blob: 769c386bd4288758635acde6ca77c401a49c61a8 [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
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700315/*
316 * sockfs_dname() is called from d_path().
317 */
318static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
319{
320 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
321 dentry->d_inode->i_ino);
322}
323
Al Viro3ba13d12009-02-20 06:02:22 +0000324static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700325 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326};
327
328/*
329 * Obtains the first available file descriptor and sets it up for use.
330 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800331 * These functions create file structures and maps them to fd space
332 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 * and file struct implicitly stored in sock->file.
334 * Note that another thread may close file descriptor before we return
335 * from this function. We use the fact that now we do not refer
336 * to socket after mapping. If one day we will need it, this
337 * function will increment ref. count on file by 1.
338 *
339 * In any case returned fd MAY BE not valid!
340 * This race condition is unavoidable
341 * with shared fd spaces, we cannot solve it inside kernel,
342 * but we take care of internal coherence yet.
343 */
344
Al Viro7cbe66b2009-08-05 19:59:08 +0400345static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346{
Al Viro7cbe66b2009-08-05 19:59:08 +0400347 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400348 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400349 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800351
Ulrich Dreppera677a032008-07-23 21:29:17 -0700352 fd = get_unused_fd_flags(flags);
Al Viro7cbe66b2009-08-05 19:59:08 +0400353 if (unlikely(fd < 0))
354 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800355
Al Viro2c48b9c2009-08-09 00:52:35 +0400356 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
357 if (unlikely(!path.dentry)) {
Al Viro7cbe66b2009-08-05 19:59:08 +0400358 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800359 return -ENOMEM;
Al Viro7cbe66b2009-08-05 19:59:08 +0400360 }
Al Viro2c48b9c2009-08-09 00:52:35 +0400361 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362
Al Viro2c48b9c2009-08-09 00:52:35 +0400363 path.dentry->d_op = &sockfs_dentry_operations;
Al Viro2c48b9c2009-08-09 00:52:35 +0400364 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400365 SOCK_INODE(sock)->i_fop = &socket_file_ops;
366
Al Viro2c48b9c2009-08-09 00:52:35 +0400367 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400368 &socket_file_ops);
369 if (unlikely(!file)) {
370 /* drop dentry, keep inode */
371 atomic_inc(&path.dentry->d_inode->i_count);
Al Viro2c48b9c2009-08-09 00:52:35 +0400372 path_put(&path);
Al Virocc3808f2009-08-06 09:43:59 +0400373 put_unused_fd(fd);
374 return -ENFILE;
375 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800376
377 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700378 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800379 file->f_pos = 0;
380 file->private_data = sock;
381
Al Viro7cbe66b2009-08-05 19:59:08 +0400382 *f = file;
383 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800384}
385
Ulrich Dreppera677a032008-07-23 21:29:17 -0700386int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387{
388 struct file *newfile;
Al Viro7cbe66b2009-08-05 19:59:08 +0400389 int fd = sock_alloc_file(sock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800390
Al Viro7cbe66b2009-08-05 19:59:08 +0400391 if (likely(fd >= 0))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392 fd_install(fd, newfile);
Al Viro7cbe66b2009-08-05 19:59:08 +0400393
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 return fd;
395}
396
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800397static struct socket *sock_from_file(struct file *file, int *err)
398{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800399 if (file->f_op == &socket_file_ops)
400 return file->private_data; /* set in sock_map_fd */
401
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800402 *err = -ENOTSOCK;
403 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800404}
405
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406/**
407 * sockfd_lookup - Go from a file number to its socket slot
408 * @fd: file handle
409 * @err: pointer to an error code return
410 *
411 * The file handle passed in is locked and the socket it is bound
412 * too is returned. If an error occurs the err pointer is overwritten
413 * with a negative errno code and NULL is returned. The function checks
414 * for both invalid handles and passing a handle which is not a socket.
415 *
416 * On a success the socket object pointer is returned.
417 */
418
419struct socket *sockfd_lookup(int fd, int *err)
420{
421 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 struct socket *sock;
423
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700424 file = fget(fd);
425 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 *err = -EBADF;
427 return NULL;
428 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700429
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800430 sock = sock_from_file(file, err);
431 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 return sock;
434}
435
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800436static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
437{
438 struct file *file;
439 struct socket *sock;
440
Hua Zhong36725582006-04-19 15:25:02 -0700441 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800442 file = fget_light(fd, fput_needed);
443 if (file) {
444 sock = sock_from_file(file, err);
445 if (sock)
446 return sock;
447 fput_light(file, *fput_needed);
448 }
449 return NULL;
450}
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452/**
453 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700454 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455 * Allocate a new inode and socket object. The two are bound together
456 * and initialised. The socket is then returned. If we are out of inodes
457 * NULL is returned.
458 */
459
460static struct socket *sock_alloc(void)
461{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700462 struct inode *inode;
463 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464
465 inode = new_inode(sock_mnt->mnt_sb);
466 if (!inode)
467 return NULL;
468
469 sock = SOCKET_I(inode);
470
Eric Dumazet29a020d2009-09-15 02:39:20 -0700471 kmemcheck_annotate_bitfield(sock, type);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700472 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100473 inode->i_uid = current_fsuid();
474 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
Eric Dumazet4e694892009-04-04 16:41:09 -0700476 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 return sock;
478}
479
480/*
481 * In theory you can't get an open on this inode, but /proc provides
482 * a back door. Remember to keep it shut otherwise you'll let the
483 * creepy crawlies in.
484 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700485
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
487{
488 return -ENXIO;
489}
490
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800491const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 .owner = THIS_MODULE,
493 .open = sock_no_open,
494};
495
496/**
497 * sock_release - close a socket
498 * @sock: socket to close
499 *
500 * The socket is released from the protocol stack if it has a release
501 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700502 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505void sock_release(struct socket *sock)
506{
507 if (sock->ops) {
508 struct module *owner = sock->ops->owner;
509
510 sock->ops->release(sock);
511 sock->ops = NULL;
512 module_put(owner);
513 }
514
515 if (sock->fasync_list)
516 printk(KERN_ERR "sock_release: fasync list not empty!\n");
517
Eric Dumazet4e694892009-04-04 16:41:09 -0700518 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 if (!sock->file) {
520 iput(SOCK_INODE(sock));
521 return;
522 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700523 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524}
525
Patrick Ohly20d49472009-02-12 05:03:38 +0000526int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
527 union skb_shared_tx *shtx)
528{
529 shtx->flags = 0;
530 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
531 shtx->hardware = 1;
532 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
533 shtx->software = 1;
534 return 0;
535}
536EXPORT_SYMBOL(sock_tx_timestamp);
537
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700538static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 struct msghdr *msg, size_t size)
540{
541 struct sock_iocb *si = kiocb_to_siocb(iocb);
542 int err;
543
544 si->sock = sock;
545 si->scm = NULL;
546 si->msg = msg;
547 si->size = size;
548
549 err = security_socket_sendmsg(sock, msg, size);
550 if (err)
551 return err;
552
553 return sock->ops->sendmsg(iocb, sock, msg, size);
554}
555
556int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
557{
558 struct kiocb iocb;
559 struct sock_iocb siocb;
560 int ret;
561
562 init_sync_kiocb(&iocb, NULL);
563 iocb.private = &siocb;
564 ret = __sock_sendmsg(&iocb, sock, msg, size);
565 if (-EIOCBQUEUED == ret)
566 ret = wait_on_sync_kiocb(&iocb);
567 return ret;
568}
569
570int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
571 struct kvec *vec, size_t num, size_t size)
572{
573 mm_segment_t oldfs = get_fs();
574 int result;
575
576 set_fs(KERNEL_DS);
577 /*
578 * the following is safe, since for compiler definitions of kvec and
579 * iovec are identical, yielding the same in-core layout and alignment
580 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700581 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 msg->msg_iovlen = num;
583 result = sock_sendmsg(sock, msg, size);
584 set_fs(oldfs);
585 return result;
586}
587
Patrick Ohly20d49472009-02-12 05:03:38 +0000588static int ktime2ts(ktime_t kt, struct timespec *ts)
589{
590 if (kt.tv64) {
591 *ts = ktime_to_timespec(kt);
592 return 1;
593 } else {
594 return 0;
595 }
596}
597
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700598/*
599 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
600 */
601void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
602 struct sk_buff *skb)
603{
Patrick Ohly20d49472009-02-12 05:03:38 +0000604 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
605 struct timespec ts[3];
606 int empty = 1;
607 struct skb_shared_hwtstamps *shhwtstamps =
608 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700609
Patrick Ohly20d49472009-02-12 05:03:38 +0000610 /* Race occurred between timestamp enabling and packet
611 receiving. Fill in the current time for now. */
612 if (need_software_tstamp && skb->tstamp.tv64 == 0)
613 __net_timestamp(skb);
614
615 if (need_software_tstamp) {
616 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
617 struct timeval tv;
618 skb_get_timestamp(skb, &tv);
619 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
620 sizeof(tv), &tv);
621 } else {
622 struct timespec ts;
623 skb_get_timestampns(skb, &ts);
624 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
625 sizeof(ts), &ts);
626 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700627 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000628
629
630 memset(ts, 0, sizeof(ts));
631 if (skb->tstamp.tv64 &&
632 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
633 skb_get_timestampns(skb, ts + 0);
634 empty = 0;
635 }
636 if (shhwtstamps) {
637 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
638 ktime2ts(shhwtstamps->syststamp, ts + 1))
639 empty = 0;
640 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
641 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
642 empty = 0;
643 }
644 if (!empty)
645 put_cmsg(msg, SOL_SOCKET,
646 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700647}
648
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300649EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
650
Neil Horman3b885782009-10-12 13:26:31 -0700651inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
652{
653 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
654 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
655 sizeof(__u32), &skb->dropcount);
656}
657
658void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
659 struct sk_buff *skb)
660{
661 sock_recv_timestamp(msg, sk, skb);
662 sock_recv_drops(msg, sk, skb);
663}
664EXPORT_SYMBOL_GPL(sock_recv_ts_and_drops);
665
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700666static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
667 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 struct sock_iocb *si = kiocb_to_siocb(iocb);
670
671 si->sock = sock;
672 si->scm = NULL;
673 si->msg = msg;
674 si->size = size;
675 si->flags = flags;
676
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
678}
679
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700680static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
