blob: 88f759adf3afba28af3b1a694220c18f5e895616 [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
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000603int 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 return 0;
613}
614EXPORT_SYMBOL(sock_tx_timestamp);
615
Anton Blanchard228e5482011-05-02 20:21:35 +0000616static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
617 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618{
619 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620
621 si->sock = sock;
622 si->scm = NULL;
623 si->msg = msg;
624 si->size = size;
625
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 return sock->ops->sendmsg(iocb, sock, msg, size);
627}
628
Anton Blanchard228e5482011-05-02 20:21:35 +0000629static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
630 struct msghdr *msg, size_t size)
631{
632 int err = security_socket_sendmsg(sock, msg, size);
633
634 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
635}
636
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
638{
639 struct kiocb iocb;
640 struct sock_iocb siocb;
641 int ret;
642
643 init_sync_kiocb(&iocb, NULL);
644 iocb.private = &siocb;
645 ret = __sock_sendmsg(&iocb, sock, msg, size);
646 if (-EIOCBQUEUED == ret)
647 ret = wait_on_sync_kiocb(&iocb);
648 return ret;
649}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700650EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
Eric Dumazet894dc242011-07-26 02:39:41 +0000652static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000653{
654 struct kiocb iocb;
655 struct sock_iocb siocb;
656 int ret;
657
658 init_sync_kiocb(&iocb, NULL);
659 iocb.private = &siocb;
660 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
661 if (-EIOCBQUEUED == ret)
662 ret = wait_on_sync_kiocb(&iocb);
663 return ret;
664}
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
667 struct kvec *vec, size_t num, size_t size)
668{
669 mm_segment_t oldfs = get_fs();
670 int result;
671
672 set_fs(KERNEL_DS);
673 /*
674 * the following is safe, since for compiler definitions of kvec and
675 * iovec are identical, yielding the same in-core layout and alignment
676 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700677 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 msg->msg_iovlen = num;
679 result = sock_sendmsg(sock, msg, size);
680 set_fs(oldfs);
681 return result;
682}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700683EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
Patrick Ohly20d49472009-02-12 05:03:38 +0000685static int ktime2ts(ktime_t kt, struct timespec *ts)
686{
687 if (kt.tv64) {
688 *ts = ktime_to_timespec(kt);
689 return 1;
690 } else {
691 return 0;
692 }
693}
694
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700695/*
696 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
697 */
698void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
699 struct sk_buff *skb)
700{
Patrick Ohly20d49472009-02-12 05:03:38 +0000701 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
702 struct timespec ts[3];
703 int empty = 1;
704 struct skb_shared_hwtstamps *shhwtstamps =
705 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700706
Patrick Ohly20d49472009-02-12 05:03:38 +0000707 /* Race occurred between timestamp enabling and packet
708 receiving. Fill in the current time for now. */
709 if (need_software_tstamp && skb->tstamp.tv64 == 0)
710 __net_timestamp(skb);
711
712 if (need_software_tstamp) {
713 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
714 struct timeval tv;
715 skb_get_timestamp(skb, &tv);
716 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
717 sizeof(tv), &tv);
718 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000719 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000720 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000721 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000722 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700723 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000724
725
726 memset(ts, 0, sizeof(ts));
727 if (skb->tstamp.tv64 &&
728 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
729 skb_get_timestampns(skb, ts + 0);
730 empty = 0;
731 }
732 if (shhwtstamps) {
733 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
734 ktime2ts(shhwtstamps->syststamp, ts + 1))
735 empty = 0;
736 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
737 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
738 empty = 0;
739 }
740 if (!empty)
741 put_cmsg(msg, SOL_SOCKET,
742 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700743}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300744EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
745
Johannes Berg6e3e9392011-11-09 10:15:42 +0100746void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
747 struct sk_buff *skb)
748{
749 int ack;
750
751 if (!sock_flag(sk, SOCK_WIFI_STATUS))
752 return;
753 if (!skb->wifi_acked_valid)
754 return;
755
756 ack = skb->wifi_acked;
757
758 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
759}
760EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
761
stephen hemminger11165f12010-10-18 14:27:29 +0000762static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
763 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700764{
765 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
766 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
767 sizeof(__u32), &skb->dropcount);
768}
769
Eric Dumazet767dd032010-04-28 19:14:43 +0000770void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700771 struct sk_buff *skb)
772{
773 sock_recv_timestamp(msg, sk, skb);
774 sock_recv_drops(msg, sk, skb);
775}
Eric Dumazet767dd032010-04-28 19:14:43 +0000776EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700777
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700778static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
779 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 struct sock_iocb *si = kiocb_to_siocb(iocb);
782
783 si->sock = sock;
784 si->scm = NULL;
785 si->msg = msg;
786 si->size = size;
787 si->flags = flags;
788
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
790}
791
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700792static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
793 struct msghdr *msg, size_t size, int flags)
794{
795 int err = security_socket_recvmsg(sock, msg, size, flags);
796
797 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
798}
799
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700800int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 size_t size, int flags)
802{
803 struct kiocb iocb;
804 struct sock_iocb siocb;
805 int ret;
806
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700807 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 iocb.private = &siocb;
809 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
810 if (-EIOCBQUEUED == ret)
811 ret = wait_on_sync_kiocb(&iocb);
812 return ret;
813}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700814EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700816static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
817 size_t size, int flags)
818{
819 struct kiocb iocb;
820 struct sock_iocb siocb;
821 int ret;
822
823 init_sync_kiocb(&iocb, NULL);
824 iocb.private = &siocb;
825 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
826 if (-EIOCBQUEUED == ret)
827 ret = wait_on_sync_kiocb(&iocb);
828 return ret;
829}
830
Martin Lucinac1249c02010-12-10 00:04:05 +0000831/**
832 * kernel_recvmsg - Receive a message from a socket (kernel space)
833 * @sock: The socket to receive the message from
834 * @msg: Received message
835 * @vec: Input s/g array for message data
836 * @num: Size of input s/g array
837 * @size: Number of bytes to read
838 * @flags: Message flags (MSG_DONTWAIT, etc...)
839 *
840 * On return the msg structure contains the scatter/gather array passed in the
841 * vec argument. The array is modified so that it consists of the unfilled
842 * portion of the original array.
843 *
844 * The returned value is the total number of bytes received, or an error.
845 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700846int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
847 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
849 mm_segment_t oldfs = get_fs();
850 int result;
851
852 set_fs(KERNEL_DS);
853 /*
854 * the following is safe, since for compiler definitions of kvec and
855 * iovec are identical, yielding the same in-core layout and alignment
856 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700857 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 result = sock_recvmsg(sock, msg, size, flags);
859 set_fs(oldfs);
860 return result;
861}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700862EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
864static void sock_aio_dtor(struct kiocb *iocb)
865{
866 kfree(iocb->private);
867}
868
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300869static ssize_t sock_sendpage(struct file *file, struct page *page,
870 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871{
872 struct socket *sock;
873 int flags;
874
Eric Dumazetb69aee02005-09-06 14:42:45 -0700875 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876
Eric Dumazet35f9c092012-04-05 03:05:35 +0000877 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
878 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
879 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880
Linus Torvaldse6949582009-08-13 08:28:36 -0700881 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882}
883
Jens Axboe9c55e012007-11-06 23:30:13 -0800884static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700885 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800886 unsigned int flags)
887{
888 struct socket *sock = file->private_data;
889
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800890 if (unlikely(!sock->ops->splice_read))
891 return -EINVAL;
892
Jens Axboe9c55e012007-11-06 23:30:13 -0800893 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
894}
895
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800896static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700897 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800898{
899 if (!is_sync_kiocb(iocb)) {
900 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
901 if (!siocb)
902 return NULL;
903 iocb->ki_dtor = sock_aio_dtor;
904 }
905
906 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800907 iocb->private = siocb;
908 return siocb;
909}
910
911static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700912 struct file *file, const struct iovec *iov,
913 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800914{
915 struct socket *sock = file->private_data;
916 size_t size = 0;
917 int i;
918
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700919 for (i = 0; i < nr_segs; i++)
920 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800921
922 msg->msg_name = NULL;
923 msg->msg_namelen = 0;
924 msg->msg_control = NULL;
925 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700926 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800927 msg->msg_iovlen = nr_segs;
928 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
929
930 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
931}
932
Badari Pulavarty027445c2006-09-30 23:28:46 -0700933static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
934 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800935{
936 struct sock_iocb siocb, *x;
937
938 if (pos != 0)
939 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700940
941 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800942 return 0;
943
Badari Pulavarty027445c2006-09-30 23:28:46 -0700944
945 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800946 if (!x)
947 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700948 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800949}
950
951static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700952 struct file *file, const struct iovec *iov,
953 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800954{
955 struct socket *sock = file->private_data;
956 size_t size = 0;
957 int i;
958
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700959 for (i = 0; i < nr_segs; i++)
960 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800961
962 msg->msg_name = NULL;
963 msg->msg_namelen = 0;
964 msg->msg_control = NULL;
965 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700966 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800967 msg->msg_iovlen = nr_segs;
968 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
969 if (sock->type == SOCK_SEQPACKET)
970 msg->msg_flags |= MSG_EOR;
971
972 return __sock_sendmsg(iocb, sock, msg, size);
973}
974
Badari Pulavarty027445c2006-09-30 23:28:46 -0700975static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
976 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800978 struct sock_iocb siocb, *x;
979
980 if (pos != 0)
981 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700982
Badari Pulavarty027445c2006-09-30 23:28:46 -0700983 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800984 if (!x)
985 return -ENOMEM;
986
Badari Pulavarty027445c2006-09-30 23:28:46 -0700987 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988}
989
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990/*
991 * Atomic setting of ioctl hooks to avoid race
992 * with module unload.
