blob: 4ca1526db7562ff7a1ae189a5b51bd6487c045ed [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/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030073#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <linux/init.h>
76#include <linux/poll.h>
77#include <linux/cache.h>
78#include <linux/module.h>
79#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070080#include <linux/mount.h>
81#include <linux/security.h>
82#include <linux/syscalls.h>
83#include <linux/compat.h>
84#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010085#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010086#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070087#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070088#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090089#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000090#include <linux/xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92#include <asm/uaccess.h>
93#include <asm/unistd.h>
94
95#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070096#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070097#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098
99#include <net/sock.h>
100#include <linux/netfilter.h>
101
Arnd Bergmann6b960182009-11-06 23:10:54 -0800102#include <linux/if_tun.h>
103#include <linux/ipv6_route.h>
104#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800105#include <linux/sockios.h>
106#include <linux/atalk.h>
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700109static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
110 unsigned long nr_segs, loff_t pos);
111static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
112 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700113static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115static int sock_close(struct inode *inode, struct file *file);
116static unsigned int sock_poll(struct file *file,
117 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800119#ifdef CONFIG_COMPAT
120static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700121 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800122#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124static ssize_t sock_sendpage(struct file *file, struct page *page,
125 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800126static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700127 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800128 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130/*
131 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
132 * in the operation structures but are done directly via the socketcall() multiplexor.
133 */
134
Arjan van de Venda7071d2007-02-12 00:55:36 -0800135static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 .owner = THIS_MODULE,
137 .llseek = no_llseek,
138 .aio_read = sock_aio_read,
139 .aio_write = sock_aio_write,
140 .poll = sock_poll,
141 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800142#ifdef CONFIG_COMPAT
143 .compat_ioctl = compat_sock_ioctl,
144#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 .mmap = sock_mmap,
146 .open = sock_no_open, /* special open code to disallow open via /proc */
147 .release = sock_close,
148 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200149 .sendpage = sock_sendpage,
150 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800151 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152};
153
154/*
155 * The protocol list. Each protocol is registered in here.
156 */
157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000159static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161/*
162 * Statistics counters of the socket lists
163 */
164
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700165static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
167/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700168 * Support routines.
169 * Move socket addresses back and forth across the kernel/user
170 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
172
Linus Torvalds1da177e2005-04-16 15:20:36 -0700173/**
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
178 *
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 */
183
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000184int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *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
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000212static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000213 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214{
215 int err;
216 int len;
217
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700218 err = get_user(len, ulen);
219 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700221 if (len > klen)
222 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700223 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500226 if (audit_sockaddr(klen, kaddr))
227 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return -EFAULT;
230 }
231 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 * "fromlen shall refer to the value before truncation.."
233 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235 return __put_user(klen, ulen);
236}
237
Christoph Lametere18b8902006-12-06 20:33:20 -0800238static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240static struct inode *sock_alloc_inode(struct super_block *sb)
241{
242 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000243 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700244
Christoph Lametere94b1762006-12-06 20:33:17 -0800245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 if (!ei)
247 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000248 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
249 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000250 kmem_cache_free(sock_inode_cachep, ei);
251 return NULL;
252 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000253 init_waitqueue_head(&wq->wait);
254 wq->fasync_list = NULL;
255 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700256
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
263 return &ei->vfs_inode;
264}
265
266static void sock_destroy_inode(struct inode *inode)
267{
Eric Dumazet43815482010-04-29 11:01:49 +0000268 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000269 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000270
271 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000272 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800273 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000274 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275}
276
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700277static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700279 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
Christoph Lametera35afb82007-05-16 22:10:57 -0700281 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284static int init_inodecache(void)
285{
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700287 sizeof(struct socket_alloc),
288 0,
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
291 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900292 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 if (sock_inode_cachep == NULL)
294 return -ENOMEM;
295 return 0;
296}
297
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700298static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700299 .alloc_inode = sock_alloc_inode,
300 .destroy_inode = sock_destroy_inode,
301 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302};
303
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700304/*
305 * sockfs_dname() is called from d_path().
306 */
307static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
308{
309 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
310 dentry->d_inode->i_ino);
311}
312
Al Viro3ba13d12009-02-20 06:02:22 +0000313static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700314 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315};
316
Al Viroc74a1cb2011-01-12 16:59:34 -0500317static struct dentry *sockfs_mount(struct file_system_type *fs_type,
318 int flags, const char *dev_name, void *data)
319{
320 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
321 &sockfs_dentry_operations, SOCKFS_MAGIC);
322}
323
324static struct vfsmount *sock_mnt __read_mostly;
325
326static struct file_system_type sock_fs_type = {
327 .name = "sockfs",
328 .mount = sockfs_mount,
329 .kill_sb = kill_anon_super,
330};
331
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332/*
333 * Obtains the first available file descriptor and sets it up for use.
334 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800335 * These functions create file structures and maps them to fd space
336 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 * and file struct implicitly stored in sock->file.
338 * Note that another thread may close file descriptor before we return
339 * from this function. We use the fact that now we do not refer
340 * to socket after mapping. If one day we will need it, this
341 * function will increment ref. count on file by 1.
342 *
343 * In any case returned fd MAY BE not valid!
344 * This race condition is unavoidable
345 * with shared fd spaces, we cannot solve it inside kernel,
346 * but we take care of internal coherence yet.
347 */
348
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700349struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
Al Viro7cbe66b2009-08-05 19:59:08 +0400351 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400352 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400353 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800354
Masatake YAMATO600e1772012-08-29 10:44:29 +0000355 if (dname) {
356 name.name = dname;
357 name.len = strlen(name.name);
358 } else if (sock->sk) {
359 name.name = sock->sk->sk_prot_creator->name;
360 name.len = strlen(name.name);
361 }
Nick Piggin4b936882011-01-07 17:50:07 +1100362 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400363 if (unlikely(!path.dentry))
364 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400365 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366
Al Viro2c48b9c2009-08-09 00:52:35 +0400367 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400368 SOCK_INODE(sock)->i_fop = &socket_file_ops;
369
Al Viro2c48b9c2009-08-09 00:52:35 +0400370 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400371 &socket_file_ops);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700372 if (unlikely(IS_ERR(file))) {
Al Virocc3808f2009-08-06 09:43:59 +0400373 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400374 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400375 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700376 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400377 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800378
379 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700380 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800381 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400382 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800383}
Al Viro56b31d12012-08-18 00:25:51 -0400384EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800385
Al Viro56b31d12012-08-18 00:25:51 -0400386static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387{
388 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400389 int fd = get_unused_fd_flags(flags);
390 if (unlikely(fd < 0))
391 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700393 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400394 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800395 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400396 return fd;
397 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400398
Al Viro28407632012-08-17 23:54:15 -0400399 put_unused_fd(fd);
400 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401}
402
John Fastabend406a3c62012-07-20 10:39:25 +0000403struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800404{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800405 if (file->f_op == &socket_file_ops)
406 return file->private_data; /* set in sock_map_fd */
407
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800408 *err = -ENOTSOCK;
409 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800410}
John Fastabend406a3c62012-07-20 10:39:25 +0000411EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800412
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700414 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 * @fd: file handle
416 * @err: pointer to an error code return
417 *
418 * The file handle passed in is locked and the socket it is bound
419 * too is returned. If an error occurs the err pointer is overwritten
420 * with a negative errno code and NULL is returned. The function checks
421 * for both invalid handles and passing a handle which is not a socket.
422 *
423 * On a success the socket object pointer is returned.
424 */
425
426struct socket *sockfd_lookup(int fd, int *err)
427{
428 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 struct socket *sock;
430
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700431 file = fget(fd);
432 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 *err = -EBADF;
434 return NULL;
435 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700436
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800437 sock = sock_from_file(file, err);
438 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 return sock;
441}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700442EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800444static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
445{
446 struct file *file;
447 struct socket *sock;
448
Hua Zhong36725582006-04-19 15:25:02 -0700449 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800450 file = fget_light(fd, fput_needed);
451 if (file) {
452 sock = sock_from_file(file, err);
453 if (sock)
454 return sock;
455 fput_light(file, *fput_needed);
456 }
457 return NULL;
458}
459
Masatake YAMATO600e1772012-08-29 10:44:29 +0000460#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
461#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
462#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
463static ssize_t sockfs_getxattr(struct dentry *dentry,
464 const char *name, void *value, size_t size)
465{
466 const char *proto_name;
467 size_t proto_size;
468 int error;
469
470 error = -ENODATA;
471 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
472 proto_name = dentry->d_name.name;
473 proto_size = strlen(proto_name);
474
475 if (value) {
476 error = -ERANGE;
477 if (proto_size + 1 > size)
478 goto out;
479
480 strncpy(value, proto_name, proto_size + 1);
481 }
482 error = proto_size + 1;
483 }
484
485out:
486 return error;
487}
488
489static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
490 size_t size)
491{
492 ssize_t len;
493 ssize_t used = 0;
494
495 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
496 if (len < 0)
497 return len;
498 used += len;
499 if (buffer) {
500 if (size < used)
501 return -ERANGE;
502 buffer += len;
503 }
504
505 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
506 used += len;
507 if (buffer) {
508 if (size < used)
509 return -ERANGE;
510 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
511 buffer += len;
512 }
513
514 return used;
515}
516
517static const struct inode_operations sockfs_inode_ops = {
518 .getxattr = sockfs_getxattr,
519 .listxattr = sockfs_listxattr,
520};
521
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522/**
523 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700524 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 * Allocate a new inode and socket object. The two are bound together
526 * and initialised. The socket is then returned. If we are out of inodes
527 * NULL is returned.
