sunrpc: Tag rpc_xprt with net
[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
20d49473
PO
538int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
539 union skb_shared_tx *shtx)
540{
541 shtx->flags = 0;
542 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
543 shtx->hardware = 1;
544 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
545 shtx->software = 1;
546 return 0;
547}
548EXPORT_SYMBOL(sock_tx_timestamp);
549
89bddce5 550static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT
551 struct msghdr *msg, size_t size)
552{
553 struct sock_iocb *si = kiocb_to_siocb(iocb);
554 int err;
555
f8451725
HX
556 sock_update_classid(sock->sk);
557
1da177e4
LT
558 si->sock = sock;
559 si->scm = NULL;
560 si->msg = msg;
561 si->size = size;
562
563 err = security_socket_sendmsg(sock, msg, size);
564 if (err)
565 return err;
566
567 return sock->ops->sendmsg(iocb, sock, msg, size);
568}
569
570int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
571{
572 struct kiocb iocb;
573 struct sock_iocb siocb;
574 int ret;
575
576 init_sync_kiocb(&iocb, NULL);
577 iocb.private = &siocb;
578 ret = __sock_sendmsg(&iocb, sock, msg, size);
579 if (-EIOCBQUEUED == ret)
580 ret = wait_on_sync_kiocb(&iocb);
581 return ret;
582}
c6d409cf 583EXPORT_SYMBOL(sock_sendmsg);
1da177e4
LT
584
585int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
586 struct kvec *vec, size_t num, size_t size)
587{
588 mm_segment_t oldfs = get_fs();
589 int result;
590
591 set_fs(KERNEL_DS);
592 /*
593 * the following is safe, since for compiler definitions of kvec and
594 * iovec are identical, yielding the same in-core layout and alignment
595 */
89bddce5 596 msg->msg_iov = (struct iovec *)vec;
1da177e4
LT
597 msg->msg_iovlen = num;
598 result = sock_sendmsg(sock, msg, size);
599 set_fs(oldfs);
600 return result;
601}
c6d409cf 602EXPORT_SYMBOL(kernel_sendmsg);
1da177e4 603
20d49473
PO
604static int ktime2ts(ktime_t kt, struct timespec *ts)
605{
606 if (kt.tv64) {
607 *ts = ktime_to_timespec(kt);
608 return 1;
609 } else {
610 return 0;
611 }
612}
613
92f37fd2
ED
614/*
615 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
616 */
617void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
618 struct sk_buff *skb)
619{
20d49473
PO
620 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
621 struct timespec ts[3];
622 int empty = 1;
623 struct skb_shared_hwtstamps *shhwtstamps =
624 skb_hwtstamps(skb);
625
626 /* Race occurred between timestamp enabling and packet
627 receiving. Fill in the current time for now. */
628 if (need_software_tstamp && skb->tstamp.tv64 == 0)
629 __net_timestamp(skb);
630
631 if (need_software_tstamp) {
632 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
633 struct timeval tv;
634 skb_get_timestamp(skb, &tv);
635 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
636 sizeof(tv), &tv);
637 } else {
842509b8 638 skb_get_timestampns(skb, &ts[0]);
20d49473 639 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
842509b8 640 sizeof(ts[0]), &ts[0]);
20d49473
PO
641 }
642 }
643
644
645 memset(ts, 0, sizeof(ts));
646 if (skb->tstamp.tv64 &&
647 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
648 skb_get_timestampns(skb, ts + 0);
649 empty = 0;
650 }
651 if (shhwtstamps) {
652 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
653 ktime2ts(shhwtstamps->syststamp, ts + 1))
654 empty = 0;
655 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
656 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
657 empty = 0;
92f37fd2 658 }
20d49473
PO
659 if (!empty)
660 put_cmsg(msg, SOL_SOCKET,
661 SCM_TIMESTAMPING, sizeof(ts), &ts);
92f37fd2 662}
7c81fd8b
ACM
663EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
664
3b885787
NH
665inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
666{
667 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
668 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
669 sizeof(__u32), &skb->dropcount);
670}
671
767dd033 672void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
3b885787
NH
673 struct sk_buff *skb)
674{
675 sock_recv_timestamp(msg, sk, skb);
676 sock_recv_drops(msg, sk, skb);
677}
767dd033 678EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
3b885787 679
a2e27255
ACM
680static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
681 struct msghdr *msg, size_t size, int flags)
1da177e4 682{
1da177e4
LT
683 struct sock_iocb *si = kiocb_to_siocb(iocb);
684
f8451725
HX
685 sock_update_classid(sock->sk);
686
1da177e4
LT
687 si->sock = sock;
688 si->scm = NULL;
689 si->msg = msg;
690 si->size = size;
691 si->flags = flags;
692
1da177e4
LT
693 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
694}
695
a2e27255
ACM
696static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
697 struct msghdr *msg, size_t size, int flags)
698{
699 int err = security_socket_recvmsg(sock, msg, size, flags);
700
701 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
702}
703
89bddce5 704int sock_recvmsg(struct socket *sock, struct msghdr *msg,
1da177e4
LT
705 size_t size, int flags)
706{
707 struct kiocb iocb;
708 struct sock_iocb siocb;
709 int ret;
710
89bddce5 711 init_sync_kiocb(&iocb, NULL);
1da177e4
LT
712 iocb.private = &siocb;
713 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
714 if (-EIOCBQUEUED == ret)
715 ret = wait_on_sync_kiocb(&iocb);
716 return ret;
717}
c6d409cf 718EXPORT_SYMBOL(sock_recvmsg);
1da177e4 719
a2e27255
ACM
720static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
721 size_t size, int flags)
722{
723 struct kiocb iocb;
724 struct sock_iocb siocb;
725 int ret;
726
727 init_sync_kiocb(&iocb, NULL);
728 iocb.private = &siocb;
729 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
730 if (-EIOCBQUEUED == ret)
731 ret = wait_on_sync_kiocb(&iocb);
732 return ret;
733}
734
89bddce5
SH
735int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
736 struct kvec *vec, size_t num, size_t size, int flags)
1da177e4
LT
737{
738 mm_segment_t oldfs = get_fs();
739 int result;
740
741 set_fs(KERNEL_DS);
742 /*
743 * the following is safe, since for compiler definitions of kvec and
744 * iovec are identical, yielding the same in-core layout and alignment
745 */
89bddce5 746 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
1da177e4
LT
747 result = sock_recvmsg(sock, msg, size, flags);
748 set_fs(oldfs);
749 return result;
750}
c6d409cf 751EXPORT_SYMBOL(kernel_recvmsg);
1da177e4
LT
752
753static void sock_aio_dtor(struct kiocb *iocb)
754{
755 kfree(iocb->private);
756}
757
ce1d4d3e
CH
758static ssize_t sock_sendpage(struct file *file, struct page *page,
759 int offset, size_t size, loff_t *ppos, int more)
1da177e4 760{
1da177e4
LT
761 struct socket *sock;
762 int flags;
763
ce1d4d3e
CH
764 sock = file->private_data;
765
766 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
767 if (more)
768 flags |= MSG_MORE;
769
e6949583 770 return kernel_sendpage(sock, page, offset, size, flags);
ce1d4d3e 771}
1da177e4 772
9c55e01c 773static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
c6d409cf 774 struct pipe_inode_info *pipe, size_t len,
9c55e01c
JA
775 unsigned int flags)
776{
777 struct socket *sock = file->private_data;
778
997b37da
RDC
779 if (unlikely(!sock->ops->splice_read))
780 return -EINVAL;
781
f8451725
HX
782 sock_update_classid(sock->sk);
783
9c55e01c
JA
784 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
785}
786
ce1d4d3e 787static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
89bddce5 788 struct sock_iocb *siocb)
ce1d4d3e
CH
789{
790 if (!is_sync_kiocb(iocb)) {
791 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
792 if (!siocb)
793 return NULL;
1da177e4
LT
794 iocb->ki_dtor = sock_aio_dtor;
795 }
1da177e4 796
ce1d4d3e 797 siocb->kiocb = iocb;
ce1d4d3e
CH
798 iocb->private = siocb;
799 return siocb;
1da177e4
LT
800}
801
ce1d4d3e 802static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
027445c3
BP
803 struct file *file, const struct iovec *iov,
804 unsigned long nr_segs)
ce1d4d3e
CH
805{
806 struct socket *sock = file->private_data;
807 size_t size = 0;
808 int i;
1da177e4 809
89bddce5
SH
810 for (i = 0; i < nr_segs; i++)
811 size += iov[i].iov_len;
1da177e4 812
ce1d4d3e
CH
813 msg->msg_name = NULL;
814 msg->msg_namelen = 0;
815 msg->msg_control = NULL;
816 msg->msg_controllen = 0;
89bddce5 817 msg->msg_iov = (struct iovec *)iov;
ce1d4d3e
CH
818 msg->msg_iovlen = nr_segs;
819 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
820
821 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
822}
823
027445c3
BP
824static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
825 unsigned long nr_segs, loff_t pos)
ce1d4d3e
CH
826{
827 struct sock_iocb siocb, *x;
828
1da177e4
LT
829 if (pos != 0)
830 return -ESPIPE;
027445c3
BP
831
832 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
1da177e4
LT
833 return 0;
834
027445c3
BP
835
836 x = alloc_sock_iocb(iocb, &siocb);
ce1d4d3e
CH
837 if (!x)
838 return -ENOMEM;
027445c3 839 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
1da177e4
LT
840}
841
ce1d4d3e 842static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
027445c3
BP
843 struct file *file, const struct iovec *iov,
844 unsigned long nr_segs)
1da177e4 845{
ce1d4d3e
CH
846 struct socket *sock = file->private_data;
847 size_t size = 0;
848 int i;
1da177e4 849
89bddce5
SH
850 for (i = 0; i < nr_segs; i++)
851 size += iov[i].iov_len;
1da177e4 852
ce1d4d3e
CH
853 msg->msg_name = NULL;
854 msg->msg_namelen = 0;
855 msg->msg_control = NULL;
856 msg->msg_controllen = 0;
89bddce5 857 msg->msg_iov = (struct iovec *)iov;
ce1d4d3e
CH
858 msg->msg_iovlen = nr_segs;
859 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
860 if (sock->type == SOCK_SEQPACKET)
861 msg->msg_flags |= MSG_EOR;
1da177e4 862
ce1d4d3e 863 return __sock_sendmsg(iocb, sock, msg, size);
1da177e4
LT
864}
865
027445c3
BP
866static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
867 unsigned long nr_segs, loff_t pos)
ce1d4d3e
CH
868{
869 struct sock_iocb siocb, *x;
1da177e4 870
ce1d4d3e
CH
871 if (pos != 0)
872 return -ESPIPE;
027445c3 873
027445c3 874 x = alloc_sock_iocb(iocb, &siocb);
ce1d4d3e
CH
875 if (!x)
876 return -ENOMEM;
1da177e4 877
027445c3 878 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
1da177e4
LT
879}
880
1da177e4
LT
881/*
882 * Atomic setting of ioctl hooks to avoid race
883 * with module unload.
