blob: 4da14cbd49b6486d08ec840e714b9f583835b685 [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>
Eliezer Tamir06021292013-06-10 11:39:50 +0300107#include <net/ll_poll.h>
108
109#ifdef CONFIG_NET_LL_RX_POLL
Eliezer Tamir2d48d672013-06-24 10:28:03 +0300110unsigned int sysctl_net_ll_read __read_mostly;
Eliezer Tamireb6db622013-06-14 16:33:25 +0300111unsigned int sysctl_net_ll_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300112#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700115static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
116 unsigned long nr_segs, loff_t pos);
117static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
118 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700119static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
121static int sock_close(struct inode *inode, struct file *file);
122static unsigned int sock_poll(struct file *file,
123 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700124static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800125#ifdef CONFIG_COMPAT
126static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700127 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800128#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130static ssize_t sock_sendpage(struct file *file, struct page *page,
131 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800132static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700133 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800134 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136/*
137 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
138 * in the operation structures but are done directly via the socketcall() multiplexor.
139 */
140
Arjan van de Venda7071d2007-02-12 00:55:36 -0800141static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 .owner = THIS_MODULE,
143 .llseek = no_llseek,
144 .aio_read = sock_aio_read,
145 .aio_write = sock_aio_write,
146 .poll = sock_poll,
147 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800148#ifdef CONFIG_COMPAT
149 .compat_ioctl = compat_sock_ioctl,
150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 .mmap = sock_mmap,
152 .open = sock_no_open, /* special open code to disallow open via /proc */
153 .release = sock_close,
154 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200155 .sendpage = sock_sendpage,
156 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800157 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158};
159
160/*
161 * The protocol list. Each protocol is registered in here.
162 */
163
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000165static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167/*
168 * Statistics counters of the socket lists
169 */
170
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700171static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700174 * Support routines.
175 * Move socket addresses back and forth across the kernel/user
176 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 */
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179/**
180 * move_addr_to_kernel - copy a socket address into kernel space
181 * @uaddr: Address in user space
182 * @kaddr: Address in kernel space
183 * @ulen: Length in user space
184 *
185 * The address is copied into kernel space. If the provided address is
186 * too long an error code of -EINVAL is returned. If the copy gives
187 * invalid addresses -EFAULT is returned. On a success 0 is returned.
188 */
189
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000190int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700192 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700194 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700196 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100198 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199}
200
201/**
202 * move_addr_to_user - copy an address to user space
203 * @kaddr: kernel space address
204 * @klen: length of address in kernel
205 * @uaddr: user space address
206 * @ulen: pointer to user length field
207 *
208 * The value pointed to by ulen on entry is the buffer length available.
209 * This is overwritten with the buffer space used. -EINVAL is returned
210 * if an overlong buffer is specified or a negative buffer size. -EFAULT
211 * is returned if either the buffer or the length field are not
212 * accessible.
213 * After copying the data up to the limit the user specifies, the true
214 * length of the data is written over the length limit the user
215 * specified. Zero is returned for a success.
216 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700217
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000218static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000219 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220{
221 int err;
222 int len;
223
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700224 err = get_user(len, ulen);
225 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700227 if (len > klen)
228 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700229 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700231 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500232 if (audit_sockaddr(klen, kaddr))
233 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700234 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 return -EFAULT;
236 }
237 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700238 * "fromlen shall refer to the value before truncation.."
239 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 */
241 return __put_user(klen, ulen);
242}
243
Christoph Lametere18b8902006-12-06 20:33:20 -0800244static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245
246static struct inode *sock_alloc_inode(struct super_block *sb)
247{
248 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000249 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700250
Christoph Lametere94b1762006-12-06 20:33:17 -0800251 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 if (!ei)
253 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000254 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000256 kmem_cache_free(sock_inode_cachep, ei);
257 return NULL;
258 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000259 init_waitqueue_head(&wq->wait);
260 wq->fasync_list = NULL;
261 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700262
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
269 return &ei->vfs_inode;
270}
271
272static void sock_destroy_inode(struct inode *inode)
273{
Eric Dumazet43815482010-04-29 11:01:49 +0000274 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000275 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000276
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000278 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800279 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000280 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281}
282
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700283static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700285 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Christoph Lametera35afb82007-05-16 22:10:57 -0700287 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290static int init_inodecache(void)
291{
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700293 sizeof(struct socket_alloc),
294 0,
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
297 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900298 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 if (sock_inode_cachep == NULL)
300 return -ENOMEM;
301 return 0;
302}
303
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700304static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700305 .alloc_inode = sock_alloc_inode,
306 .destroy_inode = sock_destroy_inode,
307 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308};
309
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700310/*
311 * sockfs_dname() is called from d_path().
312 */
313static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314{
315 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316 dentry->d_inode->i_ino);
317}
318
Al Viro3ba13d12009-02-20 06:02:22 +0000319static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700320 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321};
322
Al Viroc74a1cb2011-01-12 16:59:34 -0500323static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324 int flags, const char *dev_name, void *data)
325{
326 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327 &sockfs_dentry_operations, SOCKFS_MAGIC);
328}
329
330static struct vfsmount *sock_mnt __read_mostly;
331
332static struct file_system_type sock_fs_type = {
333 .name = "sockfs",
334 .mount = sockfs_mount,
335 .kill_sb = kill_anon_super,
336};
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338/*
339 * Obtains the first available file descriptor and sets it up for use.
340 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
348 *
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
353 */
354
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700355struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356{
Al Viro7cbe66b2009-08-05 19:59:08 +0400357 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400358 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400359 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800360
Masatake YAMATO600e1772012-08-29 10:44:29 +0000361 if (dname) {
362 name.name = dname;
363 name.len = strlen(name.name);
364 } else if (sock->sk) {
365 name.name = sock->sk->sk_prot_creator->name;
366 name.len = strlen(name.name);
367 }
Nick Piggin4b936882011-01-07 17:50:07 +1100368 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400369 if (unlikely(!path.dentry))
370 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400371 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
Al Viro2c48b9c2009-08-09 00:52:35 +0400373 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400374 SOCK_INODE(sock)->i_fop = &socket_file_ops;
375
Al Viro2c48b9c2009-08-09 00:52:35 +0400376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400377 &socket_file_ops);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700378 if (unlikely(IS_ERR(file))) {
Al Virocc3808f2009-08-06 09:43:59 +0400379 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400380 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400381 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700382 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400383 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800384
385 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400388 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800389}
Al Viro56b31d12012-08-18 00:25:51 -0400390EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800391
Al Viro56b31d12012-08-18 00:25:51 -0400392static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393{
394 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400395 int fd = get_unused_fd_flags(flags);
396 if (unlikely(fd < 0))
397 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800398
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700399 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400400 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400402 return fd;
403 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400404
Al Viro28407632012-08-17 23:54:15 -0400405 put_unused_fd(fd);
406 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407}
408
John Fastabend406a3c62012-07-20 10:39:25 +0000409struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800410{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800411 if (file->f_op == &socket_file_ops)
412 return file->private_data; /* set in sock_map_fd */
413
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800414 *err = -ENOTSOCK;
415 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800416}
John Fastabend406a3c62012-07-20 10:39:25 +0000417EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800418
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700420 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * @fd: file handle
422 * @err: pointer to an error code return
423 *
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
428 *
429 * On a success the socket object pointer is returned.
430 */
431
432struct socket *sockfd_lookup(int fd, int *err)
433{
434 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 struct socket *sock;
436
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700437 file = fget(fd);
438 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 *err = -EBADF;
440 return NULL;
441 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700442
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800443 sock = sock_from_file(file, err);
444 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 return sock;
447}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700448EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800450static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
451{
452 struct file *file;
453 struct socket *sock;
454
Hua Zhong36725582006-04-19 15:25:02 -0700455 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800456 file = fget_light(fd, fput_needed);
457 if (file) {
458 sock = sock_from_file(file, err);
459 if (sock)
460 return sock;
461 fput_light(file, *fput_needed);
462 }
463 return NULL;
464}
465
Masatake YAMATO600e1772012-08-29 10:44:29 +0000466#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469static ssize_t sockfs_getxattr(struct dentry *dentry,
470 const char *name, void *value, size_t size)
471{
472 const char *proto_name;
473 size_t proto_size;
474 int error;
475
476 error = -ENODATA;
477 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478 proto_name = dentry->d_name.name;
479 proto_size = strlen(proto_name);
480
481 if (value) {
482 error = -ERANGE;
483 if (proto_size + 1 > size)
484 goto out;
485
486 strncpy(value, proto_name, proto_size + 1);
487 }
488 error = proto_size + 1;
489 }
490
491out:
492 return error;
493}
494
495static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496 size_t size)
497{
498 ssize_t len;
499 ssize_t used = 0;
500
501 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
502 if (len < 0)
503 return len;
504 used += len;
505 if (buffer) {
506 if (size < used)
507 return -ERANGE;
508 buffer += len;
509 }
510
511 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512 used += len;
513 if (buffer) {
514 if (size < used)
515 return -ERANGE;
516 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517 buffer += len;
518 }
519
520 return used;
521}
522
523static const struct inode_operations sockfs_inode_ops = {
524 .getxattr = sockfs_getxattr,
525 .listxattr = sockfs_listxattr,
526};
527
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528/**
529 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700530 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 * Allocate a new inode and socket object. The two are bound together
532 * and initialised. The socket is then returned. If we are out of inodes
533 * NULL is returned.