993 */
994
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800995static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700996static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997
Eric W. Biederman881d9662007-09-17 11:56:21 -0700998void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001000 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001002 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003}
1004EXPORT_SYMBOL(brioctl_set);
1005
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001006static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001007static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008
Eric W. Biederman881d9662007-09-17 11:56:21 -07001009void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001011 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001013 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014}
1015EXPORT_SYMBOL(vlan_ioctl_set);
1016
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001017static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001018static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001020void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001022 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001024 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025}
1026EXPORT_SYMBOL(dlci_ioctl_set);
1027
Arnd Bergmann6b960182009-11-06 23:10:54 -08001028static long sock_do_ioctl(struct net *net, struct socket *sock,
1029 unsigned int cmd, unsigned long arg)
1030{
1031 int err;
1032 void __user *argp = (void __user *)arg;
1033
1034 err = sock->ops->ioctl(sock, cmd, arg);
1035
1036 /*
1037 * If this ioctl is unknown try to hand it down
1038 * to the NIC driver.
1039 */
1040 if (err == -ENOIOCTLCMD)
1041 err = dev_ioctl(net, cmd, argp);
1042
1043 return err;
1044}
1045
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046/*
1047 * With an ioctl, arg may well be a user mode pointer, but we don't know
1048 * what to do with it - that's up to the protocol still.
1049 */
1050
1051static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1052{
1053 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001054 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 void __user *argp = (void __user *)arg;
1056 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001057 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
Eric Dumazetb69aee02005-09-06 14:42:45 -07001059 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001060 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001061 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001063 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001065#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001067 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001069#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001070 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 case FIOSETOWN:
1072 case SIOCSPGRP:
1073 err = -EFAULT;
1074 if (get_user(pid, (int __user *)argp))
1075 break;
1076 err = f_setown(sock->file, pid, 1);
1077 break;
1078 case FIOGETOWN:
1079 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001080 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001081 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 break;
1083 case SIOCGIFBR:
1084 case SIOCSIFBR:
1085 case SIOCBRADDBR:
1086 case SIOCBRDELBR:
1087 err = -ENOPKG;
1088 if (!br_ioctl_hook)
1089 request_module("bridge");
1090
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001091 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001092 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001093 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001094 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 break;
1096 case SIOCGIFVLAN:
1097 case SIOCSIFVLAN:
1098 err = -ENOPKG;
1099 if (!vlan_ioctl_hook)
1100 request_module("8021q");
1101
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001102 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001104 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001105 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 case SIOCADDDLCI:
1108 case SIOCDELDLCI:
1109 err = -ENOPKG;
1110 if (!dlci_ioctl_hook)
1111 request_module("dlci");
1112
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001113 mutex_lock(&dlci_ioctl_mutex);
1114 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001116 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 break;
1118 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001119 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001121 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 return err;
1123}
1124
1125int sock_create_lite(int family, int type, int protocol, struct socket **res)
1126{
1127 int err;
1128 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 err = security_socket_create(family, type, protocol, 1);
1131 if (err)
1132 goto out;
1133
1134 sock = sock_alloc();
1135 if (!sock) {
1136 err = -ENOMEM;
1137 goto out;
1138 }
1139
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001141 err = security_socket_post_create(sock, family, type, protocol, 1);
1142 if (err)
1143 goto out_release;
1144
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145out:
1146 *res = sock;
1147 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001148out_release:
1149 sock_release(sock);
1150 sock = NULL;
1151 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001153EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
1155/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001156static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157{
1158 struct socket *sock;
1159
1160 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001161 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001163 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 return sock->ops->poll(file, sock, wait);
1165}
1166
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001167static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001169 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170
1171 return sock->ops->mmap(file, sock, vma);
1172}
1173
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001174static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175{
1176 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001177 * It was possible the inode is NULL we were
1178 * closing an unfinished socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 */
1180
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001181 if (!inode) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 printk(KERN_DEBUG "sock_close: NULL inode\n");
1183 return 0;
1184 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 sock_release(SOCKET_I(inode));
1186 return 0;
1187}
1188
1189/*
1190 * Update the socket async list
1191 *
1192 * Fasync_list locking strategy.
1193 *
1194 * 1. fasync_list is modified only under process context socket lock
1195 * i.e. under semaphore.
1196 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001197 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 */
1199
1200static int sock_fasync(int fd, struct file *filp, int on)
1201{
Eric Dumazet989a2972010-04-14 09:55:35 +00001202 struct socket *sock = filp->private_data;
1203 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001204 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205
Eric Dumazet989a2972010-04-14 09:55:35 +00001206 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208
1209 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001210 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1211 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212
Eric Dumazeteaefd112011-02-18 03:26:36 +00001213 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001214 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001215 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001216 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217
Eric Dumazet989a2972010-04-14 09:55:35 +00001218 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 return 0;
1220}
1221
Eric Dumazet43815482010-04-29 11:01:49 +00001222/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223
1224int sock_wake_async(struct socket *sock, int how, int band)
1225{
Eric Dumazet43815482010-04-29 11:01:49 +00001226 struct socket_wq *wq;
1227
1228 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001230 rcu_read_lock();
1231 wq = rcu_dereference(sock->wq);
1232 if (!wq || !wq->fasync_list) {
1233 rcu_read_unlock();
1234 return -1;
1235 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001236 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001237 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1239 break;
1240 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001241 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1243 break;
1244 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001245 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001246call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001247 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001249 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001250 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 }
Eric Dumazet43815482010-04-29 11:01:49 +00001252 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 return 0;
1254}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001255EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256
Pavel Emelyanov721db932010-09-29 16:06:32 +04001257int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001258 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259{
1260 int err;
1261 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001262 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263
1264 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001265 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 */
1267 if (family < 0 || family >= NPROTO)
1268 return -EAFNOSUPPORT;
1269 if (type < 0 || type >= SOCK_MAX)
1270 return -EINVAL;
1271
1272 /* Compatibility.
1273
1274 This uglymoron is moved from INET layer to here to avoid
1275 deadlock in module load.
1276 */
1277 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001278 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 if (!warned) {
1280 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001281 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1282 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 }
1284 family = PF_PACKET;
1285 }
1286
1287 err = security_socket_create(family, type, protocol, kern);
1288 if (err)
1289 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001290
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001291 /*
1292 * Allocate the socket and allow the family to set things up. if
1293 * the protocol is 0, the family is instructed to select an appropriate
1294 * default.
