blob: b94c3dd71015f8443b122706fcd08388ebe57247 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
Jesper Juhl02c30a82005-05-05 16:16:16 -07007 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
Stephen Hemminger89bddce2006-09-01 00:19:31 -070045 * Tigran Aivazian : Made listen(2) backlog sanity checks
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 * 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
Stephen Hemminger89bddce2006-09-01 00:19:31 -070056 * paradigm.
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070066#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080071#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030074#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070088#include <linux/nsproxy.h>
Nick Black1fd73172009-09-22 16:43:33 -070089#include <linux/magic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
91#include <asm/uaccess.h>
92#include <asm/unistd.h>
93
94#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070095#include <net/wext.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97#include <net/sock.h>
98#include <linux/netfilter.h>
99
Arnd Bergmann6b960182009-11-06 23:10:54 -0800100#include <linux/if_tun.h>
101#include <linux/ipv6_route.h>
102#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800103#include <linux/sockios.h>
104#include <linux/atalk.h>
105
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700107static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
108 unsigned long nr_segs, loff_t pos);
109static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700111static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
113static int sock_close(struct inode *inode, struct file *file);
114static unsigned int sock_poll(struct file *file,
115 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700116static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800117#ifdef CONFIG_COMPAT
118static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700119 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800120#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122static ssize_t sock_sendpage(struct file *file, struct page *page,
123 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800124static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
125 struct pipe_inode_info *pipe, size_t len,
126 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128/*
129 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
130 * in the operation structures but are done directly via the socketcall() multiplexor.
131 */
132
Arjan van de Venda7071d2007-02-12 00:55:36 -0800133static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134 .owner = THIS_MODULE,
135 .llseek = no_llseek,
136 .aio_read = sock_aio_read,
137 .aio_write = sock_aio_write,
138 .poll = sock_poll,
139 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800140#ifdef CONFIG_COMPAT
141 .compat_ioctl = compat_sock_ioctl,
142#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 .mmap = sock_mmap,
144 .open = sock_no_open, /* special open code to disallow open via /proc */
145 .release = sock_close,
146 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200147 .sendpage = sock_sendpage,
148 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800149 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150};
151
152/*
153 * The protocol list. Each protocol is registered in here.
154 */
155
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700157static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159/*
160 * Statistics counters of the socket lists
161 */
162
163static DEFINE_PER_CPU(int, sockets_in_use) = 0;
164
165/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700166 * Support routines.
167 * Move socket addresses back and forth across the kernel/user
168 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169 */
170
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700171#define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 16 for IP, 16 for IPX,
173 24 for IPv6,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700174 about 80 for AX.25
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 must be at least one bigger than
176 the AF_UNIX size (see net/unix/af_unix.c
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700177 :unix_mkname()).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700179
Linus Torvalds1da177e2005-04-16 15:20:36 -0700180/**
181 * move_addr_to_kernel - copy a socket address into kernel space
182 * @uaddr: Address in user space
183 * @kaddr: Address in kernel space
184 * @ulen: Length in user space
185 *
186 * The address is copied into kernel space. If the provided address is
187 * too long an error code of -EINVAL is returned. If the copy gives
188 * invalid addresses -EFAULT is returned. On a success 0 is returned.
189 */
190
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700191int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700193 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700195 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700196 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700197 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100199 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200}
201
202/**
203 * move_addr_to_user - copy an address to user space
204 * @kaddr: kernel space address
205 * @klen: length of address in kernel
206 * @uaddr: user space address
207 * @ulen: pointer to user length field
208 *
209 * The value pointed to by ulen on entry is the buffer length available.
210 * This is overwritten with the buffer space used. -EINVAL is returned
211 * if an overlong buffer is specified or a negative buffer size. -EFAULT
212 * is returned if either the buffer or the length field are not
213 * accessible.
214 * After copying the data up to the limit the user specifies, the true
215 * length of the data is written over the length limit the user
216 * specified. Zero is returned for a success.
217 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700219int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700220 int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221{
222 int err;
223 int len;
224
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 err = get_user(len, ulen);
226 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (len > klen)
229 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700230 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500233 if (audit_sockaddr(klen, kaddr))
234 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700235 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 return -EFAULT;
237 }
238 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700239 * "fromlen shall refer to the value before truncation.."
240 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 */
242 return __put_user(klen, ulen);
243}
244
Christoph Lametere18b8902006-12-06 20:33:20 -0800245static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
247static struct inode *sock_alloc_inode(struct super_block *sb)
248{
249 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700250
Christoph Lametere94b1762006-12-06 20:33:17 -0800251 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 if (!ei)
253 return NULL;
254 init_waitqueue_head(&ei->socket.wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700255
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 ei->socket.fasync_list = NULL;
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
263 return &ei->vfs_inode;
264}
265
266static void sock_destroy_inode(struct inode *inode)
267{
268 kmem_cache_free(sock_inode_cachep,
269 container_of(inode, struct socket_alloc, vfs_inode));
270}
271
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700272static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700274 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275
Christoph Lametera35afb82007-05-16 22:10:57 -0700276 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700278
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279static int init_inodecache(void)
280{
281 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700282 sizeof(struct socket_alloc),
283 0,
284 (SLAB_HWCACHE_ALIGN |
285 SLAB_RECLAIM_ACCOUNT |
286 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900287 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 if (sock_inode_cachep == NULL)
289 return -ENOMEM;
290 return 0;
291}
292
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700293static const struct super_operations sockfs_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 .alloc_inode = sock_alloc_inode,
295 .destroy_inode =sock_destroy_inode,
296 .statfs = simple_statfs,
297};
298
David Howells454e2392006-06-23 02:02:57 -0700299static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700300 int flags, const char *dev_name, void *data,
301 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302{
David Howells454e2392006-06-23 02:02:57 -0700303 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
304 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305}
306
Eric Dumazetba899662005-08-26 12:05:31 -0700307static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308
309static struct file_system_type sock_fs_type = {
310 .name = "sockfs",
311 .get_sb = sockfs_get_sb,
312 .kill_sb = kill_anon_super,
313};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700314
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315static int sockfs_delete_dentry(struct dentry *dentry)
316{
Eric Dumazet304e61e2006-12-06 20:38:49 -0800317 /*
318 * At creation time, we pretended this dentry was hashed
319 * (by clearing DCACHE_UNHASHED bit in d_flags)
320 * At delete time, we restore the truth : not hashed.
321 * (so that dput() can proceed correctly)
322 */
323 dentry->d_flags |= DCACHE_UNHASHED;
324 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325}
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700326
327/*
328 * sockfs_dname() is called from d_path().
329 */
330static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
331{
332 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
333 dentry->d_inode->i_ino);
334}
335
Al Viro3ba13d12009-02-20 06:02:22 +0000336static const struct dentry_operations sockfs_dentry_operations = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700337 .d_delete = sockfs_delete_dentry,
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700338 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339};
340
341/*
342 * Obtains the first available file descriptor and sets it up for use.
343 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
351 *
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
356 */
357
Ulrich Dreppera677a032008-07-23 21:29:17 -0700358static int sock_alloc_fd(struct file **filep, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
360 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800361
Ulrich Dreppera677a032008-07-23 21:29:17 -0700362 fd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800363 if (likely(fd >= 0)) {
364 struct file *file = get_empty_filp();
365
366 *filep = file;
367 if (unlikely(!file)) {
368 put_unused_fd(fd);
369 return -ENFILE;
370 }
371 } else
372 *filep = NULL;
373 return fd;
374}
375
Ulrich Drepper77d27202008-07-23 21:29:35 -0700376static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800377{
Dave Hansence8d2cd2007-10-16 23:31:13 -0700378 struct dentry *dentry;
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700379 struct qstr name = { .name = "" };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380
Dave Hansence8d2cd2007-10-16 23:31:13 -0700381 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
382 if (unlikely(!dentry))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800383 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384
Dave Hansence8d2cd2007-10-16 23:31:13 -0700385 dentry->d_op = &sockfs_dentry_operations;
Eric Dumazet304e61e2006-12-06 20:38:49 -0800386 /*
387 * We dont want to push this dentry into global dentry hash table.
388 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
389 * This permits a working /proc/$pid/fd/XXX on sockets
390 */
Dave Hansence8d2cd2007-10-16 23:31:13 -0700391 dentry->d_flags &= ~DCACHE_UNHASHED;
392 d_instantiate(dentry, SOCK_INODE(sock));
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393
394 sock->file = file;
Dave Hansence8d2cd2007-10-16 23:31:13 -0700395 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
396 &socket_file_ops);
397 SOCK_INODE(sock)->i_fop = &socket_file_ops;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700398 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800399 file->f_pos = 0;
400 file->private_data = sock;
401
402 return 0;
403}
404
Ulrich Dreppera677a032008-07-23 21:29:17 -0700405int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800406{
407 struct file *newfile;
Ulrich Dreppera677a032008-07-23 21:29:17 -0700408 int fd = sock_alloc_fd(&newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800409
410 if (likely(fd >= 0)) {
Ulrich Drepper77d27202008-07-23 21:29:35 -0700411 int err = sock_attach_fd(sock, newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800412
413 if (unlikely(err < 0)) {
414 put_filp(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800416 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800418 fd_install(fd, newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 return fd;
421}
422
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800423static struct socket *sock_from_file(struct file *file, int *err)
424{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800425 if (file->f_op == &socket_file_ops)
426 return file->private_data; /* set in sock_map_fd */
427
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800428 *err = -ENOTSOCK;
429 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800430}
431
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432/**
433 * sockfd_lookup - Go from a file number to its socket slot
434 * @fd: file handle
435 * @err: pointer to an error code return
436 *
437 * The file handle passed in is locked and the socket it is bound
438 * too is returned. If an error occurs the err pointer is overwritten
439 * with a negative errno code and NULL is returned. The function checks
440 * for both invalid handles and passing a handle which is not a socket.
