blob: 717a5f1c8792baca7fd6079ca075851e1849fcdd [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
Jesper Juhl02c30a82005-05-05 16:16:16 -07007 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
Stephen Hemminger89bddce2006-09-01 00:19:31 -070045 * Tigran Aivazian : Made listen(2) backlog sanity checks
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -070056 * paradigm.
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070066#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080071#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/wanrouter.h>
73#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030074#include <linux/if_frad.h>
75#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070088#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070089#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090090#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070096#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070097#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
Arnd Bergmann6b960182009-11-06 23:10:54 -0800102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700121 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800122#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700127 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800128 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
Arjan van de Venda7071d2007-02-12 00:55:36 -0800135static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 .owner = THIS_MODULE,
137 .llseek = no_llseek,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
140 .poll = sock_poll,
141 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800142#ifdef CONFIG_COMPAT
143 .compat_ioctl = compat_sock_ioctl,
144#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 .mmap = sock_mmap,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800151 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158static DEFINE_SPINLOCK(net_family_lock);
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -0700159static const struct net_proto_family *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161/*
162 * Statistics counters of the socket lists
163 */
164
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700165static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
167/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700187 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700188 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700190 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100192 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193}
194
195/**
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
201 *
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
206 * accessible.
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
210 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700211
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700212int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700213 int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 int err;
216 int len;
217
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218 err = get_user(len, ulen);
219 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700221 if (len > klen)
222 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700223 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return -EFAULT;
230 }
231 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235 return __put_user(klen, ulen);
236}
237
Christoph Lametere18b8902006-12-06 20:33:20 -0800238static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700243
Christoph Lametere94b1762006-12-06 20:33:17 -0800244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 if (!ei)
246 return NULL;
Eric Dumazet43815482010-04-29 11:01:49 +0000247 ei->socket.wq = kmalloc(sizeof(struct socket_wq), GFP_KERNEL);
248 if (!ei->socket.wq) {
249 kmem_cache_free(sock_inode_cachep, ei);
250 return NULL;
251 }
252 init_waitqueue_head(&ei->socket.wq->wait);
253 ei->socket.wq->fasync_list = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700254
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 ei->socket.state = SS_UNCONNECTED;
256 ei->socket.flags = 0;
257 ei->socket.ops = NULL;
258 ei->socket.sk = NULL;
259 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
261 return &ei->vfs_inode;
262}
263
Eric Dumazet43815482010-04-29 11:01:49 +0000264
265static void wq_free_rcu(struct rcu_head *head)
266{
267 struct socket_wq *wq = container_of(head, struct socket_wq, rcu);
268
269 kfree(wq);
270}
271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272static void sock_destroy_inode(struct inode *inode)
273{
Eric Dumazet43815482010-04-29 11:01:49 +0000274 struct socket_alloc *ei;
275
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
277 call_rcu(&ei->socket.wq->rcu, wq_free_rcu);
278 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279}
280
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700281static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700283 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284
Christoph Lametera35afb82007-05-16 22:10:57 -0700285 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700287
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288static int init_inodecache(void)
289{
290 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700291 sizeof(struct socket_alloc),
292 0,
293 (SLAB_HWCACHE_ALIGN |
294 SLAB_RECLAIM_ACCOUNT |
295 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900296 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 if (sock_inode_cachep == NULL)
298 return -ENOMEM;
299 return 0;
300}
301
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700302static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306};
307
David Howells454e2392006-06-23 02:02:57 -0700308static int sockfs_get_sb(struct file_system_type *fs_type,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700309 int flags, const char *dev_name, void *data,
310 struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311{
David Howells454e2392006-06-23 02:02:57 -0700312 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
313 mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314}
315
Eric Dumazetba899662005-08-26 12:05:31 -0700316static struct vfsmount *sock_mnt __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
318static struct file_system_type sock_fs_type = {
319 .name = "sockfs",
320 .get_sb = sockfs_get_sb,
321 .kill_sb = kill_anon_super,
322};
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700323
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700324/*
325 * sockfs_dname() is called from d_path().
326 */
327static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
328{
329 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
330 dentry->d_inode->i_ino);
331}
332
Al Viro3ba13d12009-02-20 06:02:22 +0000333static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700334 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335};
336
337/*
338 * Obtains the first available file descriptor and sets it up for use.
339 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
347 *
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
352 */
353
Al Viro7cbe66b2009-08-05 19:59:08 +0400354static int sock_alloc_file(struct socket *sock, struct file **f, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355{
Al Viro7cbe66b2009-08-05 19:59:08 +0400356 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400357 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400358 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 int fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800360
Ulrich Dreppera677a032008-07-23 21:29:17 -0700361 fd = get_unused_fd_flags(flags);
Al Viro7cbe66b2009-08-05 19:59:08 +0400362 if (unlikely(fd < 0))
363 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800364
Al Viro2c48b9c2009-08-09 00:52:35 +0400365 path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
366 if (unlikely(!path.dentry)) {
Al Viro7cbe66b2009-08-05 19:59:08 +0400367 put_unused_fd(fd);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800368 return -ENOMEM;
Al Viro7cbe66b2009-08-05 19:59:08 +0400369 }
Al Viro2c48b9c2009-08-09 00:52:35 +0400370 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
Al Viro2c48b9c2009-08-09 00:52:35 +0400372 path.dentry->d_op = &sockfs_dentry_operations;
Al Viro2c48b9c2009-08-09 00:52:35 +0400373 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400374 SOCK_INODE(sock)->i_fop = &socket_file_ops;
375
Al Viro2c48b9c2009-08-09 00:52:35 +0400376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400377 &socket_file_ops);
378 if (unlikely(!file)) {
379 /* drop dentry, keep inode */
380 atomic_inc(&path.dentry->d_inode->i_count);
Al Viro2c48b9c2009-08-09 00:52:35 +0400381 path_put(&path);
Al Virocc3808f2009-08-06 09:43:59 +0400382 put_unused_fd(fd);
383 return -ENFILE;
384 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800385
386 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700387 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800388 file->f_pos = 0;
389 file->private_data = sock;
390
Al Viro7cbe66b2009-08-05 19:59:08 +0400391 *f = file;
392 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393}
394
Ulrich Dreppera677a032008-07-23 21:29:17 -0700395int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800396{
397 struct file *newfile;
Al Viro7cbe66b2009-08-05 19:59:08 +0400398 int fd = sock_alloc_file(sock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800399
Al Viro7cbe66b2009-08-05 19:59:08 +0400400 if (likely(fd >= 0))
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401 fd_install(fd, newfile);
Al Viro7cbe66b2009-08-05 19:59:08 +0400402
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 return fd;
404}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700405EXPORT_SYMBOL(sock_map_fd);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800407static struct socket *sock_from_file(struct file *file, int *err)
408{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800409 if (file->f_op == &socket_file_ops)
410 return file->private_data; /* set in sock_map_fd */
411
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800412 *err = -ENOTSOCK;
413 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800414}
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700417 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 * @fd: file handle
419 * @err: pointer to an error code return
420 *
421 * The file handle passed in is locked and the socket it is bound
422 * too is returned. If an error occurs the err pointer is overwritten
423 * with a negative errno code and NULL is returned. The function checks
424 * for both invalid handles and passing a handle which is not a socket.
425 *
426 * On a success the socket object pointer is returned.
427 */
428
429struct socket *sockfd_lookup(int fd, int *err)
430{
431 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 struct socket *sock;
433
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700434 file = fget(fd);
435 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 *err = -EBADF;
437 return NULL;
438 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700439
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800440 sock = sock_from_file(file, err);
441 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 return sock;
444}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700445EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800447static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
448{
449 struct file *file;
450 struct socket *sock;
451
Hua Zhong36725582006-04-19 15:25:02 -0700452 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800453 file = fget_light(fd, fput_needed);
454 if (file) {
455 sock = sock_from_file(file, err);
456 if (sock)
457 return sock;
458 fput_light(file, *fput_needed);
459 }
460 return NULL;
461}
462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463/**
464 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700465 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 * Allocate a new inode and socket object. The two are bound together
467 * and initialised. The socket is then returned. If we are out of inodes
468 * NULL is returned.