1295 */
1296 sock = sock_alloc();
1297 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001298 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001299 return -ENFILE; /* Not exactly a match, but its the
1300 closest posix thing */
1301 }
1302
1303 sock->type = type;
1304
Johannes Berg95a5afc2008-10-16 15:24:51 -07001305#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001306 /* Attempt to load a protocol module if the find failed.
1307 *
1308 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 * requested real, full-featured networking support upon configuration.
1310 * Otherwise module support will break!
1311 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001312 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001313 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314#endif
1315
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001316 rcu_read_lock();
1317 pf = rcu_dereference(net_families[family]);
1318 err = -EAFNOSUPPORT;
1319 if (!pf)
1320 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
1322 /*
1323 * We will call the ->create function, that possibly is in a loadable
1324 * module, so we have to bump that loadable module refcnt first.
1325 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001326 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 goto out_release;
1328
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001329 /* Now protected by module ref count */
1330 rcu_read_unlock();
1331
Eric Paris3f378b62009-11-05 22:18:14 -08001332 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001333 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001335
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 /*
1337 * Now to bump the refcnt of the [loadable] module that owns this
1338 * socket at sock_release time we decrement its refcnt.
1339 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001340 if (!try_module_get(sock->ops->owner))
1341 goto out_module_busy;
1342
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 /*
1344 * Now that we're done with the ->create function, the [loadable]
1345 * module can have its refcnt decremented
1346 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001347 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001348 err = security_socket_post_create(sock, family, type, protocol, kern);
1349 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001350 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001351 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001353 return 0;
1354
1355out_module_busy:
1356 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001358 sock->ops = NULL;
1359 module_put(pf->owner);
1360out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001362 return err;
1363
1364out_release:
1365 rcu_read_unlock();
1366 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001368EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369
1370int sock_create(int family, int type, int protocol, struct socket **res)
1371{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001372 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001374EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
1376int sock_create_kern(int family, int type, int protocol, struct socket **res)
1377{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001378 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001380EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001382SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383{
1384 int retval;
1385 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001386 int flags;
1387
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001388 /* Check the SOCK_* constants for consistency. */
1389 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1390 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1391 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1392 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1393
Ulrich Dreppera677a032008-07-23 21:29:17 -07001394 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001395 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001396 return -EINVAL;
1397 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001399 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1400 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1401
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 retval = sock_create(family, type, protocol, &sock);
1403 if (retval < 0)
1404 goto out;
1405
Ulrich Drepper77d27202008-07-23 21:29:35 -07001406 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 if (retval < 0)
1408 goto out_release;
1409
1410out:
1411 /* It may be already another descriptor 8) Not kernel problem. */
1412 return retval;
1413
1414out_release:
1415 sock_release(sock);
1416 return retval;
1417}
1418
1419/*
1420 * Create a pair of connected sockets.
1421 */
1422
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001423SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1424 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425{
1426 struct socket *sock1, *sock2;
1427 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001428 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001429 int flags;
1430
1431 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001432 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001433 return -EINVAL;
1434 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001436 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1437 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1438
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 /*
1440 * Obtain the first socket and check if the underlying protocol
1441 * supports the socketpair call.
1442 */
1443
1444 err = sock_create(family, type, protocol, &sock1);
1445 if (err < 0)
1446 goto out;
1447
1448 err = sock_create(family, type, protocol, &sock2);
1449 if (err < 0)
1450 goto out_release_1;
1451
1452 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001453 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 goto out_release_both;
1455
Al Viro28407632012-08-17 23:54:15 -04001456 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001457 if (unlikely(fd1 < 0)) {
1458 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001460 }
Al Viro28407632012-08-17 23:54:15 -04001461 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001462 if (unlikely(fd2 < 0)) {
1463 err = fd2;
Al Viro28407632012-08-17 23:54:15 -04001464 put_unused_fd(fd1);
1465 goto out_release_both;
1466 }
1467
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001468 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001469 if (unlikely(IS_ERR(newfile1))) {
1470 err = PTR_ERR(newfile1);
1471 put_unused_fd(fd1);
1472 put_unused_fd(fd2);
1473 goto out_release_both;
1474 }
1475
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001476 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001477 if (IS_ERR(newfile2)) {
1478 err = PTR_ERR(newfile2);
Al Viro198de4d2009-08-05 19:29:23 +04001479 fput(newfile1);
1480 put_unused_fd(fd1);
Al Viro28407632012-08-17 23:54:15 -04001481 put_unused_fd(fd2);
Al Viro198de4d2009-08-05 19:29:23 +04001482 sock_release(sock2);
1483 goto out;
Al Virodb349502007-02-07 01:48:00 -05001484 }
1485
Al Viro157cf642008-12-14 04:57:47 -05001486 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001487 fd_install(fd1, newfile1);
1488 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 /* fd1 and fd2 may be already another descriptors.
1490 * Not kernel problem.
1491 */
1492
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001493 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494 if (!err)
1495 err = put_user(fd2, &usockvec[1]);
1496 if (!err)
1497 return 0;
1498
1499 sys_close(fd2);
1500 sys_close(fd1);
1501 return err;
1502
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001504 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001506 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507out:
1508 return err;
1509}
1510
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511/*
1512 * Bind a name to a socket. Nothing much to do here since it's
1513 * the protocol's responsibility to handle the local address.
1514 *
1515 * We move the socket address to kernel space before we call
1516 * the protocol layer (having also checked the address is ok).
1517 */
1518
Heiko Carstens20f37032009-01-14 14:14:23 +01001519SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520{
1521 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001522 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001523 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001525 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001526 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001527 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001528 if (err >= 0) {
1529 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001530 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001531 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001532 if (!err)
1533 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001534 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001535 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001537 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001538 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 return err;
1540}
1541
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542/*
1543 * Perform a listen. Basically, we allow the protocol to do anything
1544 * necessary for a listen, and if that works, we mark the socket as
1545 * ready for listening.
1546 */
1547
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001548SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549{
1550 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001551 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001552 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001553
1554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1555 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001556 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001557 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001558 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559
1560 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001561 if (!err)
1562 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001564 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 }
1566 return err;
1567}
1568
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569/*
1570 * For accept, we attempt to create a new socket, set up the link
1571 * with the client, wake up the client, then return the new
1572 * connected fd. We collect the address of the connector in kernel
1573 * space and move it to user at the very end. This is unclean because
1574 * we open the socket then return an error.
1575 *
1576 * 1003.1g adds the ability to recvmsg() to query connection pending
1577 * status to recvmsg. We need to add that support in a way thats
1578 * clean when we restucture accept also.
1579 */
1580
Heiko Carstens20f37032009-01-14 14:14:23 +01001581SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1582 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583{
1584 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001585 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001586 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001587 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588
Ulrich Drepper77d27202008-07-23 21:29:35 -07001589 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001590 return -EINVAL;
1591
1592 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1593 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1594
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001595 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596 if (!sock)
1597 goto out;
1598
1599 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001600 newsock = sock_alloc();
1601 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 goto out_put;
1603
1604 newsock->type = sock->type;
1605 newsock->ops = sock->ops;
1606
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 /*
1608 * We don't need try_module_get here, as the listening socket (sock)
1609 * has the protocol module (sock->ops->owner) held.
1610 */
1611 __module_get(newsock->ops->owner);
1612
Al Viro28407632012-08-17 23:54:15 -04001613 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001614 if (unlikely(newfd < 0)) {
1615 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001616 sock_release(newsock);
1617 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001618 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001619 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001620 if (unlikely(IS_ERR(newfile))) {
1621 err = PTR_ERR(newfile);
1622 put_unused_fd(newfd);
1623 sock_release(newsock);
1624 goto out_put;
1625 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001626
Frank Filza79af592005-09-27 15:23:38 -07001627 err = security_socket_accept(sock, newsock);
1628 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001629 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001630
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1632 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001633 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634
1635 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001636 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001637 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001639 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001641 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001642 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001644 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 }
1646
1647 /* File flags are not inherited via accept() unlike another OSes. */
1648
David S. Miller39d8c1b2006-03-20 17:13:49 -08001649 fd_install(newfd, newfile);
1650 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001653 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654out:
1655 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001656out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001657 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001658 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 goto out_put;
1660}
1661
Heiko Carstens20f37032009-01-14 14:14:23 +01001662SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1663 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001664{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001665 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001666}
1667
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668/*
1669 * Attempt to connect to a socket with the server address. The address
1670 * is in user space so we verify it is OK and move it to kernel space.
