include/net/raw.h: Convert raw_seq_private macro to inline
[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)
89bddce5 45 * Tigran Aivazian : Made listen(2) backlog sanity checks
1da177e4
LT
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
89bddce5 56 * paradigm.
1da177e4
LT
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
1da177e4 61#include <linux/mm.h>
1da177e4
LT
62#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
aaca0bdc 66#include <linux/thread_info.h>
55737fda 67#include <linux/rcupdate.h>
1da177e4
LT
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
4a3e2f71 71#include <linux/mutex.h>
1da177e4
LT
72#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
20380731
ACM
74#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
1da177e4
LT
76#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
1da177e4
LT
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
3ec3b2fb 86#include <linux/audit.h>
d86b5e0e 87#include <linux/wireless.h>
1b8d7ae4 88#include <linux/nsproxy.h>
1fd7317d 89#include <linux/magic.h>
5a0e3ad6 90#include <linux/slab.h>
1da177e4
LT
91
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
87de87d5 96#include <net/wext.h>
f8451725 97#include <net/cls_cgroup.h>
1da177e4
LT
98
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
6b96018b
AB
102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
6b96018b
AB
105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
1da177e4 108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
027445c3
BP
109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
89bddce5 113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
1da177e4
LT
114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
89bddce5 118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
89bbfc95
SP
119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
89bddce5 121 unsigned int cmd, unsigned long arg);
89bbfc95 122#endif
1da177e4 123static int sock_fasync(int fd, struct file *filp, int on);
1da177e4
LT
124static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
9c55e01c 126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
c6d409cf 127 struct pipe_inode_info *pipe, size_t len,
9c55e01c 128 unsigned int flags);
1da177e4 129
1da177e4
LT
130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
da7071d7 135static const struct file_operations socket_file_ops = {
1da177e4
LT
136 .owner = THIS_MODULE,
137 .llseek = no_llseek,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
140 .poll = sock_poll,
141 .unlocked_ioctl = sock_ioctl,
89bbfc95
SP
142#ifdef CONFIG_COMPAT
143 .compat_ioctl = compat_sock_ioctl,
144#endif
1da177e4
LT
145 .mmap = sock_mmap,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
5274f052
JA
149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
9c55e01c 151 .splice_read = sock_splice_read,
1da177e4
LT
152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
1da177e4 158static DEFINE_SPINLOCK(net_family_lock);
f0fd27d4 159static const struct net_proto_family *net_families[NPROTO] __read_mostly;
1da177e4 160
1da177e4
LT
161/*
162 * Statistics counters of the socket lists
163 */
164
c6d409cf 165static DEFINE_PER_CPU(int, sockets_in_use);
1da177e4
LT
166
167/*
89bddce5
SH
168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
1da177e4
LT
171 */
172
1da177e4
LT
173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
230b1839 184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
1da177e4 185{
230b1839 186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
1da177e4 187 return -EINVAL;
89bddce5 188 if (ulen == 0)
1da177e4 189 return 0;
89bddce5 190 if (copy_from_user(kaddr, uaddr, ulen))
1da177e4 191 return -EFAULT;
3ec3b2fb 192 return audit_sockaddr(ulen, kaddr);
1da177e4
LT
193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
89bddce5 211
230b1839 212int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
89bddce5 213 int __user *ulen)
1da177e4
LT
214{
215 int err;
216 int len;
217
89bddce5
SH
218 err = get_user(len, ulen);
219 if (err)
1da177e4 220 return err;
89bddce5
SH
221 if (len > klen)
222 len = klen;
230b1839 223 if (len < 0 || len > sizeof(struct sockaddr_storage))
1da177e4 224 return -EINVAL;
89bddce5 225 if (len) {
d6fe3945
SG
226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
89bddce5 228 if (copy_to_user(uaddr, kaddr, len))
1da177e4
LT
229 return -EFAULT;
230 }
231 /*
89bddce5
SH
232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
1da177e4
LT
234 */
235 return __put_user(klen, ulen);
236}
237
e18b890b 238static struct kmem_cache *sock_inode_cachep __read_mostly;
1da177e4
LT
239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
89bddce5 243
e94b1766 244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
1da177e4
LT
245 if (!ei)
246 return NULL;
43815482
ED
247 ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
248 if (!ei->socket.wq) {
249 kmem_cache_free(sock_inode_cachep, ei);
250 return NULL;
251 }
252 init_waitqueue_head(&ei->socket.wq->wait);
253 ei->socket.wq->fasync_list = NULL;
89bddce5 254
1da177e4
LT
255 ei->socket.state = SS_UNCONNECTED;
256 ei->socket.flags = 0;
257 ei->socket.ops = NULL;
258 ei->socket.sk = NULL;
259 ei->socket.file = NULL;
1da177e4
LT
260
261 return &ei->vfs_inode;
262}
263
43815482
ED
264
265static void wq_free_rcu(struct rcu_head *head)
266{
267 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
268
269 kfree(wq);
270}
271
1da177e4
LT
272static void sock_destroy_inode(struct inode *inode)
273{
43815482
ED
274 struct socket_alloc *ei;
275
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
277 call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
278 kmem_cache_free(sock_inode_cachep, ei);
1da177e4
LT
279}
280
51cc5068 281static void init_once(void *foo)
1da177e4 282{
89bddce5 283 struct socket_alloc *ei = (struct socket_alloc *)foo;
1da177e4 284
a35afb83 285 inode_init_once(&ei->vfs_inode);
1da177e4 286}
89bddce5 287
1da177e4
LT
288static int init_inodecache(void)
289{
290 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
89bddce5
SH
291 sizeof(struct socket_alloc),
292 0,
293 (SLAB_HWCACHE_ALIGN |
294 SLAB_RECLAIM_ACCOUNT |
295 SLAB_MEM_SPREAD),
20c2df83 296 init_once);
1da177e4
LT
297 if (sock_inode_cachep == NULL)
298 return -ENOMEM;
299 return 0;
300}
301
b87221de 302static const struct super_operations sockfs_ops = {
c6d409cf
ED
303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
1da177e4
LT
306};
307
454e2398 308static int sockfs_get_sb(struct file_system_type *fs_type,
89bddce5
SH
309 int flags, const char *dev_name, void *data,
310 struct vfsmount *mnt)
1da177e4 311{
454e2398
DH
312 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
313 mnt);
1da177e4
LT
314}
315
ba89966c 316static struct vfsmount *sock_mnt __read_mostly;
1da177e4
LT
317
318static struct file_system_type sock_fs_type = {
319 .name = "sockfs",
320 .get_sb = sockfs_get_sb,
321 .kill_sb = kill_anon_super,
322};
89bddce5 323
c23fbb6b
ED
324/*
325 * sockfs_dname() is called from d_path().
326 */
327static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
328{
329 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
330 dentry->d_inode->i_ino);
331}
332
3ba13d17 333static const struct dentry_operations sockfs_dentry_operations = {
c23fbb6b 334 .d_dname = sockfs_dname,
1da177e4
LT
335};
336
337/*
338 * Obtains the first available file descriptor and sets it up for use.
339 *
39d8c1b6
DM
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
1da177e4
LT
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
347 *
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
352 */
353
7cbe66b6 354static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
1da177e4 355{
7cbe66b6 356 struct qstr name = { .name = "" };
2c48b9c4 357 struct path path;
7cbe66b6 358 struct file *file;
1da177e4 359 int fd;
1da177e4 360
a677a039 361 fd = get_unused_fd_flags(flags);
7cbe66b6
AV
362 if (unlikely(fd < 0))
363 return fd;
1da177e4 364
2c48b9c4
AV
365 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
366 if (unlikely(!path.dentry)) {
7cbe66b6 367 put_unused_fd(fd);
39d8c1b6 368 return -ENOMEM;
7cbe66b6 369 }
2c48b9c4 370 path.mnt = mntget(sock_mnt);
39d8c1b6 371
2c48b9c4 372 path.dentry->d_op = &sockfs_dentry_operations;
2c48b9c4 373 d_instantiate(path.dentry, SOCK_INODE(sock));
cc3808f8 374 SOCK_INODE(sock)->i_fop = &socket_file_ops;
39d8c1b6 375
2c48b9c4 376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
ce8d2cdf 377 &socket_file_ops);
cc3808f8
AV
378 if (unlikely(!file)) {
379 /* drop dentry, keep inode */
380 atomic_inc(&path.dentry->d_inode->i_count);
2c48b9c4 381 path_put(&path);
cc3808f8
AV
382 put_unused_fd(fd);
383 return -ENFILE;
384 }
385
386 sock->file = file;
77d27200 387 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
39d8c1b6
DM
388 file->f_pos = 0;
389 file->private_data = sock;
1da177e4 390
7cbe66b6
AV
391 *f = file;
392 return fd;
39d8c1b6
DM
393}
394
a677a039 395int sock_map_fd(struct socket *sock, int flags)
39d8c1b6
DM
396{
397 struct file *newfile;
7cbe66b6 398 int fd = sock_alloc_file(sock, &newfile, flags);
39d8c1b6 399
7cbe66b6 400 if (likely(fd >= 0))
39d8c1b6 401 fd_install(fd, newfile);
7cbe66b6 402
1da177e4
LT
403 return fd;
404}
c6d409cf 405EXPORT_SYMBOL(sock_map_fd);
1da177e4 406
6cb153ca
BL
407static struct socket *sock_from_file(struct file *file, int *err)
408{
6cb153ca
BL
409 if (file->f_op == &socket_file_ops)
410 return file->private_data; /* set in sock_map_fd */
411
23bb80d2
ED
412 *err = -ENOTSOCK;
413 return NULL;
6cb153ca
BL
414}
415
1da177e4 416/**
c6d409cf 417 * sockfd_lookup - Go from a file number to its socket slot
1da177e4
LT
418 * @fd: file handle
419 * @err: pointer to an error code return
420 *
421 * The file handle passed in is locked and the socket it is bound
422 * too is returned. If an error occurs the err pointer is overwritten
423 * with a negative errno code and NULL is returned. The function checks
424 * for both invalid handles and passing a handle which is not a socket.
425 *
426 * On a success the socket object pointer is returned.
427 */
428
429struct socket *sockfd_lookup(int fd, int *err)
430{
431 struct file *file;
1da177e4
LT
432 struct socket *sock;
433
89bddce5
SH
434 file = fget(fd);
435 if (!file) {
1da177e4
LT
436 *err = -EBADF;
437 return NULL;
438 }
89bddce5 439
6cb153ca
BL
440 sock = sock_from_file(file, err);
441 if (!sock)
1da177e4 442 fput(file);
6cb153ca
BL
443 return sock;
444}
c6d409cf 445EXPORT_SYMBOL(sockfd_lookup);
1da177e4 446
6cb153ca
BL
447static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
448{
449 struct file *file;
450 struct socket *sock;
451
3672558c 452 *err = -EBADF;
6cb153ca
BL
453 file = fget_light(fd, fput_needed);
454 if (file) {
455 sock = sock_from_file(file, err);
456 if (sock)
457 return sock;
458 fput_light(file, *fput_needed);
1da177e4 459 }
6cb153ca 460 return NULL;
1da177e4
LT
461}
462
463/**
464 * sock_alloc - allocate a socket
89bddce5 465 *
1da177e4
LT
466 * Allocate a new inode and socket object. The two are bound together
467 * and initialised. The socket is then returned. If we are out of inodes
468 * NULL is returned.
469 */
470
471static struct socket *sock_alloc(void)
472{
89bddce5
SH
473 struct inode *inode;
474 struct socket *sock;
1da177e4
LT
475
476 inode = new_inode(sock_mnt->mnt_sb);
477 if (!inode)
478 return NULL;
479
480 sock = SOCKET_I(inode);
481
29a020d3 482 kmemcheck_annotate_bitfield(sock, type);
89bddce5 483 inode->i_mode = S_IFSOCK | S_IRWXUGO;
8192b0c4
DH
484 inode->i_uid = current_fsuid();
485 inode->i_gid = current_fsgid();
1da177e4 486
4e69489a 487 percpu_add(sockets_in_use, 1);
1da177e4
LT
488 return sock;
489}
490
491/*
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
495 */
89bddce5 496
1da177e4
LT
497static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
498{
499 return -ENXIO;
500}
501
4b6f5d20 502const struct file_operations bad_sock_fops = {
1da177e4
LT
503 .owner = THIS_MODULE,
504 .open = sock_no_open,
505};
506
507/**
508 * sock_release - close a socket
509 * @sock: socket to close
510 *
511 * The socket is released from the protocol stack if it has a release
512 * callback, and the inode is then released if the socket is bound to
89bddce5 513 * an inode not a file.