534 */
535
536static struct socket *sock_alloc(void)
537{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700538 struct inode *inode;
539 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200541 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 if (!inode)
543 return NULL;
544
545 sock = SOCKET_I(inode);
546
Eric Dumazet29a020d2009-09-15 02:39:20 -0700547 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400548 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700549 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100550 inode->i_uid = current_fsuid();
551 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000552 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
Alex Shi19e8d692012-05-14 14:15:31 -0700554 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return sock;
556}
557
558/*
559 * In theory you can't get an open on this inode, but /proc provides
560 * a back door. Remember to keep it shut otherwise you'll let the
561 * creepy crawlies in.
562 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
565{
566 return -ENXIO;
567}
568
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800569const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 .owner = THIS_MODULE,
571 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200572 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573};
574
575/**
576 * sock_release - close a socket
577 * @sock: socket to close
578 *
579 * The socket is released from the protocol stack if it has a release
580 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700581 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700583
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584void sock_release(struct socket *sock)
585{
586 if (sock->ops) {
587 struct module *owner = sock->ops->owner;
588
589 sock->ops->release(sock);
590 sock->ops = NULL;
591 module_put(owner);
592 }
593
Eric Dumazeteaefd112011-02-18 03:26:36 +0000594 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 printk(KERN_ERR "sock_release: fasync list not empty!\n");
596
Mikulas Patockab09e7862012-07-19 06:13:36 +0000597 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
598 return;
599
Alex Shi19e8d692012-05-14 14:15:31 -0700600 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 if (!sock->file) {
602 iput(SOCK_INODE(sock));
603 return;
604 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700605 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700607EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000609void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000610{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000611 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000612 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000613 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000614 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000615 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100616 if (sock_flag(sk, SOCK_WIFI_STATUS))
617 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000618}
619EXPORT_SYMBOL(sock_tx_timestamp);
620
Anton Blanchard228e5482011-05-02 20:21:35 +0000621static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
622 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623{
624 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626 si->sock = sock;
627 si->scm = NULL;
628 si->msg = msg;
629 si->size = size;
630
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 return sock->ops->sendmsg(iocb, sock, msg, size);
632}
633
Anton Blanchard228e5482011-05-02 20:21:35 +0000634static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
635 struct msghdr *msg, size_t size)
636{
637 int err = security_socket_sendmsg(sock, msg, size);
638
639 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
640}
641
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
643{
644 struct kiocb iocb;
645 struct sock_iocb siocb;
646 int ret;
647
648 init_sync_kiocb(&iocb, NULL);
649 iocb.private = &siocb;
650 ret = __sock_sendmsg(&iocb, sock, msg, size);
651 if (-EIOCBQUEUED == ret)
652 ret = wait_on_sync_kiocb(&iocb);
653 return ret;
654}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700655EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Eric Dumazet894dc242011-07-26 02:39:41 +0000657static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000658{
659 struct kiocb iocb;
660 struct sock_iocb siocb;
661 int ret;
662
663 init_sync_kiocb(&iocb, NULL);
664 iocb.private = &siocb;
665 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
666 if (-EIOCBQUEUED == ret)
667 ret = wait_on_sync_kiocb(&iocb);
668 return ret;
669}
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
672 struct kvec *vec, size_t num, size_t size)
673{
674 mm_segment_t oldfs = get_fs();
675 int result;
676
677 set_fs(KERNEL_DS);
678 /*
679 * the following is safe, since for compiler definitions of kvec and
680 * iovec are identical, yielding the same in-core layout and alignment
681 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700682 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 msg->msg_iovlen = num;
684 result = sock_sendmsg(sock, msg, size);
685 set_fs(oldfs);
686 return result;
687}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700688EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700690/*
691 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
692 */
693void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
694 struct sk_buff *skb)
695{
Patrick Ohly20d49472009-02-12 05:03:38 +0000696 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
697 struct timespec ts[3];
698 int empty = 1;
699 struct skb_shared_hwtstamps *shhwtstamps =
700 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700701
Patrick Ohly20d49472009-02-12 05:03:38 +0000702 /* Race occurred between timestamp enabling and packet
703 receiving. Fill in the current time for now. */
704 if (need_software_tstamp && skb->tstamp.tv64 == 0)
705 __net_timestamp(skb);
706
707 if (need_software_tstamp) {
708 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
709 struct timeval tv;
710 skb_get_timestamp(skb, &tv);
711 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
712 sizeof(tv), &tv);
713 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000714 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000715 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000716 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000717 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700718 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000719
720
721 memset(ts, 0, sizeof(ts));
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000722 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
723 ktime_to_timespec_cond(skb->tstamp, ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000724 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000725 if (shhwtstamps) {
726 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000727 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
Patrick Ohly20d49472009-02-12 05:03:38 +0000728 empty = 0;
729 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000730 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
Patrick Ohly20d49472009-02-12 05:03:38 +0000731 empty = 0;
732 }
733 if (!empty)
734 put_cmsg(msg, SOL_SOCKET,
735 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700736}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300737EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
738
Johannes Berg6e3e9392011-11-09 10:15:42 +0100739void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
740 struct sk_buff *skb)
741{
742 int ack;
743
744 if (!sock_flag(sk, SOCK_WIFI_STATUS))
745 return;
746 if (!skb->wifi_acked_valid)
747 return;
748
749 ack = skb->wifi_acked;
750
751 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
752}
753EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
754
stephen hemminger11165f12010-10-18 14:27:29 +0000755static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
756 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700757{
758 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
759 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
760 sizeof(__u32), &skb->dropcount);
761}
762
Eric Dumazet767dd032010-04-28 19:14:43 +0000763void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700764 struct sk_buff *skb)
765{
766 sock_recv_timestamp(msg, sk, skb);
767 sock_recv_drops(msg, sk, skb);
768}
Eric Dumazet767dd032010-04-28 19:14:43 +0000769EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700770
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700771static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
772 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 struct sock_iocb *si = kiocb_to_siocb(iocb);
775
776 si->sock = sock;
777 si->scm = NULL;
778 si->msg = msg;
779 si->size = size;
780 si->flags = flags;
781
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
783}
784
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700785static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
786 struct msghdr *msg, size_t size, int flags)
787{
788 int err = security_socket_recvmsg(sock, msg, size, flags);
789
790 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
791}
792
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700793int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 size_t size, int flags)
795{
796 struct kiocb iocb;
797 struct sock_iocb siocb;
798 int ret;
799
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700800 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 iocb.private = &siocb;
802 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
803 if (-EIOCBQUEUED == ret)
804 ret = wait_on_sync_kiocb(&iocb);
805 return ret;
806}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700807EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700809static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
810 size_t size, int flags)
811{
812 struct kiocb iocb;
813 struct sock_iocb siocb;
814 int ret;
815
816 init_sync_kiocb(&iocb, NULL);
817 iocb.private = &siocb;
818 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
819 if (-EIOCBQUEUED == ret)
820 ret = wait_on_sync_kiocb(&iocb);
821 return ret;
822}
823
Martin Lucinac1249c02010-12-10 00:04:05 +0000824/**
825 * kernel_recvmsg - Receive a message from a socket (kernel space)
826 * @sock: The socket to receive the message from
827 * @msg: Received message
828 * @vec: Input s/g array for message data
829 * @num: Size of input s/g array
830 * @size: Number of bytes to read
831 * @flags: Message flags (MSG_DONTWAIT, etc...)
832 *
833 * On return the msg structure contains the scatter/gather array passed in the
834 * vec argument. The array is modified so that it consists of the unfilled
835 * portion of the original array.
836 *
837 * The returned value is the total number of bytes received, or an error.
838 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700839int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
840 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
842 mm_segment_t oldfs = get_fs();
843 int result;
844
845 set_fs(KERNEL_DS);
846 /*
847 * the following is safe, since for compiler definitions of kvec and
848 * iovec are identical, yielding the same in-core layout and alignment
849 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700850 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 result = sock_recvmsg(sock, msg, size, flags);
852 set_fs(oldfs);
853 return result;
854}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700855EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
857static void sock_aio_dtor(struct kiocb *iocb)
858{
859 kfree(iocb->private);
860}
861
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300862static ssize_t sock_sendpage(struct file *file, struct page *page,
863 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864{
865 struct socket *sock;
866 int flags;
867
Eric Dumazetb69aee02005-09-06 14:42:45 -0700868 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
Eric Dumazet35f9c092012-04-05 03:05:35 +0000870 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
871 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
872 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
Linus Torvaldse6949582009-08-13 08:28:36 -0700874 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875}
876
Jens Axboe9c55e012007-11-06 23:30:13 -0800877static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700878 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800879 unsigned int flags)
880{
881 struct socket *sock = file->private_data;
882
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800883 if (unlikely(!sock->ops->splice_read))
884 return -EINVAL;
885
Jens Axboe9c55e012007-11-06 23:30:13 -0800886 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
887}
888
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800889static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700890 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800891{
892 if (!is_sync_kiocb(iocb)) {
893 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
894 if (!siocb)
895 return NULL;
896 iocb->ki_dtor = sock_aio_dtor;
897 }
898
899 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800900 iocb->private = siocb;
901 return siocb;
902}
903
904static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700905 struct file *file, const struct iovec *iov,
906 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800907{
908 struct socket *sock = file->private_data;
909 size_t size = 0;
910 int i;
911
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700912 for (i = 0; i < nr_segs; i++)
913 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800914
915 msg->msg_name = NULL;
916 msg->msg_namelen = 0;
917 msg->msg_control = NULL;
918 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700919 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800920 msg->msg_iovlen = nr_segs;
921 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
922
923 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
924}
925
Badari Pulavarty027445c2006-09-30 23:28:46 -0700926static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
927 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800928{
929 struct sock_iocb siocb, *x;
930
931 if (pos != 0)
932 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700933
934 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800935 return 0;
936
Badari Pulavarty027445c2006-09-30 23:28:46 -0700937
938 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800939 if (!x)
940 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700941 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800942}
943
944static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700945 struct file *file, const struct iovec *iov,
946 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800947{
948 struct socket *sock = file->private_data;
949 size_t size = 0;
950 int i;
951
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700952 for (i = 0; i < nr_segs; i++)
953 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800954
955 msg->msg_name = NULL;
956 msg->msg_namelen = 0;
957 msg->msg_control = NULL;
958 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700959 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800960 msg->msg_iovlen = nr_segs;
961 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
962 if (sock->type == SOCK_SEQPACKET)
963 msg->msg_flags |= MSG_EOR;
964
965 return __sock_sendmsg(iocb, sock, msg, size);
966}
967
Badari Pulavarty027445c2006-09-30 23:28:46 -0700968static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
969 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800971 struct sock_iocb siocb, *x;
972
973 if (pos != 0)
974 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700975
Badari Pulavarty027445c2006-09-30 23:28:46 -0700976 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800977 if (!x)
978 return -ENOMEM;
979
Badari Pulavarty027445c2006-09-30 23:28:46 -0700980 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981}
982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983/*
984 * Atomic setting of ioctl hooks to avoid race
985 * with module unload.