1671 *
1672 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1673 * break bindings
1674 *
1675 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1676 * other SEQPACKET protocols that take time to connect() as it doesn't
1677 * include the -EINPROGRESS status for such sockets.
1678 */
1679
Heiko Carstens20f37032009-01-14 14:14:23 +01001680SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1681 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682{
1683 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001684 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001685 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001687 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 if (!sock)
1689 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001690 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 if (err < 0)
1692 goto out_put;
1693
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001694 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001695 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 if (err)
1697 goto out_put;
1698
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001699 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 sock->file->f_flags);
1701out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001702 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703out:
1704 return err;
1705}
1706
1707/*
1708 * Get the local address ('name') of a socket object. Move the obtained
1709 * name to user space.
1710 */
1711
Heiko Carstens20f37032009-01-14 14:14:23 +01001712SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1713 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714{
1715 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001716 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001717 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001718
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001719 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 if (!sock)
1721 goto out;
1722
1723 err = security_socket_getsockname(sock);
1724 if (err)
1725 goto out_put;
1726
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001727 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 if (err)
1729 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001730 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001733 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734out:
1735 return err;
1736}
1737
1738/*
1739 * Get the remote address ('name') of a socket object. Move the obtained
1740 * name to user space.
1741 */
1742
Heiko Carstens20f37032009-01-14 14:14:23 +01001743SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1744 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745{
1746 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001747 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001748 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001750 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1751 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 err = security_socket_getpeername(sock);
1753 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001754 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 return err;
1756 }
1757
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001758 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001759 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001760 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001762 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001763 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001764 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 }
1766 return err;
1767}
1768
1769/*
1770 * Send a datagram to a given address. We move the address into kernel
1771 * space and check the user space data area is readable before invoking
1772 * the protocol.
1773 */
1774
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001775SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001776 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001777 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778{
1779 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001780 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 int err;
1782 struct msghdr msg;
1783 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001784 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001785
Linus Torvalds253eacc2010-10-30 16:43:10 -07001786 if (len > INT_MAX)
1787 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001788 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1789 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001790 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001791
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001792 iov.iov_base = buff;
1793 iov.iov_len = len;
1794 msg.msg_name = NULL;
1795 msg.msg_iov = &iov;
1796 msg.msg_iovlen = 1;
1797 msg.msg_control = NULL;
1798 msg.msg_controllen = 0;
1799 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001800 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001801 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 if (err < 0)
1803 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001804 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001805 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 }
1807 if (sock->file->f_flags & O_NONBLOCK)
1808 flags |= MSG_DONTWAIT;
1809 msg.msg_flags = flags;
1810 err = sock_sendmsg(sock, &msg, len);
1811
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001812out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001813 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001814out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 return err;
1816}
1817
1818/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001819 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 */
1821
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001822SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001823 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824{
1825 return sys_sendto(fd, buff, len, flags, NULL, 0);
1826}
1827
1828/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001829 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 * sender. We verify the buffers are writable and if needed move the
1831 * sender address from kernel to user space.
1832 */
1833
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001834SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001835 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001836 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837{
1838 struct socket *sock;
1839 struct iovec iov;
1840 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001841 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001842 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001843 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844
Linus Torvalds253eacc2010-10-30 16:43:10 -07001845 if (size > INT_MAX)
1846 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001847 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001849 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001851 msg.msg_control = NULL;
1852 msg.msg_controllen = 0;
1853 msg.msg_iovlen = 1;
1854 msg.msg_iov = &iov;
1855 iov.iov_len = size;
1856 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001857 msg.msg_name = (struct sockaddr *)&address;
1858 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 if (sock->file->f_flags & O_NONBLOCK)
1860 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001861 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001863 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001864 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001865 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001866 if (err2 < 0)
1867 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001869
1870 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001871out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 return err;
1873}
1874
1875/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001876 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877 */
1878
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001879asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001880 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881{
1882 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1883}
1884
1885/*
1886 * Set a socket option. Because we don't know the option lengths we have
1887 * to pass the user mode parameter for the protocols to sort out.
1888 */
1889
Heiko Carstens20f37032009-01-14 14:14:23 +01001890SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1891 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001893 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 struct socket *sock;
1895
1896 if (optlen < 0)
1897 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001898
1899 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1900 if (sock != NULL) {
1901 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001902 if (err)
1903 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904
1905 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001906 err =
1907 sock_setsockopt(sock, level, optname, optval,
1908 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910 err =
1911 sock->ops->setsockopt(sock, level, optname, optval,
1912 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001913out_put:
1914 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 }
1916 return err;
1917}
1918
1919/*
1920 * Get a socket option. Because we don't know the option lengths we have
1921 * to pass a user mode parameter for the protocols to sort out.
1922 */
1923
Heiko Carstens20f37032009-01-14 14:14:23 +01001924SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1925 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001927 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928 struct socket *sock;
1929
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001930 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1931 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001932 err = security_socket_getsockopt(sock, level, optname);
1933 if (err)
1934 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935
1936 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001937 err =
1938 sock_getsockopt(sock, level, optname, optval,
1939 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001941 err =
1942 sock->ops->getsockopt(sock, level, optname, optval,
1943 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001944out_put:
1945 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 }
1947 return err;
1948}
1949
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950/*
1951 * Shutdown a socket.
1952 */
1953
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001954SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001956 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 struct socket *sock;
1958
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001959 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1960 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001962 if (!err)
1963 err = sock->ops->shutdown(sock, how);
1964 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 }
1966 return err;
1967}
1968
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001969/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 * fields which are the same type (int / unsigned) on our platforms.
1971 */
1972#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1973#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1974#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1975
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001976struct used_address {
1977 struct sockaddr_storage name;
1978 unsigned int name_len;
1979};
1980
Anton Blanchard228e5482011-05-02 20:21:35 +00001981static int __sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001982 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001983 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001985 struct compat_msghdr __user *msg_compat =
1986 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001987 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001989 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001990 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1991 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001993 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001994
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 err = -EFAULT;
1996 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001997 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001999 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 return -EFAULT;
2001
Anton Blanchard228e5482011-05-02 20:21:35 +00002002 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002003 err = -EMSGSIZE;
2004 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2005 goto out;
2006 err = -ENOMEM;
2007 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2008 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00002010 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 }
2012
2013 /* This will also move the address data into kernel space */
2014 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002015 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002017 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002018 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 goto out_freeiov;
2020 total_len = err;
2021
2022 err = -ENOBUFS;
2023
Anton Blanchard228e5482011-05-02 20:21:35 +00002024 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002026 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002028 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002029 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002030 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 if (err)
2032 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002033 ctl_buf = msg_sys->msg_control;
2034 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002036 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002038 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 goto out_freeiov;
2040 }
2041 err = -EFAULT;
2042 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002043 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2045 * checking falls down on this.
2046 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002047 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002048 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002049 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002051 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002053 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054
2055 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002056 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002057 /*
2058 * If this is sendmmsg() and current destination address is same as
2059 * previously succeeded address, omit asking LSM's decision.
2060 * used_address->name_len is initialized to UINT_MAX so that the first
2061 * destination address never matches.
2062 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002063 if (used_address && msg_sys->msg_name &&
2064 used_address->name_len == msg_sys->msg_namelen &&
2065 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002066 used_address->name_len)) {
2067 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2068 goto out_freectl;
2069 }
2070 err = sock_sendmsg(sock, msg_sys, total_len);
2071 /*
2072 * If this is sendmmsg() and sending to current destination address was
2073 * successful, remember it.
