blob: 2ca51c719ef984cdadef749008456cf7bd5e1ae4 [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>
Masatake YAMATO600e1772012-08-29 10:44:29 +000091#include <linux/xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93#include <asm/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070097#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070098#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
Arnd Bergmann6b960182009-11-06 23:10:54 -0800103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800106#include <linux/sockios.h>
107#include <linux/atalk.h>
108
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700110static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
111 unsigned long nr_segs, loff_t pos);
112static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
113 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700114static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
116static int sock_close(struct inode *inode, struct file *file);
117static unsigned int sock_poll(struct file *file,
118 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700119static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800120#ifdef CONFIG_COMPAT
121static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700122 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800123#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125static ssize_t sock_sendpage(struct file *file, struct page *page,
126 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800127static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700128 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800129 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/*
132 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
133 * in the operation structures but are done directly via the socketcall() multiplexor.
134 */
135
Arjan van de Venda7071d2007-02-12 00:55:36 -0800136static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 .owner = THIS_MODULE,
138 .llseek = no_llseek,
139 .aio_read = sock_aio_read,
140 .aio_write = sock_aio_write,
141 .poll = sock_poll,
142 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800143#ifdef CONFIG_COMPAT
144 .compat_ioctl = compat_sock_ioctl,
145#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 .mmap = sock_mmap,
147 .open = sock_no_open, /* special open code to disallow open via /proc */
148 .release = sock_close,
149 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200150 .sendpage = sock_sendpage,
151 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800152 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153};
154
155/*
156 * The protocol list. Each protocol is registered in here.
157 */
158
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000160static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162/*
163 * Statistics counters of the socket lists
164 */
165
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700166static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
168/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700169 * Support routines.
170 * Move socket addresses back and forth across the kernel/user
171 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 */
173
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174/**
175 * move_addr_to_kernel - copy a socket address into kernel space
176 * @uaddr: Address in user space
177 * @kaddr: Address in kernel space
178 * @ulen: Length in user space
179 *
180 * The address is copied into kernel space. If the provided address is
181 * too long an error code of -EINVAL is returned. If the copy gives
182 * invalid addresses -EFAULT is returned. On a success 0 is returned.
183 */
184
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000185int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700187 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700189 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700191 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100193 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700194}
195
196/**
197 * move_addr_to_user - copy an address to user space
198 * @kaddr: kernel space address
199 * @klen: length of address in kernel
200 * @uaddr: user space address
201 * @ulen: pointer to user length field
202 *
203 * The value pointed to by ulen on entry is the buffer length available.
204 * This is overwritten with the buffer space used. -EINVAL is returned
205 * if an overlong buffer is specified or a negative buffer size. -EFAULT
206 * is returned if either the buffer or the length field are not
207 * accessible.
208 * After copying the data up to the limit the user specifies, the true
209 * length of the data is written over the length limit the user
210 * specified. Zero is returned for a success.
211 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700212
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000213static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000214 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215{
216 int err;
217 int len;
218
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700219 err = get_user(len, ulen);
220 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700222 if (len > klen)
223 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700224 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700226 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500227 if (audit_sockaddr(klen, kaddr))
228 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700229 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return -EFAULT;
231 }
232 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700233 * "fromlen shall refer to the value before truncation.."
234 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 */
236 return __put_user(klen, ulen);
237}
238
Christoph Lametere18b8902006-12-06 20:33:20 -0800239static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
241static struct inode *sock_alloc_inode(struct super_block *sb)
242{
243 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000244 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700245
Christoph Lametere94b1762006-12-06 20:33:17 -0800246 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 if (!ei)
248 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000249 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
250 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000251 kmem_cache_free(sock_inode_cachep, ei);
252 return NULL;
253 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000254 init_waitqueue_head(&wq->wait);
255 wq->fasync_list = NULL;
256 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700257
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258 ei->socket.state = SS_UNCONNECTED;
259 ei->socket.flags = 0;
260 ei->socket.ops = NULL;
261 ei->socket.sk = NULL;
262 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263
264 return &ei->vfs_inode;
265}
266
267static void sock_destroy_inode(struct inode *inode)
268{
Eric Dumazet43815482010-04-29 11:01:49 +0000269 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000270 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000271
272 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000273 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800274 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000275 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276}
277
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700278static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700280 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
Christoph Lametera35afb82007-05-16 22:10:57 -0700282 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285static int init_inodecache(void)
286{
287 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700288 sizeof(struct socket_alloc),
289 0,
290 (SLAB_HWCACHE_ALIGN |
291 SLAB_RECLAIM_ACCOUNT |
292 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900293 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 if (sock_inode_cachep == NULL)
295 return -ENOMEM;
296 return 0;
297}
298
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700299static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700300 .alloc_inode = sock_alloc_inode,
301 .destroy_inode = sock_destroy_inode,
302 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303};
304
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700305/*
306 * sockfs_dname() is called from d_path().
307 */
308static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
309{
310 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
311 dentry->d_inode->i_ino);
312}
313
Al Viro3ba13d12009-02-20 06:02:22 +0000314static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700315 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316};
317
Al Viroc74a1cb2011-01-12 16:59:34 -0500318static struct dentry *sockfs_mount(struct file_system_type *fs_type,
319 int flags, const char *dev_name, void *data)
320{
321 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
322 &sockfs_dentry_operations, SOCKFS_MAGIC);
323}
324
325static struct vfsmount *sock_mnt __read_mostly;
326
327static struct file_system_type sock_fs_type = {
328 .name = "sockfs",
329 .mount = sockfs_mount,
330 .kill_sb = kill_anon_super,
331};
332
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333/*
334 * Obtains the first available file descriptor and sets it up for use.
335 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800336 * These functions create file structures and maps them to fd space
337 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 * and file struct implicitly stored in sock->file.
339 * Note that another thread may close file descriptor before we return
340 * from this function. We use the fact that now we do not refer
341 * to socket after mapping. If one day we will need it, this
342 * function will increment ref. count on file by 1.
343 *
344 * In any case returned fd MAY BE not valid!
345 * This race condition is unavoidable
346 * with shared fd spaces, we cannot solve it inside kernel,
347 * but we take care of internal coherence yet.
348 */
349
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700350struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351{
Al Viro7cbe66b2009-08-05 19:59:08 +0400352 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400353 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400354 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800355
Masatake YAMATO600e1772012-08-29 10:44:29 +0000356 if (dname) {
357 name.name = dname;
358 name.len = strlen(name.name);
359 } else if (sock->sk) {
360 name.name = sock->sk->sk_prot_creator->name;
361 name.len = strlen(name.name);
362 }
Nick Piggin4b936882011-01-07 17:50:07 +1100363 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400364 if (unlikely(!path.dentry))
365 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400366 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367
Al Viro2c48b9c2009-08-09 00:52:35 +0400368 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400369 SOCK_INODE(sock)->i_fop = &socket_file_ops;
370
Al Viro2c48b9c2009-08-09 00:52:35 +0400371 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400372 &socket_file_ops);
373 if (unlikely(!file)) {
374 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400375 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400376 path_put(&path);
Al Viro28407632012-08-17 23:54:15 -0400377 return ERR_PTR(-ENFILE);
Al Virocc3808f2009-08-06 09:43:59 +0400378 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800379
380 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700381 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800382 file->f_pos = 0;
383 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400384 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800385}
Al Viro56b31d12012-08-18 00:25:51 -0400386EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387
Al Viro56b31d12012-08-18 00:25:51 -0400388static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800389{
390 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400391 int fd = get_unused_fd_flags(flags);
392 if (unlikely(fd < 0))
393 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800394
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700395 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400396 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800397 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400398 return fd;
399 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400400
Al Viro28407632012-08-17 23:54:15 -0400401 put_unused_fd(fd);
402 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403}
404
John Fastabend406a3c62012-07-20 10:39:25 +0000405struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800406{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800407 if (file->f_op == &socket_file_ops)
408 return file->private_data; /* set in sock_map_fd */
409
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800410 *err = -ENOTSOCK;
411 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800412}
John Fastabend406a3c62012-07-20 10:39:25 +0000413EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800414
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
Masatake YAMATO600e1772012-08-29 10:44:29 +0000462#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
463#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
464#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
465static ssize_t sockfs_getxattr(struct dentry *dentry,
466 const char *name, void *value, size_t size)
467{
468 const char *proto_name;
469 size_t proto_size;
470 int error;
471
472 error = -ENODATA;
473 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
474 proto_name = dentry->d_name.name;
475 proto_size = strlen(proto_name);
476
477 if (value) {
478 error = -ERANGE;
479 if (proto_size + 1 > size)
480 goto out;
481
482 strncpy(value, proto_name, proto_size + 1);
483 }
484 error = proto_size + 1;
485 }
486
487out:
488 return error;
489}
490
491static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
492 size_t size)
493{
494 ssize_t len;
495 ssize_t used = 0;
496
497 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
498 if (len < 0)
499 return len;
500 used += len;
501 if (buffer) {
502 if (size < used)
503 return -ERANGE;
504 buffer += len;
505 }
506
507 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
508 used += len;
509 if (buffer) {
510 if (size < used)
511 return -ERANGE;
512 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
513 buffer += len;
514 }
515
516 return used;
517}
518
519static const struct inode_operations sockfs_inode_ops = {
520 .getxattr = sockfs_getxattr,
521 .listxattr = sockfs_listxattr,
522};
523
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524/**
525 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700526 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 * Allocate a new inode and socket object. The two are bound together
528 * and initialised. The socket is then returned. If we are out of inodes
529 * NULL is returned.
