blob: 3ca2fd9e37200e3e12f6606065acb8c982f25528 [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 Black1fd7317d2009-09-22 16:43:33 -070089#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090090#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070096#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070097#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
Arnd Bergmann6b960182009-11-06 23:10:54 -0800102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700121 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800122#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700127 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800128 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
Arjan van de Venda7071d2007-02-12 00:55:36 -0800135static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 .owner = THIS_MODULE,
137 .llseek = no_llseek,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
140 .poll = sock_poll,
141 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800142#ifdef CONFIG_COMPAT
143 .compat_ioctl = compat_sock_ioctl,
144#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 .mmap = sock_mmap,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800151 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700159static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161/*
162 * Statistics counters of the socket lists
163 */
164
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700165static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
167/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700188 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700190 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100192 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700211
stephen hemminger11165f12010-10-18 14:27:29 +0000212static int move_addr_to_user(struct sockaddr *kaddr, int klen,
213 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 int err;
216 int len;
217
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218 err = get_user(len, ulen);
219 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700221 if (len > klen)
222 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700223 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return -EFAULT;
230 }
231 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235 return __put_user(klen, ulen);
236}
237
Christoph Lametere18b8902006-12-06 20:33:20 -0800238static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700243
Christoph Lametere94b1762006-12-06 20:33:17 -0800244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 if (!ei)
246 return NULL;
Eric Dumazet43815482010-04-29 11:01:49 +0000247 ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
248 if (!ei->socket.wq) {
249 kmem_cache_free(sock_inode_cachep, ei);
250 return NULL;
251 }
252 init_waitqueue_head(&ei->socket.wq->wait);
253 ei->socket.wq->fasync_list = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 ei->socket.state = SS_UNCONNECTED;
256 ei->socket.flags = 0;
257 ei->socket.ops = NULL;
258 ei->socket.sk = NULL;
259 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
261 return &ei->vfs_inode;
262}
263
Eric Dumazet43815482010-04-29 11:01:49 +0000264
265static void wq_free_rcu(struct rcu_head *head)
266{
267 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
268
269 kfree(wq);
270}
271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272static void sock_destroy_inode(struct inode *inode)
273{
Eric Dumazet43815482010-04-29 11:01:49 +0000274 struct socket_alloc *ei;
275
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
277 call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
278 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279}
280
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700281static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700283 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
Christoph Lametera35afb82007-05-16 22:10:57 -0700285 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700287
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288static int init_inodecache(void)
289{
290 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700291 sizeof(struct socket_alloc),
292 0,
293 (SLAB_HWCACHE_ALIGN |
294 SLAB_RECLAIM_ACCOUNT |
295 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900296 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 if (sock_inode_cachep == NULL)
298 return -ENOMEM;
299 return 0;
300}
301
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700302static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306};
307
Al Viro51139ad2010-07-25 23:47:46 +0400308static struct dentry *sockfs_mount(struct file_system_type *fs_type,
309 int flags, const char *dev_name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310{
Al Viro51139ad2010-07-25 23:47:46 +0400311 return mount_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312}
313
Eric Dumazetba899662005-08-26 12:05:31 -0700314static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
316static struct file_system_type sock_fs_type = {
317 .name = "sockfs",
Al Viro51139ad2010-07-25 23:47:46 +0400318 .mount = sockfs_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 .kill_sb = kill_anon_super,
320};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700321
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700322/*
323 * sockfs_dname() is called from d_path().
324 */
325static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
326{
327 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
328 dentry->d_inode->i_ino);
329}
330
Al Viro3ba13d12009-02-20 06:02:22 +0000331static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700332 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333};
334
335/*
336 * Obtains the first available file descriptor and sets it up for use.
337 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800338 * These functions create file structures and maps them to fd space
339 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 * and file struct implicitly stored in sock->file.
341 * Note that another thread may close file descriptor before we return
342 * from this function. We use the fact that now we do not refer
343 * to socket after mapping. If one day we will need it, this
344 * function will increment ref. count on file by 1.
345 *
346 * In any case returned fd MAY BE not valid!
347 * This race condition is unavoidable
348 * with shared fd spaces, we cannot solve it inside kernel,
349 * but we take care of internal coherence yet.
350 */
351
Al Viro7cbe66b2009-08-05 19:59:08 +0400352static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353{
Al Viro7cbe66b2009-08-05 19:59:08 +0400354 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400355 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400356 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800358
Ulrich Dreppera677a032008-07-23 21:29:17 -0700359 fd = get_unused_fd_flags(flags);
Al Viro7cbe66b2009-08-05 19:59:08 +0400360 if (unlikely(fd < 0))
361 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800362
Al Viro2c48b9c2009-08-09 00:52:35 +0400363 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
364 if (unlikely(!path.dentry)) {
Al Viro7cbe66b2009-08-05 19:59:08 +0400365 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800366 return -ENOMEM;
Al Viro7cbe66b2009-08-05 19:59:08 +0400367 }
Al Viro2c48b9c2009-08-09 00:52:35 +0400368 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369
Al Viro2c48b9c2009-08-09 00:52:35 +0400370 path.dentry->d_op = &sockfs_dentry_operations;
Al Viro2c48b9c2009-08-09 00:52:35 +0400371 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400372 SOCK_INODE(sock)->i_fop = &socket_file_ops;
373
Al Viro2c48b9c2009-08-09 00:52:35 +0400374 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400375 &socket_file_ops);
376 if (unlikely(!file)) {
377 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400378 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400379 path_put(&path);
Al Virocc3808f2009-08-06 09:43:59 +0400380 put_unused_fd(fd);
381 return -ENFILE;
382 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800383
384 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700385 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800386 file->f_pos = 0;
387 file->private_data = sock;
388
Al Viro7cbe66b2009-08-05 19:59:08 +0400389 *f = file;
390 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800391}
392
Ulrich Dreppera677a032008-07-23 21:29:17 -0700393int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800394{
395 struct file *newfile;
Al Viro7cbe66b2009-08-05 19:59:08 +0400396 int fd = sock_alloc_file(sock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800397
Al Viro7cbe66b2009-08-05 19:59:08 +0400398 if (likely(fd >= 0))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800399 fd_install(fd, newfile);
Al Viro7cbe66b2009-08-05 19:59:08 +0400400
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 return fd;
402}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700403EXPORT_SYMBOL(sock_map_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800405static struct socket *sock_from_file(struct file *file, int *err)
406{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800407 if (file->f_op == &socket_file_ops)
408 return file->private_data; /* set in sock_map_fd */
409
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800410 *err = -ENOTSOCK;
411 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800412}
413
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700415 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 * @fd: file handle
417 * @err: pointer to an error code return
418 *
419 * The file handle passed in is locked and the socket it is bound
420 * too is returned. If an error occurs the err pointer is overwritten
421 * with a negative errno code and NULL is returned. The function checks
422 * for both invalid handles and passing a handle which is not a socket.
423 *
424 * On a success the socket object pointer is returned.
425 */
426
427struct socket *sockfd_lookup(int fd, int *err)
428{
429 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 struct socket *sock;
431
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700432 file = fget(fd);
433 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 *err = -EBADF;
435 return NULL;
436 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700437
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800438 sock = sock_from_file(file, err);
439 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 return sock;
442}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700443EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800445static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
446{
447 struct file *file;
448 struct socket *sock;
449
Hua Zhong36725582006-04-19 15:25:02 -0700450 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451 file = fget_light(fd, fput_needed);
452 if (file) {
453 sock = sock_from_file(file, err);
454 if (sock)
455 return sock;
456 fput_light(file, *fput_needed);
457 }
458 return NULL;
459}
460
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461/**
462 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700463 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 * Allocate a new inode and socket object. The two are bound together
465 * and initialised. The socket is then returned. If we are out of inodes
466 * NULL is returned.
467 */
468
469static struct socket *sock_alloc(void)
470{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700471 struct inode *inode;
472 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
474 inode = new_inode(sock_mnt->mnt_sb);
475 if (!inode)
476 return NULL;
477
478 sock = SOCKET_I(inode);
479
Eric Dumazet29a020d2009-09-15 02:39:20 -0700480 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400481 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700482 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100483 inode->i_uid = current_fsuid();
484 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Eric Dumazet4e694892009-04-04 16:41:09 -0700486 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 return sock;
488}
489
490/*
491 * In theory you can't get an open on this inode, but /proc provides
492 * a back door. Remember to keep it shut otherwise you'll let the
493 * creepy crawlies in.
