blob: ccc576a6a5083135001666fcc08523034549540a [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);
Eric Dumazet190683a2010-11-10 10:50:44 +0000159static const struct net_proto_family __rcu *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
Eric Dumazet43815482010-04-29 11:01:49 +0000265
Eric Dumazet43815482010-04-29 11:01:49 +0000266static void wq_free_rcu(struct rcu_head *head)
267{
268 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
269
270 kfree(wq);
271}
272
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273static void sock_destroy_inode(struct inode *inode)
274{
Eric Dumazet43815482010-04-29 11:01:49 +0000275 struct socket_alloc *ei;
276
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
278 call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
279 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280}
281
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700282static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700284 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
Christoph Lametera35afb82007-05-16 22:10:57 -0700286 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700288
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289static int init_inodecache(void)
290{
291 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700292 sizeof(struct socket_alloc),
293 0,
294 (SLAB_HWCACHE_ALIGN |
295 SLAB_RECLAIM_ACCOUNT |
296 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900297 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 if (sock_inode_cachep == NULL)
299 return -ENOMEM;
300 return 0;
301}
302
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700303static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700304 .alloc_inode = sock_alloc_inode,
305 .destroy_inode = sock_destroy_inode,
306 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307};
308
Al Viro51139ad2010-07-25 23:47:46 +0400309static struct dentry *sockfs_mount(struct file_system_type *fs_type,
310 int flags, const char *dev_name, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311{
Al Viro51139ad2010-07-25 23:47:46 +0400312 return mount_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313}
314
Eric Dumazetba899662005-08-26 12:05:31 -0700315static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316
317static struct file_system_type sock_fs_type = {
318 .name = "sockfs",
Al Viro51139ad2010-07-25 23:47:46 +0400319 .mount = sockfs_mount,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320 .kill_sb = kill_anon_super,
321};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700322
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700323/*
324 * sockfs_dname() is called from d_path().
325 */
326static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
327{
328 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
329 dentry->d_inode->i_ino);
330}
331
Al Viro3ba13d12009-02-20 06:02:22 +0000332static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700333 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334};
335
336/*
337 * Obtains the first available file descriptor and sets it up for use.
338 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800339 * These functions create file structures and maps them to fd space
340 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341 * and file struct implicitly stored in sock->file.
342 * Note that another thread may close file descriptor before we return
343 * from this function. We use the fact that now we do not refer
344 * to socket after mapping. If one day we will need it, this
345 * function will increment ref. count on file by 1.
346 *
347 * In any case returned fd MAY BE not valid!
348 * This race condition is unavoidable
349 * with shared fd spaces, we cannot solve it inside kernel,
350 * but we take care of internal coherence yet.
351 */
352
Al Viro7cbe66b2009-08-05 19:59:08 +0400353static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354{
Al Viro7cbe66b2009-08-05 19:59:08 +0400355 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400356 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400357 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800359
Ulrich Dreppera677a032008-07-23 21:29:17 -0700360 fd = get_unused_fd_flags(flags);
Al Viro7cbe66b2009-08-05 19:59:08 +0400361 if (unlikely(fd < 0))
362 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800363
Nick Piggin4b936882011-01-07 17:50:07 +1100364 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro2c48b9c2009-08-09 00:52:35 +0400365 if (unlikely(!path.dentry)) {
Al Viro7cbe66b2009-08-05 19:59:08 +0400366 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800367 return -ENOMEM;
Al Viro7cbe66b2009-08-05 19:59:08 +0400368 }
Al Viro2c48b9c2009-08-09 00:52:35 +0400369 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370
Nick Pigginfb045ad2011-01-07 17:49:55 +1100371 d_set_d_op(path.dentry, &sockfs_dentry_operations);
Al Viro2c48b9c2009-08-09 00:52:35 +0400372 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400373 SOCK_INODE(sock)->i_fop = &socket_file_ops;
374
Al Viro2c48b9c2009-08-09 00:52:35 +0400375 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400376 &socket_file_ops);
377 if (unlikely(!file)) {
378 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400379 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400380 path_put(&path);
Al Virocc3808f2009-08-06 09:43:59 +0400381 put_unused_fd(fd);
382 return -ENFILE;
383 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800384
385 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387 file->f_pos = 0;
388 file->private_data = sock;
389
Al Viro7cbe66b2009-08-05 19:59:08 +0400390 *f = file;
391 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392}
393
Ulrich Dreppera677a032008-07-23 21:29:17 -0700394int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800395{
396 struct file *newfile;
Al Viro7cbe66b2009-08-05 19:59:08 +0400397 int fd = sock_alloc_file(sock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800398
Al Viro7cbe66b2009-08-05 19:59:08 +0400399 if (likely(fd >= 0))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800400 fd_install(fd, newfile);
Al Viro7cbe66b2009-08-05 19:59:08 +0400401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 return fd;
403}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700404EXPORT_SYMBOL(sock_map_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800406static struct socket *sock_from_file(struct file *file, int *err)
407{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800408 if (file->f_op == &socket_file_ops)
409 return file->private_data; /* set in sock_map_fd */
410
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800411 *err = -ENOTSOCK;
412 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800413}
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700416 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 * @fd: file handle
418 * @err: pointer to an error code return
419 *
420 * The file handle passed in is locked and the socket it is bound
421 * too is returned. If an error occurs the err pointer is overwritten
422 * with a negative errno code and NULL is returned. The function checks
423 * for both invalid handles and passing a handle which is not a socket.
424 *
425 * On a success the socket object pointer is returned.
426 */
427
428struct socket *sockfd_lookup(int fd, int *err)
429{
430 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 struct socket *sock;
432
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700433 file = fget(fd);
434 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 *err = -EBADF;
436 return NULL;
437 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700438
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800439 sock = sock_from_file(file, err);
440 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 return sock;
443}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700444EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800446static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
447{
448 struct file *file;
449 struct socket *sock;
450
Hua Zhong36725582006-04-19 15:25:02 -0700451 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800452 file = fget_light(fd, fput_needed);
453 if (file) {
454 sock = sock_from_file(file, err);
455 if (sock)
456 return sock;
457 fput_light(file, *fput_needed);
458 }
459 return NULL;
460}
461
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462/**
463 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700464 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 * Allocate a new inode and socket object. The two are bound together
466 * and initialised. The socket is then returned. If we are out of inodes
467 * NULL is returned.