528 */
529
530static struct socket *sock_alloc(void)
531{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700532 struct inode *inode;
533 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200535 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 if (!inode)
537 return NULL;
538
539 sock = SOCKET_I(inode);
540
Eric Dumazet29a020d2009-09-15 02:39:20 -0700541 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400542 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700543 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100544 inode->i_uid = current_fsuid();
545 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000546 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
Alex Shi19e8d692012-05-14 14:15:31 -0700548 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 return sock;
550}
551
552/*
553 * In theory you can't get an open on this inode, but /proc provides
554 * a back door. Remember to keep it shut otherwise you'll let the
555 * creepy crawlies in.
556 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
559{
560 return -ENXIO;
561}
562
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800563const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 .owner = THIS_MODULE,
565 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200566 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567};
568
569/**
570 * sock_release - close a socket
571 * @sock: socket to close
572 *
573 * The socket is released from the protocol stack if it has a release
574 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700575 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578void sock_release(struct socket *sock)
579{
580 if (sock->ops) {
581 struct module *owner = sock->ops->owner;
582
583 sock->ops->release(sock);
584 sock->ops = NULL;
585 module_put(owner);
586 }
587
Eric Dumazeteaefd112011-02-18 03:26:36 +0000588 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 printk(KERN_ERR "sock_release: fasync list not empty!\n");
590
Mikulas Patockab09e7862012-07-19 06:13:36 +0000591 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
592 return;
593
Alex Shi19e8d692012-05-14 14:15:31 -0700594 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 if (!sock->file) {
596 iput(SOCK_INODE(sock));
597 return;
598 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700599 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700601EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000603void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000604{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000605 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000606 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000607 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000608 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000609 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100610 if (sock_flag(sk, SOCK_WIFI_STATUS))
611 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000612}
613EXPORT_SYMBOL(sock_tx_timestamp);
614
Anton Blanchard228e5482011-05-02 20:21:35 +0000615static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
616 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617{
618 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619
620 si->sock = sock;
621 si->scm = NULL;
622 si->msg = msg;
623 si->size = size;
624
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 return sock->ops->sendmsg(iocb, sock, msg, size);
626}
627
Anton Blanchard228e5482011-05-02 20:21:35 +0000628static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
629 struct msghdr *msg, size_t size)
630{
631 int err = security_socket_sendmsg(sock, msg, size);
632
633 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
634}
635
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
637{
638 struct kiocb iocb;
639 struct sock_iocb siocb;
640 int ret;
641
642 init_sync_kiocb(&iocb, NULL);
643 iocb.private = &siocb;
644 ret = __sock_sendmsg(&iocb, sock, msg, size);
645 if (-EIOCBQUEUED == ret)
646 ret = wait_on_sync_kiocb(&iocb);
647 return ret;
648}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700649EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Eric Dumazet894dc242011-07-26 02:39:41 +0000651static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000652{
653 struct kiocb iocb;
654 struct sock_iocb siocb;
655 int ret;
656
657 init_sync_kiocb(&iocb, NULL);
658 iocb.private = &siocb;
659 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
660 if (-EIOCBQUEUED == ret)
661 ret = wait_on_sync_kiocb(&iocb);
662 return ret;
663}
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
666 struct kvec *vec, size_t num, size_t size)
667{
668 mm_segment_t oldfs = get_fs();
669 int result;
670
671 set_fs(KERNEL_DS);
672 /*
673 * the following is safe, since for compiler definitions of kvec and
674 * iovec are identical, yielding the same in-core layout and alignment
675 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700676 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 msg->msg_iovlen = num;
678 result = sock_sendmsg(sock, msg, size);
679 set_fs(oldfs);
680 return result;
681}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700682EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700684/*
685 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
686 */
687void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
688 struct sk_buff *skb)
689{
Patrick Ohly20d49472009-02-12 05:03:38 +0000690 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
691 struct timespec ts[3];
692 int empty = 1;
693 struct skb_shared_hwtstamps *shhwtstamps =
694 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700695
Patrick Ohly20d49472009-02-12 05:03:38 +0000696 /* Race occurred between timestamp enabling and packet
697 receiving. Fill in the current time for now. */
698 if (need_software_tstamp && skb->tstamp.tv64 == 0)
699 __net_timestamp(skb);
700
701 if (need_software_tstamp) {
702 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
703 struct timeval tv;
704 skb_get_timestamp(skb, &tv);
705 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
706 sizeof(tv), &tv);
707 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000708 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000709 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000710 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000711 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700712 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000713
714
715 memset(ts, 0, sizeof(ts));
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000716 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
717 ktime_to_timespec_cond(skb->tstamp, ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000718 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000719 if (shhwtstamps) {
720 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000721 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
Patrick Ohly20d49472009-02-12 05:03:38 +0000722 empty = 0;
723 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000724 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
Patrick Ohly20d49472009-02-12 05:03:38 +0000725 empty = 0;
726 }
727 if (!empty)
728 put_cmsg(msg, SOL_SOCKET,
729 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700730}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300731EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
732
Johannes Berg6e3e9392011-11-09 10:15:42 +0100733void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
734 struct sk_buff *skb)
735{
736 int ack;
737
738 if (!sock_flag(sk, SOCK_WIFI_STATUS))
739 return;
740 if (!skb->wifi_acked_valid)
741 return;
742
743 ack = skb->wifi_acked;
744
745 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
746}
747EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
748
stephen hemminger11165f12010-10-18 14:27:29 +0000749static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
750 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700751{
752 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
753 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
754 sizeof(__u32), &skb->dropcount);
755}
756
Eric Dumazet767dd032010-04-28 19:14:43 +0000757void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700758 struct sk_buff *skb)
759{
760 sock_recv_timestamp(msg, sk, skb);
761 sock_recv_drops(msg, sk, skb);
762}
Eric Dumazet767dd032010-04-28 19:14:43 +0000763EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700764
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700765static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
766 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 struct sock_iocb *si = kiocb_to_siocb(iocb);
769
770 si->sock = sock;
771 si->scm = NULL;
772 si->msg = msg;
773 si->size = size;
774 si->flags = flags;
775
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
777}
778
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700779static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
780 struct msghdr *msg, size_t size, int flags)
781{
782 int err = security_socket_recvmsg(sock, msg, size, flags);
783
784 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
785}
786
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700787int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 size_t size, int flags)
789{
790 struct kiocb iocb;
791 struct sock_iocb siocb;
792 int ret;
793
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700794 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 iocb.private = &siocb;
796 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
797 if (-EIOCBQUEUED == ret)
798 ret = wait_on_sync_kiocb(&iocb);
799 return ret;
800}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700801EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700803static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
804 size_t size, int flags)
805{
806 struct kiocb iocb;
807 struct sock_iocb siocb;
808 int ret;
809
810 init_sync_kiocb(&iocb, NULL);
811 iocb.private = &siocb;
812 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
813 if (-EIOCBQUEUED == ret)
814 ret = wait_on_sync_kiocb(&iocb);
815 return ret;
816}
817
Martin Lucinac1249c02010-12-10 00:04:05 +0000818/**
819 * kernel_recvmsg - Receive a message from a socket (kernel space)
820 * @sock: The socket to receive the message from
821 * @msg: Received message
822 * @vec: Input s/g array for message data
823 * @num: Size of input s/g array
824 * @size: Number of bytes to read
825 * @flags: Message flags (MSG_DONTWAIT, etc...)
826 *
827 * On return the msg structure contains the scatter/gather array passed in the
828 * vec argument. The array is modified so that it consists of the unfilled
829 * portion of the original array.
830 *
831 * The returned value is the total number of bytes received, or an error.
832 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700833int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
834 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
836 mm_segment_t oldfs = get_fs();
837 int result;
838
839 set_fs(KERNEL_DS);
840 /*
841 * the following is safe, since for compiler definitions of kvec and
842 * iovec are identical, yielding the same in-core layout and alignment
843 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700844 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845 result = sock_recvmsg(sock, msg, size, flags);
846 set_fs(oldfs);
847 return result;
848}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700849EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850
851static void sock_aio_dtor(struct kiocb *iocb)
852{
853 kfree(iocb->private);
854}
855
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300856static ssize_t sock_sendpage(struct file *file, struct page *page,
857 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858{
859 struct socket *sock;
860 int flags;
861
Eric Dumazetb69aee02005-09-06 14:42:45 -0700862 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
Eric Dumazet35f9c092012-04-05 03:05:35 +0000864 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
865 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
866 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
Linus Torvaldse6949582009-08-13 08:28:36 -0700868 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869}
870
Jens Axboe9c55e012007-11-06 23:30:13 -0800871static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700872 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800873 unsigned int flags)
874{
875 struct socket *sock = file->private_data;
876
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800877 if (unlikely(!sock->ops->splice_read))
878 return -EINVAL;
879
Jens Axboe9c55e012007-11-06 23:30:13 -0800880 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
881}
882
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800883static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700884 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800885{
886 if (!is_sync_kiocb(iocb)) {
887 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
888 if (!siocb)
889 return NULL;
890 iocb->ki_dtor = sock_aio_dtor;
891 }
892
893 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800894 iocb->private = siocb;
895 return siocb;
896}
897
898static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700899 struct file *file, const struct iovec *iov,
900 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800901{
902 struct socket *sock = file->private_data;
903 size_t size = 0;
904 int i;
905
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700906 for (i = 0; i < nr_segs; i++)
907 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800908
909 msg->msg_name = NULL;
910 msg->msg_namelen = 0;
911 msg->msg_control = NULL;
912 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700913 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800914 msg->msg_iovlen = nr_segs;
915 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
916
917 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
918}
919
Badari Pulavarty027445c2006-09-30 23:28:46 -0700920static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
921 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800922{
923 struct sock_iocb siocb, *x;
924
925 if (pos != 0)
926 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700927
928 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800929 return 0;
930
Badari Pulavarty027445c2006-09-30 23:28:46 -0700931
932 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800933 if (!x)
934 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700935 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800936}
937
938static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700939 struct file *file, const struct iovec *iov,
940 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800941{
942 struct socket *sock = file->private_data;
943 size_t size = 0;
944 int i;
945
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700946 for (i = 0; i < nr_segs; i++)
947 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800948
949 msg->msg_name = NULL;
950 msg->msg_namelen = 0;
951 msg->msg_control = NULL;
952 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700953 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800954 msg->msg_iovlen = nr_segs;
955 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
956 if (sock->type == SOCK_SEQPACKET)
957 msg->msg_flags |= MSG_EOR;
958
959 return __sock_sendmsg(iocb, sock, msg, size);
960}
961
Badari Pulavarty027445c2006-09-30 23:28:46 -0700962static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
963 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800965 struct sock_iocb siocb, *x;
966
967 if (pos != 0)
968 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700969
Badari Pulavarty027445c2006-09-30 23:28:46 -0700970 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800971 if (!x)
972 return -ENOMEM;
973
Badari Pulavarty027445c2006-09-30 23:28:46 -0700974 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975}
976
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977/*
978 * Atomic setting of ioctl hooks to avoid race
979 * with module unload.
