blob: e83c416708af5554f89c98723bba150b8750b3f0 [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 Tamir076bb0c2013-07-10 17:13:17 +0300107#include <net/busy_poll.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300108
Cong Wange0d10952013-08-01 11:10:25 +0800109#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300110unsigned int sysctl_net_busy_read __read_mostly;
111unsigned int sysctl_net_busy_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
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100224 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 err = get_user(len, ulen);
226 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (len > klen)
229 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100230 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500233 if (audit_sockaddr(klen, kaddr))
234 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700235 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 return -EFAULT;
237 }
238 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700239 * "fromlen shall refer to the value before truncation.."
240 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 */
242 return __put_user(klen, ulen);
243}
244
Christoph Lametere18b8902006-12-06 20:33:20 -0800245static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
247static struct inode *sock_alloc_inode(struct super_block *sb)
248{
249 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000250 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700251
Christoph Lametere94b1762006-12-06 20:33:17 -0800252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 if (!ei)
254 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000255 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000257 kmem_cache_free(sock_inode_cachep, ei);
258 return NULL;
259 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000260 init_waitqueue_head(&wq->wait);
261 wq->fasync_list = NULL;
262 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700263
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 ei->socket.state = SS_UNCONNECTED;
265 ei->socket.flags = 0;
266 ei->socket.ops = NULL;
267 ei->socket.sk = NULL;
268 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
270 return &ei->vfs_inode;
271}
272
273static void sock_destroy_inode(struct inode *inode)
274{
Eric Dumazet43815482010-04-29 11:01:49 +0000275 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000276 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000277
278 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000279 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800280 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000281 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282}
283
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700284static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700286 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Christoph Lametera35afb82007-05-16 22:10:57 -0700288 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700290
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291static int init_inodecache(void)
292{
293 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700294 sizeof(struct socket_alloc),
295 0,
296 (SLAB_HWCACHE_ALIGN |
297 SLAB_RECLAIM_ACCOUNT |
298 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900299 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 if (sock_inode_cachep == NULL)
301 return -ENOMEM;
302 return 0;
303}
304
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700305static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700306 .alloc_inode = sock_alloc_inode,
307 .destroy_inode = sock_destroy_inode,
308 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309};
310
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700311/*
312 * sockfs_dname() is called from d_path().
313 */
314static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315{
316 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 dentry->d_inode->i_ino);
318}
319
Al Viro3ba13d12009-02-20 06:02:22 +0000320static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700321 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322};
323
Al Viroc74a1cb2011-01-12 16:59:34 -0500324static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 int flags, const char *dev_name, void *data)
326{
327 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 &sockfs_dentry_operations, SOCKFS_MAGIC);
329}
330
331static struct vfsmount *sock_mnt __read_mostly;
332
333static struct file_system_type sock_fs_type = {
334 .name = "sockfs",
335 .mount = sockfs_mount,
336 .kill_sb = kill_anon_super,
337};
338
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339/*
340 * Obtains the first available file descriptor and sets it up for use.
341 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
349 *
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
354 */
355
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700356struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357{
Al Viro7cbe66b2009-08-05 19:59:08 +0400358 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400359 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400360 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800361
Masatake YAMATO600e1772012-08-29 10:44:29 +0000362 if (dname) {
363 name.name = dname;
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
368 }
Nick Piggin4b936882011-01-07 17:50:07 +1100369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400370 if (unlikely(!path.dentry))
371 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400372 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373
Al Viro2c48b9c2009-08-09 00:52:35 +0400374 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400375 SOCK_INODE(sock)->i_fop = &socket_file_ops;
376
Al Viro2c48b9c2009-08-09 00:52:35 +0400377 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400378 &socket_file_ops);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700379 if (unlikely(IS_ERR(file))) {
Al Virocc3808f2009-08-06 09:43:59 +0400380 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400381 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400382 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700383 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400384 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800385
386 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700387 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800388 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400389 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800390}
Al Viro56b31d12012-08-18 00:25:51 -0400391EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392
Al Viro56b31d12012-08-18 00:25:51 -0400393static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800394{
395 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400396 int fd = get_unused_fd_flags(flags);
397 if (unlikely(fd < 0))
398 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800399
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700400 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400401 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800402 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400403 return fd;
404 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400405
Al Viro28407632012-08-17 23:54:15 -0400406 put_unused_fd(fd);
407 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408}
409
John Fastabend406a3c62012-07-20 10:39:25 +0000410struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800411{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800412 if (file->f_op == &socket_file_ops)
413 return file->private_data; /* set in sock_map_fd */
414
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800415 *err = -ENOTSOCK;
416 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800417}
John Fastabend406a3c62012-07-20 10:39:25 +0000418EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800419
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700421 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 * @fd: file handle
423 * @err: pointer to an error code return
424 *
425 * The file handle passed in is locked and the socket it is bound
426 * too is returned. If an error occurs the err pointer is overwritten
427 * with a negative errno code and NULL is returned. The function checks
428 * for both invalid handles and passing a handle which is not a socket.
429 *
430 * On a success the socket object pointer is returned.
431 */
432
433struct socket *sockfd_lookup(int fd, int *err)
434{
435 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 struct socket *sock;
437
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700438 file = fget(fd);
439 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 *err = -EBADF;
441 return NULL;
442 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700443
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800444 sock = sock_from_file(file, err);
445 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 return sock;
448}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700449EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
452{
453 struct file *file;
454 struct socket *sock;
455
Hua Zhong36725582006-04-19 15:25:02 -0700456 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800457 file = fget_light(fd, fput_needed);
458 if (file) {
459 sock = sock_from_file(file, err);
460 if (sock)
461 return sock;
462 fput_light(file, *fput_needed);
463 }
464 return NULL;
465}
466
Masatake YAMATO600e1772012-08-29 10:44:29 +0000467#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
472{
473 const char *proto_name;
474 size_t proto_size;
475 int error;
476
477 error = -ENODATA;
478 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 proto_name = dentry->d_name.name;
480 proto_size = strlen(proto_name);
481
482 if (value) {
483 error = -ERANGE;
484 if (proto_size + 1 > size)
485 goto out;
486
487 strncpy(value, proto_name, proto_size + 1);
488 }
489 error = proto_size + 1;
490 }
491
492out:
493 return error;
494}
495
496static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
497 size_t size)
498{
499 ssize_t len;
500 ssize_t used = 0;
501
502 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
503 if (len < 0)
504 return len;
505 used += len;
506 if (buffer) {
507 if (size < used)
508 return -ERANGE;
509 buffer += len;
510 }
511
512 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
513 used += len;
514 if (buffer) {
515 if (size < used)
516 return -ERANGE;
517 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
518 buffer += len;
519 }
520
521 return used;
522}
523
524static const struct inode_operations sockfs_inode_ops = {
525 .getxattr = sockfs_getxattr,
526 .listxattr = sockfs_listxattr,
527};
528
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529/**
530 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700531 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
534 * NULL is returned.
535 */
536
537static struct socket *sock_alloc(void)
538{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700539 struct inode *inode;
540 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200542 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 if (!inode)
544 return NULL;
545
546 sock = SOCKET_I(inode);
547
Eric Dumazet29a020d2009-09-15 02:39:20 -0700548 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400549 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700550 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100551 inode->i_uid = current_fsuid();
552 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000553 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
Alex Shi19e8d692012-05-14 14:15:31 -0700555 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 return sock;
557}
558
559/*
560 * In theory you can't get an open on this inode, but /proc provides
561 * a back door. Remember to keep it shut otherwise you'll let the
562 * creepy crawlies in.