986 */
987
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800988static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700989static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Eric W. Biederman881d9662007-09-17 11:56:21 -0700991void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800993 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800995 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996}
997EXPORT_SYMBOL(brioctl_set);
998
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800999static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001000static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
Eric W. Biederman881d9662007-09-17 11:56:21 -07001002void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001004 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001006 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007}
1008EXPORT_SYMBOL(vlan_ioctl_set);
1009
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001010static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001011static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001013void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001015 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001017 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018}
1019EXPORT_SYMBOL(dlci_ioctl_set);
1020
Arnd Bergmann6b960182009-11-06 23:10:54 -08001021static long sock_do_ioctl(struct net *net, struct socket *sock,
1022 unsigned int cmd, unsigned long arg)
1023{
1024 int err;
1025 void __user *argp = (void __user *)arg;
1026
1027 err = sock->ops->ioctl(sock, cmd, arg);
1028
1029 /*
1030 * If this ioctl is unknown try to hand it down
1031 * to the NIC driver.
1032 */
1033 if (err == -ENOIOCTLCMD)
1034 err = dev_ioctl(net, cmd, argp);
1035
1036 return err;
1037}
1038
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039/*
1040 * With an ioctl, arg may well be a user mode pointer, but we don't know
1041 * what to do with it - that's up to the protocol still.
1042 */
1043
1044static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1045{
1046 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001047 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 void __user *argp = (void __user *)arg;
1049 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001050 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
Eric Dumazetb69aee02005-09-06 14:42:45 -07001052 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001053 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001054 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001056 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001058#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001060 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001062#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001063 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064 case FIOSETOWN:
1065 case SIOCSPGRP:
1066 err = -EFAULT;
1067 if (get_user(pid, (int __user *)argp))
1068 break;
1069 err = f_setown(sock->file, pid, 1);
1070 break;
1071 case FIOGETOWN:
1072 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001073 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001074 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 break;
1076 case SIOCGIFBR:
1077 case SIOCSIFBR:
1078 case SIOCBRADDBR:
1079 case SIOCBRDELBR:
1080 err = -ENOPKG;
1081 if (!br_ioctl_hook)
1082 request_module("bridge");
1083
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001084 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001085 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001086 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001087 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 break;
1089 case SIOCGIFVLAN:
1090 case SIOCSIFVLAN:
1091 err = -ENOPKG;
1092 if (!vlan_ioctl_hook)
1093 request_module("8021q");
1094
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001095 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001097 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001098 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 case SIOCADDDLCI:
1101 case SIOCDELDLCI:
1102 err = -ENOPKG;
1103 if (!dlci_ioctl_hook)
1104 request_module("dlci");
1105
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001106 mutex_lock(&dlci_ioctl_mutex);
1107 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001109 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 break;
1111 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001112 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001114 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 return err;
1116}
1117
1118int sock_create_lite(int family, int type, int protocol, struct socket **res)
1119{
1120 int err;
1121 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 err = security_socket_create(family, type, protocol, 1);
1124 if (err)
1125 goto out;
1126
1127 sock = sock_alloc();
1128 if (!sock) {
1129 err = -ENOMEM;
1130 goto out;
1131 }
1132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001134 err = security_socket_post_create(sock, family, type, protocol, 1);
1135 if (err)
1136 goto out_release;
1137
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138out:
1139 *res = sock;
1140 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001141out_release:
1142 sock_release(sock);
1143 sock = NULL;
1144 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001146EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001149static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150{
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001151 unsigned int ll_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 struct socket *sock;
1153
1154 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001155 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001157 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001158
1159 if (sk_valid_ll(sock->sk)) {
1160 /* this socket can poll_ll so tell the system call */
1161 ll_flag = POLL_LL;
1162
1163 /* once, only if requested by syscall */
1164 if (wait && (wait->_key & POLL_LL))
1165 sk_poll_ll(sock->sk, 1);
1166 }
1167
1168 return ll_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169}
1170
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001171static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001173 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174
1175 return sock->ops->mmap(file, sock, vma);
1176}
1177
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001178static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 sock_release(SOCKET_I(inode));
1181 return 0;
1182}
1183
1184/*
1185 * Update the socket async list
1186 *
1187 * Fasync_list locking strategy.
1188 *
1189 * 1. fasync_list is modified only under process context socket lock
1190 * i.e. under semaphore.
1191 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001192 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 */
1194
1195static int sock_fasync(int fd, struct file *filp, int on)
1196{
Eric Dumazet989a2972010-04-14 09:55:35 +00001197 struct socket *sock = filp->private_data;
1198 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001199 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
Eric Dumazet989a2972010-04-14 09:55:35 +00001201 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
1204 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001205 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1206 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
Eric Dumazeteaefd112011-02-18 03:26:36 +00001208 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001209 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001210 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001211 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212
Eric Dumazet989a2972010-04-14 09:55:35 +00001213 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 return 0;
1215}
1216
Eric Dumazet43815482010-04-29 11:01:49 +00001217/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218
1219int sock_wake_async(struct socket *sock, int how, int band)
1220{
Eric Dumazet43815482010-04-29 11:01:49 +00001221 struct socket_wq *wq;
1222
1223 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001225 rcu_read_lock();
1226 wq = rcu_dereference(sock->wq);
1227 if (!wq || !wq->fasync_list) {
1228 rcu_read_unlock();
1229 return -1;
1230 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001231 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001232 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1234 break;
1235 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001236 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1238 break;
1239 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001240 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001241call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001242 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001244 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001245 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 }
Eric Dumazet43815482010-04-29 11:01:49 +00001247 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 return 0;
1249}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001250EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
Pavel Emelyanov721db932010-09-29 16:06:32 +04001252int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001253 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254{
1255 int err;
1256 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001257 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
1259 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001260 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 */
1262 if (family < 0 || family >= NPROTO)
1263 return -EAFNOSUPPORT;
1264 if (type < 0 || type >= SOCK_MAX)
1265 return -EINVAL;
1266
1267 /* Compatibility.
1268
1269 This uglymoron is moved from INET layer to here to avoid
1270 deadlock in module load.
1271 */
1272 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001273 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 if (!warned) {
1275 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001276 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1277 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 }
1279 family = PF_PACKET;
1280 }
1281
1282 err = security_socket_create(family, type, protocol, kern);
1283 if (err)
1284 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001285
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001286 /*
1287 * Allocate the socket and allow the family to set things up. if
1288 * the protocol is 0, the family is instructed to select an appropriate
1289 * default.
1290 */
1291 sock = sock_alloc();
1292 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001293 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001294 return -ENFILE; /* Not exactly a match, but its the
1295 closest posix thing */
1296 }
1297
1298 sock->type = type;
1299
Johannes Berg95a5afc2008-10-16 15:24:51 -07001300#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001301 /* Attempt to load a protocol module if the find failed.
1302 *
1303 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 * requested real, full-featured networking support upon configuration.
1305 * Otherwise module support will break!
1306 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001307 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001308 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309#endif
1310
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001311 rcu_read_lock();
1312 pf = rcu_dereference(net_families[family]);
1313 err = -EAFNOSUPPORT;
1314 if (!pf)
1315 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316
1317 /*
1318 * We will call the ->create function, that possibly is in a loadable
1319 * module, so we have to bump that loadable module refcnt first.
1320 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001321 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 goto out_release;
1323
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001324 /* Now protected by module ref count */
1325 rcu_read_unlock();
1326
Eric Paris3f378b62009-11-05 22:18:14 -08001327 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001328 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001330
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 /*
1332 * Now to bump the refcnt of the [loadable] module that owns this
1333 * socket at sock_release time we decrement its refcnt.
1334 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001335 if (!try_module_get(sock->ops->owner))
1336 goto out_module_busy;
1337
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 /*
1339 * Now that we're done with the ->create function, the [loadable]
1340 * module can have its refcnt decremented
1341 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001342 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001343 err = security_socket_post_create(sock, family, type, protocol, kern);
1344 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001345 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001346 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001348 return 0;
1349
1350out_module_busy:
1351 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001353 sock->ops = NULL;
1354 module_put(pf->owner);
1355out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001357 return err;
1358
1359out_release:
1360 rcu_read_unlock();
1361 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001363EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
1365int sock_create(int family, int type, int protocol, struct socket **res)
1366{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001367 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001369EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370
1371int sock_create_kern(int family, int type, int protocol, struct socket **res)
1372{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001373 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001375EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001377SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378{
1379 int retval;
1380 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001381 int flags;
1382
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001383 /* Check the SOCK_* constants for consistency. */
1384 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1385 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1386 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1387 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1388
Ulrich Dreppera677a032008-07-23 21:29:17 -07001389 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001390 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001391 return -EINVAL;
1392 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001394 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1395 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1396
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 retval = sock_create(family, type, protocol, &sock);
1398 if (retval < 0)
1399 goto out;
1400
Ulrich Drepper77d27202008-07-23 21:29:35 -07001401 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 if (retval < 0)
1403 goto out_release;
1404
1405out:
1406 /* It may be already another descriptor 8) Not kernel problem. */
1407 return retval;
1408
1409out_release:
1410 sock_release(sock);
1411 return retval;
1412}
1413
1414/*
1415 * Create a pair of connected sockets.
