blob: ebed4b68f768a1afccea45ea7dc22ef555f3f679 [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
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700224 err = get_user(len, ulen);
225 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700227 if (len > klen)
228 len = klen;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700229 if (len < 0 || len > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700231 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500232 if (audit_sockaddr(klen, kaddr))
233 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700234 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 return -EFAULT;
236 }
237 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700238 * "fromlen shall refer to the value before truncation.."
239 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 */
241 return __put_user(klen, ulen);
242}
243
Christoph Lametere18b8902006-12-06 20:33:20 -0800244static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245
246static struct inode *sock_alloc_inode(struct super_block *sb)
247{
248 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000249 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700250
Christoph Lametere94b1762006-12-06 20:33:17 -0800251 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 if (!ei)
253 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000254 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
255 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000256 kmem_cache_free(sock_inode_cachep, ei);
257 return NULL;
258 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000259 init_waitqueue_head(&wq->wait);
260 wq->fasync_list = NULL;
261 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700262
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
269 return &ei->vfs_inode;
270}
271
272static void sock_destroy_inode(struct inode *inode)
273{
Eric Dumazet43815482010-04-29 11:01:49 +0000274 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000275 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000276
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000278 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800279 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000280 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281}
282
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700283static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700285 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Christoph Lametera35afb82007-05-16 22:10:57 -0700287 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290static int init_inodecache(void)
291{
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700293 sizeof(struct socket_alloc),
294 0,
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
297 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900298 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 if (sock_inode_cachep == NULL)
300 return -ENOMEM;
301 return 0;
302}
303
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700304static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700305 .alloc_inode = sock_alloc_inode,
306 .destroy_inode = sock_destroy_inode,
307 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308};
309
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700310/*
311 * sockfs_dname() is called from d_path().
312 */
313static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314{
315 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316 dentry->d_inode->i_ino);
317}
318
Al Viro3ba13d12009-02-20 06:02:22 +0000319static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700320 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321};
322
Al Viroc74a1cb2011-01-12 16:59:34 -0500323static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324 int flags, const char *dev_name, void *data)
325{
326 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327 &sockfs_dentry_operations, SOCKFS_MAGIC);
328}
329
330static struct vfsmount *sock_mnt __read_mostly;
331
332static struct file_system_type sock_fs_type = {
333 .name = "sockfs",
334 .mount = sockfs_mount,
335 .kill_sb = kill_anon_super,
336};
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338/*
339 * Obtains the first available file descriptor and sets it up for use.
340 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
348 *
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
353 */
354
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700355struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356{
Al Viro7cbe66b2009-08-05 19:59:08 +0400357 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400358 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400359 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800360
Masatake YAMATO600e1772012-08-29 10:44:29 +0000361 if (dname) {
362 name.name = dname;
363 name.len = strlen(name.name);
364 } else if (sock->sk) {
365 name.name = sock->sk->sk_prot_creator->name;
366 name.len = strlen(name.name);
367 }
Nick Piggin4b936882011-01-07 17:50:07 +1100368 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400369 if (unlikely(!path.dentry))
370 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400371 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
Al Viro2c48b9c2009-08-09 00:52:35 +0400373 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400374 SOCK_INODE(sock)->i_fop = &socket_file_ops;
375
Al Viro2c48b9c2009-08-09 00:52:35 +0400376 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400377 &socket_file_ops);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700378 if (unlikely(IS_ERR(file))) {
Al Virocc3808f2009-08-06 09:43:59 +0400379 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400380 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400381 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700382 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400383 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800384
385 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700386 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400388 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800389}
Al Viro56b31d12012-08-18 00:25:51 -0400390EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800391
Al Viro56b31d12012-08-18 00:25:51 -0400392static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800393{
394 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400395 int fd = get_unused_fd_flags(flags);
396 if (unlikely(fd < 0))
397 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800398
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700399 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400400 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400402 return fd;
403 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400404
Al Viro28407632012-08-17 23:54:15 -0400405 put_unused_fd(fd);
406 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407}
408
John Fastabend406a3c62012-07-20 10:39:25 +0000409struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800410{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800411 if (file->f_op == &socket_file_ops)
412 return file->private_data; /* set in sock_map_fd */
413
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800414 *err = -ENOTSOCK;
415 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800416}
John Fastabend406a3c62012-07-20 10:39:25 +0000417EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800418
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700420 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * @fd: file handle
422 * @err: pointer to an error code return
423 *
424 * The file handle passed in is locked and the socket it is bound
425 * too is returned. If an error occurs the err pointer is overwritten
426 * with a negative errno code and NULL is returned. The function checks
427 * for both invalid handles and passing a handle which is not a socket.
428 *
429 * On a success the socket object pointer is returned.
430 */
431
432struct socket *sockfd_lookup(int fd, int *err)
433{
434 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 struct socket *sock;
436
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700437 file = fget(fd);
438 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 *err = -EBADF;
440 return NULL;
441 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700442
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800443 sock = sock_from_file(file, err);
444 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 return sock;
447}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700448EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800450static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
451{
452 struct file *file;
453 struct socket *sock;
454
Hua Zhong36725582006-04-19 15:25:02 -0700455 *err = -EBADF;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800456 file = fget_light(fd, fput_needed);
457 if (file) {
458 sock = sock_from_file(file, err);
459 if (sock)
460 return sock;
461 fput_light(file, *fput_needed);
462 }
463 return NULL;
464}
465
Masatake YAMATO600e1772012-08-29 10:44:29 +0000466#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469static ssize_t sockfs_getxattr(struct dentry *dentry,
470 const char *name, void *value, size_t size)
471{
472 const char *proto_name;
473 size_t proto_size;
474 int error;
475
476 error = -ENODATA;
477 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478 proto_name = dentry->d_name.name;
479 proto_size = strlen(proto_name);
480
481 if (value) {
482 error = -ERANGE;
483 if (proto_size + 1 > size)
484 goto out;
485
486 strncpy(value, proto_name, proto_size + 1);
487 }
488 error = proto_size + 1;
489 }
490
491out:
492 return error;
493}
494
495static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496 size_t size)
497{
498 ssize_t len;
499 ssize_t used = 0;
500
501 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
502 if (len < 0)
503 return len;
504 used += len;
505 if (buffer) {
506 if (size < used)
507 return -ERANGE;
508 buffer += len;
509 }
510
511 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512 used += len;
513 if (buffer) {
514 if (size < used)
515 return -ERANGE;
516 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517 buffer += len;
518 }
519
520 return used;
521}
522
523static const struct inode_operations sockfs_inode_ops = {
524 .getxattr = sockfs_getxattr,
525 .listxattr = sockfs_listxattr,
526};
527
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528/**
529 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700530 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 * Allocate a new inode and socket object. The two are bound together
532 * and initialised. The socket is then returned. If we are out of inodes
533 * NULL is returned.
534 */
535
536static struct socket *sock_alloc(void)
537{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700538 struct inode *inode;
539 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200541 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 if (!inode)
543 return NULL;
544
545 sock = SOCKET_I(inode);
546
Eric Dumazet29a020d2009-09-15 02:39:20 -0700547 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400548 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700549 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100550 inode->i_uid = current_fsuid();
551 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000552 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
Alex Shi19e8d692012-05-14 14:15:31 -0700554 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return sock;
556}
557
558/*
559 * In theory you can't get an open on this inode, but /proc provides
560 * a back door. Remember to keep it shut otherwise you'll let the
561 * creepy crawlies in.