563 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
566{
567 return -ENXIO;
568}
569
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800570const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 .owner = THIS_MODULE,
572 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200573 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574};
575
576/**
577 * sock_release - close a socket
578 * @sock: socket to close
579 *
580 * The socket is released from the protocol stack if it has a release
581 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700582 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585void sock_release(struct socket *sock)
586{
587 if (sock->ops) {
588 struct module *owner = sock->ops->owner;
589
590 sock->ops->release(sock);
591 sock->ops = NULL;
592 module_put(owner);
593 }
594
Eric Dumazeteaefd112011-02-18 03:26:36 +0000595 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 printk(KERN_ERR "sock_release: fasync list not empty!\n");
597
Mikulas Patockab09e7862012-07-19 06:13:36 +0000598 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
599 return;
600
Alex Shi19e8d692012-05-14 14:15:31 -0700601 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 if (!sock->file) {
603 iput(SOCK_INODE(sock));
604 return;
605 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700606 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700608EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000610void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000611{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000612 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000613 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000614 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000615 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000616 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100617 if (sock_flag(sk, SOCK_WIFI_STATUS))
618 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000619}
620EXPORT_SYMBOL(sock_tx_timestamp);
621
Anton Blanchard228e5482011-05-02 20:21:35 +0000622static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
623 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
625 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626
627 si->sock = sock;
628 si->scm = NULL;
629 si->msg = msg;
630 si->size = size;
631
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 return sock->ops->sendmsg(iocb, sock, msg, size);
633}
634
Anton Blanchard228e5482011-05-02 20:21:35 +0000635static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
636 struct msghdr *msg, size_t size)
637{
638 int err = security_socket_sendmsg(sock, msg, size);
639
640 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
641}
642
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
644{
645 struct kiocb iocb;
646 struct sock_iocb siocb;
647 int ret;
648
649 init_sync_kiocb(&iocb, NULL);
650 iocb.private = &siocb;
651 ret = __sock_sendmsg(&iocb, sock, msg, size);
652 if (-EIOCBQUEUED == ret)
653 ret = wait_on_sync_kiocb(&iocb);
654 return ret;
655}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700656EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Eric Dumazet894dc242011-07-26 02:39:41 +0000658static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000659{
660 struct kiocb iocb;
661 struct sock_iocb siocb;
662 int ret;
663
664 init_sync_kiocb(&iocb, NULL);
665 iocb.private = &siocb;
666 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
667 if (-EIOCBQUEUED == ret)
668 ret = wait_on_sync_kiocb(&iocb);
669 return ret;
670}
671
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
673 struct kvec *vec, size_t num, size_t size)
674{
675 mm_segment_t oldfs = get_fs();
676 int result;
677
678 set_fs(KERNEL_DS);
679 /*
680 * the following is safe, since for compiler definitions of kvec and
681 * iovec are identical, yielding the same in-core layout and alignment
682 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700683 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 msg->msg_iovlen = num;
685 result = sock_sendmsg(sock, msg, size);
686 set_fs(oldfs);
687 return result;
688}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700689EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700691/*
692 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
693 */
694void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
695 struct sk_buff *skb)
696{
Patrick Ohly20d49472009-02-12 05:03:38 +0000697 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
698 struct timespec ts[3];
699 int empty = 1;
700 struct skb_shared_hwtstamps *shhwtstamps =
701 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700702
Patrick Ohly20d49472009-02-12 05:03:38 +0000703 /* Race occurred between timestamp enabling and packet
704 receiving. Fill in the current time for now. */
705 if (need_software_tstamp && skb->tstamp.tv64 == 0)
706 __net_timestamp(skb);
707
708 if (need_software_tstamp) {
709 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
710 struct timeval tv;
711 skb_get_timestamp(skb, &tv);
712 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
713 sizeof(tv), &tv);
714 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000715 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000716 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000717 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000718 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700719 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000720
721
722 memset(ts, 0, sizeof(ts));
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000723 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
724 ktime_to_timespec_cond(skb->tstamp, ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000725 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000726 if (shhwtstamps) {
727 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000728 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
Patrick Ohly20d49472009-02-12 05:03:38 +0000729 empty = 0;
730 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000731 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
Patrick Ohly20d49472009-02-12 05:03:38 +0000732 empty = 0;
733 }
734 if (!empty)
735 put_cmsg(msg, SOL_SOCKET,
736 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700737}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300738EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
739
Johannes Berg6e3e9392011-11-09 10:15:42 +0100740void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
741 struct sk_buff *skb)
742{
743 int ack;
744
745 if (!sock_flag(sk, SOCK_WIFI_STATUS))
746 return;
747 if (!skb->wifi_acked_valid)
748 return;
749
750 ack = skb->wifi_acked;
751
752 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
753}
754EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
755
stephen hemminger11165f12010-10-18 14:27:29 +0000756static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
757 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700758{
759 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
760 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
761 sizeof(__u32), &skb->dropcount);
762}
763
Eric Dumazet767dd032010-04-28 19:14:43 +0000764void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700765 struct sk_buff *skb)
766{
767 sock_recv_timestamp(msg, sk, skb);
768 sock_recv_drops(msg, sk, skb);
769}
Eric Dumazet767dd032010-04-28 19:14:43 +0000770EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700771
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700772static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
773 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 struct sock_iocb *si = kiocb_to_siocb(iocb);
776
777 si->sock = sock;
778 si->scm = NULL;
779 si->msg = msg;
780 si->size = size;
781 si->flags = flags;
782
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
784}
785
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700786static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
787 struct msghdr *msg, size_t size, int flags)
788{
789 int err = security_socket_recvmsg(sock, msg, size, flags);
790
791 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
792}
793
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700794int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 size_t size, int flags)
796{
797 struct kiocb iocb;
798 struct sock_iocb siocb;
799 int ret;
800
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700801 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 iocb.private = &siocb;
803 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
804 if (-EIOCBQUEUED == ret)
805 ret = wait_on_sync_kiocb(&iocb);
806 return ret;
807}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700808EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700810static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
811 size_t size, int flags)
812{
813 struct kiocb iocb;
814 struct sock_iocb siocb;
815 int ret;
816
817 init_sync_kiocb(&iocb, NULL);
818 iocb.private = &siocb;
819 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
820 if (-EIOCBQUEUED == ret)
821 ret = wait_on_sync_kiocb(&iocb);
822 return ret;
823}
824
Martin Lucinac1249c02010-12-10 00:04:05 +0000825/**
826 * kernel_recvmsg - Receive a message from a socket (kernel space)
827 * @sock: The socket to receive the message from
828 * @msg: Received message
829 * @vec: Input s/g array for message data
830 * @num: Size of input s/g array
831 * @size: Number of bytes to read
832 * @flags: Message flags (MSG_DONTWAIT, etc...)
833 *
834 * On return the msg structure contains the scatter/gather array passed in the
835 * vec argument. The array is modified so that it consists of the unfilled
836 * portion of the original array.
837 *
838 * The returned value is the total number of bytes received, or an error.
839 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700840int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
841 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842{
843 mm_segment_t oldfs = get_fs();
844 int result;
845
846 set_fs(KERNEL_DS);
847 /*
848 * the following is safe, since for compiler definitions of kvec and
849 * iovec are identical, yielding the same in-core layout and alignment
850 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700851 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 result = sock_recvmsg(sock, msg, size, flags);
853 set_fs(oldfs);
854 return result;
855}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700856EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300858static ssize_t sock_sendpage(struct file *file, struct page *page,
859 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860{
861 struct socket *sock;
862 int flags;
863
Eric Dumazetb69aee02005-09-06 14:42:45 -0700864 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
Eric Dumazet35f9c092012-04-05 03:05:35 +0000866 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
867 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
868 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
Linus Torvaldse6949582009-08-13 08:28:36 -0700870 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871}
872
Jens Axboe9c55e012007-11-06 23:30:13 -0800873static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700874 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800875 unsigned int flags)
876{
877 struct socket *sock = file->private_data;
878
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800879 if (unlikely(!sock->ops->splice_read))
880 return -EINVAL;
881
Jens Axboe9c55e012007-11-06 23:30:13 -0800882 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
883}
884
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800885static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700886 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800887{
Kent Overstreetd29c4452013-03-18 11:09:26 -0700888 if (!is_sync_kiocb(iocb))
889 BUG();
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800890
891 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800892 iocb->private = siocb;
893 return siocb;
894}
895
896static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700897 struct file *file, const struct iovec *iov,
898 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800899{
900 struct socket *sock = file->private_data;
901 size_t size = 0;
902 int i;
903
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700904 for (i = 0; i < nr_segs; i++)
905 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800906
907 msg->msg_name = NULL;
908 msg->msg_namelen = 0;
909 msg->msg_control = NULL;
910 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700911 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800912 msg->msg_iovlen = nr_segs;
913 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
914
915 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
916}
917
Badari Pulavarty027445c2006-09-30 23:28:46 -0700918static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
919 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800920{
921 struct sock_iocb siocb, *x;
922
923 if (pos != 0)
924 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700925
Kent Overstreet73a70752013-05-09 15:03:42 -0700926 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800927 return 0;
928
Badari Pulavarty027445c2006-09-30 23:28:46 -0700929
930 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800931 if (!x)
932 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700933 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800934}
935
936static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700937 struct file *file, const struct iovec *iov,
938 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800939{
940 struct socket *sock = file->private_data;
941 size_t size = 0;
942 int i;
943
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700944 for (i = 0; i < nr_segs; i++)
945 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800946
947 msg->msg_name = NULL;
948 msg->msg_namelen = 0;
949 msg->msg_control = NULL;
950 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700951 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800952 msg->msg_iovlen = nr_segs;
953 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
954 if (sock->type == SOCK_SEQPACKET)
955 msg->msg_flags |= MSG_EOR;
956
957 return __sock_sendmsg(iocb, sock, msg, size);
958}
959
Badari Pulavarty027445c2006-09-30 23:28:46 -0700960static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
961 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800963 struct sock_iocb siocb, *x;
964
965 if (pos != 0)
966 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700967
Badari Pulavarty027445c2006-09-30 23:28:46 -0700968 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800969 if (!x)
970 return -ENOMEM;
971
Badari Pulavarty027445c2006-09-30 23:28:46 -0700972 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973}
974
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975/*
976 * Atomic setting of ioctl hooks to avoid race
977 * with module unload.
978 */
979
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800980static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700981static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982
Eric W. Biederman881d9662007-09-17 11:56:21 -0700983void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800985 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800987 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988}
989EXPORT_SYMBOL(brioctl_set);
990
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800991static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700992static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993
Eric W. Biederman881d9662007-09-17 11:56:21 -0700994void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800996 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800998 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999}
1000EXPORT_SYMBOL(vlan_ioctl_set);
1001
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001002static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001003static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001005void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001007 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001009 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010}
1011EXPORT_SYMBOL(dlci_ioctl_set);
1012
Arnd Bergmann6b960182009-11-06 23:10:54 -08001013static long sock_do_ioctl(struct net *net, struct socket *sock,
1014 unsigned int cmd, unsigned long arg)
1015{
1016 int err;
1017 void __user *argp = (void __user *)arg;
1018
1019 err = sock->ops->ioctl(sock, cmd, arg);
1020
1021 /*
1022 * If this ioctl is unknown try to hand it down
1023 * to the NIC driver.