441 *
442 * On a success the socket object pointer is returned.
443 */
444
445struct socket *sockfd_lookup(int fd, int *err)
446{
447 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 struct socket *sock;
449
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700450 file = fget(fd);
451 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 *err = -EBADF;
453 return NULL;
454 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700455
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800456 sock = sock_from_file(file, err);
457 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 return sock;
460}
461
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800462static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
463{
464 struct file *file;
465 struct socket *sock;
466
Hua Zhong36725582006-04-19 15:25:02 -0700467 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800468 file = fget_light(fd, fput_needed);
469 if (file) {
470 sock = sock_from_file(file, err);
471 if (sock)
472 return sock;
473 fput_light(file, *fput_needed);
474 }
475 return NULL;
476}
477
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478/**
479 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700480 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 * Allocate a new inode and socket object. The two are bound together
482 * and initialised. The socket is then returned. If we are out of inodes
483 * NULL is returned.
484 */
485
486static struct socket *sock_alloc(void)
487{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700488 struct inode *inode;
489 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
491 inode = new_inode(sock_mnt->mnt_sb);
492 if (!inode)
493 return NULL;
494
495 sock = SOCKET_I(inode);
496
Eric Dumazet29a020d2009-09-15 02:39:20 -0700497 kmemcheck_annotate_bitfield(sock, type);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700498 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100499 inode->i_uid = current_fsuid();
500 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
Eric Dumazet4e694892009-04-04 16:41:09 -0700502 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 return sock;
504}
505
506/*
507 * In theory you can't get an open on this inode, but /proc provides
508 * a back door. Remember to keep it shut otherwise you'll let the
509 * creepy crawlies in.
510 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
513{
514 return -ENXIO;
515}
516
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800517const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 .owner = THIS_MODULE,
519 .open = sock_no_open,
520};
521
522/**
523 * sock_release - close a socket
524 * @sock: socket to close
525 *
526 * The socket is released from the protocol stack if it has a release
527 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700528 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700530
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531void sock_release(struct socket *sock)
532{
533 if (sock->ops) {
534 struct module *owner = sock->ops->owner;
535
536 sock->ops->release(sock);
537 sock->ops = NULL;
538 module_put(owner);
539 }
540
541 if (sock->fasync_list)
542 printk(KERN_ERR "sock_release: fasync list not empty!\n");
543
Eric Dumazet4e694892009-04-04 16:41:09 -0700544 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 if (!sock->file) {
546 iput(SOCK_INODE(sock));
547 return;
548 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700549 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550}
551
Patrick Ohly20d49472009-02-12 05:03:38 +0000552int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
553 union skb_shared_tx *shtx)
554{
555 shtx->flags = 0;
556 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
557 shtx->hardware = 1;
558 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
559 shtx->software = 1;
560 return 0;
561}
562EXPORT_SYMBOL(sock_tx_timestamp);
563
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700564static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 struct msghdr *msg, size_t size)
566{
567 struct sock_iocb *si = kiocb_to_siocb(iocb);
568 int err;
569
570 si->sock = sock;
571 si->scm = NULL;
572 si->msg = msg;
573 si->size = size;
574
575 err = security_socket_sendmsg(sock, msg, size);
576 if (err)
577 return err;
578
579 return sock->ops->sendmsg(iocb, sock, msg, size);
580}
581
582int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
583{
584 struct kiocb iocb;
585 struct sock_iocb siocb;
586 int ret;
587
588 init_sync_kiocb(&iocb, NULL);
589 iocb.private = &siocb;
590 ret = __sock_sendmsg(&iocb, sock, msg, size);
591 if (-EIOCBQUEUED == ret)
592 ret = wait_on_sync_kiocb(&iocb);
593 return ret;
594}
595
596int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
597 struct kvec *vec, size_t num, size_t size)
598{
599 mm_segment_t oldfs = get_fs();
600 int result;
601
602 set_fs(KERNEL_DS);
603 /*
604 * the following is safe, since for compiler definitions of kvec and
605 * iovec are identical, yielding the same in-core layout and alignment
606 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700607 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608 msg->msg_iovlen = num;
609 result = sock_sendmsg(sock, msg, size);
610 set_fs(oldfs);
611 return result;
612}
613
Patrick Ohly20d49472009-02-12 05:03:38 +0000614static int ktime2ts(ktime_t kt, struct timespec *ts)
615{
616 if (kt.tv64) {
617 *ts = ktime_to_timespec(kt);
618 return 1;
619 } else {
620 return 0;
621 }
622}
623
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700624/*
625 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
626 */
627void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
628 struct sk_buff *skb)
629{
Patrick Ohly20d49472009-02-12 05:03:38 +0000630 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
631 struct timespec ts[3];
632 int empty = 1;
633 struct skb_shared_hwtstamps *shhwtstamps =
634 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700635
Patrick Ohly20d49472009-02-12 05:03:38 +0000636 /* Race occurred between timestamp enabling and packet
637 receiving. Fill in the current time for now. */
638 if (need_software_tstamp && skb->tstamp.tv64 == 0)
639 __net_timestamp(skb);
640
641 if (need_software_tstamp) {
642 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
643 struct timeval tv;
644 skb_get_timestamp(skb, &tv);
645 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
646 sizeof(tv), &tv);
647 } else {
648 struct timespec ts;
649 skb_get_timestampns(skb, &ts);
650 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
651 sizeof(ts), &ts);
652 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700653 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000654
655
656 memset(ts, 0, sizeof(ts));
657 if (skb->tstamp.tv64 &&
658 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
659 skb_get_timestampns(skb, ts + 0);
660 empty = 0;
661 }
662 if (shhwtstamps) {
663 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
664 ktime2ts(shhwtstamps->syststamp, ts + 1))
665 empty = 0;
666 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
667 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
668 empty = 0;
669 }
670 if (!empty)
671 put_cmsg(msg, SOL_SOCKET,
672 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700673}
674
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300675EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
676
Neil Horman3b885782009-10-12 13:26:31 -0700677inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
678{
679 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
680 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
681 sizeof(__u32), &skb->dropcount);
682}
683
684void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
685 struct sk_buff *skb)
686{
687 sock_recv_timestamp(msg, sk, skb);
688 sock_recv_drops(msg, sk, skb);
689}
690EXPORT_SYMBOL_GPL(sock_recv_ts_and_drops);
691
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700692static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
693 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695 struct sock_iocb *si = kiocb_to_siocb(iocb);
696
697 si->sock = sock;
698 si->scm = NULL;
699 si->msg = msg;
700 si->size = size;
701 si->flags = flags;
702
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
704}
705
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700706static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
707 struct msghdr *msg, size_t size, int flags)
708{
709 int err = security_socket_recvmsg(sock, msg, size, flags);
710
711 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
712}
713
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700714int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 size_t size, int flags)
716{
717 struct kiocb iocb;
718 struct sock_iocb siocb;
719 int ret;
720
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700721 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722 iocb.private = &siocb;
723 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
724 if (-EIOCBQUEUED == ret)
725 ret = wait_on_sync_kiocb(&iocb);
726 return ret;
727}
728
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700729static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
730 size_t size, int flags)
731{
732 struct kiocb iocb;
733 struct sock_iocb siocb;
734 int ret;
735
736 init_sync_kiocb(&iocb, NULL);
737 iocb.private = &siocb;
738 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
739 if (-EIOCBQUEUED == ret)
740 ret = wait_on_sync_kiocb(&iocb);
741 return ret;
742}
743
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700744int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
745 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746{
747 mm_segment_t oldfs = get_fs();
748 int result;
749
750 set_fs(KERNEL_DS);
751 /*
752 * the following is safe, since for compiler definitions of kvec and
753 * iovec are identical, yielding the same in-core layout and alignment
754 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700755 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 result = sock_recvmsg(sock, msg, size, flags);
757 set_fs(oldfs);
758 return result;
759}
760
761static void sock_aio_dtor(struct kiocb *iocb)
762{
763 kfree(iocb->private);
764}
765
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300766static ssize_t sock_sendpage(struct file *file, struct page *page,
767 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768{
769 struct socket *sock;
770 int flags;
771
Eric Dumazetb69aee02005-09-06 14:42:45 -0700772 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
774 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
775 if (more)
776 flags |= MSG_MORE;
777
Linus Torvaldse6949582009-08-13 08:28:36 -0700778 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779}
780
Jens Axboe9c55e012007-11-06 23:30:13 -0800781static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
782 struct pipe_inode_info *pipe, size_t len,
783 unsigned int flags)
784{
785 struct socket *sock = file->private_data;
786
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800787 if (unlikely(!sock->ops->splice_read))
788 return -EINVAL;
789
Jens Axboe9c55e012007-11-06 23:30:13 -0800790 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
791}
792
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800793static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700794 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800795{
796 if (!is_sync_kiocb(iocb)) {
797 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
798 if (!siocb)
799 return NULL;
800 iocb->ki_dtor = sock_aio_dtor;
801 }
802
803 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800804 iocb->private = siocb;
805 return siocb;
806}
807
808static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700809 struct file *file, const struct iovec *iov,
810 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800811{
812 struct socket *sock = file->private_data;
813 size_t size = 0;
814 int i;
815
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700816 for (i = 0; i < nr_segs; i++)
817 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800818
819 msg->msg_name = NULL;
820 msg->msg_namelen = 0;
821 msg->msg_control = NULL;
822 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700823 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800824 msg->msg_iovlen = nr_segs;
825 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
826
827 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
828}
829
Badari Pulavarty027445c2006-09-30 23:28:46 -0700830static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
831 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800832{
833 struct sock_iocb siocb, *x;
834
835 if (pos != 0)
836 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700837
838 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800839 return 0;
840
Badari Pulavarty027445c2006-09-30 23:28:46 -0700841
842 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800843 if (!x)
844 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700845 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800846}
847
848static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700849 struct file *file, const struct iovec *iov,
850 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800851{
852 struct socket *sock = file->private_data;
853 size_t size = 0;
854 int i;
855
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700856 for (i = 0; i < nr_segs; i++)
857 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800858
859 msg->msg_name = NULL;
860 msg->msg_namelen = 0;
861 msg->msg_control = NULL;
862 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700863 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800864 msg->msg_iovlen = nr_segs;
865 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
866 if (sock->type == SOCK_SEQPACKET)
867 msg->msg_flags |= MSG_EOR;
868
869 return __sock_sendmsg(iocb, sock, msg, size);
870}
871
Badari Pulavarty027445c2006-09-30 23:28:46 -0700872static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
873 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800875 struct sock_iocb siocb, *x;
876
877 if (pos != 0)
878 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700879
Badari Pulavarty027445c2006-09-30 23:28:46 -0700880 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800881 if (!x)
882 return -ENOMEM;
883
Badari Pulavarty027445c2006-09-30 23:28:46 -0700884 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885}
886
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887/*
888 * Atomic setting of ioctl hooks to avoid race
889 * with module unload.
