Merge branch 'upstream-fixes'
[linux-2.6-block.git] / net / socket.c
CommitLineData
1da177e4
LT
1/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
02c30a84 7 * Ross Biro
1da177e4
LT
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61#include <linux/config.h>
62#include <linux/mm.h>
63#include <linux/smp_lock.h>
64#include <linux/socket.h>
65#include <linux/file.h>
66#include <linux/net.h>
67#include <linux/interrupt.h>
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
71#include <linux/wanrouter.h>
72#include <linux/if_bridge.h>
20380731
ACM
73#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
1da177e4
LT
75#include <linux/init.h>
76#include <linux/poll.h>
77#include <linux/cache.h>
78#include <linux/module.h>
79#include <linux/highmem.h>
80#include <linux/divert.h>
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
3ec3b2fb 86#include <linux/audit.h>
1da177e4
LT
87
88#ifdef CONFIG_NET_RADIO
89#include <linux/wireless.h> /* Note : will define WIRELESS_EXT */
90#endif /* CONFIG_NET_RADIO */
91
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
96
97#include <net/sock.h>
98#include <linux/netfilter.h>
99
100static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
101static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
102 size_t size, loff_t pos);
103static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
104 size_t size, loff_t pos);
105static int sock_mmap(struct file *file, struct vm_area_struct * vma);
106
107static int sock_close(struct inode *inode, struct file *file);
108static unsigned int sock_poll(struct file *file,
109 struct poll_table_struct *wait);
110static long sock_ioctl(struct file *file,
111 unsigned int cmd, unsigned long arg);
112static int sock_fasync(int fd, struct file *filp, int on);
113static ssize_t sock_readv(struct file *file, const struct iovec *vector,
114 unsigned long count, loff_t *ppos);
115static ssize_t sock_writev(struct file *file, const struct iovec *vector,
116 unsigned long count, loff_t *ppos);
117static ssize_t sock_sendpage(struct file *file, struct page *page,
118 int offset, size_t size, loff_t *ppos, int more);
119
120
121/*
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
124 */
125
126static struct file_operations socket_file_ops = {
127 .owner = THIS_MODULE,
128 .llseek = no_llseek,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
131 .poll = sock_poll,
132 .unlocked_ioctl = sock_ioctl,
133 .mmap = sock_mmap,
134 .open = sock_no_open, /* special open code to disallow open via /proc */
135 .release = sock_close,
136 .fasync = sock_fasync,
137 .readv = sock_readv,
138 .writev = sock_writev,
139 .sendpage = sock_sendpage
140};
141
142/*
143 * The protocol list. Each protocol is registered in here.
144 */
145
146static struct net_proto_family *net_families[NPROTO];
147
148#if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
149static atomic_t net_family_lockct = ATOMIC_INIT(0);
150static DEFINE_SPINLOCK(net_family_lock);
151
152/* The strategy is: modifications net_family vector are short, do not
153 sleep and veeery rare, but read access should be free of any exclusive
154 locks.
155 */
156
157static void net_family_write_lock(void)
158{
159 spin_lock(&net_family_lock);
160 while (atomic_read(&net_family_lockct) != 0) {
161 spin_unlock(&net_family_lock);
162
163 yield();
164
165 spin_lock(&net_family_lock);
166 }
167}
168
169static __inline__ void net_family_write_unlock(void)
170{
171 spin_unlock(&net_family_lock);
172}
173
174static __inline__ void net_family_read_lock(void)
175{
176 atomic_inc(&net_family_lockct);
177 spin_unlock_wait(&net_family_lock);
178}
179
180static __inline__ void net_family_read_unlock(void)
181{
182 atomic_dec(&net_family_lockct);
183}
184
185#else
186#define net_family_write_lock() do { } while(0)
187#define net_family_write_unlock() do { } while(0)
188#define net_family_read_lock() do { } while(0)
189#define net_family_read_unlock() do { } while(0)
190#endif
191
192
193/*
194 * Statistics counters of the socket lists
195 */
196
197static DEFINE_PER_CPU(int, sockets_in_use) = 0;
198
199/*
200 * Support routines. Move socket addresses back and forth across the kernel/user
201 * divide and look after the messy bits.
202 */
203
204#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
205 16 for IP, 16 for IPX,
206 24 for IPv6,
207 about 80 for AX.25
208 must be at least one bigger than
209 the AF_UNIX size (see net/unix/af_unix.c
210 :unix_mkname()).
211 */
212
213/**
214 * move_addr_to_kernel - copy a socket address into kernel space
215 * @uaddr: Address in user space
216 * @kaddr: Address in kernel space
217 * @ulen: Length in user space
218 *
219 * The address is copied into kernel space. If the provided address is
220 * too long an error code of -EINVAL is returned. If the copy gives
221 * invalid addresses -EFAULT is returned. On a success 0 is returned.
222 */
223
224int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
225{
226 if(ulen<0||ulen>MAX_SOCK_ADDR)
227 return -EINVAL;
228 if(ulen==0)
229 return 0;
230 if(copy_from_user(kaddr,uaddr,ulen))
231 return -EFAULT;
3ec3b2fb 232 return audit_sockaddr(ulen, kaddr);
1da177e4
LT
233}
234
235/**
236 * move_addr_to_user - copy an address to user space
237 * @kaddr: kernel space address
238 * @klen: length of address in kernel
239 * @uaddr: user space address
240 * @ulen: pointer to user length field
241 *
242 * The value pointed to by ulen on entry is the buffer length available.
243 * This is overwritten with the buffer space used. -EINVAL is returned
244 * if an overlong buffer is specified or a negative buffer size. -EFAULT
245 * is returned if either the buffer or the length field are not
246 * accessible.
247 * After copying the data up to the limit the user specifies, the true
248 * length of the data is written over the length limit the user
249 * specified. Zero is returned for a success.
250 */
251
252int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
253{
254 int err;
255 int len;
256
257 if((err=get_user(len, ulen)))
258 return err;
259 if(len>klen)
260 len=klen;
261 if(len<0 || len> MAX_SOCK_ADDR)
262 return -EINVAL;
263 if(len)
264 {
265 if(copy_to_user(uaddr,kaddr,len))
266 return -EFAULT;
267 }
268 /*
269 * "fromlen shall refer to the value before truncation.."