1024 */
1025 if (err == -ENOIOCTLCMD)
1026 err = dev_ioctl(net, cmd, argp);
1027
1028 return err;
1029}
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031/*
1032 * With an ioctl, arg may well be a user mode pointer, but we don't know
1033 * what to do with it - that's up to the protocol still.
1034 */
1035
1036static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1037{
1038 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001039 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 void __user *argp = (void __user *)arg;
1041 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001042 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
Eric Dumazetb69aee02005-09-06 14:42:45 -07001044 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001045 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001046 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001048 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001050#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001052 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001054#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001055 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 case FIOSETOWN:
1057 case SIOCSPGRP:
1058 err = -EFAULT;
1059 if (get_user(pid, (int __user *)argp))
1060 break;
1061 err = f_setown(sock->file, pid, 1);
1062 break;
1063 case FIOGETOWN:
1064 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001065 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001066 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 break;
1068 case SIOCGIFBR:
1069 case SIOCSIFBR:
1070 case SIOCBRADDBR:
1071 case SIOCBRDELBR:
1072 err = -ENOPKG;
1073 if (!br_ioctl_hook)
1074 request_module("bridge");
1075
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001076 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001077 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001078 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001079 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 break;
1081 case SIOCGIFVLAN:
1082 case SIOCSIFVLAN:
1083 err = -ENOPKG;
1084 if (!vlan_ioctl_hook)
1085 request_module("8021q");
1086
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001087 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001089 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001090 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092 case SIOCADDDLCI:
1093 case SIOCDELDLCI:
1094 err = -ENOPKG;
1095 if (!dlci_ioctl_hook)
1096 request_module("dlci");
1097
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001098 mutex_lock(&dlci_ioctl_mutex);
1099 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001101 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 break;
1103 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001104 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001106 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 return err;
1108}
1109
1110int sock_create_lite(int family, int type, int protocol, struct socket **res)
1111{
1112 int err;
1113 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001114
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 err = security_socket_create(family, type, protocol, 1);
1116 if (err)
1117 goto out;
1118
1119 sock = sock_alloc();
1120 if (!sock) {
1121 err = -ENOMEM;
1122 goto out;
1123 }
1124
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001126 err = security_socket_post_create(sock, family, type, protocol, 1);
1127 if (err)
1128 goto out_release;
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130out:
1131 *res = sock;
1132 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001133out_release:
1134 sock_release(sock);
1135 sock = NULL;
1136 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001138EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
1140/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001141static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142{
Eliezer Tamircbf55002013-07-08 16:20:34 +03001143 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 struct socket *sock;
1145
1146 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001147 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001149 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001150
Eliezer Tamircbf55002013-07-08 16:20:34 +03001151 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001152 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001153 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001154
1155 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001156 if (wait && (wait->_key & POLL_BUSY_LOOP))
1157 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001158 }
1159
Eliezer Tamircbf55002013-07-08 16:20:34 +03001160 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161}
1162
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001163static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001165 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
1167 return sock->ops->mmap(file, sock, vma);
1168}
1169
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001170static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 sock_release(SOCKET_I(inode));
1173 return 0;
1174}
1175
1176/*
1177 * Update the socket async list
1178 *
1179 * Fasync_list locking strategy.
1180 *
1181 * 1. fasync_list is modified only under process context socket lock
1182 * i.e. under semaphore.
1183 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001184 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 */
1186
1187static int sock_fasync(int fd, struct file *filp, int on)
1188{
Eric Dumazet989a2972010-04-14 09:55:35 +00001189 struct socket *sock = filp->private_data;
1190 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001191 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192
Eric Dumazet989a2972010-04-14 09:55:35 +00001193 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195
1196 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001197 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1198 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Eric Dumazeteaefd112011-02-18 03:26:36 +00001200 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001201 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001202 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001203 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204
Eric Dumazet989a2972010-04-14 09:55:35 +00001205 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 return 0;
1207}
1208
Eric Dumazet43815482010-04-29 11:01:49 +00001209/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210
1211int sock_wake_async(struct socket *sock, int how, int band)
1212{
Eric Dumazet43815482010-04-29 11:01:49 +00001213 struct socket_wq *wq;
1214
1215 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001217 rcu_read_lock();
1218 wq = rcu_dereference(sock->wq);
1219 if (!wq || !wq->fasync_list) {
1220 rcu_read_unlock();
1221 return -1;
1222 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001223 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001224 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1226 break;
1227 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001228 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1230 break;
1231 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001232 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001233call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001234 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001236 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001237 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 }
Eric Dumazet43815482010-04-29 11:01:49 +00001239 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 return 0;
1241}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001242EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243
Pavel Emelyanov721db932010-09-29 16:06:32 +04001244int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001245 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246{
1247 int err;
1248 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001249 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250
1251 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001252 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 */
1254 if (family < 0 || family >= NPROTO)
1255 return -EAFNOSUPPORT;
1256 if (type < 0 || type >= SOCK_MAX)
1257 return -EINVAL;
1258
1259 /* Compatibility.
1260
1261 This uglymoron is moved from INET layer to here to avoid
1262 deadlock in module load.
1263 */
1264 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001265 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 if (!warned) {
1267 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001268 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1269 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 }
1271 family = PF_PACKET;
1272 }
1273
1274 err = security_socket_create(family, type, protocol, kern);
1275 if (err)
1276 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001277
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001278 /*
1279 * Allocate the socket and allow the family to set things up. if
1280 * the protocol is 0, the family is instructed to select an appropriate
1281 * default.
1282 */
1283 sock = sock_alloc();
1284 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001285 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001286 return -ENFILE; /* Not exactly a match, but its the
1287 closest posix thing */
1288 }
1289
1290 sock->type = type;
1291
Johannes Berg95a5afc2008-10-16 15:24:51 -07001292#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001293 /* Attempt to load a protocol module if the find failed.
1294 *
1295 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 * requested real, full-featured networking support upon configuration.
1297 * Otherwise module support will break!
1298 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001299 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001300 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301#endif
1302
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001303 rcu_read_lock();
1304 pf = rcu_dereference(net_families[family]);
1305 err = -EAFNOSUPPORT;
1306 if (!pf)
1307 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
1309 /*
1310 * We will call the ->create function, that possibly is in a loadable
1311 * module, so we have to bump that loadable module refcnt first.
1312 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001313 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 goto out_release;
1315
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001316 /* Now protected by module ref count */
1317 rcu_read_unlock();
1318
Eric Paris3f378b62009-11-05 22:18:14 -08001319 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001320 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001322
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 /*
1324 * Now to bump the refcnt of the [loadable] module that owns this
1325 * socket at sock_release time we decrement its refcnt.
1326 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001327 if (!try_module_get(sock->ops->owner))
1328 goto out_module_busy;
1329
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 /*
1331 * Now that we're done with the ->create function, the [loadable]
1332 * module can have its refcnt decremented
1333 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001334 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001335 err = security_socket_post_create(sock, family, type, protocol, kern);
1336 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001337 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001338 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001340 return 0;
1341
1342out_module_busy:
1343 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001345 sock->ops = NULL;
1346 module_put(pf->owner);
1347out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001349 return err;
1350
1351out_release:
1352 rcu_read_unlock();
1353 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001355EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
1357int sock_create(int family, int type, int protocol, struct socket **res)
1358{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001359 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001361EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
1363int sock_create_kern(int family, int type, int protocol, struct socket **res)
1364{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001365 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001367EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001369SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370{
1371 int retval;
1372 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001373 int flags;
1374
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001375 /* Check the SOCK_* constants for consistency. */
1376 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1377 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1378 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1379 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1380
Ulrich Dreppera677a032008-07-23 21:29:17 -07001381 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001382 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001383 return -EINVAL;
1384 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001386 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1387 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1388
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 retval = sock_create(family, type, protocol, &sock);
1390 if (retval < 0)
1391 goto out;
1392
Ulrich Drepper77d27202008-07-23 21:29:35 -07001393 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 if (retval < 0)
1395 goto out_release;
1396
1397out:
1398 /* It may be already another descriptor 8) Not kernel problem. */
1399 return retval;
1400
1401out_release:
1402 sock_release(sock);
1403 return retval;
1404}
1405
1406/*
1407 * Create a pair of connected sockets.
1408 */
1409
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001410SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1411 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412{
1413 struct socket *sock1, *sock2;
1414 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001415 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001416 int flags;
1417
1418 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001419 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001420 return -EINVAL;
1421 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001423 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1424 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1425
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 /*
1427 * Obtain the first socket and check if the underlying protocol
1428 * supports the socketpair call.