890 */
891
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800892static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700893static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Eric W. Biederman881d9662007-09-17 11:56:21 -0700895void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800897 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800899 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700901
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902EXPORT_SYMBOL(brioctl_set);
903
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800904static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700905static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
Eric W. Biederman881d9662007-09-17 11:56:21 -0700907void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800909 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800911 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700913
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914EXPORT_SYMBOL(vlan_ioctl_set);
915
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800916static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700917static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700919void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800921 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800923 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700925
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926EXPORT_SYMBOL(dlci_ioctl_set);
927
Arnd Bergmann6b960182009-11-06 23:10:54 -0800928static long sock_do_ioctl(struct net *net, struct socket *sock,
929 unsigned int cmd, unsigned long arg)
930{
931 int err;
932 void __user *argp = (void __user *)arg;
933
934 err = sock->ops->ioctl(sock, cmd, arg);
935
936 /*
937 * If this ioctl is unknown try to hand it down
938 * to the NIC driver.
939 */
940 if (err == -ENOIOCTLCMD)
941 err = dev_ioctl(net, cmd, argp);
942
943 return err;
944}
945
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946/*
947 * With an ioctl, arg may well be a user mode pointer, but we don't know
948 * what to do with it - that's up to the protocol still.
949 */
950
951static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
952{
953 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700954 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 void __user *argp = (void __user *)arg;
956 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700957 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958
Eric Dumazetb69aee02005-09-06 14:42:45 -0700959 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700960 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900961 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700963 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200965#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700967 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200969#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700970 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 case FIOSETOWN:
972 case SIOCSPGRP:
973 err = -EFAULT;
974 if (get_user(pid, (int __user *)argp))
975 break;
976 err = f_setown(sock->file, pid, 1);
977 break;
978 case FIOGETOWN:
979 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700980 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700981 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 break;
983 case SIOCGIFBR:
984 case SIOCSIFBR:
985 case SIOCBRADDBR:
986 case SIOCBRDELBR:
987 err = -ENOPKG;
988 if (!br_ioctl_hook)
989 request_module("bridge");
990
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800991 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700992 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700993 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800994 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 break;
996 case SIOCGIFVLAN:
997 case SIOCSIFVLAN:
998 err = -ENOPKG;
999 if (!vlan_ioctl_hook)
1000 request_module("8021q");
1001
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001002 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001004 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001005 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 case SIOCADDDLCI:
1008 case SIOCDELDLCI:
1009 err = -ENOPKG;
1010 if (!dlci_ioctl_hook)
1011 request_module("dlci");
1012
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001013 mutex_lock(&dlci_ioctl_mutex);
1014 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001016 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 break;
1018 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001019 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001021 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 return err;
1023}
1024
1025int sock_create_lite(int family, int type, int protocol, struct socket **res)
1026{
1027 int err;
1028 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001029
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 err = security_socket_create(family, type, protocol, 1);
1031 if (err)
1032 goto out;
1033
1034 sock = sock_alloc();
1035 if (!sock) {
1036 err = -ENOMEM;
1037 goto out;
1038 }
1039
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001041 err = security_socket_post_create(sock, family, type, protocol, 1);
1042 if (err)
1043 goto out_release;
1044
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045out:
1046 *res = sock;
1047 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001048out_release:
1049 sock_release(sock);
1050 sock = NULL;
1051 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052}
1053
1054/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001055static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056{
1057 struct socket *sock;
1058
1059 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001060 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001062 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063 return sock->ops->poll(file, sock, wait);
1064}
1065
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001066static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001068 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069
1070 return sock->ops->mmap(file, sock, vma);
1071}
1072
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001073static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
1075 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001076 * It was possible the inode is NULL we were
1077 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 */
1079
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001080 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 printk(KERN_DEBUG "sock_close: NULL inode\n");
1082 return 0;
1083 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 sock_release(SOCKET_I(inode));
1085 return 0;
1086}
1087
1088/*
1089 * Update the socket async list
1090 *
1091 * Fasync_list locking strategy.
1092 *
1093 * 1. fasync_list is modified only under process context socket lock
1094 * i.e. under semaphore.
1095 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1096 * or under socket lock.
1097 * 3. fasync_list can be used from softirq context, so that
1098 * modification under socket lock have to be enhanced with
1099 * write_lock_bh(&sk->sk_callback_lock).
1100 * --ANK (990710)
1101 */
1102
1103static int sock_fasync(int fd, struct file *filp, int on)
1104{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001105 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 struct socket *sock;
1107 struct sock *sk;
1108
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001109 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001110 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001111 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 return -ENOMEM;
1113 }
1114
Eric Dumazetb69aee02005-09-06 14:42:45 -07001115 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001117 sk = sock->sk;
1118 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 kfree(fna);
1120 return -EINVAL;
1121 }
1122
1123 lock_sock(sk);
1124
Jonathan Corbet76398422009-02-01 14:26:59 -07001125 spin_lock(&filp->f_lock);
1126 if (on)
1127 filp->f_flags |= FASYNC;
1128 else
1129 filp->f_flags &= ~FASYNC;
1130 spin_unlock(&filp->f_lock);
1131
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001132 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001134 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1135 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 break;
1137
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001138 if (on) {
1139 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001141 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 write_unlock_bh(&sk->sk_callback_lock);
1143
1144 kfree(fna);
1145 goto out;
1146 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001147 fna->fa_file = filp;
1148 fna->fa_fd = fd;
1149 fna->magic = FASYNC_MAGIC;
1150 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001152 sock->fasync_list = fna;
Eric Dumazetbcdce712009-10-06 17:28:29 -07001153 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001155 } else {
1156 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001158 *prev = fa->fa_next;
Eric Dumazetbcdce712009-10-06 17:28:29 -07001159 if (!sock->fasync_list)
1160 sock_reset_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 write_unlock_bh(&sk->sk_callback_lock);
1162 kfree(fa);
1163 }
1164 }
1165
1166out:
1167 release_sock(sock->sk);
1168 return 0;
1169}
1170
1171/* This function may be called only under socket lock or callback_lock */
1172
1173int sock_wake_async(struct socket *sock, int how, int band)
1174{
1175 if (!sock || !sock->fasync_list)
1176 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001177 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001178 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1180 break;
1181 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001182 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1184 break;
1185 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001186 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001187call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 __kill_fasync(sock->fasync_list, SIGIO, band);
1189 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001190 case SOCK_WAKE_URG:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 __kill_fasync(sock->fasync_list, SIGURG, band);
1192 }
1193 return 0;
1194}
1195
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001196static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001197 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198{
1199 int err;
1200 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001201 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202
1203 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001204 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 */
1206 if (family < 0 || family >= NPROTO)
1207 return -EAFNOSUPPORT;
1208 if (type < 0 || type >= SOCK_MAX)
1209 return -EINVAL;
1210
1211 /* Compatibility.
1212
1213 This uglymoron is moved from INET layer to here to avoid
1214 deadlock in module load.
1215 */
1216 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001217 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 if (!warned) {
1219 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001220 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1221 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 }
1223 family = PF_PACKET;
1224 }
1225
1226 err = security_socket_create(family, type, protocol, kern);
1227 if (err)
1228 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001229
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001230 /*
1231 * Allocate the socket and allow the family to set things up. if
1232 * the protocol is 0, the family is instructed to select an appropriate
1233 * default.
1234 */
1235 sock = sock_alloc();
1236 if (!sock) {
1237 if (net_ratelimit())
1238 printk(KERN_WARNING "socket: no more sockets\n");
1239 return -ENFILE; /* Not exactly a match, but its the
1240 closest posix thing */
1241 }
1242
1243 sock->type = type;
1244
Johannes Berg95a5afc2008-10-16 15:24:51 -07001245#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001246 /* Attempt to load a protocol module if the find failed.
1247 *
1248 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 * requested real, full-featured networking support upon configuration.
1250 * Otherwise module support will break!