494 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700495
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
497{
498 return -ENXIO;
499}
500
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800501const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 .owner = THIS_MODULE,
503 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200504 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505};
506
507/**
508 * sock_release - close a socket
509 * @sock: socket to close
510 *
511 * The socket is released from the protocol stack if it has a release
512 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700513 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516void sock_release(struct socket *sock)
517{
518 if (sock->ops) {
519 struct module *owner = sock->ops->owner;
520
521 sock->ops->release(sock);
522 sock->ops = NULL;
523 module_put(owner);
524 }
525
Eric Dumazet43815482010-04-29 11:01:49 +0000526 if (sock->wq->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 printk(KERN_ERR "sock_release: fasync list not empty!\n");
528
Eric Dumazet4e694892009-04-04 16:41:09 -0700529 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 if (!sock->file) {
531 iput(SOCK_INODE(sock));
532 return;
533 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700534 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700536EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000538int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000539{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000540 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000541 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000542 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000543 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000544 *tx_flags |= SKBTX_SW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000545 return 0;
546}
547EXPORT_SYMBOL(sock_tx_timestamp);
548
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700549static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 struct msghdr *msg, size_t size)
551{
552 struct sock_iocb *si = kiocb_to_siocb(iocb);
553 int err;
554
Herbert Xuf8451722010-05-24 00:12:34 -0700555 sock_update_classid(sock->sk);
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 si->sock = sock;
558 si->scm = NULL;
559 si->msg = msg;
560 si->size = size;
561
562 err = security_socket_sendmsg(sock, msg, size);
563 if (err)
564 return err;
565
566 return sock->ops->sendmsg(iocb, sock, msg, size);
567}
568
569int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
570{
571 struct kiocb iocb;
572 struct sock_iocb siocb;
573 int ret;
574
575 init_sync_kiocb(&iocb, NULL);
576 iocb.private = &siocb;
577 ret = __sock_sendmsg(&iocb, sock, msg, size);
578 if (-EIOCBQUEUED == ret)
579 ret = wait_on_sync_kiocb(&iocb);
580 return ret;
581}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700582EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
584int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
585 struct kvec *vec, size_t num, size_t size)
586{
587 mm_segment_t oldfs = get_fs();
588 int result;
589
590 set_fs(KERNEL_DS);
591 /*
592 * the following is safe, since for compiler definitions of kvec and
593 * iovec are identical, yielding the same in-core layout and alignment
594 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700595 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 msg->msg_iovlen = num;
597 result = sock_sendmsg(sock, msg, size);
598 set_fs(oldfs);
599 return result;
600}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700601EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
Patrick Ohly20d49472009-02-12 05:03:38 +0000603static int ktime2ts(ktime_t kt, struct timespec *ts)
604{
605 if (kt.tv64) {
606 *ts = ktime_to_timespec(kt);
607 return 1;
608 } else {
609 return 0;
610 }
611}
612
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700613/*
614 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
615 */
616void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
617 struct sk_buff *skb)
618{
Patrick Ohly20d49472009-02-12 05:03:38 +0000619 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
620 struct timespec ts[3];
621 int empty = 1;
622 struct skb_shared_hwtstamps *shhwtstamps =
623 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700624
Patrick Ohly20d49472009-02-12 05:03:38 +0000625 /* Race occurred between timestamp enabling and packet
626 receiving. Fill in the current time for now. */
627 if (need_software_tstamp && skb->tstamp.tv64 == 0)
628 __net_timestamp(skb);
629
630 if (need_software_tstamp) {
631 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
632 struct timeval tv;
633 skb_get_timestamp(skb, &tv);
634 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
635 sizeof(tv), &tv);
636 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000637 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000638 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000639 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000640 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700641 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000642
643
644 memset(ts, 0, sizeof(ts));
645 if (skb->tstamp.tv64 &&
646 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
647 skb_get_timestampns(skb, ts + 0);
648 empty = 0;
649 }
650 if (shhwtstamps) {
651 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
652 ktime2ts(shhwtstamps->syststamp, ts + 1))
653 empty = 0;
654 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
655 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
656 empty = 0;
657 }
658 if (!empty)
659 put_cmsg(msg, SOL_SOCKET,
660 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700661}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300662EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
663
stephen hemminger11165f12010-10-18 14:27:29 +0000664static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
665 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700666{
667 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
668 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
669 sizeof(__u32), &skb->dropcount);
670}
671
Eric Dumazet767dd032010-04-28 19:14:43 +0000672void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700673 struct sk_buff *skb)
674{
675 sock_recv_timestamp(msg, sk, skb);
676 sock_recv_drops(msg, sk, skb);
677}
Eric Dumazet767dd032010-04-28 19:14:43 +0000678EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700679
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700680static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
681 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 struct sock_iocb *si = kiocb_to_siocb(iocb);
684
Herbert Xuf8451722010-05-24 00:12:34 -0700685 sock_update_classid(sock->sk);
686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 si->sock = sock;
688 si->scm = NULL;
689 si->msg = msg;
690 si->size = size;
691 si->flags = flags;
692
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
694}
695
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700696static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
697 struct msghdr *msg, size_t size, int flags)
698{
699 int err = security_socket_recvmsg(sock, msg, size, flags);
700
701 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
702}
703
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700704int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 size_t size, int flags)
706{
707 struct kiocb iocb;
708 struct sock_iocb siocb;
709 int ret;
710
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700711 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712 iocb.private = &siocb;
713 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
714 if (-EIOCBQUEUED == ret)
715 ret = wait_on_sync_kiocb(&iocb);
716 return ret;
717}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700718EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700720static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
721 size_t size, int flags)
722{
723 struct kiocb iocb;
724 struct sock_iocb siocb;
725 int ret;
726
727 init_sync_kiocb(&iocb, NULL);
728 iocb.private = &siocb;
729 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
730 if (-EIOCBQUEUED == ret)
731 ret = wait_on_sync_kiocb(&iocb);
732 return ret;
733}
734
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700735int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
736 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700737{
738 mm_segment_t oldfs = get_fs();
739 int result;
740
741 set_fs(KERNEL_DS);
742 /*
743 * the following is safe, since for compiler definitions of kvec and
744 * iovec are identical, yielding the same in-core layout and alignment
745 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700746 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 result = sock_recvmsg(sock, msg, size, flags);
748 set_fs(oldfs);
749 return result;
750}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700751EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752
753static void sock_aio_dtor(struct kiocb *iocb)
754{
755 kfree(iocb->private);
756}
757
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300758static ssize_t sock_sendpage(struct file *file, struct page *page,
759 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760{
761 struct socket *sock;
762 int flags;
763
Eric Dumazetb69aee02005-09-06 14:42:45 -0700764 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765
766 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
767 if (more)
768 flags |= MSG_MORE;
769
Linus Torvaldse6949582009-08-13 08:28:36 -0700770 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771}
772
Jens Axboe9c55e012007-11-06 23:30:13 -0800773static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700774 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800775 unsigned int flags)
776{
777 struct socket *sock = file->private_data;
778
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800779 if (unlikely(!sock->ops->splice_read))
780 return -EINVAL;
781
Herbert Xuf8451722010-05-24 00:12:34 -0700782 sock_update_classid(sock->sk);
783
Jens Axboe9c55e012007-11-06 23:30:13 -0800784 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
785}
786
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800787static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700788 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800789{
790 if (!is_sync_kiocb(iocb)) {
791 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
792 if (!siocb)
793 return NULL;
794 iocb->ki_dtor = sock_aio_dtor;
795 }
796
797 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800798 iocb->private = siocb;
799 return siocb;
800}
801
802static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700803 struct file *file, const struct iovec *iov,
804 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800805{
806 struct socket *sock = file->private_data;
807 size_t size = 0;
808 int i;
809
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700810 for (i = 0; i < nr_segs; i++)
811 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800812
813 msg->msg_name = NULL;
814 msg->msg_namelen = 0;
815 msg->msg_control = NULL;
816 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700817 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800818 msg->msg_iovlen = nr_segs;
819 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
820
821 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
822}
823
Badari Pulavarty027445c2006-09-30 23:28:46 -0700824static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
825 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800826{
827 struct sock_iocb siocb, *x;
828
829 if (pos != 0)
830 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700831
832 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800833 return 0;
834
Badari Pulavarty027445c2006-09-30 23:28:46 -0700835
836 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800837 if (!x)
838 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700839 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800840}
841
842static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700843 struct file *file, const struct iovec *iov,
844 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800845{
846 struct socket *sock = file->private_data;
847 size_t size = 0;
848 int i;
849
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700850 for (i = 0; i < nr_segs; i++)
851 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800852
853 msg->msg_name = NULL;
854 msg->msg_namelen = 0;
855 msg->msg_control = NULL;
856 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700857 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800858 msg->msg_iovlen = nr_segs;
859 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
860 if (sock->type == SOCK_SEQPACKET)
861 msg->msg_flags |= MSG_EOR;
862
863 return __sock_sendmsg(iocb, sock, msg, size);
864}
865
Badari Pulavarty027445c2006-09-30 23:28:46 -0700866static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
867 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800869 struct sock_iocb siocb, *x;
870
871 if (pos != 0)
872 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700873
Badari Pulavarty027445c2006-09-30 23:28:46 -0700874 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800875 if (!x)
876 return -ENOMEM;
877
Badari Pulavarty027445c2006-09-30 23:28:46 -0700878 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879}
880
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881/*
882 * Atomic setting of ioctl hooks to avoid race
883 * with module unload.