270 * 1003.1g
271 */
272 return __put_user(klen, ulen);
273}
274
275#define SOCKFS_MAGIC 0x534F434B
276
ba89966c 277static kmem_cache_t * sock_inode_cachep __read_mostly;
1da177e4
LT
278
279static struct inode *sock_alloc_inode(struct super_block *sb)
280{
281 struct socket_alloc *ei;
282 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
283 if (!ei)
284 return NULL;
285 init_waitqueue_head(&ei->socket.wait);
286
287 ei->socket.fasync_list = NULL;
288 ei->socket.state = SS_UNCONNECTED;
289 ei->socket.flags = 0;
290 ei->socket.ops = NULL;
291 ei->socket.sk = NULL;
292 ei->socket.file = NULL;
293 ei->socket.flags = 0;
294
295 return &ei->vfs_inode;
296}
297
298static void sock_destroy_inode(struct inode *inode)
299{
300 kmem_cache_free(sock_inode_cachep,
301 container_of(inode, struct socket_alloc, vfs_inode));
302}
303
304static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
305{
306 struct socket_alloc *ei = (struct socket_alloc *) foo;
307
308 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
309 SLAB_CTOR_CONSTRUCTOR)
310 inode_init_once(&ei->vfs_inode);
311}
312
313static int init_inodecache(void)
314{
315 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
316 sizeof(struct socket_alloc),
317 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
318 init_once, NULL);
319 if (sock_inode_cachep == NULL)
320 return -ENOMEM;
321 return 0;
322}
323
324static struct super_operations sockfs_ops = {
325 .alloc_inode = sock_alloc_inode,
326 .destroy_inode =sock_destroy_inode,
327 .statfs = simple_statfs,
328};
329
330static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
331 int flags, const char *dev_name, void *data)
332{
333 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
334}
335
ba89966c 336static struct vfsmount *sock_mnt __read_mostly;
1da177e4
LT
337
338static struct file_system_type sock_fs_type = {
339 .name = "sockfs",
340 .get_sb = sockfs_get_sb,
341 .kill_sb = kill_anon_super,
342};
343static int sockfs_delete_dentry(struct dentry *dentry)
344{
345 return 1;
346}
347static struct dentry_operations sockfs_dentry_operations = {
348 .d_delete = sockfs_delete_dentry,
349};
350
351/*
352 * Obtains the first available file descriptor and sets it up for use.
353 *
354 * This function creates file structure and maps it to fd space
355 * of current process. On success it returns file descriptor
356 * and file struct implicitly stored in sock->file.
357 * Note that another thread may close file descriptor before we return
358 * from this function. We use the fact that now we do not refer
359 * to socket after mapping. If one day we will need it, this
360 * function will increment ref. count on file by 1.
361 *
362 * In any case returned fd MAY BE not valid!
363 * This race condition is unavoidable
364 * with shared fd spaces, we cannot solve it inside kernel,
365 * but we take care of internal coherence yet.
366 */
367
368int sock_map_fd(struct socket *sock)
369{
370 int fd;
371 struct qstr this;
372 char name[32];
373
374 /*
375 * Find a file descriptor suitable for return to the user.
376 */
377
378 fd = get_unused_fd();
379 if (fd >= 0) {
380 struct file *file = get_empty_filp();
381
382 if (!file) {
383 put_unused_fd(fd);
384 fd = -ENFILE;
385 goto out;
386 }
387
f31f5f05 388 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
1da177e4 389 this.name = name;
1da177e4
LT
390 this.hash = SOCK_INODE(sock)->i_ino;
391
392 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
393 if (!file->f_dentry) {
394 put_filp(file);
395 put_unused_fd(fd);
396 fd = -ENOMEM;
397 goto out;
398 }
399 file->f_dentry->d_op = &sockfs_dentry_operations;
400 d_add(file->f_dentry, SOCK_INODE(sock));
401 file->f_vfsmnt = mntget(sock_mnt);
402 file->f_mapping = file->f_dentry->d_inode->i_mapping;
403
404 sock->file = file;
405 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
406 file->f_mode = FMODE_READ | FMODE_WRITE;
407 file->f_flags = O_RDWR;
408 file->f_pos = 0;
07dc3f07 409 file->private_data = sock;
1da177e4
LT
410 fd_install(fd, file);
411 }
412
413out:
414 return fd;
415}
416
417/**
418 * sockfd_lookup - Go from a file number to its socket slot
419 * @fd: file handle
420 * @err: pointer to an error code return
421 *
422 * The file handle passed in is locked and the socket it is bound
423 * too is returned. If an error occurs the err pointer is overwritten
424 * with a negative errno code and NULL is returned. The function checks
425 * for both invalid handles and passing a handle which is not a socket.
426 *
427 * On a success the socket object pointer is returned.
428 */
429
430struct socket *sockfd_lookup(int fd, int *err)
431{
432 struct file *file;
433 struct inode *inode;
434 struct socket *sock;
435
436 if (!(file = fget(fd)))
437 {
438 *err = -EBADF;
439 return NULL;
440 }
441
07dc3f07
BL
442 if (file->f_op == &socket_file_ops)
443 return file->private_data; /* set in sock_map_fd */
444
1da177e4
LT
445 inode = file->f_dentry->d_inode;
446 if (!S_ISSOCK(inode->i_mode)) {
447 *err = -ENOTSOCK;
448 fput(file);
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
460/**
461 * sock_alloc - allocate a socket
462 *
463 * Allocate a new inode and socket object. The two are bound together
464 * and initialised. The socket is then returned. If we are out of inodes
465 * NULL is returned.
466 */
467
468static struct socket *sock_alloc(void)
469{
470 struct inode * inode;
471 struct socket * sock;
472
473 inode = new_inode(sock_mnt->mnt_sb);
474 if (!inode)
475 return NULL;
476
477 sock = SOCKET_I(inode);
478
479 inode->i_mode = S_IFSOCK|S_IRWXUGO;
480 inode->i_uid = current->fsuid;
481 inode->i_gid = current->fsgid;
482
483 get_cpu_var(sockets_in_use)++;
484 put_cpu_var(sockets_in_use);
485 return sock;
486}
487
488/*
489 * In theory you can't get an open on this inode, but /proc provides
490 * a back door. Remember to keep it shut otherwise you'll let the
491 * creepy crawlies in.
492 */
493
494static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
495{
496 return -ENXIO;
497}
498
499struct file_operations bad_sock_fops = {
500 .owner = THIS_MODULE,
501 .open = sock_no_open,
502};
503
504/**
505 * sock_release - close a socket
506 * @sock: socket to close
507 *
508 * The socket is released from the protocol stack if it has a release
509 * callback, and the inode is then released if the socket is bound to
510 * an inode not a file.