1251 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001252 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001253 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254#endif
1255
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001256 rcu_read_lock();
1257 pf = rcu_dereference(net_families[family]);
1258 err = -EAFNOSUPPORT;
1259 if (!pf)
1260 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261
1262 /*
1263 * We will call the ->create function, that possibly is in a loadable
1264 * module, so we have to bump that loadable module refcnt first.
1265 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001266 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 goto out_release;
1268
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001269 /* Now protected by module ref count */
1270 rcu_read_unlock();
1271
Eric Paris3f378b62009-11-05 22:18:14 -08001272 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001273 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001275
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 /*
1277 * Now to bump the refcnt of the [loadable] module that owns this
1278 * socket at sock_release time we decrement its refcnt.
1279 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001280 if (!try_module_get(sock->ops->owner))
1281 goto out_module_busy;
1282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 /*
1284 * Now that we're done with the ->create function, the [loadable]
1285 * module can have its refcnt decremented
1286 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001287 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001288 err = security_socket_post_create(sock, family, type, protocol, kern);
1289 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001290 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001291 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001293 return 0;
1294
1295out_module_busy:
1296 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001298 sock->ops = NULL;
1299 module_put(pf->owner);
1300out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001302 return err;
1303
1304out_release:
1305 rcu_read_unlock();
1306 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307}
1308
1309int sock_create(int family, int type, int protocol, struct socket **res)
1310{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001311 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312}
1313
1314int sock_create_kern(int family, int type, int protocol, struct socket **res)
1315{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001316 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317}
1318
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001319SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320{
1321 int retval;
1322 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001323 int flags;
1324
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001325 /* Check the SOCK_* constants for consistency. */
1326 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1327 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1328 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1329 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1330
Ulrich Dreppera677a032008-07-23 21:29:17 -07001331 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001332 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001333 return -EINVAL;
1334 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001336 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1337 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1338
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 retval = sock_create(family, type, protocol, &sock);
1340 if (retval < 0)
1341 goto out;
1342
Ulrich Drepper77d27202008-07-23 21:29:35 -07001343 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 if (retval < 0)
1345 goto out_release;
1346
1347out:
1348 /* It may be already another descriptor 8) Not kernel problem. */
1349 return retval;
1350
1351out_release:
1352 sock_release(sock);
1353 return retval;
1354}
1355
1356/*
1357 * Create a pair of connected sockets.
1358 */
1359
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001360SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1361 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362{
1363 struct socket *sock1, *sock2;
1364 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001365 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001366 int flags;
1367
1368 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001369 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001370 return -EINVAL;
1371 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001373 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1374 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1375
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 /*
1377 * Obtain the first socket and check if the underlying protocol
1378 * supports the socketpair call.
1379 */
1380
1381 err = sock_create(family, type, protocol, &sock1);
1382 if (err < 0)
1383 goto out;
1384
1385 err = sock_create(family, type, protocol, &sock2);
1386 if (err < 0)
1387 goto out_release_1;
1388
1389 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001390 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 goto out_release_both;
1392
Ulrich Dreppera677a032008-07-23 21:29:17 -07001393 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001394 if (unlikely(fd1 < 0)) {
1395 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001397 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
Ulrich Dreppera677a032008-07-23 21:29:17 -07001399 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
Al Virodb349502007-02-07 01:48:00 -05001400 if (unlikely(fd2 < 0)) {
David S. Millerbf3c23d2007-10-29 21:54:02 -07001401 err = fd2;
Al Virodb349502007-02-07 01:48:00 -05001402 put_filp(newfile1);
1403 put_unused_fd(fd1);
1404 goto out_release_both;
1405 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406
Ulrich Drepper77d27202008-07-23 21:29:35 -07001407 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
Al Virodb349502007-02-07 01:48:00 -05001408 if (unlikely(err < 0)) {
1409 goto out_fd2;
1410 }
1411
Ulrich Drepper77d27202008-07-23 21:29:35 -07001412 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
Al Virodb349502007-02-07 01:48:00 -05001413 if (unlikely(err < 0)) {
1414 fput(newfile1);
1415 goto out_fd1;
1416 }
1417
Al Viro157cf642008-12-14 04:57:47 -05001418 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001419 fd_install(fd1, newfile1);
1420 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 /* fd1 and fd2 may be already another descriptors.
1422 * Not kernel problem.
1423 */
1424
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001425 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 if (!err)
1427 err = put_user(fd2, &usockvec[1]);
1428 if (!err)
1429 return 0;
1430
1431 sys_close(fd2);
1432 sys_close(fd1);
1433 return err;
1434
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001436 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001438 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439out:
1440 return err;
Al Virodb349502007-02-07 01:48:00 -05001441
1442out_fd2:
1443 put_filp(newfile1);
1444 sock_release(sock1);
1445out_fd1:
1446 put_filp(newfile2);
1447 sock_release(sock2);
Al Virodb349502007-02-07 01:48:00 -05001448 put_unused_fd(fd1);
1449 put_unused_fd(fd2);
1450 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451}
1452
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453/*
1454 * Bind a name to a socket. Nothing much to do here since it's
1455 * the protocol's responsibility to handle the local address.
1456 *
1457 * We move the socket address to kernel space before we call
1458 * the protocol layer (having also checked the address is ok).
1459 */
1460
Heiko Carstens20f37032009-01-14 14:14:23 +01001461SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462{
1463 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001464 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001465 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001467 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001468 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001469 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001470 if (err >= 0) {
1471 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001472 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001473 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001474 if (!err)
1475 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001476 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001477 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001479 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001480 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 return err;
1482}
1483
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484/*
1485 * Perform a listen. Basically, we allow the protocol to do anything
1486 * necessary for a listen, and if that works, we mark the socket as
1487 * ready for listening.
1488 */
1489
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001490SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491{
1492 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001493 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001494 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001495
1496 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1497 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001498 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001499 if ((unsigned)backlog > somaxconn)
1500 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501
1502 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001503 if (!err)
1504 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001506 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 }
1508 return err;
1509}
1510
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511/*
1512 * For accept, we attempt to create a new socket, set up the link
1513 * with the client, wake up the client, then return the new
1514 * connected fd. We collect the address of the connector in kernel
1515 * space and move it to user at the very end. This is unclean because
1516 * we open the socket then return an error.
1517 *
1518 * 1003.1g adds the ability to recvmsg() to query connection pending
1519 * status to recvmsg. We need to add that support in a way thats
1520 * clean when we restucture accept also.
1521 */
1522
Heiko Carstens20f37032009-01-14 14:14:23 +01001523SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1524 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525{
1526 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001527 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001528 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001529 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530
Ulrich Drepper77d27202008-07-23 21:29:35 -07001531 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001532 return -EINVAL;
1533
1534 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1535 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1536
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001537 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 if (!sock)
1539 goto out;
1540
1541 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001542 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 goto out_put;
1544
1545 newsock->type = sock->type;
1546 newsock->ops = sock->ops;
1547
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 /*
1549 * We don't need try_module_get here, as the listening socket (sock)
1550 * has the protocol module (sock->ops->owner) held.
1551 */
1552 __module_get(newsock->ops->owner);
1553
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001554 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001555 if (unlikely(newfd < 0)) {
1556 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001557 sock_release(newsock);
1558 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001559 }
1560
Ulrich Drepper77d27202008-07-23 21:29:35 -07001561 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001562 if (err < 0)
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001563 goto out_fd_simple;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001564
Frank Filza79af592005-09-27 15:23:38 -07001565 err = security_socket_accept(sock, newsock);
1566 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001567 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001568
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1570 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001571 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
1573 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001574 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001575 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001577 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001579 err = move_addr_to_user((struct sockaddr *)&address,
1580 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001582 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 }
1584
1585 /* File flags are not inherited via accept() unlike another OSes. */
1586
David S. Miller39d8c1b2006-03-20 17:13:49 -08001587 fd_install(newfd, newfile);
1588 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001591 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592out:
1593 return err;
Alexey Dobriyan79f4f642007-03-26 14:09:52 -07001594out_fd_simple:
1595 sock_release(newsock);
1596 put_filp(newfile);
1597 put_unused_fd(newfd);
1598 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001599out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001600 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001601 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 goto out_put;
1603}
1604
Heiko Carstens20f37032009-01-14 14:14:23 +01001605SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1606 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001607{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001608 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001609}
1610
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611/*
1612 * Attempt to connect to a socket with the server address. The address
1613 * is in user space so we verify it is OK and move it to kernel space.
1614 *
1615 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1616 * break bindings
1617 *
1618 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1619 * other SEQPACKET protocols that take time to connect() as it doesn't
1620 * include the -EINPROGRESS status for such sockets.
1621 */
1622
Heiko Carstens20f37032009-01-14 14:14:23 +01001623SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1624 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625{
1626 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001627 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001628 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001630 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 if (!sock)
1632 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001633 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634 if (err < 0)
1635 goto out_put;
1636
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001637 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001638 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 if (err)
1640 goto out_put;
1641
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001642 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 sock->file->f_flags);
1644out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001645 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646out:
1647 return err;
1648}
1649
1650/*
1651 * Get the local address ('name') of a socket object. Move the obtained
1652 * name to user space.
1653 */
1654
Heiko Carstens20f37032009-01-14 14:14:23 +01001655SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1656 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657{
1658 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001659 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001660 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001661
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001662 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 if (!sock)
1664 goto out;
1665
1666 err = security_socket_getsockname(sock);
1667 if (err)
1668 goto out_put;
1669
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001670 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 if (err)
1672 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001673 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674
1675out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001676 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677out:
1678 return err;
1679}
1680
1681/*
1682 * Get the remote address ('name') of a socket object. Move the obtained
1683 * name to user space.