530 */
531
532static struct socket *sock_alloc(void)
533{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700534 struct inode *inode;
535 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200537 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 if (!inode)
539 return NULL;
540
541 sock = SOCKET_I(inode);
542
Eric Dumazet29a020d2009-09-15 02:39:20 -0700543 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400544 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700545 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100546 inode->i_uid = current_fsuid();
547 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000548 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
Alex Shi19e8d692012-05-14 14:15:31 -0700550 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 return sock;
552}
553
554/*
555 * In theory you can't get an open on this inode, but /proc provides
556 * a back door. Remember to keep it shut otherwise you'll let the
557 * creepy crawlies in.
558 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700559
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
561{
562 return -ENXIO;
563}
564
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800565const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 .owner = THIS_MODULE,
567 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200568 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569};
570
571/**
572 * sock_release - close a socket
573 * @sock: socket to close
574 *
575 * The socket is released from the protocol stack if it has a release
576 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700577 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700579
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580void sock_release(struct socket *sock)
581{
582 if (sock->ops) {
583 struct module *owner = sock->ops->owner;
584
585 sock->ops->release(sock);
586 sock->ops = NULL;
587 module_put(owner);
588 }
589
Eric Dumazeteaefd112011-02-18 03:26:36 +0000590 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 printk(KERN_ERR "sock_release: fasync list not empty!\n");
592
Mikulas Patockab09e7862012-07-19 06:13:36 +0000593 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
594 return;
595
Alex Shi19e8d692012-05-14 14:15:31 -0700596 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 if (!sock->file) {
598 iput(SOCK_INODE(sock));
599 return;
600 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700601 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700603EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000605int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000606{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000607 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000608 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000609 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000610 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000611 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100612 if (sock_flag(sk, SOCK_WIFI_STATUS))
613 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000614 return 0;
615}
616EXPORT_SYMBOL(sock_tx_timestamp);
617
Anton Blanchard228e5482011-05-02 20:21:35 +0000618static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
619 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620{
621 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623 si->sock = sock;
624 si->scm = NULL;
625 si->msg = msg;
626 si->size = size;
627
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628 return sock->ops->sendmsg(iocb, sock, msg, size);
629}
630
Anton Blanchard228e5482011-05-02 20:21:35 +0000631static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
632 struct msghdr *msg, size_t size)
633{
634 int err = security_socket_sendmsg(sock, msg, size);
635
636 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
637}
638
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
640{
641 struct kiocb iocb;
642 struct sock_iocb siocb;
643 int ret;
644
645 init_sync_kiocb(&iocb, NULL);
646 iocb.private = &siocb;
647 ret = __sock_sendmsg(&iocb, sock, msg, size);
648 if (-EIOCBQUEUED == ret)
649 ret = wait_on_sync_kiocb(&iocb);
650 return ret;
651}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700652EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Eric Dumazet894dc242011-07-26 02:39:41 +0000654static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000655{
656 struct kiocb iocb;
657 struct sock_iocb siocb;
658 int ret;
659
660 init_sync_kiocb(&iocb, NULL);
661 iocb.private = &siocb;
662 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
663 if (-EIOCBQUEUED == ret)
664 ret = wait_on_sync_kiocb(&iocb);
665 return ret;
666}
667
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
669 struct kvec *vec, size_t num, size_t size)
670{
671 mm_segment_t oldfs = get_fs();
672 int result;
673
674 set_fs(KERNEL_DS);
675 /*
676 * the following is safe, since for compiler definitions of kvec and
677 * iovec are identical, yielding the same in-core layout and alignment
678 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700679 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 msg->msg_iovlen = num;
681 result = sock_sendmsg(sock, msg, size);
682 set_fs(oldfs);
683 return result;
684}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700685EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686
Patrick Ohly20d49472009-02-12 05:03:38 +0000687static int ktime2ts(ktime_t kt, struct timespec *ts)
688{
689 if (kt.tv64) {
690 *ts = ktime_to_timespec(kt);
691 return 1;
692 } else {
693 return 0;
694 }
695}
696
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700697/*
698 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
699 */
700void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
701 struct sk_buff *skb)
702{
Patrick Ohly20d49472009-02-12 05:03:38 +0000703 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
704 struct timespec ts[3];
705 int empty = 1;
706 struct skb_shared_hwtstamps *shhwtstamps =
707 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700708
Patrick Ohly20d49472009-02-12 05:03:38 +0000709 /* Race occurred between timestamp enabling and packet
710 receiving. Fill in the current time for now. */
711 if (need_software_tstamp && skb->tstamp.tv64 == 0)
712 __net_timestamp(skb);
713
714 if (need_software_tstamp) {
715 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
716 struct timeval tv;
717 skb_get_timestamp(skb, &tv);
718 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
719 sizeof(tv), &tv);
720 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000721 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000722 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000723 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000724 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700725 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000726
727
728 memset(ts, 0, sizeof(ts));
729 if (skb->tstamp.tv64 &&
730 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
731 skb_get_timestampns(skb, ts + 0);
732 empty = 0;
733 }
734 if (shhwtstamps) {
735 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
736 ktime2ts(shhwtstamps->syststamp, ts + 1))
737 empty = 0;
738 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
739 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
740 empty = 0;
741 }
742 if (!empty)
743 put_cmsg(msg, SOL_SOCKET,
744 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700745}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300746EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
747
Johannes Berg6e3e9392011-11-09 10:15:42 +0100748void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
749 struct sk_buff *skb)
750{
751 int ack;
752
753 if (!sock_flag(sk, SOCK_WIFI_STATUS))
754 return;
755 if (!skb->wifi_acked_valid)
756 return;
757
758 ack = skb->wifi_acked;
759
760 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
761}
762EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
763
stephen hemminger11165f12010-10-18 14:27:29 +0000764static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
765 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700766{
767 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
768 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
769 sizeof(__u32), &skb->dropcount);
770}
771
Eric Dumazet767dd032010-04-28 19:14:43 +0000772void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700773 struct sk_buff *skb)
774{
775 sock_recv_timestamp(msg, sk, skb);
776 sock_recv_drops(msg, sk, skb);
777}
Eric Dumazet767dd032010-04-28 19:14:43 +0000778EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700779
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700780static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
781 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 struct sock_iocb *si = kiocb_to_siocb(iocb);
784
785 si->sock = sock;
786 si->scm = NULL;
787 si->msg = msg;
788 si->size = size;
789 si->flags = flags;
790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
792}
793
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700794static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
795 struct msghdr *msg, size_t size, int flags)
796{
797 int err = security_socket_recvmsg(sock, msg, size, flags);
798
799 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
800}
801
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700802int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803 size_t size, int flags)
804{
805 struct kiocb iocb;
806 struct sock_iocb siocb;
807 int ret;
808
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700809 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810 iocb.private = &siocb;
811 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
812 if (-EIOCBQUEUED == ret)
813 ret = wait_on_sync_kiocb(&iocb);
814 return ret;
815}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700816EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700818static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
819 size_t size, int flags)
820{
821 struct kiocb iocb;
822 struct sock_iocb siocb;
823 int ret;
824
825 init_sync_kiocb(&iocb, NULL);
826 iocb.private = &siocb;
827 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
828 if (-EIOCBQUEUED == ret)
829 ret = wait_on_sync_kiocb(&iocb);
830 return ret;
831}
832
Martin Lucinac1249c02010-12-10 00:04:05 +0000833/**
834 * kernel_recvmsg - Receive a message from a socket (kernel space)
835 * @sock: The socket to receive the message from
836 * @msg: Received message
837 * @vec: Input s/g array for message data
838 * @num: Size of input s/g array
839 * @size: Number of bytes to read
840 * @flags: Message flags (MSG_DONTWAIT, etc...)
841 *
842 * On return the msg structure contains the scatter/gather array passed in the
843 * vec argument. The array is modified so that it consists of the unfilled
844 * portion of the original array.
845 *
846 * The returned value is the total number of bytes received, or an error.
847 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700848int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
849 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
851 mm_segment_t oldfs = get_fs();
852 int result;
853
854 set_fs(KERNEL_DS);
855 /*
856 * the following is safe, since for compiler definitions of kvec and
857 * iovec are identical, yielding the same in-core layout and alignment
858 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700859 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 result = sock_recvmsg(sock, msg, size, flags);
861 set_fs(oldfs);
862 return result;
863}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700864EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
866static void sock_aio_dtor(struct kiocb *iocb)
867{
868 kfree(iocb->private);
869}
870
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300871static ssize_t sock_sendpage(struct file *file, struct page *page,
872 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873{
874 struct socket *sock;
875 int flags;
876
Eric Dumazetb69aee02005-09-06 14:42:45 -0700877 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Eric Dumazet35f9c092012-04-05 03:05:35 +0000879 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
880 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
881 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882
Linus Torvaldse6949582009-08-13 08:28:36 -0700883 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884}
885
Jens Axboe9c55e012007-11-06 23:30:13 -0800886static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700887 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800888 unsigned int flags)
889{
890 struct socket *sock = file->private_data;
891
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800892 if (unlikely(!sock->ops->splice_read))
893 return -EINVAL;
894
Jens Axboe9c55e012007-11-06 23:30:13 -0800895 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
896}
897
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800898static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700899 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800900{
901 if (!is_sync_kiocb(iocb)) {
902 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
903 if (!siocb)
904 return NULL;
905 iocb->ki_dtor = sock_aio_dtor;
906 }
907
908 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800909 iocb->private = siocb;
910 return siocb;
911}
912
913static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700914 struct file *file, const struct iovec *iov,
915 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800916{
917 struct socket *sock = file->private_data;
918 size_t size = 0;
919 int i;
920
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700921 for (i = 0; i < nr_segs; i++)
922 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800923
924 msg->msg_name = NULL;
925 msg->msg_namelen = 0;
926 msg->msg_control = NULL;
927 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700928 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800929 msg->msg_iovlen = nr_segs;
930 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
931
932 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
933}
934
Badari Pulavarty027445c2006-09-30 23:28:46 -0700935static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
936 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800937{
938 struct sock_iocb siocb, *x;
939
940 if (pos != 0)
941 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700942
943 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800944 return 0;
945
Badari Pulavarty027445c2006-09-30 23:28:46 -0700946
947 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800948 if (!x)
949 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700950 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800951}
952
953static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700954 struct file *file, const struct iovec *iov,
955 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800956{
957 struct socket *sock = file->private_data;
958 size_t size = 0;
959 int i;
960
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700961 for (i = 0; i < nr_segs; i++)
962 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800963
964 msg->msg_name = NULL;
965 msg->msg_namelen = 0;
966 msg->msg_control = NULL;
967 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700968 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800969 msg->msg_iovlen = nr_segs;
970 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
971 if (sock->type == SOCK_SEQPACKET)
972 msg->msg_flags |= MSG_EOR;
973
974 return __sock_sendmsg(iocb, sock, msg, size);
975}
976
Badari Pulavarty027445c2006-09-30 23:28:46 -0700977static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
978 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800980 struct sock_iocb siocb, *x;
981
982 if (pos != 0)
983 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700984
Badari Pulavarty027445c2006-09-30 23:28:46 -0700985 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800986 if (!x)
987 return -ENOMEM;
988
Badari Pulavarty027445c2006-09-30 23:28:46 -0700989 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990}
991
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992/*
993 * Atomic setting of ioctl hooks to avoid race
994 * with module unload.