884 */
885
4a3e2f71 886static DEFINE_MUTEX(br_ioctl_mutex);
c6d409cf 887static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
1da177e4 888
881d966b 889void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
1da177e4 890{
4a3e2f71 891 mutex_lock(&br_ioctl_mutex);
1da177e4 892 br_ioctl_hook = hook;
4a3e2f71 893 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
894}
895EXPORT_SYMBOL(brioctl_set);
896
4a3e2f71 897static DEFINE_MUTEX(vlan_ioctl_mutex);
881d966b 898static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1da177e4 899
881d966b 900void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1da177e4 901{
4a3e2f71 902 mutex_lock(&vlan_ioctl_mutex);
1da177e4 903 vlan_ioctl_hook = hook;
4a3e2f71 904 mutex_unlock(&vlan_ioctl_mutex);
1da177e4
LT
905}
906EXPORT_SYMBOL(vlan_ioctl_set);
907
4a3e2f71 908static DEFINE_MUTEX(dlci_ioctl_mutex);
89bddce5 909static int (*dlci_ioctl_hook) (unsigned int, void __user *);
1da177e4 910
89bddce5 911void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
1da177e4 912{
4a3e2f71 913 mutex_lock(&dlci_ioctl_mutex);
1da177e4 914 dlci_ioctl_hook = hook;
4a3e2f71 915 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
916}
917EXPORT_SYMBOL(dlci_ioctl_set);
918
6b96018b
AB
919static long sock_do_ioctl(struct net *net, struct socket *sock,
920 unsigned int cmd, unsigned long arg)
921{
922 int err;
923 void __user *argp = (void __user *)arg;
924
925 err = sock->ops->ioctl(sock, cmd, arg);
926
927 /*
928 * If this ioctl is unknown try to hand it down
929 * to the NIC driver.
930 */
931 if (err == -ENOIOCTLCMD)
932 err = dev_ioctl(net, cmd, argp);
933
934 return err;
935}
936
1da177e4
LT
937/*
938 * With an ioctl, arg may well be a user mode pointer, but we don't know
939 * what to do with it - that's up to the protocol still.
940 */
941
942static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
943{
944 struct socket *sock;
881d966b 945 struct sock *sk;
1da177e4
LT
946 void __user *argp = (void __user *)arg;
947 int pid, err;
881d966b 948 struct net *net;
1da177e4 949
b69aee04 950 sock = file->private_data;
881d966b 951 sk = sock->sk;
3b1e0a65 952 net = sock_net(sk);
1da177e4 953 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
881d966b 954 err = dev_ioctl(net, cmd, argp);
1da177e4 955 } else
3d23e349 956#ifdef CONFIG_WEXT_CORE
1da177e4 957 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
881d966b 958 err = dev_ioctl(net, cmd, argp);
1da177e4 959 } else
3d23e349 960#endif
89bddce5 961 switch (cmd) {
1da177e4
LT
962 case FIOSETOWN:
963 case SIOCSPGRP:
964 err = -EFAULT;
965 if (get_user(pid, (int __user *)argp))
966 break;
967 err = f_setown(sock->file, pid, 1);
968 break;
969 case FIOGETOWN:
970 case SIOCGPGRP:
609d7fa9 971 err = put_user(f_getown(sock->file),
89bddce5 972 (int __user *)argp);
1da177e4
LT
973 break;
974 case SIOCGIFBR:
975 case SIOCSIFBR:
976 case SIOCBRADDBR:
977 case SIOCBRDELBR:
978 err = -ENOPKG;
979 if (!br_ioctl_hook)
980 request_module("bridge");
981
4a3e2f71 982 mutex_lock(&br_ioctl_mutex);
89bddce5 983 if (br_ioctl_hook)
881d966b 984 err = br_ioctl_hook(net, cmd, argp);
4a3e2f71 985 mutex_unlock(&br_ioctl_mutex);
1da177e4
LT
986 break;
987 case SIOCGIFVLAN:
988 case SIOCSIFVLAN:
989 err = -ENOPKG;
990 if (!vlan_ioctl_hook)
991 request_module("8021q");
992
4a3e2f71 993 mutex_lock(&vlan_ioctl_mutex);
1da177e4 994 if (vlan_ioctl_hook)
881d966b 995 err = vlan_ioctl_hook(net, argp);
4a3e2f71 996 mutex_unlock(&vlan_ioctl_mutex);
1da177e4 997 break;
1da177e4
LT
998 case SIOCADDDLCI:
999 case SIOCDELDLCI:
1000 err = -ENOPKG;
1001 if (!dlci_ioctl_hook)
1002 request_module("dlci");
1003
7512cbf6
PE
1004 mutex_lock(&dlci_ioctl_mutex);
1005 if (dlci_ioctl_hook)
1da177e4 1006 err = dlci_ioctl_hook(cmd, argp);
7512cbf6 1007 mutex_unlock(&dlci_ioctl_mutex);
1da177e4
LT
1008 break;
1009 default:
6b96018b 1010 err = sock_do_ioctl(net, sock, cmd, arg);
1da177e4 1011 break;
89bddce5 1012 }
1da177e4
LT
1013 return err;
1014}
1015
1016int sock_create_lite(int family, int type, int protocol, struct socket **res)
1017{
1018 int err;
1019 struct socket *sock = NULL;
89bddce5 1020
1da177e4
LT
1021 err = security_socket_create(family, type, protocol, 1);
1022 if (err)
1023 goto out;
1024
1025 sock = sock_alloc();
1026 if (!sock) {
1027 err = -ENOMEM;
1028 goto out;
1029 }
1030
1da177e4 1031 sock->type = type;
7420ed23
VY
1032 err = security_socket_post_create(sock, family, type, protocol, 1);
1033 if (err)
1034 goto out_release;
1035
1da177e4
LT
1036out:
1037 *res = sock;
1038 return err;
7420ed23
VY
1039out_release:
1040 sock_release(sock);
1041 sock = NULL;
1042 goto out;
1da177e4 1043}
c6d409cf 1044EXPORT_SYMBOL(sock_create_lite);
1da177e4
LT
1045
1046/* No kernel lock held - perfect */
89bddce5 1047static unsigned int sock_poll(struct file *file, poll_table *wait)
1da177e4
LT
1048{
1049 struct socket *sock;
1050
1051 /*
89bddce5 1052 * We can't return errors to poll, so it's either yes or no.
1da177e4 1053 */
b69aee04 1054 sock = file->private_data;
1da177e4
LT
1055 return sock->ops->poll(file, sock, wait);
1056}
1057
89bddce5 1058static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1da177e4 1059{
b69aee04 1060 struct socket *sock = file->private_data;
1da177e4
LT
1061
1062 return sock->ops->mmap(file, sock, vma);
1063}
1064
20380731 1065static int sock_close(struct inode *inode, struct file *filp)
1da177e4
LT
1066{
1067 /*
89bddce5
SH
1068 * It was possible the inode is NULL we were
1069 * closing an unfinished socket.
1da177e4
LT
1070 */
1071
89bddce5 1072 if (!inode) {
1da177e4
LT
1073 printk(KERN_DEBUG "sock_close: NULL inode\n");
1074 return 0;
1075 }
1da177e4
LT
1076 sock_release(SOCKET_I(inode));
1077 return 0;
1078}
1079
1080/*
1081 * Update the socket async list
1082 *
1083 * Fasync_list locking strategy.
1084 *
1085 * 1. fasync_list is modified only under process context socket lock
1086 * i.e. under semaphore.
1087 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
989a2979 1088 * or under socket lock
1da177e4
LT
1089 */
1090
1091static int sock_fasync(int fd, struct file *filp, int on)
1092{
989a2979
ED
1093 struct socket *sock = filp->private_data;
1094 struct sock *sk = sock->sk;
1da177e4 1095
989a2979 1096 if (sk == NULL)
1da177e4 1097 return -EINVAL;
1da177e4
LT
1098
1099 lock_sock(sk);
1100
43815482 1101 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
1da177e4 1102
43815482 1103 if (!sock->wq->fasync_list)
989a2979
ED
1104 sock_reset_flag(sk, SOCK_FASYNC);
1105 else
bcdce719 1106 sock_set_flag(sk, SOCK_FASYNC);
1da177e4 1107
989a2979 1108 release_sock(sk);
1da177e4
LT
1109 return 0;
1110}
1111
43815482 1112/* This function may be called only under socket lock or callback_lock or rcu_lock */
1da177e4
LT
1113
1114int sock_wake_async(struct socket *sock, int how, int band)
1115{
43815482
ED
1116 struct socket_wq *wq;
1117
1118 if (!sock)
1119 return -1;
1120 rcu_read_lock();
1121 wq = rcu_dereference(sock->wq);
1122 if (!wq || !wq->fasync_list) {
1123 rcu_read_unlock();
1da177e4 1124 return -1;
43815482 1125 }
89bddce5 1126 switch (how) {
8d8ad9d7 1127 case SOCK_WAKE_WAITD:
1da177e4
LT
1128 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1129 break;
1130 goto call_kill;
8d8ad9d7 1131 case SOCK_WAKE_SPACE:
1da177e4
LT
1132 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1133 break;
1134 /* fall through */
8d8ad9d7 1135 case SOCK_WAKE_IO:
89bddce5 1136call_kill:
43815482 1137 kill_fasync(&wq->fasync_list, SIGIO, band);
1da177e4 1138 break;
8d8ad9d7 1139 case SOCK_WAKE_URG:
43815482 1140 kill_fasync(&wq->fasync_list, SIGURG, band);
1da177e4 1141 }
43815482 1142 rcu_read_unlock();
1da177e4
LT
1143 return 0;
1144}
c6d409cf 1145EXPORT_SYMBOL(sock_wake_async);
1da177e4 1146
1b8d7ae4 1147static int __sock_create(struct net *net, int family, int type, int protocol,
89bddce5 1148 struct socket **res, int kern)
1da177e4
LT
1149{
1150 int err;
1151 struct socket *sock;
55737fda 1152 const struct net_proto_family *pf;
1da177e4
LT
1153
1154 /*
89bddce5 1155 * Check protocol is in range
1da177e4
LT
1156 */
1157 if (family < 0 || family >= NPROTO)
1158 return -EAFNOSUPPORT;
1159 if (type < 0 || type >= SOCK_MAX)
1160 return -EINVAL;
1161
1162 /* Compatibility.
1163
1164 This uglymoron is moved from INET layer to here to avoid
1165 deadlock in module load.
1166 */
1167 if (family == PF_INET && type == SOCK_PACKET) {
89bddce5 1168 static int warned;
1da177e4
LT
1169 if (!warned) {
1170 warned = 1;
89bddce5
SH
1171 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1172 current->comm);
1da177e4
LT
1173 }
1174 family = PF_PACKET;
1175 }
1176
1177 err = security_socket_create(family, type, protocol, kern);
1178 if (err)
1179 return err;
89bddce5 1180
55737fda
SH
1181 /*
1182 * Allocate the socket and allow the family to set things up. if
1183 * the protocol is 0, the family is instructed to select an appropriate
1184 * default.