980 */
981
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800982static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700983static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984
Eric W. Biederman881d9662007-09-17 11:56:21 -0700985void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800987 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800989 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990}
991EXPORT_SYMBOL(brioctl_set);
992
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800993static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700994static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995
Eric W. Biederman881d9662007-09-17 11:56:21 -0700996void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800998 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001000 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001}
1002EXPORT_SYMBOL(vlan_ioctl_set);
1003
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001004static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001005static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001007void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001009 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001011 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012}
1013EXPORT_SYMBOL(dlci_ioctl_set);
1014
Arnd Bergmann6b960182009-11-06 23:10:54 -08001015static long sock_do_ioctl(struct net *net, struct socket *sock,
1016 unsigned int cmd, unsigned long arg)
1017{
1018 int err;
1019 void __user *argp = (void __user *)arg;
1020
1021 err = sock->ops->ioctl(sock, cmd, arg);
1022
1023 /*
1024 * If this ioctl is unknown try to hand it down
1025 * to the NIC driver.
1026 */
1027 if (err == -ENOIOCTLCMD)
1028 err = dev_ioctl(net, cmd, argp);
1029
1030 return err;
1031}
1032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033/*
1034 * With an ioctl, arg may well be a user mode pointer, but we don't know
1035 * what to do with it - that's up to the protocol still.
1036 */
1037
1038static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1039{
1040 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001041 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 void __user *argp = (void __user *)arg;
1043 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001044 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045
Eric Dumazetb69aee02005-09-06 14:42:45 -07001046 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001047 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001048 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001050 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001052#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001054 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001056#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001057 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 case FIOSETOWN:
1059 case SIOCSPGRP:
1060 err = -EFAULT;
1061 if (get_user(pid, (int __user *)argp))
1062 break;
1063 err = f_setown(sock->file, pid, 1);
1064 break;
1065 case FIOGETOWN:
1066 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001067 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001068 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 break;
1070 case SIOCGIFBR:
1071 case SIOCSIFBR:
1072 case SIOCBRADDBR:
1073 case SIOCBRDELBR:
1074 err = -ENOPKG;
1075 if (!br_ioctl_hook)
1076 request_module("bridge");
1077
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001078 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001079 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001080 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001081 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 break;
1083 case SIOCGIFVLAN:
1084 case SIOCSIFVLAN:
1085 err = -ENOPKG;
1086 if (!vlan_ioctl_hook)
1087 request_module("8021q");
1088
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001089 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001091 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001092 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 case SIOCADDDLCI:
1095 case SIOCDELDLCI:
1096 err = -ENOPKG;
1097 if (!dlci_ioctl_hook)
1098 request_module("dlci");
1099
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001100 mutex_lock(&dlci_ioctl_mutex);
1101 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001103 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 break;
1105 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001106 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001108 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 return err;
1110}
1111
1112int sock_create_lite(int family, int type, int protocol, struct socket **res)
1113{
1114 int err;
1115 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001116
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 err = security_socket_create(family, type, protocol, 1);
1118 if (err)
1119 goto out;
1120
1121 sock = sock_alloc();
1122 if (!sock) {
1123 err = -ENOMEM;
1124 goto out;
1125 }
1126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001128 err = security_socket_post_create(sock, family, type, protocol, 1);
1129 if (err)
1130 goto out_release;
1131
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132out:
1133 *res = sock;
1134 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001135out_release:
1136 sock_release(sock);
1137 sock = NULL;
1138 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001140EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141
1142/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001143static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144{
1145 struct socket *sock;
1146
1147 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001148 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001150 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 return sock->ops->poll(file, sock, wait);
1152}
1153
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001154static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001156 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157
1158 return sock->ops->mmap(file, sock, vma);
1159}
1160
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001161static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 sock_release(SOCKET_I(inode));
1164 return 0;
1165}
1166
1167/*
1168 * Update the socket async list
1169 *
1170 * Fasync_list locking strategy.
1171 *
1172 * 1. fasync_list is modified only under process context socket lock
1173 * i.e. under semaphore.
1174 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001175 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 */
1177
1178static int sock_fasync(int fd, struct file *filp, int on)
1179{
Eric Dumazet989a2972010-04-14 09:55:35 +00001180 struct socket *sock = filp->private_data;
1181 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001182 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Eric Dumazet989a2972010-04-14 09:55:35 +00001184 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
1187 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001188 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1189 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190
Eric Dumazeteaefd112011-02-18 03:26:36 +00001191 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001192 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001193 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001194 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195
Eric Dumazet989a2972010-04-14 09:55:35 +00001196 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 return 0;
1198}
1199
Eric Dumazet43815482010-04-29 11:01:49 +00001200/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201
1202int sock_wake_async(struct socket *sock, int how, int band)
1203{
Eric Dumazet43815482010-04-29 11:01:49 +00001204 struct socket_wq *wq;
1205
1206 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001208 rcu_read_lock();
1209 wq = rcu_dereference(sock->wq);
1210 if (!wq || !wq->fasync_list) {
1211 rcu_read_unlock();
1212 return -1;
1213 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001214 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001215 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1217 break;
1218 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001219 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1221 break;
1222 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001223 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001224call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001225 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001227 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001228 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 }
Eric Dumazet43815482010-04-29 11:01:49 +00001230 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 return 0;
1232}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001233EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
Pavel Emelyanov721db932010-09-29 16:06:32 +04001235int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001236 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237{
1238 int err;
1239 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001240 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
1242 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001243 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 */
1245 if (family < 0 || family >= NPROTO)
1246 return -EAFNOSUPPORT;
1247 if (type < 0 || type >= SOCK_MAX)
1248 return -EINVAL;
1249
1250 /* Compatibility.
1251
1252 This uglymoron is moved from INET layer to here to avoid
1253 deadlock in module load.
1254 */
1255 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001256 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 if (!warned) {
1258 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001259 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1260 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 }
1262 family = PF_PACKET;
1263 }
1264
1265 err = security_socket_create(family, type, protocol, kern);
1266 if (err)
1267 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001268
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001269 /*
1270 * Allocate the socket and allow the family to set things up. if
1271 * the protocol is 0, the family is instructed to select an appropriate
1272 * default.
1273 */
1274 sock = sock_alloc();
1275 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001276 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001277 return -ENFILE; /* Not exactly a match, but its the
1278 closest posix thing */
1279 }
1280
1281 sock->type = type;
1282
Johannes Berg95a5afc2008-10-16 15:24:51 -07001283#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001284 /* Attempt to load a protocol module if the find failed.
1285 *
1286 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 * requested real, full-featured networking support upon configuration.
1288 * Otherwise module support will break!
1289 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001290 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001291 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292#endif
1293
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001294 rcu_read_lock();
1295 pf = rcu_dereference(net_families[family]);
1296 err = -EAFNOSUPPORT;
1297 if (!pf)
1298 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299
1300 /*
1301 * We will call the ->create function, that possibly is in a loadable
1302 * module, so we have to bump that loadable module refcnt first.
1303 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001304 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 goto out_release;
1306
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001307 /* Now protected by module ref count */
1308 rcu_read_unlock();
1309
Eric Paris3f378b62009-11-05 22:18:14 -08001310 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001311 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001313
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 /*
1315 * Now to bump the refcnt of the [loadable] module that owns this
1316 * socket at sock_release time we decrement its refcnt.