1416 */
1417
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001418SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1419 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420{
1421 struct socket *sock1, *sock2;
1422 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001423 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001424 int flags;
1425
1426 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001427 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001428 return -EINVAL;
1429 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001431 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1432 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1433
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 /*
1435 * Obtain the first socket and check if the underlying protocol
1436 * supports the socketpair call.
1437 */
1438
1439 err = sock_create(family, type, protocol, &sock1);
1440 if (err < 0)
1441 goto out;
1442
1443 err = sock_create(family, type, protocol, &sock2);
1444 if (err < 0)
1445 goto out_release_1;
1446
1447 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001448 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 goto out_release_both;
1450
Al Viro28407632012-08-17 23:54:15 -04001451 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001452 if (unlikely(fd1 < 0)) {
1453 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001455 }
Al Viro28407632012-08-17 23:54:15 -04001456 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001457 if (unlikely(fd2 < 0)) {
1458 err = fd2;
Al Viro28407632012-08-17 23:54:15 -04001459 put_unused_fd(fd1);
1460 goto out_release_both;
1461 }
1462
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001463 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001464 if (unlikely(IS_ERR(newfile1))) {
1465 err = PTR_ERR(newfile1);
1466 put_unused_fd(fd1);
1467 put_unused_fd(fd2);
1468 goto out_release_both;
1469 }
1470
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001471 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001472 if (IS_ERR(newfile2)) {
1473 err = PTR_ERR(newfile2);
Al Viro198de4d2009-08-05 19:29:23 +04001474 fput(newfile1);
1475 put_unused_fd(fd1);
Al Viro28407632012-08-17 23:54:15 -04001476 put_unused_fd(fd2);
Al Viro198de4d2009-08-05 19:29:23 +04001477 sock_release(sock2);
1478 goto out;
Al Virodb349502007-02-07 01:48:00 -05001479 }
1480
Al Viro157cf642008-12-14 04:57:47 -05001481 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001482 fd_install(fd1, newfile1);
1483 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 /* fd1 and fd2 may be already another descriptors.
1485 * Not kernel problem.
1486 */
1487
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001488 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 if (!err)
1490 err = put_user(fd2, &usockvec[1]);
1491 if (!err)
1492 return 0;
1493
1494 sys_close(fd2);
1495 sys_close(fd1);
1496 return err;
1497
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001499 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001501 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502out:
1503 return err;
1504}
1505
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506/*
1507 * Bind a name to a socket. Nothing much to do here since it's
1508 * the protocol's responsibility to handle the local address.
1509 *
1510 * We move the socket address to kernel space before we call
1511 * the protocol layer (having also checked the address is ok).
1512 */
1513
Heiko Carstens20f37032009-01-14 14:14:23 +01001514SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515{
1516 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001517 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001518 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001520 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001521 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001522 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001523 if (err >= 0) {
1524 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001525 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001526 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001527 if (!err)
1528 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001529 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001530 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001532 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001533 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 return err;
1535}
1536
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537/*
1538 * Perform a listen. Basically, we allow the protocol to do anything
1539 * necessary for a listen, and if that works, we mark the socket as
1540 * ready for listening.
1541 */
1542
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001543SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544{
1545 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001546 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001547 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001548
1549 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1550 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001551 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001552 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001553 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554
1555 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001556 if (!err)
1557 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001559 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 }
1561 return err;
1562}
1563
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564/*
1565 * For accept, we attempt to create a new socket, set up the link
1566 * with the client, wake up the client, then return the new
1567 * connected fd. We collect the address of the connector in kernel
1568 * space and move it to user at the very end. This is unclean because
1569 * we open the socket then return an error.
1570 *
1571 * 1003.1g adds the ability to recvmsg() to query connection pending
1572 * status to recvmsg. We need to add that support in a way thats
1573 * clean when we restucture accept also.
1574 */
1575
Heiko Carstens20f37032009-01-14 14:14:23 +01001576SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1577 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578{
1579 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001580 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001581 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001582 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
Ulrich Drepper77d27202008-07-23 21:29:35 -07001584 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001585 return -EINVAL;
1586
1587 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1588 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1589
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001590 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 if (!sock)
1592 goto out;
1593
1594 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001595 newsock = sock_alloc();
1596 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 goto out_put;
1598
1599 newsock->type = sock->type;
1600 newsock->ops = sock->ops;
1601
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 /*
1603 * We don't need try_module_get here, as the listening socket (sock)
1604 * has the protocol module (sock->ops->owner) held.
1605 */
1606 __module_get(newsock->ops->owner);
1607
Al Viro28407632012-08-17 23:54:15 -04001608 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001609 if (unlikely(newfd < 0)) {
1610 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001611 sock_release(newsock);
1612 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001613 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001614 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001615 if (unlikely(IS_ERR(newfile))) {
1616 err = PTR_ERR(newfile);
1617 put_unused_fd(newfd);
1618 sock_release(newsock);
1619 goto out_put;
1620 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001621
Frank Filza79af592005-09-27 15:23:38 -07001622 err = security_socket_accept(sock, newsock);
1623 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001624 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001625
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1627 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001628 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629
1630 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001631 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001632 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001634 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001636 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001637 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001639 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 }
1641
1642 /* File flags are not inherited via accept() unlike another OSes. */
1643
David S. Miller39d8c1b2006-03-20 17:13:49 -08001644 fd_install(newfd, newfile);
1645 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001648 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649out:
1650 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001651out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001652 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001653 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 goto out_put;
1655}
1656
Heiko Carstens20f37032009-01-14 14:14:23 +01001657SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1658 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001659{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001660 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001661}
1662
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663/*
1664 * Attempt to connect to a socket with the server address. The address
1665 * is in user space so we verify it is OK and move it to kernel space.
1666 *
1667 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1668 * break bindings
1669 *
1670 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1671 * other SEQPACKET protocols that take time to connect() as it doesn't
1672 * include the -EINPROGRESS status for such sockets.
1673 */
1674
Heiko Carstens20f37032009-01-14 14:14:23 +01001675SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1676 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677{
1678 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001679 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001680 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001682 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (!sock)
1684 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001685 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 if (err < 0)
1687 goto out_put;
1688
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001689 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001690 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 if (err)
1692 goto out_put;
1693
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001694 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 sock->file->f_flags);
1696out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001697 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698out:
1699 return err;
1700}
1701
1702/*
1703 * Get the local address ('name') of a socket object. Move the obtained
1704 * name to user space.
1705 */
1706
Heiko Carstens20f37032009-01-14 14:14:23 +01001707SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1708 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709{
1710 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001711 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001712 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001713
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001714 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 if (!sock)
1716 goto out;
1717
1718 err = security_socket_getsockname(sock);
1719 if (err)
1720 goto out_put;
1721
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001722 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 if (err)
1724 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001725 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
1727out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001728 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729out:
1730 return err;
1731}
1732
1733/*
1734 * Get the remote address ('name') of a socket object. Move the obtained
1735 * name to user space.
1736 */
1737
Heiko Carstens20f37032009-01-14 14:14:23 +01001738SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1739 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740{
1741 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001742 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001743 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001745 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1746 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 err = security_socket_getpeername(sock);
1748 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001749 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 return err;
1751 }
1752
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001753 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001754 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001755 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001757 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001758 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001759 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 }
1761 return err;
1762}
1763
1764/*
1765 * Send a datagram to a given address. We move the address into kernel
1766 * space and check the user space data area is readable before invoking
1767 * the protocol.
1768 */
1769
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001770SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001771 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001772 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773{
1774 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001775 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 int err;
1777 struct msghdr msg;
1778 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001779 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001780
Linus Torvalds253eacc2010-10-30 16:43:10 -07001781 if (len > INT_MAX)
1782 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001783 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1784 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001785 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001786
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001787 iov.iov_base = buff;
1788 iov.iov_len = len;
1789 msg.msg_name = NULL;
1790 msg.msg_iov = &iov;
1791 msg.msg_iovlen = 1;
1792 msg.msg_control = NULL;
1793 msg.msg_controllen = 0;
1794 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001795 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001796 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001797 if (err < 0)
1798 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001799 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001800 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 }
1802 if (sock->file->f_flags & O_NONBLOCK)
1803 flags |= MSG_DONTWAIT;
1804 msg.msg_flags = flags;
1805 err = sock_sendmsg(sock, &msg, len);
1806
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001807out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001808 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001809out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 return err;
1811}
1812
1813/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001814 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 */
1816
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001817SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001818 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819{
1820 return sys_sendto(fd, buff, len, flags, NULL, 0);
1821}
1822
1823/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001824 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 * sender. We verify the buffers are writable and if needed move the
1826 * sender address from kernel to user space.