1684 */
1685
Heiko Carstens20f37032009-01-14 14:14:23 +01001686SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1687 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688{
1689 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001690 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001691 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001693 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1694 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 err = security_socket_getpeername(sock);
1696 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001697 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 return err;
1699 }
1700
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001701 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001702 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001703 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001705 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001706 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001707 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 }
1709 return err;
1710}
1711
1712/*
1713 * Send a datagram to a given address. We move the address into kernel
1714 * space and check the user space data area is readable before invoking
1715 * the protocol.
1716 */
1717
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001718SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1719 unsigned, flags, struct sockaddr __user *, addr,
1720 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721{
1722 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001723 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 int err;
1725 struct msghdr msg;
1726 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001727 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001728
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001729 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1730 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001731 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001732
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001733 iov.iov_base = buff;
1734 iov.iov_len = len;
1735 msg.msg_name = NULL;
1736 msg.msg_iov = &iov;
1737 msg.msg_iovlen = 1;
1738 msg.msg_control = NULL;
1739 msg.msg_controllen = 0;
1740 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001741 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001742 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (err < 0)
1744 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001745 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001746 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 }
1748 if (sock->file->f_flags & O_NONBLOCK)
1749 flags |= MSG_DONTWAIT;
1750 msg.msg_flags = flags;
1751 err = sock_sendmsg(sock, &msg, len);
1752
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001753out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001754 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001755out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 return err;
1757}
1758
1759/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001760 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 */
1762
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001763SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1764 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765{
1766 return sys_sendto(fd, buff, len, flags, NULL, 0);
1767}
1768
1769/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001770 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 * sender. We verify the buffers are writable and if needed move the
1772 * sender address from kernel to user space.
1773 */
1774
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001775SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1776 unsigned, flags, struct sockaddr __user *, addr,
1777 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778{
1779 struct socket *sock;
1780 struct iovec iov;
1781 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001782 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001783 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001784 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001786 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001788 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001790 msg.msg_control = NULL;
1791 msg.msg_controllen = 0;
1792 msg.msg_iovlen = 1;
1793 msg.msg_iov = &iov;
1794 iov.iov_len = size;
1795 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001796 msg.msg_name = (struct sockaddr *)&address;
1797 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 if (sock->file->f_flags & O_NONBLOCK)
1799 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001800 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001802 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001803 err2 = move_addr_to_user((struct sockaddr *)&address,
1804 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001805 if (err2 < 0)
1806 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001808
1809 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001810out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 return err;
1812}
1813
1814/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001815 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 */
1817
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001818asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1819 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820{
1821 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1822}
1823
1824/*
1825 * Set a socket option. Because we don't know the option lengths we have
1826 * to pass the user mode parameter for the protocols to sort out.
1827 */
1828
Heiko Carstens20f37032009-01-14 14:14:23 +01001829SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1830 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001832 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 struct socket *sock;
1834
1835 if (optlen < 0)
1836 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001837
1838 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1839 if (sock != NULL) {
1840 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001841 if (err)
1842 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843
1844 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001845 err =
1846 sock_setsockopt(sock, level, optname, optval,
1847 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001849 err =
1850 sock->ops->setsockopt(sock, level, optname, optval,
1851 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001852out_put:
1853 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 }
1855 return err;
1856}
1857
1858/*
1859 * Get a socket option. Because we don't know the option lengths we have
1860 * to pass a user mode parameter for the protocols to sort out.
1861 */
1862
Heiko Carstens20f37032009-01-14 14:14:23 +01001863SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1864 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001866 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 struct socket *sock;
1868
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001869 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1870 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001871 err = security_socket_getsockopt(sock, level, optname);
1872 if (err)
1873 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
1875 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001876 err =
1877 sock_getsockopt(sock, level, optname, optval,
1878 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001880 err =
1881 sock->ops->getsockopt(sock, level, optname, optval,
1882 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001883out_put:
1884 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 }
1886 return err;
1887}
1888
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889/*
1890 * Shutdown a socket.
1891 */
1892
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001893SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001895 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 struct socket *sock;
1897
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001898 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1899 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001901 if (!err)
1902 err = sock->ops->shutdown(sock, how);
1903 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 }
1905 return err;
1906}
1907
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 * fields which are the same type (int / unsigned) on our platforms.
1910 */
1911#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1912#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1913#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1914
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915/*
1916 * BSD sendmsg interface
1917 */
1918
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001919SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001921 struct compat_msghdr __user *msg_compat =
1922 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001924 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001926 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001927 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1928 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 unsigned char *ctl_buf = ctl;
1930 struct msghdr msg_sys;
1931 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001932 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001933
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 err = -EFAULT;
1935 if (MSG_CMSG_COMPAT & flags) {
1936 if (get_compat_msghdr(&msg_sys, msg_compat))
1937 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001938 }
1939 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 return -EFAULT;
1941
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001942 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001943 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 goto out;
1945
1946 /* do not move before msg_sys is valid */
1947 err = -EMSGSIZE;
1948 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1949 goto out_put;
1950
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001951 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 err = -ENOMEM;
1953 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1954 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1955 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1956 if (!iov)
1957 goto out_put;
1958 }
1959
1960 /* This will also move the address data into kernel space */
1961 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001962 err = verify_compat_iovec(&msg_sys, iov,
1963 (struct sockaddr *)&address,
1964 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001966 err = verify_iovec(&msg_sys, iov,
1967 (struct sockaddr *)&address,
1968 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001969 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 goto out_freeiov;
1971 total_len = err;
1972
1973 err = -ENOBUFS;
1974
1975 if (msg_sys.msg_controllen > INT_MAX)
1976 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001977 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001979 err =
1980 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1981 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 if (err)
1983 goto out_freeiov;
1984 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001985 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001987 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001989 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 goto out_freeiov;
1991 }
1992 err = -EFAULT;
1993 /*
1994 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1995 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1996 * checking falls down on this.
1997 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001998 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1999 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 goto out_freectl;
2001 msg_sys.msg_control = ctl_buf;
2002 }
2003 msg_sys.msg_flags = flags;
2004
2005 if (sock->file->f_flags & O_NONBLOCK)
2006 msg_sys.msg_flags |= MSG_DONTWAIT;
2007 err = sock_sendmsg(sock, &msg_sys, total_len);
2008
2009out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002010 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2012out_freeiov:
2013 if (iov != iovstack)
2014 sock_kfree_s(sock->sk, iov, iov_size);
2015out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002016 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002017out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 return err;
2019}
2020
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002021static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
2022 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002024 struct compat_msghdr __user *msg_compat =
2025 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002027 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 unsigned long cmsg_ptr;
2029 int err, iov_size, total_len, len;
2030
2031 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002032 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
2034 /* user mode address pointers */
2035 struct sockaddr __user *uaddr;
2036 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002037
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002039 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002041 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002042 else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002043 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002046 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2047 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002048
2049 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002051 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
2052 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2054 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002055 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 }
2057
2058 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002059 * Save the user-mode address (verify_iovec will change the
2060 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002062
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002063 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 uaddr_len = COMPAT_NAMELEN(msg);
2065 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002066 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002067 (struct sockaddr *)&addr,
2068 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069 } else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002070 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002071 (struct sockaddr *)&addr,
2072 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 if (err < 0)
2074 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002075 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002077 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2078 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002079
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080 if (sock->file->f_flags & O_NONBLOCK)
2081 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002082 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2083 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 if (err < 0)
2085 goto out_freeiov;
2086 len = err;
2087
2088 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002089 err = move_addr_to_user((struct sockaddr *)&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002090 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002091 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 if (err < 0)
2093 goto out_freeiov;
2094 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002095 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002096 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 if (err)
2098 goto out_freeiov;
2099 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002100 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 &msg_compat->msg_controllen);
2102 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002103 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 &msg->msg_controllen);
2105 if (err)
2106 goto out_freeiov;
2107 err = len;
2108
2109out_freeiov:
2110 if (iov != iovstack)
2111 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002112out:
2113 return err;
2114}
2115
2116/*
2117 * BSD recvmsg interface
2118 */
2119
2120SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2121 unsigned int, flags)
2122{
2123 int fput_needed, err;
2124 struct msghdr msg_sys;
2125 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2126
2127 if (!sock)
2128 goto out;
2129
2130 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2131
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002132 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133out:
2134 return err;
2135}
2136
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002137/*
2138 * Linux recvmmsg interface
2139 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002141int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2142 unsigned int flags, struct timespec *timeout)
2143{
2144 int fput_needed, err, datagrams;
2145 struct socket *sock;
2146 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002147 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002148 struct msghdr msg_sys;
2149 struct timespec end_time;
2150
2151 if (timeout &&
2152 poll_select_set_timeout(&end_time, timeout->tv_sec,
2153 timeout->tv_nsec))
2154 return -EINVAL;
2155
2156 datagrams = 0;
2157
2158 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2159 if (!sock)
2160 return err;
2161
2162 err = sock_error(sock->sk);
2163 if (err)
2164 goto out_put;
2165
2166 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002167 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002168
2169 while (datagrams < vlen) {
2170 /*
2171 * No need to ask LSM for more than the first datagram.