469 */
470
471static struct socket *sock_alloc(void)
472{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700473 struct inode *inode;
474 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475
476 inode = new_inode(sock_mnt->mnt_sb);
477 if (!inode)
478 return NULL;
479
480 sock = SOCKET_I(inode);
481
Eric Dumazet29a020d2009-09-15 02:39:20 -0700482 kmemcheck_annotate_bitfield(sock, type);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700483 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100484 inode->i_uid = current_fsuid();
485 inode->i_gid = current_fsgid();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Eric Dumazet4e694892009-04-04 16:41:09 -0700487 percpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 return sock;
489}
490
491/*
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
495 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
498{
499 return -ENXIO;
500}
501
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800502const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 .owner = THIS_MODULE,
504 .open = sock_no_open,
505};
506
507/**
508 * sock_release - close a socket
509 * @sock: socket to close
510 *
511 * The socket is released from the protocol stack if it has a release
512 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700513 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516void sock_release(struct socket *sock)
517{
518 if (sock->ops) {
519 struct module *owner = sock->ops->owner;
520
521 sock->ops->release(sock);
522 sock->ops = NULL;
523 module_put(owner);
524 }
525
Eric Dumazet43815482010-04-29 11:01:49 +0000526 if (sock->wq->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 printk(KERN_ERR "sock_release: fasync list not empty!\n");
528
Eric Dumazet4e694892009-04-04 16:41:09 -0700529 percpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 if (!sock->file) {
531 iput(SOCK_INODE(sock));
532 return;
533 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700534 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700536EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000538int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000539{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000540 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000541 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000542 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000543 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000544 *tx_flags |= SKBTX_SW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000545 return 0;
546}
547EXPORT_SYMBOL(sock_tx_timestamp);
548
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700549static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 struct msghdr *msg, size_t size)
551{
552 struct sock_iocb *si = kiocb_to_siocb(iocb);
553 int err;
554
Herbert Xuf8451722010-05-24 00:12:34 -0700555 sock_update_classid(sock->sk);
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 si->sock = sock;
558 si->scm = NULL;
559 si->msg = msg;
560 si->size = size;
561
562 err = security_socket_sendmsg(sock, msg, size);
563 if (err)
564 return err;
565
566 return sock->ops->sendmsg(iocb, sock, msg, size);
567}
568
569int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
570{
571 struct kiocb iocb;
572 struct sock_iocb siocb;
573 int ret;
574
575 init_sync_kiocb(&iocb, NULL);
576 iocb.private = &siocb;
577 ret = __sock_sendmsg(&iocb, sock, msg, size);
578 if (-EIOCBQUEUED == ret)
579 ret = wait_on_sync_kiocb(&iocb);
580 return ret;
581}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700582EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
584int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
585 struct kvec *vec, size_t num, size_t size)
586{
587 mm_segment_t oldfs = get_fs();
588 int result;
589
590 set_fs(KERNEL_DS);
591 /*
592 * the following is safe, since for compiler definitions of kvec and
593 * iovec are identical, yielding the same in-core layout and alignment
594 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700595 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 msg->msg_iovlen = num;
597 result = sock_sendmsg(sock, msg, size);
598 set_fs(oldfs);
599 return result;
600}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700601EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
Patrick Ohly20d49472009-02-12 05:03:38 +0000603static int ktime2ts(ktime_t kt, struct timespec *ts)
604{
605 if (kt.tv64) {
606 *ts = ktime_to_timespec(kt);
607 return 1;
608 } else {
609 return 0;
610 }
611}
612
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700613/*
614 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
615 */
616void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
617 struct sk_buff *skb)
618{
Patrick Ohly20d49472009-02-12 05:03:38 +0000619 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
620 struct timespec ts[3];
621 int empty = 1;
622 struct skb_shared_hwtstamps *shhwtstamps =
623 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700624
Patrick Ohly20d49472009-02-12 05:03:38 +0000625 /* Race occurred between timestamp enabling and packet
626 receiving. Fill in the current time for now. */
627 if (need_software_tstamp && skb->tstamp.tv64 == 0)
628 __net_timestamp(skb);
629
630 if (need_software_tstamp) {
631 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
632 struct timeval tv;
633 skb_get_timestamp(skb, &tv);
634 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
635 sizeof(tv), &tv);
636 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000637 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000638 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000639 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000640 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700641 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000642
643
644 memset(ts, 0, sizeof(ts));
645 if (skb->tstamp.tv64 &&
646 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
647 skb_get_timestampns(skb, ts + 0);
648 empty = 0;
649 }
650 if (shhwtstamps) {
651 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
652 ktime2ts(shhwtstamps->syststamp, ts + 1))
653 empty = 0;
654 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
655 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
656 empty = 0;
657 }
658 if (!empty)
659 put_cmsg(msg, SOL_SOCKET,
660 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700661}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300662EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
663
Neil Horman3b885782009-10-12 13:26:31 -0700664inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
665{
666 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
667 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
668 sizeof(__u32), &skb->dropcount);
669}
670
Eric Dumazet767dd032010-04-28 19:14:43 +0000671void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700672 struct sk_buff *skb)
673{
674 sock_recv_timestamp(msg, sk, skb);
675 sock_recv_drops(msg, sk, skb);
676}
Eric Dumazet767dd032010-04-28 19:14:43 +0000677EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700678
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700679static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
680 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 struct sock_iocb *si = kiocb_to_siocb(iocb);
683
Herbert Xuf8451722010-05-24 00:12:34 -0700684 sock_update_classid(sock->sk);
685
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 si->sock = sock;
687 si->scm = NULL;
688 si->msg = msg;
689 si->size = size;
690 si->flags = flags;
691
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
693}
694
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700695static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
696 struct msghdr *msg, size_t size, int flags)
697{
698 int err = security_socket_recvmsg(sock, msg, size, flags);
699
700 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
701}
702
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700703int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 size_t size, int flags)
705{
706 struct kiocb iocb;
707 struct sock_iocb siocb;
708 int ret;
709
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700710 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 iocb.private = &siocb;
712 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
713 if (-EIOCBQUEUED == ret)
714 ret = wait_on_sync_kiocb(&iocb);
715 return ret;
716}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700717EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700719static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
720 size_t size, int flags)
721{
722 struct kiocb iocb;
723 struct sock_iocb siocb;
724 int ret;
725
726 init_sync_kiocb(&iocb, NULL);
727 iocb.private = &siocb;
728 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
729 if (-EIOCBQUEUED == ret)
730 ret = wait_on_sync_kiocb(&iocb);
731 return ret;
732}
733
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700734int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
735 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736{
737 mm_segment_t oldfs = get_fs();
738 int result;
739
740 set_fs(KERNEL_DS);
741 /*
742 * the following is safe, since for compiler definitions of kvec and
743 * iovec are identical, yielding the same in-core layout and alignment
744 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700745 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 result = sock_recvmsg(sock, msg, size, flags);
747 set_fs(oldfs);
748 return result;
749}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700750EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751
752static void sock_aio_dtor(struct kiocb *iocb)
753{
754 kfree(iocb->private);
755}
756
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300757static ssize_t sock_sendpage(struct file *file, struct page *page,
758 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759{
760 struct socket *sock;
761 int flags;
762
Eric Dumazetb69aee02005-09-06 14:42:45 -0700763 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
765 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
766 if (more)
767 flags |= MSG_MORE;
768
Linus Torvaldse6949582009-08-13 08:28:36 -0700769 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770}
771
Jens Axboe9c55e012007-11-06 23:30:13 -0800772static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700773 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800774 unsigned int flags)
775{
776 struct socket *sock = file->private_data;
777
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800778 if (unlikely(!sock->ops->splice_read))
779 return -EINVAL;
780
Herbert Xuf8451722010-05-24 00:12:34 -0700781 sock_update_classid(sock->sk);
782
Jens Axboe9c55e012007-11-06 23:30:13 -0800783 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
784}
785
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800786static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700787 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800788{
789 if (!is_sync_kiocb(iocb)) {
790 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
791 if (!siocb)
792 return NULL;
793 iocb->ki_dtor = sock_aio_dtor;
794 }
795
796 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800797 iocb->private = siocb;
798 return siocb;
799}
800
801static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700802 struct file *file, const struct iovec *iov,
803 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800804{
805 struct socket *sock = file->private_data;
806 size_t size = 0;
807 int i;
808
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700809 for (i = 0; i < nr_segs; i++)
810 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800811
812 msg->msg_name = NULL;
813 msg->msg_namelen = 0;
814 msg->msg_control = NULL;
815 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700816 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800817 msg->msg_iovlen = nr_segs;
818 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
819
820 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
821}
822
Badari Pulavarty027445c2006-09-30 23:28:46 -0700823static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
824 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800825{
826 struct sock_iocb siocb, *x;
827
828 if (pos != 0)
829 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700830
831 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800832 return 0;
833
Badari Pulavarty027445c2006-09-30 23:28:46 -0700834
835 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800836 if (!x)
837 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700838 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800839}
840
841static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700842 struct file *file, const struct iovec *iov,
843 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800844{
845 struct socket *sock = file->private_data;
846 size_t size = 0;
847 int i;
848
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700849 for (i = 0; i < nr_segs; i++)
850 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800851
852 msg->msg_name = NULL;
853 msg->msg_namelen = 0;
854 msg->msg_control = NULL;
855 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700856 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800857 msg->msg_iovlen = nr_segs;
858 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
859 if (sock->type == SOCK_SEQPACKET)
860 msg->msg_flags |= MSG_EOR;
861
862 return __sock_sendmsg(iocb, sock, msg, size);
863}
864
Badari Pulavarty027445c2006-09-30 23:28:46 -0700865static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
866 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800868 struct sock_iocb siocb, *x;
869
870 if (pos != 0)
871 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700872
Badari Pulavarty027445c2006-09-30 23:28:46 -0700873 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800874 if (!x)
875 return -ENOMEM;
876
Badari Pulavarty027445c2006-09-30 23:28:46 -0700877 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878}
879
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880/*
881 * Atomic setting of ioctl hooks to avoid race
882 * with module unload.