884 */
885
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800886static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700887static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888
Eric W. Biederman881d9662007-09-17 11:56:21 -0700889void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800891 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800893 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894}
895EXPORT_SYMBOL(brioctl_set);
896
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800897static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700898static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899
Eric W. Biederman881d9662007-09-17 11:56:21 -0700900void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800902 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800904 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905}
906EXPORT_SYMBOL(vlan_ioctl_set);
907
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800908static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700909static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700911void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800913 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800915 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917EXPORT_SYMBOL(dlci_ioctl_set);
918
Arnd Bergmann6b960182009-11-06 23:10:54 -0800919static long sock_do_ioctl(struct net *net, struct socket *sock,
920 unsigned int cmd, unsigned long arg)
921{
922 int err;
923 void __user *argp = (void __user *)arg;
924
925 err = sock->ops->ioctl(sock, cmd, arg);
926
927 /*
928 * If this ioctl is unknown try to hand it down
929 * to the NIC driver.
930 */
931 if (err == -ENOIOCTLCMD)
932 err = dev_ioctl(net, cmd, argp);
933
934 return err;
935}
936
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937/*
938 * With an ioctl, arg may well be a user mode pointer, but we don't know
939 * what to do with it - that's up to the protocol still.
940 */
941
942static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
943{
944 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700945 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946 void __user *argp = (void __user *)arg;
947 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700948 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
Eric Dumazetb69aee02005-09-06 14:42:45 -0700950 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700951 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900952 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700954 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200956#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700958 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200960#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700961 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 case FIOSETOWN:
963 case SIOCSPGRP:
964 err = -EFAULT;
965 if (get_user(pid, (int __user *)argp))
966 break;
967 err = f_setown(sock->file, pid, 1);
968 break;
969 case FIOGETOWN:
970 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700971 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700972 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 break;
974 case SIOCGIFBR:
975 case SIOCSIFBR:
976 case SIOCBRADDBR:
977 case SIOCBRDELBR:
978 err = -ENOPKG;
979 if (!br_ioctl_hook)
980 request_module("bridge");
981
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800982 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700983 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700984 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800985 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 break;
987 case SIOCGIFVLAN:
988 case SIOCSIFVLAN:
989 err = -ENOPKG;
990 if (!vlan_ioctl_hook)
991 request_module("8021q");
992
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800993 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700995 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800996 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 case SIOCADDDLCI:
999 case SIOCDELDLCI:
1000 err = -ENOPKG;
1001 if (!dlci_ioctl_hook)
1002 request_module("dlci");
1003
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001004 mutex_lock(&dlci_ioctl_mutex);
1005 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001007 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 break;
1009 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001010 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001012 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 return err;
1014}
1015
1016int sock_create_lite(int family, int type, int protocol, struct socket **res)
1017{
1018 int err;
1019 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001020
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 err = security_socket_create(family, type, protocol, 1);
1022 if (err)
1023 goto out;
1024
1025 sock = sock_alloc();
1026 if (!sock) {
1027 err = -ENOMEM;
1028 goto out;
1029 }
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001032 err = security_socket_post_create(sock, family, type, protocol, 1);
1033 if (err)
1034 goto out_release;
1035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036out:
1037 *res = sock;
1038 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001039out_release:
1040 sock_release(sock);
1041 sock = NULL;
1042 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001044EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045
1046/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001047static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048{
1049 struct socket *sock;
1050
1051 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001052 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001054 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 return sock->ops->poll(file, sock, wait);
1056}
1057
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001058static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001060 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
1062 return sock->ops->mmap(file, sock, vma);
1063}
1064
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001065static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066{
1067 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001068 * It was possible the inode is NULL we were
1069 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 */
1071
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001072 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 printk(KERN_DEBUG "sock_close: NULL inode\n");
1074 return 0;
1075 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 sock_release(SOCKET_I(inode));
1077 return 0;
1078}
1079
1080/*
1081 * Update the socket async list
1082 *
1083 * Fasync_list locking strategy.
1084 *
1085 * 1. fasync_list is modified only under process context socket lock
1086 * i.e. under semaphore.
1087 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001088 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 */
1090
1091static int sock_fasync(int fd, struct file *filp, int on)
1092{
Eric Dumazet989a2972010-04-14 09:55:35 +00001093 struct socket *sock = filp->private_data;
1094 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
Eric Dumazet989a2972010-04-14 09:55:35 +00001096 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098
1099 lock_sock(sk);
1100
Eric Dumazet43815482010-04-29 11:01:49 +00001101 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
Eric Dumazet989a2972010-04-14 09:55:35 +00001102
Eric Dumazet43815482010-04-29 11:01:49 +00001103 if (!sock->wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001104 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001105 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001106 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Eric Dumazet989a2972010-04-14 09:55:35 +00001108 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 return 0;
1110}
1111
Eric Dumazet43815482010-04-29 11:01:49 +00001112/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113
1114int sock_wake_async(struct socket *sock, int how, int band)
1115{
Eric Dumazet43815482010-04-29 11:01:49 +00001116 struct socket_wq *wq;
1117
1118 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001120 rcu_read_lock();
1121 wq = rcu_dereference(sock->wq);
1122 if (!wq || !wq->fasync_list) {
1123 rcu_read_unlock();
1124 return -1;
1125 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001126 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001127 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1129 break;
1130 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001131 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1133 break;
1134 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001135 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001136call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001137 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001139 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001140 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 }
Eric Dumazet43815482010-04-29 11:01:49 +00001142 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 return 0;
1144}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001145EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Pavel Emelyanov721db932010-09-29 16:06:32 +04001147int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001148 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149{
1150 int err;
1151 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001152 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
1154 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001155 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156 */
1157 if (family < 0 || family >= NPROTO)
1158 return -EAFNOSUPPORT;
1159 if (type < 0 || type >= SOCK_MAX)
1160 return -EINVAL;
1161
1162 /* Compatibility.
1163
1164 This uglymoron is moved from INET layer to here to avoid
1165 deadlock in module load.
1166 */
1167 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001168 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 if (!warned) {
1170 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001171 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1172 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 }
1174 family = PF_PACKET;
1175 }
1176
1177 err = security_socket_create(family, type, protocol, kern);
1178 if (err)
1179 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001180
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001181 /*
1182 * Allocate the socket and allow the family to set things up. if
1183 * the protocol is 0, the family is instructed to select an appropriate
1184 * default.
1185 */
1186 sock = sock_alloc();
1187 if (!sock) {
1188 if (net_ratelimit())
1189 printk(KERN_WARNING "socket: no more sockets\n");
1190 return -ENFILE; /* Not exactly a match, but its the
1191 closest posix thing */
1192 }
1193
1194 sock->type = type;
1195
Johannes Berg95a5afc2008-10-16 15:24:51 -07001196#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001197 /* Attempt to load a protocol module if the find failed.
1198 *
1199 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 * requested real, full-featured networking support upon configuration.
1201 * Otherwise module support will break!
1202 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001203 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001204 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205#endif
1206
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001207 rcu_read_lock();
1208 pf = rcu_dereference(net_families[family]);
1209 err = -EAFNOSUPPORT;
1210 if (!pf)
1211 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212
1213 /*
1214 * We will call the ->create function, that possibly is in a loadable
1215 * module, so we have to bump that loadable module refcnt first.
1216 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001217 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 goto out_release;
1219
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001220 /* Now protected by module ref count */
1221 rcu_read_unlock();
1222
Eric Paris3f378b62009-11-05 22:18:14 -08001223 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001224 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001226
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 /*
1228 * Now to bump the refcnt of the [loadable] module that owns this
1229 * socket at sock_release time we decrement its refcnt.