562 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
565{
566 return -ENXIO;
567}
568
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800569const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 .owner = THIS_MODULE,
571 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200572 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573};
574
575/**
576 * sock_release - close a socket
577 * @sock: socket to close
578 *
579 * The socket is released from the protocol stack if it has a release
580 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700581 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700583
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584void sock_release(struct socket *sock)
585{
586 if (sock->ops) {
587 struct module *owner = sock->ops->owner;
588
589 sock->ops->release(sock);
590 sock->ops = NULL;
591 module_put(owner);
592 }
593
Eric Dumazeteaefd112011-02-18 03:26:36 +0000594 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 printk(KERN_ERR "sock_release: fasync list not empty!\n");
596
Mikulas Patockab09e7862012-07-19 06:13:36 +0000597 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
598 return;
599
Alex Shi19e8d692012-05-14 14:15:31 -0700600 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 if (!sock->file) {
602 iput(SOCK_INODE(sock));
603 return;
604 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700605 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700607EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000609void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000610{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000611 *tx_flags = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000612 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000613 *tx_flags |= SKBTX_HW_TSTAMP;
Patrick Ohly20d49472009-02-12 05:03:38 +0000614 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000615 *tx_flags |= SKBTX_SW_TSTAMP;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100616 if (sock_flag(sk, SOCK_WIFI_STATUS))
617 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000618}
619EXPORT_SYMBOL(sock_tx_timestamp);
620
Anton Blanchard228e5482011-05-02 20:21:35 +0000621static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
622 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623{
624 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626 si->sock = sock;
627 si->scm = NULL;
628 si->msg = msg;
629 si->size = size;
630
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 return sock->ops->sendmsg(iocb, sock, msg, size);
632}
633
Anton Blanchard228e5482011-05-02 20:21:35 +0000634static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
635 struct msghdr *msg, size_t size)
636{
637 int err = security_socket_sendmsg(sock, msg, size);
638
639 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
640}
641
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
643{
644 struct kiocb iocb;
645 struct sock_iocb siocb;
646 int ret;
647
648 init_sync_kiocb(&iocb, NULL);
649 iocb.private = &siocb;
650 ret = __sock_sendmsg(&iocb, sock, msg, size);
651 if (-EIOCBQUEUED == ret)
652 ret = wait_on_sync_kiocb(&iocb);
653 return ret;
654}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700655EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Eric Dumazet894dc242011-07-26 02:39:41 +0000657static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000658{
659 struct kiocb iocb;
660 struct sock_iocb siocb;
661 int ret;
662
663 init_sync_kiocb(&iocb, NULL);
664 iocb.private = &siocb;
665 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
666 if (-EIOCBQUEUED == ret)
667 ret = wait_on_sync_kiocb(&iocb);
668 return ret;
669}
670
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
672 struct kvec *vec, size_t num, size_t size)
673{
674 mm_segment_t oldfs = get_fs();
675 int result;
676
677 set_fs(KERNEL_DS);
678 /*
679 * the following is safe, since for compiler definitions of kvec and
680 * iovec are identical, yielding the same in-core layout and alignment
681 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700682 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 msg->msg_iovlen = num;
684 result = sock_sendmsg(sock, msg, size);
685 set_fs(oldfs);
686 return result;
687}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700688EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700690/*
691 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
692 */
693void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
694 struct sk_buff *skb)
695{
Patrick Ohly20d49472009-02-12 05:03:38 +0000696 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
697 struct timespec ts[3];
698 int empty = 1;
699 struct skb_shared_hwtstamps *shhwtstamps =
700 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700701
Patrick Ohly20d49472009-02-12 05:03:38 +0000702 /* Race occurred between timestamp enabling and packet
703 receiving. Fill in the current time for now. */
704 if (need_software_tstamp && skb->tstamp.tv64 == 0)
705 __net_timestamp(skb);
706
707 if (need_software_tstamp) {
708 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
709 struct timeval tv;
710 skb_get_timestamp(skb, &tv);
711 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
712 sizeof(tv), &tv);
713 } else {
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000714 skb_get_timestampns(skb, &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000715 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Hagen Paul Pfeifer842509b2010-04-06 05:39:52 +0000716 sizeof(ts[0]), &ts[0]);
Patrick Ohly20d49472009-02-12 05:03:38 +0000717 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700718 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000719
720
721 memset(ts, 0, sizeof(ts));
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000722 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE) &&
723 ktime_to_timespec_cond(skb->tstamp, ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000724 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000725 if (shhwtstamps) {
726 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000727 ktime_to_timespec_cond(shhwtstamps->syststamp, ts + 1))
Patrick Ohly20d49472009-02-12 05:03:38 +0000728 empty = 0;
729 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
Daniel Borkmann6e94d1e2013-04-16 01:29:10 +0000730 ktime_to_timespec_cond(shhwtstamps->hwtstamp, ts + 2))
Patrick Ohly20d49472009-02-12 05:03:38 +0000731 empty = 0;
732 }
733 if (!empty)
734 put_cmsg(msg, SOL_SOCKET,
735 SCM_TIMESTAMPING, sizeof(ts), &ts);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700736}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300737EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
738
Johannes Berg6e3e9392011-11-09 10:15:42 +0100739void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
740 struct sk_buff *skb)
741{
742 int ack;
743
744 if (!sock_flag(sk, SOCK_WIFI_STATUS))
745 return;
746 if (!skb->wifi_acked_valid)
747 return;
748
749 ack = skb->wifi_acked;
750
751 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
752}
753EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
754
stephen hemminger11165f12010-10-18 14:27:29 +0000755static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
756 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700757{
758 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
759 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
760 sizeof(__u32), &skb->dropcount);
761}
762
Eric Dumazet767dd032010-04-28 19:14:43 +0000763void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700764 struct sk_buff *skb)
765{
766 sock_recv_timestamp(msg, sk, skb);
767 sock_recv_drops(msg, sk, skb);
768}
Eric Dumazet767dd032010-04-28 19:14:43 +0000769EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700770
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700771static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
772 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 struct sock_iocb *si = kiocb_to_siocb(iocb);
775
776 si->sock = sock;
777 si->scm = NULL;
778 si->msg = msg;
779 si->size = size;
780 si->flags = flags;
781
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
783}
784
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700785static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
786 struct msghdr *msg, size_t size, int flags)
787{
788 int err = security_socket_recvmsg(sock, msg, size, flags);
789
790 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
791}
792
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700793int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 size_t size, int flags)
795{
796 struct kiocb iocb;
797 struct sock_iocb siocb;
798 int ret;
799
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700800 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 iocb.private = &siocb;
802 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
803 if (-EIOCBQUEUED == ret)
804 ret = wait_on_sync_kiocb(&iocb);
805 return ret;
806}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700807EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700809static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
810 size_t size, int flags)
811{
812 struct kiocb iocb;
813 struct sock_iocb siocb;
814 int ret;
815
816 init_sync_kiocb(&iocb, NULL);
817 iocb.private = &siocb;
818 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
819 if (-EIOCBQUEUED == ret)
820 ret = wait_on_sync_kiocb(&iocb);
821 return ret;
822}
823
Martin Lucinac1249c02010-12-10 00:04:05 +0000824/**
825 * kernel_recvmsg - Receive a message from a socket (kernel space)
826 * @sock: The socket to receive the message from
827 * @msg: Received message
828 * @vec: Input s/g array for message data
829 * @num: Size of input s/g array
830 * @size: Number of bytes to read
831 * @flags: Message flags (MSG_DONTWAIT, etc...)
832 *
833 * On return the msg structure contains the scatter/gather array passed in the
834 * vec argument. The array is modified so that it consists of the unfilled
835 * portion of the original array.
836 *
837 * The returned value is the total number of bytes received, or an error.
838 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700839int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
840 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
842 mm_segment_t oldfs = get_fs();
843 int result;
844
845 set_fs(KERNEL_DS);
846 /*
847 * the following is safe, since for compiler definitions of kvec and
848 * iovec are identical, yielding the same in-core layout and alignment
849 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700850 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 result = sock_recvmsg(sock, msg, size, flags);
852 set_fs(oldfs);
853 return result;
854}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700855EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300857static ssize_t sock_sendpage(struct file *file, struct page *page,
858 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859{
860 struct socket *sock;
861 int flags;
862
Eric Dumazetb69aee02005-09-06 14:42:45 -0700863 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
Eric Dumazet35f9c092012-04-05 03:05:35 +0000865 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
866 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
867 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
Linus Torvaldse6949582009-08-13 08:28:36 -0700869 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870}
871
Jens Axboe9c55e012007-11-06 23:30:13 -0800872static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700873 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800874 unsigned int flags)
875{
876 struct socket *sock = file->private_data;
877
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800878 if (unlikely(!sock->ops->splice_read))
879 return -EINVAL;
880
Jens Axboe9c55e012007-11-06 23:30:13 -0800881 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
882}
883
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800884static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700885 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800886{
Kent Overstreetd29c4452013-03-18 11:09:26 -0700887 if (!is_sync_kiocb(iocb))
888 BUG();
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800889
890 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800891 iocb->private = siocb;
892 return siocb;
893}
894
895static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700896 struct file *file, const struct iovec *iov,
897 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800898{
899 struct socket *sock = file->private_data;
900 size_t size = 0;
901 int i;
902
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700903 for (i = 0; i < nr_segs; i++)
904 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800905
906 msg->msg_name = NULL;
907 msg->msg_namelen = 0;
908 msg->msg_control = NULL;
909 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700910 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800911 msg->msg_iovlen = nr_segs;
912 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
913
914 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
915}
916
Badari Pulavarty027445c2006-09-30 23:28:46 -0700917static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
918 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800919{
920 struct sock_iocb siocb, *x;
921
922 if (pos != 0)
923 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700924
Kent Overstreet73a70752013-05-09 15:03:42 -0700925 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800926 return 0;
927
Badari Pulavarty027445c2006-09-30 23:28:46 -0700928
929 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800930 if (!x)
931 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700932 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800933}
934
935static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700936 struct file *file, const struct iovec *iov,
937 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800938{
939 struct socket *sock = file->private_data;
940 size_t size = 0;
941 int i;
942
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700943 for (i = 0; i < nr_segs; i++)
944 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800945
946 msg->msg_name = NULL;
947 msg->msg_namelen = 0;
948 msg->msg_control = NULL;
949 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700950 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800951 msg->msg_iovlen = nr_segs;
952 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
953 if (sock->type == SOCK_SEQPACKET)
954 msg->msg_flags |= MSG_EOR;
955
956 return __sock_sendmsg(iocb, sock, msg, size);
957}
958
Badari Pulavarty027445c2006-09-30 23:28:46 -0700959static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
960 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800962 struct sock_iocb siocb, *x;
963
964 if (pos != 0)
965 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700966
Badari Pulavarty027445c2006-09-30 23:28:46 -0700967 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800968 if (!x)
969 return -ENOMEM;
970
Badari Pulavarty027445c2006-09-30 23:28:46 -0700971 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972}
973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974/*
975 * Atomic setting of ioctl hooks to avoid race
976 * with module unload.