468 */
469
470static struct socket *sock_alloc(void)
471{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700472 struct inode *inode;
473 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
475 inode = new_inode(sock_mnt->mnt_sb);
476 if (!inode)
477 return NULL;
478
479 sock = SOCKET_I(inode);
480
Eric Dumazet29a020d2009-09-15 02:39:20 -0700481 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400482 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700483 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100484 inode->i_uid = current_fsuid();
485 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Eric Dumazet4e694892009-04-04 16:41:09 -0700487 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 return sock;
489}
490
491/*
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
495 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
498{
499 return -ENXIO;
500}
501
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800502const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 .owner = THIS_MODULE,
504 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200505 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506};
507
508/**
509 * sock_release - close a socket
510 * @sock: socket to close
511 *
512 * The socket is released from the protocol stack if it has a release
513 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700514 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517void sock_release(struct socket *sock)
518{
519 if (sock->ops) {
520 struct module *owner = sock->ops->owner;
521
522 sock->ops->release(sock);
523 sock->ops = NULL;
524 module_put(owner);
525 }
526
Eric Dumazet43815482010-04-29 11:01:49 +0000527 if (sock->wq->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 printk(KERN_ERR "sock_release: fasync list not empty!\n");
529
Eric Dumazet4e694892009-04-04 16:41:09 -0700530 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 if (!sock->file) {
532 iput(SOCK_INODE(sock));
533 return;
534 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700535 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700537EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000539int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000540{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000541 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000542 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000543 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000544 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000545 *tx_flags |= SKBTX_SW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000546 return 0;
547}
548EXPORT_SYMBOL(sock_tx_timestamp);
549
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700550static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 struct msghdr *msg, size_t size)
552{
553 struct sock_iocb *si = kiocb_to_siocb(iocb);
554 int err;
555
Herbert Xuf8451722010-05-24 00:12:34 -0700556 sock_update_classid(sock->sk);
557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 si->sock = sock;
559 si->scm = NULL;
560 si->msg = msg;
561 si->size = size;
562
563 err = security_socket_sendmsg(sock, msg, size);
564 if (err)
565 return err;
566
567 return sock->ops->sendmsg(iocb, sock, msg, size);
568}
569
570int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
571{
572 struct kiocb iocb;
573 struct sock_iocb siocb;
574 int ret;
575
576 init_sync_kiocb(&iocb, NULL);
577 iocb.private = &siocb;
578 ret = __sock_sendmsg(&iocb, sock, msg, size);
579 if (-EIOCBQUEUED == ret)
580 ret = wait_on_sync_kiocb(&iocb);
581 return ret;
582}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700583EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
585int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
586 struct kvec *vec, size_t num, size_t size)
587{
588 mm_segment_t oldfs = get_fs();
589 int result;
590
591 set_fs(KERNEL_DS);
592 /*
593 * the following is safe, since for compiler definitions of kvec and
594 * iovec are identical, yielding the same in-core layout and alignment
595 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700596 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 msg->msg_iovlen = num;
598 result = sock_sendmsg(sock, msg, size);
599 set_fs(oldfs);
600 return result;
601}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700602EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
Patrick Ohly20d49472009-02-12 05:03:38 +0000604static int ktime2ts(ktime_t kt, struct timespec *ts)
605{
606 if (kt.tv64) {
607 *ts = ktime_to_timespec(kt);
608 return 1;
609 } else {
610 return 0;
611 }
612}
613
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700614/*
615 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
616 */
617void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
618 struct sk_buff *skb)
619{
Patrick Ohly20d49472009-02-12 05:03:38 +0000620 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
621 struct timespec ts[3];
622 int empty = 1;
623 struct skb_shared_hwtstamps *shhwtstamps =
624 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700625
Patrick Ohly20d49472009-02-12 05:03:38 +0000626 /* Race occurred between timestamp enabling and packet
627 receiving. Fill in the current time for now. */
628 if (need_software_tstamp && skb->tstamp.tv64 == 0)
629 __net_timestamp(skb);
630
631 if (need_software_tstamp) {
632 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
633 struct timeval tv;
634 skb_get_timestamp(skb, &tv);
635 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
636 sizeof(tv), &tv);
637 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000638 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000639 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000640 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000641 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700642 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000643
644
645 memset(ts, 0, sizeof(ts));
646 if (skb->tstamp.tv64 &&
647 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
648 skb_get_timestampns(skb, ts + 0);
649 empty = 0;
650 }
651 if (shhwtstamps) {
652 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
653 ktime2ts(shhwtstamps->syststamp, ts + 1))
654 empty = 0;
655 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
656 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
657 empty = 0;
658 }
659 if (!empty)
660 put_cmsg(msg, SOL_SOCKET,
661 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700662}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300663EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
664
stephen hemminger11165f12010-10-18 14:27:29 +0000665static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
666 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700667{
668 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
669 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
670 sizeof(__u32), &skb->dropcount);
671}
672
Eric Dumazet767dd032010-04-28 19:14:43 +0000673void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700674 struct sk_buff *skb)
675{
676 sock_recv_timestamp(msg, sk, skb);
677 sock_recv_drops(msg, sk, skb);
678}
Eric Dumazet767dd032010-04-28 19:14:43 +0000679EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700680
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700681static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
682 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 struct sock_iocb *si = kiocb_to_siocb(iocb);
685
Herbert Xuf8451722010-05-24 00:12:34 -0700686 sock_update_classid(sock->sk);
687
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688 si->sock = sock;
689 si->scm = NULL;
690 si->msg = msg;
691 si->size = size;
692 si->flags = flags;
693
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
695}
696
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700697static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
698 struct msghdr *msg, size_t size, int flags)
699{
700 int err = security_socket_recvmsg(sock, msg, size, flags);
701
702 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
703}
704
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700705int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706 size_t size, int flags)
707{
708 struct kiocb iocb;
709 struct sock_iocb siocb;
710 int ret;
711
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700712 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 iocb.private = &siocb;
714 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
715 if (-EIOCBQUEUED == ret)
716 ret = wait_on_sync_kiocb(&iocb);
717 return ret;
718}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700719EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700721static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
722 size_t size, int flags)
723{
724 struct kiocb iocb;
725 struct sock_iocb siocb;
726 int ret;
727
728 init_sync_kiocb(&iocb, NULL);
729 iocb.private = &siocb;
730 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
731 if (-EIOCBQUEUED == ret)
732 ret = wait_on_sync_kiocb(&iocb);
733 return ret;
734}
735
Martin Lucinac1249c02010-12-10 00:04:05 +0000736/**
737 * kernel_recvmsg - Receive a message from a socket (kernel space)
738 * @sock: The socket to receive the message from
739 * @msg: Received message
740 * @vec: Input s/g array for message data
741 * @num: Size of input s/g array
742 * @size: Number of bytes to read
743 * @flags: Message flags (MSG_DONTWAIT, etc...)
744 *
745 * On return the msg structure contains the scatter/gather array passed in the
746 * vec argument. The array is modified so that it consists of the unfilled
747 * portion of the original array.
748 *
749 * The returned value is the total number of bytes received, or an error.
750 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700751int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
752 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753{
754 mm_segment_t oldfs = get_fs();
755 int result;
756
757 set_fs(KERNEL_DS);
758 /*
759 * the following is safe, since for compiler definitions of kvec and
760 * iovec are identical, yielding the same in-core layout and alignment
761 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700762 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 result = sock_recvmsg(sock, msg, size, flags);
764 set_fs(oldfs);
765 return result;
766}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700767EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
769static void sock_aio_dtor(struct kiocb *iocb)
770{
771 kfree(iocb->private);
772}
773
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300774static ssize_t sock_sendpage(struct file *file, struct page *page,
775 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776{
777 struct socket *sock;
778 int flags;
779
Eric Dumazetb69aee02005-09-06 14:42:45 -0700780 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
783 if (more)
784 flags |= MSG_MORE;
785
Linus Torvaldse6949582009-08-13 08:28:36 -0700786 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787}
788
Jens Axboe9c55e012007-11-06 23:30:13 -0800789static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700790 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800791 unsigned int flags)
792{
793 struct socket *sock = file->private_data;
794
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800795 if (unlikely(!sock->ops->splice_read))
796 return -EINVAL;
797
Herbert Xuf8451722010-05-24 00:12:34 -0700798 sock_update_classid(sock->sk);
799
Jens Axboe9c55e012007-11-06 23:30:13 -0800800 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
801}
802
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800803static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700804 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800805{
806 if (!is_sync_kiocb(iocb)) {
807 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
808 if (!siocb)
809 return NULL;
810 iocb->ki_dtor = sock_aio_dtor;
811 }
812
813 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800814 iocb->private = siocb;
815 return siocb;
816}
817
818static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700819 struct file *file, const struct iovec *iov,
820 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800821{
822 struct socket *sock = file->private_data;
823 size_t size = 0;
824 int i;
825
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700826 for (i = 0; i < nr_segs; i++)
827 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800828
829 msg->msg_name = NULL;
830 msg->msg_namelen = 0;
831 msg->msg_control = NULL;
832 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700833 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800834 msg->msg_iovlen = nr_segs;
835 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
836
837 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
838}
839
Badari Pulavarty027445c2006-09-30 23:28:46 -0700840static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
841 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800842{
843 struct sock_iocb siocb, *x;
844
845 if (pos != 0)
846 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700847
848 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800849 return 0;
850
Badari Pulavarty027445c2006-09-30 23:28:46 -0700851
852 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800853 if (!x)
854 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700855 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800856}
857
858static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700859 struct file *file, const struct iovec *iov,
860 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800861{
862 struct socket *sock = file->private_data;
863 size_t size = 0;
864 int i;
865
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700866 for (i = 0; i < nr_segs; i++)
867 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800868
869 msg->msg_name = NULL;
870 msg->msg_namelen = 0;
871 msg->msg_control = NULL;
872 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700873 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800874 msg->msg_iovlen = nr_segs;
875 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
876 if (sock->type == SOCK_SEQPACKET)
877 msg->msg_flags |= MSG_EOR;
878
879 return __sock_sendmsg(iocb, sock, msg, size);
880}
881
Badari Pulavarty027445c2006-09-30 23:28:46 -0700882static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
883 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800885 struct sock_iocb siocb, *x;
886
887 if (pos != 0)
888 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700889
Badari Pulavarty027445c2006-09-30 23:28:46 -0700890 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800891 if (!x)
892 return -ENOMEM;
893
Badari Pulavarty027445c2006-09-30 23:28:46 -0700894 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895}
896
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897/*
898 * Atomic setting of ioctl hooks to avoid race
899 * with module unload.