1429 */
1430
1431 err = sock_create(family, type, protocol, &sock1);
1432 if (err < 0)
1433 goto out;
1434
1435 err = sock_create(family, type, protocol, &sock2);
1436 if (err < 0)
1437 goto out_release_1;
1438
1439 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001440 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 goto out_release_both;
1442
Al Viro28407632012-08-17 23:54:15 -04001443 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001444 if (unlikely(fd1 < 0)) {
1445 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001447 }
Al Viro28407632012-08-17 23:54:15 -04001448 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001449 if (unlikely(fd2 < 0)) {
1450 err = fd2;
Al Viro28407632012-08-17 23:54:15 -04001451 put_unused_fd(fd1);
1452 goto out_release_both;
1453 }
1454
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001455 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001456 if (unlikely(IS_ERR(newfile1))) {
1457 err = PTR_ERR(newfile1);
1458 put_unused_fd(fd1);
1459 put_unused_fd(fd2);
1460 goto out_release_both;
1461 }
1462
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001463 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001464 if (IS_ERR(newfile2)) {
1465 err = PTR_ERR(newfile2);
Al Viro198de4d2009-08-05 19:29:23 +04001466 fput(newfile1);
1467 put_unused_fd(fd1);
Al Viro28407632012-08-17 23:54:15 -04001468 put_unused_fd(fd2);
Al Viro198de4d2009-08-05 19:29:23 +04001469 sock_release(sock2);
1470 goto out;
Al Virodb349502007-02-07 01:48:00 -05001471 }
1472
Al Viro157cf642008-12-14 04:57:47 -05001473 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001474 fd_install(fd1, newfile1);
1475 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 /* fd1 and fd2 may be already another descriptors.
1477 * Not kernel problem.
1478 */
1479
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001480 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 if (!err)
1482 err = put_user(fd2, &usockvec[1]);
1483 if (!err)
1484 return 0;
1485
1486 sys_close(fd2);
1487 sys_close(fd1);
1488 return err;
1489
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001491 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001493 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494out:
1495 return err;
1496}
1497
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498/*
1499 * Bind a name to a socket. Nothing much to do here since it's
1500 * the protocol's responsibility to handle the local address.
1501 *
1502 * We move the socket address to kernel space before we call
1503 * the protocol layer (having also checked the address is ok).
1504 */
1505
Heiko Carstens20f37032009-01-14 14:14:23 +01001506SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507{
1508 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001509 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001510 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001512 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001513 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001514 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001515 if (err >= 0) {
1516 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001517 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001518 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001519 if (!err)
1520 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001521 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001522 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001524 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001525 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 return err;
1527}
1528
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529/*
1530 * Perform a listen. Basically, we allow the protocol to do anything
1531 * necessary for a listen, and if that works, we mark the socket as
1532 * ready for listening.
1533 */
1534
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001535SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536{
1537 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001538 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001539 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001540
1541 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1542 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001543 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001544 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001545 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546
1547 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001548 if (!err)
1549 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001551 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552 }
1553 return err;
1554}
1555
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556/*
1557 * For accept, we attempt to create a new socket, set up the link
1558 * with the client, wake up the client, then return the new
1559 * connected fd. We collect the address of the connector in kernel
1560 * space and move it to user at the very end. This is unclean because
1561 * we open the socket then return an error.
1562 *
1563 * 1003.1g adds the ability to recvmsg() to query connection pending
1564 * status to recvmsg. We need to add that support in a way thats
1565 * clean when we restucture accept also.
1566 */
1567
Heiko Carstens20f37032009-01-14 14:14:23 +01001568SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1569 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570{
1571 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001572 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001573 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001574 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575
Ulrich Drepper77d27202008-07-23 21:29:35 -07001576 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001577 return -EINVAL;
1578
1579 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1580 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1581
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001582 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 if (!sock)
1584 goto out;
1585
1586 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001587 newsock = sock_alloc();
1588 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 goto out_put;
1590
1591 newsock->type = sock->type;
1592 newsock->ops = sock->ops;
1593
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594 /*
1595 * We don't need try_module_get here, as the listening socket (sock)
1596 * has the protocol module (sock->ops->owner) held.
1597 */
1598 __module_get(newsock->ops->owner);
1599
Al Viro28407632012-08-17 23:54:15 -04001600 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001601 if (unlikely(newfd < 0)) {
1602 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001603 sock_release(newsock);
1604 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001605 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001606 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001607 if (unlikely(IS_ERR(newfile))) {
1608 err = PTR_ERR(newfile);
1609 put_unused_fd(newfd);
1610 sock_release(newsock);
1611 goto out_put;
1612 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001613
Frank Filza79af592005-09-27 15:23:38 -07001614 err = security_socket_accept(sock, newsock);
1615 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001616 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001617
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1619 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001620 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621
1622 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001623 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001624 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001626 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001628 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001629 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001631 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 }
1633
1634 /* File flags are not inherited via accept() unlike another OSes. */
1635
David S. Miller39d8c1b2006-03-20 17:13:49 -08001636 fd_install(newfd, newfile);
1637 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001640 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641out:
1642 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001643out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001644 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001645 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 goto out_put;
1647}
1648
Heiko Carstens20f37032009-01-14 14:14:23 +01001649SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1650 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001651{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001652 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001653}
1654
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655/*
1656 * Attempt to connect to a socket with the server address. The address
1657 * is in user space so we verify it is OK and move it to kernel space.
1658 *
1659 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1660 * break bindings
1661 *
1662 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1663 * other SEQPACKET protocols that take time to connect() as it doesn't
1664 * include the -EINPROGRESS status for such sockets.
1665 */
1666
Heiko Carstens20f37032009-01-14 14:14:23 +01001667SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1668 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669{
1670 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001671 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001672 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001674 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 if (!sock)
1676 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001677 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 if (err < 0)
1679 goto out_put;
1680
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001681 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001682 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (err)
1684 goto out_put;
1685
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001686 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 sock->file->f_flags);
1688out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001689 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690out:
1691 return err;
1692}
1693
1694/*
1695 * Get the local address ('name') of a socket object. Move the obtained
1696 * name to user space.
1697 */
1698
Heiko Carstens20f37032009-01-14 14:14:23 +01001699SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1700 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701{
1702 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001703 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001704 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001705
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001706 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 if (!sock)
1708 goto out;
1709
1710 err = security_socket_getsockname(sock);
1711 if (err)
1712 goto out_put;
1713
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001714 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 if (err)
1716 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001717 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718
1719out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001720 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721out:
1722 return err;
1723}
1724
1725/*
1726 * Get the remote address ('name') of a socket object. Move the obtained
1727 * name to user space.
1728 */
1729
Heiko Carstens20f37032009-01-14 14:14:23 +01001730SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1731 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732{
1733 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001734 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001735 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001737 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1738 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 err = security_socket_getpeername(sock);
1740 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001741 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 return err;
1743 }
1744
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001745 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001746 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001747 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001749 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001750 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001751 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 }
1753 return err;
1754}
1755
1756/*
1757 * Send a datagram to a given address. We move the address into kernel
1758 * space and check the user space data area is readable before invoking
1759 * the protocol.
1760 */
1761
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001762SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001763 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001764 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765{
1766 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001767 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 int err;
1769 struct msghdr msg;
1770 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001771 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001772
Linus Torvalds253eacc2010-10-30 16:43:10 -07001773 if (len > INT_MAX)
1774 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001775 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1776 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001777 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001778
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001779 iov.iov_base = buff;
1780 iov.iov_len = len;
1781 msg.msg_name = NULL;
1782 msg.msg_iov = &iov;
1783 msg.msg_iovlen = 1;
1784 msg.msg_control = NULL;
1785 msg.msg_controllen = 0;
1786 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001787 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001788 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789 if (err < 0)
1790 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001791 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001792 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 }
1794 if (sock->file->f_flags & O_NONBLOCK)
1795 flags |= MSG_DONTWAIT;
1796 msg.msg_flags = flags;
1797 err = sock_sendmsg(sock, &msg, len);
1798
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001799out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001800 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001801out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 return err;
1803}
1804
1805/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001806 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 */
1808
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001809SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001810 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811{
1812 return sys_sendto(fd, buff, len, flags, NULL, 0);
1813}
1814
1815/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001816 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 * sender. We verify the buffers are writable and if needed move the
1818 * sender address from kernel to user space.
1819 */
1820
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001821SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001822 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001823 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824{
1825 struct socket *sock;
1826 struct iovec iov;
1827 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001828 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001829 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001830 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831
Linus Torvalds253eacc2010-10-30 16:43:10 -07001832 if (size > INT_MAX)
1833 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001834 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001836 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001838 msg.msg_control = NULL;
1839 msg.msg_controllen = 0;
1840 msg.msg_iovlen = 1;
1841 msg.msg_iov = &iov;
1842 iov.iov_len = size;
1843 iov.iov_base = ubuf;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001844 /* Save some cycles and don't copy the address if not needed */
1845 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1846 /* We assume all kernel code knows the size of sockaddr_storage */
1847 msg.msg_namelen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 if (sock->file->f_flags & O_NONBLOCK)
1849 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001850 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001852 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001853 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001854 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001855 if (err2 < 0)
1856 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001858
1859 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001860out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 return err;
1862}
1863
1864/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001865 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 */
1867
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001868asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001869 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870{
1871 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1872}
1873
1874/*
1875 * Set a socket option. Because we don't know the option lengths we have
1876 * to pass the user mode parameter for the protocols to sort out.
1877 */
1878
Heiko Carstens20f37032009-01-14 14:14:23 +01001879SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1880 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001882 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 struct socket *sock;
1884
1885 if (optlen < 0)
1886 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001887
1888 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1889 if (sock != NULL) {
1890 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001891 if (err)
1892 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
1894 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001895 err =
1896 sock_setsockopt(sock, level, optname, optval,
1897 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001899 err =
1900 sock->ops->setsockopt(sock, level, optname, optval,
1901 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001902out_put:
1903 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 }
1905 return err;
1906}
1907
1908/*
1909 * Get a socket option. Because we don't know the option lengths we have
1910 * to pass a user mode parameter for the protocols to sort out.