511 */
512
513void sock_release(struct socket *sock)
514{
515 if (sock->ops) {
516 struct module *owner = sock->ops->owner;
517
518 sock->ops->release(sock);
519 sock->ops = NULL;
520 module_put(owner);
521 }
522
523 if (sock->fasync_list)
524 printk(KERN_ERR "sock_release: fasync list not empty!\n");
525
526 get_cpu_var(sockets_in_use)--;
527 put_cpu_var(sockets_in_use);
528 if (!sock->file) {
529 iput(SOCK_INODE(sock));
530 return;
531 }
532 sock->file=NULL;
533}
534
535static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
536 struct msghdr *msg, size_t size)
537{
538 struct sock_iocb *si = kiocb_to_siocb(iocb);
539 int err;
540
541 si->sock = sock;
542 si->scm = NULL;
543 si->msg = msg;
544 si->size = size;
545
546 err = security_socket_sendmsg(sock, msg, size);
547 if (err)
548 return err;
549
550 return sock->ops->sendmsg(iocb, sock, msg, size);
551}
552
553int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
554{
555 struct kiocb iocb;
556 struct sock_iocb siocb;
557 int ret;
558
559 init_sync_kiocb(&iocb, NULL);
560 iocb.private = &siocb;
561 ret = __sock_sendmsg(&iocb, sock, msg, size);
562 if (-EIOCBQUEUED == ret)
563 ret = wait_on_sync_kiocb(&iocb);
564 return ret;
565}
566
567int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
568 struct kvec *vec, size_t num, size_t size)
569{
570 mm_segment_t oldfs = get_fs();
571 int result;
572
573 set_fs(KERNEL_DS);
574 /*
575 * the following is safe, since for compiler definitions of kvec and
576 * iovec are identical, yielding the same in-core layout and alignment
577 */
578 msg->msg_iov = (struct iovec *)vec,
579 msg->msg_iovlen = num;
580 result = sock_sendmsg(sock, msg, size);
581 set_fs(oldfs);
582 return result;
583}
584
585static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
586 struct msghdr *msg, size_t size, int flags)
587{
588 int err;
589 struct sock_iocb *si = kiocb_to_siocb(iocb);
590
591 si->sock = sock;
592 si->scm = NULL;
593 si->msg = msg;
594 si->size = size;
595 si->flags = flags;
596
597 err = security_socket_recvmsg(sock, msg, size, flags);
598 if (err)
599 return err;
600
601 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
602}
603
604int sock_recvmsg(struct socket *sock, struct msghdr *msg,
605 size_t size, int flags)
606{
607 struct kiocb iocb;
608 struct sock_iocb siocb;
609 int ret;
610
611 init_sync_kiocb(&iocb, NULL);
612 iocb.private = &siocb;
613 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
614 if (-EIOCBQUEUED == ret)
615 ret = wait_on_sync_kiocb(&iocb);
616 return ret;
617}
618
619int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
620 struct kvec *vec, size_t num,
621 size_t size, int flags)
622{
623 mm_segment_t oldfs = get_fs();
624 int result;
625
626 set_fs(KERNEL_DS);
627 /*
628 * the following is safe, since for compiler definitions of kvec and
629 * iovec are identical, yielding the same in-core layout and alignment
630 */
631 msg->msg_iov = (struct iovec *)vec,
632 msg->msg_iovlen = num;
633 result = sock_recvmsg(sock, msg, size, flags);
634 set_fs(oldfs);
635 return result;
636}
637
638static void sock_aio_dtor(struct kiocb *iocb)
639{
640 kfree(iocb->private);
641}
642
ce1d4d3e
CH
643static ssize_t sock_sendpage(struct file *file, struct page *page,
644 int offset, size_t size, loff_t *ppos, int more)
1da177e4 645{
1da177e4
LT
646 struct socket *sock;
647 int flags;
648
ce1d4d3e
CH
649 sock = file->private_data;
650
651 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
652 if (more)
653 flags |= MSG_MORE;
654
655 return sock->ops->sendpage(sock, page, offset, size, flags);
656}
1da177e4 657
ce1d4d3e
CH
658static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
659 char __user *ubuf, size_t size, struct sock_iocb *siocb)
660{
661 if (!is_sync_kiocb(iocb)) {
662 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
663 if (!siocb)
664 return NULL;
1da177e4
LT
665 iocb->ki_dtor = sock_aio_dtor;
666 }
1da177e4 667
ce1d4d3e
CH
668 siocb->kiocb = iocb;
669 siocb->async_iov.iov_base = ubuf;
670 siocb->async_iov.iov_len = size;
1da177e4 671
ce1d4d3e
CH
672 iocb->private = siocb;
673 return siocb;
1da177e4
LT
674}
675
ce1d4d3e
CH
676static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
677 struct file *file, struct iovec *iov, unsigned long nr_segs)
678{
679 struct socket *sock = file->private_data;
680 size_t size = 0;
681 int i;
1da177e4 682
ce1d4d3e
CH
683 for (i = 0 ; i < nr_segs ; i++)
684 size += iov[i].iov_len;
1da177e4 685
ce1d4d3e
CH
686 msg->msg_name = NULL;
687 msg->msg_namelen = 0;
688 msg->msg_control = NULL;
689 msg->msg_controllen = 0;
690 msg->msg_iov = (struct iovec *) iov;
691 msg->msg_iovlen = nr_segs;
692 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
693
694 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
695}
696
697static ssize_t sock_readv(struct file *file, const struct iovec *iov,
698 unsigned long nr_segs, loff_t *ppos)
1da177e4 699{
ce1d4d3e
CH
700 struct kiocb iocb;
701 struct sock_iocb siocb;
702 struct msghdr msg;
703 int ret;
704
705 init_sync_kiocb(&iocb, NULL);
706 iocb.private = &siocb;
707
708 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
709 if (-EIOCBQUEUED == ret)
710 ret = wait_on_sync_kiocb(&iocb);
711 return ret;
712}
713
714static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
715 size_t count, loff_t pos)
716{
717 struct sock_iocb siocb, *x;
718
1da177e4
LT
719 if (pos != 0)
720 return -ESPIPE;
ce1d4d3e 721 if (count == 0) /* Match SYS5 behaviour */
1da177e4
LT
722 return 0;
723
ce1d4d3e
CH
724 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
725 if (!x)
726 return -ENOMEM;
727 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
728 &x->async_iov, 1);
1da177e4
LT
729}
730
ce1d4d3e
CH
731static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
732 struct file *file, struct iovec *iov, unsigned long nr_segs)
1da177e4 733{
ce1d4d3e
CH
734 struct socket *sock = file->private_data;
735 size_t size = 0;
736 int i;
1da177e4 737
ce1d4d3e
CH
738 for (i = 0 ; i < nr_segs ; i++)
739 size += iov[i].iov_len;
1da177e4 740
ce1d4d3e
CH
741 msg->msg_name = NULL;
742 msg->msg_namelen = 0;
743 msg->msg_control = NULL;
744 msg->msg_controllen = 0;
745 msg->msg_iov = (struct iovec *) iov;
746 msg->msg_iovlen = nr_segs;
747 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
748 if (sock->type == SOCK_SEQPACKET)
749 msg->msg_flags |= MSG_EOR;
1da177e4 750
ce1d4d3e 751 return __sock_sendmsg(iocb, sock, msg, size);
1da177e4
LT
752}
753
ce1d4d3e
CH
754static ssize_t sock_writev(struct file *file, const struct iovec *iov,
755 unsigned long nr_segs, loff_t *ppos)
1da177e4
LT
756{
757 struct msghdr msg;
ce1d4d3e
CH
758 struct kiocb iocb;
759 struct sock_iocb siocb;
760 int ret;
1da177e4 761
ce1d4d3e
CH
762 init_sync_kiocb(&iocb, NULL);
763 iocb.private = &siocb;
1da177e4 764
ce1d4d3e
CH
765 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
766 if (-EIOCBQUEUED == ret)
767 ret = wait_on_sync_kiocb(&iocb);
768 return ret;
769}
1da177e4 770
ce1d4d3e
CH
771static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
772 size_t count, loff_t pos)
773{
774 struct sock_iocb siocb, *x;
1da177e4 775
ce1d4d3e
CH
776 if (pos != 0)
777 return -ESPIPE;
778 if (count == 0) /* Match SYS5 behaviour */
779 return 0;
1da177e4 780
ce1d4d3e
CH
781 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
782 if (!x)
783 return -ENOMEM;
1da177e4 784
ce1d4d3e
CH
785 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
786 &x->async_iov, 1);
1da177e4
LT
787}
788
789
790/*
791 * Atomic setting of ioctl hooks to avoid race
792 * with module unload.