1317 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001318 if (!try_module_get(sock->ops->owner))
1319 goto out_module_busy;
1320
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 /*
1322 * Now that we're done with the ->create function, the [loadable]
1323 * module can have its refcnt decremented
1324 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001325 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001326 err = security_socket_post_create(sock, family, type, protocol, kern);
1327 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001328 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001329 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001331 return 0;
1332
1333out_module_busy:
1334 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001336 sock->ops = NULL;
1337 module_put(pf->owner);
1338out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001340 return err;
1341
1342out_release:
1343 rcu_read_unlock();
1344 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001346EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347
1348int sock_create(int family, int type, int protocol, struct socket **res)
1349{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001350 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001352EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353
1354int sock_create_kern(int family, int type, int protocol, struct socket **res)
1355{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001356 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001358EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001360SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361{
1362 int retval;
1363 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001364 int flags;
1365
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001366 /* Check the SOCK_* constants for consistency. */
1367 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1368 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1369 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1370 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1371
Ulrich Dreppera677a032008-07-23 21:29:17 -07001372 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001373 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001374 return -EINVAL;
1375 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001377 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1378 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1379
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 retval = sock_create(family, type, protocol, &sock);
1381 if (retval < 0)
1382 goto out;
1383
Ulrich Drepper77d27202008-07-23 21:29:35 -07001384 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 if (retval < 0)
1386 goto out_release;
1387
1388out:
1389 /* It may be already another descriptor 8) Not kernel problem. */
1390 return retval;
1391
1392out_release:
1393 sock_release(sock);
1394 return retval;
1395}
1396
1397/*
1398 * Create a pair of connected sockets.
1399 */
1400
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001401SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1402 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403{
1404 struct socket *sock1, *sock2;
1405 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001406 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001407 int flags;
1408
1409 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001410 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001411 return -EINVAL;
1412 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001414 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1415 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1416
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417 /*
1418 * Obtain the first socket and check if the underlying protocol
1419 * supports the socketpair call.
1420 */
1421
1422 err = sock_create(family, type, protocol, &sock1);
1423 if (err < 0)
1424 goto out;
1425
1426 err = sock_create(family, type, protocol, &sock2);
1427 if (err < 0)
1428 goto out_release_1;
1429
1430 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001431 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 goto out_release_both;
1433
Al Viro28407632012-08-17 23:54:15 -04001434 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001435 if (unlikely(fd1 < 0)) {
1436 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001438 }
Al Viro28407632012-08-17 23:54:15 -04001439 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001440 if (unlikely(fd2 < 0)) {
1441 err = fd2;
Al Viro28407632012-08-17 23:54:15 -04001442 put_unused_fd(fd1);
1443 goto out_release_both;
1444 }
1445
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001446 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001447 if (unlikely(IS_ERR(newfile1))) {
1448 err = PTR_ERR(newfile1);
1449 put_unused_fd(fd1);
1450 put_unused_fd(fd2);
1451 goto out_release_both;
1452 }
1453
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001454 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001455 if (IS_ERR(newfile2)) {
1456 err = PTR_ERR(newfile2);
Al Viro198de4d2009-08-05 19:29:23 +04001457 fput(newfile1);
1458 put_unused_fd(fd1);
Al Viro28407632012-08-17 23:54:15 -04001459 put_unused_fd(fd2);
Al Viro198de4d2009-08-05 19:29:23 +04001460 sock_release(sock2);
1461 goto out;
Al Virodb349502007-02-07 01:48:00 -05001462 }
1463
Al Viro157cf642008-12-14 04:57:47 -05001464 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001465 fd_install(fd1, newfile1);
1466 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 /* fd1 and fd2 may be already another descriptors.
1468 * Not kernel problem.
1469 */
1470
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001471 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 if (!err)
1473 err = put_user(fd2, &usockvec[1]);
1474 if (!err)
1475 return 0;
1476
1477 sys_close(fd2);
1478 sys_close(fd1);
1479 return err;
1480
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001482 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001484 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485out:
1486 return err;
1487}
1488
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489/*
1490 * Bind a name to a socket. Nothing much to do here since it's
1491 * the protocol's responsibility to handle the local address.
1492 *
1493 * We move the socket address to kernel space before we call
1494 * the protocol layer (having also checked the address is ok).
1495 */
1496
Heiko Carstens20f37032009-01-14 14:14:23 +01001497SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498{
1499 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001500 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001501 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001503 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001504 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001505 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001506 if (err >= 0) {
1507 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001508 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001509 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001510 if (!err)
1511 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001512 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001513 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001515 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001516 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 return err;
1518}
1519
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520/*
1521 * Perform a listen. Basically, we allow the protocol to do anything
1522 * necessary for a listen, and if that works, we mark the socket as
1523 * ready for listening.
1524 */
1525
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001526SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527{
1528 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001529 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001530 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001531
1532 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1533 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001534 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001535 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001536 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537
1538 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001539 if (!err)
1540 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001542 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 }
1544 return err;
1545}
1546
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547/*
1548 * For accept, we attempt to create a new socket, set up the link
1549 * with the client, wake up the client, then return the new
1550 * connected fd. We collect the address of the connector in kernel
1551 * space and move it to user at the very end. This is unclean because
1552 * we open the socket then return an error.
1553 *
1554 * 1003.1g adds the ability to recvmsg() to query connection pending
1555 * status to recvmsg. We need to add that support in a way thats
1556 * clean when we restucture accept also.
1557 */
1558
Heiko Carstens20f37032009-01-14 14:14:23 +01001559SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1560 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561{
1562 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001563 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001564 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001565 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566
Ulrich Drepper77d27202008-07-23 21:29:35 -07001567 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001568 return -EINVAL;
1569
1570 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1571 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1572
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001573 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 if (!sock)
1575 goto out;
1576
1577 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001578 newsock = sock_alloc();
1579 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 goto out_put;
1581
1582 newsock->type = sock->type;
1583 newsock->ops = sock->ops;
1584
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 /*
1586 * We don't need try_module_get here, as the listening socket (sock)
1587 * has the protocol module (sock->ops->owner) held.
1588 */
1589 __module_get(newsock->ops->owner);
1590
Al Viro28407632012-08-17 23:54:15 -04001591 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001592 if (unlikely(newfd < 0)) {
1593 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001594 sock_release(newsock);
1595 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001596 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001597 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001598 if (unlikely(IS_ERR(newfile))) {
1599 err = PTR_ERR(newfile);
1600 put_unused_fd(newfd);
1601 sock_release(newsock);
1602 goto out_put;
1603 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001604
Frank Filza79af592005-09-27 15:23:38 -07001605 err = security_socket_accept(sock, newsock);
1606 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001607 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001608
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1610 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001611 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612
1613 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001614 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001615 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001617 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001619 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001620 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001622 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 }
1624
1625 /* File flags are not inherited via accept() unlike another OSes. */
1626
David S. Miller39d8c1b2006-03-20 17:13:49 -08001627 fd_install(newfd, newfile);
1628 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001631 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632out:
1633 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001634out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001635 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001636 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 goto out_put;
1638}
1639
Heiko Carstens20f37032009-01-14 14:14:23 +01001640SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1641 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001642{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001643 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001644}
1645
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646/*
1647 * Attempt to connect to a socket with the server address. The address
1648 * is in user space so we verify it is OK and move it to kernel space.
1649 *
1650 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1651 * break bindings
1652 *
1653 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1654 * other SEQPACKET protocols that take time to connect() as it doesn't
1655 * include the -EINPROGRESS status for such sockets.
1656 */
1657
Heiko Carstens20f37032009-01-14 14:14:23 +01001658SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1659 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660{
1661 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001662 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001663 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001665 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 if (!sock)
1667 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001668 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 if (err < 0)
1670 goto out_put;
1671
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001672 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001673 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 if (err)
1675 goto out_put;
1676
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001677 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 sock->file->f_flags);
1679out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001680 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681out:
1682 return err;
1683}
1684
1685/*
1686 * Get the local address ('name') of a socket object. Move the obtained
1687 * name to user space.
1688 */
1689
Heiko Carstens20f37032009-01-14 14:14:23 +01001690SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1691 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692{
1693 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001694 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001695 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001696
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001697 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 if (!sock)
1699 goto out;
1700
1701 err = security_socket_getsockname(sock);
1702 if (err)
1703 goto out_put;
1704
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001705 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 if (err)
1707 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001708 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709
1710out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001711 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712out:
1713 return err;
1714}
1715
1716/*
1717 * Get the remote address ('name') of a socket object. Move the obtained
1718 * name to user space.
1719 */
1720
Heiko Carstens20f37032009-01-14 14:14:23 +01001721SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1722 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723{
1724 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001725 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001726 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001728 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1729 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 err = security_socket_getpeername(sock);
1731 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001732 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 return err;
1734 }
1735
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001736 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001737 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001738 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001740 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001741 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001742 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 }
1744 return err;
1745}
1746
1747/*
1748 * Send a datagram to a given address. We move the address into kernel
1749 * space and check the user space data area is readable before invoking
1750 * the protocol.
1751 */
1752
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001753SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001754 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001755 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756{
1757 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001758 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 int err;
1760 struct msghdr msg;
1761 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001762 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001763
Linus Torvalds253eacc2010-10-30 16:43:10 -07001764 if (len > INT_MAX)
1765 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001766 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1767 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001768 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001769
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001770 iov.iov_base = buff;
1771 iov.iov_len = len;
1772 msg.msg_name = NULL;
1773 msg.msg_iov = &iov;
1774 msg.msg_iovlen = 1;
1775 msg.msg_control = NULL;
1776 msg.msg_controllen = 0;
1777 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001778 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001779 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 if (err < 0)
1781 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001782 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001783 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 }
1785 if (sock->file->f_flags & O_NONBLOCK)
1786 flags |= MSG_DONTWAIT;
1787 msg.msg_flags = flags;
1788 err = sock_sendmsg(sock, &msg, len);
1789
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001790out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001791 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001792out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 return err;
1794}
1795
1796/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 */
1799
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001800SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001801 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802{
1803 return sys_sendto(fd, buff, len, flags, NULL, 0);
1804}
1805
1806/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001807 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 * sender. We verify the buffers are writable and if needed move the
1809 * sender address from kernel to user space.