1185 */
1186 sock = sock_alloc();
1187 if (!sock) {
1188 if (net_ratelimit())
1189 printk(KERN_WARNING "socket: no more sockets\n");
1190 return -ENFILE; /* Not exactly a match, but its the
1191 closest posix thing */
1192 }
1193
1194 sock->type = type;
1195
95a5afca 1196#ifdef CONFIG_MODULES
89bddce5
SH
1197 /* Attempt to load a protocol module if the find failed.
1198 *
1199 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1da177e4
LT
1200 * requested real, full-featured networking support upon configuration.
1201 * Otherwise module support will break!
1202 */
55737fda 1203 if (net_families[family] == NULL)
89bddce5 1204 request_module("net-pf-%d", family);
1da177e4
LT
1205#endif
1206
55737fda
SH
1207 rcu_read_lock();
1208 pf = rcu_dereference(net_families[family]);
1209 err = -EAFNOSUPPORT;
1210 if (!pf)
1211 goto out_release;
1da177e4
LT
1212
1213 /*
1214 * We will call the ->create function, that possibly is in a loadable
1215 * module, so we have to bump that loadable module refcnt first.
1216 */
55737fda 1217 if (!try_module_get(pf->owner))
1da177e4
LT
1218 goto out_release;
1219
55737fda
SH
1220 /* Now protected by module ref count */
1221 rcu_read_unlock();
1222
3f378b68 1223 err = pf->create(net, sock, protocol, kern);
55737fda 1224 if (err < 0)
1da177e4 1225 goto out_module_put;
a79af59e 1226
1da177e4
LT
1227 /*
1228 * Now to bump the refcnt of the [loadable] module that owns this
1229 * socket at sock_release time we decrement its refcnt.
1230 */
55737fda
SH
1231 if (!try_module_get(sock->ops->owner))
1232 goto out_module_busy;
1233
1da177e4
LT
1234 /*
1235 * Now that we're done with the ->create function, the [loadable]
1236 * module can have its refcnt decremented
1237 */
55737fda 1238 module_put(pf->owner);
7420ed23
VY
1239 err = security_socket_post_create(sock, family, type, protocol, kern);
1240 if (err)
3b185525 1241 goto out_sock_release;
55737fda 1242 *res = sock;
1da177e4 1243
55737fda
SH
1244 return 0;
1245
1246out_module_busy:
1247 err = -EAFNOSUPPORT;
1da177e4 1248out_module_put:
55737fda
SH
1249 sock->ops = NULL;
1250 module_put(pf->owner);
1251out_sock_release:
1da177e4 1252 sock_release(sock);
55737fda
SH
1253 return err;
1254
1255out_release:
1256 rcu_read_unlock();
1257 goto out_sock_release;
1da177e4
LT
1258}
1259
1260int sock_create(int family, int type, int protocol, struct socket **res)
1261{
1b8d7ae4 1262 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1da177e4 1263}
c6d409cf 1264EXPORT_SYMBOL(sock_create);
1da177e4
LT
1265
1266int sock_create_kern(int family, int type, int protocol, struct socket **res)
1267{
1b8d7ae4 1268 return __sock_create(&init_net, family, type, protocol, res, 1);
1da177e4 1269}
c6d409cf 1270EXPORT_SYMBOL(sock_create_kern);
1da177e4 1271
3e0fa65f 1272SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1da177e4
LT
1273{
1274 int retval;
1275 struct socket *sock;
a677a039
UD
1276 int flags;
1277
e38b36f3
UD
1278 /* Check the SOCK_* constants for consistency. */
1279 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1280 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1281 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1282 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1283
a677a039 1284 flags = type & ~SOCK_TYPE_MASK;
77d27200 1285 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
a677a039
UD
1286 return -EINVAL;
1287 type &= SOCK_TYPE_MASK;
1da177e4 1288
aaca0bdc
UD
1289 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1290 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1291
1da177e4
LT
1292 retval = sock_create(family, type, protocol, &sock);
1293 if (retval < 0)
1294 goto out;
1295
77d27200 1296 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1da177e4
LT
1297 if (retval < 0)
1298 goto out_release;
1299
1300out:
1301 /* It may be already another descriptor 8) Not kernel problem. */
1302 return retval;
1303
1304out_release:
1305 sock_release(sock);
1306 return retval;
1307}
1308
1309/*
1310 * Create a pair of connected sockets.
1311 */
1312
3e0fa65f
HC
1313SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1314 int __user *, usockvec)
1da177e4
LT
1315{
1316 struct socket *sock1, *sock2;
1317 int fd1, fd2, err;
db349509 1318 struct file *newfile1, *newfile2;
a677a039
UD
1319 int flags;
1320
1321 flags = type & ~SOCK_TYPE_MASK;
77d27200 1322 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
a677a039
UD
1323 return -EINVAL;
1324 type &= SOCK_TYPE_MASK;
1da177e4 1325
aaca0bdc
UD
1326 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1327 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1328
1da177e4
LT
1329 /*
1330 * Obtain the first socket and check if the underlying protocol
1331 * supports the socketpair call.
1332 */
1333
1334 err = sock_create(family, type, protocol, &sock1);
1335 if (err < 0)
1336 goto out;
1337
1338 err = sock_create(family, type, protocol, &sock2);
1339 if (err < 0)
1340 goto out_release_1;
1341
1342 err = sock1->ops->socketpair(sock1, sock2);
89bddce5 1343 if (err < 0)
1da177e4
LT
1344 goto out_release_both;
1345
7cbe66b6 1346 fd1 = sock_alloc_file(sock1, &newfile1, flags);
bf3c23d1
DM
1347 if (unlikely(fd1 < 0)) {
1348 err = fd1;
db349509 1349 goto out_release_both;
bf3c23d1 1350 }
1da177e4 1351
7cbe66b6 1352 fd2 = sock_alloc_file(sock2, &newfile2, flags);
198de4d7
AV
1353 if (unlikely(fd2 < 0)) {
1354 err = fd2;
1355 fput(newfile1);
1356 put_unused_fd(fd1);
1357 sock_release(sock2);
1358 goto out;
db349509
AV
1359 }
1360
157cf649 1361 audit_fd_pair(fd1, fd2);
db349509
AV
1362 fd_install(fd1, newfile1);
1363 fd_install(fd2, newfile2);
1da177e4
LT
1364 /* fd1 and fd2 may be already another descriptors.
1365 * Not kernel problem.
1366 */
1367
89bddce5 1368 err = put_user(fd1, &usockvec[0]);
1da177e4
LT
1369 if (!err)
1370 err = put_user(fd2, &usockvec[1]);
1371 if (!err)
1372 return 0;
1373
1374 sys_close(fd2);
1375 sys_close(fd1);
1376 return err;
1377
1da177e4 1378out_release_both:
89bddce5 1379 sock_release(sock2);
1da177e4 1380out_release_1:
89bddce5 1381 sock_release(sock1);
1da177e4
LT
1382out:
1383 return err;
1384}
1385
1da177e4
LT
1386/*
1387 * Bind a name to a socket. Nothing much to do here since it's
1388 * the protocol's responsibility to handle the local address.
1389 *
1390 * We move the socket address to kernel space before we call
1391 * the protocol layer (having also checked the address is ok).
1392 */
1393
20f37034 1394SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1da177e4
LT
1395{
1396 struct socket *sock;
230b1839 1397 struct sockaddr_storage address;
6cb153ca 1398 int err, fput_needed;
1da177e4 1399
89bddce5 1400 sock = sockfd_lookup_light(fd, &err, &fput_needed);
e71a4783 1401 if (sock) {
230b1839 1402 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
89bddce5
SH
1403 if (err >= 0) {
1404 err = security_socket_bind(sock,
230b1839 1405 (struct sockaddr *)&address,
89bddce5 1406 addrlen);
6cb153ca
BL
1407 if (!err)
1408 err = sock->ops->bind(sock,
89bddce5 1409 (struct sockaddr *)
230b1839 1410 &address, addrlen);
1da177e4 1411 }
6cb153ca 1412 fput_light(sock->file, fput_needed);
89bddce5 1413 }
1da177e4
LT
1414 return err;
1415}
1416
1da177e4
LT
1417/*
1418 * Perform a listen. Basically, we allow the protocol to do anything
1419 * necessary for a listen, and if that works, we mark the socket as
1420 * ready for listening.
1421 */
1422
3e0fa65f 1423SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1da177e4
LT
1424{
1425 struct socket *sock;
6cb153ca 1426 int err, fput_needed;
b8e1f9b5 1427 int somaxconn;
89bddce5
SH
1428
1429 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1430 if (sock) {
8efa6e93 1431 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
b8e1f9b5
PE
1432 if ((unsigned)backlog > somaxconn)
1433 backlog = somaxconn;
1da177e4
LT
1434
1435 err = security_socket_listen(sock, backlog);
6cb153ca
BL
1436 if (!err)
1437 err = sock->ops->listen(sock, backlog);
1da177e4 1438
6cb153ca 1439 fput_light(sock->file, fput_needed);
1da177e4
LT
1440 }
1441 return err;
1442}
1443
1da177e4
LT
1444/*
1445 * For accept, we attempt to create a new socket, set up the link
1446 * with the client, wake up the client, then return the new
1447 * connected fd. We collect the address of the connector in kernel
1448 * space and move it to user at the very end. This is unclean because
1449 * we open the socket then return an error.
1450 *
1451 * 1003.1g adds the ability to recvmsg() to query connection pending
1452 * status to recvmsg. We need to add that support in a way thats
1453 * clean when we restucture accept also.
1454 */
1455
20f37034
HC
1456SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1457 int __user *, upeer_addrlen, int, flags)
1da177e4
LT
1458{
1459 struct socket *sock, *newsock;
39d8c1b6 1460 struct file *newfile;
6cb153ca 1461 int err, len, newfd, fput_needed;
230b1839 1462 struct sockaddr_storage address;
1da177e4 1463
77d27200 1464 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
aaca0bdc
UD
1465 return -EINVAL;
1466
1467 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1468 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1469
6cb153ca 1470 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1471 if (!sock)
1472 goto out;
1473
1474 err = -ENFILE;
c6d409cf
ED
1475 newsock = sock_alloc();
1476 if (!newsock)
1da177e4
LT
1477 goto out_put;
1478
1479 newsock->type = sock->type;
1480 newsock->ops = sock->ops;
1481
1da177e4
LT
1482 /*
1483 * We don't need try_module_get here, as the listening socket (sock)
1484 * has the protocol module (sock->ops->owner) held.