883 */
884
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800885static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700886static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887
Eric W. Biederman881d9662007-09-17 11:56:21 -0700888void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800890 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800892 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893}
894EXPORT_SYMBOL(brioctl_set);
895
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800896static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700897static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
Eric W. Biederman881d9662007-09-17 11:56:21 -0700899void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800901 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800903 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904}
905EXPORT_SYMBOL(vlan_ioctl_set);
906
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800907static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700908static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700910void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800912 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800914 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916EXPORT_SYMBOL(dlci_ioctl_set);
917
Arnd Bergmann6b960182009-11-06 23:10:54 -0800918static long sock_do_ioctl(struct net *net, struct socket *sock,
919 unsigned int cmd, unsigned long arg)
920{
921 int err;
922 void __user *argp = (void __user *)arg;
923
924 err = sock->ops->ioctl(sock, cmd, arg);
925
926 /*
927 * If this ioctl is unknown try to hand it down
928 * to the NIC driver.
929 */
930 if (err == -ENOIOCTLCMD)
931 err = dev_ioctl(net, cmd, argp);
932
933 return err;
934}
935
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936/*
937 * With an ioctl, arg may well be a user mode pointer, but we don't know
938 * what to do with it - that's up to the protocol still.
939 */
940
941static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
942{
943 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700944 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 void __user *argp = (void __user *)arg;
946 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700947 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948
Eric Dumazetb69aee02005-09-06 14:42:45 -0700949 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700950 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900951 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700953 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200955#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700957 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200959#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700960 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 case FIOSETOWN:
962 case SIOCSPGRP:
963 err = -EFAULT;
964 if (get_user(pid, (int __user *)argp))
965 break;
966 err = f_setown(sock->file, pid, 1);
967 break;
968 case FIOGETOWN:
969 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700970 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700971 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 break;
973 case SIOCGIFBR:
974 case SIOCSIFBR:
975 case SIOCBRADDBR:
976 case SIOCBRDELBR:
977 err = -ENOPKG;
978 if (!br_ioctl_hook)
979 request_module("bridge");
980
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800981 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700982 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700983 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800984 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 break;
986 case SIOCGIFVLAN:
987 case SIOCSIFVLAN:
988 err = -ENOPKG;
989 if (!vlan_ioctl_hook)
990 request_module("8021q");
991
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800992 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700994 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800995 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 case SIOCADDDLCI:
998 case SIOCDELDLCI:
999 err = -ENOPKG;
1000 if (!dlci_ioctl_hook)
1001 request_module("dlci");
1002
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001003 mutex_lock(&dlci_ioctl_mutex);
1004 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001006 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 break;
1008 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001009 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001011 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 return err;
1013}
1014
1015int sock_create_lite(int family, int type, int protocol, struct socket **res)
1016{
1017 int err;
1018 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 err = security_socket_create(family, type, protocol, 1);
1021 if (err)
1022 goto out;
1023
1024 sock = sock_alloc();
1025 if (!sock) {
1026 err = -ENOMEM;
1027 goto out;
1028 }
1029
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001031 err = security_socket_post_create(sock, family, type, protocol, 1);
1032 if (err)
1033 goto out_release;
1034
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035out:
1036 *res = sock;
1037 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001038out_release:
1039 sock_release(sock);
1040 sock = NULL;
1041 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001043EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
1045/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001046static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047{
1048 struct socket *sock;
1049
1050 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001051 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001053 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 return sock->ops->poll(file, sock, wait);
1055}
1056
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001057static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001059 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
1061 return sock->ops->mmap(file, sock, vma);
1062}
1063
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001064static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065{
1066 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001067 * It was possible the inode is NULL we were
1068 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 */
1070
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001071 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 printk(KERN_DEBUG "sock_close: NULL inode\n");
1073 return 0;
1074 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 sock_release(SOCKET_I(inode));
1076 return 0;
1077}
1078
1079/*
1080 * Update the socket async list
1081 *
1082 * Fasync_list locking strategy.
1083 *
1084 * 1. fasync_list is modified only under process context socket lock
1085 * i.e. under semaphore.
1086 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001087 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 */
1089
1090static int sock_fasync(int fd, struct file *filp, int on)
1091{
Eric Dumazet989a2972010-04-14 09:55:35 +00001092 struct socket *sock = filp->private_data;
1093 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
Eric Dumazet989a2972010-04-14 09:55:35 +00001095 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097
1098 lock_sock(sk);
1099
Eric Dumazet43815482010-04-29 11:01:49 +00001100 fasync_helper(fd, filp, on, &sock->wq->fasync_list);
Eric Dumazet989a2972010-04-14 09:55:35 +00001101
Eric Dumazet43815482010-04-29 11:01:49 +00001102 if (!sock->wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001103 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001104 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001105 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106
Eric Dumazet989a2972010-04-14 09:55:35 +00001107 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 return 0;
1109}
1110
Eric Dumazet43815482010-04-29 11:01:49 +00001111/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
1113int sock_wake_async(struct socket *sock, int how, int band)
1114{
Eric Dumazet43815482010-04-29 11:01:49 +00001115 struct socket_wq *wq;
1116
1117 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001119 rcu_read_lock();
1120 wq = rcu_dereference(sock->wq);
1121 if (!wq || !wq->fasync_list) {
1122 rcu_read_unlock();
1123 return -1;
1124 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001125 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001126 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1128 break;
1129 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001130 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1132 break;
1133 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001134 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001135call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001136 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001138 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001139 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 }
Eric Dumazet43815482010-04-29 11:01:49 +00001141 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 return 0;
1143}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001144EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001146static int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001147 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148{
1149 int err;
1150 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001151 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152
1153 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001154 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 */
1156 if (family < 0 || family >= NPROTO)
1157 return -EAFNOSUPPORT;
1158 if (type < 0 || type >= SOCK_MAX)
1159 return -EINVAL;
1160
1161 /* Compatibility.
1162
1163 This uglymoron is moved from INET layer to here to avoid
1164 deadlock in module load.
1165 */
1166 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001167 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 if (!warned) {
1169 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001170 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1171 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 }
1173 family = PF_PACKET;
1174 }
1175
1176 err = security_socket_create(family, type, protocol, kern);
1177 if (err)
1178 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001179
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001180 /*
1181 * Allocate the socket and allow the family to set things up. if
1182 * the protocol is 0, the family is instructed to select an appropriate
1183 * default.
1184 */
1185 sock = sock_alloc();
1186 if (!sock) {
1187 if (net_ratelimit())
1188 printk(KERN_WARNING "socket: no more sockets\n");
1189 return -ENFILE; /* Not exactly a match, but its the
1190 closest posix thing */
1191 }
1192
1193 sock->type = type;
1194
Johannes Berg95a5afc2008-10-16 15:24:51 -07001195#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001196 /* Attempt to load a protocol module if the find failed.
1197 *
1198 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 * requested real, full-featured networking support upon configuration.
1200 * Otherwise module support will break!
1201 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001202 if (net_families[family] == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001203 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204#endif
1205
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001206 rcu_read_lock();
1207 pf = rcu_dereference(net_families[family]);
1208 err = -EAFNOSUPPORT;
1209 if (!pf)
1210 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211
1212 /*
1213 * We will call the ->create function, that possibly is in a loadable
1214 * module, so we have to bump that loadable module refcnt first.
1215 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001216 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 goto out_release;
1218
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001219 /* Now protected by module ref count */
1220 rcu_read_unlock();
1221
Eric Paris3f378b62009-11-05 22:18:14 -08001222 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001223 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001225
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 /*
1227 * Now to bump the refcnt of the [loadable] module that owns this
1228 * socket at sock_release time we decrement its refcnt.