1230 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001231 if (!try_module_get(sock->ops->owner))
1232 goto out_module_busy;
1233
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 /*
1235 * Now that we're done with the ->create function, the [loadable]
1236 * module can have its refcnt decremented
1237 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001238 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001239 err = security_socket_post_create(sock, family, type, protocol, kern);
1240 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001241 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001242 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001244 return 0;
1245
1246out_module_busy:
1247 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001249 sock->ops = NULL;
1250 module_put(pf->owner);
1251out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001253 return err;
1254
1255out_release:
1256 rcu_read_unlock();
1257 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001259EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260
1261int sock_create(int family, int type, int protocol, struct socket **res)
1262{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001263 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001265EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
1267int sock_create_kern(int family, int type, int protocol, struct socket **res)
1268{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001269 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001271EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001273SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274{
1275 int retval;
1276 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001277 int flags;
1278
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001279 /* Check the SOCK_* constants for consistency. */
1280 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1281 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1282 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1283 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1284
Ulrich Dreppera677a032008-07-23 21:29:17 -07001285 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001286 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001287 return -EINVAL;
1288 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001290 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1291 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1292
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 retval = sock_create(family, type, protocol, &sock);
1294 if (retval < 0)
1295 goto out;
1296
Ulrich Drepper77d27202008-07-23 21:29:35 -07001297 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 if (retval < 0)
1299 goto out_release;
1300
1301out:
1302 /* It may be already another descriptor 8) Not kernel problem. */
1303 return retval;
1304
1305out_release:
1306 sock_release(sock);
1307 return retval;
1308}
1309
1310/*
1311 * Create a pair of connected sockets.
1312 */
1313
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001314SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1315 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316{
1317 struct socket *sock1, *sock2;
1318 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001319 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001320 int flags;
1321
1322 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001323 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001324 return -EINVAL;
1325 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001327 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1328 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1329
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 /*
1331 * Obtain the first socket and check if the underlying protocol
1332 * supports the socketpair call.
1333 */
1334
1335 err = sock_create(family, type, protocol, &sock1);
1336 if (err < 0)
1337 goto out;
1338
1339 err = sock_create(family, type, protocol, &sock2);
1340 if (err < 0)
1341 goto out_release_1;
1342
1343 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001344 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 goto out_release_both;
1346
Al Viro7cbe66b2009-08-05 19:59:08 +04001347 fd1 = sock_alloc_file(sock1, &newfile1, flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001348 if (unlikely(fd1 < 0)) {
1349 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001351 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Al Viro7cbe66b2009-08-05 19:59:08 +04001353 fd2 = sock_alloc_file(sock2, &newfile2, flags);
Al Viro198de4d2009-08-05 19:29:23 +04001354 if (unlikely(fd2 < 0)) {
1355 err = fd2;
1356 fput(newfile1);
1357 put_unused_fd(fd1);
1358 sock_release(sock2);
1359 goto out;
Al Virodb349502007-02-07 01:48:00 -05001360 }
1361
Al Viro157cf642008-12-14 04:57:47 -05001362 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001363 fd_install(fd1, newfile1);
1364 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 /* fd1 and fd2 may be already another descriptors.
1366 * Not kernel problem.
1367 */
1368
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001369 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 if (!err)
1371 err = put_user(fd2, &usockvec[1]);
1372 if (!err)
1373 return 0;
1374
1375 sys_close(fd2);
1376 sys_close(fd1);
1377 return err;
1378
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001380 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001382 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383out:
1384 return err;
1385}
1386
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387/*
1388 * Bind a name to a socket. Nothing much to do here since it's
1389 * the protocol's responsibility to handle the local address.
1390 *
1391 * We move the socket address to kernel space before we call
1392 * the protocol layer (having also checked the address is ok).
1393 */
1394
Heiko Carstens20f37032009-01-14 14:14:23 +01001395SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396{
1397 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001398 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001399 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001401 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001402 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001403 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001404 if (err >= 0) {
1405 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001406 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001407 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001408 if (!err)
1409 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001410 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001411 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001413 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001414 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 return err;
1416}
1417
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418/*
1419 * Perform a listen. Basically, we allow the protocol to do anything
1420 * necessary for a listen, and if that works, we mark the socket as
1421 * ready for listening.
1422 */
1423
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001424SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425{
1426 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001427 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001428 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001429
1430 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1431 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001432 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001433 if ((unsigned)backlog > somaxconn)
1434 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
1436 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001437 if (!err)
1438 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001440 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 }
1442 return err;
1443}
1444
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445/*
1446 * For accept, we attempt to create a new socket, set up the link
1447 * with the client, wake up the client, then return the new
1448 * connected fd. We collect the address of the connector in kernel
1449 * space and move it to user at the very end. This is unclean because
1450 * we open the socket then return an error.
1451 *
1452 * 1003.1g adds the ability to recvmsg() to query connection pending
1453 * status to recvmsg. We need to add that support in a way thats
1454 * clean when we restucture accept also.
1455 */
1456
Heiko Carstens20f37032009-01-14 14:14:23 +01001457SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1458 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459{
1460 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001461 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001462 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001463 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Ulrich Drepper77d27202008-07-23 21:29:35 -07001465 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001466 return -EINVAL;
1467
1468 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1469 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1470
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001471 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 if (!sock)
1473 goto out;
1474
1475 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001476 newsock = sock_alloc();
1477 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 goto out_put;
1479
1480 newsock->type = sock->type;
1481 newsock->ops = sock->ops;
1482
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 /*
1484 * We don't need try_module_get here, as the listening socket (sock)
1485 * has the protocol module (sock->ops->owner) held.
1486 */
1487 __module_get(newsock->ops->owner);
1488
Al Viro7cbe66b2009-08-05 19:59:08 +04001489 newfd = sock_alloc_file(newsock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001490 if (unlikely(newfd < 0)) {
1491 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001492 sock_release(newsock);
1493 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001494 }
1495
Frank Filza79af592005-09-27 15:23:38 -07001496 err = security_socket_accept(sock, newsock);
1497 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001498 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001499
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1501 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001502 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503
1504 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001505 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001506 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001508 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001510 err = move_addr_to_user((struct sockaddr *)&address,
1511 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001513 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 }
1515
1516 /* File flags are not inherited via accept() unlike another OSes. */
1517
David S. Miller39d8c1b2006-03-20 17:13:49 -08001518 fd_install(newfd, newfile);
1519 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001522 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523out:
1524 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001525out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001526 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001527 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 goto out_put;
1529}
1530
Heiko Carstens20f37032009-01-14 14:14:23 +01001531SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1532 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001533{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001534 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001535}
1536
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537/*
1538 * Attempt to connect to a socket with the server address. The address
1539 * is in user space so we verify it is OK and move it to kernel space.
1540 *
1541 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1542 * break bindings
1543 *
1544 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1545 * other SEQPACKET protocols that take time to connect() as it doesn't
1546 * include the -EINPROGRESS status for such sockets.
1547 */
1548
Heiko Carstens20f37032009-01-14 14:14:23 +01001549SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1550 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551{
1552 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001553 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001554 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001556 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 if (!sock)
1558 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001559 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 if (err < 0)
1561 goto out_put;
1562
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001563 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001564 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 if (err)
1566 goto out_put;
1567
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001568 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 sock->file->f_flags);
1570out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001571 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572out:
1573 return err;
1574}
1575
1576/*
1577 * Get the local address ('name') of a socket object. Move the obtained
1578 * name to user space.
1579 */
1580
Heiko Carstens20f37032009-01-14 14:14:23 +01001581SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1582 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583{
1584 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001585 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001586 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001587
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001588 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 if (!sock)
1590 goto out;
1591
1592 err = security_socket_getsockname(sock);
1593 if (err)
1594 goto out_put;
1595
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001596 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 if (err)
1598 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001599 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600
1601out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001602 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603out:
1604 return err;
1605}
1606
1607/*
1608 * Get the remote address ('name') of a socket object. Move the obtained
1609 * name to user space.
1610 */
1611
Heiko Carstens20f37032009-01-14 14:14:23 +01001612SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1613 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614{
1615 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001616 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001617 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001619 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1620 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 err = security_socket_getpeername(sock);
1622 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001623 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 return err;
1625 }
1626
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001627 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001628 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001629 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001631 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001632 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001633 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634 }
1635 return err;
1636}
1637
1638/*
1639 * Send a datagram to a given address. We move the address into kernel
1640 * space and check the user space data area is readable before invoking
1641 * the protocol.