793 */
794
795static DECLARE_MUTEX(br_ioctl_mutex);
796static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
797
798void brioctl_set(int (*hook)(unsigned int, void __user *))
799{
800 down(&br_ioctl_mutex);
801 br_ioctl_hook = hook;
802 up(&br_ioctl_mutex);
803}
804EXPORT_SYMBOL(brioctl_set);
805
806static DECLARE_MUTEX(vlan_ioctl_mutex);
807static int (*vlan_ioctl_hook)(void __user *arg);
808
809void vlan_ioctl_set(int (*hook)(void __user *))
810{
811 down(&vlan_ioctl_mutex);
812 vlan_ioctl_hook = hook;
813 up(&vlan_ioctl_mutex);
814}
815EXPORT_SYMBOL(vlan_ioctl_set);
816
817static DECLARE_MUTEX(dlci_ioctl_mutex);
818static int (*dlci_ioctl_hook)(unsigned int, void __user *);
819
820void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
821{
822 down(&dlci_ioctl_mutex);
823 dlci_ioctl_hook = hook;
824 up(&dlci_ioctl_mutex);
825}
826EXPORT_SYMBOL(dlci_ioctl_set);
827
828/*
829 * With an ioctl, arg may well be a user mode pointer, but we don't know
830 * what to do with it - that's up to the protocol still.
831 */
832
833static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
834{
835 struct socket *sock;
836 void __user *argp = (void __user *)arg;
837 int pid, err;
838
b69aee04 839 sock = file->private_data;
1da177e4
LT
840 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
841 err = dev_ioctl(cmd, argp);
842 } else
843#ifdef WIRELESS_EXT
844 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
845 err = dev_ioctl(cmd, argp);
846 } else
847#endif /* WIRELESS_EXT */
848 switch (cmd) {
849 case FIOSETOWN:
850 case SIOCSPGRP:
851 err = -EFAULT;
852 if (get_user(pid, (int __user *)argp))
853 break;
854 err = f_setown(sock->file, pid, 1);
855 break;
856 case FIOGETOWN:
857 case SIOCGPGRP:
858 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
859 break;
860 case SIOCGIFBR:
861 case SIOCSIFBR:
862 case SIOCBRADDBR:
863 case SIOCBRDELBR:
864 err = -ENOPKG;
865 if (!br_ioctl_hook)
866 request_module("bridge");
867
868 down(&br_ioctl_mutex);
869 if (br_ioctl_hook)
870 err = br_ioctl_hook(cmd, argp);
871 up(&br_ioctl_mutex);
872 break;
873 case SIOCGIFVLAN:
874 case SIOCSIFVLAN:
875 err = -ENOPKG;
876 if (!vlan_ioctl_hook)
877 request_module("8021q");
878
879 down(&vlan_ioctl_mutex);
880 if (vlan_ioctl_hook)
881 err = vlan_ioctl_hook(argp);
882 up(&vlan_ioctl_mutex);
883 break;
884 case SIOCGIFDIVERT:
885 case SIOCSIFDIVERT:
886 /* Convert this to call through a hook */
887 err = divert_ioctl(cmd, argp);
888 break;
889 case SIOCADDDLCI:
890 case SIOCDELDLCI:
891 err = -ENOPKG;
892 if (!dlci_ioctl_hook)
893 request_module("dlci");
894
895 if (dlci_ioctl_hook) {
896 down(&dlci_ioctl_mutex);
897 err = dlci_ioctl_hook(cmd, argp);
898 up(&dlci_ioctl_mutex);
899 }
900 break;
901 default:
902 err = sock->ops->ioctl(sock, cmd, arg);
b5e5fa5e
CH
903
904 /*
905 * If this ioctl is unknown try to hand it down
906 * to the NIC driver.
907 */
908 if (err == -ENOIOCTLCMD)
909 err = dev_ioctl(cmd, argp);
1da177e4
LT
910 break;
911 }
912 return err;
913}
914
915int sock_create_lite(int family, int type, int protocol, struct socket **res)
916{
917 int err;
918 struct socket *sock = NULL;
919
920 err = security_socket_create(family, type, protocol, 1);
921 if (err)
922 goto out;
923
924 sock = sock_alloc();
925 if (!sock) {
926 err = -ENOMEM;
927 goto out;
928 }
929
930 security_socket_post_create(sock, family, type, protocol, 1);
931 sock->type = type;
932out:
933 *res = sock;
934 return err;
935}
936
937/* No kernel lock held - perfect */
938static unsigned int sock_poll(struct file *file, poll_table * wait)
939{
940 struct socket *sock;
941
942 /*
943 * We can't return errors to poll, so it's either yes or no.
944 */
b69aee04 945 sock = file->private_data;
1da177e4
LT
946 return sock->ops->poll(file, sock, wait);
947}
948
949static int sock_mmap(struct file * file, struct vm_area_struct * vma)
950{
b69aee04 951 struct socket *sock = file->private_data;
1da177e4
LT
952
953 return sock->ops->mmap(file, sock, vma);
954}
955
20380731 956static int sock_close(struct inode *inode, struct file *filp)
1da177e4
LT
957{
958 /*
959 * It was possible the inode is NULL we were
960 * closing an unfinished socket.
961 */
962
963 if (!inode)
964 {
965 printk(KERN_DEBUG "sock_close: NULL inode\n");
966 return 0;
967 }
968 sock_fasync(-1, filp, 0);
969 sock_release(SOCKET_I(inode));
970 return 0;
971}
972
973/*
974 * Update the socket async list
975 *
976 * Fasync_list locking strategy.
977 *
978 * 1. fasync_list is modified only under process context socket lock
979 * i.e. under semaphore.
980 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
981 * or under socket lock.
982 * 3. fasync_list can be used from softirq context, so that
983 * modification under socket lock have to be enhanced with
984 * write_lock_bh(&sk->sk_callback_lock).
985 * --ANK (990710)
986 */
987
988static int sock_fasync(int fd, struct file *filp, int on)
989{
990 struct fasync_struct *fa, *fna=NULL, **prev;
991 struct socket *sock;
992 struct sock *sk;
993
994 if (on)
995 {
8b3a7005 996 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1da177e4
LT
997 if(fna==NULL)
998 return -ENOMEM;
999 }
1000
b69aee04 1001 sock = filp->private_data;
1da177e4
LT
1002
1003 if ((sk=sock->sk) == NULL) {
1004 kfree(fna);
1005 return -EINVAL;
1006 }
1007
1008 lock_sock(sk);
1009
1010 prev=&(sock->fasync_list);
1011
1012 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1013 if (fa->fa_file==filp)
1014 break;
1015
1016 if(on)
1017 {
1018 if(fa!=NULL)
1019 {
1020 write_lock_bh(&sk->sk_callback_lock);
1021 fa->fa_fd=fd;
1022 write_unlock_bh(&sk->sk_callback_lock);
1023
1024 kfree(fna);
1025 goto out;
1026 }
1027 fna->fa_file=filp;
1028 fna->fa_fd=fd;
1029 fna->magic=FASYNC_MAGIC;
1030 fna->fa_next=sock->fasync_list;
1031 write_lock_bh(&sk->sk_callback_lock);
1032 sock->fasync_list=fna;
1033 write_unlock_bh(&sk->sk_callback_lock);
1034 }
1035 else
1036 {
1037 if (fa!=NULL)
1038 {
1039 write_lock_bh(&sk->sk_callback_lock);
1040 *prev=fa->fa_next;
1041 write_unlock_bh(&sk->sk_callback_lock);
1042 kfree(fa);
1043 }
1044 }
1045
1046out:
1047 release_sock(sock->sk);
1048 return 0;
1049}
1050
1051/* This function may be called only under socket lock or callback_lock */
1052
1053int sock_wake_async(struct socket *sock, int how, int band)
1054{
1055 if (!sock || !sock->fasync_list)
1056 return -1;
1057 switch (how)
1058 {
1059 case 1:
1060
1061 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1062 break;
1063 goto call_kill;
1064 case 2:
1065 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1066 break;
1067 /* fall through */
1068 case 0:
1069 call_kill:
1070 __kill_fasync(sock->fasync_list, SIGIO, band);
1071 break;
1072 case 3:
1073 __kill_fasync(sock->fasync_list, SIGURG, band);
1074 }
1075 return 0;
1076}
1077
1078static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1079{
1080 int err;
1081 struct socket *sock;
1082
1083 /*
1084 * Check protocol is in range
1085 */
1086 if (family < 0 || family >= NPROTO)
1087 return -EAFNOSUPPORT;
1088 if (type < 0 || type >= SOCK_MAX)
1089 return -EINVAL;
1090
1091 /* Compatibility.