681 struct msghdr *msg, size_t size, int flags)
682{
683 int err = security_socket_recvmsg(sock, msg, size, flags);
684
685 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
686}
687
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700688int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 size_t size, int flags)
690{
691 struct kiocb iocb;
692 struct sock_iocb siocb;
693 int ret;
694
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700695 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 iocb.private = &siocb;
697 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
698 if (-EIOCBQUEUED == ret)
699 ret = wait_on_sync_kiocb(&iocb);
700 return ret;
701}
702
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700703static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
704 size_t size, int flags)
705{
706 struct kiocb iocb;
707 struct sock_iocb siocb;
708 int ret;
709
710 init_sync_kiocb(&iocb, NULL);
711 iocb.private = &siocb;
712 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
713 if (-EIOCBQUEUED == ret)
714 ret = wait_on_sync_kiocb(&iocb);
715 return ret;
716}
717
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700718int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
719 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720{
721 mm_segment_t oldfs = get_fs();
722 int result;
723
724 set_fs(KERNEL_DS);
725 /*
726 * the following is safe, since for compiler definitions of kvec and
727 * iovec are identical, yielding the same in-core layout and alignment
728 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700729 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 result = sock_recvmsg(sock, msg, size, flags);
731 set_fs(oldfs);
732 return result;
733}
734
735static void sock_aio_dtor(struct kiocb *iocb)
736{
737 kfree(iocb->private);
738}
739
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300740static ssize_t sock_sendpage(struct file *file, struct page *page,
741 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742{
743 struct socket *sock;
744 int flags;
745
Eric Dumazetb69aee02005-09-06 14:42:45 -0700746 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
748 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
749 if (more)
750 flags |= MSG_MORE;
751
Linus Torvaldse6949582009-08-13 08:28:36 -0700752 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753}
754
Jens Axboe9c55e012007-11-06 23:30:13 -0800755static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
756 struct pipe_inode_info *pipe, size_t len,
757 unsigned int flags)
758{
759 struct socket *sock = file->private_data;
760
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800761 if (unlikely(!sock->ops->splice_read))
762 return -EINVAL;
763
Jens Axboe9c55e012007-11-06 23:30:13 -0800764 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
765}
766
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800767static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700768 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800769{
770 if (!is_sync_kiocb(iocb)) {
771 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
772 if (!siocb)
773 return NULL;
774 iocb->ki_dtor = sock_aio_dtor;
775 }
776
777 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800778 iocb->private = siocb;
779 return siocb;
780}
781
782static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700783 struct file *file, const struct iovec *iov,
784 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800785{
786 struct socket *sock = file->private_data;
787 size_t size = 0;
788 int i;
789
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700790 for (i = 0; i < nr_segs; i++)
791 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800792
793 msg->msg_name = NULL;
794 msg->msg_namelen = 0;
795 msg->msg_control = NULL;
796 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700797 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800798 msg->msg_iovlen = nr_segs;
799 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
800
801 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
802}
803
Badari Pulavarty027445c2006-09-30 23:28:46 -0700804static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
805 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800806{
807 struct sock_iocb siocb, *x;
808
809 if (pos != 0)
810 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700811
812 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800813 return 0;
814
Badari Pulavarty027445c2006-09-30 23:28:46 -0700815
816 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800817 if (!x)
818 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700819 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800820}
821
822static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700823 struct file *file, const struct iovec *iov,
824 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800825{
826 struct socket *sock = file->private_data;
827 size_t size = 0;
828 int i;
829
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700830 for (i = 0; i < nr_segs; i++)
831 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800832
833 msg->msg_name = NULL;
834 msg->msg_namelen = 0;
835 msg->msg_control = NULL;
836 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700837 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800838 msg->msg_iovlen = nr_segs;
839 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
840 if (sock->type == SOCK_SEQPACKET)
841 msg->msg_flags |= MSG_EOR;
842
843 return __sock_sendmsg(iocb, sock, msg, size);
844}
845
Badari Pulavarty027445c2006-09-30 23:28:46 -0700846static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
847 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800849 struct sock_iocb siocb, *x;
850
851 if (pos != 0)
852 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700853
Badari Pulavarty027445c2006-09-30 23:28:46 -0700854 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800855 if (!x)
856 return -ENOMEM;
857
Badari Pulavarty027445c2006-09-30 23:28:46 -0700858 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859}
860
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861/*
862 * Atomic setting of ioctl hooks to avoid race
863 * with module unload.
864 */
865
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800866static DEFINE_MUTEX(br_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700867static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
Eric W. Biederman881d9662007-09-17 11:56:21 -0700869void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800871 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800873 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700875
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876EXPORT_SYMBOL(brioctl_set);
877
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800878static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700879static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880
Eric W. Biederman881d9662007-09-17 11:56:21 -0700881void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800883 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800885 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888EXPORT_SYMBOL(vlan_ioctl_set);
889
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800890static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700891static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700893void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800895 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800897 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700899
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900EXPORT_SYMBOL(dlci_ioctl_set);
901
Arnd Bergmann6b960182009-11-06 23:10:54 -0800902static long sock_do_ioctl(struct net *net, struct socket *sock,
903 unsigned int cmd, unsigned long arg)
904{
905 int err;
906 void __user *argp = (void __user *)arg;
907
908 err = sock->ops->ioctl(sock, cmd, arg);
909
910 /*
911 * If this ioctl is unknown try to hand it down
912 * to the NIC driver.
913 */
914 if (err == -ENOIOCTLCMD)
915 err = dev_ioctl(net, cmd, argp);
916
917 return err;
918}
919
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920/*
921 * With an ioctl, arg may well be a user mode pointer, but we don't know
922 * what to do with it - that's up to the protocol still.
923 */
924
925static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
926{
927 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700928 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 void __user *argp = (void __user *)arg;
930 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700931 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932
Eric Dumazetb69aee02005-09-06 14:42:45 -0700933 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700934 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900935 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700937 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200939#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700941 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200943#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700944 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 case FIOSETOWN:
946 case SIOCSPGRP:
947 err = -EFAULT;
948 if (get_user(pid, (int __user *)argp))
949 break;
950 err = f_setown(sock->file, pid, 1);
951 break;
952 case FIOGETOWN:
953 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700954 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700955 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 break;
957 case SIOCGIFBR:
958 case SIOCSIFBR:
959 case SIOCBRADDBR:
960 case SIOCBRDELBR:
961 err = -ENOPKG;
962 if (!br_ioctl_hook)
963 request_module("bridge");
964
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800965 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700966 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700967 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800968 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 break;
970 case SIOCGIFVLAN:
971 case SIOCSIFVLAN:
972 err = -ENOPKG;
973 if (!vlan_ioctl_hook)
974 request_module("8021q");
975
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800976 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700978 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800979 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 case SIOCADDDLCI:
982 case SIOCDELDLCI:
983 err = -ENOPKG;
984 if (!dlci_ioctl_hook)
985 request_module("dlci");
986
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700987 mutex_lock(&dlci_ioctl_mutex);
988 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700990 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 break;
992 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -0800993 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700995 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 return err;
997}
998
999int sock_create_lite(int family, int type, int protocol, struct socket **res)
1000{
1001 int err;
1002 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001003
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 err = security_socket_create(family, type, protocol, 1);
1005 if (err)
1006 goto out;
1007
1008 sock = sock_alloc();
1009 if (!sock) {
1010 err = -ENOMEM;
1011 goto out;
1012 }
1013
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001015 err = security_socket_post_create(sock, family, type, protocol, 1);
1016 if (err)
1017 goto out_release;
1018
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019out:
1020 *res = sock;
1021 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001022out_release:
1023 sock_release(sock);
1024 sock = NULL;
1025 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026}
1027
1028/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001029static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030{
1031 struct socket *sock;
1032
1033 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001034 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001036 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 return sock->ops->poll(file, sock, wait);
1038}
1039
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001040static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001042 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
1044 return sock->ops->mmap(file, sock, vma);
1045}
1046
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001047static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048{
1049 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001050 * It was possible the inode is NULL we were
1051 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 */
1053
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001054 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 printk(KERN_DEBUG "sock_close: NULL inode\n");
1056 return 0;
1057 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 sock_release(SOCKET_I(inode));
1059 return 0;
1060}
1061
1062/*
1063 * Update the socket async list
1064 *
1065 * Fasync_list locking strategy.