1642 */
1643
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001644SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1645 unsigned, flags, struct sockaddr __user *, addr,
1646 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647{
1648 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001649 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 int err;
1651 struct msghdr msg;
1652 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001653 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001654
Linus Torvalds253eacc2010-10-30 16:43:10 -07001655 if (len > INT_MAX)
1656 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001657 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1658 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001659 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001660
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001661 iov.iov_base = buff;
1662 iov.iov_len = len;
1663 msg.msg_name = NULL;
1664 msg.msg_iov = &iov;
1665 msg.msg_iovlen = 1;
1666 msg.msg_control = NULL;
1667 msg.msg_controllen = 0;
1668 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001669 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001670 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 if (err < 0)
1672 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001673 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001674 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 }
1676 if (sock->file->f_flags & O_NONBLOCK)
1677 flags |= MSG_DONTWAIT;
1678 msg.msg_flags = flags;
1679 err = sock_sendmsg(sock, &msg, len);
1680
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001681out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001682 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001683out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 return err;
1685}
1686
1687/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001688 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 */
1690
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001691SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1692 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693{
1694 return sys_sendto(fd, buff, len, flags, NULL, 0);
1695}
1696
1697/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001698 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 * sender. We verify the buffers are writable and if needed move the
1700 * sender address from kernel to user space.
1701 */
1702
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001703SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1704 unsigned, flags, struct sockaddr __user *, addr,
1705 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706{
1707 struct socket *sock;
1708 struct iovec iov;
1709 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001710 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001711 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001712 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Linus Torvalds253eacc2010-10-30 16:43:10 -07001714 if (size > INT_MAX)
1715 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001716 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001718 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001720 msg.msg_control = NULL;
1721 msg.msg_controllen = 0;
1722 msg.msg_iovlen = 1;
1723 msg.msg_iov = &iov;
1724 iov.iov_len = size;
1725 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001726 msg.msg_name = (struct sockaddr *)&address;
1727 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 if (sock->file->f_flags & O_NONBLOCK)
1729 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001730 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001732 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001733 err2 = move_addr_to_user((struct sockaddr *)&address,
1734 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001735 if (err2 < 0)
1736 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001738
1739 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001740out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 return err;
1742}
1743
1744/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001745 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 */
1747
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001748asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1749 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
1751 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1752}
1753
1754/*
1755 * Set a socket option. Because we don't know the option lengths we have
1756 * to pass the user mode parameter for the protocols to sort out.
1757 */
1758
Heiko Carstens20f37032009-01-14 14:14:23 +01001759SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1760 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001762 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 struct socket *sock;
1764
1765 if (optlen < 0)
1766 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001767
1768 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769 if (sock != NULL) {
1770 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001771 if (err)
1772 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
1774 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001775 err =
1776 sock_setsockopt(sock, level, optname, optval,
1777 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001779 err =
1780 sock->ops->setsockopt(sock, level, optname, optval,
1781 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001782out_put:
1783 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 }
1785 return err;
1786}
1787
1788/*
1789 * Get a socket option. Because we don't know the option lengths we have
1790 * to pass a user mode parameter for the protocols to sort out.
1791 */
1792
Heiko Carstens20f37032009-01-14 14:14:23 +01001793SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1794 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001796 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 struct socket *sock;
1798
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001799 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1800 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001801 err = security_socket_getsockopt(sock, level, optname);
1802 if (err)
1803 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804
1805 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001806 err =
1807 sock_getsockopt(sock, level, optname, optval,
1808 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001810 err =
1811 sock->ops->getsockopt(sock, level, optname, optval,
1812 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001813out_put:
1814 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 }
1816 return err;
1817}
1818
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819/*
1820 * Shutdown a socket.
1821 */
1822
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001823SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001825 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 struct socket *sock;
1827
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1829 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001831 if (!err)
1832 err = sock->ops->shutdown(sock, how);
1833 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 }
1835 return err;
1836}
1837
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001838/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 * fields which are the same type (int / unsigned) on our platforms.
1840 */
1841#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1842#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1843#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1844
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845/*
1846 * BSD sendmsg interface
1847 */
1848
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001849SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001851 struct compat_msghdr __user *msg_compat =
1852 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001854 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001856 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001857 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1858 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 unsigned char *ctl_buf = ctl;
1860 struct msghdr msg_sys;
1861 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001862 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001863
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 err = -EFAULT;
1865 if (MSG_CMSG_COMPAT & flags) {
1866 if (get_compat_msghdr(&msg_sys, msg_compat))
1867 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001868 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 return -EFAULT;
1870
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001871 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001872 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 goto out;
1874
1875 /* do not move before msg_sys is valid */
1876 err = -EMSGSIZE;
1877 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1878 goto out_put;
1879
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001880 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 err = -ENOMEM;
1882 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1883 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1884 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1885 if (!iov)
1886 goto out_put;
1887 }
1888
1889 /* This will also move the address data into kernel space */
1890 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001891 err = verify_compat_iovec(&msg_sys, iov,
1892 (struct sockaddr *)&address,
1893 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001895 err = verify_iovec(&msg_sys, iov,
1896 (struct sockaddr *)&address,
1897 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001898 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 goto out_freeiov;
1900 total_len = err;
1901
1902 err = -ENOBUFS;
1903
1904 if (msg_sys.msg_controllen > INT_MAX)
1905 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001906 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908 err =
1909 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1910 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 if (err)
1912 goto out_freeiov;
1913 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001914 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001916 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001918 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 goto out_freeiov;
1920 }
1921 err = -EFAULT;
1922 /*
1923 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1924 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1925 * checking falls down on this.
1926 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001927 if (copy_from_user(ctl_buf,
1928 (void __user __force *)msg_sys.msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001929 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 goto out_freectl;
1931 msg_sys.msg_control = ctl_buf;
1932 }
1933 msg_sys.msg_flags = flags;
1934
1935 if (sock->file->f_flags & O_NONBLOCK)
1936 msg_sys.msg_flags |= MSG_DONTWAIT;
1937 err = sock_sendmsg(sock, &msg_sys, total_len);
1938
1939out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001940 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1942out_freeiov:
1943 if (iov != iovstack)
1944 sock_kfree_s(sock->sk, iov, iov_size);
1945out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001946 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001947out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 return err;
1949}
1950
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001951static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1952 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001954 struct compat_msghdr __user *msg_compat =
1955 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001957 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 unsigned long cmsg_ptr;
1959 int err, iov_size, total_len, len;
1960
1961 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001962 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963
1964 /* user mode address pointers */
1965 struct sockaddr __user *uaddr;
1966 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001967
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001969 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001971 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001972 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001975 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1976 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001977
1978 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001980 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1981 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1983 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001984 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 }
1986
1987 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001988 * Save the user-mode address (verify_iovec will change the
1989 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001991
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001992 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 uaddr_len = COMPAT_NAMELEN(msg);
1994 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001995 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001996 (struct sockaddr *)&addr,
1997 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 } else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001999 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002000 (struct sockaddr *)&addr,
2001 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 if (err < 0)
2003 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002004 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002006 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2007 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002008
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 if (sock->file->f_flags & O_NONBLOCK)
2010 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002011 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2012 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 if (err < 0)
2014 goto out_freeiov;
2015 len = err;
2016
2017 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002018 err = move_addr_to_user((struct sockaddr *)&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002019 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 if (err < 0)
2022 goto out_freeiov;
2023 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002024 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002025 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 if (err)
2027 goto out_freeiov;
2028 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002029 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 &msg_compat->msg_controllen);
2031 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002032 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 &msg->msg_controllen);
2034 if (err)
2035 goto out_freeiov;
2036 err = len;
2037
2038out_freeiov:
2039 if (iov != iovstack)
2040 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002041out:
2042 return err;
2043}
2044
2045/*
2046 * BSD recvmsg interface
2047 */
2048
2049SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2050 unsigned int, flags)
2051{
2052 int fput_needed, err;
2053 struct msghdr msg_sys;
2054 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2055
2056 if (!sock)
2057 goto out;
2058
2059 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2060
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002061 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062out:
2063 return err;
2064}
2065
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002066/*
2067 * Linux recvmmsg interface
2068 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002070int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2071 unsigned int flags, struct timespec *timeout)
2072{
2073 int fput_needed, err, datagrams;
2074 struct socket *sock;
2075 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002076 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002077 struct msghdr msg_sys;
2078 struct timespec end_time;
2079
2080 if (timeout &&
2081 poll_select_set_timeout(&end_time, timeout->tv_sec,
2082 timeout->tv_nsec))
2083 return -EINVAL;
2084
2085 datagrams = 0;
2086
2087 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2088 if (!sock)
2089 return err;
2090
2091 err = sock_error(sock->sk);
2092 if (err)
2093 goto out_put;
2094
2095 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002096 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002097
2098 while (datagrams < vlen) {
2099 /*
2100 * No need to ask LSM for more than the first datagram.