1810 */
1811
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001812SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001813 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001814 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815{
1816 struct socket *sock;
1817 struct iovec iov;
1818 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001819 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001820 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001821 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
Linus Torvalds253eacc2010-10-30 16:43:10 -07001823 if (size > INT_MAX)
1824 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001825 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001827 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001829 msg.msg_control = NULL;
1830 msg.msg_controllen = 0;
1831 msg.msg_iovlen = 1;
1832 msg.msg_iov = &iov;
1833 iov.iov_len = size;
1834 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001835 msg.msg_name = (struct sockaddr *)&address;
1836 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 if (sock->file->f_flags & O_NONBLOCK)
1838 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001839 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001841 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001842 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001843 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001844 if (err2 < 0)
1845 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001847
1848 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001849out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 return err;
1851}
1852
1853/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001854 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 */
1856
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001857asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001858 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859{
1860 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1861}
1862
1863/*
1864 * Set a socket option. Because we don't know the option lengths we have
1865 * to pass the user mode parameter for the protocols to sort out.
1866 */
1867
Heiko Carstens20f37032009-01-14 14:14:23 +01001868SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1869 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001871 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 struct socket *sock;
1873
1874 if (optlen < 0)
1875 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001876
1877 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1878 if (sock != NULL) {
1879 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001880 if (err)
1881 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882
1883 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001884 err =
1885 sock_setsockopt(sock, level, optname, optval,
1886 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001888 err =
1889 sock->ops->setsockopt(sock, level, optname, optval,
1890 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001891out_put:
1892 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 }
1894 return err;
1895}
1896
1897/*
1898 * Get a socket option. Because we don't know the option lengths we have
1899 * to pass a user mode parameter for the protocols to sort out.
1900 */
1901
Heiko Carstens20f37032009-01-14 14:14:23 +01001902SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1903 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001905 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 struct socket *sock;
1907
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1909 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001910 err = security_socket_getsockopt(sock, level, optname);
1911 if (err)
1912 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
1914 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001915 err =
1916 sock_getsockopt(sock, level, optname, optval,
1917 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001919 err =
1920 sock->ops->getsockopt(sock, level, optname, optval,
1921 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001922out_put:
1923 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924 }
1925 return err;
1926}
1927
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928/*
1929 * Shutdown a socket.
1930 */
1931
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001932SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001934 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 struct socket *sock;
1936
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001937 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1938 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001940 if (!err)
1941 err = sock->ops->shutdown(sock, how);
1942 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 }
1944 return err;
1945}
1946
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001947/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 * fields which are the same type (int / unsigned) on our platforms.
1949 */
1950#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1951#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1952#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1953
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001954struct used_address {
1955 struct sockaddr_storage name;
1956 unsigned int name_len;
1957};
1958
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001959static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001960 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001961 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001963 struct compat_msghdr __user *msg_compat =
1964 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001965 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001967 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001968 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1969 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001971 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001972
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 err = -EFAULT;
1974 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001975 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001977 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 return -EFAULT;
1979
Anton Blanchard228e5482011-05-02 20:21:35 +00001980 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02001981 err = -EMSGSIZE;
1982 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1983 goto out;
1984 err = -ENOMEM;
1985 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
1986 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00001988 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 }
1990
1991 /* This will also move the address data into kernel space */
1992 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001993 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001995 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001996 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 goto out_freeiov;
1998 total_len = err;
1999
2000 err = -ENOBUFS;
2001
Anton Blanchard228e5482011-05-02 20:21:35 +00002002 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002004 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002006 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002007 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002008 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 if (err)
2010 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002011 ctl_buf = msg_sys->msg_control;
2012 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002014 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002016 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 goto out_freeiov;
2018 }
2019 err = -EFAULT;
2020 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002021 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2023 * checking falls down on this.
2024 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002025 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002026 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002027 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002029 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002031 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032
2033 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002034 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002035 /*
2036 * If this is sendmmsg() and current destination address is same as
2037 * previously succeeded address, omit asking LSM's decision.
2038 * used_address->name_len is initialized to UINT_MAX so that the first
2039 * destination address never matches.
2040 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002041 if (used_address && msg_sys->msg_name &&
2042 used_address->name_len == msg_sys->msg_namelen &&
2043 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002044 used_address->name_len)) {
2045 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2046 goto out_freectl;
2047 }
2048 err = sock_sendmsg(sock, msg_sys, total_len);
2049 /*
2050 * If this is sendmmsg() and sending to current destination address was
2051 * successful, remember it.
2052 */
2053 if (used_address && err >= 0) {
2054 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002055 if (msg_sys->msg_name)
2056 memcpy(&used_address->name, msg_sys->msg_name,
2057 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002058 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059
2060out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002061 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2063out_freeiov:
2064 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002065 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002066out:
2067 return err;
2068}
2069
2070/*
2071 * BSD sendmsg interface
2072 */
2073
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002074long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002075{
2076 int fput_needed, err;
2077 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002078 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002079
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002080 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002081 if (!sock)
2082 goto out;
2083
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002084 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002085
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002086 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002087out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 return err;
2089}
2090
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002091SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2092{
2093 if (flags & MSG_CMSG_COMPAT)
2094 return -EINVAL;
2095 return __sys_sendmsg(fd, msg, flags);
2096}
2097
Anton Blanchard228e5482011-05-02 20:21:35 +00002098/*
2099 * Linux sendmmsg interface
2100 */
2101
2102int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2103 unsigned int flags)
2104{
2105 int fput_needed, err, datagrams;
2106 struct socket *sock;
2107 struct mmsghdr __user *entry;
2108 struct compat_mmsghdr __user *compat_entry;
2109 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002110 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002111
Anton Blanchard98382f42011-08-04 14:07:39 +00002112 if (vlen > UIO_MAXIOV)
2113 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002114
2115 datagrams = 0;
2116
2117 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2118 if (!sock)
2119 return err;
2120
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002121 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002122 entry = mmsg;
2123 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002124 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002125
2126 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002127 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002128 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2129 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002130 if (err < 0)
2131 break;
2132 err = __put_user(err, &compat_entry->msg_len);
2133 ++compat_entry;
2134 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002135 err = ___sys_sendmsg(sock,
2136 (struct msghdr __user *)entry,
2137 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002138 if (err < 0)
2139 break;
2140 err = put_user(err, &entry->msg_len);
2141 ++entry;
2142 }
2143
2144 if (err)
2145 break;
2146 ++datagrams;
2147 }
2148
Anton Blanchard228e5482011-05-02 20:21:35 +00002149 fput_light(sock->file, fput_needed);
2150
Anton Blanchard728ffb82011-08-04 14:07:38 +00002151 /* We only return an error if no datagrams were able to be sent */
2152 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002153 return datagrams;
2154
Anton Blanchard228e5482011-05-02 20:21:35 +00002155 return err;
2156}
2157
2158SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2159 unsigned int, vlen, unsigned int, flags)
2160{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002161 if (flags & MSG_CMSG_COMPAT)
2162 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002163 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2164}
2165
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002166static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002167 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002169 struct compat_msghdr __user *msg_compat =
2170 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002172 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002174 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002175
2176 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002177 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178
2179 /* user mode address pointers */
2180 struct sockaddr __user *uaddr;
2181 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002182
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002184 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002186 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002187 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002189 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002190 err = -EMSGSIZE;
2191 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2192 goto out;
2193 err = -ENOMEM;
2194 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2195 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002197 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 }
2199
2200 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002201 * Save the user-mode address (verify_iovec will change the
2202 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002204
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002205 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206 uaddr_len = COMPAT_NAMELEN(msg);
2207 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002208 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002210 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 if (err < 0)
2212 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002213 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002215 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2216 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002217
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 if (sock->file->f_flags & O_NONBLOCK)
2219 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002220 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2221 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222 if (err < 0)
2223 goto out_freeiov;
2224 len = err;
2225
2226 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002227 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002228 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002229 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230 if (err < 0)
2231 goto out_freeiov;
2232 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002233 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002234 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235 if (err)
2236 goto out_freeiov;
2237 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002238 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239 &msg_compat->msg_controllen);
2240 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002241 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 &msg->msg_controllen);
2243 if (err)
2244 goto out_freeiov;
2245 err = len;
2246
2247out_freeiov:
2248 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002249 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002250out:
2251 return err;
2252}
2253
2254/*
2255 * BSD recvmsg interface
2256 */
2257
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002258long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002259{
2260 int fput_needed, err;
2261 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002262 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002263
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002264 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002265 if (!sock)
2266 goto out;
2267
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002268 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002269
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002270 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271out:
2272 return err;
2273}
2274
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002275SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2276 unsigned int, flags)
2277{
2278 if (flags & MSG_CMSG_COMPAT)
2279 return -EINVAL;
2280 return __sys_recvmsg(fd, msg, flags);
2281}
2282
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002283/*
2284 * Linux recvmmsg interface
2285 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002287int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2288 unsigned int flags, struct timespec *timeout)
2289{
2290 int fput_needed, err, datagrams;
2291 struct socket *sock;
2292 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002293 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002294 struct msghdr msg_sys;
2295 struct timespec end_time;
2296
2297 if (timeout &&
2298 poll_select_set_timeout(&end_time, timeout->tv_sec,
2299 timeout->tv_nsec))
2300 return -EINVAL;
2301
2302 datagrams = 0;
2303
2304 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2305 if (!sock)
2306 return err;
2307
2308 err = sock_error(sock->sk);
2309 if (err)
2310 goto out_put;
2311
2312 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002313 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002314
2315 while (datagrams < vlen) {
2316 /*
2317 * No need to ask LSM for more than the first datagram.