1092
1093 This uglymoron is moved from INET layer to here to avoid
1094 deadlock in module load.
1095 */
1096 if (family == PF_INET && type == SOCK_PACKET) {
1097 static int warned;
1098 if (!warned) {
1099 warned = 1;
1100 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1101 }
1102 family = PF_PACKET;
1103 }
1104
1105 err = security_socket_create(family, type, protocol, kern);
1106 if (err)
1107 return err;
1108
1109#if defined(CONFIG_KMOD)
1110 /* Attempt to load a protocol module if the find failed.
1111 *
1112 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1113 * requested real, full-featured networking support upon configuration.
1114 * Otherwise module support will break!
1115 */
1116 if (net_families[family]==NULL)
1117 {
1118 request_module("net-pf-%d",family);
1119 }
1120#endif
1121
1122 net_family_read_lock();
1123 if (net_families[family] == NULL) {
1124 err = -EAFNOSUPPORT;
1125 goto out;
1126 }
1127
1128/*
1129 * Allocate the socket and allow the family to set things up. if
1130 * the protocol is 0, the family is instructed to select an appropriate
1131 * default.
1132 */
1133
1134 if (!(sock = sock_alloc())) {
1135 printk(KERN_WARNING "socket: no more sockets\n");
1136 err = -ENFILE; /* Not exactly a match, but its the
1137 closest posix thing */
1138 goto out;
1139 }
1140
1141 sock->type = type;
1142
1143 /*
1144 * We will call the ->create function, that possibly is in a loadable
1145 * module, so we have to bump that loadable module refcnt first.
1146 */
1147 err = -EAFNOSUPPORT;
1148 if (!try_module_get(net_families[family]->owner))
1149 goto out_release;
1150
a79af59e
FF
1151 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1152 sock->ops = NULL;
1da177e4 1153 goto out_module_put;
a79af59e
FF
1154 }
1155
1da177e4
LT
1156 /*
1157 * Now to bump the refcnt of the [loadable] module that owns this
1158 * socket at sock_release time we decrement its refcnt.
1159 */
1160 if (!try_module_get(sock->ops->owner)) {
1161 sock->ops = NULL;
1162 goto out_module_put;
1163 }
1164 /*
1165 * Now that we're done with the ->create function, the [loadable]
1166 * module can have its refcnt decremented
1167 */
1168 module_put(net_families[family]->owner);
1169 *res = sock;
1170 security_socket_post_create(sock, family, type, protocol, kern);
1171
1172out:
1173 net_family_read_unlock();
1174 return err;
1175out_module_put:
1176 module_put(net_families[family]->owner);
1177out_release:
1178 sock_release(sock);
1179 goto out;
1180}
1181
1182int sock_create(int family, int type, int protocol, struct socket **res)
1183{
1184 return __sock_create(family, type, protocol, res, 0);
1185}
1186
1187int sock_create_kern(int family, int type, int protocol, struct socket **res)
1188{
1189 return __sock_create(family, type, protocol, res, 1);
1190}
1191
1192asmlinkage long sys_socket(int family, int type, int protocol)
1193{
1194 int retval;
1195 struct socket *sock;
1196
1197 retval = sock_create(family, type, protocol, &sock);
1198 if (retval < 0)
1199 goto out;
1200
1201 retval = sock_map_fd(sock);
1202 if (retval < 0)
1203 goto out_release;
1204
1205out:
1206 /* It may be already another descriptor 8) Not kernel problem. */
1207 return retval;
1208
1209out_release:
1210 sock_release(sock);
1211 return retval;
1212}
1213
1214/*
1215 * Create a pair of connected sockets.
1216 */
1217
1218asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1219{
1220 struct socket *sock1, *sock2;
1221 int fd1, fd2, err;
1222
1223 /*
1224 * Obtain the first socket and check if the underlying protocol
1225 * supports the socketpair call.
1226 */
1227
1228 err = sock_create(family, type, protocol, &sock1);
1229 if (err < 0)
1230 goto out;
1231
1232 err = sock_create(family, type, protocol, &sock2);
1233 if (err < 0)
1234 goto out_release_1;
1235
1236 err = sock1->ops->socketpair(sock1, sock2);
1237 if (err < 0)
1238 goto out_release_both;
1239
1240 fd1 = fd2 = -1;
1241
1242 err = sock_map_fd(sock1);
1243 if (err < 0)
1244 goto out_release_both;
1245 fd1 = err;
1246
1247 err = sock_map_fd(sock2);
1248 if (err < 0)
1249 goto out_close_1;
1250 fd2 = err;
1251
1252 /* fd1 and fd2 may be already another descriptors.
1253 * Not kernel problem.
1254 */
1255
1256 err = put_user(fd1, &usockvec[0]);
1257 if (!err)
1258 err = put_user(fd2, &usockvec[1]);
1259 if (!err)
1260 return 0;
1261
1262 sys_close(fd2);
1263 sys_close(fd1);
1264 return err;
1265
1266out_close_1:
1267 sock_release(sock2);
1268 sys_close(fd1);
1269 return err;
1270
1271out_release_both:
1272 sock_release(sock2);
1273out_release_1:
1274 sock_release(sock1);
1275out:
1276 return err;
1277}
1278
1279
1280/*
1281 * Bind a name to a socket. Nothing much to do here since it's
1282 * the protocol's responsibility to handle the local address.
1283 *
1284 * We move the socket address to kernel space before we call
1285 * the protocol layer (having also checked the address is ok).
1286 */
1287
1288asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1289{
1290 struct socket *sock;
1291 char address[MAX_SOCK_ADDR];
1292 int err;
1293
1294 if((sock = sockfd_lookup(fd,&err))!=NULL)
1295 {
1296 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1297 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1298 if (err) {
1299 sockfd_put(sock);
1300 return err;
1301 }
1302 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1303 }
1304 sockfd_put(sock);
1305 }
1306 return err;
1307}
1308
1309
1310/*
1311 * Perform a listen. Basically, we allow the protocol to do anything
1312 * necessary for a listen, and if that works, we mark the socket as
1313 * ready for listening.