2101 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002102 if (MSG_CMSG_COMPAT & flags) {
2103 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2104 &msg_sys, flags, datagrams);
2105 if (err < 0)
2106 break;
2107 err = __put_user(err, &compat_entry->msg_len);
2108 ++compat_entry;
2109 } else {
2110 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2111 &msg_sys, flags, datagrams);
2112 if (err < 0)
2113 break;
2114 err = put_user(err, &entry->msg_len);
2115 ++entry;
2116 }
2117
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002118 if (err)
2119 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002120 ++datagrams;
2121
Brandon L Black71c5c152010-03-26 16:18:03 +00002122 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2123 if (flags & MSG_WAITFORONE)
2124 flags |= MSG_DONTWAIT;
2125
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002126 if (timeout) {
2127 ktime_get_ts(timeout);
2128 *timeout = timespec_sub(end_time, *timeout);
2129 if (timeout->tv_sec < 0) {
2130 timeout->tv_sec = timeout->tv_nsec = 0;
2131 break;
2132 }
2133
2134 /* Timeout, return less than vlen datagrams */
2135 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2136 break;
2137 }
2138
2139 /* Out of band data, return right away */
2140 if (msg_sys.msg_flags & MSG_OOB)
2141 break;
2142 }
2143
2144out_put:
2145 fput_light(sock->file, fput_needed);
2146
2147 if (err == 0)
2148 return datagrams;
2149
2150 if (datagrams != 0) {
2151 /*
2152 * We may return less entries than requested (vlen) if the
2153 * sock is non block and there aren't enough datagrams...
2154 */
2155 if (err != -EAGAIN) {
2156 /*
2157 * ... or if recvmsg returns an error after we
2158 * received some datagrams, where we record the
2159 * error to return on the next call or if the
2160 * app asks about it using getsockopt(SO_ERROR).
2161 */
2162 sock->sk->sk_err = -err;
2163 }
2164
2165 return datagrams;
2166 }
2167
2168 return err;
2169}
2170
2171SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2172 unsigned int, vlen, unsigned int, flags,
2173 struct timespec __user *, timeout)
2174{
2175 int datagrams;
2176 struct timespec timeout_sys;
2177
2178 if (!timeout)
2179 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2180
2181 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2182 return -EFAULT;
2183
2184 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2185
2186 if (datagrams > 0 &&
2187 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2188 datagrams = -EFAULT;
2189
2190 return datagrams;
2191}
2192
2193#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194/* Argument list sizes for sys_socketcall */
2195#define AL(x) ((x) * sizeof(unsigned long))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002196static const unsigned char nargs[20] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002197 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2198 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2199 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2200 AL(4), AL(5)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002201};
2202
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203#undef AL
2204
2205/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002206 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207 *
2208 * Argument checking cleaned up. Saved 20% in size.
2209 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002210 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 */
2212
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002213SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214{
2215 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002216 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002218 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002220 if (call < 1 || call > SYS_RECVMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 return -EINVAL;
2222
Arjan van de Ven47379052009-09-28 12:57:44 -07002223 len = nargs[call];
2224 if (len > sizeof(a))
2225 return -EINVAL;
2226
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002228 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002230
Al Virof3298dc2008-12-10 03:16:51 -05002231 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002232
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002233 a0 = a[0];
2234 a1 = a[1];
2235
2236 switch (call) {
2237 case SYS_SOCKET:
2238 err = sys_socket(a0, a1, a[2]);
2239 break;
2240 case SYS_BIND:
2241 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2242 break;
2243 case SYS_CONNECT:
2244 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2245 break;
2246 case SYS_LISTEN:
2247 err = sys_listen(a0, a1);
2248 break;
2249 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002250 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2251 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002252 break;
2253 case SYS_GETSOCKNAME:
2254 err =
2255 sys_getsockname(a0, (struct sockaddr __user *)a1,
2256 (int __user *)a[2]);
2257 break;
2258 case SYS_GETPEERNAME:
2259 err =
2260 sys_getpeername(a0, (struct sockaddr __user *)a1,
2261 (int __user *)a[2]);
2262 break;
2263 case SYS_SOCKETPAIR:
2264 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2265 break;
2266 case SYS_SEND:
2267 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2268 break;
2269 case SYS_SENDTO:
2270 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2271 (struct sockaddr __user *)a[4], a[5]);
2272 break;
2273 case SYS_RECV:
2274 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2275 break;
2276 case SYS_RECVFROM:
2277 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2278 (struct sockaddr __user *)a[4],
2279 (int __user *)a[5]);
2280 break;
2281 case SYS_SHUTDOWN:
2282 err = sys_shutdown(a0, a1);
2283 break;
2284 case SYS_SETSOCKOPT:
2285 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2286 break;
2287 case SYS_GETSOCKOPT:
2288 err =
2289 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2290 (int __user *)a[4]);
2291 break;
2292 case SYS_SENDMSG:
2293 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2294 break;
2295 case SYS_RECVMSG:
2296 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2297 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002298 case SYS_RECVMMSG:
2299 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2300 (struct timespec __user *)a[4]);
2301 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002302 case SYS_ACCEPT4:
2303 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2304 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002305 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002306 default:
2307 err = -EINVAL;
2308 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 }
2310 return err;
2311}
2312
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002313#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002315/**
2316 * sock_register - add a socket protocol handler
2317 * @ops: description of protocol
2318 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319 * This function is called by a protocol handler that wants to
2320 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002321 * socket interface. The value ops->family coresponds to the
2322 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002324int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325{
2326 int err;
2327
2328 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002329 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2330 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331 return -ENOBUFS;
2332 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002333
2334 spin_lock(&net_family_lock);
2335 if (net_families[ops->family])
2336 err = -EEXIST;
2337 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002338 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339 err = 0;
2340 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002341 spin_unlock(&net_family_lock);
2342
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002343 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 return err;
2345}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002346EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002348/**
2349 * sock_unregister - remove a protocol handler
2350 * @family: protocol family to remove
2351 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 * This function is called by a protocol handler that wants to
2353 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002354 * new socket creation.
2355 *
2356 * If protocol handler is a module, then it can use module reference
2357 * counts to protect against new references. If protocol handler is not
2358 * a module then it needs to provide its own protection in
2359 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002361void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002363 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002365 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002366 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002367 spin_unlock(&net_family_lock);
2368
2369 synchronize_rcu();
2370
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002371 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002373EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374
Andi Kleen77d76ea2005-12-22 12:43:42 -08002375static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376{
2377 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002378 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002380
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381 sk_init();
2382
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002384 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385 */
2386 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387
2388 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002389 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 */
2391
2392 init_inodecache();
2393 register_filesystem(&sock_fs_type);
2394 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002395
2396 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397 */
2398
2399#ifdef CONFIG_NETFILTER
2400 netfilter_init();
2401#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002402
Richard Cochranc1f19b52010-07-17 08:49:36 +00002403#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2404 skb_timestamping_init();
2405#endif
2406
David S. Millercbeb3212005-12-22 12:58:55 -08002407 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408}
2409
Andi Kleen77d76ea2005-12-22 12:43:42 -08002410core_initcall(sock_init); /* early initcall */
2411
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412#ifdef CONFIG_PROC_FS
2413void socket_seq_show(struct seq_file *seq)
2414{
2415 int cpu;
2416 int counter = 0;
2417
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002418 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002419 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002420
2421 /* It can be negative, by the way. 8) */
2422 if (counter < 0)
2423 counter = 0;
2424
2425 seq_printf(seq, "sockets: used %d\n", counter);
2426}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002427#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002429#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002430static int do_siocgstamp(struct net *net, struct socket *sock,
2431 unsigned int cmd, struct compat_timeval __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002432{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002433 mm_segment_t old_fs = get_fs();
2434 struct timeval ktv;
2435 int err;
2436
2437 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002438 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002439 set_fs(old_fs);
2440 if (!err) {
2441 err = put_user(ktv.tv_sec, &up->tv_sec);
2442 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2443 }
2444 return err;
2445}
2446
Arnd Bergmann6b960182009-11-06 23:10:54 -08002447static int do_siocgstampns(struct net *net, struct socket *sock,
2448 unsigned int cmd, struct compat_timespec __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002449{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002450 mm_segment_t old_fs = get_fs();
2451 struct timespec kts;
2452 int err;
2453
2454 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002455 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002456 set_fs(old_fs);
2457 if (!err) {
2458 err = put_user(kts.tv_sec, &up->tv_sec);
2459 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2460 }
2461 return err;
2462}
2463
Arnd Bergmann6b960182009-11-06 23:10:54 -08002464static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002465{
2466 struct ifreq __user *uifr;
2467 int err;
2468
2469 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002470 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002471 return -EFAULT;
2472
Arnd Bergmann6b960182009-11-06 23:10:54 -08002473 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002474 if (err)
2475 return err;
2476
Arnd Bergmann6b960182009-11-06 23:10:54 -08002477 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002478 return -EFAULT;
2479
2480 return 0;
2481}
2482
Arnd Bergmann6b960182009-11-06 23:10:54 -08002483static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002484{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002485 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002486 struct ifconf ifc;
2487 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002488 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002489 struct ifreq __user *ifr;
2490 unsigned int i, j;
2491 int err;
2492
Arnd Bergmann6b960182009-11-06 23:10:54 -08002493 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002494 return -EFAULT;
2495
2496 if (ifc32.ifcbuf == 0) {
2497 ifc32.ifc_len = 0;
2498 ifc.ifc_len = 0;
2499 ifc.ifc_req = NULL;
2500 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2501 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002502 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2503 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002504 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2505 ifc.ifc_len = len;
2506 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2507 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002508 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002509 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002510 return -EFAULT;
2511 ifr++;
2512 ifr32++;
2513 }
2514 }
2515 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2516 return -EFAULT;
2517
Arnd Bergmann6b960182009-11-06 23:10:54 -08002518 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002519 if (err)
2520 return err;
2521
2522 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2523 return -EFAULT;
2524
2525 ifr = ifc.ifc_req;
2526 ifr32 = compat_ptr(ifc32.ifcbuf);
2527 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002528 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2529 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2530 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002531 return -EFAULT;
2532 ifr32++;
2533 ifr++;
2534 }
2535
2536 if (ifc32.ifcbuf == 0) {
2537 /* Translate from 64-bit structure multiple to
2538 * a 32-bit one.