977 */
978
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800979static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700980static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981
Eric W. Biederman881d9662007-09-17 11:56:21 -0700982void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800984 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800986 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987}
988EXPORT_SYMBOL(brioctl_set);
989
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800990static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700991static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992
Eric W. Biederman881d9662007-09-17 11:56:21 -0700993void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800995 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800997 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998}
999EXPORT_SYMBOL(vlan_ioctl_set);
1000
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001001static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001002static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001004void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001006 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001008 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
1010EXPORT_SYMBOL(dlci_ioctl_set);
1011
Arnd Bergmann6b960182009-11-06 23:10:54 -08001012static long sock_do_ioctl(struct net *net, struct socket *sock,
1013 unsigned int cmd, unsigned long arg)
1014{
1015 int err;
1016 void __user *argp = (void __user *)arg;
1017
1018 err = sock->ops->ioctl(sock, cmd, arg);
1019
1020 /*
1021 * If this ioctl is unknown try to hand it down
1022 * to the NIC driver.
1023 */
1024 if (err == -ENOIOCTLCMD)
1025 err = dev_ioctl(net, cmd, argp);
1026
1027 return err;
1028}
1029
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030/*
1031 * With an ioctl, arg may well be a user mode pointer, but we don't know
1032 * what to do with it - that's up to the protocol still.
1033 */
1034
1035static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1036{
1037 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001038 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 void __user *argp = (void __user *)arg;
1040 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001041 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042
Eric Dumazetb69aee02005-09-06 14:42:45 -07001043 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001044 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001045 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001047 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001049#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001051 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001053#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001054 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 case FIOSETOWN:
1056 case SIOCSPGRP:
1057 err = -EFAULT;
1058 if (get_user(pid, (int __user *)argp))
1059 break;
1060 err = f_setown(sock->file, pid, 1);
1061 break;
1062 case FIOGETOWN:
1063 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001064 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001065 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066 break;
1067 case SIOCGIFBR:
1068 case SIOCSIFBR:
1069 case SIOCBRADDBR:
1070 case SIOCBRDELBR:
1071 err = -ENOPKG;
1072 if (!br_ioctl_hook)
1073 request_module("bridge");
1074
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001075 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001076 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001077 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001078 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 break;
1080 case SIOCGIFVLAN:
1081 case SIOCSIFVLAN:
1082 err = -ENOPKG;
1083 if (!vlan_ioctl_hook)
1084 request_module("8021q");
1085
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001086 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001088 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001089 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 case SIOCADDDLCI:
1092 case SIOCDELDLCI:
1093 err = -ENOPKG;
1094 if (!dlci_ioctl_hook)
1095 request_module("dlci");
1096
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001097 mutex_lock(&dlci_ioctl_mutex);
1098 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001100 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 break;
1102 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001103 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001105 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 return err;
1107}
1108
1109int sock_create_lite(int family, int type, int protocol, struct socket **res)
1110{
1111 int err;
1112 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001113
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 err = security_socket_create(family, type, protocol, 1);
1115 if (err)
1116 goto out;
1117
1118 sock = sock_alloc();
1119 if (!sock) {
1120 err = -ENOMEM;
1121 goto out;
1122 }
1123
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001125 err = security_socket_post_create(sock, family, type, protocol, 1);
1126 if (err)
1127 goto out_release;
1128
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129out:
1130 *res = sock;
1131 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001132out_release:
1133 sock_release(sock);
1134 sock = NULL;
1135 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001137EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
1139/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001140static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141{
Eliezer Tamircbf55002013-07-08 16:20:34 +03001142 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 struct socket *sock;
1144
1145 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001146 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001148 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001149
Eliezer Tamircbf55002013-07-08 16:20:34 +03001150 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001151 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001152 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001153
1154 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001155 if (wait && (wait->_key & POLL_BUSY_LOOP))
1156 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001157 }
1158
Eliezer Tamircbf55002013-07-08 16:20:34 +03001159 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160}
1161
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001162static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001164 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165
1166 return sock->ops->mmap(file, sock, vma);
1167}
1168
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001169static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 sock_release(SOCKET_I(inode));
1172 return 0;
1173}
1174
1175/*
1176 * Update the socket async list
1177 *
1178 * Fasync_list locking strategy.
1179 *
1180 * 1. fasync_list is modified only under process context socket lock
1181 * i.e. under semaphore.
1182 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001183 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 */
1185
1186static int sock_fasync(int fd, struct file *filp, int on)
1187{
Eric Dumazet989a2972010-04-14 09:55:35 +00001188 struct socket *sock = filp->private_data;
1189 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001190 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
Eric Dumazet989a2972010-04-14 09:55:35 +00001192 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
1195 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001196 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1197 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
Eric Dumazeteaefd112011-02-18 03:26:36 +00001199 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001200 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001201 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001202 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
Eric Dumazet989a2972010-04-14 09:55:35 +00001204 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 return 0;
1206}
1207
Eric Dumazet43815482010-04-29 11:01:49 +00001208/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
1210int sock_wake_async(struct socket *sock, int how, int band)
1211{
Eric Dumazet43815482010-04-29 11:01:49 +00001212 struct socket_wq *wq;
1213
1214 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001216 rcu_read_lock();
1217 wq = rcu_dereference(sock->wq);
1218 if (!wq || !wq->fasync_list) {
1219 rcu_read_unlock();
1220 return -1;
1221 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001222 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001223 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1225 break;
1226 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001227 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1229 break;
1230 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001231 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001232call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001233 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001235 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001236 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 }
Eric Dumazet43815482010-04-29 11:01:49 +00001238 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 return 0;
1240}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001241EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242
Pavel Emelyanov721db932010-09-29 16:06:32 +04001243int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001244 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245{
1246 int err;
1247 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001248 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249
1250 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001251 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 */
1253 if (family < 0 || family >= NPROTO)
1254 return -EAFNOSUPPORT;
1255 if (type < 0 || type >= SOCK_MAX)
1256 return -EINVAL;
1257
1258 /* Compatibility.
1259
1260 This uglymoron is moved from INET layer to here to avoid
1261 deadlock in module load.
1262 */
1263 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001264 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 if (!warned) {
1266 warned = 1;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001267 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1268 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 }
1270 family = PF_PACKET;
1271 }
1272
1273 err = security_socket_create(family, type, protocol, kern);
1274 if (err)
1275 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001276
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001277 /*
1278 * Allocate the socket and allow the family to set things up. if
1279 * the protocol is 0, the family is instructed to select an appropriate
1280 * default.
1281 */
1282 sock = sock_alloc();
1283 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001284 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001285 return -ENFILE; /* Not exactly a match, but its the
1286 closest posix thing */
1287 }
1288
1289 sock->type = type;
1290
Johannes Berg95a5afc2008-10-16 15:24:51 -07001291#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001292 /* Attempt to load a protocol module if the find failed.
1293 *
1294 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 * requested real, full-featured networking support upon configuration.
1296 * Otherwise module support will break!
1297 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001298 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001299 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300#endif
1301
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001302 rcu_read_lock();
1303 pf = rcu_dereference(net_families[family]);
1304 err = -EAFNOSUPPORT;
1305 if (!pf)
1306 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307
1308 /*
1309 * We will call the ->create function, that possibly is in a loadable
1310 * module, so we have to bump that loadable module refcnt first.
1311 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001312 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 goto out_release;
1314
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001315 /* Now protected by module ref count */
1316 rcu_read_unlock();
1317
Eric Paris3f378b62009-11-05 22:18:14 -08001318 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001319 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001321
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 /*
1323 * Now to bump the refcnt of the [loadable] module that owns this
1324 * socket at sock_release time we decrement its refcnt.
1325 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001326 if (!try_module_get(sock->ops->owner))
1327 goto out_module_busy;
1328
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 /*
1330 * Now that we're done with the ->create function, the [loadable]
1331 * module can have its refcnt decremented
1332 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001333 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001334 err = security_socket_post_create(sock, family, type, protocol, kern);
1335 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001336 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001337 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001339 return 0;
1340
1341out_module_busy:
1342 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001344 sock->ops = NULL;
1345 module_put(pf->owner);
1346out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001348 return err;
1349
1350out_release:
1351 rcu_read_unlock();
1352 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001354EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
1356int sock_create(int family, int type, int protocol, struct socket **res)
1357{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001358 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001360EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
1362int sock_create_kern(int family, int type, int protocol, struct socket **res)
1363{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001364 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001366EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001368SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369{
1370 int retval;
1371 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001372 int flags;
1373
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001374 /* Check the SOCK_* constants for consistency. */
1375 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1376 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1377 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1378 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1379
Ulrich Dreppera677a032008-07-23 21:29:17 -07001380 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001381 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001382 return -EINVAL;
1383 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001385 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1386 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1387
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 retval = sock_create(family, type, protocol, &sock);
1389 if (retval < 0)
1390 goto out;
1391
Ulrich Drepper77d27202008-07-23 21:29:35 -07001392 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393 if (retval < 0)
1394 goto out_release;
1395
1396out:
1397 /* It may be already another descriptor 8) Not kernel problem. */
1398 return retval;
1399
1400out_release:
1401 sock_release(sock);
1402 return retval;
1403}
1404
1405/*
1406 * Create a pair of connected sockets.
1407 */
1408
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001409SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1410 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411{
1412 struct socket *sock1, *sock2;
1413 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001414 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001415 int flags;
1416
1417 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001418 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001419 return -EINVAL;
1420 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001422 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1423 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1424
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 /*
1426 * Obtain the first socket and check if the underlying protocol
1427 * supports the socketpair call.