1066 *
1067 * 1. fasync_list is modified only under process context socket lock
1068 * i.e. under semaphore.
1069 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1070 * or under socket lock.
1071 * 3. fasync_list can be used from softirq context, so that
1072 * modification under socket lock have to be enhanced with
1073 * write_lock_bh(&sk->sk_callback_lock).
1074 * --ANK (990710)
1075 */
1076
1077static int sock_fasync(int fd, struct file *filp, int on)
1078{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001079 struct fasync_struct *fa, *fna = NULL, **prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 struct socket *sock;
1081 struct sock *sk;
1082
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001083 if (on) {
Kris Katterjohn8b3a7002006-01-11 15:56:43 -08001084 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001085 if (fna == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 return -ENOMEM;
1087 }
1088
Eric Dumazetb69aee02005-09-06 14:42:45 -07001089 sock = filp->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001091 sk = sock->sk;
1092 if (sk == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 kfree(fna);
1094 return -EINVAL;
1095 }
1096
1097 lock_sock(sk);
1098
Jonathan Corbet76398422009-02-01 14:26:59 -07001099 spin_lock(&filp->f_lock);
1100 if (on)
1101 filp->f_flags |= FASYNC;
1102 else
1103 filp->f_flags &= ~FASYNC;
1104 spin_unlock(&filp->f_lock);
1105
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001106 prev = &(sock->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001108 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1109 if (fa->fa_file == filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 break;
1111
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001112 if (on) {
1113 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001115 fa->fa_fd = fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 write_unlock_bh(&sk->sk_callback_lock);
1117
1118 kfree(fna);
1119 goto out;
1120 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001121 fna->fa_file = filp;
1122 fna->fa_fd = fd;
1123 fna->magic = FASYNC_MAGIC;
1124 fna->fa_next = sock->fasync_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001126 sock->fasync_list = fna;
Eric Dumazetbcdce712009-10-06 17:28:29 -07001127 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 write_unlock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001129 } else {
1130 if (fa != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 write_lock_bh(&sk->sk_callback_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001132 *prev = fa->fa_next;
Eric Dumazetbcdce712009-10-06 17:28:29 -07001133 if (!sock->fasync_list)
1134 sock_reset_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 write_unlock_bh(&sk->sk_callback_lock);
1136 kfree(fa);
1137 }
1138 }
1139
1140out:
1141 release_sock(sock->sk);
1142 return 0;
1143}
1144
1145/* This function may be called only under socket lock or callback_lock */
1146
1147int sock_wake_async(struct socket *sock, int how, int band)
1148{
1149 if (!sock || !sock->fasync_list)
1150 return -1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001151 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001152 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1154 break;
1155 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001156 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1158 break;
1159 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001160 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001161call_kill:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 __kill_fasync(sock->fasync_list, SIGIO, band);
1163 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001164 case SOCK_WAKE_URG:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 __kill_fasync(sock->fasync_list, SIGURG, band);
1166 }
1167 return 0;
1168}
1169
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001170static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001171 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
1173 int err;
1174 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001175 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176
1177 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001178 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 */
1180 if (family < 0 || family >= NPROTO)
1181 return -EAFNOSUPPORT;
1182 if (type < 0 || type >= SOCK_MAX)
1183 return -EINVAL;
1184
1185 /* Compatibility.
1186
1187 This uglymoron is moved from INET layer to here to avoid
1188 deadlock in module load.
1189 */
1190 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001191 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 if (!warned) {
1193 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001194 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1195 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 }
1197 family = PF_PACKET;
1198 }
1199
1200 err = security_socket_create(family, type, protocol, kern);
1201 if (err)
1202 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001203
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001204 /*
1205 * Allocate the socket and allow the family to set things up. if
1206 * the protocol is 0, the family is instructed to select an appropriate
1207 * default.
1208 */
1209 sock = sock_alloc();
1210 if (!sock) {
1211 if (net_ratelimit())
1212 printk(KERN_WARNING "socket: no more sockets\n");
1213 return -ENFILE; /* Not exactly a match, but its the
1214 closest posix thing */
1215 }
1216
1217 sock->type = type;
1218
Johannes Berg95a5afc2008-10-16 15:24:51 -07001219#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001220 /* Attempt to load a protocol module if the find failed.
1221 *
1222 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 * requested real, full-featured networking support upon configuration.
1224 * Otherwise module support will break!
1225 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001226 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001227 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228#endif
1229
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001230 rcu_read_lock();
1231 pf = rcu_dereference(net_families[family]);
1232 err = -EAFNOSUPPORT;
1233 if (!pf)
1234 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
1236 /*
1237 * We will call the ->create function, that possibly is in a loadable
1238 * module, so we have to bump that loadable module refcnt first.
1239 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001240 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 goto out_release;
1242
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001243 /* Now protected by module ref count */
1244 rcu_read_unlock();
1245
Eric Paris3f378b62009-11-05 22:18:14 -08001246 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001247 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001249
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 /*
1251 * Now to bump the refcnt of the [loadable] module that owns this
1252 * socket at sock_release time we decrement its refcnt.
1253 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001254 if (!try_module_get(sock->ops->owner))
1255 goto out_module_busy;
1256
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 /*
1258 * Now that we're done with the ->create function, the [loadable]
1259 * module can have its refcnt decremented
1260 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001261 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001262 err = security_socket_post_create(sock, family, type, protocol, kern);
1263 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001264 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001265 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001267 return 0;
1268
1269out_module_busy:
1270 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001272 sock->ops = NULL;
1273 module_put(pf->owner);
1274out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001276 return err;
1277
1278out_release:
1279 rcu_read_unlock();
1280 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281}
1282
1283int sock_create(int family, int type, int protocol, struct socket **res)
1284{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001285 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286}
1287
1288int sock_create_kern(int family, int type, int protocol, struct socket **res)
1289{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001290 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291}
1292
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001293SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294{
1295 int retval;
1296 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001297 int flags;
1298
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001299 /* Check the SOCK_* constants for consistency. */
1300 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1301 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1302 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1303 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1304
Ulrich Dreppera677a032008-07-23 21:29:17 -07001305 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001306 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001307 return -EINVAL;
1308 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001310 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1311 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 retval = sock_create(family, type, protocol, &sock);
1314 if (retval < 0)
1315 goto out;
1316
Ulrich Drepper77d27202008-07-23 21:29:35 -07001317 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 if (retval < 0)
1319 goto out_release;
1320
1321out:
1322 /* It may be already another descriptor 8) Not kernel problem. */
1323 return retval;
1324
1325out_release:
1326 sock_release(sock);
1327 return retval;
1328}
1329
1330/*
1331 * Create a pair of connected sockets.
1332 */
1333
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001334SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1335 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336{
1337 struct socket *sock1, *sock2;
1338 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001339 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001340 int flags;
1341
1342 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001343 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001344 return -EINVAL;
1345 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001347 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1348 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 /*
1351 * Obtain the first socket and check if the underlying protocol
1352 * supports the socketpair call.
1353 */
1354
1355 err = sock_create(family, type, protocol, &sock1);
1356 if (err < 0)
1357 goto out;
1358
1359 err = sock_create(family, type, protocol, &sock2);
1360 if (err < 0)
1361 goto out_release_1;
1362
1363 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001364 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 goto out_release_both;
1366
Al Viro7cbe66b2009-08-05 19:59:08 +04001367 fd1 = sock_alloc_file(sock1, &newfile1, flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001368 if (unlikely(fd1 < 0)) {
1369 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001371 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
Al Viro7cbe66b2009-08-05 19:59:08 +04001373 fd2 = sock_alloc_file(sock2, &newfile2, flags);
Al Viro198de4d2009-08-05 19:29:23 +04001374 if (unlikely(fd2 < 0)) {
1375 err = fd2;
1376 fput(newfile1);
1377 put_unused_fd(fd1);
1378 sock_release(sock2);
1379 goto out;
Al Virodb349502007-02-07 01:48:00 -05001380 }
1381
Al Viro157cf642008-12-14 04:57:47 -05001382 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001383 fd_install(fd1, newfile1);
1384 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 /* fd1 and fd2 may be already another descriptors.