1314 */
1315
1316int sysctl_somaxconn = SOMAXCONN;
1317
1318asmlinkage long sys_listen(int fd, int backlog)
1319{
1320 struct socket *sock;
1321 int err;
1322
1323 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1324 if ((unsigned) backlog > sysctl_somaxconn)
1325 backlog = sysctl_somaxconn;
1326
1327 err = security_socket_listen(sock, backlog);
1328 if (err) {
1329 sockfd_put(sock);
1330 return err;
1331 }
1332
1333 err=sock->ops->listen(sock, backlog);
1334 sockfd_put(sock);
1335 }
1336 return err;
1337}
1338
1339
1340/*
1341 * For accept, we attempt to create a new socket, set up the link
1342 * with the client, wake up the client, then return the new
1343 * connected fd. We collect the address of the connector in kernel
1344 * space and move it to user at the very end. This is unclean because
1345 * we open the socket then return an error.
1346 *
1347 * 1003.1g adds the ability to recvmsg() to query connection pending
1348 * status to recvmsg. We need to add that support in a way thats
1349 * clean when we restucture accept also.
1350 */
1351
1352asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1353{
1354 struct socket *sock, *newsock;
1355 int err, len;
1356 char address[MAX_SOCK_ADDR];
1357
1358 sock = sockfd_lookup(fd, &err);
1359 if (!sock)
1360 goto out;
1361
1362 err = -ENFILE;
1363 if (!(newsock = sock_alloc()))
1364 goto out_put;
1365
1366 newsock->type = sock->type;
1367 newsock->ops = sock->ops;
1368
1da177e4
LT
1369 /*
1370 * We don't need try_module_get here, as the listening socket (sock)
1371 * has the protocol module (sock->ops->owner) held.
1372 */
1373 __module_get(newsock->ops->owner);
1374
a79af59e
FF
1375 err = security_socket_accept(sock, newsock);
1376 if (err)
1377 goto out_release;
1378
1da177e4
LT
1379 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1380 if (err < 0)
1381 goto out_release;
1382
1383 if (upeer_sockaddr) {
1384 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1385 err = -ECONNABORTED;
1386 goto out_release;
1387 }
1388 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1389 if (err < 0)
1390 goto out_release;
1391 }
1392
1393 /* File flags are not inherited via accept() unlike another OSes. */
1394
1395 if ((err = sock_map_fd(newsock)) < 0)
1396 goto out_release;
1397
1398 security_socket_post_accept(sock, newsock);
1399
1400out_put:
1401 sockfd_put(sock);
1402out:
1403 return err;
1404out_release:
1405 sock_release(newsock);
1406 goto out_put;
1407}
1408
1409
1410/*
1411 * Attempt to connect to a socket with the server address. The address
1412 * is in user space so we verify it is OK and move it to kernel space.
1413 *
1414 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1415 * break bindings
1416 *
1417 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1418 * other SEQPACKET protocols that take time to connect() as it doesn't
1419 * include the -EINPROGRESS status for such sockets.
1420 */
1421
1422asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1423{
1424 struct socket *sock;
1425 char address[MAX_SOCK_ADDR];
1426 int err;
1427
1428 sock = sockfd_lookup(fd, &err);
1429 if (!sock)
1430 goto out;
1431 err = move_addr_to_kernel(uservaddr, addrlen, address);
1432 if (err < 0)
1433 goto out_put;
1434
1435 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1436 if (err)
1437 goto out_put;
1438
1439 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1440 sock->file->f_flags);
1441out_put:
1442 sockfd_put(sock);
1443out:
1444 return err;
1445}
1446
1447/*
1448 * Get the local address ('name') of a socket object. Move the obtained
1449 * name to user space.
1450 */
1451
1452asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1453{
1454 struct socket *sock;
1455 char address[MAX_SOCK_ADDR];
1456 int len, err;
1457
1458 sock = sockfd_lookup(fd, &err);
1459 if (!sock)
1460 goto out;
1461
1462 err = security_socket_getsockname(sock);
1463 if (err)
1464 goto out_put;
1465
1466 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1467 if (err)
1468 goto out_put;
1469 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1470
1471out_put:
1472 sockfd_put(sock);
1473out:
1474 return err;
1475}
1476
1477/*
1478 * Get the remote address ('name') of a socket object. Move the obtained
1479 * name to user space.
1480 */
1481
1482asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1483{
1484 struct socket *sock;
1485 char address[MAX_SOCK_ADDR];
1486 int len, err;
1487
1488 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1489 {
1490 err = security_socket_getpeername(sock);
1491 if (err) {
1492 sockfd_put(sock);
1493 return err;
1494 }
1495
1496 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1497 if (!err)
1498 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1499 sockfd_put(sock);
1500 }
1501 return err;
1502}
1503
1504/*
1505 * Send a datagram to a given address. We move the address into kernel
1506 * space and check the user space data area is readable before invoking
1507 * the protocol.
1508 */
1509
1510asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1511 struct sockaddr __user *addr, int addr_len)
1512{
1513 struct socket *sock;
1514 char address[MAX_SOCK_ADDR];
1515 int err;
1516 struct msghdr msg;
1517 struct iovec iov;
1518
1519 sock = sockfd_lookup(fd, &err);
1520 if (!sock)
1521 goto out;
1522 iov.iov_base=buff;
1523 iov.iov_len=len;
1524 msg.msg_name=NULL;
1525 msg.msg_iov=&iov;
1526 msg.msg_iovlen=1;
1527 msg.msg_control=NULL;
1528 msg.msg_controllen=0;
1529 msg.msg_namelen=0;
1530 if(addr)
1531 {
1532 err = move_addr_to_kernel(addr, addr_len, address);
1533 if (err < 0)
1534 goto out_put;
1535 msg.msg_name=address;
1536 msg.msg_namelen=addr_len;
1537 }
1538 if (sock->file->f_flags & O_NONBLOCK)
1539 flags |= MSG_DONTWAIT;
1540 msg.msg_flags = flags;
1541 err = sock_sendmsg(sock, &msg, len);
1542
1543out_put:
1544 sockfd_put(sock);
1545out:
1546 return err;
1547}
1548
1549/*
1550 * Send a datagram down a socket.
1551 */
1552
1553asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1554{
1555 return sys_sendto(fd, buff, len, flags, NULL, 0);
1556}
1557
1558/*
1559 * Receive a frame from the socket and optionally record the address of the
1560 * sender. We verify the buffers are writable and if needed move the
1561 * sender address from kernel to user space.
1562 */
1563
1564asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1565 struct sockaddr __user *addr, int __user *addr_len)
1566{
1567 struct socket *sock;
1568 struct iovec iov;
1569 struct msghdr msg;
1570 char address[MAX_SOCK_ADDR];
1571 int err,err2;
1572
1573 sock = sockfd_lookup(fd, &err);
1574 if (!sock)
1575 goto out;
1576
1577 msg.msg_control=NULL;
1578 msg.msg_controllen=0;
1579 msg.msg_iovlen=1;
1580 msg.msg_iov=&iov;
1581 iov.iov_len=size;
1582 iov.iov_base=ubuf;
1583 msg.msg_name=address;
1584 msg.msg_namelen=MAX_SOCK_ADDR;
1585 if (sock->file->f_flags & O_NONBLOCK)
1586 flags |= MSG_DONTWAIT;
1587 err=sock_recvmsg(sock, &msg, size, flags);
1588
1589 if(err >= 0 && addr != NULL)
1590 {
1591 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1592 if(err2<0)
1593 err=err2;
1594 }
1595 sockfd_put(sock);
1596out:
1597 return err;
1598}
1599
1600/*
1601 * Receive a datagram from a socket.