1428 */
1429
1430 err = sock_create(family, type, protocol, &sock1);
1431 if (err < 0)
1432 goto out;
1433
1434 err = sock_create(family, type, protocol, &sock2);
1435 if (err < 0)
1436 goto out_release_1;
1437
1438 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001439 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 goto out_release_both;
1441
Al Viro28407632012-08-17 23:54:15 -04001442 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001443 if (unlikely(fd1 < 0)) {
1444 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001446 }
Al Viro28407632012-08-17 23:54:15 -04001447 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001448 if (unlikely(fd2 < 0)) {
1449 err = fd2;
Al Viro28407632012-08-17 23:54:15 -04001450 put_unused_fd(fd1);
1451 goto out_release_both;
1452 }
1453
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001454 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001455 if (unlikely(IS_ERR(newfile1))) {
1456 err = PTR_ERR(newfile1);
1457 put_unused_fd(fd1);
1458 put_unused_fd(fd2);
1459 goto out_release_both;
1460 }
1461
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001462 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001463 if (IS_ERR(newfile2)) {
1464 err = PTR_ERR(newfile2);
Al Viro198de4d2009-08-05 19:29:23 +04001465 fput(newfile1);
1466 put_unused_fd(fd1);
Al Viro28407632012-08-17 23:54:15 -04001467 put_unused_fd(fd2);
Al Viro198de4d2009-08-05 19:29:23 +04001468 sock_release(sock2);
1469 goto out;
Al Virodb349502007-02-07 01:48:00 -05001470 }
1471
Al Viro157cf642008-12-14 04:57:47 -05001472 audit_fd_pair(fd1, fd2);
Al Virodb349502007-02-07 01:48:00 -05001473 fd_install(fd1, newfile1);
1474 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 /* fd1 and fd2 may be already another descriptors.
1476 * Not kernel problem.
1477 */
1478
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001479 err = put_user(fd1, &usockvec[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 if (!err)
1481 err = put_user(fd2, &usockvec[1]);
1482 if (!err)
1483 return 0;
1484
1485 sys_close(fd2);
1486 sys_close(fd1);
1487 return err;
1488
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001490 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001492 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493out:
1494 return err;
1495}
1496
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497/*
1498 * Bind a name to a socket. Nothing much to do here since it's
1499 * the protocol's responsibility to handle the local address.
1500 *
1501 * We move the socket address to kernel space before we call
1502 * the protocol layer (having also checked the address is ok).
1503 */
1504
Heiko Carstens20f37032009-01-14 14:14:23 +01001505SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506{
1507 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001508 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001509 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001511 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001512 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001513 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001514 if (err >= 0) {
1515 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001516 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001517 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001518 if (!err)
1519 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001520 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001521 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001523 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001524 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 return err;
1526}
1527
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528/*
1529 * Perform a listen. Basically, we allow the protocol to do anything
1530 * necessary for a listen, and if that works, we mark the socket as
1531 * ready for listening.
1532 */
1533
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001534SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535{
1536 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001537 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001538 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001539
1540 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1541 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001542 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001543 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001544 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545
1546 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001547 if (!err)
1548 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001550 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 }
1552 return err;
1553}
1554
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555/*
1556 * For accept, we attempt to create a new socket, set up the link
1557 * with the client, wake up the client, then return the new
1558 * connected fd. We collect the address of the connector in kernel
1559 * space and move it to user at the very end. This is unclean because
1560 * we open the socket then return an error.
1561 *
1562 * 1003.1g adds the ability to recvmsg() to query connection pending
1563 * status to recvmsg. We need to add that support in a way thats
1564 * clean when we restucture accept also.
1565 */
1566
Heiko Carstens20f37032009-01-14 14:14:23 +01001567SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1568 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569{
1570 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001571 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001572 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001573 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574
Ulrich Drepper77d27202008-07-23 21:29:35 -07001575 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001576 return -EINVAL;
1577
1578 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1579 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1580
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001581 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 if (!sock)
1583 goto out;
1584
1585 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001586 newsock = sock_alloc();
1587 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 goto out_put;
1589
1590 newsock->type = sock->type;
1591 newsock->ops = sock->ops;
1592
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 /*
1594 * We don't need try_module_get here, as the listening socket (sock)
1595 * has the protocol module (sock->ops->owner) held.
1596 */
1597 __module_get(newsock->ops->owner);
1598
Al Viro28407632012-08-17 23:54:15 -04001599 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001600 if (unlikely(newfd < 0)) {
1601 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001602 sock_release(newsock);
1603 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001604 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001605 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001606 if (unlikely(IS_ERR(newfile))) {
1607 err = PTR_ERR(newfile);
1608 put_unused_fd(newfd);
1609 sock_release(newsock);
1610 goto out_put;
1611 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001612
Frank Filza79af592005-09-27 15:23:38 -07001613 err = security_socket_accept(sock, newsock);
1614 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001615 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001616
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1618 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001619 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620
1621 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001622 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001623 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001625 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001627 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001628 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001630 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 }
1632
1633 /* File flags are not inherited via accept() unlike another OSes. */
1634
David S. Miller39d8c1b2006-03-20 17:13:49 -08001635 fd_install(newfd, newfile);
1636 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001639 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640out:
1641 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001642out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001643 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001644 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 goto out_put;
1646}
1647
Heiko Carstens20f37032009-01-14 14:14:23 +01001648SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1649 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001650{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001651 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001652}
1653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654/*
1655 * Attempt to connect to a socket with the server address. The address
1656 * is in user space so we verify it is OK and move it to kernel space.
1657 *
1658 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1659 * break bindings
1660 *
1661 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1662 * other SEQPACKET protocols that take time to connect() as it doesn't
1663 * include the -EINPROGRESS status for such sockets.
1664 */
1665
Heiko Carstens20f37032009-01-14 14:14:23 +01001666SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1667 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668{
1669 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001670 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001671 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001673 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 if (!sock)
1675 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001676 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 if (err < 0)
1678 goto out_put;
1679
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001680 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001681 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 if (err)
1683 goto out_put;
1684
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001685 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 sock->file->f_flags);
1687out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001688 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689out:
1690 return err;
1691}
1692
1693/*
1694 * Get the local address ('name') of a socket object. Move the obtained
1695 * name to user space.
1696 */
1697
Heiko Carstens20f37032009-01-14 14:14:23 +01001698SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1699 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700{
1701 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001702 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001703 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001704
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001705 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 if (!sock)
1707 goto out;
1708
1709 err = security_socket_getsockname(sock);
1710 if (err)
1711 goto out_put;
1712
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001713 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (err)
1715 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001716 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717
1718out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001719 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720out:
1721 return err;
1722}
1723
1724/*
1725 * Get the remote address ('name') of a socket object. Move the obtained
1726 * name to user space.
1727 */
1728
Heiko Carstens20f37032009-01-14 14:14:23 +01001729SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1730 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
1732 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001733 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001734 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001736 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1737 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 err = security_socket_getpeername(sock);
1739 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001740 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 return err;
1742 }
1743
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001744 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001745 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001746 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001748 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001749 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001750 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 }
1752 return err;
1753}
1754
1755/*
1756 * Send a datagram to a given address. We move the address into kernel
1757 * space and check the user space data area is readable before invoking
1758 * the protocol.
1759 */
1760
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001761SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001762 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001763 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764{
1765 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001766 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 int err;
1768 struct msghdr msg;
1769 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001770 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001771
Linus Torvalds253eacc2010-10-30 16:43:10 -07001772 if (len > INT_MAX)
1773 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001774 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1775 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001776 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001777
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001778 iov.iov_base = buff;
1779 iov.iov_len = len;
1780 msg.msg_name = NULL;
1781 msg.msg_iov = &iov;
1782 msg.msg_iovlen = 1;
1783 msg.msg_control = NULL;
1784 msg.msg_controllen = 0;
1785 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001786 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001787 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788 if (err < 0)
1789 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001790 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001791 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 }
1793 if (sock->file->f_flags & O_NONBLOCK)
1794 flags |= MSG_DONTWAIT;
1795 msg.msg_flags = flags;
1796 err = sock_sendmsg(sock, &msg, len);
1797
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001798out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001799 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001800out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 return err;
1802}
1803
1804/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001805 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 */
1807
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001808SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001809 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810{
1811 return sys_sendto(fd, buff, len, flags, NULL, 0);
1812}
1813
1814/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001815 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 * sender. We verify the buffers are writable and if needed move the
1817 * sender address from kernel to user space.
1818 */
1819
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001820SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001821 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001822 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823{
1824 struct socket *sock;
1825 struct iovec iov;
1826 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001827 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001829 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830
Linus Torvalds253eacc2010-10-30 16:43:10 -07001831 if (size > INT_MAX)
1832 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001833 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001835 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001837 msg.msg_control = NULL;
1838 msg.msg_controllen = 0;
1839 msg.msg_iovlen = 1;
1840 msg.msg_iov = &iov;
1841 iov.iov_len = size;
1842 iov.iov_base = ubuf;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001843 msg.msg_name = (struct sockaddr *)&address;
1844 msg.msg_namelen = sizeof(address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 if (sock->file->f_flags & O_NONBLOCK)
1846 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001847 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001849 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001850 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001851 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001852 if (err2 < 0)
1853 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001855
1856 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001857out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001858 return err;
1859}
1860
1861/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001862 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 */
1864
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001865asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001866 unsigned int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867{
1868 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1869}
1870
1871/*
1872 * Set a socket option. Because we don't know the option lengths we have
1873 * to pass the user mode parameter for the protocols to sort out.