1386 * Not kernel problem.
1387 */
1388
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001389 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 if (!err)
1391 err = put_user(fd2, &usockvec[1]);
1392 if (!err)
1393 return 0;
1394
1395 sys_close(fd2);
1396 sys_close(fd1);
1397 return err;
1398
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001400 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001402 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403out:
1404 return err;
1405}
1406
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407/*
1408 * Bind a name to a socket. Nothing much to do here since it's
1409 * the protocol's responsibility to handle the local address.
1410 *
1411 * We move the socket address to kernel space before we call
1412 * the protocol layer (having also checked the address is ok).
1413 */
1414
Heiko Carstens20f37032009-01-14 14:14:23 +01001415SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416{
1417 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001418 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001419 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001421 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001422 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001423 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001424 if (err >= 0) {
1425 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001426 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001427 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001428 if (!err)
1429 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001430 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001431 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001433 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001434 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 return err;
1436}
1437
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438/*
1439 * Perform a listen. Basically, we allow the protocol to do anything
1440 * necessary for a listen, and if that works, we mark the socket as
1441 * ready for listening.
1442 */
1443
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001444SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445{
1446 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001447 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001448 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001449
1450 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1451 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001452 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001453 if ((unsigned)backlog > somaxconn)
1454 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455
1456 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001457 if (!err)
1458 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001460 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 }
1462 return err;
1463}
1464
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465/*
1466 * For accept, we attempt to create a new socket, set up the link
1467 * with the client, wake up the client, then return the new
1468 * connected fd. We collect the address of the connector in kernel
1469 * space and move it to user at the very end. This is unclean because
1470 * we open the socket then return an error.
1471 *
1472 * 1003.1g adds the ability to recvmsg() to query connection pending
1473 * status to recvmsg. We need to add that support in a way thats
1474 * clean when we restucture accept also.
1475 */
1476
Heiko Carstens20f37032009-01-14 14:14:23 +01001477SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1478 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479{
1480 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001481 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001482 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001483 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484
Ulrich Drepper77d27202008-07-23 21:29:35 -07001485 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001486 return -EINVAL;
1487
1488 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1489 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1490
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001491 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 if (!sock)
1493 goto out;
1494
1495 err = -ENFILE;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001496 if (!(newsock = sock_alloc()))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 goto out_put;
1498
1499 newsock->type = sock->type;
1500 newsock->ops = sock->ops;
1501
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 /*
1503 * We don't need try_module_get here, as the listening socket (sock)
1504 * has the protocol module (sock->ops->owner) held.
1505 */
1506 __module_get(newsock->ops->owner);
1507
Al Viro7cbe66b2009-08-05 19:59:08 +04001508 newfd = sock_alloc_file(newsock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001509 if (unlikely(newfd < 0)) {
1510 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001511 sock_release(newsock);
1512 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001513 }
1514
Frank Filza79af592005-09-27 15:23:38 -07001515 err = security_socket_accept(sock, newsock);
1516 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001517 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001518
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1520 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001521 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
1523 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001524 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001525 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001527 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001529 err = move_addr_to_user((struct sockaddr *)&address,
1530 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001532 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 }
1534
1535 /* File flags are not inherited via accept() unlike another OSes. */
1536
David S. Miller39d8c1b2006-03-20 17:13:49 -08001537 fd_install(newfd, newfile);
1538 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001541 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542out:
1543 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001544out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001545 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001546 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547 goto out_put;
1548}
1549
Heiko Carstens20f37032009-01-14 14:14:23 +01001550SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1551 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001552{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001553 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001554}
1555
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556/*
1557 * Attempt to connect to a socket with the server address. The address
1558 * is in user space so we verify it is OK and move it to kernel space.
1559 *
1560 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1561 * break bindings
1562 *
1563 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1564 * other SEQPACKET protocols that take time to connect() as it doesn't
1565 * include the -EINPROGRESS status for such sockets.
1566 */
1567
Heiko Carstens20f37032009-01-14 14:14:23 +01001568SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1569 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570{
1571 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001572 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001573 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001575 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 if (!sock)
1577 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001578 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579 if (err < 0)
1580 goto out_put;
1581
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001582 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001583 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 if (err)
1585 goto out_put;
1586
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001587 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 sock->file->f_flags);
1589out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001590 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591out:
1592 return err;
1593}
1594
1595/*
1596 * Get the local address ('name') of a socket object. Move the obtained
1597 * name to user space.
1598 */
1599
Heiko Carstens20f37032009-01-14 14:14:23 +01001600SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1601 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602{
1603 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001604 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001605 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001606
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001607 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 if (!sock)
1609 goto out;
1610
1611 err = security_socket_getsockname(sock);
1612 if (err)
1613 goto out_put;
1614
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001615 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 if (err)
1617 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001618 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619
1620out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001621 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622out:
1623 return err;
1624}
1625
1626/*
1627 * Get the remote address ('name') of a socket object. Move the obtained
1628 * name to user space.
1629 */
1630
Heiko Carstens20f37032009-01-14 14:14:23 +01001631SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1632 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633{
1634 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001635 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001636 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001638 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1639 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 err = security_socket_getpeername(sock);
1641 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001642 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 return err;
1644 }
1645
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001646 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001647 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001648 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001650 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001651 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001652 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 }
1654 return err;
1655}
1656
1657/*
1658 * Send a datagram to a given address. We move the address into kernel
1659 * space and check the user space data area is readable before invoking
1660 * the protocol.
1661 */
1662
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001663SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1664 unsigned, flags, struct sockaddr __user *, addr,
1665 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666{
1667 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001668 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 int err;
1670 struct msghdr msg;
1671 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001672 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001673
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001674 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1675 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001676 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001677
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001678 iov.iov_base = buff;
1679 iov.iov_len = len;
1680 msg.msg_name = NULL;
1681 msg.msg_iov = &iov;
1682 msg.msg_iovlen = 1;
1683 msg.msg_control = NULL;
1684 msg.msg_controllen = 0;
1685 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001686 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001687 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 if (err < 0)
1689 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001690 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001691 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 }
1693 if (sock->file->f_flags & O_NONBLOCK)
1694 flags |= MSG_DONTWAIT;
1695 msg.msg_flags = flags;
1696 err = sock_sendmsg(sock, &msg, len);
1697
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001698out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001699 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001700out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 return err;
1702}
1703
1704/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001705 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 */
1707
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001708SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1709 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710{
1711 return sys_sendto(fd, buff, len, flags, NULL, 0);
1712}
1713
1714/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001715 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 * sender. We verify the buffers are writable and if needed move the
1717 * sender address from kernel to user space.
1718 */
1719
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001720SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1721 unsigned, flags, struct sockaddr __user *, addr,
1722 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723{
1724 struct socket *sock;
1725 struct iovec iov;
1726 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001727 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001728 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001729 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001731 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001733 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001735 msg.msg_control = NULL;
1736 msg.msg_controllen = 0;
1737 msg.msg_iovlen = 1;
1738 msg.msg_iov = &iov;
1739 iov.iov_len = size;
1740 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001741 msg.msg_name = (struct sockaddr *)&address;
1742 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 if (sock->file->f_flags & O_NONBLOCK)
1744 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001745 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001747 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001748 err2 = move_addr_to_user((struct sockaddr *)&address,
1749 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001750 if (err2 < 0)
1751 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001753
1754 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 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 */
1762
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001763asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1764 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765{
1766 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1767}
1768
1769/*
1770 * Set a socket option. Because we don't know the option lengths we have
1771 * to pass the user mode parameter for the protocols to sort out.
1772 */
1773
Heiko Carstens20f37032009-01-14 14:14:23 +01001774SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1775 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001777 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 struct socket *sock;
1779
1780 if (optlen < 0)
1781 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001782
1783 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1784 if (sock != NULL) {
1785 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001786 if (err)
1787 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788
1789 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001790 err =
1791 sock_setsockopt(sock, level, optname, optval,
1792 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001794 err =
1795 sock->ops->setsockopt(sock, level, optname, optval,
1796 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001797out_put:
1798 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 }
1800 return err;
1801}
1802
1803/*
1804 * Get a socket option. Because we don't know the option lengths we have
1805 * to pass a user mode parameter for the protocols to sort out.