995 */
996
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800997static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700998static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
Eric W. Biederman881d9662007-09-17 11:56:21 -07001000void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001002 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001004 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005}
1006EXPORT_SYMBOL(brioctl_set);
1007
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001008static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001009static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Eric W. Biederman881d9662007-09-17 11:56:21 -07001011void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001013 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001015 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016}
1017EXPORT_SYMBOL(vlan_ioctl_set);
1018
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001019static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001020static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001022void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001024 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001026 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027}
1028EXPORT_SYMBOL(dlci_ioctl_set);
1029
Arnd Bergmann6b960182009-11-06 23:10:54 -08001030static long sock_do_ioctl(struct net *net, struct socket *sock,
1031 unsigned int cmd, unsigned long arg)
1032{
1033 int err;
1034 void __user *argp = (void __user *)arg;
1035
1036 err = sock->ops->ioctl(sock, cmd, arg);
1037
1038 /*
1039 * If this ioctl is unknown try to hand it down
1040 * to the NIC driver.
1041 */
1042 if (err == -ENOIOCTLCMD)
1043 err = dev_ioctl(net, cmd, argp);
1044
1045 return err;
1046}
1047
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048/*
1049 * With an ioctl, arg may well be a user mode pointer, but we don't know
1050 * what to do with it - that's up to the protocol still.
1051 */
1052
1053static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1054{
1055 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001056 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 void __user *argp = (void __user *)arg;
1058 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001059 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
Eric Dumazetb69aee02005-09-06 14:42:45 -07001061 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001062 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001063 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001065 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001067#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001069 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001071#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001072 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 case FIOSETOWN:
1074 case SIOCSPGRP:
1075 err = -EFAULT;
1076 if (get_user(pid, (int __user *)argp))
1077 break;
1078 err = f_setown(sock->file, pid, 1);
1079 break;
1080 case FIOGETOWN:
1081 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001082 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001083 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 break;
1085 case SIOCGIFBR:
1086 case SIOCSIFBR:
1087 case SIOCBRADDBR:
1088 case SIOCBRDELBR:
1089 err = -ENOPKG;
1090 if (!br_ioctl_hook)
1091 request_module("bridge");
1092
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001093 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001094 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001095 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001096 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 break;
1098 case SIOCGIFVLAN:
1099 case SIOCSIFVLAN:
1100 err = -ENOPKG;
1101 if (!vlan_ioctl_hook)
1102 request_module("8021q");
1103
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001104 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001106 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001107 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 case SIOCADDDLCI:
1110 case SIOCDELDLCI:
1111 err = -ENOPKG;
1112 if (!dlci_ioctl_hook)
1113 request_module("dlci");
1114
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001115 mutex_lock(&dlci_ioctl_mutex);
1116 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001118 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 break;
1120 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001121 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001123 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 return err;
1125}
1126
1127int sock_create_lite(int family, int type, int protocol, struct socket **res)
1128{
1129 int err;
1130 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001131
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 err = security_socket_create(family, type, protocol, 1);
1133 if (err)
1134 goto out;
1135
1136 sock = sock_alloc();
1137 if (!sock) {
1138 err = -ENOMEM;
1139 goto out;
1140 }
1141
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001143 err = security_socket_post_create(sock, family, type, protocol, 1);
1144 if (err)
1145 goto out_release;
1146
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147out:
1148 *res = sock;
1149 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001150out_release:
1151 sock_release(sock);
1152 sock = NULL;
1153 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001155EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
1157/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001158static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159{
1160 struct socket *sock;
1161
1162 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001163 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001165 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 return sock->ops->poll(file, sock, wait);
1167}
1168
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001169static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001171 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
1173 return sock->ops->mmap(file, sock, vma);
1174}
1175
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001176static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177{
1178 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001179 * It was possible the inode is NULL we were
1180 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 */
1182
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001183 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 printk(KERN_DEBUG "sock_close: NULL inode\n");
1185 return 0;
1186 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 sock_release(SOCKET_I(inode));
1188 return 0;
1189}
1190
1191/*
1192 * Update the socket async list
1193 *
1194 * Fasync_list locking strategy.
1195 *
1196 * 1. fasync_list is modified only under process context socket lock
1197 * i.e. under semaphore.
1198 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001199 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 */
1201
1202static int sock_fasync(int fd, struct file *filp, int on)
1203{
Eric Dumazet989a2972010-04-14 09:55:35 +00001204 struct socket *sock = filp->private_data;
1205 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001206 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
Eric Dumazet989a2972010-04-14 09:55:35 +00001208 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210
1211 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001212 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1213 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214
Eric Dumazeteaefd112011-02-18 03:26:36 +00001215 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001216 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001217 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001218 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219
Eric Dumazet989a2972010-04-14 09:55:35 +00001220 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 return 0;
1222}
1223
Eric Dumazet43815482010-04-29 11:01:49 +00001224/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225
1226int sock_wake_async(struct socket *sock, int how, int band)
1227{
Eric Dumazet43815482010-04-29 11:01:49 +00001228 struct socket_wq *wq;
1229
1230 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001232 rcu_read_lock();
1233 wq = rcu_dereference(sock->wq);
1234 if (!wq || !wq->fasync_list) {
1235 rcu_read_unlock();
1236 return -1;
1237 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001238 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001239 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1241 break;
1242 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001243 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1245 break;
1246 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001247 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001248call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001249 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001251 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001252 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 }
Eric Dumazet43815482010-04-29 11:01:49 +00001254 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 return 0;
1256}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001257EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
Pavel Emelyanov721db932010-09-29 16:06:32 +04001259int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001260 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261{
1262 int err;
1263 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001264 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
1266 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001267 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 */
1269 if (family < 0 || family >= NPROTO)
1270 return -EAFNOSUPPORT;
1271 if (type < 0 || type >= SOCK_MAX)
1272 return -EINVAL;
1273
1274 /* Compatibility.
1275
1276 This uglymoron is moved from INET layer to here to avoid
1277 deadlock in module load.
1278 */
1279 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001280 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 if (!warned) {
1282 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001283 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1284 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 }
1286 family = PF_PACKET;
1287 }
1288
1289 err = security_socket_create(family, type, protocol, kern);
1290 if (err)
1291 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001292
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001293 /*
1294 * Allocate the socket and allow the family to set things up. if
1295 * the protocol is 0, the family is instructed to select an appropriate
1296 * default.
1297 */
1298 sock = sock_alloc();
1299 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001300 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001301 return -ENFILE; /* Not exactly a match, but its the
1302 closest posix thing */
1303 }
1304
1305 sock->type = type;
1306
Johannes Berg95a5afc2008-10-16 15:24:51 -07001307#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001308 /* Attempt to load a protocol module if the find failed.
1309 *
1310 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 * requested real, full-featured networking support upon configuration.
1312 * Otherwise module support will break!
1313 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001314 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001315 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316#endif
1317
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001318 rcu_read_lock();
1319 pf = rcu_dereference(net_families[family]);
1320 err = -EAFNOSUPPORT;
1321 if (!pf)
1322 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323
1324 /*
1325 * We will call the ->create function, that possibly is in a loadable
1326 * module, so we have to bump that loadable module refcnt first.
1327 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001328 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 goto out_release;
1330
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001331 /* Now protected by module ref count */
1332 rcu_read_unlock();
1333
Eric Paris3f378b62009-11-05 22:18:14 -08001334 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001335 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001337
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 /*
1339 * Now to bump the refcnt of the [loadable] module that owns this
1340 * socket at sock_release time we decrement its refcnt.