1874 */
1875
Heiko Carstens20f37032009-01-14 14:14:23 +01001876SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1877 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001879 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 struct socket *sock;
1881
1882 if (optlen < 0)
1883 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001884
1885 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1886 if (sock != NULL) {
1887 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001888 if (err)
1889 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001890
1891 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001892 err =
1893 sock_setsockopt(sock, level, optname, optval,
1894 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001896 err =
1897 sock->ops->setsockopt(sock, level, optname, optval,
1898 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001899out_put:
1900 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 }
1902 return err;
1903}
1904
1905/*
1906 * Get a socket option. Because we don't know the option lengths we have
1907 * to pass a user mode parameter for the protocols to sort out.
1908 */
1909
Heiko Carstens20f37032009-01-14 14:14:23 +01001910SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1911 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001913 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914 struct socket *sock;
1915
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001916 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1917 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001918 err = security_socket_getsockopt(sock, level, optname);
1919 if (err)
1920 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
1922 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001923 err =
1924 sock_getsockopt(sock, level, optname, optval,
1925 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001927 err =
1928 sock->ops->getsockopt(sock, level, optname, optval,
1929 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001930out_put:
1931 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932 }
1933 return err;
1934}
1935
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936/*
1937 * Shutdown a socket.
1938 */
1939
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001940SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001942 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 struct socket *sock;
1944
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001945 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1946 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001948 if (!err)
1949 err = sock->ops->shutdown(sock, how);
1950 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 }
1952 return err;
1953}
1954
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001955/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 * fields which are the same type (int / unsigned) on our platforms.
1957 */
1958#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1959#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1960#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1961
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001962struct used_address {
1963 struct sockaddr_storage name;
1964 unsigned int name_len;
1965};
1966
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001967static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001968 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001969 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001971 struct compat_msghdr __user *msg_compat =
1972 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001973 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001975 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001976 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1977 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02001979 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001980
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 err = -EFAULT;
1982 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00001983 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 return -EFAULT;
Anton Blanchard228e5482011-05-02 20:21:35 +00001985 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 return -EFAULT;
1987
Anton Blanchard228e5482011-05-02 20:21:35 +00001988 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02001989 err = -EMSGSIZE;
1990 if (msg_sys->msg_iovlen > UIO_MAXIOV)
1991 goto out;
1992 err = -ENOMEM;
1993 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
1994 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00001996 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997 }
1998
1999 /* This will also move the address data into kernel space */
2000 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002001 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002003 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002004 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005 goto out_freeiov;
2006 total_len = err;
2007
2008 err = -ENOBUFS;
2009
Anton Blanchard228e5482011-05-02 20:21:35 +00002010 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002012 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002014 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002015 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002016 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 if (err)
2018 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002019 ctl_buf = msg_sys->msg_control;
2020 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002022 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002024 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 goto out_freeiov;
2026 }
2027 err = -EFAULT;
2028 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002029 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2031 * checking falls down on this.
2032 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002033 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002034 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002035 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002037 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002039 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
2041 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002042 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002043 /*
2044 * If this is sendmmsg() and current destination address is same as
2045 * previously succeeded address, omit asking LSM's decision.
2046 * used_address->name_len is initialized to UINT_MAX so that the first
2047 * destination address never matches.
2048 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002049 if (used_address && msg_sys->msg_name &&
2050 used_address->name_len == msg_sys->msg_namelen &&
2051 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002052 used_address->name_len)) {
2053 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2054 goto out_freectl;
2055 }
2056 err = sock_sendmsg(sock, msg_sys, total_len);
2057 /*
2058 * If this is sendmmsg() and sending to current destination address was
2059 * successful, remember it.
2060 */
2061 if (used_address && err >= 0) {
2062 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002063 if (msg_sys->msg_name)
2064 memcpy(&used_address->name, msg_sys->msg_name,
2065 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002066 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067
2068out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002069 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2071out_freeiov:
2072 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002073 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002074out:
2075 return err;
2076}
2077
2078/*
2079 * BSD sendmsg interface
2080 */
2081
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002082long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002083{
2084 int fput_needed, err;
2085 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002086 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002087
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002088 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002089 if (!sock)
2090 goto out;
2091
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002092 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002093
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002094 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002095out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 return err;
2097}
2098
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002099SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2100{
2101 if (flags & MSG_CMSG_COMPAT)
2102 return -EINVAL;
2103 return __sys_sendmsg(fd, msg, flags);
2104}
2105
Anton Blanchard228e5482011-05-02 20:21:35 +00002106/*
2107 * Linux sendmmsg interface
2108 */
2109
2110int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2111 unsigned int flags)
2112{
2113 int fput_needed, err, datagrams;
2114 struct socket *sock;
2115 struct mmsghdr __user *entry;
2116 struct compat_mmsghdr __user *compat_entry;
2117 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002118 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002119
Anton Blanchard98382f42011-08-04 14:07:39 +00002120 if (vlen > UIO_MAXIOV)
2121 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002122
2123 datagrams = 0;
2124
2125 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2126 if (!sock)
2127 return err;
2128
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002129 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002130 entry = mmsg;
2131 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002132 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002133
2134 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002135 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002136 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2137 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002138 if (err < 0)
2139 break;
2140 err = __put_user(err, &compat_entry->msg_len);
2141 ++compat_entry;
2142 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002143 err = ___sys_sendmsg(sock,
2144 (struct msghdr __user *)entry,
2145 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002146 if (err < 0)
2147 break;
2148 err = put_user(err, &entry->msg_len);
2149 ++entry;
2150 }
2151
2152 if (err)
2153 break;
2154 ++datagrams;
2155 }
2156
Anton Blanchard228e5482011-05-02 20:21:35 +00002157 fput_light(sock->file, fput_needed);
2158
Anton Blanchard728ffb82011-08-04 14:07:38 +00002159 /* We only return an error if no datagrams were able to be sent */
2160 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002161 return datagrams;
2162
Anton Blanchard228e5482011-05-02 20:21:35 +00002163 return err;
2164}
2165
2166SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2167 unsigned int, vlen, unsigned int, flags)
2168{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002169 if (flags & MSG_CMSG_COMPAT)
2170 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002171 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2172}
2173
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002174static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002175 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002177 struct compat_msghdr __user *msg_compat =
2178 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002180 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002182 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183
2184 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002185 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186
2187 /* user mode address pointers */
2188 struct sockaddr __user *uaddr;
2189 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002190
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002192 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 return -EFAULT;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002194 } else if (copy_from_user(msg_sys, msg, sizeof(struct msghdr)))
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002195 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002197 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002198 err = -EMSGSIZE;
2199 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2200 goto out;
2201 err = -ENOMEM;
2202 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2203 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002205 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206 }
2207
2208 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002209 * Save the user-mode address (verify_iovec will change the
2210 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002212
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002213 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 uaddr_len = COMPAT_NAMELEN(msg);
2215 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002216 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002218 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 if (err < 0)
2220 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002221 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002223 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2224 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002225
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226 if (sock->file->f_flags & O_NONBLOCK)
2227 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002228 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2229 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230 if (err < 0)
2231 goto out_freeiov;
2232 len = err;
2233
2234 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002235 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002236 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002237 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238 if (err < 0)
2239 goto out_freeiov;
2240 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002241 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002242 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243 if (err)
2244 goto out_freeiov;
2245 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002246 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247 &msg_compat->msg_controllen);
2248 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002249 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 &msg->msg_controllen);
2251 if (err)
2252 goto out_freeiov;
2253 err = len;
2254
2255out_freeiov:
2256 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002257 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002258out:
2259 return err;
2260}
2261
2262/*
2263 * BSD recvmsg interface
2264 */
2265
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002266long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002267{
2268 int fput_needed, err;
2269 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002270 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002271
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002272 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002273 if (!sock)
2274 goto out;
2275
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002276 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002277
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002278 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279out:
2280 return err;
2281}
2282
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002283SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2284 unsigned int, flags)
2285{
2286 if (flags & MSG_CMSG_COMPAT)
2287 return -EINVAL;
2288 return __sys_recvmsg(fd, msg, flags);
2289}
2290
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002291/*
2292 * Linux recvmmsg interface
2293 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002295int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2296 unsigned int flags, struct timespec *timeout)
2297{
2298 int fput_needed, err, datagrams;
2299 struct socket *sock;
2300 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002301 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002302 struct msghdr msg_sys;
2303 struct timespec end_time;
2304
2305 if (timeout &&
2306 poll_select_set_timeout(&end_time, timeout->tv_sec,
2307 timeout->tv_nsec))
2308 return -EINVAL;
2309
2310 datagrams = 0;
2311
2312 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2313 if (!sock)
2314 return err;
2315
2316 err = sock_error(sock->sk);
2317 if (err)
2318 goto out_put;
2319
2320 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002321 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002322
2323 while (datagrams < vlen) {
2324 /*
2325 * No need to ask LSM for more than the first datagram.