1806 */
1807
Heiko Carstens20f37032009-01-14 14:14:23 +01001808SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1809 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001811 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 struct socket *sock;
1813
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001814 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1815 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001816 err = security_socket_getsockopt(sock, level, optname);
1817 if (err)
1818 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819
1820 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001821 err =
1822 sock_getsockopt(sock, level, optname, optval,
1823 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001825 err =
1826 sock->ops->getsockopt(sock, level, optname, optval,
1827 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001828out_put:
1829 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 }
1831 return err;
1832}
1833
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834/*
1835 * Shutdown a socket.
1836 */
1837
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001838SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001840 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841 struct socket *sock;
1842
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001843 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1844 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001846 if (!err)
1847 err = sock->ops->shutdown(sock, how);
1848 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 }
1850 return err;
1851}
1852
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001853/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 * fields which are the same type (int / unsigned) on our platforms.
1855 */
1856#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1857#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1858#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1859
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860/*
1861 * BSD sendmsg interface
1862 */
1863
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001864SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001866 struct compat_msghdr __user *msg_compat =
1867 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001869 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001871 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001872 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1873 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 unsigned char *ctl_buf = ctl;
1875 struct msghdr msg_sys;
1876 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001877 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001878
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 err = -EFAULT;
1880 if (MSG_CMSG_COMPAT & flags) {
1881 if (get_compat_msghdr(&msg_sys, msg_compat))
1882 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001883 }
1884 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 return -EFAULT;
1886
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001887 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001888 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 goto out;
1890
1891 /* do not move before msg_sys is valid */
1892 err = -EMSGSIZE;
1893 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1894 goto out_put;
1895
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001896 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 err = -ENOMEM;
1898 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1899 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1900 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1901 if (!iov)
1902 goto out_put;
1903 }
1904
1905 /* This will also move the address data into kernel space */
1906 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001907 err = verify_compat_iovec(&msg_sys, iov,
1908 (struct sockaddr *)&address,
1909 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001911 err = verify_iovec(&msg_sys, iov,
1912 (struct sockaddr *)&address,
1913 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001914 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 goto out_freeiov;
1916 total_len = err;
1917
1918 err = -ENOBUFS;
1919
1920 if (msg_sys.msg_controllen > INT_MAX)
1921 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001922 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001924 err =
1925 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1926 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 if (err)
1928 goto out_freeiov;
1929 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001930 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001932 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001934 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 goto out_freeiov;
1936 }
1937 err = -EFAULT;
1938 /*
1939 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1940 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1941 * checking falls down on this.
1942 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001943 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1944 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 goto out_freectl;
1946 msg_sys.msg_control = ctl_buf;
1947 }
1948 msg_sys.msg_flags = flags;
1949
1950 if (sock->file->f_flags & O_NONBLOCK)
1951 msg_sys.msg_flags |= MSG_DONTWAIT;
1952 err = sock_sendmsg(sock, &msg_sys, total_len);
1953
1954out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001955 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1957out_freeiov:
1958 if (iov != iovstack)
1959 sock_kfree_s(sock->sk, iov, iov_size);
1960out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001961 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001962out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 return err;
1964}
1965
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001966static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1967 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001969 struct compat_msghdr __user *msg_compat =
1970 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001972 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 unsigned long cmsg_ptr;
1974 int err, iov_size, total_len, len;
1975
1976 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001977 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978
1979 /* user mode address pointers */
1980 struct sockaddr __user *uaddr;
1981 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001982
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001984 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 return -EFAULT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001986 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001987 else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001988 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001991 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1992 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001993
1994 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001996 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1997 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1999 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002000 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 }
2002
2003 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002004 * Save the user-mode address (verify_iovec will change the
2005 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002007
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002008 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 uaddr_len = COMPAT_NAMELEN(msg);
2010 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002011 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002012 (struct sockaddr *)&addr,
2013 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 } else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002015 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002016 (struct sockaddr *)&addr,
2017 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 if (err < 0)
2019 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002022 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2023 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 if (sock->file->f_flags & O_NONBLOCK)
2026 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002027 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2028 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 if (err < 0)
2030 goto out_freeiov;
2031 len = err;
2032
2033 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002034 err = move_addr_to_user((struct sockaddr *)&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002035 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002036 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 if (err < 0)
2038 goto out_freeiov;
2039 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002040 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002041 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 if (err)
2043 goto out_freeiov;
2044 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002045 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 &msg_compat->msg_controllen);
2047 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002048 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 &msg->msg_controllen);
2050 if (err)
2051 goto out_freeiov;
2052 err = len;
2053
2054out_freeiov:
2055 if (iov != iovstack)
2056 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002057out:
2058 return err;
2059}
2060
2061/*
2062 * BSD recvmsg interface
2063 */
2064
2065SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2066 unsigned int, flags)
2067{
2068 int fput_needed, err;
2069 struct msghdr msg_sys;
2070 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2071
2072 if (!sock)
2073 goto out;
2074
2075 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2076
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002077 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078out:
2079 return err;
2080}
2081
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002082/*
2083 * Linux recvmmsg interface
2084 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002086int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2087 unsigned int flags, struct timespec *timeout)
2088{
2089 int fput_needed, err, datagrams;
2090 struct socket *sock;
2091 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002092 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002093 struct msghdr msg_sys;
2094 struct timespec end_time;
2095
2096 if (timeout &&
2097 poll_select_set_timeout(&end_time, timeout->tv_sec,
2098 timeout->tv_nsec))
2099 return -EINVAL;
2100
2101 datagrams = 0;
2102
2103 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2104 if (!sock)
2105 return err;
2106
2107 err = sock_error(sock->sk);
2108 if (err)
2109 goto out_put;
2110
2111 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002112 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002113
2114 while (datagrams < vlen) {
2115 /*
2116 * No need to ask LSM for more than the first datagram.
2117 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002118 if (MSG_CMSG_COMPAT & flags) {
2119 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2120 &msg_sys, flags, datagrams);
2121 if (err < 0)
2122 break;
2123 err = __put_user(err, &compat_entry->msg_len);
2124 ++compat_entry;
2125 } else {
2126 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2127 &msg_sys, flags, datagrams);
2128 if (err < 0)
2129 break;
2130 err = put_user(err, &entry->msg_len);
2131 ++entry;
2132 }
2133
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002134 if (err)
2135 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002136 ++datagrams;
2137
2138 if (timeout) {
2139 ktime_get_ts(timeout);
2140 *timeout = timespec_sub(end_time, *timeout);
2141 if (timeout->tv_sec < 0) {
2142 timeout->tv_sec = timeout->tv_nsec = 0;
2143 break;
2144 }
2145
2146 /* Timeout, return less than vlen datagrams */
2147 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2148 break;
2149 }
2150
2151 /* Out of band data, return right away */
2152 if (msg_sys.msg_flags & MSG_OOB)
2153 break;
2154 }
2155
2156out_put:
2157 fput_light(sock->file, fput_needed);
2158
2159 if (err == 0)
2160 return datagrams;
2161
2162 if (datagrams != 0) {
2163 /*
2164 * We may return less entries than requested (vlen) if the
2165 * sock is non block and there aren't enough datagrams...
2166 */
2167 if (err != -EAGAIN) {
2168 /*
2169 * ... or if recvmsg returns an error after we
2170 * received some datagrams, where we record the
2171 * error to return on the next call or if the
2172 * app asks about it using getsockopt(SO_ERROR).
2173 */
2174 sock->sk->sk_err = -err;
2175 }
2176
2177 return datagrams;
2178 }
2179
2180 return err;
2181}
2182
2183SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2184 unsigned int, vlen, unsigned int, flags,
2185 struct timespec __user *, timeout)
2186{
2187 int datagrams;
2188 struct timespec timeout_sys;
2189
2190 if (!timeout)
2191 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2192
2193 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2194 return -EFAULT;
2195
2196 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2197
2198 if (datagrams > 0 &&
2199 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2200 datagrams = -EFAULT;
2201
2202 return datagrams;
2203}
2204
2205#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206/* Argument list sizes for sys_socketcall */
2207#define AL(x) ((x) * sizeof(unsigned long))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002208static const unsigned char nargs[20] = {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002209 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2210 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002211 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002212 AL(4),AL(5)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002213};
2214
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215#undef AL
2216
2217/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002218 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 *
2220 * Argument checking cleaned up. Saved 20% in size.