1341 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001342 if (!try_module_get(sock->ops->owner))
1343 goto out_module_busy;
1344
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 /*
1346 * Now that we're done with the ->create function, the [loadable]
1347 * module can have its refcnt decremented
1348 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001349 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001350 err = security_socket_post_create(sock, family, type, protocol, kern);
1351 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001352 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001353 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001355 return 0;
1356
1357out_module_busy:
1358 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001360 sock->ops = NULL;
1361 module_put(pf->owner);
1362out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001364 return err;
1365
1366out_release:
1367 rcu_read_unlock();
1368 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001370EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371
1372int sock_create(int family, int type, int protocol, struct socket **res)
1373{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001374 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001376EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377
1378int sock_create_kern(int family, int type, int protocol, struct socket **res)
1379{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001380 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001382EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001384SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385{
1386 int retval;
1387 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001388 int flags;
1389
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001390 /* Check the SOCK_* constants for consistency. */
1391 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1392 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1393 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1394 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1395
Ulrich Dreppera677a032008-07-23 21:29:17 -07001396 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001397 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001398 return -EINVAL;
1399 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001401 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1402 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1403
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 retval = sock_create(family, type, protocol, &sock);
1405 if (retval < 0)
1406 goto out;
1407
Ulrich Drepper77d27202008-07-23 21:29:35 -07001408 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 if (retval < 0)
1410 goto out_release;
1411
1412out:
1413 /* It may be already another descriptor 8) Not kernel problem. */
1414 return retval;
1415
1416out_release:
1417 sock_release(sock);
1418 return retval;
1419}
1420
1421/*
1422 * Create a pair of connected sockets.
1423 */
1424
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001425SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1426 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427{
1428 struct socket *sock1, *sock2;
1429 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001430 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001431 int flags;
1432
1433 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001434 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001435 return -EINVAL;
1436 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001438 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1439 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1440
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 /*
1442 * Obtain the first socket and check if the underlying protocol
1443 * supports the socketpair call.
1444 */
1445
1446 err = sock_create(family, type, protocol, &sock1);
1447 if (err < 0)
1448 goto out;
1449
1450 err = sock_create(family, type, protocol, &sock2);
1451 if (err < 0)
1452 goto out_release_1;
1453
1454 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001455 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 goto out_release_both;
1457
Al Viro28407632012-08-17 23:54:15 -04001458 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001459 if (unlikely(fd1 < 0)) {
1460 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001462 }
Al Viro28407632012-08-17 23:54:15 -04001463 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001464 if (unlikely(fd2 < 0)) {
1465 err = fd2;
Al Viro28407632012-08-17 23:54:15 -04001466 put_unused_fd(fd1);
1467 goto out_release_both;
1468 }
1469
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001470 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001471 if (unlikely(IS_ERR(newfile1))) {
1472 err = PTR_ERR(newfile1);
1473 put_unused_fd(fd1);
1474 put_unused_fd(fd2);
1475 goto out_release_both;
1476 }
1477
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001478 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001479 if (IS_ERR(newfile2)) {
1480 err = PTR_ERR(newfile2);
Al Viro198de4d2009-08-05 19:29:23 +04001481 fput(newfile1);
1482 put_unused_fd(fd1);
Al Viro28407632012-08-17 23:54:15 -04001483 put_unused_fd(fd2);
Al Viro198de4d2009-08-05 19:29:23 +04001484 sock_release(sock2);
1485 goto out;
Al Virodb349502007-02-07 01:48:00 -05001486 }
1487
Al Viro157cf642008-12-14 04:57:47 -05001488 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001489 fd_install(fd1, newfile1);
1490 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 /* fd1 and fd2 may be already another descriptors.
1492 * Not kernel problem.
1493 */
1494
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001495 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 if (!err)
1497 err = put_user(fd2, &usockvec[1]);
1498 if (!err)
1499 return 0;
1500
1501 sys_close(fd2);
1502 sys_close(fd1);
1503 return err;
1504
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001506 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001508 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509out:
1510 return err;
1511}
1512
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513/*
1514 * Bind a name to a socket. Nothing much to do here since it's
1515 * the protocol's responsibility to handle the local address.
1516 *
1517 * We move the socket address to kernel space before we call
1518 * the protocol layer (having also checked the address is ok).
1519 */
1520
Heiko Carstens20f37032009-01-14 14:14:23 +01001521SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522{
1523 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001524 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001525 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001527 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001528 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001529 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001530 if (err >= 0) {
1531 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001532 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001533 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001534 if (!err)
1535 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001536 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001537 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001539 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 return err;
1542}
1543
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544/*
1545 * Perform a listen. Basically, we allow the protocol to do anything
1546 * necessary for a listen, and if that works, we mark the socket as
1547 * ready for listening.
1548 */
1549
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001550SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551{
1552 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001553 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001554 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001555
1556 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1557 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001558 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001559 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001560 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561
1562 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001563 if (!err)
1564 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001566 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 }
1568 return err;
1569}
1570
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571/*
1572 * For accept, we attempt to create a new socket, set up the link
1573 * with the client, wake up the client, then return the new
1574 * connected fd. We collect the address of the connector in kernel
1575 * space and move it to user at the very end. This is unclean because
1576 * we open the socket then return an error.
1577 *
1578 * 1003.1g adds the ability to recvmsg() to query connection pending
1579 * status to recvmsg. We need to add that support in a way thats
1580 * clean when we restucture accept also.
1581 */
1582
Heiko Carstens20f37032009-01-14 14:14:23 +01001583SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1584 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585{
1586 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001587 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001588 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001589 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590
Ulrich Drepper77d27202008-07-23 21:29:35 -07001591 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001592 return -EINVAL;
1593
1594 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1595 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1596
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001597 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 if (!sock)
1599 goto out;
1600
1601 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001602 newsock = sock_alloc();
1603 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 goto out_put;
1605
1606 newsock->type = sock->type;
1607 newsock->ops = sock->ops;
1608
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 /*
1610 * We don't need try_module_get here, as the listening socket (sock)
1611 * has the protocol module (sock->ops->owner) held.
1612 */
1613 __module_get(newsock->ops->owner);
1614
Al Viro28407632012-08-17 23:54:15 -04001615 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001616 if (unlikely(newfd < 0)) {
1617 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001618 sock_release(newsock);
1619 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001620 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001621 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001622 if (unlikely(IS_ERR(newfile))) {
1623 err = PTR_ERR(newfile);
1624 put_unused_fd(newfd);
1625 sock_release(newsock);
1626 goto out_put;
1627 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001628
Frank Filza79af592005-09-27 15:23:38 -07001629 err = security_socket_accept(sock, newsock);
1630 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001631 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001632
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1634 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001635 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636
1637 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001638 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001639 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001641 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001643 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001644 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001646 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 }
1648
1649 /* File flags are not inherited via accept() unlike another OSes. */
1650
David S. Miller39d8c1b2006-03-20 17:13:49 -08001651 fd_install(newfd, newfile);
1652 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001655 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656out:
1657 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001658out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001659 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001660 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 goto out_put;
1662}
1663
Heiko Carstens20f37032009-01-14 14:14:23 +01001664SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1665 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001666{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001667 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001668}
1669
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670/*
1671 * Attempt to connect to a socket with the server address. The address
1672 * is in user space so we verify it is OK and move it to kernel space.
1673 *
1674 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1675 * break bindings
1676 *
1677 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1678 * other SEQPACKET protocols that take time to connect() as it doesn't
1679 * include the -EINPROGRESS status for such sockets.
1680 */
1681
Heiko Carstens20f37032009-01-14 14:14:23 +01001682SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1683 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684{
1685 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001686 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001687 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001689 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 if (!sock)
1691 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001692 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 if (err < 0)
1694 goto out_put;
1695
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001696 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001697 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 if (err)
1699 goto out_put;
1700
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001701 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 sock->file->f_flags);
1703out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001704 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705out:
1706 return err;
1707}
1708
1709/*
1710 * Get the local address ('name') of a socket object. Move the obtained
1711 * name to user space.
1712 */
1713
Heiko Carstens20f37032009-01-14 14:14:23 +01001714SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1715 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716{
1717 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001718 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001719 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001720
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001721 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 if (!sock)
1723 goto out;
1724
1725 err = security_socket_getsockname(sock);
1726 if (err)
1727 goto out_put;
1728
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001729 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 if (err)
1731 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001732 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
1734out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001735 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736out:
1737 return err;
1738}
1739
1740/*
1741 * Get the remote address ('name') of a socket object. Move the obtained
1742 * name to user space.
1743 */
1744
Heiko Carstens20f37032009-01-14 14:14:23 +01001745SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1746 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747{
1748 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001749 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001750 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001752 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1753 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 err = security_socket_getpeername(sock);
1755 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001756 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 return err;
1758 }
1759
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001760 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001761 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001762 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001764 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001765 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001766 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 }
1768 return err;
1769}
1770
1771/*
1772 * Send a datagram to a given address. We move the address into kernel
1773 * space and check the user space data area is readable before invoking
1774 * the protocol.
1775 */
1776
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001777SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001778 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001779 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780{
1781 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001782 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 int err;
1784 struct msghdr msg;
1785 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001786 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001787
Linus Torvalds253eacc2010-10-30 16:43:10 -07001788 if (len > INT_MAX)
1789 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001790 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1791 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001792 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001793
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001794 iov.iov_base = buff;
1795 iov.iov_len = len;
1796 msg.msg_name = NULL;
1797 msg.msg_iov = &iov;
1798 msg.msg_iovlen = 1;
1799 msg.msg_control = NULL;
1800 msg.msg_controllen = 0;
1801 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001802 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001803 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 if (err < 0)
1805 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001806 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001807 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 }
1809 if (sock->file->f_flags & O_NONBLOCK)
1810 flags |= MSG_DONTWAIT;
1811 msg.msg_flags = flags;
1812 err = sock_sendmsg(sock, &msg, len);
1813
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001814out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001815 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001816out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 return err;
1818}
1819
1820/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001821 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822 */
1823
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001824SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001825 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826{
1827 return sys_sendto(fd, buff, len, flags, NULL, 0);
1828}
1829
1830/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001831 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 * sender. We verify the buffers are writable and if needed move the
1833 * sender address from kernel to user space.