1da177e4 514 */
89bddce5 515
1da177e4
LT
516void sock_release(struct socket *sock)
517{
518 if (sock->ops) {
519 struct module *owner = sock->ops->owner;
520
521 sock->ops->release(sock);
522 sock->ops = NULL;
523 module_put(owner);
524 }
525
43815482 526 if (sock->wq->fasync_list)
1da177e4
LT
527 printk(KERN_ERR "sock_release: fasync list not empty!\n");
528
4e69489a 529 percpu_sub(sockets_in_use, 1);
1da177e4
LT
530 if (!sock->file) {
531 iput(SOCK_INODE(sock));
532 return;
533 }
89bddce5 534 sock->file = NULL;
1da177e4 535}
c6d409cf 536EXPORT_SYMBOL(sock_release);
1da177e4 537
2244d07b 538int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
20d49473 539{
2244d07b 540 *tx_flags = 0;
20d49473 541 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
2244d07b 542 *tx_flags |= SKBTX_HW_TSTAMP;
20d49473 543 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
2244d07b 544 *tx_flags |= SKBTX_SW_TSTAMP;
20d49473
PO
545 return 0;
546}
547EXPORT_SYMBOL(sock_tx_timestamp);
548
89bddce5 549static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT
550 struct msghdr *msg, size_t size)
551{
552 struct sock_iocb *si = kiocb_to_siocb(iocb);
553 int err;
554
f8451725
HX
555 sock_update_classid(sock->sk);
556
1da177e4
LT
557 si->sock = sock;
558 si->scm = NULL;
559 si->msg = msg;
560 si->size = size;
561
562 err = security_socket_sendmsg(sock, msg, size);
563 if (err)
564 return err;
565
566 return sock->ops->sendmsg(iocb, sock, msg, size);
567}
568
569int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
570{
571 struct kiocb iocb;
572 struct sock_iocb siocb;
573 int ret;
574
575 init_sync_kiocb(&iocb, NULL);
576 iocb.private = &siocb;
577 ret = __sock_sendmsg(&iocb, sock, msg, size);
578 if (-EIOCBQUEUED == ret)
579 ret = wait_on_sync_kiocb(&iocb);
580 return ret;
581}
c6d409cf 582EXPORT_SYMBOL(sock_sendmsg);
1da177e4
LT
583
584int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
585 struct kvec *vec, size_t num, size_t size)
586{
587 mm_segment_t oldfs = get_fs();
588 int result;
589
590 set_fs(KERNEL_DS);
591 /*
592 * the following is safe, since for compiler definitions of kvec and
593 * iovec are identical, yielding the same in-core layout and alignment
594 */
89bddce5 595 msg->msg_iov = (struct iovec *)vec;
1da177e4
LT
596 msg->msg_iovlen = num;
597 result = sock_sendmsg(sock, msg, size);
598 set_fs(oldfs);
599 return result;
600}
c6d409cf 601EXPORT_SYMBOL(kernel_sendmsg);
1da177e4 602
20d49473
PO
603static int ktime2ts(ktime_t kt, struct timespec *ts)
604{
605 if (kt.tv64) {
606 *ts = ktime_to_timespec(kt);
607 return 1;
608 } else {
609 return 0;
610 }
611}
612
92f37fd2
ED
613/*
614 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
615 */
616void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
617 struct sk_buff *skb)
618{
20d49473
PO
619 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
620 struct timespec ts[3];
621 int empty = 1;
622 struct skb_shared_hwtstamps *shhwtstamps =
623 skb_hwtstamps(skb);
624
625 /* Race occurred between timestamp enabling and packet
626 receiving. Fill in the current time for now. */
627 if (need_software_tstamp && skb->tstamp.tv64 == 0)
628 __net_timestamp(skb);
629
630 if (need_software_tstamp) {
631 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
632 struct timeval tv;
633 skb_get_timestamp(skb, &tv);
634 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
635 sizeof(tv), &tv);
636 } else {
842509b8 637 skb_get_timestampns(skb, &ts[0]);
20d49473 638 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
842509b8 639 sizeof(ts[0]), &ts[0]);
20d49473
PO
640 }
641 }
642
643
644 memset(ts, 0, sizeof(ts));
645 if (skb->tstamp.tv64 &&
646 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
647 skb_get_timestampns(skb, ts + 0);
648 empty = 0;
649 }
650 if (shhwtstamps) {
651 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
652 ktime2ts(shhwtstamps->syststamp, ts + 1))
653 empty = 0;
654 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
655 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
656 empty = 0;
92f37fd2 657 }
20d49473
PO
658 if (!empty)
659 put_cmsg(msg, SOL_SOCKET,
660 SCM_TIMESTAMPING, sizeof(ts), &ts);
92f37fd2 661}
7c81fd8b
ACM
662EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
663
3b885787
NH
664inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
665{
666 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
667 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
668 sizeof(__u32), &skb->dropcount);
669}
670
767dd033 671void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
3b885787
NH
672 struct sk_buff *skb)
673{
674 sock_recv_timestamp(msg, sk, skb);
675 sock_recv_drops(msg, sk, skb);
676}
767dd033 677EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
3b885787 678
a2e27255
ACM
679static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
680 struct msghdr *msg, size_t size, int flags)
1da177e4 681{
1da177e4
LT
682 struct sock_iocb *si = kiocb_to_siocb(iocb);
683
f8451725
HX
684 sock_update_classid(sock->sk);
685
1da177e4
LT
686 si->sock = sock;
687 si->scm = NULL;
688 si->msg = msg;
689 si->size = size;
690 si->flags = flags;
691
1da177e4
LT
692 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
693}
694
a2e27255
ACM
695static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
696 struct msghdr *msg, size_t size, int flags)
697{
698 int err = security_socket_recvmsg(sock, msg, size, flags);
699
700 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
701}
702
89bddce5 703int sock_recvmsg(struct socket *sock, struct msghdr *msg,
1da177e4
LT
704 size_t size, int flags)
705{
706 struct kiocb iocb;
707 struct sock_iocb siocb;
708 int ret;
709
89bddce5 710 init_sync_kiocb(&iocb, NULL);
1da177e4
LT
711 iocb.private = &siocb;
712 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
713 if (-EIOCBQUEUED == ret)
714 ret = wait_on_sync_kiocb(&iocb);
715 return ret;
716}
c6d409cf 717EXPORT_SYMBOL(sock_recvmsg);
1da177e4 718
a2e27255
ACM
719static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
720 size_t size, int flags)
721{
722 struct kiocb iocb;
723 struct sock_iocb siocb;
724 int ret;
725
726 init_sync_kiocb(&iocb, NULL);
727 iocb.private = &siocb;
728 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
729 if (-EIOCBQUEUED == ret)
730 ret = wait_on_sync_kiocb(&iocb);
731 return ret;
732}
733
89bddce5
SH
734int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
735 struct kvec *vec, size_t num, size_t size, int flags)
1da177e4
LT
736{
737 mm_segment_t oldfs = get_fs();
738 int result;
739
740 set_fs(KERNEL_DS);
741 /*
742 * the following is safe, since for compiler definitions of kvec and
743 * iovec are identical, yielding the same in-core layout and alignment
744 */
89bddce5 745 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
1da177e4
LT
746 result = sock_recvmsg(sock, msg, size, flags);
747 set_fs(oldfs);
748 return result;
749}
c6d409cf 750EXPORT_SYMBOL(kernel_recvmsg);
1da177e4
LT
751
752static void sock_aio_dtor(struct kiocb *iocb)
753{
754 kfree(iocb->private);
755}
756
ce1d4d3e
CH
757static ssize_t sock_sendpage(struct file *file, struct page *page,
758 int offset, size_t size, loff_t *ppos, int more)
1da177e4 759{
1da177e4
LT
760 struct socket *sock;
761 int flags;
762
ce1d4d3e
CH
763 sock = file->private_data;
764
765 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
766 if (more)
767 flags |= MSG_MORE;
768
e6949583 769 return kernel_sendpage(sock, page, offset, size, flags);
ce1d4d3e 770}
1da177e4 771
9c55e01c 772static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
c6d409cf 773 struct pipe_inode_info *pipe, size_t len,
9c55e01c
JA
774 unsigned int flags)
775{
776 struct socket *sock = file->private_data;
777
997b37da
RDC
778 if (unlikely(!sock->ops->splice_read))
779 return -EINVAL;
780
f8451725
HX
781 sock_update_classid(sock->sk);
782
9c55e01c
JA
783 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
784}
785
ce1d4d3e 786static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
89bddce5 787 struct sock_iocb *siocb)
ce1d4d3e
CH
788{
789 if (!is_sync_kiocb(iocb)) {
790 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
791 if (!siocb)
792 return NULL;
1da177e4
LT
793 iocb->ki_dtor = sock_aio_dtor;
794 }
1da177e4 795
ce1d4d3e 796 siocb->kiocb = iocb;
ce1d4d3e
CH
797 iocb->private = siocb;
798 return siocb;
1da177e4
LT
799}
800
ce1d4d3e 801static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
027445c3
BP
802 struct file *file, const struct iovec *iov,
803 unsigned long nr_segs)
ce1d4d3e
CH
804{
805 struct socket *sock = file->private_data;
806 size_t size = 0;
807 int i;
1da177e4 808
89bddce5
SH
809 for (i = 0; i < nr_segs; i++)
810 size += iov[i].iov_len;
1da177e4 811
ce1d4d3e
CH
812 msg->msg_name = NULL;
813 msg->msg_namelen = 0;
814 msg->msg_control = NULL;
815 msg->msg_controllen = 0;
89bddce5 816 msg->msg_iov = (struct iovec *)iov;
ce1d4d3e
CH
817 msg->msg_iovlen = nr_segs;
818 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
819
820 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
821}
822
027445c3
BP
823static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
824 unsigned long nr_segs, loff_t pos)
ce1d4d3e
CH
825{
826 struct sock_iocb siocb, *x;
827
1da177e4
LT
828 if (pos != 0)
829 return -ESPIPE;
027445c3
BP
830
831 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
1da177e4
LT
832 return 0;
833
027445c3
BP
834
835 x = alloc_sock_iocb(iocb, &siocb);
ce1d4d3e
CH
836 if (!x)
837 return -ENOMEM;
027445c3 838 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
1da177e4
LT
839}
840
ce1d4d3e 841static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
027445c3
BP
842 struct file *file, const struct iovec *iov,
843 unsigned long nr_segs)
1da177e4 844{
ce1d4d3e
CH
845 struct socket *sock = file->private_data;
846 size_t size = 0;
847 int i;
1da177e4 848
89bddce5
SH
849 for (i = 0; i < nr_segs; i++)
850 size += iov[i].iov_len;
1da177e4 851
ce1d4d3e
CH
852 msg->msg_name = NULL;
853 msg->msg_namelen = 0;
854 msg->msg_control = NULL;
855 msg->msg_controllen = 0;
89bddce5 856 msg->msg_iov = (struct iovec *)iov;
ce1d4d3e
CH
857 msg->msg_iovlen = nr_segs;
858 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
859 if (sock->type == SOCK_SEQPACKET)
860 msg->msg_flags |= MSG_EOR;
1da177e4 861
ce1d4d3e 862 return __sock_sendmsg(iocb, sock, msg, size);
1da177e4
LT
863}
864
027445c3
BP
865static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
866 unsigned long nr_segs, loff_t pos)
ce1d4d3e
CH
867{
868 struct sock_iocb siocb, *x;
1da177e4 869
ce1d4d3e
CH
870 if (pos != 0)
871 return -ESPIPE;
027445c3 872
027445c3 873 x = alloc_sock_iocb(iocb, &siocb);
ce1d4d3e
CH
874 if (!x)
875 return -ENOMEM;
1da177e4 876
027445c3 877 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
1da177e4
LT
878}
879
1da177e4
LT
880/*
881 * Atomic setting of ioctl hooks to avoid race
882 * with module unload.