2326 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002327 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002328 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2329 &msg_sys, flags & ~MSG_WAITFORONE,
2330 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002331 if (err < 0)
2332 break;
2333 err = __put_user(err, &compat_entry->msg_len);
2334 ++compat_entry;
2335 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002336 err = ___sys_recvmsg(sock,
2337 (struct msghdr __user *)entry,
2338 &msg_sys, flags & ~MSG_WAITFORONE,
2339 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002340 if (err < 0)
2341 break;
2342 err = put_user(err, &entry->msg_len);
2343 ++entry;
2344 }
2345
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002346 if (err)
2347 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002348 ++datagrams;
2349
Brandon L Black71c5c152010-03-26 16:18:03 +00002350 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2351 if (flags & MSG_WAITFORONE)
2352 flags |= MSG_DONTWAIT;
2353
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002354 if (timeout) {
2355 ktime_get_ts(timeout);
2356 *timeout = timespec_sub(end_time, *timeout);
2357 if (timeout->tv_sec < 0) {
2358 timeout->tv_sec = timeout->tv_nsec = 0;
2359 break;
2360 }
2361
2362 /* Timeout, return less than vlen datagrams */
2363 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2364 break;
2365 }
2366
2367 /* Out of band data, return right away */
2368 if (msg_sys.msg_flags & MSG_OOB)
2369 break;
2370 }
2371
2372out_put:
2373 fput_light(sock->file, fput_needed);
2374
2375 if (err == 0)
2376 return datagrams;
2377
2378 if (datagrams != 0) {
2379 /*
2380 * We may return less entries than requested (vlen) if the
2381 * sock is non block and there aren't enough datagrams...
2382 */
2383 if (err != -EAGAIN) {
2384 /*
2385 * ... or if recvmsg returns an error after we
2386 * received some datagrams, where we record the
2387 * error to return on the next call or if the
2388 * app asks about it using getsockopt(SO_ERROR).
2389 */
2390 sock->sk->sk_err = -err;
2391 }
2392
2393 return datagrams;
2394 }
2395
2396 return err;
2397}
2398
2399SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2400 unsigned int, vlen, unsigned int, flags,
2401 struct timespec __user *, timeout)
2402{
2403 int datagrams;
2404 struct timespec timeout_sys;
2405
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002406 if (flags & MSG_CMSG_COMPAT)
2407 return -EINVAL;
2408
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002409 if (!timeout)
2410 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2411
2412 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2413 return -EFAULT;
2414
2415 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2416
2417 if (datagrams > 0 &&
2418 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2419 datagrams = -EFAULT;
2420
2421 return datagrams;
2422}
2423
2424#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425/* Argument list sizes for sys_socketcall */
2426#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002427static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002428 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2429 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2430 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002431 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002432};
2433
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434#undef AL
2435
2436/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002437 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438 *
2439 * Argument checking cleaned up. Saved 20% in size.
2440 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002441 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 */
2443
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002444SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445{
Chen Gang2950fa92013-04-07 16:55:23 +08002446 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002447 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002449 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450
Anton Blanchard228e5482011-05-02 20:21:35 +00002451 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 return -EINVAL;
2453
Arjan van de Ven47379052009-09-28 12:57:44 -07002454 len = nargs[call];
2455 if (len > sizeof(a))
2456 return -EINVAL;
2457
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002459 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002461
Chen Gang2950fa92013-04-07 16:55:23 +08002462 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2463 if (err)
2464 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002465
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002466 a0 = a[0];
2467 a1 = a[1];
2468
2469 switch (call) {
2470 case SYS_SOCKET:
2471 err = sys_socket(a0, a1, a[2]);
2472 break;
2473 case SYS_BIND:
2474 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2475 break;
2476 case SYS_CONNECT:
2477 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2478 break;
2479 case SYS_LISTEN:
2480 err = sys_listen(a0, a1);
2481 break;
2482 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002483 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2484 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002485 break;
2486 case SYS_GETSOCKNAME:
2487 err =
2488 sys_getsockname(a0, (struct sockaddr __user *)a1,
2489 (int __user *)a[2]);
2490 break;
2491 case SYS_GETPEERNAME:
2492 err =
2493 sys_getpeername(a0, (struct sockaddr __user *)a1,
2494 (int __user *)a[2]);
2495 break;
2496 case SYS_SOCKETPAIR:
2497 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2498 break;
2499 case SYS_SEND:
2500 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2501 break;
2502 case SYS_SENDTO:
2503 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2504 (struct sockaddr __user *)a[4], a[5]);
2505 break;
2506 case SYS_RECV:
2507 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2508 break;
2509 case SYS_RECVFROM:
2510 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2511 (struct sockaddr __user *)a[4],
2512 (int __user *)a[5]);
2513 break;
2514 case SYS_SHUTDOWN:
2515 err = sys_shutdown(a0, a1);
2516 break;
2517 case SYS_SETSOCKOPT:
2518 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2519 break;
2520 case SYS_GETSOCKOPT:
2521 err =
2522 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2523 (int __user *)a[4]);
2524 break;
2525 case SYS_SENDMSG:
2526 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2527 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002528 case SYS_SENDMMSG:
2529 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2530 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002531 case SYS_RECVMSG:
2532 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2533 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002534 case SYS_RECVMMSG:
2535 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2536 (struct timespec __user *)a[4]);
2537 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002538 case SYS_ACCEPT4:
2539 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2540 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002541 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002542 default:
2543 err = -EINVAL;
2544 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545 }
2546 return err;
2547}
2548
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002549#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002551/**
2552 * sock_register - add a socket protocol handler
2553 * @ops: description of protocol
2554 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555 * This function is called by a protocol handler that wants to
2556 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002557 * socket interface. The value ops->family coresponds to the
2558 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002560int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561{
2562 int err;
2563
2564 if (ops->family >= NPROTO) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002565 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2566 NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567 return -ENOBUFS;
2568 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002569
2570 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002571 if (rcu_dereference_protected(net_families[ops->family],
2572 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002573 err = -EEXIST;
2574 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002575 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576 err = 0;
2577 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002578 spin_unlock(&net_family_lock);
2579
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002580 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002581 return err;
2582}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002583EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002585/**
2586 * sock_unregister - remove a protocol handler
2587 * @family: protocol family to remove
2588 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002589 * This function is called by a protocol handler that wants to
2590 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002591 * new socket creation.
2592 *
2593 * If protocol handler is a module, then it can use module reference
2594 * counts to protect against new references. If protocol handler is not
2595 * a module then it needs to provide its own protection in
2596 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002598void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002599{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002600 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002602 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002603 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002604 spin_unlock(&net_family_lock);
2605
2606 synchronize_rcu();
2607
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002608 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002610EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611
Andi Kleen77d76ea2005-12-22 12:43:42 -08002612static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002613{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002614 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002615 /*
2616 * Initialize the network sysctl infrastructure.
2617 */
2618 err = net_sysctl_init();
2619 if (err)
2620 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002621
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002623 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624 */
2625 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002626
2627 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002628 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 */
2630
2631 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002632
2633 err = register_filesystem(&sock_fs_type);
2634 if (err)
2635 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002636 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002637 if (IS_ERR(sock_mnt)) {
2638 err = PTR_ERR(sock_mnt);
2639 goto out_mount;
2640 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002641
2642 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643 */
2644
2645#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002646 err = netfilter_init();
2647 if (err)
2648 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002649#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002650
Richard Cochranc1f19b52010-07-17 08:49:36 +00002651#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2652 skb_timestamping_init();
2653#endif
2654
Nick Pigginb3e19d92011-01-07 17:50:11 +11002655out:
2656 return err;
2657
2658out_mount:
2659 unregister_filesystem(&sock_fs_type);
2660out_fs:
2661 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662}
2663
Andi Kleen77d76ea2005-12-22 12:43:42 -08002664core_initcall(sock_init); /* early initcall */
2665
Linus Torvalds1da177e2005-04-16 15:20:36 -07002666#ifdef CONFIG_PROC_FS
2667void socket_seq_show(struct seq_file *seq)
2668{
2669 int cpu;
2670 int counter = 0;
2671
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002672 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002673 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002674
2675 /* It can be negative, by the way. 8) */
2676 if (counter < 0)
2677 counter = 0;
2678
2679 seq_printf(seq, "sockets: used %d\n", counter);
2680}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002681#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002682
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002683#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002684static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002685 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002686{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002687 mm_segment_t old_fs = get_fs();
2688 struct timeval ktv;
2689 int err;
2690
2691 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002692 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002693 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002694 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002695 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002696
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002697 return err;
2698}
2699
Arnd Bergmann6b960182009-11-06 23:10:54 -08002700static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002701 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002702{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002703 mm_segment_t old_fs = get_fs();
2704 struct timespec kts;
2705 int err;
2706
2707 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002708 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002709 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002710 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002711 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002712
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002713 return err;
2714}
2715
Arnd Bergmann6b960182009-11-06 23:10:54 -08002716static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002717{
2718 struct ifreq __user *uifr;
2719 int err;
2720
2721 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002722 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002723 return -EFAULT;
2724
Arnd Bergmann6b960182009-11-06 23:10:54 -08002725 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002726 if (err)
2727 return err;
2728
Arnd Bergmann6b960182009-11-06 23:10:54 -08002729 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002730 return -EFAULT;
2731
2732 return 0;
2733}
2734
Arnd Bergmann6b960182009-11-06 23:10:54 -08002735static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002736{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002737 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002738 struct ifconf ifc;
2739 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002740 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002741 struct ifreq __user *ifr;
2742 unsigned int i, j;
2743 int err;
2744
Arnd Bergmann6b960182009-11-06 23:10:54 -08002745 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002746 return -EFAULT;
2747
Mathias Krause43da5f22012-08-15 11:31:57 +00002748 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002749 if (ifc32.ifcbuf == 0) {
2750 ifc32.ifc_len = 0;
2751 ifc.ifc_len = 0;
2752 ifc.ifc_req = NULL;
2753 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2754 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002755 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2756 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002757 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2758 ifc.ifc_len = len;
2759 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2760 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002761 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002762 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002763 return -EFAULT;
2764 ifr++;
2765 ifr32++;
2766 }
2767 }
2768 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2769 return -EFAULT;
2770
Arnd Bergmann6b960182009-11-06 23:10:54 -08002771 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002772 if (err)
2773 return err;
2774
2775 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2776 return -EFAULT;
2777
2778 ifr = ifc.ifc_req;
2779 ifr32 = compat_ptr(ifc32.ifcbuf);
2780 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002781 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2782 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2783 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002784 return -EFAULT;
2785 ifr32++;
2786 ifr++;
2787 }
2788
2789 if (ifc32.ifcbuf == 0) {
2790 /* Translate from 64-bit structure multiple to
2791 * a 32-bit one.