900 */
901
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800902static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700903static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
Eric W. Biederman881d9662007-09-17 11:56:21 -0700905void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800907 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800909 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910}
911EXPORT_SYMBOL(brioctl_set);
912
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800913static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700914static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915
Eric W. Biederman881d9662007-09-17 11:56:21 -0700916void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800918 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800920 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921}
922EXPORT_SYMBOL(vlan_ioctl_set);
923
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800924static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700925static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700927void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800929 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800931 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932}
933EXPORT_SYMBOL(dlci_ioctl_set);
934
Arnd Bergmann6b960182009-11-06 23:10:54 -0800935static long sock_do_ioctl(struct net *net, struct socket *sock,
936 unsigned int cmd, unsigned long arg)
937{
938 int err;
939 void __user *argp = (void __user *)arg;
940
941 err = sock->ops->ioctl(sock, cmd, arg);
942
943 /*
944 * If this ioctl is unknown try to hand it down
945 * to the NIC driver.
946 */
947 if (err == -ENOIOCTLCMD)
948 err = dev_ioctl(net, cmd, argp);
949
950 return err;
951}
952
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953/*
954 * With an ioctl, arg may well be a user mode pointer, but we don't know
955 * what to do with it - that's up to the protocol still.
956 */
957
958static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
959{
960 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700961 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962 void __user *argp = (void __user *)arg;
963 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700964 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
Eric Dumazetb69aee02005-09-06 14:42:45 -0700966 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700967 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900968 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700970 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200972#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700974 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200976#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700977 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 case FIOSETOWN:
979 case SIOCSPGRP:
980 err = -EFAULT;
981 if (get_user(pid, (int __user *)argp))
982 break;
983 err = f_setown(sock->file, pid, 1);
984 break;
985 case FIOGETOWN:
986 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700987 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700988 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 break;
990 case SIOCGIFBR:
991 case SIOCSIFBR:
992 case SIOCBRADDBR:
993 case SIOCBRDELBR:
994 err = -ENOPKG;
995 if (!br_ioctl_hook)
996 request_module("bridge");
997
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800998 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700999 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001000 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001001 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 break;
1003 case SIOCGIFVLAN:
1004 case SIOCSIFVLAN:
1005 err = -ENOPKG;
1006 if (!vlan_ioctl_hook)
1007 request_module("8021q");
1008
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001009 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001011 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001012 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 case SIOCADDDLCI:
1015 case SIOCDELDLCI:
1016 err = -ENOPKG;
1017 if (!dlci_ioctl_hook)
1018 request_module("dlci");
1019
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001020 mutex_lock(&dlci_ioctl_mutex);
1021 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001023 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 break;
1025 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001026 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001028 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 return err;
1030}
1031
1032int sock_create_lite(int family, int type, int protocol, struct socket **res)
1033{
1034 int err;
1035 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001036
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 err = security_socket_create(family, type, protocol, 1);
1038 if (err)
1039 goto out;
1040
1041 sock = sock_alloc();
1042 if (!sock) {
1043 err = -ENOMEM;
1044 goto out;
1045 }
1046
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001048 err = security_socket_post_create(sock, family, type, protocol, 1);
1049 if (err)
1050 goto out_release;
1051
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052out:
1053 *res = sock;
1054 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001055out_release:
1056 sock_release(sock);
1057 sock = NULL;
1058 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001060EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
1062/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001063static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064{
1065 struct socket *sock;
1066
1067 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001068 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001070 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 return sock->ops->poll(file, sock, wait);
1072}
1073
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001074static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001076 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077
1078 return sock->ops->mmap(file, sock, vma);
1079}
1080
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001081static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082{
1083 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001084 * It was possible the inode is NULL we were
1085 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 */
1087
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001088 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 printk(KERN_DEBUG "sock_close: NULL inode\n");
1090 return 0;
1091 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 sock_release(SOCKET_I(inode));
1093 return 0;
1094}
1095
1096/*
1097 * Update the socket async list
1098 *
1099 * Fasync_list locking strategy.
1100 *
1101 * 1. fasync_list is modified only under process context socket lock
1102 * i.e. under semaphore.
1103 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001104 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 */
1106
1107static int sock_fasync(int fd, struct file *filp, int on)
1108{
Eric Dumazet989a2972010-04-14 09:55:35 +00001109 struct socket *sock = filp->private_data;
1110 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111
Eric Dumazet989a2972010-04-14 09:55:35 +00001112 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114
1115 lock_sock(sk);
1116
Eric Dumazet43815482010-04-29 11:01:49 +00001117 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
Eric Dumazet989a2972010-04-14 09:55:35 +00001118
Eric Dumazet43815482010-04-29 11:01:49 +00001119 if (!sock->wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001120 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001121 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001122 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123
Eric Dumazet989a2972010-04-14 09:55:35 +00001124 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 return 0;
1126}
1127
Eric Dumazet43815482010-04-29 11:01:49 +00001128/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129
1130int sock_wake_async(struct socket *sock, int how, int band)
1131{
Eric Dumazet43815482010-04-29 11:01:49 +00001132 struct socket_wq *wq;
1133
1134 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001136 rcu_read_lock();
1137 wq = rcu_dereference(sock->wq);
1138 if (!wq || !wq->fasync_list) {
1139 rcu_read_unlock();
1140 return -1;
1141 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001142 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001143 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1145 break;
1146 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001147 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1149 break;
1150 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001151 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001152call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001153 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001155 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001156 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 }
Eric Dumazet43815482010-04-29 11:01:49 +00001158 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 return 0;
1160}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001161EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
Pavel Emelyanov721db932010-09-29 16:06:32 +04001163int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001164 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165{
1166 int err;
1167 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001168 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
1170 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001171 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 */
1173 if (family < 0 || family >= NPROTO)
1174 return -EAFNOSUPPORT;
1175 if (type < 0 || type >= SOCK_MAX)
1176 return -EINVAL;
1177
1178 /* Compatibility.
1179
1180 This uglymoron is moved from INET layer to here to avoid
1181 deadlock in module load.
1182 */
1183 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001184 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 if (!warned) {
1186 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001187 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1188 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 }
1190 family = PF_PACKET;
1191 }
1192
1193 err = security_socket_create(family, type, protocol, kern);
1194 if (err)
1195 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001196
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001197 /*
1198 * Allocate the socket and allow the family to set things up. if
1199 * the protocol is 0, the family is instructed to select an appropriate
1200 * default.
1201 */
1202 sock = sock_alloc();
1203 if (!sock) {
1204 if (net_ratelimit())
1205 printk(KERN_WARNING "socket: no more sockets\n");
1206 return -ENFILE; /* Not exactly a match, but its the
1207 closest posix thing */
1208 }
1209
1210 sock->type = type;
1211
Johannes Berg95a5afc2008-10-16 15:24:51 -07001212#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001213 /* Attempt to load a protocol module if the find failed.
1214 *
1215 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 * requested real, full-featured networking support upon configuration.
1217 * Otherwise module support will break!
1218 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001219 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001220 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221#endif
1222
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001223 rcu_read_lock();
1224 pf = rcu_dereference(net_families[family]);
1225 err = -EAFNOSUPPORT;
1226 if (!pf)
1227 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228
1229 /*
1230 * We will call the ->create function, that possibly is in a loadable
1231 * module, so we have to bump that loadable module refcnt first.
1232 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001233 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 goto out_release;
1235
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001236 /* Now protected by module ref count */
1237 rcu_read_unlock();
1238
Eric Paris3f378b62009-11-05 22:18:14 -08001239 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001240 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001242
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 /*
1244 * Now to bump the refcnt of the [loadable] module that owns this
1245 * socket at sock_release time we decrement its refcnt.
1246 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001247 if (!try_module_get(sock->ops->owner))
1248 goto out_module_busy;
1249
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 /*
1251 * Now that we're done with the ->create function, the [loadable]
1252 * module can have its refcnt decremented
1253 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001254 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001255 err = security_socket_post_create(sock, family, type, protocol, kern);
1256 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001257 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001258 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001260 return 0;
1261
1262out_module_busy:
1263 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001265 sock->ops = NULL;
1266 module_put(pf->owner);
1267out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001269 return err;
1270
1271out_release:
1272 rcu_read_unlock();
1273 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001275EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276
1277int sock_create(int family, int type, int protocol, struct socket **res)
1278{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001279 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001281EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282
1283int sock_create_kern(int family, int type, int protocol, struct socket **res)
1284{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001285 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001287EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001289SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290{
1291 int retval;
1292 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001293 int flags;
1294
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001295 /* Check the SOCK_* constants for consistency. */
1296 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1297 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1298 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1299 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1300
Ulrich Dreppera677a032008-07-23 21:29:17 -07001301 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001302 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001303 return -EINVAL;
1304 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001306 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1307 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1308
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 retval = sock_create(family, type, protocol, &sock);
1310 if (retval < 0)
1311 goto out;
1312
Ulrich Drepper77d27202008-07-23 21:29:35 -07001313 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 if (retval < 0)
1315 goto out_release;
1316
1317out:
1318 /* It may be already another descriptor 8) Not kernel problem. */
1319 return retval;
1320
1321out_release:
1322 sock_release(sock);
1323 return retval;
1324}
1325
1326/*
1327 * Create a pair of connected sockets.