1827 */
1828
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001829SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001830 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001831 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832{
1833 struct socket *sock;
1834 struct iovec iov;
1835 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001836 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001837 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001838 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839
Linus Torvalds253eacc2010-10-30 16:43:10 -07001840 if (size > INT_MAX)
1841 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001842 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001844 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001846 msg.msg_control = NULL;
1847 msg.msg_controllen = 0;
1848 msg.msg_iovlen = 1;
1849 msg.msg_iov = &iov;
1850 iov.iov_len = size;
1851 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001852 msg.msg_name = (struct sockaddr *)&address;
1853 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 if (sock->file->f_flags & O_NONBLOCK)
1855 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001856 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001858 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001859 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001860 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001861 if (err2 < 0)
1862 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001864
1865 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001866out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 return err;
1868}
1869
1870/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001871 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 */
1873
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001874asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001875 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876{
1877 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1878}
1879
1880/*
1881 * Set a socket option. Because we don't know the option lengths we have
1882 * to pass the user mode parameter for the protocols to sort out.
1883 */
1884
Heiko Carstens20f37032009-01-14 14:14:23 +01001885SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1886 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001888 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 struct socket *sock;
1890
1891 if (optlen < 0)
1892 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001893
1894 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1895 if (sock != NULL) {
1896 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001897 if (err)
1898 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899
1900 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001901 err =
1902 sock_setsockopt(sock, level, optname, optval,
1903 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001905 err =
1906 sock->ops->setsockopt(sock, level, optname, optval,
1907 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001908out_put:
1909 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 }
1911 return err;
1912}
1913
1914/*
1915 * Get a socket option. Because we don't know the option lengths we have
1916 * to pass a user mode parameter for the protocols to sort out.
1917 */
1918
Heiko Carstens20f37032009-01-14 14:14:23 +01001919SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1920 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001922 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 struct socket *sock;
1924
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001925 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1926 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001927 err = security_socket_getsockopt(sock, level, optname);
1928 if (err)
1929 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930
1931 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001932 err =
1933 sock_getsockopt(sock, level, optname, optval,
1934 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001936 err =
1937 sock->ops->getsockopt(sock, level, optname, optval,
1938 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001939out_put:
1940 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 }
1942 return err;
1943}
1944
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945/*
1946 * Shutdown a socket.
1947 */
1948
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001949SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001951 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 struct socket *sock;
1953
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001954 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1955 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001957 if (!err)
1958 err = sock->ops->shutdown(sock, how);
1959 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 }
1961 return err;
1962}
1963
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001964/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 * fields which are the same type (int / unsigned) on our platforms.
1966 */
1967#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1968#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1969#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1970
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001971struct used_address {
1972 struct sockaddr_storage name;
1973 unsigned int name_len;
1974};
1975
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001976static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001977 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001978 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001980 struct compat_msghdr __user *msg_compat =
1981 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001982 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001984 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001985 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1986 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001988 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001989
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 err = -EFAULT;
1991 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001992 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001994 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 return -EFAULT;
1996
Anton Blanchard228e5482011-05-02 20:21:35 +00001997 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02001998 err = -EMSGSIZE;
1999 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2000 goto out;
2001 err = -ENOMEM;
2002 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2003 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00002005 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 }
2007
2008 /* This will also move the address data into kernel space */
2009 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002010 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002012 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002013 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 goto out_freeiov;
2015 total_len = err;
2016
2017 err = -ENOBUFS;
2018
Anton Blanchard228e5482011-05-02 20:21:35 +00002019 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002021 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002023 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002024 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002025 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 if (err)
2027 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002028 ctl_buf = msg_sys->msg_control;
2029 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002031 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002033 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034 goto out_freeiov;
2035 }
2036 err = -EFAULT;
2037 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002038 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2040 * checking falls down on this.
2041 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002042 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002043 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002044 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002046 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002048 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
2050 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002051 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002052 /*
2053 * If this is sendmmsg() and current destination address is same as
2054 * previously succeeded address, omit asking LSM's decision.
2055 * used_address->name_len is initialized to UINT_MAX so that the first
2056 * destination address never matches.
2057 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002058 if (used_address && msg_sys->msg_name &&
2059 used_address->name_len == msg_sys->msg_namelen &&
2060 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002061 used_address->name_len)) {
2062 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2063 goto out_freectl;
2064 }
2065 err = sock_sendmsg(sock, msg_sys, total_len);
2066 /*
2067 * If this is sendmmsg() and sending to current destination address was
2068 * successful, remember it.
2069 */
2070 if (used_address && err >= 0) {
2071 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002072 if (msg_sys->msg_name)
2073 memcpy(&used_address->name, msg_sys->msg_name,
2074 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002075 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
2077out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002078 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2080out_freeiov:
2081 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002082 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002083out:
2084 return err;
2085}
2086
2087/*
2088 * BSD sendmsg interface
2089 */
2090
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002091long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002092{
2093 int fput_needed, err;
2094 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002095 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002096
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002097 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002098 if (!sock)
2099 goto out;
2100
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002101 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002102
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002103 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002104out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 return err;
2106}
2107
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002108SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2109{
2110 if (flags & MSG_CMSG_COMPAT)
2111 return -EINVAL;
2112 return __sys_sendmsg(fd, msg, flags);
2113}
2114
Anton Blanchard228e5482011-05-02 20:21:35 +00002115/*
2116 * Linux sendmmsg interface
2117 */
2118
2119int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2120 unsigned int flags)
2121{
2122 int fput_needed, err, datagrams;
2123 struct socket *sock;
2124 struct mmsghdr __user *entry;
2125 struct compat_mmsghdr __user *compat_entry;
2126 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002127 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002128
Anton Blanchard98382f42011-08-04 14:07:39 +00002129 if (vlen > UIO_MAXIOV)
2130 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002131
2132 datagrams = 0;
2133
2134 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2135 if (!sock)
2136 return err;
2137
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002138 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002139 entry = mmsg;
2140 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002141 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002142
2143 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002144 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002145 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2146 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002147 if (err < 0)
2148 break;
2149 err = __put_user(err, &compat_entry->msg_len);
2150 ++compat_entry;
2151 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002152 err = ___sys_sendmsg(sock,
2153 (struct msghdr __user *)entry,
2154 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002155 if (err < 0)
2156 break;
2157 err = put_user(err, &entry->msg_len);
2158 ++entry;
2159 }
2160
2161 if (err)
2162 break;
2163 ++datagrams;
2164 }
2165
Anton Blanchard228e5482011-05-02 20:21:35 +00002166 fput_light(sock->file, fput_needed);
2167
Anton Blanchard728ffb82011-08-04 14:07:38 +00002168 /* We only return an error if no datagrams were able to be sent */
2169 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002170 return datagrams;
2171
Anton Blanchard228e5482011-05-02 20:21:35 +00002172 return err;
2173}
2174
2175SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2176 unsigned int, vlen, unsigned int, flags)
2177{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002178 if (flags & MSG_CMSG_COMPAT)
2179 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002180 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2181}
2182
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002183static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002184 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002185{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002186 struct compat_msghdr __user *msg_compat =
2187 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002189 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002191 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192
2193 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002194 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195
2196 /* user mode address pointers */
2197 struct sockaddr __user *uaddr;
2198 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002199
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002201 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002203 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002204 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002206 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002207 err = -EMSGSIZE;
2208 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2209 goto out;
2210 err = -ENOMEM;
2211 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2212 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002214 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215 }
2216
2217 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002218 * Save the user-mode address (verify_iovec will change the
2219 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002221
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002222 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 uaddr_len = COMPAT_NAMELEN(msg);
2224 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002225 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002227 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228 if (err < 0)
2229 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002230 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002232 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2233 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002234
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235 if (sock->file->f_flags & O_NONBLOCK)
2236 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002237 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2238 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239 if (err < 0)
2240 goto out_freeiov;
2241 len = err;
2242
2243 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002244 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002245 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002246 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247 if (err < 0)
2248 goto out_freeiov;
2249 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002250 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002251 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252 if (err)
2253 goto out_freeiov;
2254 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002255 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256 &msg_compat->msg_controllen);
2257 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002258 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 &msg->msg_controllen);
2260 if (err)
2261 goto out_freeiov;
2262 err = len;
2263
2264out_freeiov:
2265 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002266 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002267out:
2268 return err;
2269}
2270
2271/*
2272 * BSD recvmsg interface
2273 */
2274
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002275long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002276{
2277 int fput_needed, err;
2278 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002279 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002280
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002281 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002282 if (!sock)
2283 goto out;
2284
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002285 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002286
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002287 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288out:
2289 return err;
2290}
2291
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002292SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2293 unsigned int, flags)
2294{
2295 if (flags & MSG_CMSG_COMPAT)
2296 return -EINVAL;
2297 return __sys_recvmsg(fd, msg, flags);
2298}
2299
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002300/*
2301 * Linux recvmmsg interface
2302 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002304int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2305 unsigned int flags, struct timespec *timeout)
2306{
2307 int fput_needed, err, datagrams;
2308 struct socket *sock;
2309 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002310 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002311 struct msghdr msg_sys;
2312 struct timespec end_time;
2313
2314 if (timeout &&
2315 poll_select_set_timeout(&end_time, timeout->tv_sec,
2316 timeout->tv_nsec))
2317 return -EINVAL;
2318
2319 datagrams = 0;
2320
2321 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2322 if (!sock)
2323 return err;
2324
2325 err = sock_error(sock->sk);
2326 if (err)
2327 goto out_put;
2328
2329 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002330 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002331
2332 while (datagrams < vlen) {
2333 /*
2334 * No need to ask LSM for more than the first datagram.