2074 */
2075 if (used_address && err >= 0) {
2076 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002077 if (msg_sys->msg_name)
2078 memcpy(&used_address->name, msg_sys->msg_name,
2079 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002080 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081
2082out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002083 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2085out_freeiov:
2086 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002087 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002088out:
2089 return err;
2090}
2091
2092/*
2093 * BSD sendmsg interface
2094 */
2095
Eric Dumazet95c96172012-04-15 05:58:06 +00002096SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002097{
2098 int fput_needed, err;
2099 struct msghdr msg_sys;
2100 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2101
2102 if (!sock)
2103 goto out;
2104
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002105 err = __sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002106
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002107 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002108out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 return err;
2110}
2111
Anton Blanchard228e5482011-05-02 20:21:35 +00002112/*
2113 * Linux sendmmsg interface
2114 */
2115
2116int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2117 unsigned int flags)
2118{
2119 int fput_needed, err, datagrams;
2120 struct socket *sock;
2121 struct mmsghdr __user *entry;
2122 struct compat_mmsghdr __user *compat_entry;
2123 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002124 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002125
Anton Blanchard98382f42011-08-04 14:07:39 +00002126 if (vlen > UIO_MAXIOV)
2127 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002128
2129 datagrams = 0;
2130
2131 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2132 if (!sock)
2133 return err;
2134
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002135 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002136 entry = mmsg;
2137 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002138 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002139
2140 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002141 if (MSG_CMSG_COMPAT & flags) {
2142 err = __sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002143 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002144 if (err < 0)
2145 break;
2146 err = __put_user(err, &compat_entry->msg_len);
2147 ++compat_entry;
2148 } else {
2149 err = __sys_sendmsg(sock, (struct msghdr __user *)entry,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002150 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002151 if (err < 0)
2152 break;
2153 err = put_user(err, &entry->msg_len);
2154 ++entry;
2155 }
2156
2157 if (err)
2158 break;
2159 ++datagrams;
2160 }
2161
Anton Blanchard228e5482011-05-02 20:21:35 +00002162 fput_light(sock->file, fput_needed);
2163
Anton Blanchard728ffb82011-08-04 14:07:38 +00002164 /* We only return an error if no datagrams were able to be sent */
2165 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002166 return datagrams;
2167
Anton Blanchard228e5482011-05-02 20:21:35 +00002168 return err;
2169}
2170
2171SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2172 unsigned int, vlen, unsigned int, flags)
2173{
2174 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2175}
2176
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002177static int __sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002178 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002180 struct compat_msghdr __user *msg_compat =
2181 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002183 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002185 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186
2187 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002188 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002189
2190 /* user mode address pointers */
2191 struct sockaddr __user *uaddr;
2192 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002193
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002195 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002197 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002198 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002200 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002201 err = -EMSGSIZE;
2202 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2203 goto out;
2204 err = -ENOMEM;
2205 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2206 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002208 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 }
2210
2211 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002212 * Save the user-mode address (verify_iovec will change the
2213 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002215
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002216 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 uaddr_len = COMPAT_NAMELEN(msg);
2218 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002219 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002221 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222 if (err < 0)
2223 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002224 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002226 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2227 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002228
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 if (sock->file->f_flags & O_NONBLOCK)
2230 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002231 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2232 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 if (err < 0)
2234 goto out_freeiov;
2235 len = err;
2236
2237 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002238 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002239 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002240 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 if (err < 0)
2242 goto out_freeiov;
2243 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002244 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002245 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246 if (err)
2247 goto out_freeiov;
2248 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002249 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 &msg_compat->msg_controllen);
2251 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002252 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 &msg->msg_controllen);
2254 if (err)
2255 goto out_freeiov;
2256 err = len;
2257
2258out_freeiov:
2259 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002260 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002261out:
2262 return err;
2263}
2264
2265/*
2266 * BSD recvmsg interface
2267 */
2268
2269SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2270 unsigned int, flags)
2271{
2272 int fput_needed, err;
2273 struct msghdr msg_sys;
2274 struct socket *sock = sockfd_lookup_light(fd, &err, &fput_needed);
2275
2276 if (!sock)
2277 goto out;
2278
2279 err = __sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2280
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002281 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282out:
2283 return err;
2284}
2285
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002286/*
2287 * Linux recvmmsg interface
2288 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002290int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2291 unsigned int flags, struct timespec *timeout)
2292{
2293 int fput_needed, err, datagrams;
2294 struct socket *sock;
2295 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002296 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002297 struct msghdr msg_sys;
2298 struct timespec end_time;
2299
2300 if (timeout &&
2301 poll_select_set_timeout(&end_time, timeout->tv_sec,
2302 timeout->tv_nsec))
2303 return -EINVAL;
2304
2305 datagrams = 0;
2306
2307 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2308 if (!sock)
2309 return err;
2310
2311 err = sock_error(sock->sk);
2312 if (err)
2313 goto out_put;
2314
2315 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002316 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002317
2318 while (datagrams < vlen) {
2319 /*
2320 * No need to ask LSM for more than the first datagram.
2321 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002322 if (MSG_CMSG_COMPAT & flags) {
2323 err = __sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002324 &msg_sys, flags & ~MSG_WAITFORONE,
2325 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002326 if (err < 0)
2327 break;
2328 err = __put_user(err, &compat_entry->msg_len);
2329 ++compat_entry;
2330 } else {
2331 err = __sys_recvmsg(sock, (struct msghdr __user *)entry,
Anton Blanchardb9eb8b82011-05-17 15:38:57 -04002332 &msg_sys, flags & ~MSG_WAITFORONE,
2333 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002334 if (err < 0)
2335 break;
2336 err = put_user(err, &entry->msg_len);
2337 ++entry;
2338 }
2339
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002340 if (err)
2341 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002342 ++datagrams;
2343
Brandon L Black71c5c152010-03-26 16:18:03 +00002344 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2345 if (flags & MSG_WAITFORONE)
2346 flags |= MSG_DONTWAIT;
2347
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002348 if (timeout) {
2349 ktime_get_ts(timeout);
2350 *timeout = timespec_sub(end_time, *timeout);
2351 if (timeout->tv_sec < 0) {
2352 timeout->tv_sec = timeout->tv_nsec = 0;
2353 break;
2354 }
2355
2356 /* Timeout, return less than vlen datagrams */
2357 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2358 break;
2359 }
2360
2361 /* Out of band data, return right away */
2362 if (msg_sys.msg_flags & MSG_OOB)
2363 break;
2364 }
2365
2366out_put:
2367 fput_light(sock->file, fput_needed);
2368
2369 if (err == 0)
2370 return datagrams;
2371
2372 if (datagrams != 0) {
2373 /*
2374 * We may return less entries than requested (vlen) if the
2375 * sock is non block and there aren't enough datagrams...
2376 */
2377 if (err != -EAGAIN) {
2378 /*
2379 * ... or if recvmsg returns an error after we
2380 * received some datagrams, where we record the
2381 * error to return on the next call or if the
2382 * app asks about it using getsockopt(SO_ERROR).
2383 */
2384 sock->sk->sk_err = -err;
2385 }
2386
2387 return datagrams;
2388 }
2389
2390 return err;
2391}
2392
2393SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2394 unsigned int, vlen, unsigned int, flags,
2395 struct timespec __user *, timeout)
2396{
2397 int datagrams;
2398 struct timespec timeout_sys;
2399
2400 if (!timeout)
2401 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2402
2403 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2404 return -EFAULT;
2405
2406 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2407
2408 if (datagrams > 0 &&
2409 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2410 datagrams = -EFAULT;
2411
2412 return datagrams;
2413}
2414
2415#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416/* Argument list sizes for sys_socketcall */
2417#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002418static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002419 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2420 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2421 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002422 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002423};
2424
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425#undef AL
2426
2427/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002428 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429 *
2430 * Argument checking cleaned up. Saved 20% in size.