1229 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001230 if (!try_module_get(sock->ops->owner))
1231 goto out_module_busy;
1232
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 /*
1234 * Now that we're done with the ->create function, the [loadable]
1235 * module can have its refcnt decremented
1236 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001237 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001238 err = security_socket_post_create(sock, family, type, protocol, kern);
1239 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001240 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001241 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001243 return 0;
1244
1245out_module_busy:
1246 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001248 sock->ops = NULL;
1249 module_put(pf->owner);
1250out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001252 return err;
1253
1254out_release:
1255 rcu_read_unlock();
1256 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257}
1258
1259int sock_create(int family, int type, int protocol, struct socket **res)
1260{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001261 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001263EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264
1265int sock_create_kern(int family, int type, int protocol, struct socket **res)
1266{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001267 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001269EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001271SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272{
1273 int retval;
1274 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001275 int flags;
1276
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001277 /* Check the SOCK_* constants for consistency. */
1278 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1279 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1280 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1281 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1282
Ulrich Dreppera677a032008-07-23 21:29:17 -07001283 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001284 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001285 return -EINVAL;
1286 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001288 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1289 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1290
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 retval = sock_create(family, type, protocol, &sock);
1292 if (retval < 0)
1293 goto out;
1294
Ulrich Drepper77d27202008-07-23 21:29:35 -07001295 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 if (retval < 0)
1297 goto out_release;
1298
1299out:
1300 /* It may be already another descriptor 8) Not kernel problem. */
1301 return retval;
1302
1303out_release:
1304 sock_release(sock);
1305 return retval;
1306}
1307
1308/*
1309 * Create a pair of connected sockets.
1310 */
1311
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001312SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1313 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314{
1315 struct socket *sock1, *sock2;
1316 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001317 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001318 int flags;
1319
1320 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001321 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001322 return -EINVAL;
1323 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001325 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1326 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1327
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 /*
1329 * Obtain the first socket and check if the underlying protocol
1330 * supports the socketpair call.
1331 */
1332
1333 err = sock_create(family, type, protocol, &sock1);
1334 if (err < 0)
1335 goto out;
1336
1337 err = sock_create(family, type, protocol, &sock2);
1338 if (err < 0)
1339 goto out_release_1;
1340
1341 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001342 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 goto out_release_both;
1344
Al Viro7cbe66b2009-08-05 19:59:08 +04001345 fd1 = sock_alloc_file(sock1, &newfile1, flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001346 if (unlikely(fd1 < 0)) {
1347 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001349 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350
Al Viro7cbe66b2009-08-05 19:59:08 +04001351 fd2 = sock_alloc_file(sock2, &newfile2, flags);
Al Viro198de4d2009-08-05 19:29:23 +04001352 if (unlikely(fd2 < 0)) {
1353 err = fd2;
1354 fput(newfile1);
1355 put_unused_fd(fd1);
1356 sock_release(sock2);
1357 goto out;
Al Virodb349502007-02-07 01:48:00 -05001358 }
1359
Al Viro157cf642008-12-14 04:57:47 -05001360 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001361 fd_install(fd1, newfile1);
1362 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 /* fd1 and fd2 may be already another descriptors.
1364 * Not kernel problem.
1365 */
1366
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001367 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 if (!err)
1369 err = put_user(fd2, &usockvec[1]);
1370 if (!err)
1371 return 0;
1372
1373 sys_close(fd2);
1374 sys_close(fd1);
1375 return err;
1376
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001378 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001380 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381out:
1382 return err;
1383}
1384
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385/*
1386 * Bind a name to a socket. Nothing much to do here since it's
1387 * the protocol's responsibility to handle the local address.
1388 *
1389 * We move the socket address to kernel space before we call
1390 * the protocol layer (having also checked the address is ok).
1391 */
1392
Heiko Carstens20f37032009-01-14 14:14:23 +01001393SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394{
1395 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001396 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001397 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001399 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001400 if (sock) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001401 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001402 if (err >= 0) {
1403 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001404 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001405 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001406 if (!err)
1407 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001408 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001409 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001411 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001412 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 return err;
1414}
1415
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416/*
1417 * Perform a listen. Basically, we allow the protocol to do anything
1418 * necessary for a listen, and if that works, we mark the socket as
1419 * ready for listening.
1420 */
1421
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001422SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423{
1424 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001425 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001426 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001427
1428 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1429 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001430 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001431 if ((unsigned)backlog > somaxconn)
1432 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
1434 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001435 if (!err)
1436 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001438 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 }
1440 return err;
1441}
1442
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443/*
1444 * For accept, we attempt to create a new socket, set up the link
1445 * with the client, wake up the client, then return the new
1446 * connected fd. We collect the address of the connector in kernel
1447 * space and move it to user at the very end. This is unclean because
1448 * we open the socket then return an error.
1449 *
1450 * 1003.1g adds the ability to recvmsg() to query connection pending
1451 * status to recvmsg. We need to add that support in a way thats
1452 * clean when we restucture accept also.
1453 */
1454
Heiko Carstens20f37032009-01-14 14:14:23 +01001455SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1456 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457{
1458 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001459 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001460 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001461 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462
Ulrich Drepper77d27202008-07-23 21:29:35 -07001463 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001464 return -EINVAL;
1465
1466 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1467 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1468
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001469 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 if (!sock)
1471 goto out;
1472
1473 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001474 newsock = sock_alloc();
1475 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 goto out_put;
1477
1478 newsock->type = sock->type;
1479 newsock->ops = sock->ops;
1480
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 /*
1482 * We don't need try_module_get here, as the listening socket (sock)
1483 * has the protocol module (sock->ops->owner) held.
1484 */
1485 __module_get(newsock->ops->owner);
1486
Al Viro7cbe66b2009-08-05 19:59:08 +04001487 newfd = sock_alloc_file(newsock, &newfile, flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001488 if (unlikely(newfd < 0)) {
1489 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001490 sock_release(newsock);
1491 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001492 }
1493
Frank Filza79af592005-09-27 15:23:38 -07001494 err = security_socket_accept(sock, newsock);
1495 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001496 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001497
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1499 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001500 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501
1502 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001503 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001504 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001506 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 }
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001508 err = move_addr_to_user((struct sockaddr *)&address,
1509 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001511 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 }
1513
1514 /* File flags are not inherited via accept() unlike another OSes. */
1515
David S. Miller39d8c1b2006-03-20 17:13:49 -08001516 fd_install(newfd, newfile);
1517 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001520 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521out:
1522 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001523out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001524 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001525 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 goto out_put;
1527}
1528
Heiko Carstens20f37032009-01-14 14:14:23 +01001529SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1530 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001531{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001532 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001533}
1534
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535/*
1536 * Attempt to connect to a socket with the server address. The address
1537 * is in user space so we verify it is OK and move it to kernel space.
1538 *
1539 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1540 * break bindings
1541 *
1542 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1543 * other SEQPACKET protocols that take time to connect() as it doesn't
1544 * include the -EINPROGRESS status for such sockets.
1545 */
1546
Heiko Carstens20f37032009-01-14 14:14:23 +01001547SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1548 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549{
1550 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001551 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001552 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 if (!sock)
1556 goto out;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001557 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 if (err < 0)
1559 goto out_put;
1560
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001561 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001562 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 if (err)
1564 goto out_put;
1565
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001566 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 sock->file->f_flags);
1568out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001569 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570out:
1571 return err;
1572}
1573
1574/*
1575 * Get the local address ('name') of a socket object. Move the obtained
1576 * name to user space.
1577 */
1578
Heiko Carstens20f37032009-01-14 14:14:23 +01001579SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1580 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581{
1582 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001583 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001584 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001585
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001586 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 if (!sock)
1588 goto out;
1589
1590 err = security_socket_getsockname(sock);
1591 if (err)
1592 goto out_put;
1593
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001594 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 if (err)
1596 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001597 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001600 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601out:
1602 return err;
1603}
1604
1605/*
1606 * Get the remote address ('name') of a socket object. Move the obtained
1607 * name to user space.
1608 */
1609
Heiko Carstens20f37032009-01-14 14:14:23 +01001610SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1611 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612{
1613 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001614 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001615 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1618 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 err = security_socket_getpeername(sock);
1620 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001621 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 return err;
1623 }
1624
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001625 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001626 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001627 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628 if (!err)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001629 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001630 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001631 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 }
1633 return err;
1634}
1635
1636/*
1637 * Send a datagram to a given address. We move the address into kernel
1638 * space and check the user space data area is readable before invoking
1639 * the protocol.
1640 */
1641
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001642SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1643 unsigned, flags, struct sockaddr __user *, addr,
1644 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645{
1646 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001647 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 int err;
1649 struct msghdr msg;
1650 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001651 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001652
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001653 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1654 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001655 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001656
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001657 iov.iov_base = buff;
1658 iov.iov_len = len;
1659 msg.msg_name = NULL;
1660 msg.msg_iov = &iov;
1661 msg.msg_iovlen = 1;
1662 msg.msg_control = NULL;
1663 msg.msg_controllen = 0;
1664 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001665 if (addr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001666 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 if (err < 0)
1668 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001669 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001670 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 }
1672 if (sock->file->f_flags & O_NONBLOCK)
1673 flags |= MSG_DONTWAIT;
1674 msg.msg_flags = flags;
1675 err = sock_sendmsg(sock, &msg, len);
1676
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001677out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001678 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001679out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 return err;
1681}
1682
1683/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001684 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 */
1686
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001687SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1688 unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689{
1690 return sys_sendto(fd, buff, len, flags, NULL, 0);
1691}
1692
1693/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001694 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 * sender. We verify the buffers are writable and if needed move the
1696 * sender address from kernel to user space.