2221 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002222 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 */
2224
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002225SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226{
2227 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002228 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002230 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002232 if (call < 1 || call > SYS_RECVMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 return -EINVAL;
2234
Arjan van de Ven47379052009-09-28 12:57:44 -07002235 len = nargs[call];
2236 if (len > sizeof(a))
2237 return -EINVAL;
2238
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002240 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002242
Al Virof3298dc2008-12-10 03:16:51 -05002243 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002244
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002245 a0 = a[0];
2246 a1 = a[1];
2247
2248 switch (call) {
2249 case SYS_SOCKET:
2250 err = sys_socket(a0, a1, a[2]);
2251 break;
2252 case SYS_BIND:
2253 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2254 break;
2255 case SYS_CONNECT:
2256 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2257 break;
2258 case SYS_LISTEN:
2259 err = sys_listen(a0, a1);
2260 break;
2261 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002262 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2263 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002264 break;
2265 case SYS_GETSOCKNAME:
2266 err =
2267 sys_getsockname(a0, (struct sockaddr __user *)a1,
2268 (int __user *)a[2]);
2269 break;
2270 case SYS_GETPEERNAME:
2271 err =
2272 sys_getpeername(a0, (struct sockaddr __user *)a1,
2273 (int __user *)a[2]);
2274 break;
2275 case SYS_SOCKETPAIR:
2276 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2277 break;
2278 case SYS_SEND:
2279 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2280 break;
2281 case SYS_SENDTO:
2282 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2283 (struct sockaddr __user *)a[4], a[5]);
2284 break;
2285 case SYS_RECV:
2286 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2287 break;
2288 case SYS_RECVFROM:
2289 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2290 (struct sockaddr __user *)a[4],
2291 (int __user *)a[5]);
2292 break;
2293 case SYS_SHUTDOWN:
2294 err = sys_shutdown(a0, a1);
2295 break;
2296 case SYS_SETSOCKOPT:
2297 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2298 break;
2299 case SYS_GETSOCKOPT:
2300 err =
2301 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2302 (int __user *)a[4]);
2303 break;
2304 case SYS_SENDMSG:
2305 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2306 break;
2307 case SYS_RECVMSG:
2308 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2309 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002310 case SYS_RECVMMSG:
2311 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2312 (struct timespec __user *)a[4]);
2313 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002314 case SYS_ACCEPT4:
2315 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2316 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002317 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002318 default:
2319 err = -EINVAL;
2320 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321 }
2322 return err;
2323}
2324
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002325#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002327/**
2328 * sock_register - add a socket protocol handler
2329 * @ops: description of protocol
2330 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331 * This function is called by a protocol handler that wants to
2332 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002333 * socket interface. The value ops->family coresponds to the
2334 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002336int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337{
2338 int err;
2339
2340 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002341 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2342 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 return -ENOBUFS;
2344 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002345
2346 spin_lock(&net_family_lock);
2347 if (net_families[ops->family])
2348 err = -EEXIST;
2349 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002350 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 err = 0;
2352 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002353 spin_unlock(&net_family_lock);
2354
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002355 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356 return err;
2357}
2358
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002359/**
2360 * sock_unregister - remove a protocol handler
2361 * @family: protocol family to remove
2362 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363 * This function is called by a protocol handler that wants to
2364 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002365 * new socket creation.
2366 *
2367 * If protocol handler is a module, then it can use module reference
2368 * counts to protect against new references. If protocol handler is not
2369 * a module then it needs to provide its own protection in
2370 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002372void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002374 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002376 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002377 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002378 spin_unlock(&net_family_lock);
2379
2380 synchronize_rcu();
2381
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002382 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383}
2384
Andi Kleen77d76ea2005-12-22 12:43:42 -08002385static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386{
2387 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002388 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002390
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391 sk_init();
2392
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002394 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395 */
2396 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397
2398 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002399 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 */
2401
2402 init_inodecache();
2403 register_filesystem(&sock_fs_type);
2404 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002405
2406 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407 */
2408
2409#ifdef CONFIG_NETFILTER
2410 netfilter_init();
2411#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002412
2413 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414}
2415
Andi Kleen77d76ea2005-12-22 12:43:42 -08002416core_initcall(sock_init); /* early initcall */
2417
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418#ifdef CONFIG_PROC_FS
2419void socket_seq_show(struct seq_file *seq)
2420{
2421 int cpu;
2422 int counter = 0;
2423
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002424 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002425 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426
2427 /* It can be negative, by the way. 8) */
2428 if (counter < 0)
2429 counter = 0;
2430
2431 seq_printf(seq, "sockets: used %d\n", counter);
2432}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002433#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002435#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002436static int do_siocgstamp(struct net *net, struct socket *sock,
2437 unsigned int cmd, struct compat_timeval __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002438{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002439 mm_segment_t old_fs = get_fs();
2440 struct timeval ktv;
2441 int err;
2442
2443 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002444 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002445 set_fs(old_fs);
2446 if (!err) {
2447 err = put_user(ktv.tv_sec, &up->tv_sec);
2448 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2449 }
2450 return err;
2451}
2452
Arnd Bergmann6b960182009-11-06 23:10:54 -08002453static int do_siocgstampns(struct net *net, struct socket *sock,
2454 unsigned int cmd, struct compat_timespec __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002455{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002456 mm_segment_t old_fs = get_fs();
2457 struct timespec kts;
2458 int err;
2459
2460 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002461 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002462 set_fs(old_fs);
2463 if (!err) {
2464 err = put_user(kts.tv_sec, &up->tv_sec);
2465 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2466 }
2467 return err;
2468}
2469
Arnd Bergmann6b960182009-11-06 23:10:54 -08002470static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002471{
2472 struct ifreq __user *uifr;
2473 int err;
2474
2475 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002476 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002477 return -EFAULT;
2478
Arnd Bergmann6b960182009-11-06 23:10:54 -08002479 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002480 if (err)
2481 return err;
2482
Arnd Bergmann6b960182009-11-06 23:10:54 -08002483 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002484 return -EFAULT;
2485
2486 return 0;
2487}
2488
Arnd Bergmann6b960182009-11-06 23:10:54 -08002489static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002490{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002491 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002492 struct ifconf ifc;
2493 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002494 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002495 struct ifreq __user *ifr;
2496 unsigned int i, j;
2497 int err;
2498
Arnd Bergmann6b960182009-11-06 23:10:54 -08002499 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002500 return -EFAULT;
2501
2502 if (ifc32.ifcbuf == 0) {
2503 ifc32.ifc_len = 0;
2504 ifc.ifc_len = 0;
2505 ifc.ifc_req = NULL;
2506 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2507 } else {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002508 size_t len =((ifc32.ifc_len / sizeof (struct compat_ifreq)) + 1) *
Arnd Bergmann7a229382009-11-06 23:00:29 -08002509 sizeof (struct ifreq);
2510 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2511 ifc.ifc_len = len;
2512 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2513 ifr32 = compat_ptr(ifc32.ifcbuf);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002514 for (i = 0; i < ifc32.ifc_len; i += sizeof (struct compat_ifreq)) {
2515 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002516 return -EFAULT;
2517 ifr++;
2518 ifr32++;
2519 }
2520 }
2521 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2522 return -EFAULT;
2523
Arnd Bergmann6b960182009-11-06 23:10:54 -08002524 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002525 if (err)
2526 return err;
2527
2528 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2529 return -EFAULT;
2530
2531 ifr = ifc.ifc_req;
2532 ifr32 = compat_ptr(ifc32.ifcbuf);
2533 for (i = 0, j = 0;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002534 i + sizeof (struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2535 i += sizeof (struct compat_ifreq), j += sizeof (struct ifreq)) {
2536 if (copy_in_user(ifr32, ifr, sizeof (struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002537 return -EFAULT;
2538 ifr32++;
2539 ifr++;
2540 }
2541
2542 if (ifc32.ifcbuf == 0) {
2543 /* Translate from 64-bit structure multiple to
2544 * a 32-bit one.
