(move_addr_to_user);
2164EXPORT_SYMBOL(sock_create);
2165EXPORT_SYMBOL(sock_create_kern);
2166EXPORT_SYMBOL(sock_create_lite);
2167EXPORT_SYMBOL(sock_map_fd);
2168EXPORT_SYMBOL(sock_recvmsg);
2169EXPORT_SYMBOL(sock_register);
2170EXPORT_SYMBOL(sock_release);
2171EXPORT_SYMBOL(sock_sendmsg);
2172EXPORT_SYMBOL(sock_unregister);
2173EXPORT_SYMBOL(sock_wake_async);
2174EXPORT_SYMBOL(sockfd_lookup);
2175EXPORT_SYMBOL(kernel_sendmsg);
2176EXPORT_SYMBOL(kernel_recvmsg);