1834 */
1835
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001836SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001837 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001838 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839{
1840 struct socket *sock;
1841 struct iovec iov;
1842 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001843 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001844 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001845 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846
Linus Torvalds253eacc2010-10-30 16:43:10 -07001847 if (size > INT_MAX)
1848 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001849 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001851 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001853 msg.msg_control = NULL;
1854 msg.msg_controllen = 0;
1855 msg.msg_iovlen = 1;
1856 msg.msg_iov = &iov;
1857 iov.iov_len = size;
1858 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001859 msg.msg_name = (struct sockaddr *)&address;
1860 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 if (sock->file->f_flags & O_NONBLOCK)
1862 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001863 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001865 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001866 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001867 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001868 if (err2 < 0)
1869 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001871
1872 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001873out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 return err;
1875}
1876
1877/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001878 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 */
1880
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001881asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001882 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883{
1884 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1885}
1886
1887/*
1888 * Set a socket option. Because we don't know the option lengths we have
1889 * to pass the user mode parameter for the protocols to sort out.
1890 */
1891
Heiko Carstens20f37032009-01-14 14:14:23 +01001892SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1893 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001895 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 struct socket *sock;
1897
1898 if (optlen < 0)
1899 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001900
1901 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1902 if (sock != NULL) {
1903 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001904 if (err)
1905 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
1907 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908 err =
1909 sock_setsockopt(sock, level, optname, optval,
1910 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001912 err =
1913 sock->ops->setsockopt(sock, level, optname, optval,
1914 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001915out_put:
1916 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 }
1918 return err;
1919}
1920
1921/*
1922 * Get a socket option. Because we don't know the option lengths we have
1923 * to pass a user mode parameter for the protocols to sort out.
1924 */
1925
Heiko Carstens20f37032009-01-14 14:14:23 +01001926SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1927 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001929 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 struct socket *sock;
1931
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001932 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1933 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001934 err = security_socket_getsockopt(sock, level, optname);
1935 if (err)
1936 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937
1938 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001939 err =
1940 sock_getsockopt(sock, level, optname, optval,
1941 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001943 err =
1944 sock->ops->getsockopt(sock, level, optname, optval,
1945 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001946out_put:
1947 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 }
1949 return err;
1950}
1951
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952/*
1953 * Shutdown a socket.
1954 */
1955
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001956SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001958 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 struct socket *sock;
1960
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001961 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1962 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001964 if (!err)
1965 err = sock->ops->shutdown(sock, how);
1966 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967 }
1968 return err;
1969}
1970
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001971/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 * fields which are the same type (int / unsigned) on our platforms.
1973 */
1974#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1975#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1976#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1977
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001978struct used_address {
1979 struct sockaddr_storage name;
1980 unsigned int name_len;
1981};
1982
Anton Blanchard228e5482011-05-02 20:21:35 +00001983static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001984 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001985 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001987 struct compat_msghdr __user *msg_compat =
1988 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001989 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001991 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001992 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1993 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001995 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001996
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 err = -EFAULT;
1998 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001999 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00002001 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 return -EFAULT;
2003
Anton Blanchard228e5482011-05-02 20:21:35 +00002004 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002005 err = -EMSGSIZE;
2006 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2007 goto out;
2008 err = -ENOMEM;
2009 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2010 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00002012 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 }
2014
2015 /* This will also move the address data into kernel space */
2016 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002017 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002019 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 goto out_freeiov;
2022 total_len = err;
2023
2024 err = -ENOBUFS;
2025
Anton Blanchard228e5482011-05-02 20:21:35 +00002026 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002028 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002030 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002031 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002032 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 if (err)
2034 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002035 ctl_buf = msg_sys->msg_control;
2036 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002038 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002040 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 goto out_freeiov;
2042 }
2043 err = -EFAULT;
2044 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002045 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2047 * checking falls down on this.
2048 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002049 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002050 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002051 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002053 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002055 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056
2057 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002058 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002059 /*
2060 * If this is sendmmsg() and current destination address is same as
2061 * previously succeeded address, omit asking LSM's decision.
2062 * used_address->name_len is initialized to UINT_MAX so that the first
2063 * destination address never matches.
2064 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002065 if (used_address && msg_sys->msg_name &&
2066 used_address->name_len == msg_sys->msg_namelen &&
2067 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002068 used_address->name_len)) {
2069 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2070 goto out_freectl;
2071 }
2072 err = sock_sendmsg(sock, msg_sys, total_len);
2073 /*
2074 * If this is sendmmsg() and sending to current destination address was
2075 * successful, remember it.
2076 */
2077 if (used_address && err >= 0) {
2078 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002079 if (msg_sys->msg_name)
2080 memcpy(&used_address->name, msg_sys->msg_name,
2081 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002082 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083
2084out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002085 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2087out_freeiov:
2088 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002089 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002090out:
2091 return err;
2092}
2093
2094/*
2095 * BSD sendmsg interface
2096 */
2097
Eric Dumazet95c96172012-04-15 05:58:06 +00002098SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002099{
2100 int fput_needed, err;
2101 struct msghdr msg_sys;
2102 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2103
2104 if (!sock)
2105 goto out;
2106
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002107 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002108
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002109 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002110out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 return err;
2112}
2113
Anton Blanchard228e5482011-05-02 20:21:35 +00002114/*
2115 * Linux sendmmsg interface
2116 */
2117
2118int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2119 unsigned int flags)
2120{
2121 int fput_needed, err, datagrams;
2122 struct socket *sock;
2123 struct mmsghdr __user *entry;
2124 struct compat_mmsghdr __user *compat_entry;
2125 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002126 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002127
Anton Blanchard98382f42011-08-04 14:07:39 +00002128 if (vlen > UIO_MAXIOV)
2129 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002130
2131 datagrams = 0;
2132
2133 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2134 if (!sock)
2135 return err;
2136
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002137 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002138 entry = mmsg;
2139 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002140 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002141
2142 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002143 if (MSG_CMSG_COMPAT & flags) {
2144 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002145 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002146 if (err < 0)
2147 break;
2148 err = __put_user(err, &compat_entry->msg_len);
2149 ++compat_entry;
2150 } else {
2151 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002152 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002153 if (err < 0)
2154 break;
2155 err = put_user(err, &entry->msg_len);
2156 ++entry;
2157 }
2158
2159 if (err)
2160 break;
2161 ++datagrams;
2162 }
2163
Anton Blanchard228e5482011-05-02 20:21:35 +00002164 fput_light(sock->file, fput_needed);
2165
Anton Blanchard728ffb82011-08-04 14:07:38 +00002166 /* We only return an error if no datagrams were able to be sent */
2167 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002168 return datagrams;
2169
Anton Blanchard228e5482011-05-02 20:21:35 +00002170 return err;
2171}
2172
2173SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2174 unsigned int, vlen, unsigned int, flags)
2175{
2176 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2177}
2178
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002179static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002180 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002182 struct compat_msghdr __user *msg_compat =
2183 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002185 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002187 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188
2189 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002190 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191
2192 /* user mode address pointers */
2193 struct sockaddr __user *uaddr;
2194 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002195
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002197 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002199 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002200 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002202 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002203 err = -EMSGSIZE;
2204 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2205 goto out;
2206 err = -ENOMEM;
2207 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2208 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002210 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 }
2212
2213 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002214 * Save the user-mode address (verify_iovec will change the
2215 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002217
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002218 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 uaddr_len = COMPAT_NAMELEN(msg);
2220 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002221 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002223 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224 if (err < 0)
2225 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002226 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002228 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2229 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002230
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 if (sock->file->f_flags & O_NONBLOCK)
2232 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002233 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2234 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235 if (err < 0)
2236 goto out_freeiov;
2237 len = err;
2238
2239 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002240 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002241 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002242 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243 if (err < 0)
2244 goto out_freeiov;
2245 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002246 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002247 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002248 if (err)
2249 goto out_freeiov;
2250 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002251 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252 &msg_compat->msg_controllen);
2253 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002254 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255 &msg->msg_controllen);
2256 if (err)
2257 goto out_freeiov;
2258 err = len;
2259
2260out_freeiov:
2261 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002262 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002263out:
2264 return err;
2265}
2266
2267/*
2268 * BSD recvmsg interface
2269 */
2270
2271SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2272 unsigned int, flags)
2273{
2274 int fput_needed, err;
2275 struct msghdr msg_sys;
2276 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2277
2278 if (!sock)
2279 goto out;
2280
2281 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2282
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002283 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284out:
2285 return err;
2286}
2287
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002288/*
2289 * Linux recvmmsg interface
2290 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002292int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2293 unsigned int flags, struct timespec *timeout)
2294{
2295 int fput_needed, err, datagrams;
2296 struct socket *sock;
2297 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002298 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002299 struct msghdr msg_sys;
2300 struct timespec end_time;
2301
2302 if (timeout &&
2303 poll_select_set_timeout(&end_time, timeout->tv_sec,
2304 timeout->tv_nsec))
2305 return -EINVAL;
2306
2307 datagrams = 0;
2308
2309 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2310 if (!sock)
2311 return err;
2312
2313 err = sock_error(sock->sk);
2314 if (err)
2315 goto out_put;
2316
2317 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002318 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002319
2320 while (datagrams < vlen) {
2321 /*
2322 * No need to ask LSM for more than the first datagram.