1328 */
1329
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001330SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1331 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332{
1333 struct socket *sock1, *sock2;
1334 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001335 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001336 int flags;
1337
1338 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001339 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001340 return -EINVAL;
1341 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001343 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1344 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1345
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 /*
1347 * Obtain the first socket and check if the underlying protocol
1348 * supports the socketpair call.
1349 */
1350
1351 err = sock_create(family, type, protocol, &sock1);
1352 if (err < 0)
1353 goto out;
1354
1355 err = sock_create(family, type, protocol, &sock2);
1356 if (err < 0)
1357 goto out_release_1;
1358
1359 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001360 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 goto out_release_both;
1362
Al Viro7cbe66b2009-08-05 19:59:08 +04001363 fd1 = sock_alloc_file(sock1, &newfile1, flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001364 if (unlikely(fd1 < 0)) {
1365 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001367 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
Al Viro7cbe66b2009-08-05 19:59:08 +04001369 fd2 = sock_alloc_file(sock2, &newfile2, flags);
Al Viro198de4d2009-08-05 19:29:23 +04001370 if (unlikely(fd2 < 0)) {
1371 err = fd2;
1372 fput(newfile1);
1373 put_unused_fd(fd1);
1374 sock_release(sock2);
1375 goto out;
Al Virodb349502007-02-07 01:48:00 -05001376 }
1377
Al Viro157cf642008-12-14 04:57:47 -05001378 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001379 fd_install(fd1, newfile1);
1380 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 /* fd1 and fd2 may be already another descriptors.
1382 * Not kernel problem.
1383 */
1384
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001385 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 if (!err)
1387 err = put_user(fd2, &usockvec[1]);
1388 if (!err)
1389 return 0;
1390
1391 sys_close(fd2);
1392 sys_close(fd1);
1393 return err;
1394
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001396 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001398 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399out:
1400 return err;
1401}
1402
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403/*
1404 * Bind a name to a socket. Nothing much to do here since it's
1405 * the protocol's responsibility to handle the local address.
1406 *
1407 * We move the socket address to kernel space before we call
1408 * the protocol layer (having also checked the address is ok).
1409 */
1410
Heiko Carstens20f37032009-01-14 14:14:23 +01001411SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412{
1413 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001414 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001415 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001417 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001418 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001419 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001420 if (err >= 0) {
1421 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001422 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001423 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001424 if (!err)
1425 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001426 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001427 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001429 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001430 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 return err;
1432}
1433
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434/*
1435 * Perform a listen. Basically, we allow the protocol to do anything
1436 * necessary for a listen, and if that works, we mark the socket as
1437 * ready for listening.
1438 */
1439
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001440SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441{
1442 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001443 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001444 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001445
1446 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1447 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001448 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001449 if ((unsigned)backlog > somaxconn)
1450 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451
1452 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001453 if (!err)
1454 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001456 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 }
1458 return err;
1459}
1460
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461/*
1462 * For accept, we attempt to create a new socket, set up the link
1463 * with the client, wake up the client, then return the new
1464 * connected fd. We collect the address of the connector in kernel
1465 * space and move it to user at the very end. This is unclean because
1466 * we open the socket then return an error.
1467 *
1468 * 1003.1g adds the ability to recvmsg() to query connection pending
1469 * status to recvmsg. We need to add that support in a way thats
1470 * clean when we restucture accept also.
1471 */
1472
Heiko Carstens20f37032009-01-14 14:14:23 +01001473SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1474 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475{
1476 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001477 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001478 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001479 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480
Ulrich Drepper77d27202008-07-23 21:29:35 -07001481 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001482 return -EINVAL;
1483
1484 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1485 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1486
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001487 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 if (!sock)
1489 goto out;
1490
1491 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001492 newsock = sock_alloc();
1493 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 goto out_put;
1495
1496 newsock->type = sock->type;
1497 newsock->ops = sock->ops;
1498
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 /*
1500 * We don't need try_module_get here, as the listening socket (sock)
1501 * has the protocol module (sock->ops->owner) held.
1502 */
1503 __module_get(newsock->ops->owner);
1504
Al Viro7cbe66b2009-08-05 19:59:08 +04001505 newfd = sock_alloc_file(newsock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001506 if (unlikely(newfd < 0)) {
1507 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001508 sock_release(newsock);
1509 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001510 }
1511
Frank Filza79af592005-09-27 15:23:38 -07001512 err = security_socket_accept(sock, newsock);
1513 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001514 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1517 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001518 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
1520 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001521 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001522 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001524 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001526 err = move_addr_to_user((struct sockaddr *)&address,
1527 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001529 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 }
1531
1532 /* File flags are not inherited via accept() unlike another OSes. */
1533
David S. Miller39d8c1b2006-03-20 17:13:49 -08001534 fd_install(newfd, newfile);
1535 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001538 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539out:
1540 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001541out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001542 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001543 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 goto out_put;
1545}
1546
Heiko Carstens20f37032009-01-14 14:14:23 +01001547SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1548 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001549{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001550 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001551}
1552
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553/*
1554 * Attempt to connect to a socket with the server address. The address
1555 * is in user space so we verify it is OK and move it to kernel space.
1556 *
1557 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1558 * break bindings
1559 *
1560 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1561 * other SEQPACKET protocols that take time to connect() as it doesn't
1562 * include the -EINPROGRESS status for such sockets.
1563 */
1564
Heiko Carstens20f37032009-01-14 14:14:23 +01001565SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1566 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567{
1568 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001569 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001570 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001572 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 if (!sock)
1574 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001575 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 if (err < 0)
1577 goto out_put;
1578
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001579 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001580 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 if (err)
1582 goto out_put;
1583
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001584 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 sock->file->f_flags);
1586out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001587 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588out:
1589 return err;
1590}
1591
1592/*
1593 * Get the local address ('name') of a socket object. Move the obtained
1594 * name to user space.
1595 */
1596
Heiko Carstens20f37032009-01-14 14:14:23 +01001597SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1598 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599{
1600 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001601 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001602 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001603
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001604 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 if (!sock)
1606 goto out;
1607
1608 err = security_socket_getsockname(sock);
1609 if (err)
1610 goto out_put;
1611
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001612 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 if (err)
1614 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001615 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
1617out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001618 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619out:
1620 return err;
1621}
1622
1623/*
1624 * Get the remote address ('name') of a socket object. Move the obtained
1625 * name to user space.
1626 */
1627
Heiko Carstens20f37032009-01-14 14:14:23 +01001628SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1629 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630{
1631 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001632 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001633 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001635 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1636 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 err = security_socket_getpeername(sock);
1638 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001639 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 return err;
1641 }
1642
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001643 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001644 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001645 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001647 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001648 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001649 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 }
1651 return err;
1652}
1653
1654/*
1655 * Send a datagram to a given address. We move the address into kernel
1656 * space and check the user space data area is readable before invoking
1657 * the protocol.
1658 */
1659
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001660SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1661 unsigned, flags, struct sockaddr __user *, addr,
1662 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663{
1664 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001665 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 int err;
1667 struct msghdr msg;
1668 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001669 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001670
Linus Torvalds253eacc2010-10-30 16:43:10 -07001671 if (len > INT_MAX)
1672 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001673 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1674 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001675 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001676
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001677 iov.iov_base = buff;
1678 iov.iov_len = len;
1679 msg.msg_name = NULL;
1680 msg.msg_iov = &iov;
1681 msg.msg_iovlen = 1;
1682 msg.msg_control = NULL;
1683 msg.msg_controllen = 0;
1684 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001685 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001686 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 if (err < 0)
1688 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001689 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001690 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 }
1692 if (sock->file->f_flags & O_NONBLOCK)
1693 flags |= MSG_DONTWAIT;
1694 msg.msg_flags = flags;
1695 err = sock_sendmsg(sock, &msg, len);
1696
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001697out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001698 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001699out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 return err;
1701}
1702
1703/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001704 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 */
1706
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001707SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1708 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709{
1710 return sys_sendto(fd, buff, len, flags, NULL, 0);
1711}
1712
1713/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001714 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 * sender. We verify the buffers are writable and if needed move the
1716 * sender address from kernel to user space.