1697 */
1698
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001699SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1700 unsigned, flags, struct sockaddr __user *, addr,
1701 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702{
1703 struct socket *sock;
1704 struct iovec iov;
1705 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001706 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001707 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001708 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001710 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001712 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001714 msg.msg_control = NULL;
1715 msg.msg_controllen = 0;
1716 msg.msg_iovlen = 1;
1717 msg.msg_iov = &iov;
1718 iov.iov_len = size;
1719 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001720 msg.msg_name = (struct sockaddr *)&address;
1721 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 if (sock->file->f_flags & O_NONBLOCK)
1723 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001724 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001726 if (err >= 0 && addr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001727 err2 = move_addr_to_user((struct sockaddr *)&address,
1728 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001729 if (err2 < 0)
1730 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001732
1733 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001734out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 return err;
1736}
1737
1738/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001739 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 */
1741
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001742asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1743 unsigned flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744{
1745 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1746}
1747
1748/*
1749 * Set a socket option. Because we don't know the option lengths we have
1750 * to pass the user mode parameter for the protocols to sort out.
1751 */
1752
Heiko Carstens20f37032009-01-14 14:14:23 +01001753SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1754 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001756 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 struct socket *sock;
1758
1759 if (optlen < 0)
1760 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001761
1762 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1763 if (sock != NULL) {
1764 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001765 if (err)
1766 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767
1768 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001769 err =
1770 sock_setsockopt(sock, level, optname, optval,
1771 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001773 err =
1774 sock->ops->setsockopt(sock, level, optname, optval,
1775 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001776out_put:
1777 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 }
1779 return err;
1780}
1781
1782/*
1783 * Get a socket option. Because we don't know the option lengths we have
1784 * to pass a user mode parameter for the protocols to sort out.
1785 */
1786
Heiko Carstens20f37032009-01-14 14:14:23 +01001787SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1788 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001790 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 struct socket *sock;
1792
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001793 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1794 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001795 err = security_socket_getsockopt(sock, level, optname);
1796 if (err)
1797 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798
1799 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001800 err =
1801 sock_getsockopt(sock, level, optname, optval,
1802 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001804 err =
1805 sock->ops->getsockopt(sock, level, optname, optval,
1806 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001807out_put:
1808 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 }
1810 return err;
1811}
1812
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813/*
1814 * Shutdown a socket.
1815 */
1816
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001817SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001819 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 struct socket *sock;
1821
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001822 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1823 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001825 if (!err)
1826 err = sock->ops->shutdown(sock, how);
1827 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 }
1829 return err;
1830}
1831
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001832/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 * fields which are the same type (int / unsigned) on our platforms.
1834 */
1835#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1836#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1837#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839/*
1840 * BSD sendmsg interface
1841 */
1842
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001843SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001845 struct compat_msghdr __user *msg_compat =
1846 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001848 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001850 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001851 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1852 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 unsigned char *ctl_buf = ctl;
1854 struct msghdr msg_sys;
1855 int err, ctl_len, iov_size, total_len;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001856 int fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001857
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 err = -EFAULT;
1859 if (MSG_CMSG_COMPAT & flags) {
1860 if (get_compat_msghdr(&msg_sys, msg_compat))
1861 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001862 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 return -EFAULT;
1864
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001865 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001866 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 goto out;
1868
1869 /* do not move before msg_sys is valid */
1870 err = -EMSGSIZE;
1871 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1872 goto out_put;
1873
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001874 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 err = -ENOMEM;
1876 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1877 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1878 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1879 if (!iov)
1880 goto out_put;
1881 }
1882
1883 /* This will also move the address data into kernel space */
1884 if (MSG_CMSG_COMPAT & flags) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001885 err = verify_compat_iovec(&msg_sys, iov,
1886 (struct sockaddr *)&address,
1887 VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888 } else
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001889 err = verify_iovec(&msg_sys, iov,
1890 (struct sockaddr *)&address,
1891 VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001892 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 goto out_freeiov;
1894 total_len = err;
1895
1896 err = -ENOBUFS;
1897
1898 if (msg_sys.msg_controllen > INT_MAX)
1899 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001900 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001902 err =
1903 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1904 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 if (err)
1906 goto out_freeiov;
1907 ctl_buf = msg_sys.msg_control;
Al Viro8920e8f2005-09-07 18:28:51 -07001908 ctl_len = msg_sys.msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001912 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 goto out_freeiov;
1914 }
1915 err = -EFAULT;
1916 /*
1917 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1918 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1919 * checking falls down on this.
1920 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001921 if (copy_from_user(ctl_buf,
1922 (void __user __force *)msg_sys.msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001923 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924 goto out_freectl;
1925 msg_sys.msg_control = ctl_buf;
1926 }
1927 msg_sys.msg_flags = flags;
1928
1929 if (sock->file->f_flags & O_NONBLOCK)
1930 msg_sys.msg_flags |= MSG_DONTWAIT;
1931 err = sock_sendmsg(sock, &msg_sys, total_len);
1932
1933out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001934 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1936out_freeiov:
1937 if (iov != iovstack)
1938 sock_kfree_s(sock->sk, iov, iov_size);
1939out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001940 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001941out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 return err;
1943}
1944
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001945static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
1946 struct msghdr *msg_sys, unsigned flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001948 struct compat_msghdr __user *msg_compat =
1949 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001951 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 unsigned long cmsg_ptr;
1953 int err, iov_size, total_len, len;
1954
1955 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001956 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
1958 /* user mode address pointers */
1959 struct sockaddr __user *uaddr;
1960 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001961
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001963 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001965 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001966 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 err = -EMSGSIZE;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001969 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1970 goto out;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001971
1972 /* Check whether to allocate the iovec area */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 err = -ENOMEM;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001974 iov_size = msg_sys->msg_iovlen * sizeof(struct iovec);
1975 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1977 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001978 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 }
1980
1981 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001982 * Save the user-mode address (verify_iovec will change the
1983 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001985
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001986 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 uaddr_len = COMPAT_NAMELEN(msg);
1988 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001989 err = verify_compat_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001990 (struct sockaddr *)&addr,
1991 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 } else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001993 err = verify_iovec(msg_sys, iov,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001994 (struct sockaddr *)&addr,
1995 VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 if (err < 0)
1997 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001998 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002000 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2001 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002002
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 if (sock->file->f_flags & O_NONBLOCK)
2004 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002005 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2006 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 if (err < 0)
2008 goto out_freeiov;
2009 len = err;
2010
2011 if (uaddr != NULL) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002012 err = move_addr_to_user((struct sockaddr *)&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002013 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002014 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 if (err < 0)
2016 goto out_freeiov;
2017 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002018 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002019 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 if (err)
2021 goto out_freeiov;
2022 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002023 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 &msg_compat->msg_controllen);
2025 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002026 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 &msg->msg_controllen);
2028 if (err)
2029 goto out_freeiov;
2030 err = len;
2031
2032out_freeiov:
2033 if (iov != iovstack)
2034 sock_kfree_s(sock->sk, iov, iov_size);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002035out:
2036 return err;
2037}
2038
2039/*
2040 * BSD recvmsg interface
2041 */
2042
2043SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2044 unsigned int, flags)
2045{
2046 int fput_needed, err;
2047 struct msghdr msg_sys;
2048 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2049
2050 if (!sock)
2051 goto out;
2052
2053 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2054
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002055 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056out:
2057 return err;
2058}
2059
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002060/*
2061 * Linux recvmmsg interface
2062 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002064int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2065 unsigned int flags, struct timespec *timeout)
2066{
2067 int fput_needed, err, datagrams;
2068 struct socket *sock;
2069 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002070 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002071 struct msghdr msg_sys;
2072 struct timespec end_time;
2073
2074 if (timeout &&
2075 poll_select_set_timeout(&end_time, timeout->tv_sec,
2076 timeout->tv_nsec))
2077 return -EINVAL;
2078
2079 datagrams = 0;
2080
2081 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2082 if (!sock)
2083 return err;
2084
2085 err = sock_error(sock->sk);
2086 if (err)
2087 goto out_put;
2088
2089 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002090 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002091
2092 while (datagrams < vlen) {
2093 /*
2094 * No need to ask LSM for more than the first datagram.