1911 */
1912
Heiko Carstens20f37032009-01-14 14:14:23 +01001913SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1914 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001916 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 struct socket *sock;
1918
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001919 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1920 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001921 err = security_socket_getsockopt(sock, level, optname);
1922 if (err)
1923 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924
1925 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001926 err =
1927 sock_getsockopt(sock, level, optname, optval,
1928 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001930 err =
1931 sock->ops->getsockopt(sock, level, optname, optval,
1932 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001933out_put:
1934 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 }
1936 return err;
1937}
1938
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939/*
1940 * Shutdown a socket.
1941 */
1942
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001943SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001945 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 struct socket *sock;
1947
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001948 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1949 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001951 if (!err)
1952 err = sock->ops->shutdown(sock, how);
1953 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 }
1955 return err;
1956}
1957
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001958/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 * fields which are the same type (int / unsigned) on our platforms.
1960 */
1961#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1962#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1963#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1964
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001965struct used_address {
1966 struct sockaddr_storage name;
1967 unsigned int name_len;
1968};
1969
Dan Carpenter1661bf32013-10-03 00:27:20 +03001970static int copy_msghdr_from_user(struct msghdr *kmsg,
1971 struct msghdr __user *umsg)
1972{
1973 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1974 return -EFAULT;
1975 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001976 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Dan Carpenter1661bf32013-10-03 00:27:20 +03001977 return 0;
1978}
1979
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001980static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001981 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001982 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001984 struct compat_msghdr __user *msg_compat =
1985 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001986 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001988 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001989 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1990 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001992 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001993
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 err = -EFAULT;
1995 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001996 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 return -EFAULT;
Dan Carpenter1661bf32013-10-03 00:27:20 +03001998 } else {
1999 err = copy_msghdr_from_user(msg_sys, msg);
2000 if (err)
2001 return err;
2002 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003
Anton Blanchard228e5482011-05-02 20:21:35 +00002004 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002005 err = -EMSGSIZE;
2006 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2007 goto out;
2008 err = -ENOMEM;
2009 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2010 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00002012 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 }
2014
2015 /* This will also move the address data into kernel space */
2016 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002017 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002019 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 goto out_freeiov;
2022 total_len = err;
2023
2024 err = -ENOBUFS;
2025
Anton Blanchard228e5482011-05-02 20:21:35 +00002026 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002028 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002030 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002031 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002032 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 if (err)
2034 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002035 ctl_buf = msg_sys->msg_control;
2036 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002038 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002040 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 goto out_freeiov;
2042 }
2043 err = -EFAULT;
2044 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002045 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2047 * checking falls down on this.
2048 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002049 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002050 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002051 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002053 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002055 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056
2057 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002058 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002059 /*
2060 * If this is sendmmsg() and current destination address is same as
2061 * previously succeeded address, omit asking LSM's decision.
2062 * used_address->name_len is initialized to UINT_MAX so that the first
2063 * destination address never matches.
2064 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002065 if (used_address && msg_sys->msg_name &&
2066 used_address->name_len == msg_sys->msg_namelen &&
2067 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002068 used_address->name_len)) {
2069 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2070 goto out_freectl;
2071 }
2072 err = sock_sendmsg(sock, msg_sys, total_len);
2073 /*
2074 * If this is sendmmsg() and sending to current destination address was
2075 * successful, remember it.
2076 */
2077 if (used_address && err >= 0) {
2078 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002079 if (msg_sys->msg_name)
2080 memcpy(&used_address->name, msg_sys->msg_name,
2081 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002082 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083
2084out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002085 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2087out_freeiov:
2088 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002089 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002090out:
2091 return err;
2092}
2093
2094/*
2095 * BSD sendmsg interface
2096 */
2097
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002098long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002099{
2100 int fput_needed, err;
2101 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002102 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002103
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002104 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002105 if (!sock)
2106 goto out;
2107
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002108 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002109
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002110 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002111out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 return err;
2113}
2114
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002115SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2116{
2117 if (flags & MSG_CMSG_COMPAT)
2118 return -EINVAL;
2119 return __sys_sendmsg(fd, msg, flags);
2120}
2121
Anton Blanchard228e5482011-05-02 20:21:35 +00002122/*
2123 * Linux sendmmsg interface
2124 */
2125
2126int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2127 unsigned int flags)
2128{
2129 int fput_needed, err, datagrams;
2130 struct socket *sock;
2131 struct mmsghdr __user *entry;
2132 struct compat_mmsghdr __user *compat_entry;
2133 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002134 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002135
Anton Blanchard98382f42011-08-04 14:07:39 +00002136 if (vlen > UIO_MAXIOV)
2137 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002138
2139 datagrams = 0;
2140
2141 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2142 if (!sock)
2143 return err;
2144
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002145 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002146 entry = mmsg;
2147 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002148 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002149
2150 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002151 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002152 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2153 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002154 if (err < 0)
2155 break;
2156 err = __put_user(err, &compat_entry->msg_len);
2157 ++compat_entry;
2158 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002159 err = ___sys_sendmsg(sock,
2160 (struct msghdr __user *)entry,
2161 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002162 if (err < 0)
2163 break;
2164 err = put_user(err, &entry->msg_len);
2165 ++entry;
2166 }
2167
2168 if (err)
2169 break;
2170 ++datagrams;
2171 }
2172
Anton Blanchard228e5482011-05-02 20:21:35 +00002173 fput_light(sock->file, fput_needed);
2174
Anton Blanchard728ffb82011-08-04 14:07:38 +00002175 /* We only return an error if no datagrams were able to be sent */
2176 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002177 return datagrams;
2178
Anton Blanchard228e5482011-05-02 20:21:35 +00002179 return err;
2180}
2181
2182SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2183 unsigned int, vlen, unsigned int, flags)
2184{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002185 if (flags & MSG_CMSG_COMPAT)
2186 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002187 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2188}
2189
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002190static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002191 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002193 struct compat_msghdr __user *msg_compat =
2194 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002196 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002198 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199
2200 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002201 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202
2203 /* user mode address pointers */
2204 struct sockaddr __user *uaddr;
2205 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002206
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002208 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 return -EFAULT;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002210 } else {
2211 err = copy_msghdr_from_user(msg_sys, msg);
2212 if (err)
2213 return err;
2214 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002216 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002217 err = -EMSGSIZE;
2218 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2219 goto out;
2220 err = -ENOMEM;
2221 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2222 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002224 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225 }
2226
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002227 /* Save the user-mode address (verify_iovec will change the
2228 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 */
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002230 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 uaddr_len = COMPAT_NAMELEN(msg);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002232 if (MSG_CMSG_COMPAT & flags)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002233 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002234 else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002235 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236 if (err < 0)
2237 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002238 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002239
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002240 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2241 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002242
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002243 /* We assume all kernel code knows the size of sockaddr_storage */
2244 msg_sys->msg_namelen = 0;
2245
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246 if (sock->file->f_flags & O_NONBLOCK)
2247 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002248 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2249 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 if (err < 0)
2251 goto out_freeiov;
2252 len = err;
2253
2254 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002255 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002256 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002257 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258 if (err < 0)
2259 goto out_freeiov;
2260 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002261 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002262 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263 if (err)
2264 goto out_freeiov;
2265 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002266 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267 &msg_compat->msg_controllen);
2268 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002269 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270 &msg->msg_controllen);
2271 if (err)
2272 goto out_freeiov;
2273 err = len;
2274
2275out_freeiov:
2276 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002277 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002278out:
2279 return err;
2280}
2281
2282/*
2283 * BSD recvmsg interface
2284 */
2285
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002286long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002287{
2288 int fput_needed, err;
2289 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002290 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002291
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002292 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002293 if (!sock)
2294 goto out;
2295
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002296 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002297
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002298 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299out:
2300 return err;
2301}
2302
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002303SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2304 unsigned int, flags)
2305{
2306 if (flags & MSG_CMSG_COMPAT)
2307 return -EINVAL;
2308 return __sys_recvmsg(fd, msg, flags);
2309}
2310
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002311/*
2312 * Linux recvmmsg interface
2313 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002315int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2316 unsigned int flags, struct timespec *timeout)
2317{
2318 int fput_needed, err, datagrams;
2319 struct socket *sock;
2320 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002321 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002322 struct msghdr msg_sys;
2323 struct timespec end_time;
2324
2325 if (timeout &&
2326 poll_select_set_timeout(&end_time, timeout->tv_sec,
2327 timeout->tv_nsec))
2328 return -EINVAL;
2329
2330 datagrams = 0;
2331
2332 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2333 if (!sock)
2334 return err;
2335
2336 err = sock_error(sock->sk);
2337 if (err)
2338 goto out_put;
2339
2340 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002341 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002342
2343 while (datagrams < vlen) {
2344 /*
2345 * No need to ask LSM for more than the first datagram.