2431 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002432 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433 */
2434
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002435SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002436{
2437 unsigned long a[6];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002438 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002440 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441
Anton Blanchard228e5482011-05-02 20:21:35 +00002442 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443 return -EINVAL;
2444
Arjan van de Ven47379052009-09-28 12:57:44 -07002445 len = nargs[call];
2446 if (len > sizeof(a))
2447 return -EINVAL;
2448
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002450 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002452
Al Virof3298dc2008-12-10 03:16:51 -05002453 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002454
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002455 a0 = a[0];
2456 a1 = a[1];
2457
2458 switch (call) {
2459 case SYS_SOCKET:
2460 err = sys_socket(a0, a1, a[2]);
2461 break;
2462 case SYS_BIND:
2463 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2464 break;
2465 case SYS_CONNECT:
2466 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2467 break;
2468 case SYS_LISTEN:
2469 err = sys_listen(a0, a1);
2470 break;
2471 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002472 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2473 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002474 break;
2475 case SYS_GETSOCKNAME:
2476 err =
2477 sys_getsockname(a0, (struct sockaddr __user *)a1,
2478 (int __user *)a[2]);
2479 break;
2480 case SYS_GETPEERNAME:
2481 err =
2482 sys_getpeername(a0, (struct sockaddr __user *)a1,
2483 (int __user *)a[2]);
2484 break;
2485 case SYS_SOCKETPAIR:
2486 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2487 break;
2488 case SYS_SEND:
2489 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2490 break;
2491 case SYS_SENDTO:
2492 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2493 (struct sockaddr __user *)a[4], a[5]);
2494 break;
2495 case SYS_RECV:
2496 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2497 break;
2498 case SYS_RECVFROM:
2499 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2500 (struct sockaddr __user *)a[4],
2501 (int __user *)a[5]);
2502 break;
2503 case SYS_SHUTDOWN:
2504 err = sys_shutdown(a0, a1);
2505 break;
2506 case SYS_SETSOCKOPT:
2507 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2508 break;
2509 case SYS_GETSOCKOPT:
2510 err =
2511 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2512 (int __user *)a[4]);
2513 break;
2514 case SYS_SENDMSG:
2515 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2516 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002517 case SYS_SENDMMSG:
2518 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2519 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002520 case SYS_RECVMSG:
2521 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2522 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002523 case SYS_RECVMMSG:
2524 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2525 (struct timespec __user *)a[4]);
2526 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002527 case SYS_ACCEPT4:
2528 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2529 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002530 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002531 default:
2532 err = -EINVAL;
2533 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534 }
2535 return err;
2536}
2537
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002538#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002539
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002540/**
2541 * sock_register - add a socket protocol handler
2542 * @ops: description of protocol
2543 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002544 * This function is called by a protocol handler that wants to
2545 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002546 * socket interface. The value ops->family coresponds to the
2547 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002549int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550{
2551 int err;
2552
2553 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002554 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2555 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556 return -ENOBUFS;
2557 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002558
2559 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002560 if (rcu_dereference_protected(net_families[ops->family],
2561 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002562 err = -EEXIST;
2563 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002564 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565 err = 0;
2566 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002567 spin_unlock(&net_family_lock);
2568
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002569 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570 return err;
2571}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002572EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002574/**
2575 * sock_unregister - remove a protocol handler
2576 * @family: protocol family to remove
2577 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578 * This function is called by a protocol handler that wants to
2579 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002580 * new socket creation.
2581 *
2582 * If protocol handler is a module, then it can use module reference
2583 * counts to protect against new references. If protocol handler is not
2584 * a module then it needs to provide its own protection in
2585 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002586 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002587void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002588{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002589 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002591 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002592 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002593 spin_unlock(&net_family_lock);
2594
2595 synchronize_rcu();
2596
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002597 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002598}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002599EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002600
Andi Kleen77d76ea2005-12-22 12:43:42 -08002601static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002602{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002603 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002604 /*
2605 * Initialize the network sysctl infrastructure.
2606 */
2607 err = net_sysctl_init();
2608 if (err)
2609 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002610
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002612 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002613 */
2614 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002615
2616 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002617 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618 */
2619
2620 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002621
2622 err = register_filesystem(&sock_fs_type);
2623 if (err)
2624 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002625 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002626 if (IS_ERR(sock_mnt)) {
2627 err = PTR_ERR(sock_mnt);
2628 goto out_mount;
2629 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002630
2631 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002632 */
2633
2634#ifdef CONFIG_NETFILTER
2635 netfilter_init();
2636#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002637
Richard Cochranc1f19b52010-07-17 08:49:36 +00002638#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2639 skb_timestamping_init();
2640#endif
2641
Nick Pigginb3e19d92011-01-07 17:50:11 +11002642out:
2643 return err;
2644
2645out_mount:
2646 unregister_filesystem(&sock_fs_type);
2647out_fs:
2648 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002649}
2650
Andi Kleen77d76ea2005-12-22 12:43:42 -08002651core_initcall(sock_init); /* early initcall */
2652
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653#ifdef CONFIG_PROC_FS
2654void socket_seq_show(struct seq_file *seq)
2655{
2656 int cpu;
2657 int counter = 0;
2658
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002659 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002660 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002661
2662 /* It can be negative, by the way. 8) */
2663 if (counter < 0)
2664 counter = 0;
2665
2666 seq_printf(seq, "sockets: used %d\n", counter);
2667}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002668#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002669
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002670#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002671static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002672 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002673{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002674 mm_segment_t old_fs = get_fs();
2675 struct timeval ktv;
2676 int err;
2677
2678 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002679 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002680 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002681 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002682 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002683
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002684 return err;
2685}
2686
Arnd Bergmann6b960182009-11-06 23:10:54 -08002687static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002688 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002689{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002690 mm_segment_t old_fs = get_fs();
2691 struct timespec kts;
2692 int err;
2693
2694 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002695 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002696 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002697 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002698 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002699
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002700 return err;
2701}
2702
Arnd Bergmann6b960182009-11-06 23:10:54 -08002703static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002704{
2705 struct ifreq __user *uifr;
2706 int err;
2707
2708 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002709 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002710 return -EFAULT;
2711
Arnd Bergmann6b960182009-11-06 23:10:54 -08002712 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002713 if (err)
2714 return err;
2715
Arnd Bergmann6b960182009-11-06 23:10:54 -08002716 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002717 return -EFAULT;
2718
2719 return 0;
2720}
2721
Arnd Bergmann6b960182009-11-06 23:10:54 -08002722static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002723{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002724 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002725 struct ifconf ifc;
2726 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002727 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002728 struct ifreq __user *ifr;
2729 unsigned int i, j;
2730 int err;
2731
Arnd Bergmann6b960182009-11-06 23:10:54 -08002732 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002733 return -EFAULT;
2734
Mathias Krause43da5f22012-08-15 11:31:57 +00002735 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002736 if (ifc32.ifcbuf == 0) {
2737 ifc32.ifc_len = 0;
2738 ifc.ifc_len = 0;
2739 ifc.ifc_req = NULL;
2740 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2741 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002742 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2743 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002744 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2745 ifc.ifc_len = len;
2746 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2747 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002748 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002749 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002750 return -EFAULT;
2751 ifr++;
2752 ifr32++;
2753 }
2754 }
2755 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2756 return -EFAULT;
2757
Arnd Bergmann6b960182009-11-06 23:10:54 -08002758 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002759 if (err)
2760 return err;
2761
2762 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2763 return -EFAULT;
2764
2765 ifr = ifc.ifc_req;
2766 ifr32 = compat_ptr(ifc32.ifcbuf);
2767 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002768 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2769 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2770 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002771 return -EFAULT;
2772 ifr32++;
2773 ifr++;
2774 }
2775
2776 if (ifc32.ifcbuf == 0) {
2777 /* Translate from 64-bit structure multiple to
2778 * a 32-bit one.
2779 */
2780 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002781 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002782 ifc32.ifc_len = i;
2783 } else {
2784 ifc32.ifc_len = i;
2785 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002786 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002787 return -EFAULT;
2788
2789 return 0;
2790}
2791
Arnd Bergmann6b960182009-11-06 23:10:54 -08002792static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002793{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002794 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2795 bool convert_in = false, convert_out = false;
2796 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2797 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002798 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002799 u32 rule_cnt = 0, actual_rule_cnt;
2800 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002801 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002802 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002803
2804 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2805 return -EFAULT;
2806
Ben Hutchings3a7da392011-03-17 07:34:32 +00002807 compat_rxnfc = compat_ptr(data);
2808
2809 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002810 return -EFAULT;
2811
Ben Hutchings3a7da392011-03-17 07:34:32 +00002812 /* Most ethtool structures are defined without padding.
2813 * Unfortunately struct ethtool_rxnfc is an exception.