2323 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002324 if (MSG_CMSG_COMPAT & flags) {
2325 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002326 &msg_sys, flags & ~MSG_WAITFORONE,
2327 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002328 if (err < 0)
2329 break;
2330 err = __put_user(err, &compat_entry->msg_len);
2331 ++compat_entry;
2332 } else {
2333 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002334 &msg_sys, flags & ~MSG_WAITFORONE,
2335 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002336 if (err < 0)
2337 break;
2338 err = put_user(err, &entry->msg_len);
2339 ++entry;
2340 }
2341
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002342 if (err)
2343 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002344 ++datagrams;
2345
Brandon L Black71c5c152010-03-26 16:18:03 +00002346 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2347 if (flags & MSG_WAITFORONE)
2348 flags |= MSG_DONTWAIT;
2349
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002350 if (timeout) {
2351 ktime_get_ts(timeout);
2352 *timeout = timespec_sub(end_time, *timeout);
2353 if (timeout->tv_sec < 0) {
2354 timeout->tv_sec = timeout->tv_nsec = 0;
2355 break;
2356 }
2357
2358 /* Timeout, return less than vlen datagrams */
2359 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2360 break;
2361 }
2362
2363 /* Out of band data, return right away */
2364 if (msg_sys.msg_flags & MSG_OOB)
2365 break;
2366 }
2367
2368out_put:
2369 fput_light(sock->file, fput_needed);
2370
2371 if (err == 0)
2372 return datagrams;
2373
2374 if (datagrams != 0) {
2375 /*
2376 * We may return less entries than requested (vlen) if the
2377 * sock is non block and there aren't enough datagrams...
2378 */
2379 if (err != -EAGAIN) {
2380 /*
2381 * ... or if recvmsg returns an error after we
2382 * received some datagrams, where we record the
2383 * error to return on the next call or if the
2384 * app asks about it using getsockopt(SO_ERROR).
2385 */
2386 sock->sk->sk_err = -err;
2387 }
2388
2389 return datagrams;
2390 }
2391
2392 return err;
2393}
2394
2395SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2396 unsigned int, vlen, unsigned int, flags,
2397 struct timespec __user *, timeout)
2398{
2399 int datagrams;
2400 struct timespec timeout_sys;
2401
2402 if (!timeout)
2403 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2404
2405 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2406 return -EFAULT;
2407
2408 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2409
2410 if (datagrams > 0 &&
2411 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2412 datagrams = -EFAULT;
2413
2414 return datagrams;
2415}
2416
2417#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418/* Argument list sizes for sys_socketcall */
2419#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002420static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002421 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2422 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2423 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002424 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002425};
2426
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427#undef AL
2428
2429/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002430 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 *
2432 * Argument checking cleaned up. Saved 20% in size.
2433 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002434 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 */
2436
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002437SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438{
2439 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002440 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002442 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443
Anton Blanchard228e5482011-05-02 20:21:35 +00002444 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445 return -EINVAL;
2446
Arjan van de Ven47379052009-09-28 12:57:44 -07002447 len = nargs[call];
2448 if (len > sizeof(a))
2449 return -EINVAL;
2450
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002452 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002454
Al Virof3298dc2008-12-10 03:16:51 -05002455 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002456
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002457 a0 = a[0];
2458 a1 = a[1];
2459
2460 switch (call) {
2461 case SYS_SOCKET:
2462 err = sys_socket(a0, a1, a[2]);
2463 break;
2464 case SYS_BIND:
2465 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2466 break;
2467 case SYS_CONNECT:
2468 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2469 break;
2470 case SYS_LISTEN:
2471 err = sys_listen(a0, a1);
2472 break;
2473 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002474 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2475 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002476 break;
2477 case SYS_GETSOCKNAME:
2478 err =
2479 sys_getsockname(a0, (struct sockaddr __user *)a1,
2480 (int __user *)a[2]);
2481 break;
2482 case SYS_GETPEERNAME:
2483 err =
2484 sys_getpeername(a0, (struct sockaddr __user *)a1,
2485 (int __user *)a[2]);
2486 break;
2487 case SYS_SOCKETPAIR:
2488 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2489 break;
2490 case SYS_SEND:
2491 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2492 break;
2493 case SYS_SENDTO:
2494 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2495 (struct sockaddr __user *)a[4], a[5]);
2496 break;
2497 case SYS_RECV:
2498 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2499 break;
2500 case SYS_RECVFROM:
2501 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2502 (struct sockaddr __user *)a[4],
2503 (int __user *)a[5]);
2504 break;
2505 case SYS_SHUTDOWN:
2506 err = sys_shutdown(a0, a1);
2507 break;
2508 case SYS_SETSOCKOPT:
2509 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2510 break;
2511 case SYS_GETSOCKOPT:
2512 err =
2513 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2514 (int __user *)a[4]);
2515 break;
2516 case SYS_SENDMSG:
2517 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2518 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002519 case SYS_SENDMMSG:
2520 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2521 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002522 case SYS_RECVMSG:
2523 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2524 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002525 case SYS_RECVMMSG:
2526 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2527 (struct timespec __user *)a[4]);
2528 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002529 case SYS_ACCEPT4:
2530 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2531 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002532 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002533 default:
2534 err = -EINVAL;
2535 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536 }
2537 return err;
2538}
2539
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002540#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002542/**
2543 * sock_register - add a socket protocol handler
2544 * @ops: description of protocol
2545 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546 * This function is called by a protocol handler that wants to
2547 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002548 * socket interface. The value ops->family coresponds to the
2549 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002551int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002552{
2553 int err;
2554
2555 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002556 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2557 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558 return -ENOBUFS;
2559 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002560
2561 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002562 if (rcu_dereference_protected(net_families[ops->family],
2563 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002564 err = -EEXIST;
2565 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002566 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567 err = 0;
2568 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002569 spin_unlock(&net_family_lock);
2570
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002571 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572 return err;
2573}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002574EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002576/**
2577 * sock_unregister - remove a protocol handler
2578 * @family: protocol family to remove
2579 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 * This function is called by a protocol handler that wants to
2581 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002582 * new socket creation.
2583 *
2584 * If protocol handler is a module, then it can use module reference
2585 * counts to protect against new references. If protocol handler is not
2586 * a module then it needs to provide its own protection in
2587 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002588 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002589void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002591 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002593 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002594 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002595 spin_unlock(&net_family_lock);
2596
2597 synchronize_rcu();
2598
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002599 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002600}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002601EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002602
Andi Kleen77d76ea2005-12-22 12:43:42 -08002603static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002604{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002605 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002606 /*
2607 * Initialize the network sysctl infrastructure.
2608 */
2609 err = net_sysctl_init();
2610 if (err)
2611 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002612
Linus Torvalds1da177e2005-04-16 15:20:36 -07002613 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002614 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002615 */
2616 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002617
2618 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002619 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002620 */
2621
2622 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002623
2624 err = register_filesystem(&sock_fs_type);
2625 if (err)
2626 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002627 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002628 if (IS_ERR(sock_mnt)) {
2629 err = PTR_ERR(sock_mnt);
2630 goto out_mount;
2631 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002632
2633 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634 */
2635
2636#ifdef CONFIG_NETFILTER
2637 netfilter_init();
2638#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002639
Richard Cochranc1f19b52010-07-17 08:49:36 +00002640#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2641 skb_timestamping_init();
2642#endif
2643
Nick Pigginb3e19d92011-01-07 17:50:11 +11002644out:
2645 return err;
2646
2647out_mount:
2648 unregister_filesystem(&sock_fs_type);
2649out_fs:
2650 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002651}
2652
Andi Kleen77d76ea2005-12-22 12:43:42 -08002653core_initcall(sock_init); /* early initcall */
2654
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655#ifdef CONFIG_PROC_FS
2656void socket_seq_show(struct seq_file *seq)
2657{
2658 int cpu;
2659 int counter = 0;
2660
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002661 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002662 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663
2664 /* It can be negative, by the way. 8) */
2665 if (counter < 0)
2666 counter = 0;
2667
2668 seq_printf(seq, "sockets: used %d\n", counter);
2669}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002670#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002671
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002672#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002673static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002674 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002675{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002676 mm_segment_t old_fs = get_fs();
2677 struct timeval ktv;
2678 int err;
2679
2680 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002681 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002682 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002683 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002684 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002685
Arnd Bergmann7a229382009-11-06 23:00:29 -08002686 return err;
2687}
2688
Arnd Bergmann6b960182009-11-06 23:10:54 -08002689static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002690 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002691{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002692 mm_segment_t old_fs = get_fs();
2693 struct timespec kts;
2694 int err;
2695
2696 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002697 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002698 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002699 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002700 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002701
Arnd Bergmann7a229382009-11-06 23:00:29 -08002702 return err;
2703}
2704
Arnd Bergmann6b960182009-11-06 23:10:54 -08002705static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002706{
2707 struct ifreq __user *uifr;
2708 int err;
2709
2710 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002711 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002712 return -EFAULT;
2713
Arnd Bergmann6b960182009-11-06 23:10:54 -08002714 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002715 if (err)
2716 return err;
2717
Arnd Bergmann6b960182009-11-06 23:10:54 -08002718 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002719 return -EFAULT;
2720
2721 return 0;
2722}
2723
Arnd Bergmann6b960182009-11-06 23:10:54 -08002724static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002725{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002726 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002727 struct ifconf ifc;
2728 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002729 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002730 struct ifreq __user *ifr;
2731 unsigned int i, j;
2732 int err;
2733
Arnd Bergmann6b960182009-11-06 23:10:54 -08002734 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002735 return -EFAULT;
2736
Mathias Krause43da5f22012-08-15 11:31:57 +00002737 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002738 if (ifc32.ifcbuf == 0) {
2739 ifc32.ifc_len = 0;
2740 ifc.ifc_len = 0;
2741 ifc.ifc_req = NULL;
2742 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2743 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002744 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2745 sizeof(struct ifreq);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002746 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2747 ifc.ifc_len = len;
2748 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2749 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002750 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002751 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002752 return -EFAULT;
2753 ifr++;
2754 ifr32++;
2755 }
2756 }
2757 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2758 return -EFAULT;
2759
Arnd Bergmann6b960182009-11-06 23:10:54 -08002760 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002761 if (err)
2762 return err;
2763
2764 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2765 return -EFAULT;
2766
2767 ifr = ifc.ifc_req;
2768 ifr32 = compat_ptr(ifc32.ifcbuf);
2769 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002770 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2771 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2772 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002773 return -EFAULT;
2774 ifr32++;
2775 ifr++;
2776 }
2777
2778 if (ifc32.ifcbuf == 0) {
2779 /* Translate from 64-bit structure multiple to
2780 * a 32-bit one.