2792 */
2793 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002794 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002795 ifc32.ifc_len = i;
2796 } else {
2797 ifc32.ifc_len = i;
2798 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002799 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002800 return -EFAULT;
2801
2802 return 0;
2803}
2804
Arnd Bergmann6b960182009-11-06 23:10:54 -08002805static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002806{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002807 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2808 bool convert_in = false, convert_out = false;
2809 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2810 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002811 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002812 u32 rule_cnt = 0, actual_rule_cnt;
2813 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002814 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002815 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002816
2817 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2818 return -EFAULT;
2819
Ben Hutchings3a7da392011-03-17 07:34:32 +00002820 compat_rxnfc = compat_ptr(data);
2821
2822 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002823 return -EFAULT;
2824
Ben Hutchings3a7da392011-03-17 07:34:32 +00002825 /* Most ethtool structures are defined without padding.
2826 * Unfortunately struct ethtool_rxnfc is an exception.
2827 */
2828 switch (ethcmd) {
2829 default:
2830 break;
2831 case ETHTOOL_GRXCLSRLALL:
2832 /* Buffer size is variable */
2833 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2834 return -EFAULT;
2835 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2836 return -ENOMEM;
2837 buf_size += rule_cnt * sizeof(u32);
2838 /* fall through */
2839 case ETHTOOL_GRXRINGS:
2840 case ETHTOOL_GRXCLSRLCNT:
2841 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002842 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002843 convert_out = true;
2844 /* fall through */
2845 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002846 buf_size += sizeof(struct ethtool_rxnfc);
2847 convert_in = true;
2848 break;
2849 }
2850
2851 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002852 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002853
2854 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2855 return -EFAULT;
2856
2857 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2858 &ifr->ifr_ifru.ifru_data))
2859 return -EFAULT;
2860
2861 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002862 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002863 * fs.ring_cookie and at the end of fs, but nowhere else.
2864 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002865 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2866 sizeof(compat_rxnfc->fs.m_ext) !=
2867 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2868 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002869 BUILD_BUG_ON(
2870 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2871 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2872 offsetof(struct ethtool_rxnfc, fs.location) -
2873 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2874
2875 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002876 (void __user *)(&rxnfc->fs.m_ext + 1) -
2877 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002878 copy_in_user(&rxnfc->fs.ring_cookie,
2879 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002880 (void __user *)(&rxnfc->fs.location + 1) -
2881 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002882 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2883 sizeof(rxnfc->rule_cnt)))
2884 return -EFAULT;
2885 }
2886
2887 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2888 if (ret)
2889 return ret;
2890
2891 if (convert_out) {
2892 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002893 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2894 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002895 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2896 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002897 (const void __user *)(&rxnfc->fs.location + 1) -
2898 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002899 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2900 sizeof(rxnfc->rule_cnt)))
2901 return -EFAULT;
2902
2903 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2904 /* As an optimisation, we only copy the actual
2905 * number of rules that the underlying
2906 * function returned. Since Mallory might
2907 * change the rule count in user memory, we
2908 * check that it is less than the rule count
2909 * originally given (as the user buffer size),
2910 * which has been range-checked.
2911 */
2912 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2913 return -EFAULT;
2914 if (actual_rule_cnt < rule_cnt)
2915 rule_cnt = actual_rule_cnt;
2916 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2917 &rxnfc->rule_locs[0],
2918 rule_cnt * sizeof(u32)))
2919 return -EFAULT;
2920 }
2921 }
2922
2923 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002924}
2925
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002926static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2927{
2928 void __user *uptr;
2929 compat_uptr_t uptr32;
2930 struct ifreq __user *uifr;
2931
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002932 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002933 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2934 return -EFAULT;
2935
2936 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2937 return -EFAULT;
2938
2939 uptr = compat_ptr(uptr32);
2940
2941 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2942 return -EFAULT;
2943
2944 return dev_ioctl(net, SIOCWANDEV, uifr);
2945}
2946
Arnd Bergmann6b960182009-11-06 23:10:54 -08002947static int bond_ioctl(struct net *net, unsigned int cmd,
2948 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002949{
2950 struct ifreq kifr;
2951 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002952 mm_segment_t old_fs;
2953 int err;
2954 u32 data;
2955 void __user *datap;
2956
2957 switch (cmd) {
2958 case SIOCBONDENSLAVE:
2959 case SIOCBONDRELEASE:
2960 case SIOCBONDSETHWADDR:
2961 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002962 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002963 return -EFAULT;
2964
2965 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002966 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002967 err = dev_ioctl(net, cmd,
2968 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002969 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002970
2971 return err;
2972 case SIOCBONDSLAVEINFOQUERY:
2973 case SIOCBONDINFOQUERY:
2974 uifr = compat_alloc_user_space(sizeof(*uifr));
2975 if (copy_in_user(&uifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2976 return -EFAULT;
2977
2978 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2979 return -EFAULT;
2980
2981 datap = compat_ptr(data);
2982 if (put_user(datap, &uifr->ifr_ifru.ifru_data))
2983 return -EFAULT;
2984
Arnd Bergmann6b960182009-11-06 23:10:54 -08002985 return dev_ioctl(net, cmd, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002986 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002987 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002988 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002989}
2990
Arnd Bergmann6b960182009-11-06 23:10:54 -08002991static int siocdevprivate_ioctl(struct net *net, unsigned int cmd,
2992 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002993{
2994 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002995 char tmp_buf[IFNAMSIZ];
2996 void __user *data64;
2997 u32 data32;
2998
2999 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3000 IFNAMSIZ))
3001 return -EFAULT;
3002 if (__get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
3003 return -EFAULT;
3004 data64 = compat_ptr(data32);
3005
3006 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3007
3008 /* Don't check these user accesses, just let that get trapped
3009 * in the ioctl handler instead.