2814 */
2815 switch (ethcmd) {
2816 default:
2817 break;
2818 case ETHTOOL_GRXCLSRLALL:
2819 /* Buffer size is variable */
2820 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2821 return -EFAULT;
2822 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2823 return -ENOMEM;
2824 buf_size += rule_cnt * sizeof(u32);
2825 /* fall through */
2826 case ETHTOOL_GRXRINGS:
2827 case ETHTOOL_GRXCLSRLCNT:
2828 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002829 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002830 convert_out = true;
2831 /* fall through */
2832 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002833 buf_size += sizeof(struct ethtool_rxnfc);
2834 convert_in = true;
2835 break;
2836 }
2837
2838 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002839 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002840
2841 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2842 return -EFAULT;
2843
2844 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2845 &ifr->ifr_ifru.ifru_data))
2846 return -EFAULT;
2847
2848 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002849 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002850 * fs.ring_cookie and at the end of fs, but nowhere else.
2851 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002852 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2853 sizeof(compat_rxnfc->fs.m_ext) !=
2854 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2855 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002856 BUILD_BUG_ON(
2857 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2858 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2859 offsetof(struct ethtool_rxnfc, fs.location) -
2860 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2861
2862 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002863 (void __user *)(&rxnfc->fs.m_ext + 1) -
2864 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002865 copy_in_user(&rxnfc->fs.ring_cookie,
2866 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002867 (void __user *)(&rxnfc->fs.location + 1) -
2868 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002869 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2870 sizeof(rxnfc->rule_cnt)))
2871 return -EFAULT;
2872 }
2873
2874 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2875 if (ret)
2876 return ret;
2877
2878 if (convert_out) {
2879 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002880 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2881 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002882 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2883 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002884 (const void __user *)(&rxnfc->fs.location + 1) -
2885 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002886 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2887 sizeof(rxnfc->rule_cnt)))
2888 return -EFAULT;
2889
2890 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2891 /* As an optimisation, we only copy the actual
2892 * number of rules that the underlying
2893 * function returned. Since Mallory might
2894 * change the rule count in user memory, we
2895 * check that it is less than the rule count
2896 * originally given (as the user buffer size),
2897 * which has been range-checked.
2898 */
2899 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2900 return -EFAULT;
2901 if (actual_rule_cnt < rule_cnt)
2902 rule_cnt = actual_rule_cnt;
2903 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2904 &rxnfc->rule_locs[0],
2905 rule_cnt * sizeof(u32)))
2906 return -EFAULT;
2907 }
2908 }
2909
2910 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002911}
2912
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002913static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2914{
2915 void __user *uptr;
2916 compat_uptr_t uptr32;
2917 struct ifreq __user *uifr;
2918
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002919 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002920 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2921 return -EFAULT;
2922
2923 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2924 return -EFAULT;
2925
2926 uptr = compat_ptr(uptr32);
2927
2928 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2929 return -EFAULT;
2930
2931 return dev_ioctl(net, SIOCWANDEV, uifr);
2932}
2933
Arnd Bergmann6b960182009-11-06 23:10:54 -08002934static int bond_ioctl(struct net *net, unsigned int cmd,
2935 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002936{
2937 struct ifreq kifr;
2938 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002939 mm_segment_t old_fs;
2940 int err;
2941 u32 data;
2942 void __user *datap;
2943
2944 switch (cmd) {
2945 case SIOCBONDENSLAVE:
2946 case SIOCBONDRELEASE:
2947 case SIOCBONDSETHWADDR:
2948 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002949 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002950 return -EFAULT;
2951
2952 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002953 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002954 err = dev_ioctl(net, cmd,
2955 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002956 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002957
2958 return err;
2959 case SIOCBONDSLAVEINFOQUERY:
2960 case SIOCBONDINFOQUERY:
2961 uifr = compat_alloc_user_space(sizeof(*uifr));
2962 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2963 return -EFAULT;
2964
2965 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2966 return -EFAULT;
2967
2968 datap = compat_ptr(data);
2969 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2970 return -EFAULT;
2971
Arnd Bergmann6b960182009-11-06 23:10:54 -08002972 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002973 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002974 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002975 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002976}
2977
Arnd Bergmann6b960182009-11-06 23:10:54 -08002978static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2979 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002980{
2981 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002982 char tmp_buf[IFNAMSIZ];
2983 void __user *data64;
2984 u32 data32;
2985
2986 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2987 IFNAMSIZ))
2988 return -EFAULT;
2989 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2990 return -EFAULT;
2991 data64 = compat_ptr(data32);
2992
2993 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2994
2995 /* Don't check these user accesses, just let that get trapped
2996 * in the ioctl handler instead.
2997 */
2998 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2999 IFNAMSIZ))
3000 return -EFAULT;
3001 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3002 return -EFAULT;
3003
Arnd Bergmann6b960182009-11-06 23:10:54 -08003004 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003005}
3006
Arnd Bergmann6b960182009-11-06 23:10:54 -08003007static int dev_ifsioc(struct net *net, struct socket *sock,
3008 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003009{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003010 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003011 int err;
3012
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003013 uifr = compat_alloc_user_space(sizeof(*uifr));
3014 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3015 return -EFAULT;
3016
3017 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3018
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003019 if (!err) {
3020 switch (cmd) {
3021 case SIOCGIFFLAGS:
3022 case SIOCGIFMETRIC:
3023 case SIOCGIFMTU:
3024 case SIOCGIFMEM:
3025 case SIOCGIFHWADDR:
3026 case SIOCGIFINDEX:
3027 case SIOCGIFADDR:
3028 case SIOCGIFBRDADDR:
3029 case SIOCGIFDSTADDR:
3030 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003031 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003032 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003033 case SIOCGMIIPHY:
3034 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003035 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003036 err = -EFAULT;
3037 break;
3038 }
3039 }
3040 return err;
3041}
3042
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003043static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3044 struct compat_ifreq __user *uifr32)
3045{
3046 struct ifreq ifr;
3047 struct compat_ifmap __user *uifmap32;
3048 mm_segment_t old_fs;
3049 int err;
3050
3051 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3052 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3053 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3054 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3055 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3056 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3057 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3058 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3059 if (err)
3060 return -EFAULT;
3061
3062 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003063 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003064 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003065 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003066
3067 if (cmd == SIOCGIFMAP && !err) {
3068 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3069 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3070 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3071 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3072 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3073 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3074 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3075 if (err)
3076 err = -EFAULT;
3077 }
3078 return err;
3079}
3080
3081static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3082{
3083 void __user *uptr;
3084 compat_uptr_t uptr32;
3085 struct ifreq __user *uifr;
3086
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003087 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003088 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3089 return -EFAULT;
3090
3091 if (get_user(uptr32, &uifr32->ifr_data))
3092 return -EFAULT;
3093
3094 uptr = compat_ptr(uptr32);
3095
3096 if (put_user(uptr, &uifr->ifr_data))
3097 return -EFAULT;
3098
3099 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3100}
3101
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003102struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003103 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003104 struct sockaddr rt_dst; /* target address */
3105 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3106 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003107 unsigned short rt_flags;
3108 short rt_pad2;
3109 u32 rt_pad3;
3110 unsigned char rt_tos;
3111 unsigned char rt_class;
3112 short rt_pad4;
3113 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003114 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003115 u32 rt_mtu; /* per route MTU/Window */
3116 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003117 unsigned short rt_irtt; /* Initial RTT */
3118};
3119
3120struct in6_rtmsg32 {
3121 struct in6_addr rtmsg_dst;
3122 struct in6_addr rtmsg_src;
3123 struct in6_addr rtmsg_gateway;
3124 u32 rtmsg_type;
3125 u16 rtmsg_dst_len;
3126 u16 rtmsg_src_len;
3127 u32 rtmsg_metric;
3128 u32 rtmsg_info;
3129 u32 rtmsg_flags;
3130 s32 rtmsg_ifindex;
3131};
3132
Arnd Bergmann6b960182009-11-06 23:10:54 -08003133static int routing_ioctl(struct net *net, struct socket *sock,
3134 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003135{
3136 int ret;
3137 void *r = NULL;
3138 struct in6_rtmsg r6;
3139 struct rtentry r4;
3140 char devname[16];
3141 u32 rtdev;
3142 mm_segment_t old_fs = get_fs();
3143