1717 */
1718
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001719SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1720 unsigned, flags, struct sockaddr __user *, addr,
1721 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722{
1723 struct socket *sock;
1724 struct iovec iov;
1725 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001726 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001727 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001728 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729
Linus Torvalds253eacc2010-10-30 16:43:10 -07001730 if (size > INT_MAX)
1731 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001732 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001734 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001736 msg.msg_control = NULL;
1737 msg.msg_controllen = 0;
1738 msg.msg_iovlen = 1;
1739 msg.msg_iov = &iov;
1740 iov.iov_len = size;
1741 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001742 msg.msg_name = (struct sockaddr *)&address;
1743 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 if (sock->file->f_flags & O_NONBLOCK)
1745 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001746 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001748 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001749 err2 = move_addr_to_user((struct sockaddr *)&address,
1750 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001751 if (err2 < 0)
1752 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001754
1755 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001756out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 return err;
1758}
1759
1760/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001761 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 */
1763
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001764asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1765 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766{
1767 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1768}
1769
1770/*
1771 * Set a socket option. Because we don't know the option lengths we have
1772 * to pass the user mode parameter for the protocols to sort out.
1773 */
1774
Heiko Carstens20f37032009-01-14 14:14:23 +01001775SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1776 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001778 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 struct socket *sock;
1780
1781 if (optlen < 0)
1782 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001783
1784 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1785 if (sock != NULL) {
1786 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001787 if (err)
1788 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789
1790 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001791 err =
1792 sock_setsockopt(sock, level, optname, optval,
1793 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001795 err =
1796 sock->ops->setsockopt(sock, level, optname, optval,
1797 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001798out_put:
1799 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 }
1801 return err;
1802}
1803
1804/*
1805 * Get a socket option. Because we don't know the option lengths we have
1806 * to pass a user mode parameter for the protocols to sort out.
1807 */
1808
Heiko Carstens20f37032009-01-14 14:14:23 +01001809SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1810 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001812 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813 struct socket *sock;
1814
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001815 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1816 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001817 err = security_socket_getsockopt(sock, level, optname);
1818 if (err)
1819 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820
1821 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001822 err =
1823 sock_getsockopt(sock, level, optname, optval,
1824 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001826 err =
1827 sock->ops->getsockopt(sock, level, optname, optval,
1828 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001829out_put:
1830 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 }
1832 return err;
1833}
1834
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835/*
1836 * Shutdown a socket.
1837 */
1838
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001839SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001841 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 struct socket *sock;
1843
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001844 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1845 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001847 if (!err)
1848 err = sock->ops->shutdown(sock, how);
1849 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 }
1851 return err;
1852}
1853
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001854/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 * fields which are the same type (int / unsigned) on our platforms.
1856 */
1857#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1858#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1859#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1860
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861/*
1862 * BSD sendmsg interface
1863 */
1864
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001865SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001867 struct compat_msghdr __user *msg_compat =
1868 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001870 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001872 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001873 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1874 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 unsigned char *ctl_buf = ctl;
1876 struct msghdr msg_sys;
1877 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001878 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001879
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 err = -EFAULT;
1881 if (MSG_CMSG_COMPAT & flags) {
1882 if (get_compat_msghdr(&msg_sys, msg_compat))
1883 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001884 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 return -EFAULT;
1886
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001887 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001888 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 goto out;
1890
1891 /* do not move before msg_sys is valid */
1892 err = -EMSGSIZE;
1893 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1894 goto out_put;
1895
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001896 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 err = -ENOMEM;
1898 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1899 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1900 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1901 if (!iov)
1902 goto out_put;
1903 }
1904
1905 /* This will also move the address data into kernel space */
1906 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001907 err = verify_compat_iovec(&msg_sys, iov,
1908 (struct sockaddr *)&address,
1909 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001911 err = verify_iovec(&msg_sys, iov,
1912 (struct sockaddr *)&address,
1913 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001914 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 goto out_freeiov;
1916 total_len = err;
1917
1918 err = -ENOBUFS;
1919
1920 if (msg_sys.msg_controllen > INT_MAX)
1921 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001922 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001924 err =
1925 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1926 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 if (err)
1928 goto out_freeiov;
1929 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001930 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001932 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001934 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 goto out_freeiov;
1936 }
1937 err = -EFAULT;
1938 /*
1939 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1940 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1941 * checking falls down on this.
1942 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001943 if (copy_from_user(ctl_buf,
1944 (void __user __force *)msg_sys.msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001945 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 goto out_freectl;
1947 msg_sys.msg_control = ctl_buf;
1948 }
1949 msg_sys.msg_flags = flags;
1950
1951 if (sock->file->f_flags & O_NONBLOCK)
1952 msg_sys.msg_flags |= MSG_DONTWAIT;
1953 err = sock_sendmsg(sock, &msg_sys, total_len);
1954
1955out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001956 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1958out_freeiov:
1959 if (iov != iovstack)
1960 sock_kfree_s(sock->sk, iov, iov_size);
1961out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001962 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001963out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 return err;
1965}
1966
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001967static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1968 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001970 struct compat_msghdr __user *msg_compat =
1971 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001973 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 unsigned long cmsg_ptr;
1975 int err, iov_size, total_len, len;
1976
1977 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001978 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979
1980 /* user mode address pointers */
1981 struct sockaddr __user *uaddr;
1982 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001983
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001985 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001987 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001988 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001991 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1992 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001993
1994 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001996 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1997 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1999 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002000 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 }
2002
2003 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002004 * Save the user-mode address (verify_iovec will change the
2005 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002007
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002008 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 uaddr_len = COMPAT_NAMELEN(msg);
2010 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002011 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002012 (struct sockaddr *)&addr,
2013 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 } else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002015 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002016 (struct sockaddr *)&addr,
2017 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 if (err < 0)
2019 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002022 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2023 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002024
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 if (sock->file->f_flags & O_NONBLOCK)
2026 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002027 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2028 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 if (err < 0)
2030 goto out_freeiov;
2031 len = err;
2032
2033 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002034 err = move_addr_to_user((struct sockaddr *)&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002035 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002036 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 if (err < 0)
2038 goto out_freeiov;
2039 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002040 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002041 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 if (err)
2043 goto out_freeiov;
2044 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002045 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 &msg_compat->msg_controllen);
2047 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002048 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 &msg->msg_controllen);
2050 if (err)
2051 goto out_freeiov;
2052 err = len;
2053
2054out_freeiov:
2055 if (iov != iovstack)
2056 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002057out:
2058 return err;
2059}
2060
2061/*
2062 * BSD recvmsg interface
2063 */
2064
2065SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2066 unsigned int, flags)
2067{
2068 int fput_needed, err;
2069 struct msghdr msg_sys;
2070 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2071
2072 if (!sock)
2073 goto out;
2074
2075 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2076
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002077 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078out:
2079 return err;
2080}
2081
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002082/*
2083 * Linux recvmmsg interface
2084 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002086int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2087 unsigned int flags, struct timespec *timeout)
2088{
2089 int fput_needed, err, datagrams;
2090 struct socket *sock;
2091 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002092 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002093 struct msghdr msg_sys;
2094 struct timespec end_time;
2095
2096 if (timeout &&
2097 poll_select_set_timeout(&end_time, timeout->tv_sec,
2098 timeout->tv_nsec))
2099 return -EINVAL;
2100
2101 datagrams = 0;
2102
2103 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2104 if (!sock)
2105 return err;
2106
2107 err = sock_error(sock->sk);
2108 if (err)
2109 goto out_put;
2110
2111 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002112 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002113
2114 while (datagrams < vlen) {
2115 /*
2116 * No need to ask LSM for more than the first datagram.