2095 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002096 if (MSG_CMSG_COMPAT & flags) {
2097 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2098 &msg_sys, flags, datagrams);
2099 if (err < 0)
2100 break;
2101 err = __put_user(err, &compat_entry->msg_len);
2102 ++compat_entry;
2103 } else {
2104 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
2105 &msg_sys, flags, datagrams);
2106 if (err < 0)
2107 break;
2108 err = put_user(err, &entry->msg_len);
2109 ++entry;
2110 }
2111
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002112 if (err)
2113 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002114 ++datagrams;
2115
Brandon L Black71c5c152010-03-26 16:18:03 +00002116 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2117 if (flags & MSG_WAITFORONE)
2118 flags |= MSG_DONTWAIT;
2119
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002120 if (timeout) {
2121 ktime_get_ts(timeout);
2122 *timeout = timespec_sub(end_time, *timeout);
2123 if (timeout->tv_sec < 0) {
2124 timeout->tv_sec = timeout->tv_nsec = 0;
2125 break;
2126 }
2127
2128 /* Timeout, return less than vlen datagrams */
2129 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2130 break;
2131 }
2132
2133 /* Out of band data, return right away */
2134 if (msg_sys.msg_flags & MSG_OOB)
2135 break;
2136 }
2137
2138out_put:
2139 fput_light(sock->file, fput_needed);
2140
2141 if (err == 0)
2142 return datagrams;
2143
2144 if (datagrams != 0) {
2145 /*
2146 * We may return less entries than requested (vlen) if the
2147 * sock is non block and there aren't enough datagrams...
2148 */
2149 if (err != -EAGAIN) {
2150 /*
2151 * ... or if recvmsg returns an error after we
2152 * received some datagrams, where we record the
2153 * error to return on the next call or if the
2154 * app asks about it using getsockopt(SO_ERROR).
2155 */
2156 sock->sk->sk_err = -err;
2157 }
2158
2159 return datagrams;
2160 }
2161
2162 return err;
2163}
2164
2165SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2166 unsigned int, vlen, unsigned int, flags,
2167 struct timespec __user *, timeout)
2168{
2169 int datagrams;
2170 struct timespec timeout_sys;
2171
2172 if (!timeout)
2173 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2174
2175 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2176 return -EFAULT;
2177
2178 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2179
2180 if (datagrams > 0 &&
2181 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2182 datagrams = -EFAULT;
2183
2184 return datagrams;
2185}
2186
2187#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188/* Argument list sizes for sys_socketcall */
2189#define AL(x) ((x) * sizeof(unsigned long))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002190static const unsigned char nargs[20] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002191 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2192 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2193 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2194 AL(4), AL(5)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002195};
2196
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197#undef AL
2198
2199/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002200 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201 *
2202 * Argument checking cleaned up. Saved 20% in size.
2203 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002204 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205 */
2206
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002207SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208{
2209 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002210 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002212 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002214 if (call < 1 || call > SYS_RECVMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215 return -EINVAL;
2216
Arjan van de Ven47379052009-09-28 12:57:44 -07002217 len = nargs[call];
2218 if (len > sizeof(a))
2219 return -EINVAL;
2220
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002222 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002224
Al Virof3298dc2008-12-10 03:16:51 -05002225 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002226
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002227 a0 = a[0];
2228 a1 = a[1];
2229
2230 switch (call) {
2231 case SYS_SOCKET:
2232 err = sys_socket(a0, a1, a[2]);
2233 break;
2234 case SYS_BIND:
2235 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2236 break;
2237 case SYS_CONNECT:
2238 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2239 break;
2240 case SYS_LISTEN:
2241 err = sys_listen(a0, a1);
2242 break;
2243 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002244 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2245 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002246 break;
2247 case SYS_GETSOCKNAME:
2248 err =
2249 sys_getsockname(a0, (struct sockaddr __user *)a1,
2250 (int __user *)a[2]);
2251 break;
2252 case SYS_GETPEERNAME:
2253 err =
2254 sys_getpeername(a0, (struct sockaddr __user *)a1,
2255 (int __user *)a[2]);
2256 break;
2257 case SYS_SOCKETPAIR:
2258 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2259 break;
2260 case SYS_SEND:
2261 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2262 break;
2263 case SYS_SENDTO:
2264 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2265 (struct sockaddr __user *)a[4], a[5]);
2266 break;
2267 case SYS_RECV:
2268 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2269 break;
2270 case SYS_RECVFROM:
2271 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2272 (struct sockaddr __user *)a[4],
2273 (int __user *)a[5]);
2274 break;
2275 case SYS_SHUTDOWN:
2276 err = sys_shutdown(a0, a1);
2277 break;
2278 case SYS_SETSOCKOPT:
2279 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2280 break;
2281 case SYS_GETSOCKOPT:
2282 err =
2283 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2284 (int __user *)a[4]);
2285 break;
2286 case SYS_SENDMSG:
2287 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2288 break;
2289 case SYS_RECVMSG:
2290 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2291 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002292 case SYS_RECVMMSG:
2293 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2294 (struct timespec __user *)a[4]);
2295 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002296 case SYS_ACCEPT4:
2297 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2298 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002299 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002300 default:
2301 err = -EINVAL;
2302 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303 }
2304 return err;
2305}
2306
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002307#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002309/**
2310 * sock_register - add a socket protocol handler
2311 * @ops: description of protocol
2312 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313 * This function is called by a protocol handler that wants to
2314 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002315 * socket interface. The value ops->family coresponds to the
2316 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002318int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319{
2320 int err;
2321
2322 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002323 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2324 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 return -ENOBUFS;
2326 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002327
2328 spin_lock(&net_family_lock);
2329 if (net_families[ops->family])
2330 err = -EEXIST;
2331 else {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002332 net_families[ops->family] = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333 err = 0;
2334 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002335 spin_unlock(&net_family_lock);
2336
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002337 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338 return err;
2339}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002340EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002342/**
2343 * sock_unregister - remove a protocol handler
2344 * @family: protocol family to remove
2345 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346 * This function is called by a protocol handler that wants to
2347 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002348 * new socket creation.
2349 *
2350 * If protocol handler is a module, then it can use module reference
2351 * counts to protect against new references. If protocol handler is not
2352 * a module then it needs to provide its own protection in
2353 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002355void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002357 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002359 spin_lock(&net_family_lock);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002360 net_families[family] = NULL;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002361 spin_unlock(&net_family_lock);
2362
2363 synchronize_rcu();
2364
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002365 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002367EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368
Andi Kleen77d76ea2005-12-22 12:43:42 -08002369static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002370{
2371 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002372 * Initialize sock SLAB cache.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002374
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375 sk_init();
2376
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002378 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379 */
2380 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
2382 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002383 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002384 */
2385
2386 init_inodecache();
2387 register_filesystem(&sock_fs_type);
2388 sock_mnt = kern_mount(&sock_fs_type);
Andi Kleen77d76ea2005-12-22 12:43:42 -08002389
2390 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391 */
2392
2393#ifdef CONFIG_NETFILTER
2394 netfilter_init();
2395#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002396
Richard Cochranc1f19b52010-07-17 08:49:36 +00002397#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2398 skb_timestamping_init();
2399#endif
2400
David S. Millercbeb3212005-12-22 12:58:55 -08002401 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402}
2403
Andi Kleen77d76ea2005-12-22 12:43:42 -08002404core_initcall(sock_init); /* early initcall */
2405
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406#ifdef CONFIG_PROC_FS
2407void socket_seq_show(struct seq_file *seq)
2408{
2409 int cpu;
2410 int counter = 0;
2411
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002412 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002413 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002414
2415 /* It can be negative, by the way. 8) */
2416 if (counter < 0)
2417 counter = 0;
2418
2419 seq_printf(seq, "sockets: used %d\n", counter);
2420}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002421#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002423#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002424static int do_siocgstamp(struct net *net, struct socket *sock,
2425 unsigned int cmd, struct compat_timeval __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002426{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002427 mm_segment_t old_fs = get_fs();
2428 struct timeval ktv;
2429 int err;
2430
2431 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002432 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002433 set_fs(old_fs);
2434 if (!err) {
2435 err = put_user(ktv.tv_sec, &up->tv_sec);
2436 err |= __put_user(ktv.tv_usec, &up->tv_usec);
2437 }
2438 return err;
2439}
2440
Arnd Bergmann6b960182009-11-06 23:10:54 -08002441static int do_siocgstampns(struct net *net, struct socket *sock,
2442 unsigned int cmd, struct compat_timespec __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002443{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002444 mm_segment_t old_fs = get_fs();
2445 struct timespec kts;
2446 int err;
2447
2448 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002449 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002450 set_fs(old_fs);
2451 if (!err) {
2452 err = put_user(kts.tv_sec, &up->tv_sec);
2453 err |= __put_user(kts.tv_nsec, &up->tv_nsec);
2454 }
2455 return err;
2456}
2457
Arnd Bergmann6b960182009-11-06 23:10:54 -08002458static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002459{
2460 struct ifreq __user *uifr;
2461 int err;
2462
2463 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002464 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002465 return -EFAULT;
2466
Arnd Bergmann6b960182009-11-06 23:10:54 -08002467 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002468 if (err)
2469 return err;
2470
Arnd Bergmann6b960182009-11-06 23:10:54 -08002471 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002472 return -EFAULT;
2473
2474 return 0;
2475}
2476
Arnd Bergmann6b960182009-11-06 23:10:54 -08002477static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002478{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002479 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002480 struct ifconf ifc;
2481 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002482 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002483 struct ifreq __user *ifr;
2484 unsigned int i, j;
2485 int err;
2486
Arnd Bergmann6b960182009-11-06 23:10:54 -08002487 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002488 return -EFAULT;
2489
2490 if (ifc32.ifcbuf == 0) {
2491 ifc32.ifc_len = 0;
2492 ifc.ifc_len = 0;
2493 ifc.ifc_req = NULL;
2494 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2495 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002496 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2497 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002498 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2499 ifc.ifc_len = len;
2500 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2501 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002502 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002503 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002504 return -EFAULT;
2505 ifr++;
2506 ifr32++;
2507 }
2508 }
2509 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2510 return -EFAULT;
2511
Arnd Bergmann6b960182009-11-06 23:10:54 -08002512 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002513 if (err)
2514 return err;
2515
2516 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2517 return -EFAULT;
2518
2519 ifr = ifc.ifc_req;
2520 ifr32 = compat_ptr(ifc32.ifcbuf);
2521 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002522 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2523 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2524 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002525 return -EFAULT;
2526 ifr32++;
2527 ifr++;
2528 }
2529
2530 if (ifc32.ifcbuf == 0) {
2531 /* Translate from 64-bit structure multiple to
2532 * a 32-bit one.