1602 */
1603
1604asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1605{
1606 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1607}
1608
1609/*
1610 * Set a socket option. Because we don't know the option lengths we have
1611 * to pass the user mode parameter for the protocols to sort out.
1612 */
1613
1614asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1615{
1616 int err;
1617 struct socket *sock;
1618
1619 if (optlen < 0)
1620 return -EINVAL;
1621
1622 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1623 {
1624 err = security_socket_setsockopt(sock,level,optname);
1625 if (err) {
1626 sockfd_put(sock);
1627 return err;
1628 }
1629
1630 if (level == SOL_SOCKET)
1631 err=sock_setsockopt(sock,level,optname,optval,optlen);
1632 else
1633 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1634 sockfd_put(sock);
1635 }
1636 return err;
1637}
1638
1639/*
1640 * Get a socket option. Because we don't know the option lengths we have
1641 * to pass a user mode parameter for the protocols to sort out.
1642 */
1643
1644asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1645{
1646 int err;
1647 struct socket *sock;
1648
1649 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1650 {
1651 err = security_socket_getsockopt(sock, level,
1652 optname);
1653 if (err) {
1654 sockfd_put(sock);
1655 return err;
1656 }
1657
1658 if (level == SOL_SOCKET)
1659 err=sock_getsockopt(sock,level,optname,optval,optlen);
1660 else
1661 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1662 sockfd_put(sock);
1663 }
1664 return err;
1665}
1666
1667
1668/*
1669 * Shutdown a socket.
1670 */
1671
1672asmlinkage long sys_shutdown(int fd, int how)
1673{
1674 int err;
1675 struct socket *sock;
1676
1677 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1678 {
1679 err = security_socket_shutdown(sock, how);
1680 if (err) {
1681 sockfd_put(sock);
1682 return err;
1683 }
1684
1685 err=sock->ops->shutdown(sock, how);
1686 sockfd_put(sock);
1687 }
1688 return err;
1689}
1690
1691/* A couple of helpful macros for getting the address of the 32/64 bit
1692 * fields which are the same type (int / unsigned) on our platforms.
1693 */
1694#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1695#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1696#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1697
1698
1699/*
1700 * BSD sendmsg interface
1701 */
1702
1703asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1704{
1705 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1706 struct socket *sock;
1707 char address[MAX_SOCK_ADDR];
1708 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
b9d717a7
AW
1709 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1710 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1711 /* 20 is size of ipv6_pktinfo */
1da177e4
LT
1712 unsigned char *ctl_buf = ctl;
1713 struct msghdr msg_sys;
1714 int err, ctl_len, iov_size, total_len;
1715
1716 err = -EFAULT;
1717 if (MSG_CMSG_COMPAT & flags) {
1718 if (get_compat_msghdr(&msg_sys, msg_compat))
1719 return -EFAULT;
1720 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1721 return -EFAULT;
1722
1723 sock = sockfd_lookup(fd, &err);
1724 if (!sock)
1725 goto out;
1726
1727 /* do not move before msg_sys is valid */
1728 err = -EMSGSIZE;
1729 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1730 goto out_put;
1731
1732 /* Check whether to allocate the iovec area*/
1733 err = -ENOMEM;
1734 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1735 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1736 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1737 if (!iov)
1738 goto out_put;
1739 }
1740
1741 /* This will also move the address data into kernel space */
1742 if (MSG_CMSG_COMPAT & flags) {
1743 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1744 } else
1745 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1746 if (err < 0)
1747 goto out_freeiov;
1748 total_len = err;
1749
1750 err = -ENOBUFS;
1751
1752 if (msg_sys.msg_controllen > INT_MAX)
1753 goto out_freeiov;
1754 ctl_len = msg_sys.msg_controllen;
1755 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
8920e8f9 1756 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
1da177e4
LT
1757 if (err)
1758 goto out_freeiov;
1759 ctl_buf = msg_sys.msg_control;
8920e8f9 1760 ctl_len = msg_sys.msg_controllen;
1da177e4
LT
1761 } else if (ctl_len) {
1762 if (ctl_len > sizeof(ctl))
1763 {
1764 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1765 if (ctl_buf == NULL)
1766 goto out_freeiov;
1767 }
1768 err = -EFAULT;
1769 /*
1770 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1771 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1772 * checking falls down on this.
1773 */
1774 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1775 goto out_freectl;
1776 msg_sys.msg_control = ctl_buf;
1777 }
1778 msg_sys.msg_flags = flags;
1779
1780 if (sock->file->f_flags & O_NONBLOCK)
1781 msg_sys.msg_flags |= MSG_DONTWAIT;
1782 err = sock_sendmsg(sock, &msg_sys, total_len);
1783
1784out_freectl:
1785 if (ctl_buf != ctl)
1786 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1787out_freeiov:
1788 if (iov != iovstack)
1789 sock_kfree_s(sock->sk, iov, iov_size);
1790out_put:
1791 sockfd_put(sock);
1792out:
1793 return err;
1794}
1795
1796/*
1797 * BSD recvmsg interface
1798 */
1799
1800asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1801{
1802 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1803 struct socket *sock;
1804 struct iovec iovstack[UIO_FASTIOV];
1805 struct iovec *iov=iovstack;
1806 struct msghdr msg_sys;
1807 unsigned long cmsg_ptr;
1808 int err, iov_size, total_len, len;
1809
1810 /* kernel mode address */
1811 char addr[MAX_SOCK_ADDR];
1812
1813 /* user mode address pointers */
1814 struct sockaddr __user *uaddr;
1815 int __user *uaddr_len;
1816
1817 if (MSG_CMSG_COMPAT & flags) {
1818 if (get_compat_msghdr(&msg_sys, msg_compat))
1819 return -EFAULT;
1820 } else
1821 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1822 return -EFAULT;
1823
1824 sock = sockfd_lookup(fd, &err);
1825 if (!sock)
1826 goto out;
1827
1828 err = -EMSGSIZE;
1829 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1830 goto out_put;
1831
1832 /* Check whether to allocate the iovec area*/
1833 err = -ENOMEM;
1834 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1835 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1836 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1837 if (!iov)
1838 goto out_put;
1839 }
1840
1841 /*
1842 * Save the user-mode address (verify_iovec will change the
1843 * kernel msghdr to use the kernel address space)
1844 */
1845
1846 uaddr = (void __user *) msg_sys.msg_name;
1847 uaddr_len = COMPAT_NAMELEN(msg);
1848 if (MSG_CMSG_COMPAT & flags) {
1849 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1850 } else
1851 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1852 if (err < 0)
1853 goto out_freeiov;
1854 total_len=err;
1855
1856 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1857 msg_sys.msg_flags = 0;
1858 if (MSG_CMSG_COMPAT & flags)
1859 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1860
1861 if (sock->file->f_flags & O_NONBLOCK)
1862 flags |= MSG_DONTWAIT;
1863 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1864 if (err < 0)
1865 goto out_freeiov;
1866 len = err;
1867
1868 if (uaddr != NULL) {
1869 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1870 if (err < 0)
1871 goto out_freeiov;
1872 }
37f7f421
DM
1873 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1874 COMPAT_FLAGS(msg));
1da177e4
LT
1875 if (err)
1876 goto out_freeiov;
1877 if (MSG_CMSG_COMPAT & flags)
1878 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1879 &msg_compat->msg_controllen);
1880 else
1881 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1882 &msg->msg_controllen);
1883 if (err)
1884 goto out_freeiov;
1885 err = len;
1886
1887out_freeiov:
1888 if (iov != iovstack)
1889 sock_kfree_s(sock->sk, iov, iov_size);
1890out_put:
1891 sockfd_put(sock);
1892out:
1893 return err;
1894}
1895
1896#ifdef __ARCH_WANT_SYS_SOCKETCALL
1897
1898/* Argument list sizes for sys_socketcall */
1899#define AL(x) ((x) * sizeof(unsigned long))
1900static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1901 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1902 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1903#undef AL
1904
1905/*
1906 * System call vectors.