1485 */
1486 __module_get(newsock->ops->owner);
1487
7cbe66b6 1488 newfd = sock_alloc_file(newsock, &newfile, flags);
39d8c1b6
DM
1489 if (unlikely(newfd < 0)) {
1490 err = newfd;
9a1875e6
DM
1491 sock_release(newsock);
1492 goto out_put;
39d8c1b6
DM
1493 }
1494
a79af59e
FF
1495 err = security_socket_accept(sock, newsock);
1496 if (err)
39d8c1b6 1497 goto out_fd;
a79af59e 1498
1da177e4
LT
1499 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1500 if (err < 0)
39d8c1b6 1501 goto out_fd;
1da177e4
LT
1502
1503 if (upeer_sockaddr) {
230b1839 1504 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
89bddce5 1505 &len, 2) < 0) {
1da177e4 1506 err = -ECONNABORTED;
39d8c1b6 1507 goto out_fd;
1da177e4 1508 }
230b1839
YH
1509 err = move_addr_to_user((struct sockaddr *)&address,
1510 len, upeer_sockaddr, upeer_addrlen);
1da177e4 1511 if (err < 0)
39d8c1b6 1512 goto out_fd;
1da177e4
LT
1513 }
1514
1515 /* File flags are not inherited via accept() unlike another OSes. */
1516
39d8c1b6
DM
1517 fd_install(newfd, newfile);
1518 err = newfd;
1da177e4 1519
1da177e4 1520out_put:
6cb153ca 1521 fput_light(sock->file, fput_needed);
1da177e4
LT
1522out:
1523 return err;
39d8c1b6 1524out_fd:
9606a216 1525 fput(newfile);
39d8c1b6 1526 put_unused_fd(newfd);
1da177e4
LT
1527 goto out_put;
1528}
1529
20f37034
HC
1530SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1531 int __user *, upeer_addrlen)
aaca0bdc 1532{
de11defe 1533 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
aaca0bdc
UD
1534}
1535
1da177e4
LT
1536/*
1537 * Attempt to connect to a socket with the server address. The address
1538 * is in user space so we verify it is OK and move it to kernel space.
1539 *
1540 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1541 * break bindings
1542 *
1543 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1544 * other SEQPACKET protocols that take time to connect() as it doesn't
1545 * include the -EINPROGRESS status for such sockets.
1546 */
1547
20f37034
HC
1548SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1549 int, addrlen)
1da177e4
LT
1550{
1551 struct socket *sock;
230b1839 1552 struct sockaddr_storage address;
6cb153ca 1553 int err, fput_needed;
1da177e4 1554
6cb153ca 1555 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1556 if (!sock)
1557 goto out;
230b1839 1558 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1da177e4
LT
1559 if (err < 0)
1560 goto out_put;
1561
89bddce5 1562 err =
230b1839 1563 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1da177e4
LT
1564 if (err)
1565 goto out_put;
1566
230b1839 1567 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1da177e4
LT
1568 sock->file->f_flags);
1569out_put:
6cb153ca 1570 fput_light(sock->file, fput_needed);
1da177e4
LT
1571out:
1572 return err;
1573}
1574
1575/*
1576 * Get the local address ('name') of a socket object. Move the obtained
1577 * name to user space.
1578 */
1579
20f37034
HC
1580SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1581 int __user *, usockaddr_len)
1da177e4
LT
1582{
1583 struct socket *sock;
230b1839 1584 struct sockaddr_storage address;
6cb153ca 1585 int len, err, fput_needed;
89bddce5 1586
6cb153ca 1587 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4
LT
1588 if (!sock)
1589 goto out;
1590
1591 err = security_socket_getsockname(sock);
1592 if (err)
1593 goto out_put;
1594
230b1839 1595 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1da177e4
LT
1596 if (err)
1597 goto out_put;
230b1839 1598 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1da177e4
LT
1599
1600out_put:
6cb153ca 1601 fput_light(sock->file, fput_needed);
1da177e4
LT
1602out:
1603 return err;
1604}
1605
1606/*
1607 * Get the remote address ('name') of a socket object. Move the obtained
1608 * name to user space.
1609 */
1610
20f37034
HC
1611SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1612 int __user *, usockaddr_len)
1da177e4
LT
1613{
1614 struct socket *sock;
230b1839 1615 struct sockaddr_storage address;
6cb153ca 1616 int len, err, fput_needed;
1da177e4 1617
89bddce5
SH
1618 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1619 if (sock != NULL) {
1da177e4
LT
1620 err = security_socket_getpeername(sock);
1621 if (err) {
6cb153ca 1622 fput_light(sock->file, fput_needed);
1da177e4
LT
1623 return err;
1624 }
1625
89bddce5 1626 err =
230b1839 1627 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
89bddce5 1628 1);
1da177e4 1629 if (!err)
230b1839 1630 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
89bddce5 1631 usockaddr_len);
6cb153ca 1632 fput_light(sock->file, fput_needed);
1da177e4
LT
1633 }
1634 return err;
1635}
1636
1637/*
1638 * Send a datagram to a given address. We move the address into kernel
1639 * space and check the user space data area is readable before invoking
1640 * the protocol.
1641 */
1642
3e0fa65f
HC
1643SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1644 unsigned, flags, struct sockaddr __user *, addr,
1645 int, addr_len)
1da177e4
LT
1646{
1647 struct socket *sock;
230b1839 1648 struct sockaddr_storage address;
1da177e4
LT
1649 int err;
1650 struct msghdr msg;
1651 struct iovec iov;
6cb153ca 1652 int fput_needed;
6cb153ca 1653
de0fa95c
PE
1654 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1655 if (!sock)
4387ff75 1656 goto out;
6cb153ca 1657
89bddce5
SH
1658 iov.iov_base = buff;
1659 iov.iov_len = len;
1660 msg.msg_name = NULL;
1661 msg.msg_iov = &iov;
1662 msg.msg_iovlen = 1;
1663 msg.msg_control = NULL;
1664 msg.msg_controllen = 0;
1665 msg.msg_namelen = 0;
6cb153ca 1666 if (addr) {
230b1839 1667 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1da177e4
LT
1668 if (err < 0)
1669 goto out_put;
230b1839 1670 msg.msg_name = (struct sockaddr *)&address;
89bddce5 1671 msg.msg_namelen = addr_len;
1da177e4
LT
1672 }
1673 if (sock->file->f_flags & O_NONBLOCK)
1674 flags |= MSG_DONTWAIT;
1675 msg.msg_flags = flags;
1676 err = sock_sendmsg(sock, &msg, len);
1677
89bddce5 1678out_put:
de0fa95c 1679 fput_light(sock->file, fput_needed);
4387ff75 1680out:
1da177e4
LT
1681 return err;
1682}
1683
1684/*
89bddce5 1685 * Send a datagram down a socket.
1da177e4
LT
1686 */
1687
3e0fa65f
HC
1688SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1689 unsigned, flags)
1da177e4
LT
1690{
1691 return sys_sendto(fd, buff, len, flags, NULL, 0);
1692}
1693
1694/*
89bddce5 1695 * Receive a frame from the socket and optionally record the address of the
1da177e4
LT
1696 * sender. We verify the buffers are writable and if needed move the
1697 * sender address from kernel to user space.
1698 */
1699
3e0fa65f
HC
1700SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1701 unsigned, flags, struct sockaddr __user *, addr,
1702 int __user *, addr_len)
1da177e4
LT
1703{
1704 struct socket *sock;
1705 struct iovec iov;
1706 struct msghdr msg;
230b1839 1707 struct sockaddr_storage address;
89bddce5 1708 int err, err2;
6cb153ca
BL
1709 int fput_needed;
1710
de0fa95c 1711 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1da177e4 1712 if (!sock)
de0fa95c 1713 goto out;
1da177e4 1714
89bddce5
SH
1715 msg.msg_control = NULL;
1716 msg.msg_controllen = 0;
1717 msg.msg_iovlen = 1;
1718 msg.msg_iov = &iov;
1719 iov.iov_len = size;
1720 iov.iov_base = ubuf;
230b1839
YH
1721 msg.msg_name = (struct sockaddr *)&address;
1722 msg.msg_namelen = sizeof(address);
1da177e4
LT
1723 if (sock->file->f_flags & O_NONBLOCK)
1724 flags |= MSG_DONTWAIT;
89bddce5 1725 err = sock_recvmsg(sock, &msg, size, flags);
1da177e4 1726
89bddce5 1727 if (err >= 0 && addr != NULL) {
230b1839
YH
1728 err2 = move_addr_to_user((struct sockaddr *)&address,
1729 msg.msg_namelen, addr, addr_len);
89bddce5
SH
1730 if (err2 < 0)
1731 err = err2;
1da177e4 1732 }
de0fa95c
PE
1733
1734 fput_light(sock->file, fput_needed);
4387ff75 1735out:
1da177e4
LT
1736 return err;
1737}
1738
1739/*
89bddce5 1740 * Receive a datagram from a socket.
1da177e4
LT
1741 */
1742
89bddce5
SH
1743asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1744 unsigned flags)
1da177e4
LT
1745{
1746 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1747}
1748
1749/*
1750 * Set a socket option. Because we don't know the option lengths we have
1751 * to pass the user mode parameter for the protocols to sort out.
1752 */
1753
20f37034
HC
1754SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1755 char __user *, optval, int, optlen)
1da177e4 1756{
6cb153ca 1757 int err, fput_needed;
1da177e4
LT
1758 struct socket *sock;
1759
1760 if (optlen < 0)
1761 return -EINVAL;
89bddce5
SH
1762
1763 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1764 if (sock != NULL) {
1765 err = security_socket_setsockopt(sock, level, optname);
6cb153ca
BL
1766 if (err)
1767 goto out_put;
1da177e4
LT
1768
1769 if (level == SOL_SOCKET)
89bddce5
SH
1770 err =
1771 sock_setsockopt(sock, level, optname, optval,
1772 optlen);
1da177e4 1773 else
89bddce5
SH
1774 err =
1775 sock->ops->setsockopt(sock, level, optname, optval,
1776 optlen);
6cb153ca
BL
1777out_put:
1778 fput_light(sock->file, fput_needed);
1da177e4
LT
1779 }
1780 return err;
1781}
1782
1783/*
1784 * Get a socket option. Because we don't know the option lengths we have
1785 * to pass a user mode parameter for the protocols to sort out.
1786 */
1787
20f37034
HC
1788SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1789 char __user *, optval, int __user *, optlen)
1da177e4 1790{
6cb153ca 1791 int err, fput_needed;
1da177e4
LT
1792 struct socket *sock;
1793
89bddce5
SH
1794 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1795 if (sock != NULL) {
6cb153ca
BL
1796 err = security_socket_getsockopt(sock, level, optname);
1797 if (err)
1798 goto out_put;
1da177e4
LT
1799
1800 if (level == SOL_SOCKET)
89bddce5
SH
1801 err =
1802 sock_getsockopt(sock, level, optname, optval,
1803 optlen);
1da177e4 1804 else
89bddce5
SH
1805 err =
1806 sock->ops->getsockopt(sock, level, optname, optval,
1807 optlen);
6cb153ca
BL
1808out_put:
1809 fput_light(sock->file, fput_needed);
1da177e4
LT
1810 }
1811 return err;
1812}
1813
1da177e4
LT
1814/*
1815 * Shutdown a socket.