2533 */
2534 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002535 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002536 ifc32.ifc_len = i;
2537 } else {
2538 ifc32.ifc_len = i;
2539 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002540 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002541 return -EFAULT;
2542
2543 return 0;
2544}
2545
Arnd Bergmann6b960182009-11-06 23:10:54 -08002546static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002547{
2548 struct ifreq __user *ifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002549 u32 data;
2550 void __user *datap;
2551
2552 ifr = compat_alloc_user_space(sizeof(*ifr));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002553
2554 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2555 return -EFAULT;
2556
2557 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2558 return -EFAULT;
2559
2560 datap = compat_ptr(data);
2561 if (put_user(datap, &ifr->ifr_ifru.ifru_data))
2562 return -EFAULT;
2563
Arnd Bergmann6b960182009-11-06 23:10:54 -08002564 return dev_ioctl(net, SIOCETHTOOL, ifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002565}
2566
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002567static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2568{
2569 void __user *uptr;
2570 compat_uptr_t uptr32;
2571 struct ifreq __user *uifr;
2572
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002573 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002574 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2575 return -EFAULT;
2576
2577 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2578 return -EFAULT;
2579
2580 uptr = compat_ptr(uptr32);
2581
2582 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2583 return -EFAULT;
2584
2585 return dev_ioctl(net, SIOCWANDEV, uifr);
2586}
2587
Arnd Bergmann6b960182009-11-06 23:10:54 -08002588static int bond_ioctl(struct net *net, unsigned int cmd,
2589 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002590{
2591 struct ifreq kifr;
2592 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002593 mm_segment_t old_fs;
2594 int err;
2595 u32 data;
2596 void __user *datap;
2597
2598 switch (cmd) {
2599 case SIOCBONDENSLAVE:
2600 case SIOCBONDRELEASE:
2601 case SIOCBONDSETHWADDR:
2602 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002603 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002604 return -EFAULT;
2605
2606 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002607 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002608 err = dev_ioctl(net, cmd, &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002609 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002610
2611 return err;
2612 case SIOCBONDSLAVEINFOQUERY:
2613 case SIOCBONDINFOQUERY:
2614 uifr = compat_alloc_user_space(sizeof(*uifr));
2615 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2616 return -EFAULT;
2617
2618 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2619 return -EFAULT;
2620
2621 datap = compat_ptr(data);
2622 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2623 return -EFAULT;
2624
Arnd Bergmann6b960182009-11-06 23:10:54 -08002625 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002626 default:
2627 return -EINVAL;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002628 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002629}
2630
Arnd Bergmann6b960182009-11-06 23:10:54 -08002631static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2632 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002633{
2634 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002635 char tmp_buf[IFNAMSIZ];
2636 void __user *data64;
2637 u32 data32;
2638
2639 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2640 IFNAMSIZ))
2641 return -EFAULT;
2642 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2643 return -EFAULT;
2644 data64 = compat_ptr(data32);
2645
2646 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2647
2648 /* Don't check these user accesses, just let that get trapped
2649 * in the ioctl handler instead.
2650 */
2651 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2652 IFNAMSIZ))
2653 return -EFAULT;
2654 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2655 return -EFAULT;
2656
Arnd Bergmann6b960182009-11-06 23:10:54 -08002657 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002658}
2659
Arnd Bergmann6b960182009-11-06 23:10:54 -08002660static int dev_ifsioc(struct net *net, struct socket *sock,
2661 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002662{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002663 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002664 int err;
2665
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002666 uifr = compat_alloc_user_space(sizeof(*uifr));
2667 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2668 return -EFAULT;
2669
2670 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2671
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002672 if (!err) {
2673 switch (cmd) {
2674 case SIOCGIFFLAGS:
2675 case SIOCGIFMETRIC:
2676 case SIOCGIFMTU:
2677 case SIOCGIFMEM:
2678 case SIOCGIFHWADDR:
2679 case SIOCGIFINDEX:
2680 case SIOCGIFADDR:
2681 case SIOCGIFBRDADDR:
2682 case SIOCGIFDSTADDR:
2683 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002684 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002685 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002686 case SIOCGMIIPHY:
2687 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002688 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002689 err = -EFAULT;
2690 break;
2691 }
2692 }
2693 return err;
2694}
2695
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002696static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2697 struct compat_ifreq __user *uifr32)
2698{
2699 struct ifreq ifr;
2700 struct compat_ifmap __user *uifmap32;
2701 mm_segment_t old_fs;
2702 int err;
2703
2704 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2705 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2706 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2707 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2708 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2709 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
2710 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
2711 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
2712 if (err)
2713 return -EFAULT;
2714
2715 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002716 set_fs(KERNEL_DS);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002717 err = dev_ioctl(net, cmd, (void __user *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002718 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002719
2720 if (cmd == SIOCGIFMAP && !err) {
2721 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2722 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2723 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2724 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2725 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
2726 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
2727 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
2728 if (err)
2729 err = -EFAULT;
2730 }
2731 return err;
2732}
2733
2734static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
2735{
2736 void __user *uptr;
2737 compat_uptr_t uptr32;
2738 struct ifreq __user *uifr;
2739
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002740 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002741 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2742 return -EFAULT;
2743
2744 if (get_user(uptr32, &uifr32->ifr_data))
2745 return -EFAULT;
2746
2747 uptr = compat_ptr(uptr32);
2748
2749 if (put_user(uptr, &uifr->ifr_data))
2750 return -EFAULT;
2751
2752 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
2753}
2754
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002755struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002756 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002757 struct sockaddr rt_dst; /* target address */
2758 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2759 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002760 unsigned short rt_flags;
2761 short rt_pad2;
2762 u32 rt_pad3;
2763 unsigned char rt_tos;
2764 unsigned char rt_class;
2765 short rt_pad4;
2766 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002767 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002768 u32 rt_mtu; /* per route MTU/Window */
2769 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002770 unsigned short rt_irtt; /* Initial RTT */
2771};
2772
2773struct in6_rtmsg32 {
2774 struct in6_addr rtmsg_dst;
2775 struct in6_addr rtmsg_src;
2776 struct in6_addr rtmsg_gateway;
2777 u32 rtmsg_type;
2778 u16 rtmsg_dst_len;
2779 u16 rtmsg_src_len;
2780 u32 rtmsg_metric;
2781 u32 rtmsg_info;
2782 u32 rtmsg_flags;
2783 s32 rtmsg_ifindex;
2784};
2785
Arnd Bergmann6b960182009-11-06 23:10:54 -08002786static int routing_ioctl(struct net *net, struct socket *sock,
2787 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002788{
2789 int ret;
2790 void *r = NULL;
2791 struct in6_rtmsg r6;
2792 struct rtentry r4;
2793 char devname[16];
2794 u32 rtdev;
2795 mm_segment_t old_fs = get_fs();
2796
Arnd Bergmann6b960182009-11-06 23:10:54 -08002797 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2798 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002799 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002800 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002801 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2802 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2803 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2804 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2805 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2806 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2807 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002808
2809 r = (void *) &r6;
2810 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002811 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002812 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002813 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002814 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
2815 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
2816 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
2817 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
2818 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
2819 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002820 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002821 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002822 r4.rt_dev = devname; devname[15] = 0;
2823 } else
2824 r4.rt_dev = NULL;
2825
2826 r = (void *) &r4;
2827 }
2828
2829 if (ret) {
2830 ret = -EFAULT;
2831 goto out;