2346 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002347 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002348 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2349 &msg_sys, flags & ~MSG_WAITFORONE,
2350 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002351 if (err < 0)
2352 break;
2353 err = __put_user(err, &compat_entry->msg_len);
2354 ++compat_entry;
2355 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002356 err = ___sys_recvmsg(sock,
2357 (struct msghdr __user *)entry,
2358 &msg_sys, flags & ~MSG_WAITFORONE,
2359 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002360 if (err < 0)
2361 break;
2362 err = put_user(err, &entry->msg_len);
2363 ++entry;
2364 }
2365
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002366 if (err)
2367 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002368 ++datagrams;
2369
Brandon L Black71c5c152010-03-26 16:18:03 +00002370 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2371 if (flags & MSG_WAITFORONE)
2372 flags |= MSG_DONTWAIT;
2373
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002374 if (timeout) {
2375 ktime_get_ts(timeout);
2376 *timeout = timespec_sub(end_time, *timeout);
2377 if (timeout->tv_sec < 0) {
2378 timeout->tv_sec = timeout->tv_nsec = 0;
2379 break;
2380 }
2381
2382 /* Timeout, return less than vlen datagrams */
2383 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2384 break;
2385 }
2386
2387 /* Out of band data, return right away */
2388 if (msg_sys.msg_flags & MSG_OOB)
2389 break;
2390 }
2391
2392out_put:
2393 fput_light(sock->file, fput_needed);
2394
2395 if (err == 0)
2396 return datagrams;
2397
2398 if (datagrams != 0) {
2399 /*
2400 * We may return less entries than requested (vlen) if the
2401 * sock is non block and there aren't enough datagrams...
2402 */
2403 if (err != -EAGAIN) {
2404 /*
2405 * ... or if recvmsg returns an error after we
2406 * received some datagrams, where we record the
2407 * error to return on the next call or if the
2408 * app asks about it using getsockopt(SO_ERROR).
2409 */
2410 sock->sk->sk_err = -err;
2411 }
2412
2413 return datagrams;
2414 }
2415
2416 return err;
2417}
2418
2419SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2420 unsigned int, vlen, unsigned int, flags,
2421 struct timespec __user *, timeout)
2422{
2423 int datagrams;
2424 struct timespec timeout_sys;
2425
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002426 if (flags & MSG_CMSG_COMPAT)
2427 return -EINVAL;
2428
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002429 if (!timeout)
2430 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2431
2432 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2433 return -EFAULT;
2434
2435 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2436
2437 if (datagrams > 0 &&
2438 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2439 datagrams = -EFAULT;
2440
2441 return datagrams;
2442}
2443
2444#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445/* Argument list sizes for sys_socketcall */
2446#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002447static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002448 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2449 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2450 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002451 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002452};
2453
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454#undef AL
2455
2456/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002457 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458 *
2459 * Argument checking cleaned up. Saved 20% in size.
2460 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002461 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 */
2463
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002464SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465{
Chen Gang2950fa92013-04-07 16:55:23 +08002466 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002467 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002469 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002470
Anton Blanchard228e5482011-05-02 20:21:35 +00002471 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472 return -EINVAL;
2473
Arjan van de Ven47379052009-09-28 12:57:44 -07002474 len = nargs[call];
2475 if (len > sizeof(a))
2476 return -EINVAL;
2477
Linus Torvalds1da177e2005-04-16 15:20:36 -07002478 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002479 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002481
Chen Gang2950fa92013-04-07 16:55:23 +08002482 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2483 if (err)
2484 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002485
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002486 a0 = a[0];
2487 a1 = a[1];
2488
2489 switch (call) {
2490 case SYS_SOCKET:
2491 err = sys_socket(a0, a1, a[2]);
2492 break;
2493 case SYS_BIND:
2494 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2495 break;
2496 case SYS_CONNECT:
2497 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2498 break;
2499 case SYS_LISTEN:
2500 err = sys_listen(a0, a1);
2501 break;
2502 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002503 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2504 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002505 break;
2506 case SYS_GETSOCKNAME:
2507 err =
2508 sys_getsockname(a0, (struct sockaddr __user *)a1,
2509 (int __user *)a[2]);
2510 break;
2511 case SYS_GETPEERNAME:
2512 err =
2513 sys_getpeername(a0, (struct sockaddr __user *)a1,
2514 (int __user *)a[2]);
2515 break;
2516 case SYS_SOCKETPAIR:
2517 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2518 break;
2519 case SYS_SEND:
2520 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2521 break;
2522 case SYS_SENDTO:
2523 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2524 (struct sockaddr __user *)a[4], a[5]);
2525 break;
2526 case SYS_RECV:
2527 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2528 break;
2529 case SYS_RECVFROM:
2530 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2531 (struct sockaddr __user *)a[4],
2532 (int __user *)a[5]);
2533 break;
2534 case SYS_SHUTDOWN:
2535 err = sys_shutdown(a0, a1);
2536 break;
2537 case SYS_SETSOCKOPT:
2538 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2539 break;
2540 case SYS_GETSOCKOPT:
2541 err =
2542 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2543 (int __user *)a[4]);
2544 break;
2545 case SYS_SENDMSG:
2546 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2547 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002548 case SYS_SENDMMSG:
2549 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2550 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002551 case SYS_RECVMSG:
2552 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2553 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002554 case SYS_RECVMMSG:
2555 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2556 (struct timespec __user *)a[4]);
2557 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002558 case SYS_ACCEPT4:
2559 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2560 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002561 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002562 default:
2563 err = -EINVAL;
2564 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565 }
2566 return err;
2567}
2568
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002569#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002571/**
2572 * sock_register - add a socket protocol handler
2573 * @ops: description of protocol
2574 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575 * This function is called by a protocol handler that wants to
2576 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002577 * socket interface. The value ops->family coresponds to the
2578 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002579 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002580int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002581{
2582 int err;
2583
2584 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002585 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2586 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587 return -ENOBUFS;
2588 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002589
2590 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002591 if (rcu_dereference_protected(net_families[ops->family],
2592 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002593 err = -EEXIST;
2594 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002595 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002596 err = 0;
2597 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002598 spin_unlock(&net_family_lock);
2599
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002600 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601 return err;
2602}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002603EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002604
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002605/**
2606 * sock_unregister - remove a protocol handler
2607 * @family: protocol family to remove
2608 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609 * This function is called by a protocol handler that wants to
2610 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002611 * new socket creation.
2612 *
2613 * If protocol handler is a module, then it can use module reference
2614 * counts to protect against new references. If protocol handler is not
2615 * a module then it needs to provide its own protection in
2616 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002617 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002618void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002619{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002620 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002621
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002622 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002623 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002624 spin_unlock(&net_family_lock);
2625
2626 synchronize_rcu();
2627
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002628 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002630EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002631
Andi Kleen77d76ea2005-12-22 12:43:42 -08002632static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002633{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002634 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002635 /*
2636 * Initialize the network sysctl infrastructure.
2637 */
2638 err = net_sysctl_init();
2639 if (err)
2640 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002641
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002643 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002644 */
2645 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002646
2647 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002648 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002649 */
2650
2651 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002652
2653 err = register_filesystem(&sock_fs_type);
2654 if (err)
2655 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002656 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002657 if (IS_ERR(sock_mnt)) {
2658 err = PTR_ERR(sock_mnt);
2659 goto out_mount;
2660 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002661
2662 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663 */
2664
2665#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002666 err = netfilter_init();
2667 if (err)
2668 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002669#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002670
Richard Cochranc1f19b52010-07-17 08:49:36 +00002671#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2672 skb_timestamping_init();
2673#endif
2674
Nick Pigginb3e19d92011-01-07 17:50:11 +11002675out:
2676 return err;
2677
2678out_mount:
2679 unregister_filesystem(&sock_fs_type);
2680out_fs:
2681 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002682}
2683
Andi Kleen77d76ea2005-12-22 12:43:42 -08002684core_initcall(sock_init); /* early initcall */
2685
Linus Torvalds1da177e2005-04-16 15:20:36 -07002686#ifdef CONFIG_PROC_FS
2687void socket_seq_show(struct seq_file *seq)
2688{
2689 int cpu;
2690 int counter = 0;
2691
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002692 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002693 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002694
2695 /* It can be negative, by the way. 8) */
2696 if (counter < 0)
2697 counter = 0;
2698
2699 seq_printf(seq, "sockets: used %d\n", counter);
2700}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002701#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002702
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002703#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002704static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002705 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002706{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002707 mm_segment_t old_fs = get_fs();
2708 struct timeval ktv;
2709 int err;
2710
2711 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002712 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002713 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002714 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002715 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002716
Arnd Bergmann7a229382009-11-06 23:00:29 -08002717 return err;
2718}
2719
Arnd Bergmann6b960182009-11-06 23:10:54 -08002720static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002721 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002722{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002723 mm_segment_t old_fs = get_fs();
2724 struct timespec kts;
2725 int err;
2726
2727 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002728 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002729 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002730 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002731 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002732
Arnd Bergmann7a229382009-11-06 23:00:29 -08002733 return err;
2734}
2735
Arnd Bergmann6b960182009-11-06 23:10:54 -08002736static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002737{
2738 struct ifreq __user *uifr;
2739 int err;
2740
2741 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002742 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002743 return -EFAULT;
2744
Arnd Bergmann6b960182009-11-06 23:10:54 -08002745 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002746 if (err)
2747 return err;
2748
Arnd Bergmann6b960182009-11-06 23:10:54 -08002749 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002750 return -EFAULT;
2751
2752 return 0;
2753}
2754
Arnd Bergmann6b960182009-11-06 23:10:54 -08002755static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002756{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002757 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002758 struct ifconf ifc;
2759 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002760 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002761 struct ifreq __user *ifr;
2762 unsigned int i, j;
2763 int err;
2764
Arnd Bergmann6b960182009-11-06 23:10:54 -08002765 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002766 return -EFAULT;
2767
Mathias Krause43da5f22012-08-15 11:31:57 +00002768 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002769 if (ifc32.ifcbuf == 0) {
2770 ifc32.ifc_len = 0;
2771 ifc.ifc_len = 0;
2772 ifc.ifc_req = NULL;
2773 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2774 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002775 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2776 sizeof(struct ifreq);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002777 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2778 ifc.ifc_len = len;
2779 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2780 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002781 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002782 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002783 return -EFAULT;
2784 ifr++;
2785 ifr32++;
2786 }
2787 }
2788 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2789 return -EFAULT;
2790
Arnd Bergmann6b960182009-11-06 23:10:54 -08002791 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002792 if (err)
2793 return err;
2794
2795 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2796 return -EFAULT;
2797
2798 ifr = ifc.ifc_req;
2799 ifr32 = compat_ptr(ifc32.ifcbuf);
2800 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002801 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2802 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2803 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002804 return -EFAULT;
2805 ifr32++;
2806 ifr++;
2807 }
2808
2809 if (ifc32.ifcbuf == 0) {
2810 /* Translate from 64-bit structure multiple to
2811 * a 32-bit one.