2318 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002319 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002320 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2321 &msg_sys, flags & ~MSG_WAITFORONE,
2322 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002323 if (err < 0)
2324 break;
2325 err = __put_user(err, &compat_entry->msg_len);
2326 ++compat_entry;
2327 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002328 err = ___sys_recvmsg(sock,
2329 (struct msghdr __user *)entry,
2330 &msg_sys, flags & ~MSG_WAITFORONE,
2331 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002332 if (err < 0)
2333 break;
2334 err = put_user(err, &entry->msg_len);
2335 ++entry;
2336 }
2337
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002338 if (err)
2339 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002340 ++datagrams;
2341
Brandon L Black71c5c152010-03-26 16:18:03 +00002342 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2343 if (flags & MSG_WAITFORONE)
2344 flags |= MSG_DONTWAIT;
2345
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002346 if (timeout) {
2347 ktime_get_ts(timeout);
2348 *timeout = timespec_sub(end_time, *timeout);
2349 if (timeout->tv_sec < 0) {
2350 timeout->tv_sec = timeout->tv_nsec = 0;
2351 break;
2352 }
2353
2354 /* Timeout, return less than vlen datagrams */
2355 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2356 break;
2357 }
2358
2359 /* Out of band data, return right away */
2360 if (msg_sys.msg_flags & MSG_OOB)
2361 break;
2362 }
2363
2364out_put:
2365 fput_light(sock->file, fput_needed);
2366
2367 if (err == 0)
2368 return datagrams;
2369
2370 if (datagrams != 0) {
2371 /*
2372 * We may return less entries than requested (vlen) if the
2373 * sock is non block and there aren't enough datagrams...
2374 */
2375 if (err != -EAGAIN) {
2376 /*
2377 * ... or if recvmsg returns an error after we
2378 * received some datagrams, where we record the
2379 * error to return on the next call or if the
2380 * app asks about it using getsockopt(SO_ERROR).
2381 */
2382 sock->sk->sk_err = -err;
2383 }
2384
2385 return datagrams;
2386 }
2387
2388 return err;
2389}
2390
2391SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2392 unsigned int, vlen, unsigned int, flags,
2393 struct timespec __user *, timeout)
2394{
2395 int datagrams;
2396 struct timespec timeout_sys;
2397
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002398 if (flags & MSG_CMSG_COMPAT)
2399 return -EINVAL;
2400
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002401 if (!timeout)
2402 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2403
2404 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2405 return -EFAULT;
2406
2407 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2408
2409 if (datagrams > 0 &&
2410 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2411 datagrams = -EFAULT;
2412
2413 return datagrams;
2414}
2415
2416#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002417/* Argument list sizes for sys_socketcall */
2418#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002419static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002420 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2421 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2422 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002423 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002424};
2425
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426#undef AL
2427
2428/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002429 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430 *
2431 * Argument checking cleaned up. Saved 20% in size.
2432 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002433 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 */
2435
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002436SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437{
Chen Gang2950fa92013-04-07 16:55:23 +08002438 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002439 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002441 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442
Anton Blanchard228e5482011-05-02 20:21:35 +00002443 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444 return -EINVAL;
2445
Arjan van de Ven47379052009-09-28 12:57:44 -07002446 len = nargs[call];
2447 if (len > sizeof(a))
2448 return -EINVAL;
2449
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002451 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002453
Chen Gang2950fa92013-04-07 16:55:23 +08002454 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2455 if (err)
2456 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002457
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002458 a0 = a[0];
2459 a1 = a[1];
2460
2461 switch (call) {
2462 case SYS_SOCKET:
2463 err = sys_socket(a0, a1, a[2]);
2464 break;
2465 case SYS_BIND:
2466 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2467 break;
2468 case SYS_CONNECT:
2469 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2470 break;
2471 case SYS_LISTEN:
2472 err = sys_listen(a0, a1);
2473 break;
2474 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002475 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2476 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002477 break;
2478 case SYS_GETSOCKNAME:
2479 err =
2480 sys_getsockname(a0, (struct sockaddr __user *)a1,
2481 (int __user *)a[2]);
2482 break;
2483 case SYS_GETPEERNAME:
2484 err =
2485 sys_getpeername(a0, (struct sockaddr __user *)a1,
2486 (int __user *)a[2]);
2487 break;
2488 case SYS_SOCKETPAIR:
2489 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2490 break;
2491 case SYS_SEND:
2492 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2493 break;
2494 case SYS_SENDTO:
2495 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2496 (struct sockaddr __user *)a[4], a[5]);
2497 break;
2498 case SYS_RECV:
2499 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2500 break;
2501 case SYS_RECVFROM:
2502 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2503 (struct sockaddr __user *)a[4],
2504 (int __user *)a[5]);
2505 break;
2506 case SYS_SHUTDOWN:
2507 err = sys_shutdown(a0, a1);
2508 break;
2509 case SYS_SETSOCKOPT:
2510 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2511 break;
2512 case SYS_GETSOCKOPT:
2513 err =
2514 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2515 (int __user *)a[4]);
2516 break;
2517 case SYS_SENDMSG:
2518 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2519 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002520 case SYS_SENDMMSG:
2521 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2522 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002523 case SYS_RECVMSG:
2524 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2525 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002526 case SYS_RECVMMSG:
2527 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2528 (struct timespec __user *)a[4]);
2529 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002530 case SYS_ACCEPT4:
2531 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2532 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002533 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002534 default:
2535 err = -EINVAL;
2536 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537 }
2538 return err;
2539}
2540
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002541#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002543/**
2544 * sock_register - add a socket protocol handler
2545 * @ops: description of protocol
2546 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002547 * This function is called by a protocol handler that wants to
2548 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002549 * socket interface. The value ops->family coresponds to the
2550 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002552int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553{
2554 int err;
2555
2556 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002557 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2558 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559 return -ENOBUFS;
2560 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002561
2562 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002563 if (rcu_dereference_protected(net_families[ops->family],
2564 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002565 err = -EEXIST;
2566 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002567 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002568 err = 0;
2569 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002570 spin_unlock(&net_family_lock);
2571
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002572 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573 return err;
2574}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002575EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002577/**
2578 * sock_unregister - remove a protocol handler
2579 * @family: protocol family to remove
2580 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002581 * This function is called by a protocol handler that wants to
2582 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002583 * new socket creation.
2584 *
2585 * If protocol handler is a module, then it can use module reference
2586 * counts to protect against new references. If protocol handler is not
2587 * a module then it needs to provide its own protection in
2588 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002589 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002590void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002591{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002592 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002594 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002595 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002596 spin_unlock(&net_family_lock);
2597
2598 synchronize_rcu();
2599
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002600 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002602EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603
Andi Kleen77d76ea2005-12-22 12:43:42 -08002604static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002605{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002606 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002607 /*
2608 * Initialize the network sysctl infrastructure.
2609 */
2610 err = net_sysctl_init();
2611 if (err)
2612 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002613
Linus Torvalds1da177e2005-04-16 15:20:36 -07002614 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002615 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002616 */
2617 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618
2619 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002620 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002621 */
2622
2623 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002624
2625 err = register_filesystem(&sock_fs_type);
2626 if (err)
2627 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002628 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002629 if (IS_ERR(sock_mnt)) {
2630 err = PTR_ERR(sock_mnt);
2631 goto out_mount;
2632 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002633
2634 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635 */
2636
2637#ifdef CONFIG_NETFILTER
2638 netfilter_init();
2639#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002640
Richard Cochranc1f19b52010-07-17 08:49:36 +00002641#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2642 skb_timestamping_init();
2643#endif
2644
Nick Pigginb3e19d92011-01-07 17:50:11 +11002645out:
2646 return err;
2647
2648out_mount:
2649 unregister_filesystem(&sock_fs_type);
2650out_fs:
2651 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652}
2653
Andi Kleen77d76ea2005-12-22 12:43:42 -08002654core_initcall(sock_init); /* early initcall */
2655
Linus Torvalds1da177e2005-04-16 15:20:36 -07002656#ifdef CONFIG_PROC_FS
2657void socket_seq_show(struct seq_file *seq)
2658{
2659 int cpu;
2660 int counter = 0;
2661
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002662 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002663 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002664
2665 /* It can be negative, by the way. 8) */
2666 if (counter < 0)
2667 counter = 0;
2668
2669 seq_printf(seq, "sockets: used %d\n", counter);
2670}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002671#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002672
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002673#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002674static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002675 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002676{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002677 mm_segment_t old_fs = get_fs();
2678 struct timeval ktv;
2679 int err;
2680
2681 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002682 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002683 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002684 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002685 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002686
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002687 return err;
2688}
2689
Arnd Bergmann6b960182009-11-06 23:10:54 -08002690static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002691 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002692{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002693 mm_segment_t old_fs = get_fs();
2694 struct timespec kts;
2695 int err;
2696
2697 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002698 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002699 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002700 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002701 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002702
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002703 return err;
2704}
2705
Arnd Bergmann6b960182009-11-06 23:10:54 -08002706static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002707{
2708 struct ifreq __user *uifr;
2709 int err;
2710
2711 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002712 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002713 return -EFAULT;
2714
Arnd Bergmann6b960182009-11-06 23:10:54 -08002715 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002716 if (err)
2717 return err;
2718
Arnd Bergmann6b960182009-11-06 23:10:54 -08002719 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002720 return -EFAULT;
2721
2722 return 0;
2723}
2724
Arnd Bergmann6b960182009-11-06 23:10:54 -08002725static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002726{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002727 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002728 struct ifconf ifc;
2729 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002730 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002731 struct ifreq __user *ifr;
2732 unsigned int i, j;
2733 int err;
2734
Arnd Bergmann6b960182009-11-06 23:10:54 -08002735 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002736 return -EFAULT;
2737
Mathias Krause43da5f22012-08-15 11:31:57 +00002738 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002739 if (ifc32.ifcbuf == 0) {
2740 ifc32.ifc_len = 0;
2741 ifc.ifc_len = 0;
2742 ifc.ifc_req = NULL;
2743 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2744 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002745 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2746 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002747 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2748 ifc.ifc_len = len;
2749 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2750 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002751 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002752 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002753 return -EFAULT;
2754 ifr++;
2755 ifr32++;
2756 }
2757 }
2758 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2759 return -EFAULT;
2760
Arnd Bergmann6b960182009-11-06 23:10:54 -08002761 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002762 if (err)
2763 return err;
2764
2765 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2766 return -EFAULT;
2767
2768 ifr = ifc.ifc_req;
2769 ifr32 = compat_ptr(ifc32.ifcbuf);
2770 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002771 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2772 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2773 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002774 return -EFAULT;
2775 ifr32++;
2776 ifr++;
2777 }
2778
2779 if (ifc32.ifcbuf == 0) {
2780 /* Translate from 64-bit structure multiple to
2781 * a 32-bit one.