883 */
884
4a3e2f71 885static DEFINE_MUTEX(br_ioctl_mutex);
c6d409cf 886static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1da177e4 887
881d966b 888void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1da177e4 889{
4a3e2f71 890 mutex_lock(&br_ioctl_mutex);
1da177e4 891 br_ioctl_hook = hook;
4a3e2f71 892 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
893}
894EXPORT_SYMBOL(brioctl_set);
895
4a3e2f71 896static DEFINE_MUTEX(vlan_ioctl_mutex);
881d966b 897static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1da177e4 898
881d966b 899void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1da177e4 900{
4a3e2f71 901 mutex_lock(&vlan_ioctl_mutex);
1da177e4 902 vlan_ioctl_hook = hook;
4a3e2f71 903 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
904}
905EXPORT_SYMBOL(vlan_ioctl_set);
906
4a3e2f71 907static DEFINE_MUTEX(dlci_ioctl_mutex);
89bddce5 908static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1da177e4 909
89bddce5 910void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1da177e4 911{
4a3e2f71 912 mutex_lock(&dlci_ioctl_mutex);
1da177e4 913 dlci_ioctl_hook = hook;
4a3e2f71 914 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
915}
916EXPORT_SYMBOL(dlci_ioctl_set);
917
6b96018b
AB
918static long sock_do_ioctl(struct net *net, struct socket *sock,
919 unsigned int cmd, unsigned long arg)
920{
921 int err;
922 void __user *argp = (void __user *)arg;
923
924 err = sock->ops->ioctl(sock, cmd, arg);
925
926 /*
927 * If this ioctl is unknown try to hand it down
928 * to the NIC driver.
929 */
930 if (err == -ENOIOCTLCMD)
931 err = dev_ioctl(net, cmd, argp);
932
933 return err;
934}
935
1da177e4
LT
936/*
937 * With an ioctl, arg may well be a user mode pointer, but we don't know
938 * what to do with it - that's up to the protocol still.
939 */
940
941static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
942{
943 struct socket *sock;
881d966b 944 struct sock *sk;
1da177e4
LT
945 void __user *argp = (void __user *)arg;
946 int pid, err;
881d966b 947 struct net *net;
1da177e4 948
b69aee04 949 sock = file->private_data;
881d966b 950 sk = sock->sk;
3b1e0a65 951 net = sock_net(sk);
1da177e4 952 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
881d966b 953 err = dev_ioctl(net, cmd, argp);
1da177e4 954 } else
3d23e349 955#ifdef CONFIG_WEXT_CORE
1da177e4 956 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
881d966b 957 err = dev_ioctl(net, cmd, argp);
1da177e4 958 } else
3d23e349 959#endif
89bddce5 960 switch (cmd) {
1da177e4
LT
961 case FIOSETOWN:
962 case SIOCSPGRP:
963 err = -EFAULT;
964 if (get_user(pid, (int __user *)argp))
965 break;
966 err = f_setown(sock->file, pid, 1);
967 break;
968 case FIOGETOWN:
969 case SIOCGPGRP:
609d7fa9 970 err = put_user(f_getown(sock->file),
89bddce5 971 (int __user *)argp);
1da177e4
LT
972 break;
973 case SIOCGIFBR:
974 case SIOCSIFBR:
975 case SIOCBRADDBR:
976 case SIOCBRDELBR:
977 err = -ENOPKG;
978 if (!br_ioctl_hook)
979 request_module("bridge");
980
4a3e2f71 981 mutex_lock(&br_ioctl_mutex);
89bddce5 982 if (br_ioctl_hook)
881d966b 983 err = br_ioctl_hook(net, cmd, argp);
4a3e2f71 984 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
985 break;
986 case SIOCGIFVLAN:
987 case SIOCSIFVLAN:
988 err = -ENOPKG;
989 if (!vlan_ioctl_hook)
990 request_module("8021q");
991
4a3e2f71 992 mutex_lock(&vlan_ioctl_mutex);
1da177e4 993 if (vlan_ioctl_hook)
881d966b 994 err = vlan_ioctl_hook(net, argp);
4a3e2f71 995 mutex_unlock(&vlan_ioctl_mutex);
1da177e4 996 break;
1da177e4
LT
997 case SIOCADDDLCI:
998 case SIOCDELDLCI:
999 err = -ENOPKG;
1000 if (!dlci_ioctl_hook)
1001 request_module("dlci");
1002
7512cbf6
PE
1003 mutex_lock(&dlci_ioctl_mutex);
1004 if (dlci_ioctl_hook)
1da177e4 1005 err = dlci_ioctl_hook(cmd, argp);
7512cbf6 1006 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
1007 break;
1008 default:
6b96018b 1009 err = sock_do_ioctl(net, sock, cmd, arg);
1da177e4 1010 break;
89bddce5 1011 }
1da177e4
LT
1012 return err;
1013}
1014
1015int sock_create_lite(int family, int type, int protocol, struct socket **res)
1016{
1017 int err;
1018 struct socket *sock = NULL;
89bddce5 1019
1da177e4
LT
1020 err = security_socket_create(family, type, protocol, 1);
1021 if (err)
1022 goto out;
1023
1024 sock = sock_alloc();
1025 if (!sock) {
1026 err = -ENOMEM;
1027 goto out;
1028 }
1029
1da177e4 1030 sock->type = type;
7420ed23
VY
1031 err = security_socket_post_create(sock, family, type, protocol, 1);
1032 if (err)
1033 goto out_release;
1034
1da177e4
LT
1035out:
1036 *res = sock;
1037 return err;
7420ed23
VY
1038out_release:
1039 sock_release(sock);
1040 sock = NULL;
1041 goto out;
1da177e4 1042}
c6d409cf 1043EXPORT_SYMBOL(sock_create_lite);
1da177e4
LT
1044
1045/* No kernel lock held - perfect */
89bddce5 1046static unsigned int sock_poll(struct file *file, poll_table *wait)
1da177e4
LT
1047{
1048 struct socket *sock;
1049
1050 /*
89bddce5 1051 * We can't return errors to poll, so it's either yes or no.
1da177e4 1052 */
b69aee04 1053 sock = file->private_data;
1da177e4
LT
1054 return sock->ops->poll(file, sock, wait);
1055}
1056
89bddce5 1057static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1da177e4 1058{
b69aee04 1059 struct socket *sock = file->private_data;
1da177e4
LT
1060
1061 return sock->ops->mmap(file, sock, vma);
1062}
1063
20380731 1064static int sock_close(struct inode *inode, struct file *filp)
1da177e4
LT
1065{
1066 /*
89bddce5
SH
1067 * It was possible the inode is NULL we were
1068 * closing an unfinished socket.
1da177e4
LT
1069 */
1070
89bddce5 1071 if (!inode) {
1da177e4
LT
1072 printk(KERN_DEBUG "sock_close: NULL inode\n");
1073 return 0;
1074 }
1da177e4
LT
1075 sock_release(SOCKET_I(inode));
1076 return 0;
1077}
1078
1079/*
1080 * Update the socket async list
1081 *
1082 * Fasync_list locking strategy.
1083 *
1084 * 1. fasync_list is modified only under process context socket lock
1085 * i.e. under semaphore.
1086 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
989a2979 1087 * or under socket lock
1da177e4
LT
1088 */
1089
1090static int sock_fasync(int fd, struct file *filp, int on)
1091{
989a2979
ED
1092 struct socket *sock = filp->private_data;
1093 struct sock *sk = sock->sk;
1da177e4 1094
989a2979 1095 if (sk == NULL)
1da177e4 1096 return -EINVAL;
1da177e4
LT
1097
1098 lock_sock(sk);
1099
43815482 1100 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
1da177e4 1101
43815482 1102 if (!sock->wq->fasync_list)
989a2979
ED
1103 sock_reset_flag(sk, SOCK_FASYNC);
1104 else
bcdce719 1105 sock_set_flag(sk, SOCK_FASYNC);
1da177e4 1106
989a2979 1107 release_sock(sk);
1da177e4
LT
1108 return 0;
1109}
1110
43815482 1111/* This function may be called only under socket lock or callback_lock or rcu_lock */
1da177e4
LT
1112
1113int sock_wake_async(struct socket *sock, int how, int band)
1114{
43815482
ED
1115 struct socket_wq *wq;
1116
1117 if (!sock)
1118 return -1;
1119 rcu_read_lock();
1120 wq = rcu_dereference(sock->wq);
1121 if (!wq || !wq->fasync_list) {
1122 rcu_read_unlock();
1da177e4 1123 return -1;
43815482 1124 }
89bddce5 1125 switch (how) {
8d8ad9d7 1126 case SOCK_WAKE_WAITD:
1da177e4
LT
1127 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1128 break;
1129 goto call_kill;
8d8ad9d7 1130 case SOCK_WAKE_SPACE:
1da177e4
LT
1131 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1132 break;
1133 /* fall through */
8d8ad9d7 1134 case SOCK_WAKE_IO:
89bddce5 1135call_kill:
43815482 1136 kill_fasync(&wq->fasync_list, SIGIO, band);
1da177e4 1137 break;
8d8ad9d7 1138 case SOCK_WAKE_URG:
43815482 1139 kill_fasync(&wq->fasync_list, SIGURG, band);
1da177e4 1140 }
43815482 1141 rcu_read_unlock();
1da177e4
LT
1142 return 0;
1143}
c6d409cf 1144EXPORT_SYMBOL(sock_wake_async);
1da177e4 1145
1b8d7ae4 1146static int __sock_create(struct net *net, int family, int type, int protocol,
89bddce5 1147 struct socket **res, int kern)
1da177e4
LT
1148{
1149 int err;
1150 struct socket *sock;
55737fda 1151 const struct net_proto_family *pf;
1da177e4
LT
1152
1153 /*
89bddce5 1154 * Check protocol is in range
1da177e4
LT
1155 */
1156 if (family < 0 || family >= NPROTO)
1157 return -EAFNOSUPPORT;
1158 if (type < 0 || type >= SOCK_MAX)
1159 return -EINVAL;
1160
1161 /* Compatibility.
1162
1163 This uglymoron is moved from INET layer to here to avoid
1164 deadlock in module load.
1165 */
1166 if (family == PF_INET && type == SOCK_PACKET) {
89bddce5 1167 static int warned;
1da177e4
LT
1168 if (!warned) {
1169 warned = 1;
89bddce5
SH
1170 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1171 current->comm);
1da177e4
LT
1172 }
1173 family = PF_PACKET;
1174 }
1175
1176 err = security_socket_create(family, type, protocol, kern);
1177 if (err)
1178 return err;
89bddce5 1179
55737fda
SH
1180 /*
1181 * Allocate the socket and allow the family to set things up. if
1182 * the protocol is 0, the family is instructed to select an appropriate
1183 * default.
1184 */
1185 sock = sock_alloc();
1186 if (!sock) {
1187 if (net_ratelimit())
1188 printk(KERN_WARNING "socket: no more sockets\n");
1189 return -ENFILE; /* Not exactly a match, but its the
1190 closest posix thing */
1191 }
1192
1193 sock->type = type;
1194
95a5afca 1195#ifdef CONFIG_MODULES
89bddce5
SH
1196 /* Attempt to load a protocol module if the find failed.
1197 *
1198 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1da177e4
LT
1199 * requested real, full-featured networking support upon configuration.
1200 * Otherwise module support will break!
1201 */
55737fda 1202 if (net_families[family] == NULL)
89bddce5 1203 request_module("net-pf-%d", family);
1da177e4
LT
1204#endif
1205
55737fda
SH
1206 rcu_read_lock();
1207 pf = rcu_dereference(net_families[family]);
1208 err = -EAFNOSUPPORT;
1209 if (!pf)
1210 goto out_release;
1da177e4
LT
1211
1212 /*
1213 * We will call the ->create function, that possibly is in a loadable
1214 * module, so we have to bump that loadable module refcnt first.
1215 */
55737fda 1216 if (!try_module_get(pf->owner))
1da177e4
LT
1217 goto out_release;
1218
55737fda
SH
1219 /* Now protected by module ref count */
1220 rcu_read_unlock();
1221
3f378b68 1222 err = pf->create(net, sock, protocol, kern);
55737fda 1223 if (err < 0)
1da177e4 1224 goto out_module_put;
a79af59e 1225
1da177e4
LT
1226 /*
1227 * Now to bump the refcnt of the [loadable] module that owns this
1228 * socket at sock_release time we decrement its refcnt.