Arnd Bergmann6b960182009-11-06 23:10:54 -08003144 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3145 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003146 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003147 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003148 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3149 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3150 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3151 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3152 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3153 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3154 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003155
3156 r = (void *) &r6;
3157 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003158 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003159 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003160 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003161 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3162 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3163 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3164 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3165 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3166 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003167 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003168 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003169 r4.rt_dev = (char __user __force *)devname;
3170 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003171 } else
3172 r4.rt_dev = NULL;
3173
3174 r = (void *) &r4;
3175 }
3176
3177 if (ret) {
3178 ret = -EFAULT;
3179 goto out;
3180 }
3181
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003182 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003183 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003184 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003185
3186out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003187 return ret;
3188}
3189
3190/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3191 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003192 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003193 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003194static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003195{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003196 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003197
Arnd Bergmann6b960182009-11-06 23:10:54 -08003198 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003199 return -EFAULT;
3200 if (tmp == BRCTL_GET_VERSION)
3201 return BRCTL_VERSION + 1;
3202 return -EINVAL;
3203}
3204
Arnd Bergmann6b960182009-11-06 23:10:54 -08003205static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3206 unsigned int cmd, unsigned long arg)
3207{
3208 void __user *argp = compat_ptr(arg);
3209 struct sock *sk = sock->sk;
3210 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003211
Arnd Bergmann6b960182009-11-06 23:10:54 -08003212 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3213 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003214
Arnd Bergmann6b960182009-11-06 23:10:54 -08003215 switch (cmd) {
3216 case SIOCSIFBR:
3217 case SIOCGIFBR:
3218 return old_bridge_ioctl(argp);
3219 case SIOCGIFNAME:
3220 return dev_ifname32(net, argp);
3221 case SIOCGIFCONF:
3222 return dev_ifconf(net, argp);
3223 case SIOCETHTOOL:
3224 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003225 case SIOCWANDEV:
3226 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003227 case SIOCGIFMAP:
3228 case SIOCSIFMAP:
3229 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003230 case SIOCBONDENSLAVE:
3231 case SIOCBONDRELEASE:
3232 case SIOCBONDSETHWADDR:
3233 case SIOCBONDSLAVEINFOQUERY:
3234 case SIOCBONDINFOQUERY:
3235 case SIOCBONDCHANGEACTIVE:
3236 return bond_ioctl(net, cmd, argp);
3237 case SIOCADDRT:
3238 case SIOCDELRT:
3239 return routing_ioctl(net, sock, cmd, argp);
3240 case SIOCGSTAMP:
3241 return do_siocgstamp(net, sock, cmd, argp);
3242 case SIOCGSTAMPNS:
3243 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003244 case SIOCSHWTSTAMP:
3245 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003246
Arnd Bergmann6b960182009-11-06 23:10:54 -08003247 case FIOSETOWN:
3248 case SIOCSPGRP:
3249 case FIOGETOWN:
3250 case SIOCGPGRP:
3251 case SIOCBRADDBR:
3252 case SIOCBRDELBR:
3253 case SIOCGIFVLAN:
3254 case SIOCSIFVLAN:
3255 case SIOCADDDLCI:
3256 case SIOCDELDLCI:
3257 return sock_ioctl(file, cmd, arg);
3258
3259 case SIOCGIFFLAGS:
3260 case SIOCSIFFLAGS:
3261 case SIOCGIFMETRIC:
3262 case SIOCSIFMETRIC:
3263 case SIOCGIFMTU:
3264 case SIOCSIFMTU:
3265 case SIOCGIFMEM:
3266 case SIOCSIFMEM:
3267 case SIOCGIFHWADDR:
3268 case SIOCSIFHWADDR:
3269 case SIOCADDMULTI:
3270 case SIOCDELMULTI:
3271 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003272 case SIOCGIFADDR:
3273 case SIOCSIFADDR:
3274 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003275 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003276 case SIOCGIFBRDADDR:
3277 case SIOCSIFBRDADDR:
3278 case SIOCGIFDSTADDR:
3279 case SIOCSIFDSTADDR:
3280 case SIOCGIFNETMASK:
3281 case SIOCSIFNETMASK:
3282 case SIOCSIFPFLAGS:
3283 case SIOCGIFPFLAGS:
3284 case SIOCGIFTXQLEN:
3285 case SIOCSIFTXQLEN:
3286 case SIOCBRADDIF:
3287 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003288 case SIOCSIFNAME:
3289 case SIOCGMIIPHY:
3290 case SIOCGMIIREG:
3291 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003292 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003293
Arnd Bergmann6b960182009-11-06 23:10:54 -08003294 case SIOCSARP:
3295 case SIOCGARP:
3296 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003297 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003298 return sock_do_ioctl(net, sock, cmd, arg);
3299 }
3300
Arnd Bergmann6b960182009-11-06 23:10:54 -08003301 return -ENOIOCTLCMD;
3302}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003303
Eric Dumazet95c96172012-04-15 05:58:06 +00003304static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003305 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003306{
3307 struct socket *sock = file->private_data;
3308 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003309 struct sock *sk;
3310 struct net *net;
3311
3312 sk = sock->sk;
3313 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003314
3315 if (sock->ops->compat_ioctl)
3316 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3317
David S. Miller87de87d2008-06-03 09:14:03 -07003318 if (ret == -ENOIOCTLCMD &&
3319 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3320 ret = compat_wext_handle_ioctl(net, cmd, arg);
3321
Arnd Bergmann6b960182009-11-06 23:10:54 -08003322 if (ret == -ENOIOCTLCMD)
3323 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3324
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003325 return ret;
3326}
3327#endif
3328
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003329int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3330{
3331 return sock->ops->bind(sock, addr, addrlen);
3332}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003333EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003334
3335int kernel_listen(struct socket *sock, int backlog)
3336{
3337 return sock->ops->listen(sock, backlog);
3338}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003339EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003340
3341int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3342{
3343 struct sock *sk = sock->sk;
3344 int err;
3345
3346 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3347 newsock);
3348 if (err < 0)
3349 goto done;
3350
3351 err = sock->ops->accept(sock, *newsock, flags);
3352 if (err < 0) {
3353 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003354 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003355 goto done;
3356 }
3357
3358 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003359 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003360
3361done:
3362 return err;
3363}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003364EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003365
3366int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003367 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003368{
3369 return sock->ops->connect(sock, addr, addrlen, flags);
3370}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003371EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003372
3373int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3374 int *addrlen)
3375{
3376 return sock->ops->getname(sock, addr, addrlen, 0);
3377}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003378EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003379
3380int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3381 int *addrlen)
3382{
3383 return sock->ops->getname(sock, addr, addrlen, 1);
3384}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003385EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003386
3387int kernel_getsockopt(struct socket *sock, int level, int optname,
3388 char *optval, int *optlen)
3389{
3390 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003391 char __user *uoptval;
3392 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003393 int err;
3394
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003395 uoptval = (char __user __force *) optval;
3396 uoptlen = (int __user __force *) optlen;
3397
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003398 set_fs(KERNEL_DS);
3399 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003400 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003401 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003402 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3403 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003404 set_fs(oldfs);
3405 return err;
3406}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003407EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003408
3409int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003410 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003411{
3412 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003413 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003414 int err;
3415
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003416 uoptval = (char __user __force *) optval;
3417
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003418 set_fs(KERNEL_DS);
3419 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003420 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003421 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003422 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003423 optlen);
3424 set_fs(oldfs);
3425 return err;
3426}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003427EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003428
3429int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3430 size_t size, int flags)
3431{
3432 if (sock->ops->sendpage)
3433 return sock->ops->sendpage(sock, page, offset, size, flags);
3434
3435 return sock_no_sendpage(sock, page, offset, size, flags);
3436}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003437EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003438
3439int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3440{
3441 mm_segment_t oldfs = get_fs();
3442 int err;
3443
3444 set_fs(KERNEL_DS);
3445 err = sock->ops->ioctl(sock, cmd, arg);
3446 set_fs(oldfs);
3447
3448 return err;
3449}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003450EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003451
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003452int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3453{
3454 return sock->ops->shutdown(sock, how);
3455}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003456EXPORT_SYMBOL(kernel_sock_shutdown);