3010 */
3011 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3012 IFNAMSIZ))
3013 return -EFAULT;
3014 if (__put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
3015 return -EFAULT;
3016
Arnd Bergmann6b960182009-11-06 23:10:54 -08003017 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003018}
3019
Arnd Bergmann6b960182009-11-06 23:10:54 -08003020static int dev_ifsioc(struct net *net, struct socket *sock,
3021 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003022{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003023 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003024 int err;
3025
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003026 uifr = compat_alloc_user_space(sizeof(*uifr));
3027 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3028 return -EFAULT;
3029
3030 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3031
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003032 if (!err) {
3033 switch (cmd) {
3034 case SIOCGIFFLAGS:
3035 case SIOCGIFMETRIC:
3036 case SIOCGIFMTU:
3037 case SIOCGIFMEM:
3038 case SIOCGIFHWADDR:
3039 case SIOCGIFINDEX:
3040 case SIOCGIFADDR:
3041 case SIOCGIFBRDADDR:
3042 case SIOCGIFDSTADDR:
3043 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003044 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003045 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003046 case SIOCGMIIPHY:
3047 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003048 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003049 err = -EFAULT;
3050 break;
3051 }
3052 }
3053 return err;
3054}
3055
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003056static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3057 struct compat_ifreq __user *uifr32)
3058{
3059 struct ifreq ifr;
3060 struct compat_ifmap __user *uifmap32;
3061 mm_segment_t old_fs;
3062 int err;
3063
3064 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3065 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003066 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3067 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3068 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3069 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3070 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3071 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003072 if (err)
3073 return -EFAULT;
3074
3075 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003076 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003077 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003078 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003079
3080 if (cmd == SIOCGIFMAP && !err) {
3081 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003082 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3083 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3084 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3085 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3086 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3087 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003088 if (err)
3089 err = -EFAULT;
3090 }
3091 return err;
3092}
3093
3094static int compat_siocshwtstamp(struct net *net, struct compat_ifreq __user *uifr32)
3095{
3096 void __user *uptr;
3097 compat_uptr_t uptr32;
3098 struct ifreq __user *uifr;
3099
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003100 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003101 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
3102 return -EFAULT;
3103
3104 if (get_user(uptr32, &uifr32->ifr_data))
3105 return -EFAULT;
3106
3107 uptr = compat_ptr(uptr32);
3108
3109 if (put_user(uptr, &uifr->ifr_data))
3110 return -EFAULT;
3111
3112 return dev_ioctl(net, SIOCSHWTSTAMP, uifr);
3113}
3114
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003115struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003116 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003117 struct sockaddr rt_dst; /* target address */
3118 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3119 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003120 unsigned short rt_flags;
3121 short rt_pad2;
3122 u32 rt_pad3;
3123 unsigned char rt_tos;
3124 unsigned char rt_class;
3125 short rt_pad4;
3126 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003127 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003128 u32 rt_mtu; /* per route MTU/Window */
3129 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003130 unsigned short rt_irtt; /* Initial RTT */
3131};
3132
3133struct in6_rtmsg32 {
3134 struct in6_addr rtmsg_dst;
3135 struct in6_addr rtmsg_src;
3136 struct in6_addr rtmsg_gateway;
3137 u32 rtmsg_type;
3138 u16 rtmsg_dst_len;
3139 u16 rtmsg_src_len;
3140 u32 rtmsg_metric;
3141 u32 rtmsg_info;
3142 u32 rtmsg_flags;
3143 s32 rtmsg_ifindex;
3144};
3145
Arnd Bergmann6b960182009-11-06 23:10:54 -08003146static int routing_ioctl(struct net *net, struct socket *sock,
3147 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003148{
3149 int ret;
3150 void *r = NULL;
3151 struct in6_rtmsg r6;
3152 struct rtentry r4;
3153 char devname[16];
3154 u32 rtdev;
3155 mm_segment_t old_fs = get_fs();
3156
Arnd Bergmann6b960182009-11-06 23:10:54 -08003157 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3158 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003159 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003160 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003161 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3162 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3163 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3164 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3165 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3166 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3167 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003168
3169 r = (void *) &r6;
3170 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003171 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003172 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003173 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003174 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3175 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3176 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3177 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3178 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3179 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003180 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003181 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003182 r4.rt_dev = (char __user __force *)devname;
3183 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003184 } else
3185 r4.rt_dev = NULL;
3186
3187 r = (void *) &r4;
3188 }
3189
3190 if (ret) {
3191 ret = -EFAULT;
3192 goto out;
3193 }
3194
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003195 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003196 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003197 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003198
3199out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003200 return ret;
3201}
3202
3203/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3204 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003205 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003206 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003207static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003208{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003209 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003210
Arnd Bergmann6b960182009-11-06 23:10:54 -08003211 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003212 return -EFAULT;
3213 if (tmp == BRCTL_GET_VERSION)
3214 return BRCTL_VERSION + 1;
3215 return -EINVAL;
3216}
3217
Arnd Bergmann6b960182009-11-06 23:10:54 -08003218static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3219 unsigned int cmd, unsigned long arg)
3220{
3221 void __user *argp = compat_ptr(arg);
3222 struct sock *sk = sock->sk;
3223 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003224
Arnd Bergmann6b960182009-11-06 23:10:54 -08003225 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3226 return siocdevprivate_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003227
Arnd Bergmann6b960182009-11-06 23:10:54 -08003228 switch (cmd) {
3229 case SIOCSIFBR:
3230 case SIOCGIFBR:
3231 return old_bridge_ioctl(argp);
3232 case SIOCGIFNAME:
3233 return dev_ifname32(net, argp);
3234 case SIOCGIFCONF:
3235 return dev_ifconf(net, argp);
3236 case SIOCETHTOOL:
3237 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003238 case SIOCWANDEV:
3239 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003240 case SIOCGIFMAP:
3241 case SIOCSIFMAP:
3242 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003243 case SIOCBONDENSLAVE:
3244 case SIOCBONDRELEASE:
3245 case SIOCBONDSETHWADDR:
3246 case SIOCBONDSLAVEINFOQUERY:
3247 case SIOCBONDINFOQUERY:
3248 case SIOCBONDCHANGEACTIVE:
3249 return bond_ioctl(net, cmd, argp);
3250 case SIOCADDRT:
3251 case SIOCDELRT:
3252 return routing_ioctl(net, sock, cmd, argp);
3253 case SIOCGSTAMP:
3254 return do_siocgstamp(net, sock, cmd, argp);
3255 case SIOCGSTAMPNS:
3256 return do_siocgstampns(net, sock, cmd, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003257 case SIOCSHWTSTAMP:
3258 return compat_siocshwtstamp(net, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003259
Arnd Bergmann6b960182009-11-06 23:10:54 -08003260 case FIOSETOWN:
3261 case SIOCSPGRP:
3262 case FIOGETOWN:
3263 case SIOCGPGRP:
3264 case SIOCBRADDBR:
3265 case SIOCBRDELBR:
3266 case SIOCGIFVLAN:
3267 case SIOCSIFVLAN:
3268 case SIOCADDDLCI:
3269 case SIOCDELDLCI:
3270 return sock_ioctl(file, cmd, arg);
3271
3272 case SIOCGIFFLAGS:
3273 case SIOCSIFFLAGS:
3274 case SIOCGIFMETRIC:
3275 case SIOCSIFMETRIC:
3276 case SIOCGIFMTU:
3277 case SIOCSIFMTU:
3278 case SIOCGIFMEM:
3279 case SIOCSIFMEM:
3280 case SIOCGIFHWADDR:
3281 case SIOCSIFHWADDR:
3282 case SIOCADDMULTI:
3283 case SIOCDELMULTI:
3284 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003285 case SIOCGIFADDR:
3286 case SIOCSIFADDR:
3287 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003288 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003289 case SIOCGIFBRDADDR:
3290 case SIOCSIFBRDADDR:
3291 case SIOCGIFDSTADDR:
3292 case SIOCSIFDSTADDR:
3293 case SIOCGIFNETMASK:
3294 case SIOCSIFNETMASK:
3295 case SIOCSIFPFLAGS:
3296 case SIOCGIFPFLAGS:
3297 case SIOCGIFTXQLEN:
3298 case SIOCSIFTXQLEN:
3299 case SIOCBRADDIF:
3300 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003301 case SIOCSIFNAME:
3302 case SIOCGMIIPHY:
3303 case SIOCGMIIREG:
3304 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003305 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003306
Arnd Bergmann6b960182009-11-06 23:10:54 -08003307 case SIOCSARP:
3308 case SIOCGARP:
3309 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003310 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003311 return sock_do_ioctl(net, sock, cmd, arg);
3312 }
3313
Arnd Bergmann6b960182009-11-06 23:10:54 -08003314 return -ENOIOCTLCMD;
3315}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003316
Eric Dumazet95c96172012-04-15 05:58:06 +00003317static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003318 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003319{
3320 struct socket *sock = file->private_data;
3321 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003322 struct sock *sk;
3323 struct net *net;
3324
3325 sk = sock->sk;
3326 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003327
3328 if (sock->ops->compat_ioctl)
3329 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3330
David S. Miller87de87d2008-06-03 09:14:03 -07003331 if (ret == -ENOIOCTLCMD &&
3332 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3333 ret = compat_wext_handle_ioctl(net, cmd, arg);
3334
Arnd Bergmann6b960182009-11-06 23:10:54 -08003335 if (ret == -ENOIOCTLCMD)
3336 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3337
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003338 return ret;
3339}
3340#endif
3341
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003342int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3343{
3344 return sock->ops->bind(sock, addr, addrlen);
3345}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003346EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003347
3348int kernel_listen(struct socket *sock, int backlog)
3349{
3350 return sock->ops->listen(sock, backlog);
3351}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003352EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003353
3354int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3355{
3356 struct sock *sk = sock->sk;
3357 int err;
3358
3359 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3360 newsock);
3361 if (err < 0)
3362 goto done;
3363
3364 err = sock->ops->accept(sock, *newsock, flags);
3365 if (err < 0) {
3366 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003367 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003368 goto done;
3369 }
3370
3371 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003372 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003373
3374done:
3375 return err;
3376}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003377EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003378
3379int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003380 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003381{
3382 return sock->ops->connect(sock, addr, addrlen, flags);
3383}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003384EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003385
3386int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3387 int *addrlen)
3388{
3389 return sock->ops->getname(sock, addr, addrlen, 0);
3390}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003391EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003392
3393int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3394 int *addrlen)
3395{
3396 return sock->ops->getname(sock, addr, addrlen, 1);
3397}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003398EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003399
3400int kernel_getsockopt(struct socket *sock, int level, int optname,
3401 char *optval, int *optlen)
3402{
3403 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003404 char __user *uoptval;
3405 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003406 int err;
3407
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003408 uoptval = (char __user __force *) optval;
3409 uoptlen = (int __user __force *) optlen;
3410
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003411 set_fs(KERNEL_DS);
3412 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003413 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003414 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003415 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3416 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003417 set_fs(oldfs);
3418 return err;
3419}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003420EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003421
3422int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003423 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003424{
3425 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003426 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003427 int err;
3428
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003429 uoptval = (char __user __force *) optval;
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_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003434 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003435 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003436 optlen);
3437 set_fs(oldfs);
3438 return err;
3439}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003440EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003441
3442int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3443 size_t size, int flags)
3444{
3445 if (sock->ops->sendpage)
3446 return sock->ops->sendpage(sock, page, offset, size, flags);
3447
3448 return sock_no_sendpage(sock, page, offset, size, flags);
3449}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003450EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003451
3452int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3453{
3454 mm_segment_t oldfs = get_fs();
3455 int err;
3456
3457 set_fs(KERNEL_DS);
3458 err = sock->ops->ioctl(sock, cmd, arg);
3459 set_fs(oldfs);
3460
3461 return err;
3462}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003463EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003464
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003465int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3466{
3467 return sock->ops->shutdown(sock, how);
3468}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003469EXPORT_SYMBOL(kernel_sock_shutdown);