2545 */
2546 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002547 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002548 ifc32.ifc_len = i;
2549 } else {
2550 ifc32.ifc_len = i;
2551 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002552 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002553 return -EFAULT;
2554
2555 return 0;
2556}
2557
Arnd Bergmann6b960182009-11-06 23:10:54 -08002558static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002559{
2560 struct ifreq __user *ifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002561 u32 data;
2562 void __user *datap;
2563
2564 ifr = compat_alloc_user_space(sizeof(*ifr));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002565
2566 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2567 return -EFAULT;
2568
2569 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2570 return -EFAULT;
2571
2572 datap = compat_ptr(data);
2573 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2574 return -EFAULT;
2575
Arnd Bergmann6b960182009-11-06 23:10:54 -08002576 return dev_ioctl(net, SIOCETHTOOL, ifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002577}
2578
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002579static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2580{
2581 void __user *uptr;
2582 compat_uptr_t uptr32;
2583 struct ifreq __user *uifr;
2584
2585 uifr = compat_alloc_user_space(sizeof (*uifr));
2586 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2587 return -EFAULT;
2588
2589 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2590 return -EFAULT;
2591
2592 uptr = compat_ptr(uptr32);
2593
2594 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2595 return -EFAULT;
2596
2597 return dev_ioctl(net, SIOCWANDEV, uifr);
2598}
2599
Arnd Bergmann6b960182009-11-06 23:10:54 -08002600static int bond_ioctl(struct net *net, unsigned int cmd,
2601 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002602{
2603 struct ifreq kifr;
2604 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002605 mm_segment_t old_fs;
2606 int err;
2607 u32 data;
2608 void __user *datap;
2609
2610 switch (cmd) {
2611 case SIOCBONDENSLAVE:
2612 case SIOCBONDRELEASE:
2613 case SIOCBONDSETHWADDR:
2614 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002615 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002616 return -EFAULT;
2617
2618 old_fs = get_fs();
2619 set_fs (KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002620 err = dev_ioctl(net, cmd, &kifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002621 set_fs (old_fs);
2622
2623 return err;
2624 case SIOCBONDSLAVEINFOQUERY:
2625 case SIOCBONDINFOQUERY:
2626 uifr = compat_alloc_user_space(sizeof(*uifr));
2627 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2628 return -EFAULT;
2629
2630 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2631 return -EFAULT;
2632
2633 datap = compat_ptr(data);
2634 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2635 return -EFAULT;
2636
Arnd Bergmann6b960182009-11-06 23:10:54 -08002637 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002638 default:
2639 return -EINVAL;
2640 };
2641}
2642
Arnd Bergmann6b960182009-11-06 23:10:54 -08002643static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2644 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002645{
2646 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002647 char tmp_buf[IFNAMSIZ];
2648 void __user *data64;
2649 u32 data32;
2650
2651 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2652 IFNAMSIZ))
2653 return -EFAULT;
2654 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2655 return -EFAULT;
2656 data64 = compat_ptr(data32);
2657
2658 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2659
2660 /* Don't check these user accesses, just let that get trapped
2661 * in the ioctl handler instead.
2662 */
2663 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2664 IFNAMSIZ))
2665 return -EFAULT;
2666 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2667 return -EFAULT;
2668
Arnd Bergmann6b960182009-11-06 23:10:54 -08002669 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002670}
2671
Arnd Bergmann6b960182009-11-06 23:10:54 -08002672static int dev_ifsioc(struct net *net, struct socket *sock,
2673 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002674{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002675 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002676 int err;
2677
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002678 uifr = compat_alloc_user_space(sizeof(*uifr));
2679 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2680 return -EFAULT;
2681
2682 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2683
Arnd Bergmann7a229382009-11-06 23:00:29 -08002684 if (!err) {
2685 switch (cmd) {
2686 case SIOCGIFFLAGS:
2687 case SIOCGIFMETRIC:
2688 case SIOCGIFMTU:
2689 case SIOCGIFMEM:
2690 case SIOCGIFHWADDR:
2691 case SIOCGIFINDEX:
2692 case SIOCGIFADDR:
2693 case SIOCGIFBRDADDR:
2694 case SIOCGIFDSTADDR:
2695 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002696 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002697 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002698 case SIOCGMIIPHY:
2699 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002700 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002701 err = -EFAULT;
2702 break;
2703 }
2704 }
2705 return err;
2706}
2707
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002708static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2709 struct compat_ifreq __user *uifr32)
2710{
2711 struct ifreq ifr;
2712 struct compat_ifmap __user *uifmap32;
2713 mm_segment_t old_fs;
2714 int err;
2715
2716 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2717 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2718 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2719 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2720 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2721 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2722 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2723 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2724 if (err)
2725 return -EFAULT;
2726
2727 old_fs = get_fs();
2728 set_fs (KERNEL_DS);
2729 err = dev_ioctl(net, cmd, (void __user *)&ifr);
2730 set_fs (old_fs);
2731
2732 if (cmd == SIOCGIFMAP && !err) {
2733 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2734 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2735 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2736 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2737 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2738 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2739 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2740 if (err)
2741 err = -EFAULT;
2742 }
2743 return err;
2744}
2745
2746static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2747{
2748 void __user *uptr;
2749 compat_uptr_t uptr32;
2750 struct ifreq __user *uifr;
2751
2752 uifr = compat_alloc_user_space(sizeof (*uifr));
2753 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2754 return -EFAULT;
2755
2756 if (get_user(uptr32, &uifr32->ifr_data))
2757 return -EFAULT;
2758
2759 uptr = compat_ptr(uptr32);
2760
2761 if (put_user(uptr, &uifr->ifr_data))
2762 return -EFAULT;
2763
2764 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2765}
2766
Arnd Bergmann7a229382009-11-06 23:00:29 -08002767struct rtentry32 {
2768 u32 rt_pad1;
2769 struct sockaddr rt_dst; /* target address */
2770 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2771 struct sockaddr rt_genmask; /* target network mask (IP) */
2772 unsigned short rt_flags;
2773 short rt_pad2;
2774 u32 rt_pad3;
2775 unsigned char rt_tos;
2776 unsigned char rt_class;
2777 short rt_pad4;
2778 short rt_metric; /* +1 for binary compatibility! */
2779 /* char * */ u32 rt_dev; /* forcing the device at add */
2780 u32 rt_mtu; /* per route MTU/Window */
2781 u32 rt_window; /* Window clamping */
2782 unsigned short rt_irtt; /* Initial RTT */
2783};
2784
2785struct in6_rtmsg32 {
2786 struct in6_addr rtmsg_dst;
2787 struct in6_addr rtmsg_src;
2788 struct in6_addr rtmsg_gateway;
2789 u32 rtmsg_type;
2790 u16 rtmsg_dst_len;
2791 u16 rtmsg_src_len;
2792 u32 rtmsg_metric;
2793 u32 rtmsg_info;
2794 u32 rtmsg_flags;
2795 s32 rtmsg_ifindex;
2796};
2797
Arnd Bergmann6b960182009-11-06 23:10:54 -08002798static int routing_ioctl(struct net *net, struct socket *sock,
2799 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002800{
2801 int ret;
2802 void *r = NULL;
2803 struct in6_rtmsg r6;
2804 struct rtentry r4;
2805 char devname[16];
2806 u32 rtdev;
2807 mm_segment_t old_fs = get_fs();
2808
Arnd Bergmann6b960182009-11-06 23:10:54 -08002809 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2810 struct in6_rtmsg32 __user *ur6 = argp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002811 ret = copy_from_user (&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2812 3 * sizeof(struct in6_addr));
2813 ret |= __get_user (r6.rtmsg_type, &(ur6->rtmsg_type));
2814 ret |= __get_user (r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2815 ret |= __get_user (r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2816 ret |= __get_user (r6.rtmsg_metric, &(ur6->rtmsg_metric));
2817 ret |= __get_user (r6.rtmsg_info, &(ur6->rtmsg_info));
2818 ret |= __get_user (r6.rtmsg_flags, &(ur6->rtmsg_flags));
2819 ret |= __get_user (r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2820
2821 r = (void *) &r6;
2822 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002823 struct rtentry32 __user *ur4 = argp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002824 ret = copy_from_user (&r4.rt_dst, &(ur4->rt_dst),
2825 3 * sizeof(struct sockaddr));
2826 ret |= __get_user (r4.rt_flags, &(ur4->rt_flags));
2827 ret |= __get_user (r4.rt_metric, &(ur4->rt_metric));
2828 ret |= __get_user (r4.rt_mtu, &(ur4->rt_mtu));
2829 ret |= __get_user (r4.rt_window, &(ur4->rt_window));