2781 */
2782 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002783 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002784 ifc32.ifc_len = i;
2785 } else {
2786 ifc32.ifc_len = i;
2787 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002788 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002789 return -EFAULT;
2790
2791 return 0;
2792}
2793
Arnd Bergmann6b960182009-11-06 23:10:54 -08002794static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002795{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002796 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2797 bool convert_in = false, convert_out = false;
2798 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2799 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002800 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002801 u32 rule_cnt = 0, actual_rule_cnt;
2802 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002803 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002804 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002805
2806 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2807 return -EFAULT;
2808
Ben Hutchings3a7da392011-03-17 07:34:32 +00002809 compat_rxnfc = compat_ptr(data);
2810
2811 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002812 return -EFAULT;
2813
Ben Hutchings3a7da392011-03-17 07:34:32 +00002814 /* Most ethtool structures are defined without padding.
2815 * Unfortunately struct ethtool_rxnfc is an exception.
2816 */
2817 switch (ethcmd) {
2818 default:
2819 break;
2820 case ETHTOOL_GRXCLSRLALL:
2821 /* Buffer size is variable */
2822 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2823 return -EFAULT;
2824 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2825 return -ENOMEM;
2826 buf_size += rule_cnt * sizeof(u32);
2827 /* fall through */
2828 case ETHTOOL_GRXRINGS:
2829 case ETHTOOL_GRXCLSRLCNT:
2830 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002831 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002832 convert_out = true;
2833 /* fall through */
2834 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002835 buf_size += sizeof(struct ethtool_rxnfc);
2836 convert_in = true;
2837 break;
2838 }
2839
2840 ifr = compat_alloc_user_space(buf_size);
2841 rxnfc = (void *)ifr + ALIGN(sizeof(struct ifreq), 8);
2842
2843 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2844 return -EFAULT;
2845
2846 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2847 &ifr->ifr_ifru.ifru_data))
2848 return -EFAULT;
2849
2850 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002851 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002852 * fs.ring_cookie and at the end of fs, but nowhere else.
2853 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002854 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2855 sizeof(compat_rxnfc->fs.m_ext) !=
2856 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2857 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002858 BUILD_BUG_ON(
2859 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2860 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2861 offsetof(struct ethtool_rxnfc, fs.location) -
2862 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2863
2864 if (copy_in_user(rxnfc, compat_rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002865 (void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002866 (void *)rxnfc) ||
2867 copy_in_user(&rxnfc->fs.ring_cookie,
2868 &compat_rxnfc->fs.ring_cookie,
2869 (void *)(&rxnfc->fs.location + 1) -
2870 (void *)&rxnfc->fs.ring_cookie) ||
2871 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2872 sizeof(rxnfc->rule_cnt)))
2873 return -EFAULT;
2874 }
2875
2876 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2877 if (ret)
2878 return ret;
2879
2880 if (convert_out) {
2881 if (copy_in_user(compat_rxnfc, rxnfc,
Alexander Duyck127fe532011-04-08 18:01:59 +00002882 (const void *)(&rxnfc->fs.m_ext + 1) -
Ben Hutchings3a7da392011-03-17 07:34:32 +00002883 (const void *)rxnfc) ||
2884 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2885 &rxnfc->fs.ring_cookie,
2886 (const void *)(&rxnfc->fs.location + 1) -
2887 (const void *)&rxnfc->fs.ring_cookie) ||
2888 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2889 sizeof(rxnfc->rule_cnt)))
2890 return -EFAULT;
2891
2892 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2893 /* As an optimisation, we only copy the actual
2894 * number of rules that the underlying
2895 * function returned. Since Mallory might
2896 * change the rule count in user memory, we
2897 * check that it is less than the rule count
2898 * originally given (as the user buffer size),
2899 * which has been range-checked.
2900 */
2901 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2902 return -EFAULT;
2903 if (actual_rule_cnt < rule_cnt)
2904 rule_cnt = actual_rule_cnt;
2905 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2906 &rxnfc->rule_locs[0],
2907 rule_cnt * sizeof(u32)))
2908 return -EFAULT;
2909 }
2910 }
2911
2912 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002913}
2914
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002915static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2916{
2917 void __user *uptr;
2918 compat_uptr_t uptr32;
2919 struct ifreq __user *uifr;
2920
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002921 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002922 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2923 return -EFAULT;
2924
2925 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2926 return -EFAULT;
2927
2928 uptr = compat_ptr(uptr32);
2929
2930 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2931 return -EFAULT;
2932
2933 return dev_ioctl(net, SIOCWANDEV, uifr);
2934}
2935
Arnd Bergmann6b960182009-11-06 23:10:54 -08002936static int bond_ioctl(struct net *net, unsigned int cmd,
2937 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002938{
2939 struct ifreq kifr;
2940 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002941 mm_segment_t old_fs;
2942 int err;
2943 u32 data;
2944 void __user *datap;
2945
2946 switch (cmd) {
2947 case SIOCBONDENSLAVE:
2948 case SIOCBONDRELEASE:
2949 case SIOCBONDSETHWADDR:
2950 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002951 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002952 return -EFAULT;
2953
2954 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002955 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002956 err = dev_ioctl(net, cmd,
2957 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002958 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002959
2960 return err;
2961 case SIOCBONDSLAVEINFOQUERY:
2962 case SIOCBONDINFOQUERY:
2963 uifr = compat_alloc_user_space(sizeof(*uifr));
2964 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2965 return -EFAULT;
2966
2967 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2968 return -EFAULT;
2969
2970 datap = compat_ptr(data);
2971 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2972 return -EFAULT;
2973
Arnd Bergmann6b960182009-11-06 23:10:54 -08002974 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002975 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002976 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002977 }
Arnd Bergmann7a229382009-11-06 23:00:29 -08002978}
2979
Arnd Bergmann6b960182009-11-06 23:10:54 -08002980static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2981 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002982{
2983 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002984 char tmp_buf[IFNAMSIZ];
2985 void __user *data64;
2986 u32 data32;
2987
2988 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2989 IFNAMSIZ))
2990 return -EFAULT;
2991 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2992 return -EFAULT;
2993 data64 = compat_ptr(data32);
2994
2995 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2996
2997 /* Don't check these user accesses, just let that get trapped
2998 * in the ioctl handler instead.