2117 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002118 if (MSG_CMSG_COMPAT & flags) {
2119 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2120 &msg_sys, flags, datagrams);
2121 if (err < 0)
2122 break;
2123 err = __put_user(err, &compat_entry->msg_len);
2124 ++compat_entry;
2125 } else {
2126 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2127 &msg_sys, flags, datagrams);
2128 if (err < 0)
2129 break;
2130 err = put_user(err, &entry->msg_len);
2131 ++entry;
2132 }
2133
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002134 if (err)
2135 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002136 ++datagrams;
2137
Brandon L Black71c5c152010-03-26 16:18:03 +00002138 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2139 if (flags & MSG_WAITFORONE)
2140 flags |= MSG_DONTWAIT;
2141
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002142 if (timeout) {
2143 ktime_get_ts(timeout);
2144 *timeout = timespec_sub(end_time, *timeout);
2145 if (timeout->tv_sec < 0) {
2146 timeout->tv_sec = timeout->tv_nsec = 0;
2147 break;
2148 }
2149
2150 /* Timeout, return less than vlen datagrams */
2151 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2152 break;
2153 }
2154
2155 /* Out of band data, return right away */
2156 if (msg_sys.msg_flags & MSG_OOB)
2157 break;
2158 }
2159
2160out_put:
2161 fput_light(sock->file, fput_needed);
2162
2163 if (err == 0)
2164 return datagrams;
2165
2166 if (datagrams != 0) {
2167 /*
2168 * We may return less entries than requested (vlen) if the
2169 * sock is non block and there aren't enough datagrams...
2170 */
2171 if (err != -EAGAIN) {
2172 /*
2173 * ... or if recvmsg returns an error after we
2174 * received some datagrams, where we record the
2175 * error to return on the next call or if the
2176 * app asks about it using getsockopt(SO_ERROR).
2177 */
2178 sock->sk->sk_err = -err;
2179 }
2180
2181 return datagrams;
2182 }
2183
2184 return err;
2185}
2186
2187SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2188 unsigned int, vlen, unsigned int, flags,
2189 struct timespec __user *, timeout)
2190{
2191 int datagrams;
2192 struct timespec timeout_sys;
2193
2194 if (!timeout)
2195 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2196
2197 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2198 return -EFAULT;
2199
2200 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2201
2202 if (datagrams > 0 &&
2203 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2204 datagrams = -EFAULT;
2205
2206 return datagrams;
2207}
2208
2209#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210/* Argument list sizes for sys_socketcall */
2211#define AL(x) ((x) * sizeof(unsigned long))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002212static const unsigned char nargs[20] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002213 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2214 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2215 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2216 AL(4), AL(5)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002217};
2218
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219#undef AL
2220
2221/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002222 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 *
2224 * Argument checking cleaned up. Saved 20% in size.
2225 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002226 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227 */
2228
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002229SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230{
2231 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002232 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002234 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002236 if (call < 1 || call > SYS_RECVMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237 return -EINVAL;
2238
Arjan van de Ven47379052009-09-28 12:57:44 -07002239 len = nargs[call];
2240 if (len > sizeof(a))
2241 return -EINVAL;
2242
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002244 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002246
Al Virof3298dc2008-12-10 03:16:51 -05002247 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002248
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002249 a0 = a[0];
2250 a1 = a[1];
2251
2252 switch (call) {
2253 case SYS_SOCKET:
2254 err = sys_socket(a0, a1, a[2]);
2255 break;
2256 case SYS_BIND:
2257 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2258 break;
2259 case SYS_CONNECT:
2260 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2261 break;
2262 case SYS_LISTEN:
2263 err = sys_listen(a0, a1);
2264 break;
2265 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002266 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2267 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002268 break;
2269 case SYS_GETSOCKNAME:
2270 err =
2271 sys_getsockname(a0, (struct sockaddr __user *)a1,
2272 (int __user *)a[2]);
2273 break;
2274 case SYS_GETPEERNAME:
2275 err =
2276 sys_getpeername(a0, (struct sockaddr __user *)a1,
2277 (int __user *)a[2]);
2278 break;
2279 case SYS_SOCKETPAIR:
2280 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2281 break;
2282 case SYS_SEND:
2283 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2284 break;
2285 case SYS_SENDTO:
2286 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2287 (struct sockaddr __user *)a[4], a[5]);
2288 break;
2289 case SYS_RECV:
2290 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2291 break;
2292 case SYS_RECVFROM:
2293 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2294 (struct sockaddr __user *)a[4],
2295 (int __user *)a[5]);
2296 break;
2297 case SYS_SHUTDOWN:
2298 err = sys_shutdown(a0, a1);
2299 break;
2300 case SYS_SETSOCKOPT:
2301 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2302 break;
2303 case SYS_GETSOCKOPT:
2304 err =
2305 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2306 (int __user *)a[4]);
2307 break;
2308 case SYS_SENDMSG:
2309 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2310 break;
2311 case SYS_RECVMSG:
2312 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2313 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002314 case SYS_RECVMMSG:
2315 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2316 (struct timespec __user *)a[4]);
2317 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002318 case SYS_ACCEPT4:
2319 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2320 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002321 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002322 default:
2323 err = -EINVAL;
2324 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 }
2326 return err;
2327}
2328
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002329#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002331/**
2332 * sock_register - add a socket protocol handler
2333 * @ops: description of protocol
2334 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335 * This function is called by a protocol handler that wants to
2336 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002337 * socket interface. The value ops->family coresponds to the
2338 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002340int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341{
2342 int err;
2343
2344 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002345 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2346 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347 return -ENOBUFS;
2348 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002349
2350 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002351 if (rcu_dereference_protected(net_families[ops->family],
2352 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002353 err = -EEXIST;
2354 else {
Eric Dumazet190683a2010-11-10 10:50:44 +00002355 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356 err = 0;
2357 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002358 spin_unlock(&net_family_lock);
2359
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002360 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361 return err;
2362}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002363EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002365/**
2366 * sock_unregister - remove a protocol handler
2367 * @family: protocol family to remove
2368 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 * This function is called by a protocol handler that wants to
2370 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002371 * new socket creation.
2372 *
2373 * If protocol handler is a module, then it can use module reference
2374 * counts to protect against new references. If protocol handler is not
2375 * a module then it needs to provide its own protection in
2376 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002378void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002380 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002382 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002383 rcu_assign_pointer(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002384 spin_unlock(&net_family_lock);
2385
2386 synchronize_rcu();
2387
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002388 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002390EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391
Andi Kleen77d76ea2005-12-22 12:43:42 -08002392static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002394 int err;
2395
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002397 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002399
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 sk_init();
2401
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002403 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404 */
2405 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406
2407 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002408 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409 */
2410
2411 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002412
2413 err = register_filesystem(&sock_fs_type);
2414 if (err)
2415 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002417 if (IS_ERR(sock_mnt)) {
2418 err = PTR_ERR(sock_mnt);
2419 goto out_mount;
2420 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002421
2422 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423 */
2424
2425#ifdef CONFIG_NETFILTER
2426 netfilter_init();
2427#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002428
Richard Cochranc1f19b52010-07-17 08:49:36 +00002429#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2430 skb_timestamping_init();
2431#endif
2432
Nick Pigginb3e19d92011-01-07 17:50:11 +11002433out:
2434 return err;
2435
2436out_mount:
2437 unregister_filesystem(&sock_fs_type);
2438out_fs:
2439 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440}
2441
Andi Kleen77d76ea2005-12-22 12:43:42 -08002442core_initcall(sock_init); /* early initcall */
2443
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444#ifdef CONFIG_PROC_FS
2445void socket_seq_show(struct seq_file *seq)
2446{
2447 int cpu;
2448 int counter = 0;
2449
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002450 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002451 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452
2453 /* It can be negative, by the way. 8) */
2454 if (counter < 0)
2455 counter = 0;
2456
2457 seq_printf(seq, "sockets: used %d\n", counter);
2458}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002459#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002461#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002462static int do_siocgstamp(struct net *net, struct socket *sock,
2463 unsigned int cmd, struct compat_timeval __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002464{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002465 mm_segment_t old_fs = get_fs();
2466 struct timeval ktv;
2467 int err;
2468
2469 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002470 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002471 set_fs(old_fs);
2472 if (!err) {
2473 err = put_user(ktv.tv_sec, &up->tv_sec);
2474 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2475 }
2476 return err;
2477}
2478
Arnd Bergmann6b960182009-11-06 23:10:54 -08002479static int do_siocgstampns(struct net *net, struct socket *sock,
2480 unsigned int cmd, struct compat_timespec __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002481{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002482 mm_segment_t old_fs = get_fs();
2483 struct timespec kts;
2484 int err;
2485
2486 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002487 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002488 set_fs(old_fs);
2489 if (!err) {
2490 err = put_user(kts.tv_sec, &up->tv_sec);
2491 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2492 }
2493 return err;
2494}
2495
Arnd Bergmann6b960182009-11-06 23:10:54 -08002496static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002497{
2498 struct ifreq __user *uifr;
2499 int err;
2500
2501 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002502 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002503 return -EFAULT;
2504
Arnd Bergmann6b960182009-11-06 23:10:54 -08002505 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002506 if (err)
2507 return err;
2508
Arnd Bergmann6b960182009-11-06 23:10:54 -08002509 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002510 return -EFAULT;
2511
2512 return 0;
2513}
2514
Arnd Bergmann6b960182009-11-06 23:10:54 -08002515static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002516{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002517 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002518 struct ifconf ifc;
2519 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002520 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002521 struct ifreq __user *ifr;
2522 unsigned int i, j;
2523 int err;
2524
Arnd Bergmann6b960182009-11-06 23:10:54 -08002525 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002526 return -EFAULT;
2527
2528 if (ifc32.ifcbuf == 0) {
2529 ifc32.ifc_len = 0;
2530 ifc.ifc_len = 0;
2531 ifc.ifc_req = NULL;
2532 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2533 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002534 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2535 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002536 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2537 ifc.ifc_len = len;
2538 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2539 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002540 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002541 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002542 return -EFAULT;
2543 ifr++;
2544 ifr32++;
2545 }
2546 }
2547 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2548 return -EFAULT;
2549
Arnd Bergmann6b960182009-11-06 23:10:54 -08002550 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002551 if (err)
2552 return err;
2553
2554 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2555 return -EFAULT;
2556
2557 ifr = ifc.ifc_req;
2558 ifr32 = compat_ptr(ifc32.ifcbuf);
2559 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002560 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2561 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2562 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002563 return -EFAULT;
2564 ifr32++;
2565 ifr++;
2566 }
2567
2568 if (ifc32.ifcbuf == 0) {
2569 /* Translate from 64-bit structure multiple to
2570 * a 32-bit one.