2539 */
2540 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002541 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002542 ifc32.ifc_len = i;
2543 } else {
2544 ifc32.ifc_len = i;
2545 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002546 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002547 return -EFAULT;
2548
2549 return 0;
2550}
2551
Arnd Bergmann6b960182009-11-06 23:10:54 -08002552static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002553{
2554 struct ifreq __user *ifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002555 u32 data;
2556 void __user *datap;
2557
2558 ifr = compat_alloc_user_space(sizeof(*ifr));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002559
2560 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2561 return -EFAULT;
2562
2563 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2564 return -EFAULT;
2565
2566 datap = compat_ptr(data);
2567 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2568 return -EFAULT;
2569
Arnd Bergmann6b960182009-11-06 23:10:54 -08002570 return dev_ioctl(net, SIOCETHTOOL, ifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002571}
2572
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002573static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2574{
2575 void __user *uptr;
2576 compat_uptr_t uptr32;
2577 struct ifreq __user *uifr;
2578
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002579 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002580 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2581 return -EFAULT;
2582
2583 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2584 return -EFAULT;
2585
2586 uptr = compat_ptr(uptr32);
2587
2588 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2589 return -EFAULT;
2590
2591 return dev_ioctl(net, SIOCWANDEV, uifr);
2592}
2593
Arnd Bergmann6b960182009-11-06 23:10:54 -08002594static int bond_ioctl(struct net *net, unsigned int cmd,
2595 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002596{
2597 struct ifreq kifr;
2598 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002599 mm_segment_t old_fs;
2600 int err;
2601 u32 data;
2602 void __user *datap;
2603
2604 switch (cmd) {
2605 case SIOCBONDENSLAVE:
2606 case SIOCBONDRELEASE:
2607 case SIOCBONDSETHWADDR:
2608 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002609 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002610 return -EFAULT;
2611
2612 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002613 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002614 err = dev_ioctl(net, cmd, &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002615 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002616
2617 return err;
2618 case SIOCBONDSLAVEINFOQUERY:
2619 case SIOCBONDINFOQUERY:
2620 uifr = compat_alloc_user_space(sizeof(*uifr));
2621 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2622 return -EFAULT;
2623
2624 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2625 return -EFAULT;
2626
2627 datap = compat_ptr(data);
2628 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2629 return -EFAULT;
2630
Arnd Bergmann6b960182009-11-06 23:10:54 -08002631 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002632 default:
2633 return -EINVAL;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002634 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002635}
2636
Arnd Bergmann6b960182009-11-06 23:10:54 -08002637static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2638 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002639{
2640 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002641 char tmp_buf[IFNAMSIZ];
2642 void __user *data64;
2643 u32 data32;
2644
2645 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2646 IFNAMSIZ))
2647 return -EFAULT;
2648 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2649 return -EFAULT;
2650 data64 = compat_ptr(data32);
2651
2652 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2653
2654 /* Don't check these user accesses, just let that get trapped
2655 * in the ioctl handler instead.
2656 */
2657 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2658 IFNAMSIZ))
2659 return -EFAULT;
2660 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2661 return -EFAULT;
2662
Arnd Bergmann6b960182009-11-06 23:10:54 -08002663 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002664}
2665
Arnd Bergmann6b960182009-11-06 23:10:54 -08002666static int dev_ifsioc(struct net *net, struct socket *sock,
2667 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002668{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002669 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002670 int err;
2671
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002672 uifr = compat_alloc_user_space(sizeof(*uifr));
2673 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2674 return -EFAULT;
2675
2676 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2677
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002678 if (!err) {
2679 switch (cmd) {
2680 case SIOCGIFFLAGS:
2681 case SIOCGIFMETRIC:
2682 case SIOCGIFMTU:
2683 case SIOCGIFMEM:
2684 case SIOCGIFHWADDR:
2685 case SIOCGIFINDEX:
2686 case SIOCGIFADDR:
2687 case SIOCGIFBRDADDR:
2688 case SIOCGIFDSTADDR:
2689 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002690 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002691 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002692 case SIOCGMIIPHY:
2693 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002694 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002695 err = -EFAULT;
2696 break;
2697 }
2698 }
2699 return err;
2700}
2701
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002702static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2703 struct compat_ifreq __user *uifr32)
2704{
2705 struct ifreq ifr;
2706 struct compat_ifmap __user *uifmap32;
2707 mm_segment_t old_fs;
2708 int err;
2709
2710 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2711 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2712 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2713 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2714 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2715 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2716 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2717 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2718 if (err)
2719 return -EFAULT;
2720
2721 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002722 set_fs(KERNEL_DS);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002723 err = dev_ioctl(net, cmd, (void __user *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002724 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002725
2726 if (cmd == SIOCGIFMAP && !err) {
2727 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2728 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2729 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2730 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2731 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2732 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2733 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2734 if (err)
2735 err = -EFAULT;
2736 }
2737 return err;
2738}
2739
2740static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2741{
2742 void __user *uptr;
2743 compat_uptr_t uptr32;
2744 struct ifreq __user *uifr;
2745
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002746 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002747 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2748 return -EFAULT;
2749
2750 if (get_user(uptr32, &uifr32->ifr_data))
2751 return -EFAULT;
2752
2753 uptr = compat_ptr(uptr32);
2754
2755 if (put_user(uptr, &uifr->ifr_data))
2756 return -EFAULT;
2757
2758 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2759}
2760
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002761struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002762 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002763 struct sockaddr rt_dst; /* target address */
2764 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2765 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002766 unsigned short rt_flags;
2767 short rt_pad2;
2768 u32 rt_pad3;
2769 unsigned char rt_tos;
2770 unsigned char rt_class;
2771 short rt_pad4;
2772 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002773 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002774 u32 rt_mtu; /* per route MTU/Window */
2775 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002776 unsigned short rt_irtt; /* Initial RTT */
2777};
2778
2779struct in6_rtmsg32 {
2780 struct in6_addr rtmsg_dst;
2781 struct in6_addr rtmsg_src;
2782 struct in6_addr rtmsg_gateway;
2783 u32 rtmsg_type;
2784 u16 rtmsg_dst_len;
2785 u16 rtmsg_src_len;
2786 u32 rtmsg_metric;
2787 u32 rtmsg_info;
2788 u32 rtmsg_flags;
2789 s32 rtmsg_ifindex;
2790};
2791
Arnd Bergmann6b960182009-11-06 23:10:54 -08002792static int routing_ioctl(struct net *net, struct socket *sock,
2793 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002794{
2795 int ret;
2796 void *r = NULL;
2797 struct in6_rtmsg r6;
2798 struct rtentry r4;
2799 char devname[16];
2800 u32 rtdev;
2801 mm_segment_t old_fs = get_fs();
2802
Arnd Bergmann6b960182009-11-06 23:10:54 -08002803 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2804 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002805 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002806 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002807 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2808 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2809 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2810 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2811 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2812 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2813 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002814
2815 r = (void *) &r6;
2816 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002817 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002818 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002819 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002820 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2821 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2822 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2823 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2824 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2825 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002826 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002827 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002828 r4.rt_dev = devname; devname[15] = 0;
2829 } else
2830 r4.rt_dev = NULL;
2831
2832 r = (void *) &r4;
2833 }
2834
2835 if (ret) {