2999 */
3000 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3001 IFNAMSIZ))
3002 return -EFAULT;
3003 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3004 return -EFAULT;
3005
Arnd Bergmann6b960182009-11-06 23:10:54 -08003006 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003007}
3008
Arnd Bergmann6b960182009-11-06 23:10:54 -08003009static int dev_ifsioc(struct net *net, struct socket *sock,
3010 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003011{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003012 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003013 int err;
3014
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003015 uifr = compat_alloc_user_space(sizeof(*uifr));
3016 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3017 return -EFAULT;
3018
3019 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3020
Arnd Bergmann7a229382009-11-06 23:00:29 -08003021 if (!err) {
3022 switch (cmd) {
3023 case SIOCGIFFLAGS:
3024 case SIOCGIFMETRIC:
3025 case SIOCGIFMTU:
3026 case SIOCGIFMEM:
3027 case SIOCGIFHWADDR:
3028 case SIOCGIFINDEX:
3029 case SIOCGIFADDR:
3030 case SIOCGIFBRDADDR:
3031 case SIOCGIFDSTADDR:
3032 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003033 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003034 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003035 case SIOCGMIIPHY:
3036 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003037 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003038 err = -EFAULT;
3039 break;
3040 }
3041 }
3042 return err;
3043}
3044
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003045static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3046 struct compat_ifreq __user *uifr32)
3047{
3048 struct ifreq ifr;
3049 struct compat_ifmap __user *uifmap32;
3050 mm_segment_t old_fs;
3051 int err;
3052
3053 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3054 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3055 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3056 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3057 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3058 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3059 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3060 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3061 if (err)
3062 return -EFAULT;
3063
3064 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003065 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003066 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003067 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003068
3069 if (cmd == SIOCGIFMAP && !err) {
3070 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3071 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3072 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3073 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3074 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3075 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3076 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3077 if (err)
3078 err = -EFAULT;
3079 }
3080 return err;
3081}
3082
3083static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3084{
3085 void __user *uptr;
3086 compat_uptr_t uptr32;
3087 struct ifreq __user *uifr;
3088
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003089 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003090 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3091 return -EFAULT;
3092
3093 if (get_user(uptr32, &uifr32->ifr_data))
3094 return -EFAULT;
3095
3096 uptr = compat_ptr(uptr32);
3097
3098 if (put_user(uptr, &uifr->ifr_data))
3099 return -EFAULT;
3100
3101 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3102}
3103
Arnd Bergmann7a229382009-11-06 23:00:29 -08003104struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003105 u32 rt_pad1;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003106 struct sockaddr rt_dst; /* target address */
3107 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3108 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003109 unsigned short rt_flags;
3110 short rt_pad2;
3111 u32 rt_pad3;
3112 unsigned char rt_tos;
3113 unsigned char rt_class;
3114 short rt_pad4;
3115 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003116 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003117 u32 rt_mtu; /* per route MTU/Window */
3118 u32 rt_window; /* Window clamping */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003119 unsigned short rt_irtt; /* Initial RTT */
3120};
3121
3122struct in6_rtmsg32 {
3123 struct in6_addr rtmsg_dst;
3124 struct in6_addr rtmsg_src;
3125 struct in6_addr rtmsg_gateway;
3126 u32 rtmsg_type;
3127 u16 rtmsg_dst_len;
3128 u16 rtmsg_src_len;
3129 u32 rtmsg_metric;
3130 u32 rtmsg_info;
3131 u32 rtmsg_flags;
3132 s32 rtmsg_ifindex;
3133};
3134
Arnd Bergmann6b960182009-11-06 23:10:54 -08003135static int routing_ioctl(struct net *net, struct socket *sock,
3136 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003137{
3138 int ret;
3139 void *r = NULL;
3140 struct in6_rtmsg r6;
3141 struct rtentry r4;
3142 char devname[16];
3143 u32 rtdev;
3144 mm_segment_t old_fs = get_fs();
3145
Arnd Bergmann6b960182009-11-06 23:10:54 -08003146 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3147 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003148 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003149 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003150 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3151 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3152 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3153 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3154 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3155 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3156 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003157
3158 r = (void *) &r6;
3159 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003160 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003161 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003162 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003163 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3164 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3165 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3166 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3167 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3168 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003169 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003170 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003171 r4.rt_dev = (char __user __force *)devname;
3172 devname[15] = 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003173 } else
3174 r4.rt_dev = NULL;
3175
3176 r = (void *) &r4;
3177 }
3178
3179 if (ret) {
3180 ret = -EFAULT;
3181 goto out;
3182 }
3183
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003184 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003185 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003186 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003187
3188out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003189 return ret;
3190}
3191
3192/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3193 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003194 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003195 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003196static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003197{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003198 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003199
Arnd Bergmann6b960182009-11-06 23:10:54 -08003200 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003201 return -EFAULT;
3202 if (tmp == BRCTL_GET_VERSION)
3203 return BRCTL_VERSION + 1;
3204 return -EINVAL;
3205}
3206
Arnd Bergmann6b960182009-11-06 23:10:54 -08003207static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3208 unsigned int cmd, unsigned long arg)
3209{
3210 void __user *argp = compat_ptr(arg);
3211 struct sock *sk = sock->sk;
3212 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003213
Arnd Bergmann6b960182009-11-06 23:10:54 -08003214 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3215 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003216
Arnd Bergmann6b960182009-11-06 23:10:54 -08003217 switch (cmd) {
3218 case SIOCSIFBR:
3219 case SIOCGIFBR:
3220 return old_bridge_ioctl(argp);
3221 case SIOCGIFNAME:
3222 return dev_ifname32(net, argp);
3223 case SIOCGIFCONF:
3224 return dev_ifconf(net, argp);
3225 case SIOCETHTOOL:
3226 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003227 case SIOCWANDEV:
3228 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003229 case SIOCGIFMAP:
3230 case SIOCSIFMAP:
3231 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003232 case SIOCBONDENSLAVE:
3233 case SIOCBONDRELEASE:
3234 case SIOCBONDSETHWADDR:
3235 case SIOCBONDSLAVEINFOQUERY:
3236 case SIOCBONDINFOQUERY:
3237 case SIOCBONDCHANGEACTIVE:
3238 return bond_ioctl(net, cmd, argp);
3239 case SIOCADDRT:
3240 case SIOCDELRT:
3241 return routing_ioctl(net, sock, cmd, argp);
3242 case SIOCGSTAMP:
3243 return do_siocgstamp(net, sock, cmd, argp);
3244 case SIOCGSTAMPNS:
3245 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003246 case SIOCSHWTSTAMP:
3247 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003248
Arnd Bergmann6b960182009-11-06 23:10:54 -08003249 case FIOSETOWN:
3250 case SIOCSPGRP:
3251 case FIOGETOWN:
3252 case SIOCGPGRP:
3253 case SIOCBRADDBR:
3254 case SIOCBRDELBR:
3255 case SIOCGIFVLAN:
3256 case SIOCSIFVLAN:
3257 case SIOCADDDLCI:
3258 case SIOCDELDLCI:
3259 return sock_ioctl(file, cmd, arg);
3260
3261 case SIOCGIFFLAGS:
3262 case SIOCSIFFLAGS:
3263 case SIOCGIFMETRIC:
3264 case SIOCSIFMETRIC:
3265 case SIOCGIFMTU:
3266 case SIOCSIFMTU:
3267 case SIOCGIFMEM:
3268 case SIOCSIFMEM:
3269 case SIOCGIFHWADDR:
3270 case SIOCSIFHWADDR:
3271 case SIOCADDMULTI:
3272 case SIOCDELMULTI:
3273 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003274 case SIOCGIFADDR:
3275 case SIOCSIFADDR:
3276 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003277 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003278 case SIOCGIFBRDADDR:
3279 case SIOCSIFBRDADDR:
3280 case SIOCGIFDSTADDR:
3281 case SIOCSIFDSTADDR:
3282 case SIOCGIFNETMASK:
3283 case SIOCSIFNETMASK:
3284 case SIOCSIFPFLAGS:
3285 case SIOCGIFPFLAGS:
3286 case SIOCGIFTXQLEN:
3287 case SIOCSIFTXQLEN:
3288 case SIOCBRADDIF:
3289 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003290 case SIOCSIFNAME:
3291 case SIOCGMIIPHY:
3292 case SIOCGMIIREG:
3293 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003294 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003295
Arnd Bergmann6b960182009-11-06 23:10:54 -08003296 case SIOCSARP:
3297 case SIOCGARP:
3298 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003299 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003300 return sock_do_ioctl(net, sock, cmd, arg);
3301 }
3302
Arnd Bergmann6b960182009-11-06 23:10:54 -08003303 return -ENOIOCTLCMD;
3304}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003305
Eric Dumazet95c96172012-04-15 05:58:06 +00003306static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003307 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003308{
3309 struct socket *sock = file->private_data;
3310 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003311 struct sock *sk;
3312 struct net *net;
3313
3314 sk = sock->sk;
3315 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003316
3317 if (sock->ops->compat_ioctl)
3318 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3319
David S. Miller87de87d2008-06-03 09:14:03 -07003320 if (ret == -ENOIOCTLCMD &&
3321 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3322 ret = compat_wext_handle_ioctl(net, cmd, arg);
3323
Arnd Bergmann6b960182009-11-06 23:10:54 -08003324 if (ret == -ENOIOCTLCMD)
3325 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3326
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003327 return ret;
3328}
3329#endif
3330
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003331int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3332{
3333 return sock->ops->bind(sock, addr, addrlen);
3334}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003335EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003336
3337int kernel_listen(struct socket *sock, int backlog)
3338{
3339 return sock->ops->listen(sock, backlog);
3340}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003341EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003342
3343int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3344{
3345 struct sock *sk = sock->sk;
3346 int err;
3347
3348 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3349 newsock);
3350 if (err < 0)
3351 goto done;
3352
3353 err = sock->ops->accept(sock, *newsock, flags);
3354 if (err < 0) {
3355 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003356 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003357 goto done;
3358 }
3359
3360 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003361 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003362
3363done:
3364 return err;
3365}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003366EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003367
3368int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003369 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003370{
3371 return sock->ops->connect(sock, addr, addrlen, flags);
3372}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003373EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003374
3375int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3376 int *addrlen)
3377{
3378 return sock->ops->getname(sock, addr, addrlen, 0);
3379}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003380EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003381
3382int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3383 int *addrlen)
3384{
3385 return sock->ops->getname(sock, addr, addrlen, 1);
3386}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003387EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003388
3389int kernel_getsockopt(struct socket *sock, int level, int optname,
3390 char *optval, int *optlen)
3391{
3392 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003393 char __user *uoptval;
3394 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003395 int err;
3396
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003397 uoptval = (char __user __force *) optval;
3398 uoptlen = (int __user __force *) optlen;
3399
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003400 set_fs(KERNEL_DS);
3401 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003402 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003403 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003404 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3405 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003406 set_fs(oldfs);
3407 return err;
3408}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003409EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003410
3411int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003412 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003413{
3414 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003415 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003416 int err;
3417
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003418 uoptval = (char __user __force *) optval;
3419
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003420 set_fs(KERNEL_DS);
3421 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003422 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003423 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003424 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003425 optlen);
3426 set_fs(oldfs);
3427 return err;
3428}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003429EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003430
3431int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3432 size_t size, int flags)
3433{
3434 if (sock->ops->sendpage)
3435 return sock->ops->sendpage(sock, page, offset, size, flags);
3436
3437 return sock_no_sendpage(sock, page, offset, size, flags);
3438}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003439EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003440
3441int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3442{
3443 mm_segment_t oldfs = get_fs();
3444 int err;
3445
3446 set_fs(KERNEL_DS);
3447 err = sock->ops->ioctl(sock, cmd, arg);
3448 set_fs(oldfs);
3449
3450 return err;
3451}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003452EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003453
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003454int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3455{
3456 return sock->ops->shutdown(sock, how);
3457}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003458EXPORT_SYMBOL(kernel_sock_shutdown);