1816 */
1817
754fe8d2 1818SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1da177e4 1819{
6cb153ca 1820 int err, fput_needed;
1da177e4
LT
1821 struct socket *sock;
1822
89bddce5
SH
1823 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1824 if (sock != NULL) {
1da177e4 1825 err = security_socket_shutdown(sock, how);
6cb153ca
BL
1826 if (!err)
1827 err = sock->ops->shutdown(sock, how);
1828 fput_light(sock->file, fput_needed);
1da177e4
LT
1829 }
1830 return err;
1831}
1832
89bddce5 1833/* A couple of helpful macros for getting the address of the 32/64 bit
1da177e4
LT
1834 * fields which are the same type (int / unsigned) on our platforms.
1835 */
1836#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1837#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1838#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1839
1da177e4
LT
1840/*
1841 * BSD sendmsg interface
1842 */
1843
3e0fa65f 1844SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1da177e4 1845{
89bddce5
SH
1846 struct compat_msghdr __user *msg_compat =
1847 (struct compat_msghdr __user *)msg;
1da177e4 1848 struct socket *sock;
230b1839 1849 struct sockaddr_storage address;
1da177e4 1850 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
b9d717a7 1851 unsigned char ctl[sizeof(struct cmsghdr) + 20]
89bddce5
SH
1852 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1853 /* 20 is size of ipv6_pktinfo */
1da177e4
LT
1854 unsigned char *ctl_buf = ctl;
1855 struct msghdr msg_sys;
1856 int err, ctl_len, iov_size, total_len;
6cb153ca 1857 int fput_needed;
89bddce5 1858
1da177e4
LT
1859 err = -EFAULT;
1860 if (MSG_CMSG_COMPAT & flags) {
1861 if (get_compat_msghdr(&msg_sys, msg_compat))
1862 return -EFAULT;
c6d409cf 1863 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1da177e4
LT
1864 return -EFAULT;
1865
6cb153ca 1866 sock = sockfd_lookup_light(fd, &err, &fput_needed);
89bddce5 1867 if (!sock)
1da177e4
LT
1868 goto out;
1869
1870 /* do not move before msg_sys is valid */
1871 err = -EMSGSIZE;
1872 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1873 goto out_put;
1874
89bddce5 1875 /* Check whether to allocate the iovec area */
1da177e4
LT
1876 err = -ENOMEM;
1877 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1878 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1879 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1880 if (!iov)
1881 goto out_put;
1882 }
1883
1884 /* This will also move the address data into kernel space */
1885 if (MSG_CMSG_COMPAT & flags) {
230b1839
YH
1886 err = verify_compat_iovec(&msg_sys, iov,
1887 (struct sockaddr *)&address,
1888 VERIFY_READ);
1da177e4 1889 } else
230b1839
YH
1890 err = verify_iovec(&msg_sys, iov,
1891 (struct sockaddr *)&address,
1892 VERIFY_READ);
89bddce5 1893 if (err < 0)
1da177e4
LT
1894 goto out_freeiov;
1895 total_len = err;
1896
1897 err = -ENOBUFS;
1898
1899 if (msg_sys.msg_controllen > INT_MAX)
1900 goto out_freeiov;
89bddce5 1901 ctl_len = msg_sys.msg_controllen;
1da177e4 1902 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
89bddce5
SH
1903 err =
1904 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1905 sizeof(ctl));
1da177e4
LT
1906 if (err)
1907 goto out_freeiov;
1908 ctl_buf = msg_sys.msg_control;
8920e8f9 1909 ctl_len = msg_sys.msg_controllen;
1da177e4 1910 } else if (ctl_len) {
89bddce5 1911 if (ctl_len > sizeof(ctl)) {
1da177e4 1912 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
89bddce5 1913 if (ctl_buf == NULL)
1da177e4
LT
1914 goto out_freeiov;
1915 }
1916 err = -EFAULT;
1917 /*
1918 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1919 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1920 * checking falls down on this.
1921 */
89bddce5
SH
1922 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1923 ctl_len))
1da177e4
LT
1924 goto out_freectl;
1925 msg_sys.msg_control = ctl_buf;
1926 }
1927 msg_sys.msg_flags = flags;
1928
1929 if (sock->file->f_flags & O_NONBLOCK)
1930 msg_sys.msg_flags |= MSG_DONTWAIT;
1931 err = sock_sendmsg(sock, &msg_sys, total_len);
1932
1933out_freectl:
89bddce5 1934 if (ctl_buf != ctl)
1da177e4
LT
1935 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1936out_freeiov:
1937 if (iov != iovstack)
1938 sock_kfree_s(sock->sk, iov, iov_size);
1939out_put:
6cb153ca 1940 fput_light(sock->file, fput_needed);
89bddce5 1941out:
1da177e4
LT
1942 return err;
1943}
1944
a2e27255
ACM
1945static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1946 struct msghdr *msg_sys, unsigned flags, int nosec)
1da177e4 1947{
89bddce5
SH
1948 struct compat_msghdr __user *msg_compat =
1949 (struct compat_msghdr __user *)msg;
1da177e4 1950 struct iovec iovstack[UIO_FASTIOV];
89bddce5 1951 struct iovec *iov = iovstack;
1da177e4
LT
1952 unsigned long cmsg_ptr;
1953 int err, iov_size, total_len, len;
1954
1955 /* kernel mode address */
230b1839 1956 struct sockaddr_storage addr;
1da177e4
LT
1957
1958 /* user mode address pointers */
1959 struct sockaddr __user *uaddr;
1960 int __user *uaddr_len;
89bddce5 1961
1da177e4 1962 if (MSG_CMSG_COMPAT & flags) {
a2e27255 1963 if (get_compat_msghdr(msg_sys, msg_compat))
1da177e4 1964 return -EFAULT;
c6d409cf 1965 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
89bddce5 1966 return -EFAULT;
1da177e4 1967
1da177e4 1968 err = -EMSGSIZE;
a2e27255
ACM
1969 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1970 goto out;
89bddce5
SH
1971
1972 /* Check whether to allocate the iovec area */
1da177e4 1973 err = -ENOMEM;
a2e27255
ACM
1974 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1975 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
1da177e4
LT
1976 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1977 if (!iov)
a2e27255 1978 goto out;
1da177e4
LT
1979 }
1980
1981 /*
89bddce5
SH
1982 * Save the user-mode address (verify_iovec will change the
1983 * kernel msghdr to use the kernel address space)
1da177e4 1984 */
89bddce5 1985
a2e27255 1986 uaddr = (__force void __user *)msg_sys->msg_name;
1da177e4
LT
1987 uaddr_len = COMPAT_NAMELEN(msg);
1988 if (MSG_CMSG_COMPAT & flags) {
a2e27255 1989 err = verify_compat_iovec(msg_sys, iov,
230b1839
YH
1990 (struct sockaddr *)&addr,
1991 VERIFY_WRITE);
1da177e4 1992 } else
a2e27255 1993 err = verify_iovec(msg_sys, iov,
230b1839
YH
1994 (struct sockaddr *)&addr,
1995 VERIFY_WRITE);
1da177e4
LT
1996 if (err < 0)
1997 goto out_freeiov;
89bddce5 1998 total_len = err;
1da177e4 1999
a2e27255
ACM
2000 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2001 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
89bddce5 2002
1da177e4
LT
2003 if (sock->file->f_flags & O_NONBLOCK)
2004 flags |= MSG_DONTWAIT;
a2e27255
ACM
2005 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2006 total_len, flags);
1da177e4
LT
2007 if (err < 0)
2008 goto out_freeiov;
2009 len = err;
2010
2011 if (uaddr != NULL) {
230b1839 2012 err = move_addr_to_user((struct sockaddr *)&addr,
a2e27255 2013 msg_sys->msg_namelen, uaddr,
89bddce5 2014 uaddr_len);
1da177e4
LT
2015 if (err < 0)
2016 goto out_freeiov;
2017 }
a2e27255 2018 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
37f7f421 2019 COMPAT_FLAGS(msg));
1da177e4
LT
2020 if (err)
2021 goto out_freeiov;
2022 if (MSG_CMSG_COMPAT & flags)
a2e27255 2023 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
1da177e4
LT
2024 &msg_compat->msg_controllen);
2025 else
a2e27255 2026 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
1da177e4
LT
2027 &msg->msg_controllen);
2028 if (err)
2029 goto out_freeiov;
2030 err = len;
2031
2032out_freeiov:
2033 if (iov != iovstack)
2034 sock_kfree_s(sock->sk, iov, iov_size);
a2e27255
ACM
2035out:
2036 return err;
2037}
2038
2039/*
2040 * BSD recvmsg interface
2041 */
2042
2043SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2044 unsigned int, flags)
2045{
2046 int fput_needed, err;
2047 struct msghdr msg_sys;
2048 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2049
2050 if (!sock)
2051 goto out;
2052
2053 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2054
6cb153ca 2055 fput_light(sock->file, fput_needed);
1da177e4
LT
2056out:
2057 return err;
2058}
2059
a2e27255
ACM
2060/*
2061 * Linux recvmmsg interface
2062 */
2063
2064int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2065 unsigned int flags, struct timespec *timeout)
2066{
2067 int fput_needed, err, datagrams;
2068 struct socket *sock;
2069 struct mmsghdr __user *entry;
d7256d0e 2070 struct compat_mmsghdr __user *compat_entry;
a2e27255
ACM
2071 struct msghdr msg_sys;
2072 struct timespec end_time;
2073
2074 if (timeout &&
2075 poll_select_set_timeout(&end_time, timeout->tv_sec,
2076 timeout->tv_nsec))
2077 return -EINVAL;
2078
2079 datagrams = 0;
2080
2081 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2082 if (!sock)
2083 return err;
2084
2085 err = sock_error(sock->sk);
2086 if (err)
2087 goto out_put;
2088
2089 entry = mmsg;
d7256d0e 2090 compat_entry = (struct compat_mmsghdr __user *)mmsg;
a2e27255
ACM
2091
2092 while (datagrams < vlen) {
2093 /*
2094 * No need to ask LSM for more than the first datagram.
2095 */
d7256d0e
JMG
2096 if (MSG_CMSG_COMPAT & flags) {
2097 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2098 &msg_sys, flags, datagrams);
2099 if (err < 0)
2100 break;
2101 err = __put_user(err, &compat_entry->msg_len);
2102 ++compat_entry;
2103 } else {
2104 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2105 &msg_sys, flags, datagrams);
2106 if (err < 0)
2107 break;
2108 err = put_user(err, &entry->msg_len);
2109 ++entry;
2110 }
2111
a2e27255
ACM
2112 if (err)
2113 break;
a2e27255
ACM
2114 ++datagrams;
2115
71c5c159
BB
2116 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2117 if (flags & MSG_WAITFORONE)
2118 flags |= MSG_DONTWAIT;
2119
a2e27255
ACM
2120 if (timeout) {
2121 ktime_get_ts(timeout);
2122 *timeout = timespec_sub(end_time, *timeout);
2123 if (timeout->tv_sec < 0) {
2124 timeout->tv_sec = timeout->tv_nsec = 0;
2125 break;
2126 }
2127
2128 /* Timeout, return less than vlen datagrams */
2129 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2130 break;
2131 }
2132
2133 /* Out of band data, return right away */
2134 if (msg_sys.msg_flags & MSG_OOB)
2135 break;
2136 }
2137
2138out_put:
2139 fput_light(sock->file, fput_needed);
1da177e4 2140
a2e27255
ACM
2141 if (err == 0)
2142 return datagrams;
2143
2144 if (datagrams != 0) {
2145 /*
2146 * We may return less entries than requested (vlen) if the
2147 * sock is non block and there aren't enough datagrams...