2172 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002173 if (MSG_CMSG_COMPAT & flags) {
2174 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2175 &msg_sys, flags, datagrams);
2176 if (err < 0)
2177 break;
2178 err = __put_user(err, &compat_entry->msg_len);
2179 ++compat_entry;
2180 } else {
2181 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2182 &msg_sys, flags, datagrams);
2183 if (err < 0)
2184 break;
2185 err = put_user(err, &entry->msg_len);
2186 ++entry;
2187 }
2188
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002189 if (err)
2190 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002191 ++datagrams;
2192
2193 if (timeout) {
2194 ktime_get_ts(timeout);
2195 *timeout = timespec_sub(end_time, *timeout);
2196 if (timeout->tv_sec < 0) {
2197 timeout->tv_sec = timeout->tv_nsec = 0;
2198 break;
2199 }
2200
2201 /* Timeout, return less than vlen datagrams */
2202 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2203 break;
2204 }
2205
2206 /* Out of band data, return right away */
2207 if (msg_sys.msg_flags & MSG_OOB)
2208 break;
2209 }
2210
2211out_put:
2212 fput_light(sock->file, fput_needed);
2213
2214 if (err == 0)
2215 return datagrams;
2216
2217 if (datagrams != 0) {
2218 /*
2219 * We may return less entries than requested (vlen) if the
2220 * sock is non block and there aren't enough datagrams...
2221 */
2222 if (err != -EAGAIN) {
2223 /*
2224 * ... or if recvmsg returns an error after we
2225 * received some datagrams, where we record the
2226 * error to return on the next call or if the
2227 * app asks about it using getsockopt(SO_ERROR).
2228 */
2229 sock->sk->sk_err = -err;
2230 }
2231
2232 return datagrams;
2233 }
2234
2235 return err;
2236}
2237
2238SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2239 unsigned int, vlen, unsigned int, flags,
2240 struct timespec __user *, timeout)
2241{
2242 int datagrams;
2243 struct timespec timeout_sys;
2244
2245 if (!timeout)
2246 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2247
2248 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2249 return -EFAULT;
2250
2251 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2252
2253 if (datagrams > 0 &&
2254 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2255 datagrams = -EFAULT;
2256
2257 return datagrams;
2258}
2259
2260#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002261/* Argument list sizes for sys_socketcall */
2262#define AL(x) ((x) * sizeof(unsigned long))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002263static const unsigned char nargs[20] = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002264 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2265 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002266 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002267 AL(4),AL(5)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002268};
2269
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270#undef AL
2271
2272/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002273 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 *
2275 * Argument checking cleaned up. Saved 20% in size.
2276 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002277 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278 */
2279
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002280SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281{
2282 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002283 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002285 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002287 if (call < 1 || call > SYS_RECVMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288 return -EINVAL;
2289
Arjan van de Ven47379052009-09-28 12:57:44 -07002290 len = nargs[call];
2291 if (len > sizeof(a))
2292 return -EINVAL;
2293
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002295 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002297
Al Virof3298dc2008-12-10 03:16:51 -05002298 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002299
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002300 a0 = a[0];
2301 a1 = a[1];
2302
2303 switch (call) {
2304 case SYS_SOCKET:
2305 err = sys_socket(a0, a1, a[2]);
2306 break;
2307 case SYS_BIND:
2308 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2309 break;
2310 case SYS_CONNECT:
2311 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2312 break;
2313 case SYS_LISTEN:
2314 err = sys_listen(a0, a1);
2315 break;
2316 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002317 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2318 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002319 break;
2320 case SYS_GETSOCKNAME:
2321 err =
2322 sys_getsockname(a0, (struct sockaddr __user *)a1,
2323 (int __user *)a[2]);
2324 break;
2325 case SYS_GETPEERNAME:
2326 err =
2327 sys_getpeername(a0, (struct sockaddr __user *)a1,
2328 (int __user *)a[2]);
2329 break;
2330 case SYS_SOCKETPAIR:
2331 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2332 break;
2333 case SYS_SEND:
2334 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2335 break;
2336 case SYS_SENDTO:
2337 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2338 (struct sockaddr __user *)a[4], a[5]);
2339 break;
2340 case SYS_RECV:
2341 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2342 break;
2343 case SYS_RECVFROM:
2344 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2345 (struct sockaddr __user *)a[4],
2346 (int __user *)a[5]);
2347 break;
2348 case SYS_SHUTDOWN:
2349 err = sys_shutdown(a0, a1);
2350 break;
2351 case SYS_SETSOCKOPT:
2352 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2353 break;
2354 case SYS_GETSOCKOPT:
2355 err =
2356 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2357 (int __user *)a[4]);
2358 break;
2359 case SYS_SENDMSG:
2360 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2361 break;
2362 case SYS_RECVMSG:
2363 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2364 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002365 case SYS_RECVMMSG:
2366 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2367 (struct timespec __user *)a[4]);
2368 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002369 case SYS_ACCEPT4:
2370 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2371 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002372 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002373 default:
2374 err = -EINVAL;
2375 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 }
2377 return err;
2378}
2379
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002380#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002382/**
2383 * sock_register - add a socket protocol handler
2384 * @ops: description of protocol
2385 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386 * This function is called by a protocol handler that wants to
2387 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002388 * socket interface. The value ops->family coresponds to the
2389 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002391int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392{
2393 int err;
2394
2395 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002396 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2397 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398 return -ENOBUFS;
2399 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002400
2401 spin_lock(&net_family_lock);
2402 if (net_families[ops->family])
2403 err = -EEXIST;
2404 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002405 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406 err = 0;
2407 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002408 spin_unlock(&net_family_lock);
2409
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002410 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411 return err;
2412}
2413
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002414/**
2415 * sock_unregister - remove a protocol handler
2416 * @family: protocol family to remove
2417 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418 * This function is called by a protocol handler that wants to
2419 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002420 * new socket creation.
2421 *
2422 * If protocol handler is a module, then it can use module reference
2423 * counts to protect against new references. If protocol handler is not
2424 * a module then it needs to provide its own protection in
2425 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002427void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002429 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002431 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002432 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002433 spin_unlock(&net_family_lock);
2434
2435 synchronize_rcu();
2436
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002437 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438}
2439
Andi Kleen77d76ea2005-12-22 12:43:42 -08002440static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441{
2442 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002443 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002445
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446 sk_init();
2447
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002449 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 */
2451 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452
2453 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002454 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 */
2456
2457 init_inodecache();
2458 register_filesystem(&sock_fs_type);
2459 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002460
2461 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 */
2463
2464#ifdef CONFIG_NETFILTER
2465 netfilter_init();
2466#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002467
2468 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469}
2470
Andi Kleen77d76ea2005-12-22 12:43:42 -08002471core_initcall(sock_init); /* early initcall */
2472
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473#ifdef CONFIG_PROC_FS
2474void socket_seq_show(struct seq_file *seq)
2475{
2476 int cpu;
2477 int counter = 0;
2478
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002479 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002480 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002481
2482 /* It can be negative, by the way. 8) */
2483 if (counter < 0)
2484 counter = 0;
2485
2486 seq_printf(seq, "sockets: used %d\n", counter);
2487}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002488#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002489
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002490#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002491static int do_siocgstamp(struct net *net, struct socket *sock,
2492 unsigned int cmd, struct compat_timeval __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002493{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002494 mm_segment_t old_fs = get_fs();
2495 struct timeval ktv;
2496 int err;
2497
2498 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002499 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002500 set_fs(old_fs);
2501 if (!err) {
2502 err = put_user(ktv.tv_sec, &up->tv_sec);
2503 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2504 }
2505 return err;
2506}
2507
Arnd Bergmann6b960182009-11-06 23:10:54 -08002508static int do_siocgstampns(struct net *net, struct socket *sock,
2509 unsigned int cmd, struct compat_timespec __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002510{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002511 mm_segment_t old_fs = get_fs();
2512 struct timespec kts;
2513 int err;
2514
2515 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002516 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002517 set_fs(old_fs);
2518 if (!err) {
2519 err = put_user(kts.tv_sec, &up->tv_sec);
2520 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2521 }
2522 return err;
2523}
2524
Arnd Bergmann6b960182009-11-06 23:10:54 -08002525static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002526{
2527 struct ifreq __user *uifr;
2528 int err;
2529
2530 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002531 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002532 return -EFAULT;
2533
Arnd Bergmann6b960182009-11-06 23:10:54 -08002534 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002535 if (err)
2536 return err;
2537
Arnd Bergmann6b960182009-11-06 23:10:54 -08002538 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002539 return -EFAULT;
2540
2541 return 0;
2542}
2543
Arnd Bergmann6b960182009-11-06 23:10:54 -08002544static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002545{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002546 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002547 struct ifconf ifc;
2548 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002549 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002550 struct ifreq __user *ifr;
2551 unsigned int i, j;
2552 int err;
2553
Arnd Bergmann6b960182009-11-06 23:10:54 -08002554 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002555 return -EFAULT;
2556
2557 if (ifc32.ifcbuf == 0) {
2558 ifc32.ifc_len = 0;
2559 ifc.ifc_len = 0;
2560 ifc.ifc_req = NULL;
2561 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2562 } else {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002563 size_t len =((ifc32.ifc_len / sizeof (struct compat_ifreq)) + 1) *
Arnd Bergmann7a229382009-11-06 23:00:29 -08002564 sizeof (struct ifreq);
2565 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2566 ifc.ifc_len = len;
2567 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2568 ifr32 = compat_ptr(ifc32.ifcbuf);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002569 for (i = 0; i < ifc32.ifc_len; i += sizeof (struct compat_ifreq)) {
2570 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002571 return -EFAULT;
2572 ifr++;
2573 ifr32++;
2574 }
2575 }
2576 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2577 return -EFAULT;
2578
Arnd Bergmann6b960182009-11-06 23:10:54 -08002579 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002580 if (err)
2581 return err;
2582
2583 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2584 return -EFAULT;
2585
2586 ifr = ifc.ifc_req;
2587 ifr32 = compat_ptr(ifc32.ifcbuf);
2588 for (i = 0, j = 0;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002589 i + sizeof (struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2590 i += sizeof (struct compat_ifreq), j += sizeof (struct ifreq)) {
2591 if (copy_in_user(ifr32, ifr, sizeof (struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002592 return -EFAULT;
2593 ifr32++;
2594 ifr++;
2595 }
2596
2597 if (ifc32.ifcbuf == 0) {
2598 /* Translate from 64-bit structure multiple to
2599 * a 32-bit one.