1229 */
55737fda
SH
1230 if (!try_module_get(sock->ops->owner))
1231 goto out_module_busy;
1232
1da177e4
LT
1233 /*
1234 * Now that we're done with the ->create function, the [loadable]
1235 * module can have its refcnt decremented
1236 */
55737fda 1237 module_put(pf->owner);
7420ed23
VY
1238 err = security_socket_post_create(sock, family, type, protocol, kern);
1239 if (err)
3b185525 1240 goto out_sock_release;
55737fda 1241 *res = sock;
1da177e4 1242
55737fda
SH
1243 return 0;
1244
1245out_module_busy:
1246 err = -EAFNOSUPPORT;
1da177e4 1247out_module_put:
55737fda
SH
1248 sock->ops = NULL;
1249 module_put(pf->owner);
1250out_sock_release:
1da177e4 1251 sock_release(sock);
55737fda
SH
1252 return err;
1253
1254out_release:
1255 rcu_read_unlock();
1256 goto out_sock_release;
1da177e4
LT
1257}
1258
1259int sock_create(int family, int type, int protocol, struct socket **res)
1260{
1b8d7ae4 1261 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1da177e4 1262}
c6d409cf 1263EXPORT_SYMBOL(sock_create);
1da177e4
LT
1264
1265int sock_create_kern(int family, int type, int protocol, struct socket **res)
1266{
1b8d7ae4 1267 return __sock_create(&init_net, family, type, protocol, res, 1);
1da177e4 1268}
c6d409cf 1269EXPORT_SYMBOL(sock_create_kern);
1da177e4 1270
3e0fa65f 1271SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1da177e4
LT
1272{
1273 int retval;
1274 struct socket *sock;
a677a039
UD
1275 int flags;
1276
e38b36f3
UD
1277 /* Check the SOCK_* constants for consistency. */
1278 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1279 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1280 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1281 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1282
a677a039 1283 flags = type & ~SOCK_TYPE_MASK;
77d27200 1284 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
a677a039
UD
1285 return -EINVAL;
1286 type &= SOCK_TYPE_MASK;
1da177e4 1287
aaca0bdc
UD
1288 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1289 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1290
1da177e4
LT
1291 retval = sock_create(family, type, protocol, &sock);
1292 if (retval < 0)
1293 goto out;
1294
77d27200 1295 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1da177e4
LT
1296 if (retval < 0)
1297 goto out_release;
1298
1299out:
1300 /* It may be already another descriptor 8) Not kernel problem. */
1301 return retval;
1302
1303out_release:
1304 sock_release(sock);
1305 return retval;
1306}
1307
1308/*
1309 * Create a pair of connected sockets.
1310 */
1311
3e0fa65f
HC
1312SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1313 int __user *, usockvec)
1da177e4
LT
1314{
1315 struct socket *sock1, *sock2;
1316 int fd1, fd2, err;
db349509 1317 struct file *newfile1, *newfile2;
a677a039
UD
1318 int flags;
1319
1320 flags = type & ~SOCK_TYPE_MASK;
77d27200 1321 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
a677a039
UD
1322 return -EINVAL;
1323 type &= SOCK_TYPE_MASK;
1da177e4 1324
aaca0bdc
UD
1325 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1326 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1327
1da177e4
LT
1328 /*
1329 * Obtain the first socket and check if the underlying protocol
1330 * supports the socketpair call.
1331 */
1332
1333 err = sock_create(family, type, protocol, &sock1);
1334 if (err < 0)
1335 goto out;
1336
1337 err = sock_create(family, type, protocol, &sock2);
1338 if (err < 0)
1339 goto out_release_1;
1340
1341 err = sock1->ops->socketpair(sock1, sock2);
89bddce5 1342 if (err < 0)
1da177e4
LT
1343 goto out_release_both;
1344
7cbe66b6 1345 fd1 = sock_alloc_file(sock1, &newfile1, flags);
bf3c23d1
DM
1346 if (unlikely(fd1 < 0)) {
1347 err = fd1;
db349509 1348 goto out_release_both;
bf3c23d1 1349 }
1da177e4 1350
7cbe66b6 1351 fd2 = sock_alloc_file(sock2, &newfile2, flags);
198de4d7
AV
1352 if (unlikely(fd2 < 0)) {
1353 err = fd2;
1354 fput(newfile1);
1355 put_unused_fd(fd1);
1356 sock_release(sock2);
1357 goto out;
db349509
AV
1358 }
1359
157cf649 1360 audit_fd_pair(fd1, fd2);
db349509
AV
1361 fd_install(fd1, newfile1);
1362 fd_install(fd2, newfile2);
1da177e4
LT
1363 /* fd1 and fd2 may be already another descriptors.
1364 * Not kernel problem.
1365 */
1366
89bddce5 1367 err = put_user(fd1, &usockvec[0]);
1da177e4
LT
1368 if (!err)
1369 err = put_user(fd2, &usockvec[1]);
1370 if (!err)
1371 return 0;
1372
1373 sys_close(fd2);
1374 sys_close(fd1);
1375 return err;
1376
1da177e4 1377out_release_both:
89bddce5 1378 sock_release(sock2);
1da177e4 1379out_release_1:
89bddce5 1380 sock_release(sock1);
1da177e4
LT
1381out:
1382 return err;
1383}
1384
1da177e4
LT
1385/*
1386 * Bind a name to a socket. Nothing much to do here since it's
1387 * the protocol's responsibility to handle the local address.
1388 *
1389 * We move the socket address to kernel space before we call
1390 * the protocol layer (having also checked the address is ok).
1391 */
1392
20f37034 1393SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1da177e4
LT
1394{
1395 struct socket *sock;
230b1839 1396 struct sockaddr_storage address;
6cb153ca 1397 int err, fput_needed;
1da177e4 1398
89bddce5 1399 sock = sockfd_lookup_light(fd, &err, &fput_needed);
e71a4783 1400 if (sock) {
230b1839 1401 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
89bddce5
SH
1402 if (err >= 0) {
1403 err = security_socket_bind(sock,
230b1839 1404 (struct sockaddr *)&address,
89bddce5 1405 addrlen);
6cb153ca
BL
1406 if (!err)
1407 err = sock->ops->bind(sock,
89bddce5 1408 (struct sockaddr *)
230b1839 1409 &address, addrlen);
1da177e4 1410 }
6cb153ca 1411 fput_light(sock->file, fput_needed);
89bddce5 1412 }
1da177e4
LT
1413 return err;
1414}
1415
1da177e4
LT
1416/*
1417 * Perform a listen. Basically, we allow the protocol to do anything
1418 * necessary for a listen, and if that works, we mark the socket as
1419 * ready for listening.
1420 */
1421
3e0fa65f 1422SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1da177e4
LT
1423{
1424 struct socket *sock;
6cb153ca 1425 int err, fput_needed;
b8e1f9b5 1426 int somaxconn;
89bddce5
SH
1427
1428 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1429 if (sock) {
8efa6e93 1430 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
b8e1f9b5
PE
1431 if ((unsigned)backlog > somaxconn)
1432 backlog = somaxconn;
1da177e4
LT
1433
1434 err = security_socket_listen(sock, backlog);
6cb153ca
BL
1435 if (!err)
1436 err = sock->ops->listen(sock, backlog);
1da177e4 1437
6cb153ca 1438 fput_light(sock->file, fput_needed);
1da177e4
LT
1439 }
1440 return err;
1441}
1442
1da177e4
LT
1443/*
1444 * For accept, we attempt to create a new socket, set up the link
1445 * with the client, wake up the client, then return the new
1446 * connected fd. We collect the address of the connector in kernel
1447 * space and move it to user at the very end. This is unclean because
1448 * we open the socket then return an error.
1449 *
1450 * 1003.1g adds the ability to recvmsg() to query connection pending
1451 * status to recvmsg. We need to add that support in a way thats
1452 * clean when we restucture accept also.
1453 */
1454
20f37034
HC
1455SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1456 int __user *, upeer_addrlen, int, flags)
1da177e4
LT
1457{
1458 struct socket *sock, *newsock;
39d8c1b6 1459 struct file *newfile;
6cb153ca 1460 int err, len, newfd, fput_needed;
230b1839 1461 struct sockaddr_storage address;
1da177e4 1462
77d27200 1463 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
aaca0bdc
UD
1464 return -EINVAL;
1465
1466 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1467 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1468
6cb153ca 1469 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1470 if (!sock)
1471 goto out;
1472
1473 err = -ENFILE;
c6d409cf
ED
1474 newsock = sock_alloc();
1475 if (!newsock)
1da177e4
LT
1476 goto out_put;
1477
1478 newsock->type = sock->type;
1479 newsock->ops = sock->ops;
1480
1da177e4
LT
1481 /*
1482 * We don't need try_module_get here, as the listening socket (sock)
1483 * has the protocol module (sock->ops->owner) held.
1484 */
1485 __module_get(newsock->ops->owner);
1486
7cbe66b6 1487 newfd = sock_alloc_file(newsock, &newfile, flags);
39d8c1b6
DM
1488 if (unlikely(newfd < 0)) {
1489 err = newfd;
9a1875e6
DM
1490 sock_release(newsock);
1491 goto out_put;
39d8c1b6
DM
1492 }
1493
a79af59e
FF
1494 err = security_socket_accept(sock, newsock);
1495 if (err)
39d8c1b6 1496 goto out_fd;
a79af59e 1497
1da177e4
LT
1498 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1499 if (err < 0)
39d8c1b6 1500 goto out_fd;
1da177e4
LT
1501
1502 if (upeer_sockaddr) {
230b1839 1503 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
89bddce5 1504 &len, 2) < 0) {
1da177e4 1505 err = -ECONNABORTED;
39d8c1b6 1506 goto out_fd;
1da177e4 1507 }
230b1839
YH
1508 err = move_addr_to_user((struct sockaddr *)&address,
1509 len, upeer_sockaddr, upeer_addrlen);
1da177e4 1510 if (err < 0)
39d8c1b6 1511 goto out_fd;
1da177e4
LT
1512 }
1513
1514 /* File flags are not inherited via accept() unlike another OSes. */
1515
39d8c1b6
DM
1516 fd_install(newfd, newfile);
1517 err = newfd;
1da177e4 1518
1da177e4 1519out_put:
6cb153ca 1520 fput_light(sock->file, fput_needed);
1da177e4
LT
1521out:
1522 return err;
39d8c1b6 1523out_fd:
9606a216 1524 fput(newfile);
39d8c1b6 1525 put_unused_fd(newfd);
1da177e4
LT
1526 goto out_put;
1527}
1528
20f37034
HC
1529SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1530 int __user *, upeer_addrlen)
aaca0bdc 1531{
de11defe 1532 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
aaca0bdc
UD
1533}
1534
1da177e4
LT
1535/*
1536 * Attempt to connect to a socket with the server address. The address
1537 * is in user space so we verify it is OK and move it to kernel space.
1538 *
1539 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1540 * break bindings
1541 *
1542 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1543 * other SEQPACKET protocols that take time to connect() as it doesn't
1544 * include the -EINPROGRESS status for such sockets.
1545 */
1546
20f37034
HC
1547SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1548 int, addrlen)
1da177e4
LT
1549{
1550 struct socket *sock;
230b1839 1551 struct sockaddr_storage address;
6cb153ca 1552 int err, fput_needed;
1da177e4 1553
6cb153ca 1554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1555 if (!sock)
1556 goto out;
230b1839 1557 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1da177e4
LT
1558 if (err < 0)
1559 goto out_put;
1560
89bddce5 1561 err =
230b1839 1562 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1da177e4
LT
1563 if (err)
1564 goto out_put;
1565
230b1839 1566 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1da177e4
LT
1567 sock->file->f_flags);
1568out_put:
6cb153ca 1569 fput_light(sock->file, fput_needed);
1da177e4
LT
1570out:
1571 return err;
1572}
1573
1574/*
1575 * Get the local address ('name') of a socket object. Move the obtained
1576 * name to user space.
1577 */
1578
20f37034
HC
1579SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1580 int __user *, usockaddr_len)
1da177e4
LT
1581{
1582 struct socket *sock;
230b1839 1583 struct sockaddr_storage address;
6cb153ca 1584 int len, err, fput_needed;
89bddce5 1585
6cb153ca 1586 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1587 if (!sock)
1588 goto out;
1589
1590 err = security_socket_getsockname(sock);
1591 if (err)
1592 goto out_put;
1593
230b1839 1594 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1da177e4
LT
1595 if (err)
1596 goto out_put;
230b1839 1597 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1da177e4
LT
1598
1599out_put:
6cb153ca 1600 fput_light(sock->file, fput_needed);
1da177e4
LT
1601out:
1602 return err;
1603}
1604
1605/*
1606 * Get the remote address ('name') of a socket object. Move the obtained
1607 * name to user space.
1608 */
1609
20f37034
HC
1610SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1611 int __user *, usockaddr_len)
1da177e4
LT
1612{
1613 struct socket *sock;
230b1839 1614 struct sockaddr_storage address;
6cb153ca 1615 int len, err, fput_needed;
1da177e4 1616
89bddce5
SH
1617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1618 if (sock != NULL) {
1da177e4
LT
1619 err = security_socket_getpeername(sock);
1620 if (err) {
6cb153ca 1621 fput_light(sock->file, fput_needed);
1da177e4
LT
1622 return err;
1623 }
1624
89bddce5 1625 err =
230b1839 1626 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
89bddce5 1627 1);
1da177e4 1628 if (!err)
230b1839 1629 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
89bddce5 1630 usockaddr_len);
6cb153ca 1631 fput_light(sock->file, fput_needed);
1da177e4
LT
1632 }
1633 return err;
1634}
1635
1636/*
1637 * Send a datagram to a given address. We move the address into kernel
1638 * space and check the user space data area is readable before invoking
1639 * the protocol.