2782 */
2783 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002784 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002785 ifc32.ifc_len = i;
2786 } else {
2787 ifc32.ifc_len = i;
2788 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002789 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002790 return -EFAULT;
2791
2792 return 0;
2793}
2794
Arnd Bergmann6b960182009-11-06 23:10:54 -08002795static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002796{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002797 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2798 bool convert_in = false, convert_out = false;
2799 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2800 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002801 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002802 u32 rule_cnt = 0, actual_rule_cnt;
2803 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002804 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002805 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002806
2807 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2808 return -EFAULT;
2809
Ben Hutchings3a7da392011-03-17 07:34:32 +00002810 compat_rxnfc = compat_ptr(data);
2811
2812 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002813 return -EFAULT;
2814
Ben Hutchings3a7da392011-03-17 07:34:32 +00002815 /* Most ethtool structures are defined without padding.
2816 * Unfortunately struct ethtool_rxnfc is an exception.
2817 */
2818 switch (ethcmd) {
2819 default:
2820 break;
2821 case ETHTOOL_GRXCLSRLALL:
2822 /* Buffer size is variable */
2823 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2824 return -EFAULT;
2825 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2826 return -ENOMEM;
2827 buf_size += rule_cnt * sizeof(u32);
2828 /* fall through */
2829 case ETHTOOL_GRXRINGS:
2830 case ETHTOOL_GRXCLSRLCNT:
2831 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002832 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002833 convert_out = true;
2834 /* fall through */
2835 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002836 buf_size += sizeof(struct ethtool_rxnfc);
2837 convert_in = true;
2838 break;
2839 }
2840
2841 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002842 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002843
2844 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2845 return -EFAULT;
2846
2847 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2848 &ifr->ifr_ifru.ifru_data))
2849 return -EFAULT;
2850
2851 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002852 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002853 * fs.ring_cookie and at the end of fs, but nowhere else.
2854 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002855 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2856 sizeof(compat_rxnfc->fs.m_ext) !=
2857 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2858 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002859 BUILD_BUG_ON(
2860 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2861 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2862 offsetof(struct ethtool_rxnfc, fs.location) -
2863 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2864
2865 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002866 (void __user *)(&rxnfc->fs.m_ext + 1) -
2867 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002868 copy_in_user(&rxnfc->fs.ring_cookie,
2869 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002870 (void __user *)(&rxnfc->fs.location + 1) -
2871 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002872 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2873 sizeof(rxnfc->rule_cnt)))
2874 return -EFAULT;
2875 }
2876
2877 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2878 if (ret)
2879 return ret;
2880
2881 if (convert_out) {
2882 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002883 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2884 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002885 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2886 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002887 (const void __user *)(&rxnfc->fs.location + 1) -
2888 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002889 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2890 sizeof(rxnfc->rule_cnt)))
2891 return -EFAULT;
2892
2893 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2894 /* As an optimisation, we only copy the actual
2895 * number of rules that the underlying
2896 * function returned. Since Mallory might
2897 * change the rule count in user memory, we
2898 * check that it is less than the rule count
2899 * originally given (as the user buffer size),
2900 * which has been range-checked.
2901 */
2902 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2903 return -EFAULT;
2904 if (actual_rule_cnt < rule_cnt)
2905 rule_cnt = actual_rule_cnt;
2906 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2907 &rxnfc->rule_locs[0],
2908 rule_cnt * sizeof(u32)))
2909 return -EFAULT;
2910 }
2911 }
2912
2913 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002914}
2915
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002916static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2917{
2918 void __user *uptr;
2919 compat_uptr_t uptr32;
2920 struct ifreq __user *uifr;
2921
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002922 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002923 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2924 return -EFAULT;
2925
2926 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2927 return -EFAULT;
2928
2929 uptr = compat_ptr(uptr32);
2930
2931 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2932 return -EFAULT;
2933
2934 return dev_ioctl(net, SIOCWANDEV, uifr);
2935}
2936
Arnd Bergmann6b960182009-11-06 23:10:54 -08002937static int bond_ioctl(struct net *net, unsigned int cmd,
2938 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002939{
2940 struct ifreq kifr;
2941 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002942 mm_segment_t old_fs;
2943 int err;
2944 u32 data;
2945 void __user *datap;
2946
2947 switch (cmd) {
2948 case SIOCBONDENSLAVE:
2949 case SIOCBONDRELEASE:
2950 case SIOCBONDSETHWADDR:
2951 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002952 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002953 return -EFAULT;
2954
2955 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002956 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002957 err = dev_ioctl(net, cmd,
2958 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002959 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002960
2961 return err;
2962 case SIOCBONDSLAVEINFOQUERY:
2963 case SIOCBONDINFOQUERY:
2964 uifr = compat_alloc_user_space(sizeof(*uifr));
2965 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2966 return -EFAULT;
2967
2968 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2969 return -EFAULT;
2970
2971 datap = compat_ptr(data);
2972 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2973 return -EFAULT;
2974
Arnd Bergmann6b960182009-11-06 23:10:54 -08002975 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002976 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002977 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002978 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002979}
2980
Arnd Bergmann6b960182009-11-06 23:10:54 -08002981static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2982 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002983{
2984 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002985 char tmp_buf[IFNAMSIZ];
2986 void __user *data64;
2987 u32 data32;
2988
2989 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2990 IFNAMSIZ))
2991 return -EFAULT;
2992 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2993 return -EFAULT;
2994 data64 = compat_ptr(data32);
2995
2996 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2997
2998 /* Don't check these user accesses, just let that get trapped
2999 * in the ioctl handler instead.