1907 *
1908 * Argument checking cleaned up. Saved 20% in size.
1909 * This function doesn't need to set the kernel lock because
1910 * it is set by the callees.
1911 */
1912
1913asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1914{
1915 unsigned long a[6];
1916 unsigned long a0,a1;
1917 int err;
1918
1919 if(call<1||call>SYS_RECVMSG)
1920 return -EINVAL;
1921
1922 /* copy_from_user should be SMP safe. */
1923 if (copy_from_user(a, args, nargs[call]))
1924 return -EFAULT;
3ec3b2fb 1925
4bcff1b3 1926 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
3ec3b2fb
DW
1927 if (err)
1928 return err;
1929
1da177e4
LT
1930 a0=a[0];
1931 a1=a[1];
1932
1933 switch(call)
1934 {
1935 case SYS_SOCKET:
1936 err = sys_socket(a0,a1,a[2]);
1937 break;
1938 case SYS_BIND:
1939 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1940 break;
1941 case SYS_CONNECT:
1942 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1943 break;
1944 case SYS_LISTEN:
1945 err = sys_listen(a0,a1);
1946 break;
1947 case SYS_ACCEPT:
1948 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1949 break;
1950 case SYS_GETSOCKNAME:
1951 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1952 break;
1953 case SYS_GETPEERNAME:
1954 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1955 break;
1956 case SYS_SOCKETPAIR:
1957 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1958 break;
1959 case SYS_SEND:
1960 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1961 break;
1962 case SYS_SENDTO:
1963 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
1964 (struct sockaddr __user *)a[4], a[5]);
1965 break;
1966 case SYS_RECV:
1967 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
1968 break;
1969 case SYS_RECVFROM:
1970 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
1971 (struct sockaddr __user *)a[4], (int __user *)a[5]);
1972 break;
1973 case SYS_SHUTDOWN:
1974 err = sys_shutdown(a0,a1);
1975 break;
1976 case SYS_SETSOCKOPT:
1977 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
1978 break;
1979 case SYS_GETSOCKOPT:
1980 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
1981 break;
1982 case SYS_SENDMSG:
1983 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
1984 break;
1985 case SYS_RECVMSG:
1986 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
1987 break;
1988 default:
1989 err = -EINVAL;
1990 break;
1991 }
1992 return err;
1993}
1994
1995#endif /* __ARCH_WANT_SYS_SOCKETCALL */
1996
1997/*
1998 * This function is called by a protocol handler that wants to
1999 * advertise its address family, and have it linked into the
2000 * SOCKET module.
2001 */
2002
2003int sock_register(struct net_proto_family *ops)
2004{
2005 int err;
2006
2007 if (ops->family >= NPROTO) {
2008 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2009 return -ENOBUFS;
2010 }
2011 net_family_write_lock();
2012 err = -EEXIST;
2013 if (net_families[ops->family] == NULL) {
2014 net_families[ops->family]=ops;
2015 err = 0;
2016 }
2017 net_family_write_unlock();
2018 printk(KERN_INFO "NET: Registered protocol family %d\n",
2019 ops->family);
2020 return err;
2021}
2022
2023/*
2024 * This function is called by a protocol handler that wants to
2025 * remove its address family, and have it unlinked from the
2026 * SOCKET module.
2027 */
2028
2029int sock_unregister(int family)
2030{
2031 if (family < 0 || family >= NPROTO)
2032 return -1;
2033
2034 net_family_write_lock();
2035 net_families[family]=NULL;
2036 net_family_write_unlock();
2037 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2038 family);
2039 return 0;
2040}
2041
77d76ea3 2042static int __init sock_init(void)
1da177e4
LT
2043{
2044 /*
2045 * Initialize sock SLAB cache.
2046 */
2047
2048 sk_init();
2049
1da177e4
LT
2050 /*
2051 * Initialize skbuff SLAB cache
2052 */
2053 skb_init();
1da177e4
LT
2054
2055 /*
2056 * Initialize the protocols module.
2057 */
2058
2059 init_inodecache();
2060 register_filesystem(&sock_fs_type);
2061 sock_mnt = kern_mount(&sock_fs_type);
77d76ea3
AK
2062
2063 /* The real protocol initialization is performed in later initcalls.
1da177e4
LT
2064 */
2065
2066#ifdef CONFIG_NETFILTER
2067 netfilter_init();
2068#endif
cbeb321a
DM
2069
2070 return 0;
1da177e4
LT
2071}
2072
77d76ea3
AK
2073core_initcall(sock_init); /* early initcall */
2074
1da177e4
LT
2075#ifdef CONFIG_PROC_FS
2076void socket_seq_show(struct seq_file *seq)
2077{
2078 int cpu;
2079 int counter = 0;
2080
2081 for (cpu = 0; cpu < NR_CPUS; cpu++)
2082 counter += per_cpu(sockets_in_use, cpu);
2083
2084 /* It can be negative, by the way. 8) */
2085 if (counter < 0)
2086 counter = 0;
2087
2088 seq_printf(seq, "sockets: used %d\n", counter);
2089}
2090#endif /* CONFIG_PROC_FS */
2091
2092/* ABI emulation layers need these two */
2093EXPORT_SYMBOL(move_addr_to_kernel);
2094EXPORT_SYMBOL(move_addr_to_user);
2095EXPORT_SYMBOL(sock_create);
2096EXPORT_SYMBOL(sock_create_kern);
2097EXPORT_SYMBOL(sock_create_lite);
2098EXPORT_SYMBOL(sock_map_fd);
2099EXPORT_SYMBOL(sock_recvmsg);
2100EXPORT_SYMBOL(sock_register);
2101EXPORT_SYMBOL(sock_release);
2102EXPORT_SYMBOL(sock_sendmsg);
2103EXPORT_SYMBOL(sock_unregister);
2104EXPORT_SYMBOL(sock_wake_async);
2105EXPORT_SYMBOL(sockfd_lookup);
2106EXPORT_SYMBOL(kernel_sendmsg);
2107EXPORT_SYMBOL(kernel_recvmsg);