1640 */
1641
3e0fa65f
HC
1642SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1643 unsigned, flags, struct sockaddr __user *, addr,
1644 int, addr_len)
1da177e4
LT
1645{
1646 struct socket *sock;
230b1839 1647 struct sockaddr_storage address;
1da177e4
LT
1648 int err;
1649 struct msghdr msg;
1650 struct iovec iov;
6cb153ca 1651 int fput_needed;
6cb153ca 1652
de0fa95c
PE
1653 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1654 if (!sock)
4387ff75 1655 goto out;
6cb153ca 1656
89bddce5
SH
1657 iov.iov_base = buff;
1658 iov.iov_len = len;
1659 msg.msg_name = NULL;
1660 msg.msg_iov = &iov;
1661 msg.msg_iovlen = 1;
1662 msg.msg_control = NULL;
1663 msg.msg_controllen = 0;
1664 msg.msg_namelen = 0;
6cb153ca 1665 if (addr) {
230b1839 1666 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1da177e4
LT
1667 if (err < 0)
1668 goto out_put;
230b1839 1669 msg.msg_name = (struct sockaddr *)&address;
89bddce5 1670 msg.msg_namelen = addr_len;
1da177e4
LT
1671 }
1672 if (sock->file->f_flags & O_NONBLOCK)
1673 flags |= MSG_DONTWAIT;
1674 msg.msg_flags = flags;
1675 err = sock_sendmsg(sock, &msg, len);
1676
89bddce5 1677out_put:
de0fa95c 1678 fput_light(sock->file, fput_needed);
4387ff75 1679out:
1da177e4
LT
1680 return err;
1681}
1682
1683/*
89bddce5 1684 * Send a datagram down a socket.
1da177e4
LT
1685 */
1686
3e0fa65f
HC
1687SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1688 unsigned, flags)
1da177e4
LT
1689{
1690 return sys_sendto(fd, buff, len, flags, NULL, 0);
1691}
1692
1693/*
89bddce5 1694 * Receive a frame from the socket and optionally record the address of the
1da177e4
LT
1695 * sender. We verify the buffers are writable and if needed move the
1696 * sender address from kernel to user space.
1697 */
1698
3e0fa65f
HC
1699SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1700 unsigned, flags, struct sockaddr __user *, addr,
1701 int __user *, addr_len)
1da177e4
LT
1702{
1703 struct socket *sock;
1704 struct iovec iov;
1705 struct msghdr msg;
230b1839 1706 struct sockaddr_storage address;
89bddce5 1707 int err, err2;
6cb153ca
BL
1708 int fput_needed;
1709
de0fa95c 1710 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4 1711 if (!sock)
de0fa95c 1712 goto out;
1da177e4 1713
89bddce5
SH
1714 msg.msg_control = NULL;
1715 msg.msg_controllen = 0;
1716 msg.msg_iovlen = 1;
1717 msg.msg_iov = &iov;
1718 iov.iov_len = size;
1719 iov.iov_base = ubuf;
230b1839
YH
1720 msg.msg_name = (struct sockaddr *)&address;
1721 msg.msg_namelen = sizeof(address);
1da177e4
LT
1722 if (sock->file->f_flags & O_NONBLOCK)
1723 flags |= MSG_DONTWAIT;
89bddce5 1724 err = sock_recvmsg(sock, &msg, size, flags);
1da177e4 1725
89bddce5 1726 if (err >= 0 && addr != NULL) {
230b1839
YH
1727 err2 = move_addr_to_user((struct sockaddr *)&address,
1728 msg.msg_namelen, addr, addr_len);
89bddce5
SH
1729 if (err2 < 0)
1730 err = err2;
1da177e4 1731 }
de0fa95c
PE
1732
1733 fput_light(sock->file, fput_needed);
4387ff75 1734out:
1da177e4
LT
1735 return err;
1736}
1737
1738/*
89bddce5 1739 * Receive a datagram from a socket.
1da177e4
LT
1740 */
1741
89bddce5
SH
1742asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1743 unsigned flags)
1da177e4
LT
1744{
1745 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1746}
1747
1748/*
1749 * Set a socket option. Because we don't know the option lengths we have
1750 * to pass the user mode parameter for the protocols to sort out.
1751 */
1752
20f37034
HC
1753SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1754 char __user *, optval, int, optlen)
1da177e4 1755{
6cb153ca 1756 int err, fput_needed;
1da177e4
LT
1757 struct socket *sock;
1758
1759 if (optlen < 0)
1760 return -EINVAL;
89bddce5
SH
1761
1762 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1763 if (sock != NULL) {
1764 err = security_socket_setsockopt(sock, level, optname);
6cb153ca
BL
1765 if (err)
1766 goto out_put;
1da177e4
LT
1767
1768 if (level == SOL_SOCKET)
89bddce5
SH
1769 err =
1770 sock_setsockopt(sock, level, optname, optval,
1771 optlen);
1da177e4 1772 else
89bddce5
SH
1773 err =
1774 sock->ops->setsockopt(sock, level, optname, optval,
1775 optlen);
6cb153ca
BL
1776out_put:
1777 fput_light(sock->file, fput_needed);
1da177e4
LT
1778 }
1779 return err;
1780}
1781
1782/*
1783 * Get a socket option. Because we don't know the option lengths we have
1784 * to pass a user mode parameter for the protocols to sort out.
1785 */
1786
20f37034
HC
1787SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1788 char __user *, optval, int __user *, optlen)
1da177e4 1789{
6cb153ca 1790 int err, fput_needed;
1da177e4
LT
1791 struct socket *sock;
1792
89bddce5
SH
1793 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1794 if (sock != NULL) {
6cb153ca
BL
1795 err = security_socket_getsockopt(sock, level, optname);
1796 if (err)
1797 goto out_put;
1da177e4
LT
1798
1799 if (level == SOL_SOCKET)
89bddce5
SH
1800 err =
1801 sock_getsockopt(sock, level, optname, optval,
1802 optlen);
1da177e4 1803 else
89bddce5
SH
1804 err =
1805 sock->ops->getsockopt(sock, level, optname, optval,
1806 optlen);
6cb153ca
BL
1807out_put:
1808 fput_light(sock->file, fput_needed);
1da177e4
LT
1809 }
1810 return err;
1811}
1812
1da177e4
LT
1813/*
1814 * Shutdown a socket.
1815 */
1816
754fe8d2 1817SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1da177e4 1818{
6cb153ca 1819 int err, fput_needed;
1da177e4
LT
1820 struct socket *sock;
1821
89bddce5
SH
1822 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1823 if (sock != NULL) {
1da177e4 1824 err = security_socket_shutdown(sock, how);
6cb153ca
BL
1825 if (!err)
1826 err = sock->ops->shutdown(sock, how);
1827 fput_light(sock->file, fput_needed);
1da177e4
LT
1828 }
1829 return err;
1830}
1831
89bddce5 1832/* A couple of helpful macros for getting the address of the 32/64 bit
1da177e4
LT
1833 * fields which are the same type (int / unsigned) on our platforms.
1834 */
1835#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1836#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1837#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1838
1da177e4
LT
1839/*
1840 * BSD sendmsg interface
1841 */
1842
3e0fa65f 1843SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1da177e4 1844{
89bddce5
SH
1845 struct compat_msghdr __user *msg_compat =
1846 (struct compat_msghdr __user *)msg;
1da177e4 1847 struct socket *sock;
230b1839 1848 struct sockaddr_storage address;
1da177e4 1849 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
b9d717a7 1850 unsigned char ctl[sizeof(struct cmsghdr) + 20]
89bddce5
SH
1851 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1852 /* 20 is size of ipv6_pktinfo */
1da177e4
LT
1853 unsigned char *ctl_buf = ctl;
1854 struct msghdr msg_sys;
1855 int err, ctl_len, iov_size, total_len;
6cb153ca 1856 int fput_needed;
89bddce5 1857
1da177e4
LT
1858 err = -EFAULT;
1859 if (MSG_CMSG_COMPAT & flags) {
1860 if (get_compat_msghdr(&msg_sys, msg_compat))
1861 return -EFAULT;
c6d409cf 1862 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1da177e4
LT
1863 return -EFAULT;
1864
6cb153ca 1865 sock = sockfd_lookup_light(fd, &err, &fput_needed);
89bddce5 1866 if (!sock)
1da177e4
LT
1867 goto out;
1868
1869 /* do not move before msg_sys is valid */
1870 err = -EMSGSIZE;
1871 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1872 goto out_put;
1873
89bddce5 1874 /* Check whether to allocate the iovec area */
1da177e4
LT
1875 err = -ENOMEM;
1876 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1877 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1878 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1879 if (!iov)
1880 goto out_put;
1881 }
1882
1883 /* This will also move the address data into kernel space */
1884 if (MSG_CMSG_COMPAT & flags) {
230b1839
YH
1885 err = verify_compat_iovec(&msg_sys, iov,
1886 (struct sockaddr *)&address,
1887 VERIFY_READ);
1da177e4 1888 } else
230b1839
YH
1889 err = verify_iovec(&msg_sys, iov,
1890 (struct sockaddr *)&address,
1891 VERIFY_READ);
89bddce5 1892 if (err < 0)
1da177e4
LT
1893 goto out_freeiov;
1894 total_len = err;
1895
1896 err = -ENOBUFS;
1897
1898 if (msg_sys.msg_controllen > INT_MAX)
1899 goto out_freeiov;
89bddce5 1900 ctl_len = msg_sys.msg_controllen;
1da177e4 1901 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
89bddce5
SH
1902 err =
1903 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1904 sizeof(ctl));
1da177e4
LT
1905 if (err)
1906 goto out_freeiov;
1907 ctl_buf = msg_sys.msg_control;
8920e8f9 1908 ctl_len = msg_sys.msg_controllen;
1da177e4 1909 } else if (ctl_len) {
89bddce5 1910 if (ctl_len > sizeof(ctl)) {
1da177e4 1911 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
89bddce5 1912 if (ctl_buf == NULL)
1da177e4
LT
1913 goto out_freeiov;
1914 }
1915 err = -EFAULT;
1916 /*
1917 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1918 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1919 * checking falls down on this.