3000 */
3001 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3002 IFNAMSIZ))
3003 return -EFAULT;
3004 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3005 return -EFAULT;
3006
Arnd Bergmann6b960182009-11-06 23:10:54 -08003007 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003008}
3009
Arnd Bergmann6b960182009-11-06 23:10:54 -08003010static int dev_ifsioc(struct net *net, struct socket *sock,
3011 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003012{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003013 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003014 int err;
3015
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003016 uifr = compat_alloc_user_space(sizeof(*uifr));
3017 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3018 return -EFAULT;
3019
3020 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3021
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003022 if (!err) {
3023 switch (cmd) {
3024 case SIOCGIFFLAGS:
3025 case SIOCGIFMETRIC:
3026 case SIOCGIFMTU:
3027 case SIOCGIFMEM:
3028 case SIOCGIFHWADDR:
3029 case SIOCGIFINDEX:
3030 case SIOCGIFADDR:
3031 case SIOCGIFBRDADDR:
3032 case SIOCGIFDSTADDR:
3033 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003034 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003035 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003036 case SIOCGMIIPHY:
3037 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003038 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003039 err = -EFAULT;
3040 break;
3041 }
3042 }
3043 return err;
3044}
3045
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003046static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3047 struct compat_ifreq __user *uifr32)
3048{
3049 struct ifreq ifr;
3050 struct compat_ifmap __user *uifmap32;
3051 mm_segment_t old_fs;
3052 int err;
3053
3054 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3055 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3056 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3057 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3058 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3059 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3060 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3061 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3062 if (err)
3063 return -EFAULT;
3064
3065 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003066 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003067 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003068 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003069
3070 if (cmd == SIOCGIFMAP && !err) {
3071 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3072 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3073 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3074 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3075 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3076 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3077 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3078 if (err)
3079 err = -EFAULT;
3080 }
3081 return err;
3082}
3083
3084static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3085{
3086 void __user *uptr;
3087 compat_uptr_t uptr32;
3088 struct ifreq __user *uifr;
3089
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003090 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003091 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3092 return -EFAULT;
3093
3094 if (get_user(uptr32, &uifr32->ifr_data))
3095 return -EFAULT;
3096
3097 uptr = compat_ptr(uptr32);
3098
3099 if (put_user(uptr, &uifr->ifr_data))
3100 return -EFAULT;
3101
3102 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3103}
3104
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003105struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003106 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003107 struct sockaddr rt_dst; /* target address */
3108 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3109 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003110 unsigned short rt_flags;
3111 short rt_pad2;
3112 u32 rt_pad3;
3113 unsigned char rt_tos;
3114 unsigned char rt_class;
3115 short rt_pad4;
3116 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003117 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003118 u32 rt_mtu; /* per route MTU/Window */
3119 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003120 unsigned short rt_irtt; /* Initial RTT */
3121};
3122
3123struct in6_rtmsg32 {
3124 struct in6_addr rtmsg_dst;
3125 struct in6_addr rtmsg_src;
3126 struct in6_addr rtmsg_gateway;
3127 u32 rtmsg_type;
3128 u16 rtmsg_dst_len;
3129 u16 rtmsg_src_len;
3130 u32 rtmsg_metric;
3131 u32 rtmsg_info;
3132 u32 rtmsg_flags;
3133 s32 rtmsg_ifindex;
3134};
3135
Arnd Bergmann6b960182009-11-06 23:10:54 -08003136static int routing_ioctl(struct net *net, struct socket *sock,
3137 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003138{
3139 int ret;
3140 void *r = NULL;
3141 struct in6_rtmsg r6;
3142 struct rtentry r4;
3143 char devname[16];
3144 u32 rtdev;
3145 mm_segment_t old_fs = get_fs();
3146
Arnd Bergmann6b960182009-11-06 23:10:54 -08003147 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3148 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003149 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003150 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003151 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3152 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3153 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3154 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3155 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3156 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3157 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003158
3159 r = (void *) &r6;
3160 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003161 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003162 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003163 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003164 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3165 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3166 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3167 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3168 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3169 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003170 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003171 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003172 r4.rt_dev = (char __user __force *)devname;
3173 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003174 } else
3175 r4.rt_dev = NULL;
3176
3177 r = (void *) &r4;
3178 }
3179
3180 if (ret) {
3181 ret = -EFAULT;
3182 goto out;
3183 }
3184
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003185 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003186 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003187 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003188
3189out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003190 return ret;
3191}
3192
3193/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3194 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003195 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003196 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003197static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003198{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003199 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003200
Arnd Bergmann6b960182009-11-06 23:10:54 -08003201 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003202 return -EFAULT;
3203 if (tmp == BRCTL_GET_VERSION)
3204 return BRCTL_VERSION + 1;
3205 return -EINVAL;
3206}
3207
Arnd Bergmann6b960182009-11-06 23:10:54 -08003208static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3209 unsigned int cmd, unsigned long arg)
3210{
3211 void __user *argp = compat_ptr(arg);
3212 struct sock *sk = sock->sk;
3213 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003214
Arnd Bergmann6b960182009-11-06 23:10:54 -08003215 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3216 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003217
Arnd Bergmann6b960182009-11-06 23:10:54 -08003218 switch (cmd) {
3219 case SIOCSIFBR:
3220 case SIOCGIFBR:
3221 return old_bridge_ioctl(argp);
3222 case SIOCGIFNAME:
3223 return dev_ifname32(net, argp);
3224 case SIOCGIFCONF:
3225 return dev_ifconf(net, argp);
3226 case SIOCETHTOOL:
3227 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003228 case SIOCWANDEV:
3229 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003230 case SIOCGIFMAP:
3231 case SIOCSIFMAP:
3232 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003233 case SIOCBONDENSLAVE:
3234 case SIOCBONDRELEASE:
3235 case SIOCBONDSETHWADDR:
3236 case SIOCBONDSLAVEINFOQUERY:
3237 case SIOCBONDINFOQUERY:
3238 case SIOCBONDCHANGEACTIVE:
3239 return bond_ioctl(net, cmd, argp);
3240 case SIOCADDRT:
3241 case SIOCDELRT:
3242 return routing_ioctl(net, sock, cmd, argp);
3243 case SIOCGSTAMP:
3244 return do_siocgstamp(net, sock, cmd, argp);
3245 case SIOCGSTAMPNS:
3246 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003247 case SIOCSHWTSTAMP:
3248 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003249
Arnd Bergmann6b960182009-11-06 23:10:54 -08003250 case FIOSETOWN:
3251 case SIOCSPGRP:
3252 case FIOGETOWN:
3253 case SIOCGPGRP:
3254 case SIOCBRADDBR:
3255 case SIOCBRDELBR:
3256 case SIOCGIFVLAN:
3257 case SIOCSIFVLAN:
3258 case SIOCADDDLCI:
3259 case SIOCDELDLCI:
3260 return sock_ioctl(file, cmd, arg);
3261
3262 case SIOCGIFFLAGS:
3263 case SIOCSIFFLAGS:
3264 case SIOCGIFMETRIC:
3265 case SIOCSIFMETRIC:
3266 case SIOCGIFMTU:
3267 case SIOCSIFMTU:
3268 case SIOCGIFMEM:
3269 case SIOCSIFMEM:
3270 case SIOCGIFHWADDR:
3271 case SIOCSIFHWADDR:
3272 case SIOCADDMULTI:
3273 case SIOCDELMULTI:
3274 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003275 case SIOCGIFADDR:
3276 case SIOCSIFADDR:
3277 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003278 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003279 case SIOCGIFBRDADDR:
3280 case SIOCSIFBRDADDR:
3281 case SIOCGIFDSTADDR:
3282 case SIOCSIFDSTADDR:
3283 case SIOCGIFNETMASK:
3284 case SIOCSIFNETMASK:
3285 case SIOCSIFPFLAGS:
3286 case SIOCGIFPFLAGS:
3287 case SIOCGIFTXQLEN:
3288 case SIOCSIFTXQLEN:
3289 case SIOCBRADDIF:
3290 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003291 case SIOCSIFNAME:
3292 case SIOCGMIIPHY:
3293 case SIOCGMIIREG:
3294 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003295 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003296
Arnd Bergmann6b960182009-11-06 23:10:54 -08003297 case SIOCSARP:
3298 case SIOCGARP:
3299 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003300 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003301 return sock_do_ioctl(net, sock, cmd, arg);
3302 }
3303
Arnd Bergmann6b960182009-11-06 23:10:54 -08003304 return -ENOIOCTLCMD;
3305}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003306
Eric Dumazet95c96172012-04-15 05:58:06 +00003307static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003308 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003309{
3310 struct socket *sock = file->private_data;
3311 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003312 struct sock *sk;
3313 struct net *net;
3314
3315 sk = sock->sk;
3316 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003317
3318 if (sock->ops->compat_ioctl)
3319 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3320
David S. Miller87de87d2008-06-03 09:14:03 -07003321 if (ret == -ENOIOCTLCMD &&
3322 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3323 ret = compat_wext_handle_ioctl(net, cmd, arg);
3324
Arnd Bergmann6b960182009-11-06 23:10:54 -08003325 if (ret == -ENOIOCTLCMD)
3326 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3327
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003328 return ret;
3329}
3330#endif
3331
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003332int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3333{
3334 return sock->ops->bind(sock, addr, addrlen);
3335}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003336EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003337
3338int kernel_listen(struct socket *sock, int backlog)
3339{
3340 return sock->ops->listen(sock, backlog);
3341}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003342EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003343
3344int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3345{
3346 struct sock *sk = sock->sk;
3347 int err;
3348
3349 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3350 newsock);
3351 if (err < 0)
3352 goto done;
3353
3354 err = sock->ops->accept(sock, *newsock, flags);
3355 if (err < 0) {
3356 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003357 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003358 goto done;
3359 }
3360
3361 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003362 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003363
3364done:
3365 return err;
3366}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003367EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003368
3369int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003370 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003371{
3372 return sock->ops->connect(sock, addr, addrlen, flags);
3373}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003374EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003375
3376int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3377 int *addrlen)
3378{
3379 return sock->ops->getname(sock, addr, addrlen, 0);
3380}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003381EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003382
3383int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3384 int *addrlen)
3385{
3386 return sock->ops->getname(sock, addr, addrlen, 1);
3387}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003388EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003389
3390int kernel_getsockopt(struct socket *sock, int level, int optname,
3391 char *optval, int *optlen)
3392{
3393 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003394 char __user *uoptval;
3395 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003396 int err;
3397
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003398 uoptval = (char __user __force *) optval;
3399 uoptlen = (int __user __force *) optlen;
3400
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003401 set_fs(KERNEL_DS);
3402 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003403 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003404 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003405 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3406 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003407 set_fs(oldfs);
3408 return err;
3409}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003410EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003411
3412int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003413 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003414{
3415 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003416 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003417 int err;
3418
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003419 uoptval = (char __user __force *) optval;
3420
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003421 set_fs(KERNEL_DS);
3422 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003423 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003424 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003425 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003426 optlen);
3427 set_fs(oldfs);
3428 return err;
3429}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003430EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003431
3432int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3433 size_t size, int flags)
3434{
3435 if (sock->ops->sendpage)
3436 return sock->ops->sendpage(sock, page, offset, size, flags);
3437
3438 return sock_no_sendpage(sock, page, offset, size, flags);
3439}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003440EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003441
3442int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3443{
3444 mm_segment_t oldfs = get_fs();
3445 int err;
3446
3447 set_fs(KERNEL_DS);
3448 err = sock->ops->ioctl(sock, cmd, arg);
3449 set_fs(oldfs);
3450
3451 return err;
3452}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003453EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003454
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003455int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3456{
3457 return sock->ops->shutdown(sock, how);
3458}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003459EXPORT_SYMBOL(kernel_sock_shutdown);