2600 */
2601 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002602 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002603 ifc32.ifc_len = i;
2604 } else {
2605 ifc32.ifc_len = i;
2606 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002607 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002608 return -EFAULT;
2609
2610 return 0;
2611}
2612
Arnd Bergmann6b960182009-11-06 23:10:54 -08002613static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002614{
2615 struct ifreq __user *ifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002616 u32 data;
2617 void __user *datap;
2618
2619 ifr = compat_alloc_user_space(sizeof(*ifr));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002620
2621 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2622 return -EFAULT;
2623
2624 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2625 return -EFAULT;
2626
2627 datap = compat_ptr(data);
2628 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2629 return -EFAULT;
2630
Arnd Bergmann6b960182009-11-06 23:10:54 -08002631 return dev_ioctl(net, SIOCETHTOOL, ifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002632}
2633
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002634static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2635{
2636 void __user *uptr;
2637 compat_uptr_t uptr32;
2638 struct ifreq __user *uifr;
2639
2640 uifr = compat_alloc_user_space(sizeof (*uifr));
2641 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2642 return -EFAULT;
2643
2644 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2645 return -EFAULT;
2646
2647 uptr = compat_ptr(uptr32);
2648
2649 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2650 return -EFAULT;
2651
2652 return dev_ioctl(net, SIOCWANDEV, uifr);
2653}
2654
Arnd Bergmann6b960182009-11-06 23:10:54 -08002655static int bond_ioctl(struct net *net, unsigned int cmd,
2656 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002657{
2658 struct ifreq kifr;
2659 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002660 mm_segment_t old_fs;
2661 int err;
2662 u32 data;
2663 void __user *datap;
2664
2665 switch (cmd) {
2666 case SIOCBONDENSLAVE:
2667 case SIOCBONDRELEASE:
2668 case SIOCBONDSETHWADDR:
2669 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002670 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002671 return -EFAULT;
2672
2673 old_fs = get_fs();
2674 set_fs (KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002675 err = dev_ioctl(net, cmd, &kifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002676 set_fs (old_fs);
2677
2678 return err;
2679 case SIOCBONDSLAVEINFOQUERY:
2680 case SIOCBONDINFOQUERY:
2681 uifr = compat_alloc_user_space(sizeof(*uifr));
2682 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2683 return -EFAULT;
2684
2685 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2686 return -EFAULT;
2687
2688 datap = compat_ptr(data);
2689 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2690 return -EFAULT;
2691
Arnd Bergmann6b960182009-11-06 23:10:54 -08002692 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002693 default:
2694 return -EINVAL;
2695 };
2696}
2697
Arnd Bergmann6b960182009-11-06 23:10:54 -08002698static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2699 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002700{
2701 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002702 char tmp_buf[IFNAMSIZ];
2703 void __user *data64;
2704 u32 data32;
2705
2706 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2707 IFNAMSIZ))
2708 return -EFAULT;
2709 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2710 return -EFAULT;
2711 data64 = compat_ptr(data32);
2712
2713 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2714
2715 /* Don't check these user accesses, just let that get trapped
2716 * in the ioctl handler instead.
2717 */
2718 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2719 IFNAMSIZ))
2720 return -EFAULT;
2721 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2722 return -EFAULT;
2723
Arnd Bergmann6b960182009-11-06 23:10:54 -08002724 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002725}
2726
Arnd Bergmann6b960182009-11-06 23:10:54 -08002727static int dev_ifsioc(struct net *net, struct socket *sock,
2728 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002729{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002730 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002731 int err;
2732
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002733 uifr = compat_alloc_user_space(sizeof(*uifr));
2734 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2735 return -EFAULT;
2736
2737 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2738
Arnd Bergmann7a229382009-11-06 23:00:29 -08002739 if (!err) {
2740 switch (cmd) {
2741 case SIOCGIFFLAGS:
2742 case SIOCGIFMETRIC:
2743 case SIOCGIFMTU:
2744 case SIOCGIFMEM:
2745 case SIOCGIFHWADDR:
2746 case SIOCGIFINDEX:
2747 case SIOCGIFADDR:
2748 case SIOCGIFBRDADDR:
2749 case SIOCGIFDSTADDR:
2750 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002751 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002752 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002753 case SIOCGMIIPHY:
2754 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002755 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002756 err = -EFAULT;
2757 break;
2758 }
2759 }
2760 return err;
2761}
2762
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002763static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2764 struct compat_ifreq __user *uifr32)
2765{
2766 struct ifreq ifr;
2767 struct compat_ifmap __user *uifmap32;
2768 mm_segment_t old_fs;
2769 int err;
2770
2771 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2772 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2773 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2774 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2775 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2776 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2777 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2778 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2779 if (err)
2780 return -EFAULT;
2781
2782 old_fs = get_fs();
2783 set_fs (KERNEL_DS);
2784 err = dev_ioctl(net, cmd, (void __user *)&ifr);
2785 set_fs (old_fs);
2786
2787 if (cmd == SIOCGIFMAP && !err) {
2788 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2789 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2790 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2791 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2792 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2793 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2794 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2795 if (err)
2796 err = -EFAULT;
2797 }
2798 return err;
2799}
2800
2801static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2802{
2803 void __user *uptr;
2804 compat_uptr_t uptr32;
2805 struct ifreq __user *uifr;
2806
2807 uifr = compat_alloc_user_space(sizeof (*uifr));
2808 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2809 return -EFAULT;
2810
2811 if (get_user(uptr32, &uifr32->ifr_data))
2812 return -EFAULT;
2813
2814 uptr = compat_ptr(uptr32);
2815
2816 if (put_user(uptr, &uifr->ifr_data))
2817 return -EFAULT;
2818
2819 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2820}
2821
Arnd Bergmann7a229382009-11-06 23:00:29 -08002822struct rtentry32 {
2823 u32 rt_pad1;
2824 struct sockaddr rt_dst; /* target address */
2825 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2826 struct sockaddr rt_genmask; /* target network mask (IP) */
2827 unsigned short rt_flags;
2828 short rt_pad2;
2829 u32 rt_pad3;
2830 unsigned char rt_tos;
2831 unsigned char rt_class;
2832 short rt_pad4;
2833 short rt_metric; /* +1 for binary compatibility! */
2834 /* char * */ u32 rt_dev; /* forcing the device at add */
2835 u32 rt_mtu; /* per route MTU/Window */
2836 u32 rt_window; /* Window clamping */
2837 unsigned short rt_irtt; /* Initial RTT */
2838};
2839
2840struct in6_rtmsg32 {
2841 struct in6_addr rtmsg_dst;
2842 struct in6_addr rtmsg_src;
2843 struct in6_addr rtmsg_gateway;
2844 u32 rtmsg_type;
2845 u16 rtmsg_dst_len;
2846 u16 rtmsg_src_len;
2847 u32 rtmsg_metric;
2848 u32 rtmsg_info;
2849 u32 rtmsg_flags;
2850 s32 rtmsg_ifindex;
2851};
2852
Arnd Bergmann6b960182009-11-06 23:10:54 -08002853static int routing_ioctl(struct net *net, struct socket *sock,
2854 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002855{
2856 int ret;
2857 void *r = NULL;
2858 struct in6_rtmsg r6;
2859 struct rtentry r4;
2860 char devname[16];
2861 u32 rtdev;
2862 mm_segment_t old_fs = get_fs();
2863
Arnd Bergmann6b960182009-11-06 23:10:54 -08002864 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2865 struct in6_rtmsg32 __user *ur6 = argp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002866 ret = copy_from_user (&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2867 3 * sizeof(struct in6_addr));
2868 ret |= __get_user (r6.rtmsg_type, &(ur6->rtmsg_type));
2869 ret |= __get_user (r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2870 ret |= __get_user (r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2871 ret |= __get_user (r6.rtmsg_metric, &(ur6->rtmsg_metric));
2872 ret |= __get_user (r6.rtmsg_info, &(ur6->rtmsg_info));
2873 ret |= __get_user (r6.rtmsg_flags, &(ur6->rtmsg_flags));
2874 ret |= __get_user (r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2875
2876 r = (void *) &r6;
2877 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002878 struct rtentry32 __user *ur4 = argp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002879 ret = copy_from_user (&r4.rt_dst, &(ur4->rt_dst),
2880 3 * sizeof(struct sockaddr));
2881 ret |= __get_user (r4.rt_flags, &(ur4->rt_flags));
2882 ret |= __get_user (r4.rt_metric, &(ur4->rt_metric));
2883 ret |= __get_user (r4.rt_mtu, &(ur4->rt_mtu));
2884 ret |= __get_user (r4.rt_window, &(ur4->rt_window));