2148 */
2149 if (err != -EAGAIN) {
2150 /*
2151 * ... or if recvmsg returns an error after we
2152 * received some datagrams, where we record the
2153 * error to return on the next call or if the
2154 * app asks about it using getsockopt(SO_ERROR).
2155 */
2156 sock->sk->sk_err = -err;
2157 }
2158
2159 return datagrams;
2160 }
2161
2162 return err;
2163}
2164
2165SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2166 unsigned int, vlen, unsigned int, flags,
2167 struct timespec __user *, timeout)
2168{
2169 int datagrams;
2170 struct timespec timeout_sys;
2171
2172 if (!timeout)
2173 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2174
2175 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2176 return -EFAULT;
2177
2178 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2179
2180 if (datagrams > 0 &&
2181 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2182 datagrams = -EFAULT;
2183
2184 return datagrams;
2185}
2186
2187#ifdef __ARCH_WANT_SYS_SOCKETCALL
1da177e4
LT
2188/* Argument list sizes for sys_socketcall */
2189#define AL(x) ((x) * sizeof(unsigned long))
a2e27255 2190static const unsigned char nargs[20] = {
c6d409cf
ED
2191 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2192 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2193 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2194 AL(4), AL(5)
89bddce5
SH
2195};
2196
1da177e4
LT
2197#undef AL
2198
2199/*
89bddce5 2200 * System call vectors.
1da177e4
LT
2201 *
2202 * Argument checking cleaned up. Saved 20% in size.
2203 * This function doesn't need to set the kernel lock because
89bddce5 2204 * it is set by the callees.
1da177e4
LT
2205 */
2206
3e0fa65f 2207SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
1da177e4
LT
2208{
2209 unsigned long a[6];
89bddce5 2210 unsigned long a0, a1;
1da177e4 2211 int err;
47379052 2212 unsigned int len;
1da177e4 2213
a2e27255 2214 if (call < 1 || call > SYS_RECVMMSG)
1da177e4
LT
2215 return -EINVAL;
2216
47379052
AV
2217 len = nargs[call];
2218 if (len > sizeof(a))
2219 return -EINVAL;
2220
1da177e4 2221 /* copy_from_user should be SMP safe. */
47379052 2222 if (copy_from_user(a, args, len))
1da177e4 2223 return -EFAULT;
3ec3b2fb 2224
f3298dc4 2225 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3ec3b2fb 2226
89bddce5
SH
2227 a0 = a[0];
2228 a1 = a[1];
2229
2230 switch (call) {
2231 case SYS_SOCKET:
2232 err = sys_socket(a0, a1, a[2]);
2233 break;
2234 case SYS_BIND:
2235 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2236 break;
2237 case SYS_CONNECT:
2238 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2239 break;
2240 case SYS_LISTEN:
2241 err = sys_listen(a0, a1);
2242 break;
2243 case SYS_ACCEPT:
de11defe
UD
2244 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2245 (int __user *)a[2], 0);
89bddce5
SH
2246 break;
2247 case SYS_GETSOCKNAME:
2248 err =
2249 sys_getsockname(a0, (struct sockaddr __user *)a1,
2250 (int __user *)a[2]);
2251 break;
2252 case SYS_GETPEERNAME:
2253 err =
2254 sys_getpeername(a0, (struct sockaddr __user *)a1,
2255 (int __user *)a[2]);
2256 break;
2257 case SYS_SOCKETPAIR:
2258 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2259 break;
2260 case SYS_SEND:
2261 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2262 break;
2263 case SYS_SENDTO:
2264 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2265 (struct sockaddr __user *)a[4], a[5]);
2266 break;
2267 case SYS_RECV:
2268 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2269 break;
2270 case SYS_RECVFROM:
2271 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2272 (struct sockaddr __user *)a[4],
2273 (int __user *)a[5]);
2274 break;
2275 case SYS_SHUTDOWN:
2276 err = sys_shutdown(a0, a1);
2277 break;
2278 case SYS_SETSOCKOPT:
2279 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2280 break;
2281 case SYS_GETSOCKOPT:
2282 err =
2283 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2284 (int __user *)a[4]);
2285 break;
2286 case SYS_SENDMSG:
2287 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2288 break;
2289 case SYS_RECVMSG:
2290 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2291 break;
a2e27255
ACM
2292 case SYS_RECVMMSG:
2293 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2294 (struct timespec __user *)a[4]);
2295 break;
de11defe
UD
2296 case SYS_ACCEPT4:
2297 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2298 (int __user *)a[2], a[3]);
aaca0bdc 2299 break;
89bddce5
SH
2300 default:
2301 err = -EINVAL;
2302 break;
1da177e4
LT
2303 }
2304 return err;
2305}
2306
89bddce5 2307#endif /* __ARCH_WANT_SYS_SOCKETCALL */
1da177e4 2308
55737fda
SH
2309/**
2310 * sock_register - add a socket protocol handler
2311 * @ops: description of protocol
2312 *
1da177e4
LT
2313 * This function is called by a protocol handler that wants to
2314 * advertise its address family, and have it linked into the
55737fda
SH
2315 * socket interface. The value ops->family coresponds to the
2316 * socket system call protocol family.
1da177e4 2317 */
f0fd27d4 2318int sock_register(const struct net_proto_family *ops)
1da177e4
LT
2319{
2320 int err;
2321
2322 if (ops->family >= NPROTO) {
89bddce5
SH
2323 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2324 NPROTO);
1da177e4
LT
2325 return -ENOBUFS;
2326 }
55737fda
SH
2327
2328 spin_lock(&net_family_lock);
2329 if (net_families[ops->family])
2330 err = -EEXIST;
2331 else {
89bddce5 2332 net_families[ops->family] = ops;
1da177e4
LT
2333 err = 0;
2334 }
55737fda
SH
2335 spin_unlock(&net_family_lock);
2336
89bddce5 2337 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
1da177e4
LT
2338 return err;
2339}
c6d409cf 2340EXPORT_SYMBOL(sock_register);
1da177e4 2341
55737fda
SH
2342/**
2343 * sock_unregister - remove a protocol handler
2344 * @family: protocol family to remove
2345 *
1da177e4
LT
2346 * This function is called by a protocol handler that wants to
2347 * remove its address family, and have it unlinked from the
55737fda
SH
2348 * new socket creation.
2349 *
2350 * If protocol handler is a module, then it can use module reference
2351 * counts to protect against new references. If protocol handler is not
2352 * a module then it needs to provide its own protection in
2353 * the ops->create routine.
1da177e4 2354 */
f0fd27d4 2355void sock_unregister(int family)
1da177e4 2356{
f0fd27d4 2357 BUG_ON(family < 0 || family >= NPROTO);
1da177e4 2358
55737fda 2359 spin_lock(&net_family_lock);
89bddce5 2360 net_families[family] = NULL;
55737fda
SH
2361 spin_unlock(&net_family_lock);
2362
2363 synchronize_rcu();
2364
89bddce5 2365 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
1da177e4 2366}
c6d409cf 2367EXPORT_SYMBOL(sock_unregister);
1da177e4 2368
77d76ea3 2369static int __init sock_init(void)
1da177e4
LT
2370{
2371 /*
89bddce5 2372 * Initialize sock SLAB cache.
1da177e4 2373 */
89bddce5 2374
1da177e4
LT
2375 sk_init();
2376
1da177e4 2377 /*
89bddce5 2378 * Initialize skbuff SLAB cache
1da177e4
LT
2379 */
2380 skb_init();
1da177e4
LT
2381
2382 /*
89bddce5 2383 * Initialize the protocols module.
1da177e4
LT
2384 */
2385
2386 init_inodecache();
2387 register_filesystem(&sock_fs_type);
2388 sock_mnt = kern_mount(&sock_fs_type);
77d76ea3
AK
2389
2390 /* The real protocol initialization is performed in later initcalls.
1da177e4
LT
2391 */
2392
2393#ifdef CONFIG_NETFILTER
2394 netfilter_init();
2395#endif
cbeb321a 2396
c1f19b51
RC
2397#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2398 skb_timestamping_init();
2399#endif
2400
cbeb321a 2401 return 0;
1da177e4
LT
2402}
2403
77d76ea3
AK
2404core_initcall(sock_init); /* early initcall */
2405
1da177e4
LT
2406#ifdef CONFIG_PROC_FS
2407void socket_seq_show(struct seq_file *seq)
2408{
2409 int cpu;
2410 int counter = 0;
2411
6f912042 2412 for_each_possible_cpu(cpu)
89bddce5 2413 counter += per_cpu(sockets_in_use, cpu);
1da177e4
LT
2414
2415 /* It can be negative, by the way. 8) */
2416 if (counter < 0)
2417 counter = 0;
2418
2419 seq_printf(seq, "sockets: used %d\n", counter);
2420}
89bddce5 2421#endif /* CONFIG_PROC_FS */
1da177e4 2422
89bbfc95 2423#ifdef CONFIG_COMPAT
6b96018b
AB
2424static int do_siocgstamp(struct net *net, struct socket *sock,
2425 unsigned int cmd, struct compat_timeval __user *up)
7a229387 2426{
7a229387
AB
2427 mm_segment_t old_fs = get_fs();
2428 struct timeval ktv;
2429 int err;
2430
2431 set_fs(KERNEL_DS);
6b96018b 2432 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
7a229387
AB
2433 set_fs(old_fs);
2434 if (!err) {
2435 err = put_user(ktv.tv_sec, &up->tv_sec);
2436 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2437 }
2438 return err;
2439}
2440
6b96018b
AB
2441static int do_siocgstampns(struct net *net, struct socket *sock,
2442 unsigned int cmd, struct compat_timespec __user *up)
7a229387 2443{
7a229387
AB
2444 mm_segment_t old_fs = get_fs();
2445 struct timespec kts;
2446 int err;
2447
2448 set_fs(KERNEL_DS);
6b96018b 2449 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
7a229387
AB
2450 set_fs(old_fs);
2451 if (!err) {
2452 err = put_user(kts.tv_sec, &up->tv_sec);
2453 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2454 }
2455 return err;
2456}
2457
6b96018b 2458static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
7a229387
AB
2459{
2460 struct ifreq __user *uifr;
2461 int err;
2462
2463 uifr = compat_alloc_user_space(sizeof(struct ifreq));
6b96018b 2464 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2465 return -EFAULT;
2466
6b96018b 2467 err = dev_ioctl(net, SIOCGIFNAME, uifr);
7a229387
AB
2468 if (err)
2469 return err;
2470
6b96018b 2471 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
7a229387
AB
2472 return -EFAULT;
2473
2474 return 0;
2475}
2476
6b96018b 2477static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
7a229387 2478{
6b96018b 2479 struct compat_ifconf ifc32;
7a229387
AB
2480 struct ifconf ifc;
2481 struct ifconf __user *uifc;
6b96018b 2482 struct compat_ifreq __user *ifr32;
7a229387
AB
2483 struct ifreq __user *ifr;
2484 unsigned int i, j;
2485 int err;
2486
6b96018b 2487 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
7a229387
AB
2488 return -EFAULT;
2489
2490 if (ifc32.ifcbuf == 0) {
2491 ifc32.ifc_len = 0;
2492 ifc.ifc_len = 0;
2493 ifc.ifc_req = NULL;
2494 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2495 } else {
c6d409cf
ED
2496 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2497 sizeof(struct ifreq);
7a229387
AB
2498 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2499 ifc.ifc_len = len;
2500 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2501 ifr32 = compat_ptr(ifc32.ifcbuf);
c6d409cf 2502 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
6b96018b 2503 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2504 return -EFAULT;
2505 ifr++;
2506 ifr32++;
2507 }
2508 }
2509 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2510 return -EFAULT;
2511
6b96018b 2512 err = dev_ioctl(net, SIOCGIFCONF, uifc);
7a229387
AB
2513 if (err)
2514 return err;
2515
2516 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2517 return -EFAULT;
2518
2519 ifr = ifc.ifc_req;
2520 ifr32 = compat_ptr(ifc32.ifcbuf);
2521 for (i = 0, j = 0;
c6d409cf
ED
2522 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2523 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2524 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
7a229387
AB
2525 return -EFAULT;
2526 ifr32++;
2527 ifr++;
2528 }
2529
2530 if (ifc32.ifcbuf == 0) {
2531 /* Translate from 64-bit structure multiple to
2532 * a 32-bit one.