2812 */
2813 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002814 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002815 ifc32.ifc_len = i;
2816 } else {
2817 ifc32.ifc_len = i;
2818 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002819 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002820 return -EFAULT;
2821
2822 return 0;
2823}
2824
Arnd Bergmann6b960182009-11-06 23:10:54 -08002825static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002826{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002827 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2828 bool convert_in = false, convert_out = false;
2829 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2830 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002831 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002832 u32 rule_cnt = 0, actual_rule_cnt;
2833 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002834 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002835 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002836
2837 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2838 return -EFAULT;
2839
Ben Hutchings3a7da392011-03-17 07:34:32 +00002840 compat_rxnfc = compat_ptr(data);
2841
2842 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002843 return -EFAULT;
2844
Ben Hutchings3a7da392011-03-17 07:34:32 +00002845 /* Most ethtool structures are defined without padding.
2846 * Unfortunately struct ethtool_rxnfc is an exception.
2847 */
2848 switch (ethcmd) {
2849 default:
2850 break;
2851 case ETHTOOL_GRXCLSRLALL:
2852 /* Buffer size is variable */
2853 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2854 return -EFAULT;
2855 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2856 return -ENOMEM;
2857 buf_size += rule_cnt * sizeof(u32);
2858 /* fall through */
2859 case ETHTOOL_GRXRINGS:
2860 case ETHTOOL_GRXCLSRLCNT:
2861 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002862 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002863 convert_out = true;
2864 /* fall through */
2865 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002866 buf_size += sizeof(struct ethtool_rxnfc);
2867 convert_in = true;
2868 break;
2869 }
2870
2871 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002872 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002873
2874 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2875 return -EFAULT;
2876
2877 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2878 &ifr->ifr_ifru.ifru_data))
2879 return -EFAULT;
2880
2881 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002882 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002883 * fs.ring_cookie and at the end of fs, but nowhere else.
2884 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002885 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2886 sizeof(compat_rxnfc->fs.m_ext) !=
2887 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2888 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002889 BUILD_BUG_ON(
2890 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2891 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2892 offsetof(struct ethtool_rxnfc, fs.location) -
2893 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2894
2895 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002896 (void __user *)(&rxnfc->fs.m_ext + 1) -
2897 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002898 copy_in_user(&rxnfc->fs.ring_cookie,
2899 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002900 (void __user *)(&rxnfc->fs.location + 1) -
2901 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002902 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2903 sizeof(rxnfc->rule_cnt)))
2904 return -EFAULT;
2905 }
2906
2907 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2908 if (ret)
2909 return ret;
2910
2911 if (convert_out) {
2912 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002913 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2914 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002915 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2916 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002917 (const void __user *)(&rxnfc->fs.location + 1) -
2918 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002919 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2920 sizeof(rxnfc->rule_cnt)))
2921 return -EFAULT;
2922
2923 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2924 /* As an optimisation, we only copy the actual
2925 * number of rules that the underlying
2926 * function returned. Since Mallory might
2927 * change the rule count in user memory, we
2928 * check that it is less than the rule count
2929 * originally given (as the user buffer size),
2930 * which has been range-checked.
2931 */
2932 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2933 return -EFAULT;
2934 if (actual_rule_cnt < rule_cnt)
2935 rule_cnt = actual_rule_cnt;
2936 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2937 &rxnfc->rule_locs[0],
2938 rule_cnt * sizeof(u32)))
2939 return -EFAULT;
2940 }
2941 }
2942
2943 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002944}
2945
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002946static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2947{
2948 void __user *uptr;
2949 compat_uptr_t uptr32;
2950 struct ifreq __user *uifr;
2951
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002952 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002953 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2954 return -EFAULT;
2955
2956 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2957 return -EFAULT;
2958
2959 uptr = compat_ptr(uptr32);
2960
2961 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2962 return -EFAULT;
2963
2964 return dev_ioctl(net, SIOCWANDEV, uifr);
2965}
2966
Arnd Bergmann6b960182009-11-06 23:10:54 -08002967static int bond_ioctl(struct net *net, unsigned int cmd,
2968 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002969{
2970 struct ifreq kifr;
2971 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002972 mm_segment_t old_fs;
2973 int err;
2974 u32 data;
2975 void __user *datap;
2976
2977 switch (cmd) {
2978 case SIOCBONDENSLAVE:
2979 case SIOCBONDRELEASE:
2980 case SIOCBONDSETHWADDR:
2981 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002982 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002983 return -EFAULT;
2984
2985 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002986 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002987 err = dev_ioctl(net, cmd,
2988 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002989 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002990
2991 return err;
2992 case SIOCBONDSLAVEINFOQUERY:
2993 case SIOCBONDINFOQUERY:
2994 uifr = compat_alloc_user_space(sizeof(*uifr));
2995 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2996 return -EFAULT;
2997
2998 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2999 return -EFAULT;
3000
3001 datap = compat_ptr(data);
3002 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
3003 return -EFAULT;
3004
Arnd Bergmann6b960182009-11-06 23:10:54 -08003005 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003006 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08003007 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003008 }
Arnd Bergmann7a229382009-11-06 23:00:29 -08003009}
3010
Arnd Bergmann6b960182009-11-06 23:10:54 -08003011static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
3012 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003013{
3014 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003015 char tmp_buf[IFNAMSIZ];
3016 void __user *data64;
3017 u32 data32;
3018
3019 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3020 IFNAMSIZ))
3021 return -EFAULT;
3022 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3023 return -EFAULT;
3024 data64 = compat_ptr(data32);
3025
3026 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3027
3028 /* Don't check these user accesses, just let that get trapped
3029 * in the ioctl handler instead.