1920 */
89bddce5
SH
1921 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1922 ctl_len))
1da177e4
LT
1923 goto out_freectl;
1924 msg_sys.msg_control = ctl_buf;
1925 }
1926 msg_sys.msg_flags = flags;
1927
1928 if (sock->file->f_flags & O_NONBLOCK)
1929 msg_sys.msg_flags |= MSG_DONTWAIT;
1930 err = sock_sendmsg(sock, &msg_sys, total_len);
1931
1932out_freectl:
89bddce5 1933 if (ctl_buf != ctl)
1da177e4
LT
1934 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1935out_freeiov:
1936 if (iov != iovstack)
1937 sock_kfree_s(sock->sk, iov, iov_size);
1938out_put:
6cb153ca 1939 fput_light(sock->file, fput_needed);
89bddce5 1940out:
1da177e4
LT
1941 return err;
1942}
1943
a2e27255
ACM
1944static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1945 struct msghdr *msg_sys, unsigned flags, int nosec)
1da177e4 1946{
89bddce5
SH
1947 struct compat_msghdr __user *msg_compat =
1948 (struct compat_msghdr __user *)msg;
1da177e4 1949 struct iovec iovstack[UIO_FASTIOV];
89bddce5 1950 struct iovec *iov = iovstack;
1da177e4
LT
1951 unsigned long cmsg_ptr;
1952 int err, iov_size, total_len, len;
1953
1954 /* kernel mode address */
230b1839 1955 struct sockaddr_storage addr;
1da177e4
LT
1956
1957 /* user mode address pointers */
1958 struct sockaddr __user *uaddr;
1959 int __user *uaddr_len;
89bddce5 1960
1da177e4 1961 if (MSG_CMSG_COMPAT & flags) {
a2e27255 1962 if (get_compat_msghdr(msg_sys, msg_compat))
1da177e4 1963 return -EFAULT;
c6d409cf 1964 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
89bddce5 1965 return -EFAULT;
1da177e4 1966
1da177e4 1967 err = -EMSGSIZE;
a2e27255
ACM
1968 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1969 goto out;
89bddce5
SH
1970
1971 /* Check whether to allocate the iovec area */
1da177e4 1972 err = -ENOMEM;
a2e27255
ACM
1973 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1974 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1da177e4
LT
1975 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1976 if (!iov)
a2e27255 1977 goto out;
1da177e4
LT
1978 }
1979
1980 /*
89bddce5
SH
1981 * Save the user-mode address (verify_iovec will change the
1982 * kernel msghdr to use the kernel address space)
1da177e4 1983 */
89bddce5 1984
a2e27255 1985 uaddr = (__force void __user *)msg_sys->msg_name;
1da177e4
LT
1986 uaddr_len = COMPAT_NAMELEN(msg);
1987 if (MSG_CMSG_COMPAT & flags) {
a2e27255 1988 err = verify_compat_iovec(msg_sys, iov,
230b1839
YH
1989 (struct sockaddr *)&addr,
1990 VERIFY_WRITE);
1da177e4 1991 } else
a2e27255 1992 err = verify_iovec(msg_sys, iov,
230b1839
YH
1993 (struct sockaddr *)&addr,
1994 VERIFY_WRITE);
1da177e4
LT
1995 if (err < 0)
1996 goto out_freeiov;
89bddce5 1997 total_len = err;
1da177e4 1998
a2e27255
ACM
1999 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2000 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
89bddce5 2001
1da177e4
LT
2002 if (sock->file->f_flags & O_NONBLOCK)
2003 flags |= MSG_DONTWAIT;
a2e27255
ACM
2004 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2005 total_len, flags);
1da177e4
LT
2006 if (err < 0)
2007 goto out_freeiov;
2008 len = err;
2009
2010 if (uaddr != NULL) {
230b1839 2011 err = move_addr_to_user((struct sockaddr *)&addr,
a2e27255 2012 msg_sys->msg_namelen, uaddr,
89bddce5 2013 uaddr_len);
1da177e4
LT
2014 if (err < 0)
2015 goto out_freeiov;
2016 }
a2e27255 2017 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
37f7f421 2018 COMPAT_FLAGS(msg));
1da177e4
LT
2019 if (err)
2020 goto out_freeiov;
2021 if (MSG_CMSG_COMPAT & flags)
a2e27255 2022 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
1da177e4
LT
2023 &msg_compat->msg_controllen);
2024 else
a2e27255 2025 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
1da177e4
LT
2026 &msg->msg_controllen);
2027 if (err)
2028 goto out_freeiov;
2029 err = len;
2030
2031out_freeiov:
2032 if (iov != iovstack)
2033 sock_kfree_s(sock->sk, iov, iov_size);
a2e27255
ACM
2034out:
2035 return err;
2036}
2037
2038/*
2039 * BSD recvmsg interface
2040 */
2041
2042SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2043 unsigned int, flags)
2044{
2045 int fput_needed, err;
2046 struct msghdr msg_sys;
2047 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2048
2049 if (!sock)
2050 goto out;
2051
2052 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2053
6cb153ca 2054 fput_light(sock->file, fput_needed);
1da177e4
LT
2055out:
2056 return err;
2057}
2058
a2e27255
ACM
2059/*
2060 * Linux recvmmsg interface
2061 */
2062
2063int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2064 unsigned int flags, struct timespec *timeout)
2065{
2066 int fput_needed, err, datagrams;
2067 struct socket *sock;
2068 struct mmsghdr __user *entry;
d7256d0e 2069 struct compat_mmsghdr __user *compat_entry;
a2e27255
ACM
2070 struct msghdr msg_sys;
2071 struct timespec end_time;
2072
2073 if (timeout &&
2074 poll_select_set_timeout(&end_time, timeout->tv_sec,
2075 timeout->tv_nsec))
2076 return -EINVAL;
2077
2078 datagrams = 0;
2079
2080 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2081 if (!sock)
2082 return err;
2083
2084 err = sock_error(sock->sk);
2085 if (err)
2086 goto out_put;
2087
2088 entry = mmsg;
d7256d0e 2089 compat_entry = (struct compat_mmsghdr __user *)mmsg;
a2e27255
ACM
2090
2091 while (datagrams < vlen) {
2092 /*
2093 * No need to ask LSM for more than the first datagram.
2094 */
d7256d0e
JMG
2095 if (MSG_CMSG_COMPAT & flags) {
2096 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2097 &msg_sys, flags, datagrams);
2098 if (err < 0)
2099 break;
2100 err = __put_user(err, &compat_entry->msg_len);
2101 ++compat_entry;
2102 } else {
2103 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2104 &msg_sys, flags, datagrams);
2105 if (err < 0)
2106 break;
2107 err = put_user(err, &entry->msg_len);
2108 ++entry;
2109 }
2110
a2e27255
ACM
2111 if (err)
2112 break;
a2e27255
ACM
2113 ++datagrams;
2114
71c5c159
BB
2115 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2116 if (flags & MSG_WAITFORONE)
2117 flags |= MSG_DONTWAIT;
2118
a2e27255
ACM
2119 if (timeout) {
2120 ktime_get_ts(timeout);
2121 *timeout = timespec_sub(end_time, *timeout);
2122 if (timeout->tv_sec < 0) {
2123 timeout->tv_sec = timeout->tv_nsec = 0;
2124 break;
2125 }
2126
2127 /* Timeout, return less than vlen datagrams */
2128 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2129 break;
2130 }
2131
2132 /* Out of band data, return right away */
2133 if (msg_sys.msg_flags & MSG_OOB)
2134 break;
2135 }
2136
2137out_put:
2138 fput_light(sock->file, fput_needed);
1da177e4 2139
a2e27255
ACM
2140 if (err == 0)
2141 return datagrams;
2142
2143 if (datagrams != 0) {
2144 /*
2145 * We may return less entries than requested (vlen) if the
2146 * sock is non block and there aren't enough datagrams...
2147 */
2148 if (err != -EAGAIN) {
2149 /*
2150 * ... or if recvmsg returns an error after we
2151 * received some datagrams, where we record the
2152 * error to return on the next call or if the
2153 * app asks about it using getsockopt(SO_ERROR).
2154 */
2155 sock->sk->sk_err = -err;
2156 }
2157
2158 return datagrams;
2159 }
2160
2161 return err;
2162}
2163
2164SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2165 unsigned int, vlen, unsigned int, flags,
2166 struct timespec __user *, timeout)
2167{
2168 int datagrams;
2169 struct timespec timeout_sys;
2170
2171 if (!timeout)
2172 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2173
2174 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2175 return -EFAULT;
2176
2177 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2178
2179 if (datagrams > 0 &&
2180 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2181 datagrams = -EFAULT;
2182
2183 return datagrams;
2184}
2185
2186#ifdef __ARCH_WANT_SYS_SOCKETCALL
1da177e4
LT
2187/* Argument list sizes for sys_socketcall */
2188#define AL(x) ((x) * sizeof(unsigned long))
a2e27255 2189static const unsigned char nargs[20] = {
c6d409cf
ED
2190 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2191 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2192 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2193 AL(4), AL(5)
89bddce5
SH
2194};
2195
1da177e4
LT
2196#undef AL
2197
2198/*
89bddce5 2199 * System call vectors.
1da177e4
LT
2200 *
2201 * Argument checking cleaned up. Saved 20% in size.
2202 * This function doesn't need to set the kernel lock because
89bddce5 2203 * it is set by the callees.
1da177e4
LT
2204 */
2205
3e0fa65f 2206SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
1da177e4
LT
2207{
2208 unsigned long a[6];
89bddce5 2209 unsigned long a0, a1;
1da177e4 2210 int err;
47379052 2211 unsigned int len;
1da177e4 2212
a2e27255 2213 if (call < 1 || call > SYS_RECVMMSG)
1da177e4
LT
2214 return -EINVAL;
2215
47379052
AV
2216 len = nargs[call];
2217 if (len > sizeof(a))
2218 return -EINVAL;
2219
1da177e4 2220 /* copy_from_user should be SMP safe. */
47379052 2221 if (copy_from_user(a, args, len))
1da177e4 2222 return -EFAULT;
3ec3b2fb 2223
f3298dc4 2224 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3ec3b2fb 2225
89bddce5
SH
2226 a0 = a[0];
2227 a1 = a[1];
2228
2229 switch (call) {
2230 case SYS_SOCKET:
2231 err = sys_socket(a0, a1, a[2]);
2232 break;
2233 case SYS_BIND:
2234 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2235 break;
2236 case SYS_CONNECT:
2237 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2238 break;
2239 case SYS_LISTEN:
2240 err = sys_listen(a0, a1);
2241 break;
2242 case SYS_ACCEPT:
de11defe
UD
2243 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2244 (int __user *)a[2], 0);
89bddce5
SH
2245 break;
2246 case SYS_GETSOCKNAME:
2247 err =
2248 sys_getsockname(a0, (struct sockaddr __user *)a1,
2249 (int __user *)a[2]);
2250 break;
2251 case SYS_GETPEERNAME:
2252 err =
2253 sys_getpeername(a0, (struct sockaddr __user *)a1,
2254 (int __user *)a[2]);
2255 break;
2256 case SYS_SOCKETPAIR:
2257 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2258 break;
2259 case SYS_SEND:
2260 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2261 break;
2262 case SYS_SENDTO:
2263 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2264 (struct sockaddr __user *)a[4], a[5]);
2265 break;
2266 case SYS_RECV:
2267 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2268 break;
2269 case SYS_RECVFROM:
2270 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2271 (struct sockaddr __user *)a[4],
2272 (int __user *)a[5]);
2273 break;
2274 case SYS_SHUTDOWN:
2275 err = sys_shutdown(a0, a1);
2276 break;
2277 case SYS_SETSOCKOPT:
2278 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2279 break;
2280 case SYS_GETSOCKOPT:
2281 err =
2282 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2283 (int __user *)a[4]);
2284 break;
2285 case SYS_SENDMSG:
2286 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2287 break;
2288 case SYS_RECVMSG:
2289 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2290 break;
a2e27255
ACM
2291 case SYS_RECVMMSG:
2292 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2293 (struct timespec __user *)a[4]);
2294 break;
de11defe
UD
2295 case SYS_ACCEPT4:
2296 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2297 (int __user *)a[2], a[3]);
aaca0bdc 2298 break;
89bddce5
SH
2299 default:
2300 err = -EINVAL;
2301 break;
1da177e4
LT
2302 }
2303 return err;
2304}
2305
89bddce5 2306#endif /* __ARCH_WANT_SYS_SOCKETCALL */
1da177e4 2307
55737fda
SH
2308/**
2309 * sock_register - add a socket protocol handler
2310 * @ops: description of protocol
2311 *
1da177e4
LT
2312 * This function is called by a protocol handler that wants to
2313 * advertise its address family, and have it linked into the
55737fda
SH
2314 * socket interface. The value ops->family coresponds to the
2315 * socket system call protocol family.
1da177e4 2316 */
f0fd27d4 2317int sock_register(const struct net_proto_family *ops)
1da177e4
LT
2318{
2319 int err;
2320
2321 if (ops->family >= NPROTO) {
89bddce5
SH
2322 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2323 NPROTO);
1da177e4
LT
2324 return -ENOBUFS;
2325 }
55737fda
SH
2326
2327 spin_lock(&net_family_lock);
2328 if (net_families[ops->family])
2329 err = -EEXIST;
2330 else {
89bddce5 2331 net_families[ops->family] = ops;
1da177e4
LT
2332 err = 0;
2333 }
55737fda
SH
2334 spin_unlock(&net_family_lock);
2335
89bddce5 2336 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
1da177e4
LT
2337 return err;
2338}
c6d409cf 2339EXPORT_SYMBOL(sock_register);
1da177e4 2340
55737fda
SH
2341/**
2342 * sock_unregister - remove a protocol handler
2343 * @family: protocol family to remove
2344 *
1da177e4
LT
2345 * This function is called by a protocol handler that wants to
2346 * remove its address family, and have it unlinked from the
55737fda
SH
2347 * new socket creation.
2348 *
2349 * If protocol handler is a module, then it can use module reference
2350 * counts to protect against new references. If protocol handler is not
2351 * a module then it needs to provide its own protection in
2352 * the ops->create routine.
1da177e4 2353 */
f0fd27d4 2354void sock_unregister(int family)
1da177e4 2355{
f0fd27d4 2356 BUG_ON(family < 0 || family >= NPROTO);
1da177e4 2357
55737fda 2358 spin_lock(&net_family_lock);
89bddce5 2359 net_families[family] = NULL;
55737fda
SH
2360 spin_unlock(&net_family_lock);
2361
2362 synchronize_rcu();
2363
89bddce5 2364 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
1da177e4 2365}
c6d409cf 2366EXPORT_SYMBOL(sock_unregister);
1da177e4 2367
77d76ea3 2368static int __init sock_init(void)
1da177e4
LT
2369{
2370 /*
89bddce5 2371 * Initialize sock SLAB cache.