2571 */
2572 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002573 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002574 ifc32.ifc_len = i;
2575 } else {
2576 ifc32.ifc_len = i;
2577 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002578 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002579 return -EFAULT;
2580
2581 return 0;
2582}
2583
Arnd Bergmann6b960182009-11-06 23:10:54 -08002584static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002585{
2586 struct ifreq __user *ifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002587 u32 data;
2588 void __user *datap;
2589
2590 ifr = compat_alloc_user_space(sizeof(*ifr));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002591
2592 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2593 return -EFAULT;
2594
2595 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2596 return -EFAULT;
2597
2598 datap = compat_ptr(data);
2599 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2600 return -EFAULT;
2601
Arnd Bergmann6b960182009-11-06 23:10:54 -08002602 return dev_ioctl(net, SIOCETHTOOL, ifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002603}
2604
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002605static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2606{
2607 void __user *uptr;
2608 compat_uptr_t uptr32;
2609 struct ifreq __user *uifr;
2610
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002611 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002612 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2613 return -EFAULT;
2614
2615 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2616 return -EFAULT;
2617
2618 uptr = compat_ptr(uptr32);
2619
2620 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2621 return -EFAULT;
2622
2623 return dev_ioctl(net, SIOCWANDEV, uifr);
2624}
2625
Arnd Bergmann6b960182009-11-06 23:10:54 -08002626static int bond_ioctl(struct net *net, unsigned int cmd,
2627 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002628{
2629 struct ifreq kifr;
2630 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002631 mm_segment_t old_fs;
2632 int err;
2633 u32 data;
2634 void __user *datap;
2635
2636 switch (cmd) {
2637 case SIOCBONDENSLAVE:
2638 case SIOCBONDRELEASE:
2639 case SIOCBONDSETHWADDR:
2640 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002641 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002642 return -EFAULT;
2643
2644 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002645 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002646 err = dev_ioctl(net, cmd, &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002647 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002648
2649 return err;
2650 case SIOCBONDSLAVEINFOQUERY:
2651 case SIOCBONDINFOQUERY:
2652 uifr = compat_alloc_user_space(sizeof(*uifr));
2653 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2654 return -EFAULT;
2655
2656 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2657 return -EFAULT;
2658
2659 datap = compat_ptr(data);
2660 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2661 return -EFAULT;
2662
Arnd Bergmann6b960182009-11-06 23:10:54 -08002663 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002664 default:
2665 return -EINVAL;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002666 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002667}
2668
Arnd Bergmann6b960182009-11-06 23:10:54 -08002669static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2670 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002671{
2672 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002673 char tmp_buf[IFNAMSIZ];
2674 void __user *data64;
2675 u32 data32;
2676
2677 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2678 IFNAMSIZ))
2679 return -EFAULT;
2680 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2681 return -EFAULT;
2682 data64 = compat_ptr(data32);
2683
2684 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2685
2686 /* Don't check these user accesses, just let that get trapped
2687 * in the ioctl handler instead.
2688 */
2689 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2690 IFNAMSIZ))
2691 return -EFAULT;
2692 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2693 return -EFAULT;
2694
Arnd Bergmann6b960182009-11-06 23:10:54 -08002695 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002696}
2697
Arnd Bergmann6b960182009-11-06 23:10:54 -08002698static int dev_ifsioc(struct net *net, struct socket *sock,
2699 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002700{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002701 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002702 int err;
2703
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002704 uifr = compat_alloc_user_space(sizeof(*uifr));
2705 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2706 return -EFAULT;
2707
2708 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2709
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002710 if (!err) {
2711 switch (cmd) {
2712 case SIOCGIFFLAGS:
2713 case SIOCGIFMETRIC:
2714 case SIOCGIFMTU:
2715 case SIOCGIFMEM:
2716 case SIOCGIFHWADDR:
2717 case SIOCGIFINDEX:
2718 case SIOCGIFADDR:
2719 case SIOCGIFBRDADDR:
2720 case SIOCGIFDSTADDR:
2721 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002722 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002723 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002724 case SIOCGMIIPHY:
2725 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002726 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002727 err = -EFAULT;
2728 break;
2729 }
2730 }
2731 return err;
2732}
2733
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002734static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2735 struct compat_ifreq __user *uifr32)
2736{
2737 struct ifreq ifr;
2738 struct compat_ifmap __user *uifmap32;
2739 mm_segment_t old_fs;
2740 int err;
2741
2742 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2743 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2744 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2745 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2746 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2747 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2748 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2749 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2750 if (err)
2751 return -EFAULT;
2752
2753 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002754 set_fs(KERNEL_DS);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002755 err = dev_ioctl(net, cmd, (void __user *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002756 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002757
2758 if (cmd == SIOCGIFMAP && !err) {
2759 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2760 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2761 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2762 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2763 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2764 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2765 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2766 if (err)
2767 err = -EFAULT;
2768 }
2769 return err;
2770}
2771
2772static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2773{
2774 void __user *uptr;
2775 compat_uptr_t uptr32;
2776 struct ifreq __user *uifr;
2777
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002778 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002779 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2780 return -EFAULT;
2781
2782 if (get_user(uptr32, &uifr32->ifr_data))
2783 return -EFAULT;
2784
2785 uptr = compat_ptr(uptr32);
2786
2787 if (put_user(uptr, &uifr->ifr_data))
2788 return -EFAULT;
2789
2790 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2791}
2792
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002793struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002794 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002795 struct sockaddr rt_dst; /* target address */
2796 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2797 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002798 unsigned short rt_flags;
2799 short rt_pad2;
2800 u32 rt_pad3;
2801 unsigned char rt_tos;
2802 unsigned char rt_class;
2803 short rt_pad4;
2804 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002805 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002806 u32 rt_mtu; /* per route MTU/Window */
2807 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002808 unsigned short rt_irtt; /* Initial RTT */
2809};
2810
2811struct in6_rtmsg32 {
2812 struct in6_addr rtmsg_dst;
2813 struct in6_addr rtmsg_src;
2814 struct in6_addr rtmsg_gateway;
2815 u32 rtmsg_type;
2816 u16 rtmsg_dst_len;
2817 u16 rtmsg_src_len;
2818 u32 rtmsg_metric;
2819 u32 rtmsg_info;
2820 u32 rtmsg_flags;
2821 s32 rtmsg_ifindex;
2822};
2823
Arnd Bergmann6b960182009-11-06 23:10:54 -08002824static int routing_ioctl(struct net *net, struct socket *sock,
2825 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002826{
2827 int ret;
2828 void *r = NULL;
2829 struct in6_rtmsg r6;
2830 struct rtentry r4;
2831 char devname[16];
2832 u32 rtdev;
2833 mm_segment_t old_fs = get_fs();
2834
Arnd Bergmann6b960182009-11-06 23:10:54 -08002835 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2836 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002837 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002838 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002839 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2840 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2841 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2842 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2843 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2844 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2845 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002846
2847 r = (void *) &r6;
2848 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002849 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002850 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002851 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002852 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2853 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2854 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2855 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2856 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2857 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002858 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002859 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002860 r4.rt_dev = devname; devname[15] = 0;
2861 } else
2862 r4.rt_dev = NULL;
2863
2864 r = (void *) &r4;
2865 }
2866
2867 if (ret) {