2885 ret |= __get_user (r4.rt_irtt, &(ur4->rt_irtt));
2886 ret |= __get_user (rtdev, &(ur4->rt_dev));
2887 if (rtdev) {
2888 ret |= copy_from_user (devname, compat_ptr(rtdev), 15);
2889 r4.rt_dev = devname; devname[15] = 0;
2890 } else
2891 r4.rt_dev = NULL;
2892
2893 r = (void *) &r4;
2894 }
2895
2896 if (ret) {
2897 ret = -EFAULT;
2898 goto out;
2899 }
2900
2901 set_fs (KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002902 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002903 set_fs (old_fs);
2904
2905out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002906 return ret;
2907}
2908
2909/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2910 * for some operations; this forces use of the newer bridge-utils that
2911 * use compatiable ioctls
2912 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002913static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002914{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002915 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002916
Arnd Bergmann6b960182009-11-06 23:10:54 -08002917 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002918 return -EFAULT;
2919 if (tmp == BRCTL_GET_VERSION)
2920 return BRCTL_VERSION + 1;
2921 return -EINVAL;
2922}
2923
Arnd Bergmann6b960182009-11-06 23:10:54 -08002924static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2925 unsigned int cmd, unsigned long arg)
2926{
2927 void __user *argp = compat_ptr(arg);
2928 struct sock *sk = sock->sk;
2929 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002930
Arnd Bergmann6b960182009-11-06 23:10:54 -08002931 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2932 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002933
Arnd Bergmann6b960182009-11-06 23:10:54 -08002934 switch (cmd) {
2935 case SIOCSIFBR:
2936 case SIOCGIFBR:
2937 return old_bridge_ioctl(argp);
2938 case SIOCGIFNAME:
2939 return dev_ifname32(net, argp);
2940 case SIOCGIFCONF:
2941 return dev_ifconf(net, argp);
2942 case SIOCETHTOOL:
2943 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002944 case SIOCWANDEV:
2945 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002946 case SIOCGIFMAP:
2947 case SIOCSIFMAP:
2948 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002949 case SIOCBONDENSLAVE:
2950 case SIOCBONDRELEASE:
2951 case SIOCBONDSETHWADDR:
2952 case SIOCBONDSLAVEINFOQUERY:
2953 case SIOCBONDINFOQUERY:
2954 case SIOCBONDCHANGEACTIVE:
2955 return bond_ioctl(net, cmd, argp);
2956 case SIOCADDRT:
2957 case SIOCDELRT:
2958 return routing_ioctl(net, sock, cmd, argp);
2959 case SIOCGSTAMP:
2960 return do_siocgstamp(net, sock, cmd, argp);
2961 case SIOCGSTAMPNS:
2962 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002963 case SIOCSHWTSTAMP:
2964 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002965
Arnd Bergmann6b960182009-11-06 23:10:54 -08002966 case FIOSETOWN:
2967 case SIOCSPGRP:
2968 case FIOGETOWN:
2969 case SIOCGPGRP:
2970 case SIOCBRADDBR:
2971 case SIOCBRDELBR:
2972 case SIOCGIFVLAN:
2973 case SIOCSIFVLAN:
2974 case SIOCADDDLCI:
2975 case SIOCDELDLCI:
2976 return sock_ioctl(file, cmd, arg);
2977
2978 case SIOCGIFFLAGS:
2979 case SIOCSIFFLAGS:
2980 case SIOCGIFMETRIC:
2981 case SIOCSIFMETRIC:
2982 case SIOCGIFMTU:
2983 case SIOCSIFMTU:
2984 case SIOCGIFMEM:
2985 case SIOCSIFMEM:
2986 case SIOCGIFHWADDR:
2987 case SIOCSIFHWADDR:
2988 case SIOCADDMULTI:
2989 case SIOCDELMULTI:
2990 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002991 case SIOCGIFADDR:
2992 case SIOCSIFADDR:
2993 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002994 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002995 case SIOCGIFBRDADDR:
2996 case SIOCSIFBRDADDR:
2997 case SIOCGIFDSTADDR:
2998 case SIOCSIFDSTADDR:
2999 case SIOCGIFNETMASK:
3000 case SIOCSIFNETMASK:
3001 case SIOCSIFPFLAGS:
3002 case SIOCGIFPFLAGS:
3003 case SIOCGIFTXQLEN:
3004 case SIOCSIFTXQLEN:
3005 case SIOCBRADDIF:
3006 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003007 case SIOCSIFNAME:
3008 case SIOCGMIIPHY:
3009 case SIOCGMIIREG:
3010 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003011 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003012
Arnd Bergmann6b960182009-11-06 23:10:54 -08003013 case SIOCSARP:
3014 case SIOCGARP:
3015 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003016 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003017 return sock_do_ioctl(net, sock, cmd, arg);
3018 }
3019
3020 /* Prevent warning from compat_sys_ioctl, these always
3021 * result in -EINVAL in the native case anyway. */
3022 switch (cmd) {
3023 case SIOCRTMSG:
3024 case SIOCGIFCOUNT:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003025 case SIOCSRARP:
3026 case SIOCGRARP:
3027 case SIOCDRARP:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003028 case SIOCSIFLINK:
3029 case SIOCGIFSLAVE:
3030 case SIOCSIFSLAVE:
3031 return -EINVAL;
Arnd Bergmann6b960182009-11-06 23:10:54 -08003032 }
3033
3034 return -ENOIOCTLCMD;
3035}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003036
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003037static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003038 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003039{
3040 struct socket *sock = file->private_data;
3041 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003042 struct sock *sk;
3043 struct net *net;
3044
3045 sk = sock->sk;
3046 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003047
3048 if (sock->ops->compat_ioctl)
3049 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3050
David S. Miller87de87d2008-06-03 09:14:03 -07003051 if (ret == -ENOIOCTLCMD &&
3052 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3053 ret = compat_wext_handle_ioctl(net, cmd, arg);
3054
Arnd Bergmann6b960182009-11-06 23:10:54 -08003055 if (ret == -ENOIOCTLCMD)
3056 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3057
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003058 return ret;
3059}
3060#endif
3061
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003062int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3063{
3064 return sock->ops->bind(sock, addr, addrlen);
3065}
3066
3067int kernel_listen(struct socket *sock, int backlog)
3068{
3069 return sock->ops->listen(sock, backlog);
3070}
3071
3072int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3073{
3074 struct sock *sk = sock->sk;
3075 int err;
3076
3077 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3078 newsock);
3079 if (err < 0)
3080 goto done;
3081
3082 err = sock->ops->accept(sock, *newsock, flags);
3083 if (err < 0) {
3084 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003085 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003086 goto done;
3087 }
3088
3089 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003090 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003091
3092done:
3093 return err;
3094}
3095
3096int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003097 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003098{
3099 return sock->ops->connect(sock, addr, addrlen, flags);
3100}
3101
3102int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3103 int *addrlen)
3104{
3105 return sock->ops->getname(sock, addr, addrlen, 0);
3106}
3107
3108int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3109 int *addrlen)
3110{
3111 return sock->ops->getname(sock, addr, addrlen, 1);
3112}
3113
3114int kernel_getsockopt(struct socket *sock, int level, int optname,
3115 char *optval, int *optlen)
3116{
3117 mm_segment_t oldfs = get_fs();
3118 int err;
3119
3120 set_fs(KERNEL_DS);
3121 if (level == SOL_SOCKET)
3122 err = sock_getsockopt(sock, level, optname, optval, optlen);
3123 else
3124 err = sock->ops->getsockopt(sock, level, optname, optval,
3125 optlen);
3126 set_fs(oldfs);
3127 return err;
3128}
3129
3130int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003131 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003132{
3133 mm_segment_t oldfs = get_fs();
3134 int err;
3135
3136 set_fs(KERNEL_DS);
3137 if (level == SOL_SOCKET)
3138 err = sock_setsockopt(sock, level, optname, optval, optlen);
3139 else
3140 err = sock->ops->setsockopt(sock, level, optname, optval,
3141 optlen);
3142 set_fs(oldfs);
3143 return err;
3144}
3145
3146int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3147 size_t size, int flags)
3148{
3149 if (sock->ops->sendpage)
3150 return sock->ops->sendpage(sock, page, offset, size, flags);
3151
3152 return sock_no_sendpage(sock, page, offset, size, flags);
3153}
3154
3155int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3156{
3157 mm_segment_t oldfs = get_fs();
3158 int err;
3159
3160 set_fs(KERNEL_DS);
3161 err = sock->ops->ioctl(sock, cmd, arg);
3162 set_fs(oldfs);
3163
3164 return err;
3165}
3166
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003167int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3168{
3169 return sock->ops->shutdown(sock, how);
3170}
3171
Linus Torvalds1da177e2005-04-16 15:20:36 -07003172EXPORT_SYMBOL(sock_create);
3173EXPORT_SYMBOL(sock_create_kern);
3174EXPORT_SYMBOL(sock_create_lite);
3175EXPORT_SYMBOL(sock_map_fd);
3176EXPORT_SYMBOL(sock_recvmsg);
3177EXPORT_SYMBOL(sock_register);
3178EXPORT_SYMBOL(sock_release);
3179EXPORT_SYMBOL(sock_sendmsg);
3180EXPORT_SYMBOL(sock_unregister);
3181EXPORT_SYMBOL(sock_wake_async);
3182EXPORT_SYMBOL(sockfd_lookup);
3183EXPORT_SYMBOL(kernel_sendmsg);
3184EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003185EXPORT_SYMBOL(kernel_bind);
3186EXPORT_SYMBOL(kernel_listen);
3187EXPORT_SYMBOL(kernel_accept);
3188EXPORT_SYMBOL(kernel_connect);
3189EXPORT_SYMBOL(kernel_getsockname);
3190EXPORT_SYMBOL(kernel_getpeername);
3191EXPORT_SYMBOL(kernel_getsockopt);
3192EXPORT_SYMBOL(kernel_setsockopt);
3193EXPORT_SYMBOL(kernel_sendpage);
3194EXPORT_SYMBOL(kernel_sock_ioctl);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003195EXPORT_SYMBOL(kernel_sock_shutdown);