2533 */
2534 i = ifc.ifc_len;
6b96018b 2535 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
7a229387
AB
2536 ifc32.ifc_len = i;
2537 } else {
2538 ifc32.ifc_len = i;
2539 }
6b96018b 2540 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
7a229387
AB
2541 return -EFAULT;
2542
2543 return 0;
2544}
2545
6b96018b 2546static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
7a229387
AB
2547{
2548 struct ifreq __user *ifr;
7a229387
AB
2549 u32 data;
2550 void __user *datap;
2551
2552 ifr = compat_alloc_user_space(sizeof(*ifr));
7a229387
AB
2553
2554 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2555 return -EFAULT;
2556
2557 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2558 return -EFAULT;
2559
2560 datap = compat_ptr(data);
2561 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2562 return -EFAULT;
2563
6b96018b 2564 return dev_ioctl(net, SIOCETHTOOL, ifr);
7a229387
AB
2565}
2566
7a50a240
AB
2567static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2568{
2569 void __user *uptr;
2570 compat_uptr_t uptr32;
2571 struct ifreq __user *uifr;
2572
c6d409cf 2573 uifr = compat_alloc_user_space(sizeof(*uifr));
7a50a240
AB
2574 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2575 return -EFAULT;
2576
2577 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2578 return -EFAULT;
2579
2580 uptr = compat_ptr(uptr32);
2581
2582 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2583 return -EFAULT;
2584
2585 return dev_ioctl(net, SIOCWANDEV, uifr);
2586}
2587
6b96018b
AB
2588static int bond_ioctl(struct net *net, unsigned int cmd,
2589 struct compat_ifreq __user *ifr32)
7a229387
AB
2590{
2591 struct ifreq kifr;
2592 struct ifreq __user *uifr;
7a229387
AB
2593 mm_segment_t old_fs;
2594 int err;
2595 u32 data;
2596 void __user *datap;
2597
2598 switch (cmd) {
2599 case SIOCBONDENSLAVE:
2600 case SIOCBONDRELEASE:
2601 case SIOCBONDSETHWADDR:
2602 case SIOCBONDCHANGEACTIVE:
6b96018b 2603 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
7a229387
AB
2604 return -EFAULT;
2605
2606 old_fs = get_fs();
c6d409cf 2607 set_fs(KERNEL_DS);
6b96018b 2608 err = dev_ioctl(net, cmd, &kifr);
c6d409cf 2609 set_fs(old_fs);
7a229387
AB
2610
2611 return err;
2612 case SIOCBONDSLAVEINFOQUERY:
2613 case SIOCBONDINFOQUERY:
2614 uifr = compat_alloc_user_space(sizeof(*uifr));
2615 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2616 return -EFAULT;
2617
2618 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2619 return -EFAULT;
2620
2621 datap = compat_ptr(data);
2622 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2623 return -EFAULT;
2624
6b96018b 2625 return dev_ioctl(net, cmd, uifr);
7a229387
AB
2626 default:
2627 return -EINVAL;
ccbd6a5a 2628 }
7a229387
AB
2629}
2630
6b96018b
AB
2631static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2632 struct compat_ifreq __user *u_ifreq32)
7a229387
AB
2633{
2634 struct ifreq __user *u_ifreq64;
7a229387
AB
2635 char tmp_buf[IFNAMSIZ];
2636 void __user *data64;
2637 u32 data32;
2638
2639 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2640 IFNAMSIZ))
2641 return -EFAULT;
2642 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2643 return -EFAULT;
2644 data64 = compat_ptr(data32);
2645
2646 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2647
2648 /* Don't check these user accesses, just let that get trapped
2649 * in the ioctl handler instead.
2650 */
2651 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2652 IFNAMSIZ))
2653 return -EFAULT;
2654 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2655 return -EFAULT;
2656
6b96018b 2657 return dev_ioctl(net, cmd, u_ifreq64);
7a229387
AB
2658}
2659
6b96018b
AB
2660static int dev_ifsioc(struct net *net, struct socket *sock,
2661 unsigned int cmd, struct compat_ifreq __user *uifr32)
7a229387 2662{
a2116ed2 2663 struct ifreq __user *uifr;
7a229387
AB
2664 int err;
2665
a2116ed2
AB
2666 uifr = compat_alloc_user_space(sizeof(*uifr));
2667 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2668 return -EFAULT;
2669
2670 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2671
7a229387
AB
2672 if (!err) {
2673 switch (cmd) {
2674 case SIOCGIFFLAGS:
2675 case SIOCGIFMETRIC:
2676 case SIOCGIFMTU:
2677 case SIOCGIFMEM:
2678 case SIOCGIFHWADDR:
2679 case SIOCGIFINDEX:
2680 case SIOCGIFADDR:
2681 case SIOCGIFBRDADDR:
2682 case SIOCGIFDSTADDR:
2683 case SIOCGIFNETMASK:
fab2532b 2684 case SIOCGIFPFLAGS:
7a229387 2685 case SIOCGIFTXQLEN:
fab2532b
AB
2686 case SIOCGMIIPHY:
2687 case SIOCGMIIREG:
a2116ed2 2688 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
7a229387
AB
2689 err = -EFAULT;
2690 break;
2691 }
2692 }
2693 return err;
2694}
2695
a2116ed2
AB
2696static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2697 struct compat_ifreq __user *uifr32)
2698{
2699 struct ifreq ifr;
2700 struct compat_ifmap __user *uifmap32;
2701 mm_segment_t old_fs;
2702 int err;
2703
2704 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2705 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2706 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2707 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2708 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2709 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2710 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2711 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2712 if (err)
2713 return -EFAULT;
2714
2715 old_fs = get_fs();
c6d409cf 2716 set_fs(KERNEL_DS);
a2116ed2 2717 err = dev_ioctl(net, cmd, (void __user *)&ifr);
c6d409cf 2718 set_fs(old_fs);
a2116ed2
AB
2719
2720 if (cmd == SIOCGIFMAP && !err) {
2721 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2722 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2723 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2724 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2725 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2726 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2727 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2728 if (err)
2729 err = -EFAULT;
2730 }
2731 return err;
2732}
2733
2734static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2735{
2736 void __user *uptr;
2737 compat_uptr_t uptr32;
2738 struct ifreq __user *uifr;
2739
c6d409cf 2740 uifr = compat_alloc_user_space(sizeof(*uifr));
a2116ed2
AB
2741 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2742 return -EFAULT;
2743
2744 if (get_user(uptr32, &uifr32->ifr_data))
2745 return -EFAULT;
2746
2747 uptr = compat_ptr(uptr32);
2748
2749 if (put_user(uptr, &uifr->ifr_data))
2750 return -EFAULT;
2751
2752 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2753}
2754
7a229387 2755struct rtentry32 {
c6d409cf 2756 u32 rt_pad1;
7a229387
AB
2757 struct sockaddr rt_dst; /* target address */
2758 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2759 struct sockaddr rt_genmask; /* target network mask (IP) */
c6d409cf
ED
2760 unsigned short rt_flags;
2761 short rt_pad2;
2762 u32 rt_pad3;
2763 unsigned char rt_tos;
2764 unsigned char rt_class;
2765 short rt_pad4;
2766 short rt_metric; /* +1 for binary compatibility! */
7a229387 2767 /* char * */ u32 rt_dev; /* forcing the device at add */
c6d409cf
ED
2768 u32 rt_mtu; /* per route MTU/Window */
2769 u32 rt_window; /* Window clamping */
7a229387
AB
2770 unsigned short rt_irtt; /* Initial RTT */
2771};
2772
2773struct in6_rtmsg32 {
2774 struct in6_addr rtmsg_dst;
2775 struct in6_addr rtmsg_src;
2776 struct in6_addr rtmsg_gateway;
2777 u32 rtmsg_type;
2778 u16 rtmsg_dst_len;
2779 u16 rtmsg_src_len;
2780 u32 rtmsg_metric;
2781 u32 rtmsg_info;
2782 u32 rtmsg_flags;
2783 s32 rtmsg_ifindex;
2784};
2785
6b96018b
AB
2786static int routing_ioctl(struct net *net, struct socket *sock,
2787 unsigned int cmd, void __user *argp)
7a229387
AB
2788{
2789 int ret;
2790 void *r = NULL;
2791 struct in6_rtmsg r6;
2792 struct rtentry r4;
2793 char devname[16];
2794 u32 rtdev;
2795 mm_segment_t old_fs = get_fs();
2796
6b96018b
AB
2797 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2798 struct in6_rtmsg32 __user *ur6 = argp;
c6d409cf 2799 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
7a229387 2800 3 * sizeof(struct in6_addr));
c6d409cf
ED
2801 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2802 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2803 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2804 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2805 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2806 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2807 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
7a229387
AB
2808
2809 r = (void *) &r6;
2810 } else { /* ipv4 */
6b96018b 2811 struct rtentry32 __user *ur4 = argp;
c6d409cf 2812 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