2868 ret = -EFAULT;
2869 goto out;
2870 }
2871
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002872 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002873 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002874 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002875
2876out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002877 return ret;
2878}
2879
2880/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2881 * for some operations; this forces use of the newer bridge-utils that
2882 * use compatiable ioctls
2883 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002884static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002885{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002886 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002887
Arnd Bergmann6b960182009-11-06 23:10:54 -08002888 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002889 return -EFAULT;
2890 if (tmp == BRCTL_GET_VERSION)
2891 return BRCTL_VERSION + 1;
2892 return -EINVAL;
2893}
2894
Arnd Bergmann6b960182009-11-06 23:10:54 -08002895static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2896 unsigned int cmd, unsigned long arg)
2897{
2898 void __user *argp = compat_ptr(arg);
2899 struct sock *sk = sock->sk;
2900 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002901
Arnd Bergmann6b960182009-11-06 23:10:54 -08002902 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2903 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002904
Arnd Bergmann6b960182009-11-06 23:10:54 -08002905 switch (cmd) {
2906 case SIOCSIFBR:
2907 case SIOCGIFBR:
2908 return old_bridge_ioctl(argp);
2909 case SIOCGIFNAME:
2910 return dev_ifname32(net, argp);
2911 case SIOCGIFCONF:
2912 return dev_ifconf(net, argp);
2913 case SIOCETHTOOL:
2914 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002915 case SIOCWANDEV:
2916 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002917 case SIOCGIFMAP:
2918 case SIOCSIFMAP:
2919 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002920 case SIOCBONDENSLAVE:
2921 case SIOCBONDRELEASE:
2922 case SIOCBONDSETHWADDR:
2923 case SIOCBONDSLAVEINFOQUERY:
2924 case SIOCBONDINFOQUERY:
2925 case SIOCBONDCHANGEACTIVE:
2926 return bond_ioctl(net, cmd, argp);
2927 case SIOCADDRT:
2928 case SIOCDELRT:
2929 return routing_ioctl(net, sock, cmd, argp);
2930 case SIOCGSTAMP:
2931 return do_siocgstamp(net, sock, cmd, argp);
2932 case SIOCGSTAMPNS:
2933 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002934 case SIOCSHWTSTAMP:
2935 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002936
Arnd Bergmann6b960182009-11-06 23:10:54 -08002937 case FIOSETOWN:
2938 case SIOCSPGRP:
2939 case FIOGETOWN:
2940 case SIOCGPGRP:
2941 case SIOCBRADDBR:
2942 case SIOCBRDELBR:
2943 case SIOCGIFVLAN:
2944 case SIOCSIFVLAN:
2945 case SIOCADDDLCI:
2946 case SIOCDELDLCI:
2947 return sock_ioctl(file, cmd, arg);
2948
2949 case SIOCGIFFLAGS:
2950 case SIOCSIFFLAGS:
2951 case SIOCGIFMETRIC:
2952 case SIOCSIFMETRIC:
2953 case SIOCGIFMTU:
2954 case SIOCSIFMTU:
2955 case SIOCGIFMEM:
2956 case SIOCSIFMEM:
2957 case SIOCGIFHWADDR:
2958 case SIOCSIFHWADDR:
2959 case SIOCADDMULTI:
2960 case SIOCDELMULTI:
2961 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002962 case SIOCGIFADDR:
2963 case SIOCSIFADDR:
2964 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002965 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002966 case SIOCGIFBRDADDR:
2967 case SIOCSIFBRDADDR:
2968 case SIOCGIFDSTADDR:
2969 case SIOCSIFDSTADDR:
2970 case SIOCGIFNETMASK:
2971 case SIOCSIFNETMASK:
2972 case SIOCSIFPFLAGS:
2973 case SIOCGIFPFLAGS:
2974 case SIOCGIFTXQLEN:
2975 case SIOCSIFTXQLEN:
2976 case SIOCBRADDIF:
2977 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002978 case SIOCSIFNAME:
2979 case SIOCGMIIPHY:
2980 case SIOCGMIIREG:
2981 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002982 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002983
Arnd Bergmann6b960182009-11-06 23:10:54 -08002984 case SIOCSARP:
2985 case SIOCGARP:
2986 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002987 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002988 return sock_do_ioctl(net, sock, cmd, arg);
2989 }
2990
2991 /* Prevent warning from compat_sys_ioctl, these always
2992 * result in -EINVAL in the native case anyway. */
2993 switch (cmd) {
2994 case SIOCRTMSG:
2995 case SIOCGIFCOUNT:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002996 case SIOCSRARP:
2997 case SIOCGRARP:
2998 case SIOCDRARP:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002999 case SIOCSIFLINK:
3000 case SIOCGIFSLAVE:
3001 case SIOCSIFSLAVE:
3002 return -EINVAL;
Arnd Bergmann6b960182009-11-06 23:10:54 -08003003 }
3004
3005 return -ENOIOCTLCMD;
3006}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003007
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003008static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003009 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003010{
3011 struct socket *sock = file->private_data;
3012 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003013 struct sock *sk;
3014 struct net *net;
3015
3016 sk = sock->sk;
3017 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003018
3019 if (sock->ops->compat_ioctl)
3020 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3021
David S. Miller87de87d2008-06-03 09:14:03 -07003022 if (ret == -ENOIOCTLCMD &&
3023 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3024 ret = compat_wext_handle_ioctl(net, cmd, arg);
3025
Arnd Bergmann6b960182009-11-06 23:10:54 -08003026 if (ret == -ENOIOCTLCMD)
3027 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3028
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003029 return ret;
3030}
3031#endif
3032
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003033int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3034{
3035 return sock->ops->bind(sock, addr, addrlen);
3036}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003037EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003038
3039int kernel_listen(struct socket *sock, int backlog)
3040{
3041 return sock->ops->listen(sock, backlog);
3042}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003043EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003044
3045int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3046{
3047 struct sock *sk = sock->sk;
3048 int err;
3049
3050 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3051 newsock);
3052 if (err < 0)
3053 goto done;
3054
3055 err = sock->ops->accept(sock, *newsock, flags);
3056 if (err < 0) {
3057 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003058 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003059 goto done;
3060 }
3061
3062 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003063 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003064
3065done:
3066 return err;
3067}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003068EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003069
3070int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003071 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003072{
3073 return sock->ops->connect(sock, addr, addrlen, flags);
3074}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003075EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003076
3077int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3078 int *addrlen)
3079{
3080 return sock->ops->getname(sock, addr, addrlen, 0);
3081}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003082EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003083
3084int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3085 int *addrlen)
3086{
3087 return sock->ops->getname(sock, addr, addrlen, 1);
3088}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003089EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003090
3091int kernel_getsockopt(struct socket *sock, int level, int optname,
3092 char *optval, int *optlen)
3093{
3094 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003095 char __user *uoptval;
3096 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003097 int err;
3098
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003099 uoptval = (char __user __force *) optval;
3100 uoptlen = (int __user __force *) optlen;
3101
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003102 set_fs(KERNEL_DS);
3103 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003104 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003105 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003106 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3107 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003108 set_fs(oldfs);
3109 return err;
3110}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003111EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003112
3113int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003114 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003115{
3116 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003117 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003118 int err;
3119
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003120 uoptval = (char __user __force *) optval;
3121
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003122 set_fs(KERNEL_DS);
3123 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003124 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003125 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003126 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003127 optlen);
3128 set_fs(oldfs);
3129 return err;
3130}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003131EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003132
3133int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3134 size_t size, int flags)
3135{
Herbert Xuf8451722010-05-24 00:12:34 -07003136 sock_update_classid(sock->sk);
3137
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003138 if (sock->ops->sendpage)
3139 return sock->ops->sendpage(sock, page, offset, size, flags);
3140
3141 return sock_no_sendpage(sock, page, offset, size, flags);
3142}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003143EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003144
3145int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3146{
3147 mm_segment_t oldfs = get_fs();
3148 int err;
3149
3150 set_fs(KERNEL_DS);
3151 err = sock->ops->ioctl(sock, cmd, arg);
3152 set_fs(oldfs);
3153
3154 return err;
3155}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003156EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003157
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003158int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3159{
3160 return sock->ops->shutdown(sock, how);
3161}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003162EXPORT_SYMBOL(kernel_sock_shutdown);