2836 ret = -EFAULT;
2837 goto out;
2838 }
2839
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002840 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002841 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002842 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002843
2844out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002845 return ret;
2846}
2847
2848/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2849 * for some operations; this forces use of the newer bridge-utils that
2850 * use compatiable ioctls
2851 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002852static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002853{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002854 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002855
Arnd Bergmann6b960182009-11-06 23:10:54 -08002856 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002857 return -EFAULT;
2858 if (tmp == BRCTL_GET_VERSION)
2859 return BRCTL_VERSION + 1;
2860 return -EINVAL;
2861}
2862
Arnd Bergmann6b960182009-11-06 23:10:54 -08002863static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2864 unsigned int cmd, unsigned long arg)
2865{
2866 void __user *argp = compat_ptr(arg);
2867 struct sock *sk = sock->sk;
2868 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002869
Arnd Bergmann6b960182009-11-06 23:10:54 -08002870 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2871 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002872
Arnd Bergmann6b960182009-11-06 23:10:54 -08002873 switch (cmd) {
2874 case SIOCSIFBR:
2875 case SIOCGIFBR:
2876 return old_bridge_ioctl(argp);
2877 case SIOCGIFNAME:
2878 return dev_ifname32(net, argp);
2879 case SIOCGIFCONF:
2880 return dev_ifconf(net, argp);
2881 case SIOCETHTOOL:
2882 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002883 case SIOCWANDEV:
2884 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002885 case SIOCGIFMAP:
2886 case SIOCSIFMAP:
2887 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002888 case SIOCBONDENSLAVE:
2889 case SIOCBONDRELEASE:
2890 case SIOCBONDSETHWADDR:
2891 case SIOCBONDSLAVEINFOQUERY:
2892 case SIOCBONDINFOQUERY:
2893 case SIOCBONDCHANGEACTIVE:
2894 return bond_ioctl(net, cmd, argp);
2895 case SIOCADDRT:
2896 case SIOCDELRT:
2897 return routing_ioctl(net, sock, cmd, argp);
2898 case SIOCGSTAMP:
2899 return do_siocgstamp(net, sock, cmd, argp);
2900 case SIOCGSTAMPNS:
2901 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002902 case SIOCSHWTSTAMP:
2903 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002904
Arnd Bergmann6b960182009-11-06 23:10:54 -08002905 case FIOSETOWN:
2906 case SIOCSPGRP:
2907 case FIOGETOWN:
2908 case SIOCGPGRP:
2909 case SIOCBRADDBR:
2910 case SIOCBRDELBR:
2911 case SIOCGIFVLAN:
2912 case SIOCSIFVLAN:
2913 case SIOCADDDLCI:
2914 case SIOCDELDLCI:
2915 return sock_ioctl(file, cmd, arg);
2916
2917 case SIOCGIFFLAGS:
2918 case SIOCSIFFLAGS:
2919 case SIOCGIFMETRIC:
2920 case SIOCSIFMETRIC:
2921 case SIOCGIFMTU:
2922 case SIOCSIFMTU:
2923 case SIOCGIFMEM:
2924 case SIOCSIFMEM:
2925 case SIOCGIFHWADDR:
2926 case SIOCSIFHWADDR:
2927 case SIOCADDMULTI:
2928 case SIOCDELMULTI:
2929 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002930 case SIOCGIFADDR:
2931 case SIOCSIFADDR:
2932 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002933 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002934 case SIOCGIFBRDADDR:
2935 case SIOCSIFBRDADDR:
2936 case SIOCGIFDSTADDR:
2937 case SIOCSIFDSTADDR:
2938 case SIOCGIFNETMASK:
2939 case SIOCSIFNETMASK:
2940 case SIOCSIFPFLAGS:
2941 case SIOCGIFPFLAGS:
2942 case SIOCGIFTXQLEN:
2943 case SIOCSIFTXQLEN:
2944 case SIOCBRADDIF:
2945 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002946 case SIOCSIFNAME:
2947 case SIOCGMIIPHY:
2948 case SIOCGMIIREG:
2949 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002950 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002951
Arnd Bergmann6b960182009-11-06 23:10:54 -08002952 case SIOCSARP:
2953 case SIOCGARP:
2954 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002955 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002956 return sock_do_ioctl(net, sock, cmd, arg);
2957 }
2958
2959 /* Prevent warning from compat_sys_ioctl, these always
2960 * result in -EINVAL in the native case anyway. */
2961 switch (cmd) {
2962 case SIOCRTMSG:
2963 case SIOCGIFCOUNT:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002964 case SIOCSRARP:
2965 case SIOCGRARP:
2966 case SIOCDRARP:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002967 case SIOCSIFLINK:
2968 case SIOCGIFSLAVE:
2969 case SIOCSIFSLAVE:
2970 return -EINVAL;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002971 }
2972
2973 return -ENOIOCTLCMD;
2974}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002975
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002976static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002977 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002978{
2979 struct socket *sock = file->private_data;
2980 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07002981 struct sock *sk;
2982 struct net *net;
2983
2984 sk = sock->sk;
2985 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002986
2987 if (sock->ops->compat_ioctl)
2988 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2989
David S. Miller87de87d2008-06-03 09:14:03 -07002990 if (ret == -ENOIOCTLCMD &&
2991 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2992 ret = compat_wext_handle_ioctl(net, cmd, arg);
2993
Arnd Bergmann6b960182009-11-06 23:10:54 -08002994 if (ret == -ENOIOCTLCMD)
2995 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2996
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002997 return ret;
2998}
2999#endif
3000
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003001int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3002{
3003 return sock->ops->bind(sock, addr, addrlen);
3004}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003005EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003006
3007int kernel_listen(struct socket *sock, int backlog)
3008{
3009 return sock->ops->listen(sock, backlog);
3010}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003011EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003012
3013int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3014{
3015 struct sock *sk = sock->sk;
3016 int err;
3017
3018 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3019 newsock);
3020 if (err < 0)
3021 goto done;
3022
3023 err = sock->ops->accept(sock, *newsock, flags);
3024 if (err < 0) {
3025 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003026 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003027 goto done;
3028 }
3029
3030 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003031 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003032
3033done:
3034 return err;
3035}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003036EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003037
3038int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003039 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003040{
3041 return sock->ops->connect(sock, addr, addrlen, flags);
3042}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003043EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003044
3045int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3046 int *addrlen)
3047{
3048 return sock->ops->getname(sock, addr, addrlen, 0);
3049}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003050EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003051
3052int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3053 int *addrlen)
3054{
3055 return sock->ops->getname(sock, addr, addrlen, 1);
3056}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003057EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003058
3059int kernel_getsockopt(struct socket *sock, int level, int optname,
3060 char *optval, int *optlen)
3061{
3062 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003063 char __user *uoptval;
3064 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003065 int err;
3066
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003067 uoptval = (char __user __force *) optval;
3068 uoptlen = (int __user __force *) optlen;
3069
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003070 set_fs(KERNEL_DS);
3071 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003072 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003073 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003074 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3075 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003076 set_fs(oldfs);
3077 return err;
3078}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003079EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003080
3081int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003082 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003083{
3084 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003085 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003086 int err;
3087
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003088 uoptval = (char __user __force *) optval;
3089
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003090 set_fs(KERNEL_DS);
3091 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003092 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003093 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003094 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003095 optlen);
3096 set_fs(oldfs);
3097 return err;
3098}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003099EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003100
3101int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3102 size_t size, int flags)
3103{
Herbert Xuf8451722010-05-24 00:12:34 -07003104 sock_update_classid(sock->sk);
3105
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003106 if (sock->ops->sendpage)
3107 return sock->ops->sendpage(sock, page, offset, size, flags);
3108
3109 return sock_no_sendpage(sock, page, offset, size, flags);
3110}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003111EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003112
3113int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3114{
3115 mm_segment_t oldfs = get_fs();
3116 int err;
3117
3118 set_fs(KERNEL_DS);
3119 err = sock->ops->ioctl(sock, cmd, arg);
3120 set_fs(oldfs);
3121
3122 return err;
3123}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003124EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003125
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003126int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3127{
3128 return sock->ops->shutdown(sock, how);
3129}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003130EXPORT_SYMBOL(kernel_sock_shutdown);