7a229387 2813 3 * sizeof(struct sockaddr));
c6d409cf
ED
2814 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2815 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2816 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2817 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2818 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2819 ret |= __get_user(rtdev, &(ur4->rt_dev));
7a229387 2820 if (rtdev) {
c6d409cf 2821 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
7a229387
AB
2822 r4.rt_dev = devname; devname[15] = 0;
2823 } else
2824 r4.rt_dev = NULL;
2825
2826 r = (void *) &r4;
2827 }
2828
2829 if (ret) {
2830 ret = -EFAULT;
2831 goto out;
2832 }
2833
c6d409cf 2834 set_fs(KERNEL_DS);
6b96018b 2835 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
c6d409cf 2836 set_fs(old_fs);
7a229387
AB
2837
2838out:
7a229387
AB
2839 return ret;
2840}
2841
2842/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2843 * for some operations; this forces use of the newer bridge-utils that
2844 * use compatiable ioctls
2845 */
6b96018b 2846static int old_bridge_ioctl(compat_ulong_t __user *argp)
7a229387 2847{
6b96018b 2848 compat_ulong_t tmp;
7a229387 2849
6b96018b 2850 if (get_user(tmp, argp))
7a229387
AB
2851 return -EFAULT;
2852 if (tmp == BRCTL_GET_VERSION)
2853 return BRCTL_VERSION + 1;
2854 return -EINVAL;
2855}
2856
6b96018b
AB
2857static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2858 unsigned int cmd, unsigned long arg)
2859{
2860 void __user *argp = compat_ptr(arg);
2861 struct sock *sk = sock->sk;
2862 struct net *net = sock_net(sk);
7a229387 2863
6b96018b
AB
2864 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2865 return siocdevprivate_ioctl(net, cmd, argp);
2866
2867 switch (cmd) {
2868 case SIOCSIFBR:
2869 case SIOCGIFBR:
2870 return old_bridge_ioctl(argp);
2871 case SIOCGIFNAME:
2872 return dev_ifname32(net, argp);
2873 case SIOCGIFCONF:
2874 return dev_ifconf(net, argp);
2875 case SIOCETHTOOL:
2876 return ethtool_ioctl(net, argp);
7a50a240
AB
2877 case SIOCWANDEV:
2878 return compat_siocwandev(net, argp);
a2116ed2
AB
2879 case SIOCGIFMAP:
2880 case SIOCSIFMAP:
2881 return compat_sioc_ifmap(net, cmd, argp);
6b96018b
AB
2882 case SIOCBONDENSLAVE:
2883 case SIOCBONDRELEASE:
2884 case SIOCBONDSETHWADDR:
2885 case SIOCBONDSLAVEINFOQUERY:
2886 case SIOCBONDINFOQUERY:
2887 case SIOCBONDCHANGEACTIVE:
2888 return bond_ioctl(net, cmd, argp);
2889 case SIOCADDRT:
2890 case SIOCDELRT:
2891 return routing_ioctl(net, sock, cmd, argp);
2892 case SIOCGSTAMP:
2893 return do_siocgstamp(net, sock, cmd, argp);
2894 case SIOCGSTAMPNS:
2895 return do_siocgstampns(net, sock, cmd, argp);
a2116ed2
AB
2896 case SIOCSHWTSTAMP:
2897 return compat_siocshwtstamp(net, argp);
6b96018b
AB
2898
2899 case FIOSETOWN:
2900 case SIOCSPGRP:
2901 case FIOGETOWN:
2902 case SIOCGPGRP:
2903 case SIOCBRADDBR:
2904 case SIOCBRDELBR:
2905 case SIOCGIFVLAN:
2906 case SIOCSIFVLAN:
2907 case SIOCADDDLCI:
2908 case SIOCDELDLCI:
2909 return sock_ioctl(file, cmd, arg);
2910
2911 case SIOCGIFFLAGS:
2912 case SIOCSIFFLAGS:
2913 case SIOCGIFMETRIC:
2914 case SIOCSIFMETRIC:
2915 case SIOCGIFMTU:
2916 case SIOCSIFMTU:
2917 case SIOCGIFMEM:
2918 case SIOCSIFMEM:
2919 case SIOCGIFHWADDR:
2920 case SIOCSIFHWADDR:
2921 case SIOCADDMULTI:
2922 case SIOCDELMULTI:
2923 case SIOCGIFINDEX:
6b96018b
AB
2924 case SIOCGIFADDR:
2925 case SIOCSIFADDR:
2926 case SIOCSIFHWBROADCAST:
6b96018b 2927 case SIOCDIFADDR:
6b96018b
AB
2928 case SIOCGIFBRDADDR:
2929 case SIOCSIFBRDADDR:
2930 case SIOCGIFDSTADDR:
2931 case SIOCSIFDSTADDR:
2932 case SIOCGIFNETMASK:
2933 case SIOCSIFNETMASK:
2934 case SIOCSIFPFLAGS:
2935 case SIOCGIFPFLAGS:
2936 case SIOCGIFTXQLEN:
2937 case SIOCSIFTXQLEN:
2938 case SIOCBRADDIF:
2939 case SIOCBRDELIF:
9177efd3
AB
2940 case SIOCSIFNAME:
2941 case SIOCGMIIPHY:
2942 case SIOCGMIIREG:
2943 case SIOCSMIIREG:
6b96018b 2944 return dev_ifsioc(net, sock, cmd, argp);
9177efd3 2945
6b96018b
AB
2946 case SIOCSARP:
2947 case SIOCGARP:
2948 case SIOCDARP:
6b96018b 2949 case SIOCATMARK:
9177efd3
AB
2950 return sock_do_ioctl(net, sock, cmd, arg);
2951 }
2952
2953 /* Prevent warning from compat_sys_ioctl, these always
2954 * result in -EINVAL in the native case anyway. */
2955 switch (cmd) {
2956 case SIOCRTMSG:
2957 case SIOCGIFCOUNT:
6b96018b
AB
2958 case SIOCSRARP:
2959 case SIOCGRARP:
2960 case SIOCDRARP:
9177efd3
AB
2961 case SIOCSIFLINK:
2962 case SIOCGIFSLAVE:
2963 case SIOCSIFSLAVE:
2964 return -EINVAL;
6b96018b
AB
2965 }
2966
2967 return -ENOIOCTLCMD;
2968}
7a229387 2969
89bbfc95 2970static long compat_sock_ioctl(struct file *file, unsigned cmd,
89bddce5 2971 unsigned long arg)
89bbfc95
SP
2972{
2973 struct socket *sock = file->private_data;
2974 int ret = -ENOIOCTLCMD;
87de87d5
DM
2975 struct sock *sk;
2976 struct net *net;
2977
2978 sk = sock->sk;
2979 net = sock_net(sk);
89bbfc95
SP
2980
2981 if (sock->ops->compat_ioctl)
2982 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2983
87de87d5
DM
2984 if (ret == -ENOIOCTLCMD &&
2985 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2986 ret = compat_wext_handle_ioctl(net, cmd, arg);
2987
6b96018b
AB
2988 if (ret == -ENOIOCTLCMD)
2989 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2990
89bbfc95
SP
2991 return ret;
2992}
2993#endif
2994
ac5a488e
SS
2995int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2996{
2997 return sock->ops->bind(sock, addr, addrlen);
2998}
c6d409cf 2999EXPORT_SYMBOL(kernel_bind);
ac5a488e
SS
3000
3001int kernel_listen(struct socket *sock, int backlog)
3002{
3003 return sock->ops->listen(sock, backlog);
3004}
c6d409cf 3005EXPORT_SYMBOL(kernel_listen);
ac5a488e
SS
3006
3007int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3008{
3009 struct sock *sk = sock->sk;
3010 int err;
3011
3012 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3013 newsock);
3014 if (err < 0)
3015 goto done;
3016
3017 err = sock->ops->accept(sock, *newsock, flags);
3018 if (err < 0) {
3019 sock_release(*newsock);
fa8705b0 3020 *newsock = NULL;
ac5a488e
SS
3021 goto done;
3022 }
3023
3024 (*newsock)->ops = sock->ops;
1b08534e 3025 __module_get((*newsock)->ops->owner);
ac5a488e
SS
3026
3027done:
3028 return err;
3029}
c6d409cf 3030EXPORT_SYMBOL(kernel_accept);
ac5a488e
SS
3031
3032int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
4768fbcb 3033 int flags)
ac5a488e
SS
3034{
3035 return sock->ops->connect(sock, addr, addrlen, flags);
3036}
c6d409cf 3037EXPORT_SYMBOL(kernel_connect);
ac5a488e
SS
3038
3039int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3040 int *addrlen)
3041{
3042 return sock->ops->getname(sock, addr, addrlen, 0);
3043}
c6d409cf 3044EXPORT_SYMBOL(kernel_getsockname);
ac5a488e
SS
3045
3046int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3047 int *addrlen)
3048{
3049 return sock->ops->getname(sock, addr, addrlen, 1);
3050}
c6d409cf 3051EXPORT_SYMBOL(kernel_getpeername);
ac5a488e
SS
3052
3053int kernel_getsockopt(struct socket *sock, int level, int optname,
3054 char *optval, int *optlen)
3055{
3056 mm_segment_t oldfs = get_fs();
3057 int err;
3058
3059 set_fs(KERNEL_DS);
3060 if (level == SOL_SOCKET)
3061 err = sock_getsockopt(sock, level, optname, optval, optlen);
3062 else
3063 err = sock->ops->getsockopt(sock, level, optname, optval,
3064 optlen);
3065 set_fs(oldfs);
3066 return err;
3067}
c6d409cf 3068EXPORT_SYMBOL(kernel_getsockopt);
ac5a488e
SS
3069
3070int kernel_setsockopt(struct socket *sock, int level, int optname,
b7058842 3071 char *optval, unsigned int optlen)
ac5a488e
SS
3072{
3073 mm_segment_t oldfs = get_fs();
3074 int err;
3075
3076 set_fs(KERNEL_DS);
3077 if (level == SOL_SOCKET)
3078 err = sock_setsockopt(sock, level, optname, optval, optlen);
3079 else
3080 err = sock->ops->setsockopt(sock, level, optname, optval,
3081 optlen);
3082 set_fs(oldfs);
3083 return err;
3084}
c6d409cf 3085EXPORT_SYMBOL(kernel_setsockopt);
ac5a488e
SS
3086
3087int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3088 size_t size, int flags)
3089{
f8451725
HX
3090 sock_update_classid(sock->sk);
3091
ac5a488e
SS
3092 if (sock->ops->sendpage)
3093 return sock->ops->sendpage(sock, page, offset, size, flags);
3094
3095 return sock_no_sendpage(sock, page, offset, size, flags);
3096}
c6d409cf 3097EXPORT_SYMBOL(kernel_sendpage);
ac5a488e
SS
3098
3099int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3100{
3101 mm_segment_t oldfs = get_fs();
3102 int err;
3103
3104 set_fs(KERNEL_DS);
3105 err = sock->ops->ioctl(sock, cmd, arg);
3106 set_fs(oldfs);
3107
3108 return err;
3109}
c6d409cf 3110EXPORT_SYMBOL(kernel_sock_ioctl);
ac5a488e 3111
91cf45f0
TM
3112int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3113{
3114 return sock->ops->shutdown(sock, how);
3115}
91cf45f0 3116EXPORT_SYMBOL(kernel_sock_shutdown);