2335 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002336 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002337 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2338 &msg_sys, flags & ~MSG_WAITFORONE,
2339 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002340 if (err < 0)
2341 break;
2342 err = __put_user(err, &compat_entry->msg_len);
2343 ++compat_entry;
2344 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002345 err = ___sys_recvmsg(sock,
2346 (struct msghdr __user *)entry,
2347 &msg_sys, flags & ~MSG_WAITFORONE,
2348 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002349 if (err < 0)
2350 break;
2351 err = put_user(err, &entry->msg_len);
2352 ++entry;
2353 }
2354
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002355 if (err)
2356 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002357 ++datagrams;
2358
Brandon L Black71c5c152010-03-26 16:18:03 +00002359 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2360 if (flags & MSG_WAITFORONE)
2361 flags |= MSG_DONTWAIT;
2362
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002363 if (timeout) {
2364 ktime_get_ts(timeout);
2365 *timeout = timespec_sub(end_time, *timeout);
2366 if (timeout->tv_sec < 0) {
2367 timeout->tv_sec = timeout->tv_nsec = 0;
2368 break;
2369 }
2370
2371 /* Timeout, return less than vlen datagrams */
2372 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2373 break;
2374 }
2375
2376 /* Out of band data, return right away */
2377 if (msg_sys.msg_flags & MSG_OOB)
2378 break;
2379 }
2380
2381out_put:
2382 fput_light(sock->file, fput_needed);
2383
2384 if (err == 0)
2385 return datagrams;
2386
2387 if (datagrams != 0) {
2388 /*
2389 * We may return less entries than requested (vlen) if the
2390 * sock is non block and there aren't enough datagrams...
2391 */
2392 if (err != -EAGAIN) {
2393 /*
2394 * ... or if recvmsg returns an error after we
2395 * received some datagrams, where we record the
2396 * error to return on the next call or if the
2397 * app asks about it using getsockopt(SO_ERROR).
2398 */
2399 sock->sk->sk_err = -err;
2400 }
2401
2402 return datagrams;
2403 }
2404
2405 return err;
2406}
2407
2408SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2409 unsigned int, vlen, unsigned int, flags,
2410 struct timespec __user *, timeout)
2411{
2412 int datagrams;
2413 struct timespec timeout_sys;
2414
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002415 if (flags & MSG_CMSG_COMPAT)
2416 return -EINVAL;
2417
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002418 if (!timeout)
2419 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2420
2421 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2422 return -EFAULT;
2423
2424 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2425
2426 if (datagrams > 0 &&
2427 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2428 datagrams = -EFAULT;
2429
2430 return datagrams;
2431}
2432
2433#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434/* Argument list sizes for sys_socketcall */
2435#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002436static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002437 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2438 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2439 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002440 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002441};
2442
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443#undef AL
2444
2445/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002446 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447 *
2448 * Argument checking cleaned up. Saved 20% in size.
2449 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002450 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 */
2452
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002453SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454{
Chen Gang2950fa92013-04-07 16:55:23 +08002455 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002456 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002458 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459
Anton Blanchard228e5482011-05-02 20:21:35 +00002460 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461 return -EINVAL;
2462
Arjan van de Ven47379052009-09-28 12:57:44 -07002463 len = nargs[call];
2464 if (len > sizeof(a))
2465 return -EINVAL;
2466
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002468 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002470
Chen Gang2950fa92013-04-07 16:55:23 +08002471 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2472 if (err)
2473 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002474
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002475 a0 = a[0];
2476 a1 = a[1];
2477
2478 switch (call) {
2479 case SYS_SOCKET:
2480 err = sys_socket(a0, a1, a[2]);
2481 break;
2482 case SYS_BIND:
2483 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2484 break;
2485 case SYS_CONNECT:
2486 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2487 break;
2488 case SYS_LISTEN:
2489 err = sys_listen(a0, a1);
2490 break;
2491 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002492 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2493 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002494 break;
2495 case SYS_GETSOCKNAME:
2496 err =
2497 sys_getsockname(a0, (struct sockaddr __user *)a1,
2498 (int __user *)a[2]);
2499 break;
2500 case SYS_GETPEERNAME:
2501 err =
2502 sys_getpeername(a0, (struct sockaddr __user *)a1,
2503 (int __user *)a[2]);
2504 break;
2505 case SYS_SOCKETPAIR:
2506 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2507 break;
2508 case SYS_SEND:
2509 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2510 break;
2511 case SYS_SENDTO:
2512 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2513 (struct sockaddr __user *)a[4], a[5]);
2514 break;
2515 case SYS_RECV:
2516 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2517 break;
2518 case SYS_RECVFROM:
2519 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2520 (struct sockaddr __user *)a[4],
2521 (int __user *)a[5]);
2522 break;
2523 case SYS_SHUTDOWN:
2524 err = sys_shutdown(a0, a1);
2525 break;
2526 case SYS_SETSOCKOPT:
2527 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2528 break;
2529 case SYS_GETSOCKOPT:
2530 err =
2531 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2532 (int __user *)a[4]);
2533 break;
2534 case SYS_SENDMSG:
2535 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2536 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002537 case SYS_SENDMMSG:
2538 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2539 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002540 case SYS_RECVMSG:
2541 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2542 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002543 case SYS_RECVMMSG:
2544 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2545 (struct timespec __user *)a[4]);
2546 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002547 case SYS_ACCEPT4:
2548 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2549 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002550 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002551 default:
2552 err = -EINVAL;
2553 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554 }
2555 return err;
2556}
2557
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002558#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002560/**
2561 * sock_register - add a socket protocol handler
2562 * @ops: description of protocol
2563 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002564 * This function is called by a protocol handler that wants to
2565 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002566 * socket interface. The value ops->family coresponds to the
2567 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002568 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002569int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570{
2571 int err;
2572
2573 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002574 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2575 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576 return -ENOBUFS;
2577 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002578
2579 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002580 if (rcu_dereference_protected(net_families[ops->family],
2581 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002582 err = -EEXIST;
2583 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002584 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002585 err = 0;
2586 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002587 spin_unlock(&net_family_lock);
2588
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002589 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590 return err;
2591}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002592EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002594/**
2595 * sock_unregister - remove a protocol handler
2596 * @family: protocol family to remove
2597 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002598 * This function is called by a protocol handler that wants to
2599 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002600 * new socket creation.
2601 *
2602 * If protocol handler is a module, then it can use module reference
2603 * counts to protect against new references. If protocol handler is not
2604 * a module then it needs to provide its own protection in
2605 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002606 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002607void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002609 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002611 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002612 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002613 spin_unlock(&net_family_lock);
2614
2615 synchronize_rcu();
2616
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002617 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002619EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002620
Andi Kleen77d76ea2005-12-22 12:43:42 -08002621static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002623 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002624 /*
2625 * Initialize the network sysctl infrastructure.
2626 */
2627 err = net_sysctl_init();
2628 if (err)
2629 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002630
Linus Torvalds1da177e2005-04-16 15:20:36 -07002631 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002632 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002633 */
2634 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635
2636 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002637 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638 */
2639
2640 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002641
2642 err = register_filesystem(&sock_fs_type);
2643 if (err)
2644 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002645 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002646 if (IS_ERR(sock_mnt)) {
2647 err = PTR_ERR(sock_mnt);
2648 goto out_mount;
2649 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002650
2651 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652 */
2653
2654#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002655 err = netfilter_init();
2656 if (err)
2657 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002659
Richard Cochranc1f19b52010-07-17 08:49:36 +00002660#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2661 skb_timestamping_init();
2662#endif
2663
Nick Pigginb3e19d92011-01-07 17:50:11 +11002664out:
2665 return err;
2666
2667out_mount:
2668 unregister_filesystem(&sock_fs_type);
2669out_fs:
2670 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002671}
2672
Andi Kleen77d76ea2005-12-22 12:43:42 -08002673core_initcall(sock_init); /* early initcall */
2674
Linus Torvalds1da177e2005-04-16 15:20:36 -07002675#ifdef CONFIG_PROC_FS
2676void socket_seq_show(struct seq_file *seq)
2677{
2678 int cpu;
2679 int counter = 0;
2680
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002681 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002682 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002683
2684 /* It can be negative, by the way. 8) */
2685 if (counter < 0)
2686 counter = 0;
2687
2688 seq_printf(seq, "sockets: used %d\n", counter);
2689}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002690#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002691
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002692#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002693static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002694 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002695{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002696 mm_segment_t old_fs = get_fs();
2697 struct timeval ktv;
2698 int err;
2699
2700 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002701 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002702 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002703 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002704 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002705
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002706 return err;
2707}
2708
Arnd Bergmann6b960182009-11-06 23:10:54 -08002709static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002710 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002711{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002712 mm_segment_t old_fs = get_fs();
2713 struct timespec kts;
2714 int err;
2715
2716 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002717 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002718 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002719 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002720 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002721
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002722 return err;
2723}
2724
Arnd Bergmann6b960182009-11-06 23:10:54 -08002725static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002726{
2727 struct ifreq __user *uifr;
2728 int err;
2729
2730 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002731 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002732 return -EFAULT;
2733
Arnd Bergmann6b960182009-11-06 23:10:54 -08002734 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002735 if (err)
2736 return err;
2737
Arnd Bergmann6b960182009-11-06 23:10:54 -08002738 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002739 return -EFAULT;
2740
2741 return 0;
2742}
2743
Arnd Bergmann6b960182009-11-06 23:10:54 -08002744static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002745{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002746 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002747 struct ifconf ifc;
2748 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002749 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002750 struct ifreq __user *ifr;
2751 unsigned int i, j;
2752 int err;
2753
Arnd Bergmann6b960182009-11-06 23:10:54 -08002754 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002755 return -EFAULT;
2756
Mathias Krause43da5f22012-08-15 11:31:57 +00002757 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002758 if (ifc32.ifcbuf == 0) {
2759 ifc32.ifc_len = 0;
2760 ifc.ifc_len = 0;
2761 ifc.ifc_req = NULL;
2762 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2763 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002764 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2765 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002766 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2767 ifc.ifc_len = len;
2768 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2769 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002770 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002771 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002772 return -EFAULT;
2773 ifr++;
2774 ifr32++;
2775 }
2776 }
2777 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2778 return -EFAULT;
2779
Arnd Bergmann6b960182009-11-06 23:10:54 -08002780 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002781 if (err)
2782 return err;
2783
2784 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2785 return -EFAULT;
2786
2787 ifr = ifc.ifc_req;
2788 ifr32 = compat_ptr(ifc32.ifcbuf);
2789 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002790 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2791 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2792 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002793 return -EFAULT;
2794 ifr32++;
2795 ifr++;
2796 }
2797
2798 if (ifc32.ifcbuf == 0) {
2799 /* Translate from 64-bit structure multiple to
2800 * a 32-bit one.