1da177e4 2372 */
89bddce5 2373
1da177e4
LT
2374 sk_init();
2375
1da177e4 2376 /*
89bddce5 2377 * Initialize skbuff SLAB cache
1da177e4
LT
2378 */
2379 skb_init();
1da177e4
LT
2380
2381 /*
89bddce5 2382 * Initialize the protocols module.
1da177e4
LT
2383 */
2384
2385 init_inodecache();
2386 register_filesystem(&sock_fs_type);
2387 sock_mnt = kern_mount(&sock_fs_type);
77d76ea3
AK
2388
2389 /* The real protocol initialization is performed in later initcalls.
1da177e4
LT
2390 */
2391
2392#ifdef CONFIG_NETFILTER
2393 netfilter_init();
2394#endif
cbeb321a 2395
c1f19b51
RC
2396#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2397 skb_timestamping_init();
2398#endif
2399
cbeb321a 2400 return 0;
1da177e4
LT
2401}
2402
77d76ea3
AK
2403core_initcall(sock_init); /* early initcall */
2404
1da177e4
LT
2405#ifdef CONFIG_PROC_FS
2406void socket_seq_show(struct seq_file *seq)
2407{
2408 int cpu;
2409 int counter = 0;
2410
6f912042 2411 for_each_possible_cpu(cpu)
89bddce5 2412 counter += per_cpu(sockets_in_use, cpu);
1da177e4
LT
2413
2414 /* It can be negative, by the way. 8) */
2415 if (counter < 0)
2416 counter = 0;
2417
2418 seq_printf(seq, "sockets: used %d\n", counter);
2419}
89bddce5 2420#endif /* CONFIG_PROC_FS */
1da177e4 2421
89bbfc95 2422#ifdef CONFIG_COMPAT
6b96018b
AB
2423static int do_siocgstamp(struct net *net, struct socket *sock,
2424 unsigned int cmd, struct compat_timeval __user *up)
7a229387 2425{
7a229387
AB
2426 mm_segment_t old_fs = get_fs();
2427 struct timeval ktv;
2428 int err;
2429
2430 set_fs(KERNEL_DS);
6b96018b 2431 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
7a229387
AB
2432 set_fs(old_fs);
2433 if (!err) {
2434 err = put_user(ktv.tv_sec, &up->tv_sec);
2435 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2436 }
2437 return err;
2438}
2439
6b96018b
AB
2440static int do_siocgstampns(struct net *net, struct socket *sock,
2441 unsigned int cmd, struct compat_timespec __user *up)
7a229387 2442{
7a229387
AB
2443 mm_segment_t old_fs = get_fs();
2444 struct timespec kts;
2445 int err;
2446
2447 set_fs(KERNEL_DS);
6b96018b 2448 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
7a229387
AB
2449 set_fs(old_fs);
2450 if (!err) {
2451 err = put_user(kts.tv_sec, &up->tv_sec);
2452 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2453 }
2454 return err;
2455}
2456
6b96018b 2457static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
7a229387
AB
2458{
2459 struct ifreq __user *uifr;
2460 int err;
2461
2462 uifr = compat_alloc_user_space(sizeof(struct ifreq));
6b96018b 2463 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2464 return -EFAULT;
2465
6b96018b 2466 err = dev_ioctl(net, SIOCGIFNAME, uifr);
7a229387
AB
2467 if (err)
2468 return err;
2469
6b96018b 2470 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
7a229387
AB
2471 return -EFAULT;
2472
2473 return 0;
2474}
2475
6b96018b 2476static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
7a229387 2477{
6b96018b 2478 struct compat_ifconf ifc32;
7a229387
AB
2479 struct ifconf ifc;
2480 struct ifconf __user *uifc;
6b96018b 2481 struct compat_ifreq __user *ifr32;
7a229387
AB
2482 struct ifreq __user *ifr;
2483 unsigned int i, j;
2484 int err;
2485
6b96018b 2486 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
7a229387
AB
2487 return -EFAULT;
2488
2489 if (ifc32.ifcbuf == 0) {
2490 ifc32.ifc_len = 0;
2491 ifc.ifc_len = 0;
2492 ifc.ifc_req = NULL;
2493 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2494 } else {
c6d409cf
ED
2495 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2496 sizeof(struct ifreq);
7a229387
AB
2497 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2498 ifc.ifc_len = len;
2499 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2500 ifr32 = compat_ptr(ifc32.ifcbuf);
c6d409cf 2501 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
6b96018b 2502 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2503 return -EFAULT;
2504 ifr++;
2505 ifr32++;
2506 }
2507 }
2508 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2509 return -EFAULT;
2510
6b96018b 2511 err = dev_ioctl(net, SIOCGIFCONF, uifc);
7a229387
AB
2512 if (err)
2513 return err;
2514
2515 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2516 return -EFAULT;
2517
2518 ifr = ifc.ifc_req;
2519 ifr32 = compat_ptr(ifc32.ifcbuf);
2520 for (i = 0, j = 0;
c6d409cf
ED
2521 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2522 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2523 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
7a229387
AB
2524 return -EFAULT;
2525 ifr32++;
2526 ifr++;
2527 }
2528
2529 if (ifc32.ifcbuf == 0) {
2530 /* Translate from 64-bit structure multiple to
2531 * a 32-bit one.
2532 */
2533 i = ifc.ifc_len;
6b96018b 2534 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
7a229387
AB
2535 ifc32.ifc_len = i;
2536 } else {
2537 ifc32.ifc_len = i;
2538 }
6b96018b 2539 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
7a229387
AB
2540 return -EFAULT;
2541
2542 return 0;
2543}
2544
6b96018b 2545static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
7a229387
AB
2546{
2547 struct ifreq __user *ifr;
7a229387
AB
2548 u32 data;
2549 void __user *datap;
2550
2551 ifr = compat_alloc_user_space(sizeof(*ifr));
7a229387
AB
2552
2553 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2554 return -EFAULT;
2555
2556 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2557 return -EFAULT;
2558
2559 datap = compat_ptr(data);
2560 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2561 return -EFAULT;
2562
6b96018b 2563 return dev_ioctl(net, SIOCETHTOOL, ifr);
7a229387
AB
2564}
2565
7a50a240
AB
2566static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2567{
2568 void __user *uptr;
2569 compat_uptr_t uptr32;
2570 struct ifreq __user *uifr;
2571
c6d409cf 2572 uifr = compat_alloc_user_space(sizeof(*uifr));
7a50a240
AB
2573 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2574 return -EFAULT;
2575
2576 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2577 return -EFAULT;
2578
2579 uptr = compat_ptr(uptr32);
2580
2581 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2582 return -EFAULT;
2583
2584 return dev_ioctl(net, SIOCWANDEV, uifr);
2585}
2586
6b96018b
AB
2587static int bond_ioctl(struct net *net, unsigned int cmd,
2588 struct compat_ifreq __user *ifr32)
7a229387
AB
2589{
2590 struct ifreq kifr;
2591 struct ifreq __user *uifr;
7a229387
AB
2592 mm_segment_t old_fs;
2593 int err;
2594 u32 data;
2595 void __user *datap;
2596
2597 switch (cmd) {
2598 case SIOCBONDENSLAVE:
2599 case SIOCBONDRELEASE:
2600 case SIOCBONDSETHWADDR:
2601 case SIOCBONDCHANGEACTIVE:
6b96018b 2602 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2603 return -EFAULT;
2604
2605 old_fs = get_fs();
c6d409cf 2606 set_fs(KERNEL_DS);
6b96018b 2607 err = dev_ioctl(net, cmd, &kifr);
c6d409cf 2608 set_fs(old_fs);
7a229387
AB
2609
2610 return err;
2611 case SIOCBONDSLAVEINFOQUERY:
2612 case SIOCBONDINFOQUERY:
2613 uifr = compat_alloc_user_space(sizeof(*uifr));
2614 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2615 return -EFAULT;
2616
2617 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2618 return -EFAULT;
2619
2620 datap = compat_ptr(data);
2621 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2622 return -EFAULT;
2623
6b96018b 2624 return dev_ioctl(net, cmd, uifr);
7a229387
AB
2625 default:
2626 return -EINVAL;
ccbd6a5a 2627 }
7a229387
AB
2628}
2629
6b96018b
AB
2630static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2631 struct compat_ifreq __user *u_ifreq32)
7a229387
AB
2632{
2633 struct ifreq __user *u_ifreq64;
7a229387
AB
2634 char tmp_buf[IFNAMSIZ];
2635 void __user *data64;
2636 u32 data32;
2637
2638 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2639 IFNAMSIZ))
2640 return -EFAULT;
2641 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2642 return -EFAULT;
2643 data64 = compat_ptr(data32);
2644
2645 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2646
2647 /* Don't check these user accesses, just let that get trapped
2648 * in the ioctl handler instead.
2649 */
2650 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2651 IFNAMSIZ))
2652 return -EFAULT;
2653 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2654 return -EFAULT;
2655
6b96018b 2656 return dev_ioctl(net, cmd, u_ifreq64);
7a229387
AB
2657}
2658
6b96018b
AB
2659static int dev_ifsioc(struct net *net, struct socket *sock,
2660 unsigned int cmd, struct compat_ifreq __user *uifr32)
7a229387 2661{
a2116ed2 2662 struct ifreq __user *uifr;
7a229387
AB
2663 int err;
2664
a2116ed2
AB
2665 uifr = compat_alloc_user_space(sizeof(*uifr));
2666 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2667 return -EFAULT;
2668
2669 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2670
7a229387
AB
2671 if (!err) {
2672 switch (cmd) {
2673 case SIOCGIFFLAGS:
2674 case SIOCGIFMETRIC:
2675 case SIOCGIFMTU:
2676 case SIOCGIFMEM:
2677 case SIOCGIFHWADDR:
2678 case SIOCGIFINDEX:
2679 case SIOCGIFADDR:
2680 case SIOCGIFBRDADDR:
2681 case SIOCGIFDSTADDR:
2682 case SIOCGIFNETMASK:
fab2532b 2683 case SIOCGIFPFLAGS:
7a229387 2684 case SIOCGIFTXQLEN:
fab2532b
AB
2685 case SIOCGMIIPHY:
2686 case SIOCGMIIREG:
a2116ed2 2687 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
7a229387
AB
2688 err = -EFAULT;
2689 break;
2690 }
2691 }
2692 return err;
2693}
2694
a2116ed2
AB
2695static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2696 struct compat_ifreq __user *uifr32)
2697{
2698 struct ifreq ifr;
2699 struct compat_ifmap __user *uifmap32;
2700 mm_segment_t old_fs;
2701 int err;
2702
2703 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2704 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2705 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2706 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2707 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2708 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2709 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2710 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2711 if (err)
2712 return -EFAULT;
2713
2714 old_fs = get_fs();
c6d409cf 2715 set_fs(KERNEL_DS);
a2116ed2 2716 err = dev_ioctl(net, cmd, (void __user *)&ifr);
c6d409cf 2717 set_fs(old_fs);
a2116ed2
AB
2718
2719 if (cmd == SIOCGIFMAP && !err) {
2720 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2721 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2722 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2723 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2724 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2725 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2726 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2727 if (err)
2728 err = -EFAULT;
2729 }
2730 return err;
2731}
2732
2733static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2734{
2735 void __user *uptr;
2736 compat_uptr_t uptr32;
2737 struct ifreq __user *uifr;
2738
c6d409cf 2739 uifr = compat_alloc_user_space(sizeof(*uifr));
a2116ed2
AB
2740 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2741 return -EFAULT;
2742
2743 if (get_user(uptr32, &uifr32->ifr_data))