2832 }
2833
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002834 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002835 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002836 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002837
2838out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002839 return ret;
2840}
2841
2842/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
2843 * for some operations; this forces use of the newer bridge-utils that
2844 * use compatiable ioctls
2845 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002846static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002847{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002848 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002849
Arnd Bergmann6b960182009-11-06 23:10:54 -08002850 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002851 return -EFAULT;
2852 if (tmp == BRCTL_GET_VERSION)
2853 return BRCTL_VERSION + 1;
2854 return -EINVAL;
2855}
2856
Arnd Bergmann6b960182009-11-06 23:10:54 -08002857static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
2858 unsigned int cmd, unsigned long arg)
2859{
2860 void __user *argp = compat_ptr(arg);
2861 struct sock *sk = sock->sk;
2862 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002863
Arnd Bergmann6b960182009-11-06 23:10:54 -08002864 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
2865 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002866
Arnd Bergmann6b960182009-11-06 23:10:54 -08002867 switch (cmd) {
2868 case SIOCSIFBR:
2869 case SIOCGIFBR:
2870 return old_bridge_ioctl(argp);
2871 case SIOCGIFNAME:
2872 return dev_ifname32(net, argp);
2873 case SIOCGIFCONF:
2874 return dev_ifconf(net, argp);
2875 case SIOCETHTOOL:
2876 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002877 case SIOCWANDEV:
2878 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002879 case SIOCGIFMAP:
2880 case SIOCSIFMAP:
2881 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002882 case SIOCBONDENSLAVE:
2883 case SIOCBONDRELEASE:
2884 case SIOCBONDSETHWADDR:
2885 case SIOCBONDSLAVEINFOQUERY:
2886 case SIOCBONDINFOQUERY:
2887 case SIOCBONDCHANGEACTIVE:
2888 return bond_ioctl(net, cmd, argp);
2889 case SIOCADDRT:
2890 case SIOCDELRT:
2891 return routing_ioctl(net, sock, cmd, argp);
2892 case SIOCGSTAMP:
2893 return do_siocgstamp(net, sock, cmd, argp);
2894 case SIOCGSTAMPNS:
2895 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002896 case SIOCSHWTSTAMP:
2897 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002898
Arnd Bergmann6b960182009-11-06 23:10:54 -08002899 case FIOSETOWN:
2900 case SIOCSPGRP:
2901 case FIOGETOWN:
2902 case SIOCGPGRP:
2903 case SIOCBRADDBR:
2904 case SIOCBRDELBR:
2905 case SIOCGIFVLAN:
2906 case SIOCSIFVLAN:
2907 case SIOCADDDLCI:
2908 case SIOCDELDLCI:
2909 return sock_ioctl(file, cmd, arg);
2910
2911 case SIOCGIFFLAGS:
2912 case SIOCSIFFLAGS:
2913 case SIOCGIFMETRIC:
2914 case SIOCSIFMETRIC:
2915 case SIOCGIFMTU:
2916 case SIOCSIFMTU:
2917 case SIOCGIFMEM:
2918 case SIOCSIFMEM:
2919 case SIOCGIFHWADDR:
2920 case SIOCSIFHWADDR:
2921 case SIOCADDMULTI:
2922 case SIOCDELMULTI:
2923 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002924 case SIOCGIFADDR:
2925 case SIOCSIFADDR:
2926 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002927 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002928 case SIOCGIFBRDADDR:
2929 case SIOCSIFBRDADDR:
2930 case SIOCGIFDSTADDR:
2931 case SIOCSIFDSTADDR:
2932 case SIOCGIFNETMASK:
2933 case SIOCSIFNETMASK:
2934 case SIOCSIFPFLAGS:
2935 case SIOCGIFPFLAGS:
2936 case SIOCGIFTXQLEN:
2937 case SIOCSIFTXQLEN:
2938 case SIOCBRADDIF:
2939 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002940 case SIOCSIFNAME:
2941 case SIOCGMIIPHY:
2942 case SIOCGMIIREG:
2943 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002944 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002945
Arnd Bergmann6b960182009-11-06 23:10:54 -08002946 case SIOCSARP:
2947 case SIOCGARP:
2948 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002949 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002950 return sock_do_ioctl(net, sock, cmd, arg);
2951 }
2952
2953 /* Prevent warning from compat_sys_ioctl, these always
2954 * result in -EINVAL in the native case anyway. */
2955 switch (cmd) {
2956 case SIOCRTMSG:
2957 case SIOCGIFCOUNT:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002958 case SIOCSRARP:
2959 case SIOCGRARP:
2960 case SIOCDRARP:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00002961 case SIOCSIFLINK:
2962 case SIOCGIFSLAVE:
2963 case SIOCSIFSLAVE:
2964 return -EINVAL;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002965 }
2966
2967 return -ENOIOCTLCMD;
2968}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002969
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002970static long compat_sock_ioctl(struct file *file, unsigned cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002971 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002972{
2973 struct socket *sock = file->private_data;
2974 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07002975 struct sock *sk;
2976 struct net *net;
2977
2978 sk = sock->sk;
2979 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002980
2981 if (sock->ops->compat_ioctl)
2982 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2983
David S. Miller87de87d2008-06-03 09:14:03 -07002984 if (ret == -ENOIOCTLCMD &&
2985 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2986 ret = compat_wext_handle_ioctl(net, cmd, arg);
2987
Arnd Bergmann6b960182009-11-06 23:10:54 -08002988 if (ret == -ENOIOCTLCMD)
2989 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
2990
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002991 return ret;
2992}
2993#endif
2994
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07002995int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2996{
2997 return sock->ops->bind(sock, addr, addrlen);
2998}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002999EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003000
3001int kernel_listen(struct socket *sock, int backlog)
3002{
3003 return sock->ops->listen(sock, backlog);
3004}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003005EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003006
3007int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3008{
3009 struct sock *sk = sock->sk;
3010 int err;
3011
3012 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3013 newsock);
3014 if (err < 0)
3015 goto done;
3016
3017 err = sock->ops->accept(sock, *newsock, flags);
3018 if (err < 0) {
3019 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003020 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003021 goto done;
3022 }
3023
3024 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003025 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003026
3027done:
3028 return err;
3029}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003030EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003031
3032int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003033 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003034{
3035 return sock->ops->connect(sock, addr, addrlen, flags);
3036}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003037EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003038
3039int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3040 int *addrlen)
3041{
3042 return sock->ops->getname(sock, addr, addrlen, 0);
3043}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003044EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003045
3046int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3047 int *addrlen)
3048{
3049 return sock->ops->getname(sock, addr, addrlen, 1);
3050}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003051EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003052
3053int kernel_getsockopt(struct socket *sock, int level, int optname,
3054 char *optval, int *optlen)
3055{
3056 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003057 char __user *uoptval;
3058 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003059 int err;
3060
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003061 uoptval = (char __user __force *) optval;
3062 uoptlen = (int __user __force *) optlen;
3063
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003064 set_fs(KERNEL_DS);
3065 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003066 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003067 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003068 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3069 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003070 set_fs(oldfs);
3071 return err;
3072}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003073EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003074
3075int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003076 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003077{
3078 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003079 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003080 int err;
3081
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003082 uoptval = (char __user __force *) optval;
3083
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003084 set_fs(KERNEL_DS);
3085 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003086 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003087 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003088 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003089 optlen);
3090 set_fs(oldfs);
3091 return err;
3092}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003093EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003094
3095int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3096 size_t size, int flags)
3097{
Herbert Xuf8451722010-05-24 00:12:34 -07003098 sock_update_classid(sock->sk);
3099
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003100 if (sock->ops->sendpage)
3101 return sock->ops->sendpage(sock, page, offset, size, flags);
3102
3103 return sock_no_sendpage(sock, page, offset, size, flags);
3104}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003105EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003106
3107int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3108{
3109 mm_segment_t oldfs = get_fs();
3110 int err;
3111
3112 set_fs(KERNEL_DS);
3113 err = sock->ops->ioctl(sock, cmd, arg);
3114 set_fs(oldfs);
3115
3116 return err;
3117}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003118EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003119
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003120int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3121{
3122 return sock->ops->shutdown(sock, how);
3123}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003124EXPORT_SYMBOL(kernel_sock_shutdown);