2801 */
2802 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002803 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002804 ifc32.ifc_len = i;
2805 } else {
2806 ifc32.ifc_len = i;
2807 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002808 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002809 return -EFAULT;
2810
2811 return 0;
2812}
2813
Arnd Bergmann6b960182009-11-06 23:10:54 -08002814static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002815{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002816 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2817 bool convert_in = false, convert_out = false;
2818 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2819 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002820 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002821 u32 rule_cnt = 0, actual_rule_cnt;
2822 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002823 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002824 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002825
2826 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2827 return -EFAULT;
2828
Ben Hutchings3a7da392011-03-17 07:34:32 +00002829 compat_rxnfc = compat_ptr(data);
2830
2831 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002832 return -EFAULT;
2833
Ben Hutchings3a7da392011-03-17 07:34:32 +00002834 /* Most ethtool structures are defined without padding.
2835 * Unfortunately struct ethtool_rxnfc is an exception.
2836 */
2837 switch (ethcmd) {
2838 default:
2839 break;
2840 case ETHTOOL_GRXCLSRLALL:
2841 /* Buffer size is variable */
2842 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2843 return -EFAULT;
2844 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2845 return -ENOMEM;
2846 buf_size += rule_cnt * sizeof(u32);
2847 /* fall through */
2848 case ETHTOOL_GRXRINGS:
2849 case ETHTOOL_GRXCLSRLCNT:
2850 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002851 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002852 convert_out = true;
2853 /* fall through */
2854 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002855 buf_size += sizeof(struct ethtool_rxnfc);
2856 convert_in = true;
2857 break;
2858 }
2859
2860 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002861 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002862
2863 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2864 return -EFAULT;
2865
2866 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2867 &ifr->ifr_ifru.ifru_data))
2868 return -EFAULT;
2869
2870 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002871 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002872 * fs.ring_cookie and at the end of fs, but nowhere else.
2873 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002874 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2875 sizeof(compat_rxnfc->fs.m_ext) !=
2876 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2877 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002878 BUILD_BUG_ON(
2879 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2880 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2881 offsetof(struct ethtool_rxnfc, fs.location) -
2882 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2883
2884 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002885 (void __user *)(&rxnfc->fs.m_ext + 1) -
2886 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002887 copy_in_user(&rxnfc->fs.ring_cookie,
2888 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002889 (void __user *)(&rxnfc->fs.location + 1) -
2890 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002891 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2892 sizeof(rxnfc->rule_cnt)))
2893 return -EFAULT;
2894 }
2895
2896 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2897 if (ret)
2898 return ret;
2899
2900 if (convert_out) {
2901 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002902 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2903 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002904 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2905 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002906 (const void __user *)(&rxnfc->fs.location + 1) -
2907 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002908 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2909 sizeof(rxnfc->rule_cnt)))
2910 return -EFAULT;
2911
2912 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2913 /* As an optimisation, we only copy the actual
2914 * number of rules that the underlying
2915 * function returned. Since Mallory might
2916 * change the rule count in user memory, we
2917 * check that it is less than the rule count
2918 * originally given (as the user buffer size),
2919 * which has been range-checked.
2920 */
2921 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2922 return -EFAULT;
2923 if (actual_rule_cnt < rule_cnt)
2924 rule_cnt = actual_rule_cnt;
2925 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2926 &rxnfc->rule_locs[0],
2927 rule_cnt * sizeof(u32)))
2928 return -EFAULT;
2929 }
2930 }
2931
2932 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002933}
2934
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002935static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2936{
2937 void __user *uptr;
2938 compat_uptr_t uptr32;
2939 struct ifreq __user *uifr;
2940
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002941 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002942 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2943 return -EFAULT;
2944
2945 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2946 return -EFAULT;
2947
2948 uptr = compat_ptr(uptr32);
2949
2950 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2951 return -EFAULT;
2952
2953 return dev_ioctl(net, SIOCWANDEV, uifr);
2954}
2955
Arnd Bergmann6b960182009-11-06 23:10:54 -08002956static int bond_ioctl(struct net *net, unsigned int cmd,
2957 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002958{
2959 struct ifreq kifr;
2960 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002961 mm_segment_t old_fs;
2962 int err;
2963 u32 data;
2964 void __user *datap;
2965
2966 switch (cmd) {
2967 case SIOCBONDENSLAVE:
2968 case SIOCBONDRELEASE:
2969 case SIOCBONDSETHWADDR:
2970 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002971 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002972 return -EFAULT;
2973
2974 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002975 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002976 err = dev_ioctl(net, cmd,
2977 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002978 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002979
2980 return err;
2981 case SIOCBONDSLAVEINFOQUERY:
2982 case SIOCBONDINFOQUERY:
2983 uifr = compat_alloc_user_space(sizeof(*uifr));
2984 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2985 return -EFAULT;
2986
2987 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2988 return -EFAULT;
2989
2990 datap = compat_ptr(data);
2991 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2992 return -EFAULT;
2993
Arnd Bergmann6b960182009-11-06 23:10:54 -08002994 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002995 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002996 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002997 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002998}
2999
Arnd Bergmann6b960182009-11-06 23:10:54 -08003000static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
3001 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003002{
3003 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003004 char tmp_buf[IFNAMSIZ];
3005 void __user *data64;
3006 u32 data32;
3007
3008 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3009 IFNAMSIZ))
3010 return -EFAULT;
3011 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3012 return -EFAULT;
3013 data64 = compat_ptr(data32);
3014
3015 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3016
3017 /* Don't check these user accesses, just let that get trapped
3018 * in the ioctl handler instead.
3019 */
3020 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3021 IFNAMSIZ))
3022 return -EFAULT;
3023 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3024 return -EFAULT;
3025
Arnd Bergmann6b960182009-11-06 23:10:54 -08003026 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003027}
3028
Arnd Bergmann6b960182009-11-06 23:10:54 -08003029static int dev_ifsioc(struct net *net, struct socket *sock,
3030 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003031{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003032 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003033 int err;
3034
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003035 uifr = compat_alloc_user_space(sizeof(*uifr));
3036 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3037 return -EFAULT;
3038
3039 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3040
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003041 if (!err) {
3042 switch (cmd) {
3043 case SIOCGIFFLAGS:
3044 case SIOCGIFMETRIC:
3045 case SIOCGIFMTU:
3046 case SIOCGIFMEM:
3047 case SIOCGIFHWADDR:
3048 case SIOCGIFINDEX:
3049 case SIOCGIFADDR:
3050 case SIOCGIFBRDADDR:
3051 case SIOCGIFDSTADDR:
3052 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003053 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003054 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003055 case SIOCGMIIPHY:
3056 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003057 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003058 err = -EFAULT;
3059 break;
3060 }
3061 }
3062 return err;
3063}
3064
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003065static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3066 struct compat_ifreq __user *uifr32)
3067{
3068 struct ifreq ifr;
3069 struct compat_ifmap __user *uifmap32;
3070 mm_segment_t old_fs;
3071 int err;
3072
3073 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3074 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3075 err |= __get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3076 err |= __get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3077 err |= __get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3078 err |= __get_user(ifr.ifr_map.irq, &uifmap32->irq);
3079 err |= __get_user(ifr.ifr_map.dma, &uifmap32->dma);
3080 err |= __get_user(ifr.ifr_map.port, &uifmap32->port);
3081 if (err)
3082 return -EFAULT;
3083
3084 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003085 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003086 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003087 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003088
3089 if (cmd == SIOCGIFMAP && !err) {
3090 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3091 err |= __put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3092 err |= __put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3093 err |= __put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3094 err |= __put_user(ifr.ifr_map.irq, &uifmap32->irq);
3095 err |= __put_user(ifr.ifr_map.dma, &uifmap32->dma);
3096 err |= __put_user(ifr.ifr_map.port, &uifmap32->port);
3097 if (err)
3098 err = -EFAULT;
3099 }
3100 return err;
3101}
3102
3103static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3104{
3105 void __user *uptr;
3106 compat_uptr_t uptr32;
3107 struct ifreq __user *uifr;
3108
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003109 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003110 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3111 return -EFAULT;
3112
3113 if (get_user(uptr32, &uifr32->ifr_data))
3114 return -EFAULT;
3115
3116 uptr = compat_ptr(uptr32);
3117
3118 if (put_user(uptr, &uifr->ifr_data))
3119 return -EFAULT;
3120
3121 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3122}
3123
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003124struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003125 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003126 struct sockaddr rt_dst; /* target address */
3127 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3128 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003129 unsigned short rt_flags;
3130 short rt_pad2;
3131 u32 rt_pad3;
3132 unsigned char rt_tos;
3133 unsigned char rt_class;
3134 short rt_pad4;
3135 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003136 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003137 u32 rt_mtu; /* per route MTU/Window */
3138 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003139 unsigned short rt_irtt; /* Initial RTT */
3140};