2744 return -EFAULT;
2745
2746 uptr = compat_ptr(uptr32);
2747
2748 if (put_user(uptr, &uifr->ifr_data))
2749 return -EFAULT;
2750
2751 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2752}
2753
7a229387 2754struct rtentry32 {
c6d409cf 2755 u32 rt_pad1;
7a229387
AB
2756 struct sockaddr rt_dst; /* target address */
2757 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2758 struct sockaddr rt_genmask; /* target network mask (IP) */
c6d409cf
ED
2759 unsigned short rt_flags;
2760 short rt_pad2;
2761 u32 rt_pad3;
2762 unsigned char rt_tos;
2763 unsigned char rt_class;
2764 short rt_pad4;
2765 short rt_metric; /* +1 for binary compatibility! */
7a229387 2766 /* char * */ u32 rt_dev; /* forcing the device at add */
c6d409cf
ED
2767 u32 rt_mtu; /* per route MTU/Window */
2768 u32 rt_window; /* Window clamping */
7a229387
AB
2769 unsigned short rt_irtt; /* Initial RTT */
2770};
2771
2772struct in6_rtmsg32 {
2773 struct in6_addr rtmsg_dst;
2774 struct in6_addr rtmsg_src;
2775 struct in6_addr rtmsg_gateway;
2776 u32 rtmsg_type;
2777 u16 rtmsg_dst_len;
2778 u16 rtmsg_src_len;
2779 u32 rtmsg_metric;
2780 u32 rtmsg_info;
2781 u32 rtmsg_flags;
2782 s32 rtmsg_ifindex;
2783};
2784
6b96018b
AB
2785static int routing_ioctl(struct net *net, struct socket *sock,
2786 unsigned int cmd, void __user *argp)
7a229387
AB
2787{
2788 int ret;
2789 void *r = NULL;
2790 struct in6_rtmsg r6;
2791 struct rtentry r4;
2792 char devname[16];
2793 u32 rtdev;
2794 mm_segment_t old_fs = get_fs();
2795
6b96018b
AB
2796 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2797 struct in6_rtmsg32 __user *ur6 = argp;
c6d409cf 2798 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
7a229387 2799 3 * sizeof(struct in6_addr));
c6d409cf
ED
2800 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2801 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2802 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2803 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2804 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2805 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2806 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
7a229387
AB
2807
2808 r = (void *) &r6;
2809 } else { /* ipv4 */
6b96018b 2810 struct rtentry32 __user *ur4 = argp;
c6d409cf 2811 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
7a229387 2812 3 * sizeof(struct sockaddr));
c6d409cf
ED
2813 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2814 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2815 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2816 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2817 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2818 ret |= __get_user(rtdev, &(ur4->rt_dev));
7a229387 2819 if (rtdev) {
c6d409cf 2820 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
7a229387
AB
2821 r4.rt_dev = devname; devname[15] = 0;
2822 } else
2823 r4.rt_dev = NULL;
2824
2825 r = (void *) &r4;
2826 }
2827
2828 if (ret) {
2829 ret = -EFAULT;
2830 goto out;
2831 }
2832
c6d409cf 2833 set_fs(KERNEL_DS);
6b96018b 2834 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
c6d409cf 2835 set_fs(old_fs);
7a229387
AB
2836
2837out:
7a229387
AB
2838 return ret;
2839}
2840
2841/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2842 * for some operations; this forces use of the newer bridge-utils that
2843 * use compatiable ioctls
2844 */
6b96018b 2845static int old_bridge_ioctl(compat_ulong_t __user *argp)
7a229387 2846{
6b96018b 2847 compat_ulong_t tmp;
7a229387 2848
6b96018b 2849 if (get_user(tmp, argp))
7a229387
AB
2850 return -EFAULT;
2851 if (tmp == BRCTL_GET_VERSION)
2852 return BRCTL_VERSION + 1;
2853 return -EINVAL;
2854}
2855
6b96018b
AB
2856static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2857 unsigned int cmd, unsigned long arg)
2858{
2859 void __user *argp = compat_ptr(arg);
2860 struct sock *sk = sock->sk;
2861 struct net *net = sock_net(sk);
7a229387 2862
6b96018b
AB
2863 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2864 return siocdevprivate_ioctl(net, cmd, argp);
2865
2866 switch (cmd) {
2867 case SIOCSIFBR:
2868 case SIOCGIFBR:
2869 return old_bridge_ioctl(argp);
2870 case SIOCGIFNAME:
2871 return dev_ifname32(net, argp);
2872 case SIOCGIFCONF:
2873 return dev_ifconf(net, argp);
2874 case SIOCETHTOOL:
2875 return ethtool_ioctl(net, argp);
7a50a240
AB
2876 case SIOCWANDEV:
2877 return compat_siocwandev(net, argp);
a2116ed2
AB
2878 case SIOCGIFMAP:
2879 case SIOCSIFMAP:
2880 return compat_sioc_ifmap(net, cmd, argp);
6b96018b
AB
2881 case SIOCBONDENSLAVE:
2882 case SIOCBONDRELEASE:
2883 case SIOCBONDSETHWADDR:
2884 case SIOCBONDSLAVEINFOQUERY:
2885 case SIOCBONDINFOQUERY:
2886 case SIOCBONDCHANGEACTIVE:
2887 return bond_ioctl(net, cmd, argp);
2888 case SIOCADDRT:
2889 case SIOCDELRT:
2890 return routing_ioctl(net, sock, cmd, argp);
2891 case SIOCGSTAMP:
2892 return do_siocgstamp(net, sock, cmd, argp);
2893 case SIOCGSTAMPNS:
2894 return do_siocgstampns(net, sock, cmd, argp);
a2116ed2
AB
2895 case SIOCSHWTSTAMP:
2896 return compat_siocshwtstamp(net, argp);
6b96018b
AB
2897
2898 case FIOSETOWN:
2899 case SIOCSPGRP:
2900 case FIOGETOWN:
2901 case SIOCGPGRP:
2902 case SIOCBRADDBR:
2903 case SIOCBRDELBR:
2904 case SIOCGIFVLAN:
2905 case SIOCSIFVLAN:
2906 case SIOCADDDLCI:
2907 case SIOCDELDLCI:
2908 return sock_ioctl(file, cmd, arg);
2909
2910 case SIOCGIFFLAGS:
2911 case SIOCSIFFLAGS:
2912 case SIOCGIFMETRIC:
2913 case SIOCSIFMETRIC:
2914 case SIOCGIFMTU:
2915 case SIOCSIFMTU:
2916 case SIOCGIFMEM:
2917 case SIOCSIFMEM:
2918 case SIOCGIFHWADDR:
2919 case SIOCSIFHWADDR:
2920 case SIOCADDMULTI:
2921 case SIOCDELMULTI:
2922 case SIOCGIFINDEX:
6b96018b
AB
2923 case SIOCGIFADDR:
2924 case SIOCSIFADDR:
2925 case SIOCSIFHWBROADCAST:
6b96018b 2926 case SIOCDIFADDR:
6b96018b
AB
2927 case SIOCGIFBRDADDR:
2928 case SIOCSIFBRDADDR:
2929 case SIOCGIFDSTADDR:
2930 case SIOCSIFDSTADDR:
2931 case SIOCGIFNETMASK:
2932 case SIOCSIFNETMASK:
2933 case SIOCSIFPFLAGS:
2934 case SIOCGIFPFLAGS:
2935 case SIOCGIFTXQLEN:
2936 case SIOCSIFTXQLEN:
2937 case SIOCBRADDIF:
2938 case SIOCBRDELIF:
9177efd3
AB
2939 case SIOCSIFNAME:
2940 case SIOCGMIIPHY:
2941 case SIOCGMIIREG:
2942 case SIOCSMIIREG:
6b96018b 2943 return dev_ifsioc(net, sock, cmd, argp);
9177efd3 2944
6b96018b
AB
2945 case SIOCSARP:
2946 case SIOCGARP:
2947 case SIOCDARP:
6b96018b 2948 case SIOCATMARK:
9177efd3
AB
2949 return sock_do_ioctl(net, sock, cmd, arg);
2950 }
2951
2952 /* Prevent warning from compat_sys_ioctl, these always
2953 * result in -EINVAL in the native case anyway. */
2954 switch (cmd) {
2955 case SIOCRTMSG:
2956 case SIOCGIFCOUNT:
6b96018b
AB
2957 case SIOCSRARP:
2958 case SIOCGRARP:
2959 case SIOCDRARP:
9177efd3
AB
2960 case SIOCSIFLINK:
2961 case SIOCGIFSLAVE:
2962 case SIOCSIFSLAVE:
2963 return -EINVAL;
6b96018b
AB
2964 }
2965
2966 return -ENOIOCTLCMD;
2967}
7a229387 2968
89bbfc95 2969static long compat_sock_ioctl(struct file *file, unsigned cmd,
89bddce5 2970 unsigned long arg)
89bbfc95
SP
2971{
2972 struct socket *sock = file->private_data;
2973 int ret = -ENOIOCTLCMD;
87de87d5
DM
2974 struct sock *sk;
2975 struct net *net;
2976
2977 sk = sock->sk;
2978 net = sock_net(sk);
89bbfc95
SP
2979
2980 if (sock->ops->compat_ioctl)
2981 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2982
87de87d5
DM
2983 if (ret == -ENOIOCTLCMD &&
2984 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2985 ret = compat_wext_handle_ioctl(net, cmd, arg);
2986
6b96018b
AB
2987 if (ret == -ENOIOCTLCMD)
2988 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2989
89bbfc95
SP
2990 return ret;
2991}
2992#endif
2993
ac5a488e
SS
2994int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2995{
2996 return sock->ops->bind(sock, addr, addrlen);
2997}
c6d409cf 2998EXPORT_SYMBOL(kernel_bind);
ac5a488e
SS
2999
3000int kernel_listen(struct socket *sock, int backlog)
3001{
3002 return sock->ops->listen(sock, backlog);
3003}
c6d409cf 3004EXPORT_SYMBOL(kernel_listen);
ac5a488e
SS
3005
3006int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3007{
3008 struct sock *sk = sock->sk;
3009 int err;
3010
3011 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3012 newsock);
3013 if (err < 0)
3014 goto done;
3015
3016 err = sock->ops->accept(sock, *newsock, flags);
3017 if (err < 0) {
3018 sock_release(*newsock);
fa8705b0 3019 *newsock = NULL;
ac5a488e
SS
3020 goto done;
3021 }
3022
3023 (*newsock)->ops = sock->ops;
1b08534e 3024 __module_get((*newsock)->ops->owner);
ac5a488e
SS
3025
3026done:
3027 return err;
3028}
c6d409cf 3029EXPORT_SYMBOL(kernel_accept);
ac5a488e
SS
3030
3031int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
4768fbcb 3032 int flags)
ac5a488e
SS
3033{
3034 return sock->ops->connect(sock, addr, addrlen, flags);
3035}
c6d409cf 3036EXPORT_SYMBOL(kernel_connect);
ac5a488e
SS
3037
3038int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3039 int *addrlen)
3040{
3041 return sock->ops->getname(sock, addr, addrlen, 0);
3042}
c6d409cf 3043EXPORT_SYMBOL(kernel_getsockname);
ac5a488e
SS
3044
3045int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3046 int *addrlen)
3047{
3048 return sock->ops->getname(sock, addr, addrlen, 1);
3049}
c6d409cf 3050EXPORT_SYMBOL(kernel_getpeername);
ac5a488e
SS
3051
3052int kernel_getsockopt(struct socket *sock, int level, int optname,
3053 char *optval, int *optlen)
3054{
3055 mm_segment_t oldfs = get_fs();
3056 int err;
3057
3058 set_fs(KERNEL_DS);
3059 if (level == SOL_SOCKET)
3060 err = sock_getsockopt(sock, level, optname, optval, optlen);
3061 else
3062 err = sock->ops->getsockopt(sock, level, optname, optval,
3063 optlen);
3064 set_fs(oldfs);
3065 return err;
3066}
c6d409cf 3067EXPORT_SYMBOL(kernel_getsockopt);
ac5a488e
SS
3068
3069int kernel_setsockopt(struct socket *sock, int level, int optname,
b7058842 3070 char *optval, unsigned int optlen)
ac5a488e
SS
3071{
3072 mm_segment_t oldfs = get_fs();
3073 int err;
3074
3075 set_fs(KERNEL_DS);
3076 if (level == SOL_SOCKET)
3077 err = sock_setsockopt(sock, level, optname, optval, optlen);
3078 else
3079 err = sock->ops->setsockopt(sock, level, optname, optval,
3080 optlen);
3081 set_fs(oldfs);
3082 return err;
3083}
c6d409cf 3084EXPORT_SYMBOL(kernel_setsockopt);
ac5a488e
SS
3085
3086int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3087 size_t size, int flags)
3088{
f8451725
HX
3089 sock_update_classid(sock->sk);
3090
ac5a488e
SS
3091 if (sock->ops->sendpage)
3092 return sock->ops->sendpage(sock, page, offset, size, flags);
3093
3094 return sock_no_sendpage(sock, page, offset, size, flags);
3095}
c6d409cf 3096EXPORT_SYMBOL(kernel_sendpage);
ac5a488e
SS
3097
3098int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3099{
3100 mm_segment_t oldfs = get_fs();
3101 int err;
3102
3103 set_fs(KERNEL_DS);
3104 err = sock->ops->ioctl(sock, cmd, arg);
3105 set_fs(oldfs);
3106
3107 return err;
3108}
c6d409cf 3109EXPORT_SYMBOL(kernel_sock_ioctl);
ac5a488e 3110
91cf45f0
TM
3111int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3112{
3113 return sock->ops->shutdown(sock, how);
3114}
91cf45f0 3115EXPORT_SYMBOL(kernel_sock_shutdown);