2830 ret |= __get_user (r4.rt_irtt, &(ur4->rt_irtt));
2831 ret |= __get_user (rtdev, &(ur4->rt_dev));
2832 if (rtdev) {
2833 ret |= copy_from_user (devname, compat_ptr(rtdev), 15);
2834 r4.rt_dev = devname; devname[15] = 0;
2835 } else
2836 r4.rt_dev = NULL;
2837
2838 r = (void *) &r4;
2839 }
2840
2841 if (ret) {
2842 ret = -EFAULT;
2843 goto out;
2844 }
2845
2846 set_fs (KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002847 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002848 set_fs (old_fs);
2849
2850out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002851 return ret;
2852}
2853
2854/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2855 * for some operations; this forces use of the newer bridge-utils that
2856 * use compatiable ioctls
2857 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002858static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002859{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002860 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002861
Arnd Bergmann6b960182009-11-06 23:10:54 -08002862 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002863 return -EFAULT;
2864 if (tmp == BRCTL_GET_VERSION)
2865 return BRCTL_VERSION + 1;
2866 return -EINVAL;
2867}
2868
Arnd Bergmann6b960182009-11-06 23:10:54 -08002869static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2870 unsigned int cmd, unsigned long arg)
2871{
2872 void __user *argp = compat_ptr(arg);
2873 struct sock *sk = sock->sk;
2874 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002875
Arnd Bergmann6b960182009-11-06 23:10:54 -08002876 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2877 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002878
Arnd Bergmann6b960182009-11-06 23:10:54 -08002879 switch (cmd) {
2880 case SIOCSIFBR:
2881 case SIOCGIFBR:
2882 return old_bridge_ioctl(argp);
2883 case SIOCGIFNAME:
2884 return dev_ifname32(net, argp);
2885 case SIOCGIFCONF:
2886 return dev_ifconf(net, argp);
2887 case SIOCETHTOOL:
2888 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002889 case SIOCWANDEV:
2890 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002891 case SIOCGIFMAP:
2892 case SIOCSIFMAP:
2893 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002894 case SIOCBONDENSLAVE:
2895 case SIOCBONDRELEASE:
2896 case SIOCBONDSETHWADDR:
2897 case SIOCBONDSLAVEINFOQUERY:
2898 case SIOCBONDINFOQUERY:
2899 case SIOCBONDCHANGEACTIVE:
2900 return bond_ioctl(net, cmd, argp);
2901 case SIOCADDRT:
2902 case SIOCDELRT:
2903 return routing_ioctl(net, sock, cmd, argp);
2904 case SIOCGSTAMP:
2905 return do_siocgstamp(net, sock, cmd, argp);
2906 case SIOCGSTAMPNS:
2907 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002908 case SIOCSHWTSTAMP:
2909 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002910
Arnd Bergmann6b960182009-11-06 23:10:54 -08002911 case FIOSETOWN:
2912 case SIOCSPGRP:
2913 case FIOGETOWN:
2914 case SIOCGPGRP:
2915 case SIOCBRADDBR:
2916 case SIOCBRDELBR:
2917 case SIOCGIFVLAN:
2918 case SIOCSIFVLAN:
2919 case SIOCADDDLCI:
2920 case SIOCDELDLCI:
2921 return sock_ioctl(file, cmd, arg);
2922
2923 case SIOCGIFFLAGS:
2924 case SIOCSIFFLAGS:
2925 case SIOCGIFMETRIC:
2926 case SIOCSIFMETRIC:
2927 case SIOCGIFMTU:
2928 case SIOCSIFMTU:
2929 case SIOCGIFMEM:
2930 case SIOCSIFMEM:
2931 case SIOCGIFHWADDR:
2932 case SIOCSIFHWADDR:
2933 case SIOCADDMULTI:
2934 case SIOCDELMULTI:
2935 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002936 case SIOCGIFADDR:
2937 case SIOCSIFADDR:
2938 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002939 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002940 case SIOCGIFBRDADDR:
2941 case SIOCSIFBRDADDR:
2942 case SIOCGIFDSTADDR:
2943 case SIOCSIFDSTADDR:
2944 case SIOCGIFNETMASK:
2945 case SIOCSIFNETMASK:
2946 case SIOCSIFPFLAGS:
2947 case SIOCGIFPFLAGS:
2948 case SIOCGIFTXQLEN:
2949 case SIOCSIFTXQLEN:
2950 case SIOCBRADDIF:
2951 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002952 case SIOCSIFNAME:
2953 case SIOCGMIIPHY:
2954 case SIOCGMIIREG:
2955 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002956 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002957
Arnd Bergmann6b960182009-11-06 23:10:54 -08002958 case SIOCSARP:
2959 case SIOCGARP:
2960 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002961 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002962 return sock_do_ioctl(net, sock, cmd, arg);
2963 }
2964
2965 /* Prevent warning from compat_sys_ioctl, these always
2966 * result in -EINVAL in the native case anyway. */
2967 switch (cmd) {
2968 case SIOCRTMSG:
2969 case SIOCGIFCOUNT:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002970 case SIOCSRARP:
2971 case SIOCGRARP:
2972 case SIOCDRARP:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002973 case SIOCSIFLINK:
2974 case SIOCGIFSLAVE:
2975 case SIOCSIFSLAVE:
2976 return -EINVAL;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002977 }
2978
2979 return -ENOIOCTLCMD;
2980}
Arnd Bergmann7a229382009-11-06 23:00:29 -08002981
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002982static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002983 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002984{
2985 struct socket *sock = file->private_data;
2986 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07002987 struct sock *sk;
2988 struct net *net;
2989
2990 sk = sock->sk;
2991 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002992
2993 if (sock->ops->compat_ioctl)
2994 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2995
David S. Miller87de87d2008-06-03 09:14:03 -07002996 if (ret == -ENOIOCTLCMD &&
2997 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2998 ret = compat_wext_handle_ioctl(net, cmd, arg);
2999
Arnd Bergmann6b960182009-11-06 23:10:54 -08003000 if (ret == -ENOIOCTLCMD)
3001 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3002
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003003 return ret;
3004}
3005#endif
3006
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003007int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3008{
3009 return sock->ops->bind(sock, addr, addrlen);
3010}
3011
3012int kernel_listen(struct socket *sock, int backlog)
3013{
3014 return sock->ops->listen(sock, backlog);
3015}
3016
3017int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3018{
3019 struct sock *sk = sock->sk;
3020 int err;
3021
3022 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3023 newsock);
3024 if (err < 0)
3025 goto done;
3026
3027 err = sock->ops->accept(sock, *newsock, flags);
3028 if (err < 0) {
3029 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003030 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003031 goto done;
3032 }
3033
3034 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003035 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003036
3037done:
3038 return err;
3039}
3040
3041int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003042 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003043{
3044 return sock->ops->connect(sock, addr, addrlen, flags);
3045}
3046
3047int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3048 int *addrlen)
3049{
3050 return sock->ops->getname(sock, addr, addrlen, 0);
3051}
3052
3053int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3054 int *addrlen)
3055{
3056 return sock->ops->getname(sock, addr, addrlen, 1);
3057}
3058
3059int kernel_getsockopt(struct socket *sock, int level, int optname,
3060 char *optval, int *optlen)
3061{
3062 mm_segment_t oldfs = get_fs();
3063 int err;
3064
3065 set_fs(KERNEL_DS);
3066 if (level == SOL_SOCKET)
3067 err = sock_getsockopt(sock, level, optname, optval, optlen);
3068 else
3069 err = sock->ops->getsockopt(sock, level, optname, optval,
3070 optlen);
3071 set_fs(oldfs);
3072 return err;
3073}
3074
3075int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003076 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003077{
3078 mm_segment_t oldfs = get_fs();
3079 int err;
3080
3081 set_fs(KERNEL_DS);
3082 if (level == SOL_SOCKET)
3083 err = sock_setsockopt(sock, level, optname, optval, optlen);
3084 else
3085 err = sock->ops->setsockopt(sock, level, optname, optval,
3086 optlen);
3087 set_fs(oldfs);
3088 return err;
3089}
3090
3091int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3092 size_t size, int flags)
3093{
3094 if (sock->ops->sendpage)
3095 return sock->ops->sendpage(sock, page, offset, size, flags);
3096
3097 return sock_no_sendpage(sock, page, offset, size, flags);
3098}
3099
3100int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3101{
3102 mm_segment_t oldfs = get_fs();
3103 int err;
3104
3105 set_fs(KERNEL_DS);
3106 err = sock->ops->ioctl(sock, cmd, arg);
3107 set_fs(oldfs);
3108
3109 return err;
3110}
3111
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003112int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3113{
3114 return sock->ops->shutdown(sock, how);
3115}
3116
Linus Torvalds1da177e2005-04-16 15:20:36 -07003117EXPORT_SYMBOL(sock_create);
3118EXPORT_SYMBOL(sock_create_kern);
3119EXPORT_SYMBOL(sock_create_lite);
3120EXPORT_SYMBOL(sock_map_fd);
3121EXPORT_SYMBOL(sock_recvmsg);
3122EXPORT_SYMBOL(sock_register);
3123EXPORT_SYMBOL(sock_release);
3124EXPORT_SYMBOL(sock_sendmsg);
3125EXPORT_SYMBOL(sock_unregister);
3126EXPORT_SYMBOL(sock_wake_async);
3127EXPORT_SYMBOL(sockfd_lookup);
3128EXPORT_SYMBOL(kernel_sendmsg);
3129EXPORT_SYMBOL(kernel_recvmsg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003130EXPORT_SYMBOL(kernel_bind);
3131EXPORT_SYMBOL(kernel_listen);
3132EXPORT_SYMBOL(kernel_accept);
3133EXPORT_SYMBOL(kernel_connect);
3134EXPORT_SYMBOL(kernel_getsockname);
3135EXPORT_SYMBOL(kernel_getpeername);
3136EXPORT_SYMBOL(kernel_getsockopt);
3137EXPORT_SYMBOL(kernel_setsockopt);
3138EXPORT_SYMBOL(kernel_sendpage);
3139EXPORT_SYMBOL(kernel_sock_ioctl);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003140EXPORT_SYMBOL(kernel_sock_shutdown);