3030 */
3031 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3032 IFNAMSIZ))
3033 return -EFAULT;
3034 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3035 return -EFAULT;
3036
Arnd Bergmann6b960182009-11-06 23:10:54 -08003037 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003038}
3039
Arnd Bergmann6b960182009-11-06 23:10:54 -08003040static int dev_ifsioc(struct net *net, struct socket *sock,
3041 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003042{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003043 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003044 int err;
3045
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003046 uifr = compat_alloc_user_space(sizeof(*uifr));
3047 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3048 return -EFAULT;
3049
3050 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3051
Arnd Bergmann7a229382009-11-06 23:00:29 -08003052 if (!err) {
3053 switch (cmd) {
3054 case SIOCGIFFLAGS:
3055 case SIOCGIFMETRIC:
3056 case SIOCGIFMTU:
3057 case SIOCGIFMEM:
3058 case SIOCGIFHWADDR:
3059 case SIOCGIFINDEX:
3060 case SIOCGIFADDR:
3061 case SIOCGIFBRDADDR:
3062 case SIOCGIFDSTADDR:
3063 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003064 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003065 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003066 case SIOCGMIIPHY:
3067 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003068 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003069 err = -EFAULT;
3070 break;
3071 }
3072 }
3073 return err;
3074}
3075
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003076static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3077 struct compat_ifreq __user *uifr32)
3078{
3079 struct ifreq ifr;
3080 struct compat_ifmap __user *uifmap32;
3081 mm_segment_t old_fs;
3082 int err;
3083
3084 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3085 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003086 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3087 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3088 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3089 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3090 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3091 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003092 if (err)
3093 return -EFAULT;
3094
3095 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003096 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003097 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003098 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003099
3100 if (cmd == SIOCGIFMAP && !err) {
3101 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003102 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3103 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3104 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3105 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3106 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3107 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003108 if (err)
3109 err = -EFAULT;
3110 }
3111 return err;
3112}
3113
3114static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3115{
3116 void __user *uptr;
3117 compat_uptr_t uptr32;
3118 struct ifreq __user *uifr;
3119
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003120 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003121 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3122 return -EFAULT;
3123
3124 if (get_user(uptr32, &uifr32->ifr_data))
3125 return -EFAULT;
3126
3127 uptr = compat_ptr(uptr32);
3128
3129 if (put_user(uptr, &uifr->ifr_data))
3130 return -EFAULT;
3131
3132 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3133}
3134
Arnd Bergmann7a229382009-11-06 23:00:29 -08003135struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003136 u32 rt_pad1;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003137 struct sockaddr rt_dst; /* target address */
3138 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3139 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003140 unsigned short rt_flags;
3141 short rt_pad2;
3142 u32 rt_pad3;
3143 unsigned char rt_tos;
3144 unsigned char rt_class;
3145 short rt_pad4;
3146 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003147 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003148 u32 rt_mtu; /* per route MTU/Window */
3149 u32 rt_window; /* Window clamping */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003150 unsigned short rt_irtt; /* Initial RTT */
3151};
3152
3153struct in6_rtmsg32 {
3154 struct in6_addr rtmsg_dst;
3155 struct in6_addr rtmsg_src;
3156 struct in6_addr rtmsg_gateway;
3157 u32 rtmsg_type;
3158 u16 rtmsg_dst_len;
3159 u16 rtmsg_src_len;
3160 u32 rtmsg_metric;
3161 u32 rtmsg_info;
3162 u32 rtmsg_flags;
3163 s32 rtmsg_ifindex;
3164};
3165
Arnd Bergmann6b960182009-11-06 23:10:54 -08003166static int routing_ioctl(struct net *net, struct socket *sock,
3167 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003168{
3169 int ret;
3170 void *r = NULL;
3171 struct in6_rtmsg r6;
3172 struct rtentry r4;
3173 char devname[16];
3174 u32 rtdev;
3175 mm_segment_t old_fs = get_fs();
3176
Arnd Bergmann6b960182009-11-06 23:10:54 -08003177 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3178 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003179 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003180 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003181 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3182 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3183 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3184 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3185 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3186 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3187 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003188
3189 r = (void *) &r6;
3190 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003191 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003192 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003193 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003194 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3195 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3196 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3197 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3198 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3199 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003200 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003201 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003202 r4.rt_dev = (char __user __force *)devname;
3203 devname[15] = 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003204 } else
3205 r4.rt_dev = NULL;
3206
3207 r = (void *) &r4;
3208 }
3209
3210 if (ret) {
3211 ret = -EFAULT;
3212 goto out;
3213 }
3214
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003215 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003216 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003217 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003218
3219out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003220 return ret;
3221}
3222
3223/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3224 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003225 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003226 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003227static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003228{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003229 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003230
Arnd Bergmann6b960182009-11-06 23:10:54 -08003231 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003232 return -EFAULT;
3233 if (tmp == BRCTL_GET_VERSION)
3234 return BRCTL_VERSION + 1;
3235 return -EINVAL;
3236}
3237
Arnd Bergmann6b960182009-11-06 23:10:54 -08003238static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3239 unsigned int cmd, unsigned long arg)
3240{
3241 void __user *argp = compat_ptr(arg);
3242 struct sock *sk = sock->sk;
3243 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003244
Arnd Bergmann6b960182009-11-06 23:10:54 -08003245 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3246 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003247
Arnd Bergmann6b960182009-11-06 23:10:54 -08003248 switch (cmd) {
3249 case SIOCSIFBR:
3250 case SIOCGIFBR:
3251 return old_bridge_ioctl(argp);
3252 case SIOCGIFNAME:
3253 return dev_ifname32(net, argp);
3254 case SIOCGIFCONF:
3255 return dev_ifconf(net, argp);
3256 case SIOCETHTOOL:
3257 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003258 case SIOCWANDEV:
3259 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003260 case SIOCGIFMAP:
3261 case SIOCSIFMAP:
3262 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003263 case SIOCBONDENSLAVE:
3264 case SIOCBONDRELEASE:
3265 case SIOCBONDSETHWADDR:
3266 case SIOCBONDSLAVEINFOQUERY:
3267 case SIOCBONDINFOQUERY:
3268 case SIOCBONDCHANGEACTIVE:
3269 return bond_ioctl(net, cmd, argp);
3270 case SIOCADDRT:
3271 case SIOCDELRT:
3272 return routing_ioctl(net, sock, cmd, argp);
3273 case SIOCGSTAMP:
3274 return do_siocgstamp(net, sock, cmd, argp);
3275 case SIOCGSTAMPNS:
3276 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003277 case SIOCSHWTSTAMP:
3278 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003279
Arnd Bergmann6b960182009-11-06 23:10:54 -08003280 case FIOSETOWN:
3281 case SIOCSPGRP:
3282 case FIOGETOWN:
3283 case SIOCGPGRP:
3284 case SIOCBRADDBR:
3285 case SIOCBRDELBR:
3286 case SIOCGIFVLAN:
3287 case SIOCSIFVLAN:
3288 case SIOCADDDLCI:
3289 case SIOCDELDLCI:
3290 return sock_ioctl(file, cmd, arg);
3291
3292 case SIOCGIFFLAGS:
3293 case SIOCSIFFLAGS:
3294 case SIOCGIFMETRIC:
3295 case SIOCSIFMETRIC:
3296 case SIOCGIFMTU:
3297 case SIOCSIFMTU:
3298 case SIOCGIFMEM:
3299 case SIOCSIFMEM:
3300 case SIOCGIFHWADDR:
3301 case SIOCSIFHWADDR:
3302 case SIOCADDMULTI:
3303 case SIOCDELMULTI:
3304 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003305 case SIOCGIFADDR:
3306 case SIOCSIFADDR:
3307 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003308 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003309 case SIOCGIFBRDADDR:
3310 case SIOCSIFBRDADDR:
3311 case SIOCGIFDSTADDR:
3312 case SIOCSIFDSTADDR:
3313 case SIOCGIFNETMASK:
3314 case SIOCSIFNETMASK:
3315 case SIOCSIFPFLAGS:
3316 case SIOCGIFPFLAGS:
3317 case SIOCGIFTXQLEN:
3318 case SIOCSIFTXQLEN:
3319 case SIOCBRADDIF:
3320 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003321 case SIOCSIFNAME:
3322 case SIOCGMIIPHY:
3323 case SIOCGMIIREG:
3324 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003325 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003326
Arnd Bergmann6b960182009-11-06 23:10:54 -08003327 case SIOCSARP:
3328 case SIOCGARP:
3329 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003330 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003331 return sock_do_ioctl(net, sock, cmd, arg);
3332 }
3333
Arnd Bergmann6b960182009-11-06 23:10:54 -08003334 return -ENOIOCTLCMD;
3335}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003336
Eric Dumazet95c96172012-04-15 05:58:06 +00003337static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003338 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003339{
3340 struct socket *sock = file->private_data;
3341 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003342 struct sock *sk;
3343 struct net *net;
3344
3345 sk = sock->sk;
3346 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003347
3348 if (sock->ops->compat_ioctl)
3349 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3350
David S. Miller87de87d2008-06-03 09:14:03 -07003351 if (ret == -ENOIOCTLCMD &&
3352 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3353 ret = compat_wext_handle_ioctl(net, cmd, arg);
3354
Arnd Bergmann6b960182009-11-06 23:10:54 -08003355 if (ret == -ENOIOCTLCMD)
3356 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3357
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003358 return ret;
3359}
3360#endif
3361
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003362int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3363{
3364 return sock->ops->bind(sock, addr, addrlen);
3365}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003366EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003367
3368int kernel_listen(struct socket *sock, int backlog)
3369{
3370 return sock->ops->listen(sock, backlog);
3371}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003372EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003373
3374int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3375{
3376 struct sock *sk = sock->sk;
3377 int err;
3378
3379 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3380 newsock);
3381 if (err < 0)
3382 goto done;
3383
3384 err = sock->ops->accept(sock, *newsock, flags);
3385 if (err < 0) {
3386 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003387 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003388 goto done;
3389 }
3390
3391 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003392 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003393
3394done:
3395 return err;
3396}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003397EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003398
3399int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003400 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003401{
3402 return sock->ops->connect(sock, addr, addrlen, flags);
3403}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003404EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003405
3406int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3407 int *addrlen)
3408{
3409 return sock->ops->getname(sock, addr, addrlen, 0);
3410}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003411EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003412
3413int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3414 int *addrlen)
3415{
3416 return sock->ops->getname(sock, addr, addrlen, 1);
3417}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003418EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003419
3420int kernel_getsockopt(struct socket *sock, int level, int optname,
3421 char *optval, int *optlen)
3422{
3423 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003424 char __user *uoptval;
3425 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003426 int err;
3427
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003428 uoptval = (char __user __force *) optval;
3429 uoptlen = (int __user __force *) optlen;
3430
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003431 set_fs(KERNEL_DS);
3432 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003433 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003434 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003435 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3436 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003437 set_fs(oldfs);
3438 return err;
3439}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003440EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003441
3442int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003443 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003444{
3445 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003446 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003447 int err;
3448
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003449 uoptval = (char __user __force *) optval;
3450
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003451 set_fs(KERNEL_DS);
3452 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003453 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003454 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003455 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003456 optlen);
3457 set_fs(oldfs);
3458 return err;
3459}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003460EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003461
3462int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3463 size_t size, int flags)
3464{
3465 if (sock->ops->sendpage)
3466 return sock->ops->sendpage(sock, page, offset, size, flags);
3467
3468 return sock_no_sendpage(sock, page, offset, size, flags);
3469}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003470EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003471
3472int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3473{
3474 mm_segment_t oldfs = get_fs();
3475 int err;
3476
3477 set_fs(KERNEL_DS);
3478 err = sock->ops->ioctl(sock, cmd, arg);
3479 set_fs(oldfs);
3480
3481 return err;
3482}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003483EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003484
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003485int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3486{
3487 return sock->ops->shutdown(sock, how);
3488}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003489EXPORT_SYMBOL(kernel_sock_shutdown);