3141
3142struct in6_rtmsg32 {
3143 struct in6_addr rtmsg_dst;
3144 struct in6_addr rtmsg_src;
3145 struct in6_addr rtmsg_gateway;
3146 u32 rtmsg_type;
3147 u16 rtmsg_dst_len;
3148 u16 rtmsg_src_len;
3149 u32 rtmsg_metric;
3150 u32 rtmsg_info;
3151 u32 rtmsg_flags;
3152 s32 rtmsg_ifindex;
3153};
3154
Arnd Bergmann6b960182009-11-06 23:10:54 -08003155static int routing_ioctl(struct net *net, struct socket *sock,
3156 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003157{
3158 int ret;
3159 void *r = NULL;
3160 struct in6_rtmsg r6;
3161 struct rtentry r4;
3162 char devname[16];
3163 u32 rtdev;
3164 mm_segment_t old_fs = get_fs();
3165
Arnd Bergmann6b960182009-11-06 23:10:54 -08003166 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3167 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003168 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003169 3 * sizeof(struct in6_addr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003170 ret |= __get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3171 ret |= __get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3172 ret |= __get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3173 ret |= __get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3174 ret |= __get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3175 ret |= __get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3176 ret |= __get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003177
3178 r = (void *) &r6;
3179 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003180 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003181 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003182 3 * sizeof(struct sockaddr));
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003183 ret |= __get_user(r4.rt_flags, &(ur4->rt_flags));
3184 ret |= __get_user(r4.rt_metric, &(ur4->rt_metric));
3185 ret |= __get_user(r4.rt_mtu, &(ur4->rt_mtu));
3186 ret |= __get_user(r4.rt_window, &(ur4->rt_window));
3187 ret |= __get_user(r4.rt_irtt, &(ur4->rt_irtt));
3188 ret |= __get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003189 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003190 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003191 r4.rt_dev = (char __user __force *)devname;
3192 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003193 } else
3194 r4.rt_dev = NULL;
3195
3196 r = (void *) &r4;
3197 }
3198
3199 if (ret) {
3200 ret = -EFAULT;
3201 goto out;
3202 }
3203
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003204 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003205 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003206 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003207
3208out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003209 return ret;
3210}
3211
3212/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3213 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003214 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003215 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003216static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003217{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003218 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003219
Arnd Bergmann6b960182009-11-06 23:10:54 -08003220 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003221 return -EFAULT;
3222 if (tmp == BRCTL_GET_VERSION)
3223 return BRCTL_VERSION + 1;
3224 return -EINVAL;
3225}
3226
Arnd Bergmann6b960182009-11-06 23:10:54 -08003227static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3228 unsigned int cmd, unsigned long arg)
3229{
3230 void __user *argp = compat_ptr(arg);
3231 struct sock *sk = sock->sk;
3232 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003233
Arnd Bergmann6b960182009-11-06 23:10:54 -08003234 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3235 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003236
Arnd Bergmann6b960182009-11-06 23:10:54 -08003237 switch (cmd) {
3238 case SIOCSIFBR:
3239 case SIOCGIFBR:
3240 return old_bridge_ioctl(argp);
3241 case SIOCGIFNAME:
3242 return dev_ifname32(net, argp);
3243 case SIOCGIFCONF:
3244 return dev_ifconf(net, argp);
3245 case SIOCETHTOOL:
3246 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003247 case SIOCWANDEV:
3248 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003249 case SIOCGIFMAP:
3250 case SIOCSIFMAP:
3251 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003252 case SIOCBONDENSLAVE:
3253 case SIOCBONDRELEASE:
3254 case SIOCBONDSETHWADDR:
3255 case SIOCBONDSLAVEINFOQUERY:
3256 case SIOCBONDINFOQUERY:
3257 case SIOCBONDCHANGEACTIVE:
3258 return bond_ioctl(net, cmd, argp);
3259 case SIOCADDRT:
3260 case SIOCDELRT:
3261 return routing_ioctl(net, sock, cmd, argp);
3262 case SIOCGSTAMP:
3263 return do_siocgstamp(net, sock, cmd, argp);
3264 case SIOCGSTAMPNS:
3265 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003266 case SIOCSHWTSTAMP:
3267 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003268
Arnd Bergmann6b960182009-11-06 23:10:54 -08003269 case FIOSETOWN:
3270 case SIOCSPGRP:
3271 case FIOGETOWN:
3272 case SIOCGPGRP:
3273 case SIOCBRADDBR:
3274 case SIOCBRDELBR:
3275 case SIOCGIFVLAN:
3276 case SIOCSIFVLAN:
3277 case SIOCADDDLCI:
3278 case SIOCDELDLCI:
3279 return sock_ioctl(file, cmd, arg);
3280
3281 case SIOCGIFFLAGS:
3282 case SIOCSIFFLAGS:
3283 case SIOCGIFMETRIC:
3284 case SIOCSIFMETRIC:
3285 case SIOCGIFMTU:
3286 case SIOCSIFMTU:
3287 case SIOCGIFMEM:
3288 case SIOCSIFMEM:
3289 case SIOCGIFHWADDR:
3290 case SIOCSIFHWADDR:
3291 case SIOCADDMULTI:
3292 case SIOCDELMULTI:
3293 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003294 case SIOCGIFADDR:
3295 case SIOCSIFADDR:
3296 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003297 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003298 case SIOCGIFBRDADDR:
3299 case SIOCSIFBRDADDR:
3300 case SIOCGIFDSTADDR:
3301 case SIOCSIFDSTADDR:
3302 case SIOCGIFNETMASK:
3303 case SIOCSIFNETMASK:
3304 case SIOCSIFPFLAGS:
3305 case SIOCGIFPFLAGS:
3306 case SIOCGIFTXQLEN:
3307 case SIOCSIFTXQLEN:
3308 case SIOCBRADDIF:
3309 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003310 case SIOCSIFNAME:
3311 case SIOCGMIIPHY:
3312 case SIOCGMIIREG:
3313 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003314 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003315
Arnd Bergmann6b960182009-11-06 23:10:54 -08003316 case SIOCSARP:
3317 case SIOCGARP:
3318 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003319 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003320 return sock_do_ioctl(net, sock, cmd, arg);
3321 }
3322
Arnd Bergmann6b960182009-11-06 23:10:54 -08003323 return -ENOIOCTLCMD;
3324}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003325
Eric Dumazet95c96172012-04-15 05:58:06 +00003326static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003327 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003328{
3329 struct socket *sock = file->private_data;
3330 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003331 struct sock *sk;
3332 struct net *net;
3333
3334 sk = sock->sk;
3335 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003336
3337 if (sock->ops->compat_ioctl)
3338 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3339
David S. Miller87de87d2008-06-03 09:14:03 -07003340 if (ret == -ENOIOCTLCMD &&
3341 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3342 ret = compat_wext_handle_ioctl(net, cmd, arg);
3343
Arnd Bergmann6b960182009-11-06 23:10:54 -08003344 if (ret == -ENOIOCTLCMD)
3345 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3346
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003347 return ret;
3348}
3349#endif
3350
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003351int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3352{
3353 return sock->ops->bind(sock, addr, addrlen);
3354}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003355EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003356
3357int kernel_listen(struct socket *sock, int backlog)
3358{
3359 return sock->ops->listen(sock, backlog);
3360}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003361EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003362
3363int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3364{
3365 struct sock *sk = sock->sk;
3366 int err;
3367
3368 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3369 newsock);
3370 if (err < 0)
3371 goto done;
3372
3373 err = sock->ops->accept(sock, *newsock, flags);
3374 if (err < 0) {
3375 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003376 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003377 goto done;
3378 }
3379
3380 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003381 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003382
3383done:
3384 return err;
3385}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003386EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003387
3388int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003389 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003390{
3391 return sock->ops->connect(sock, addr, addrlen, flags);
3392}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003393EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003394
3395int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3396 int *addrlen)
3397{
3398 return sock->ops->getname(sock, addr, addrlen, 0);
3399}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003400EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003401
3402int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3403 int *addrlen)
3404{
3405 return sock->ops->getname(sock, addr, addrlen, 1);
3406}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003407EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003408
3409int kernel_getsockopt(struct socket *sock, int level, int optname,
3410 char *optval, int *optlen)
3411{
3412 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003413 char __user *uoptval;
3414 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003415 int err;
3416
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003417 uoptval = (char __user __force *) optval;
3418 uoptlen = (int __user __force *) optlen;
3419
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003420 set_fs(KERNEL_DS);
3421 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003422 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003423 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003424 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3425 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003426 set_fs(oldfs);
3427 return err;
3428}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003429EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003430
3431int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003432 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003433{
3434 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003435 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003436 int err;
3437
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003438 uoptval = (char __user __force *) optval;
3439
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003440 set_fs(KERNEL_DS);
3441 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003442 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003443 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003444 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003445 optlen);
3446 set_fs(oldfs);
3447 return err;
3448}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003449EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003450
3451int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3452 size_t size, int flags)
3453{
3454 if (sock->ops->sendpage)
3455 return sock->ops->sendpage(sock, page, offset, size, flags);
3456
3457 return sock_no_sendpage(sock, page, offset, size, flags);
3458}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003459EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003460
3461int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3462{
3463 mm_segment_t oldfs = get_fs();
3464 int err;
3465
3466 set_fs(KERNEL_DS);
3467 err = sock->ops->ioctl(sock, cmd, arg);
3468 set_fs(oldfs);
3469
3470 return err;
3471}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003472EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003473
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003474int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3475{
3476 return sock->ops->shutdown(sock, how);
3477}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003478EXPORT_SYMBOL(kernel_sock_shutdown);