blob: bbedbfcb42c2505fceb57fa058f262d90e1670ed [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>
Daniel Borkmann408eccc2014-04-01 16:20:23 +020075#include <linux/ptp_classify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070088#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070089#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090090#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000091#include <linux/xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93#include <asm/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070097#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070098#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
Arnd Bergmann6b960182009-11-06 23:10:54 -0800103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800106#include <linux/sockios.h>
107#include <linux/atalk.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300108#include <net/busy_poll.h>
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400109#include <linux/errqueue.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300110
Cong Wange0d10952013-08-01 11:10:25 +0800111#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300112unsigned int sysctl_net_busy_read __read_mostly;
113unsigned int sysctl_net_busy_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300114#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800115
Al Viro8ae5e032014-11-28 19:40:50 -0500116static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
120static int sock_close(struct inode *inode, struct file *file);
121static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700123static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800124#ifdef CONFIG_COMPAT
125static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700126 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800127#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800131static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700132 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800133 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135/*
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
138 */
139
Arjan van de Venda7071d2007-02-12 00:55:36 -0800140static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 .owner = THIS_MODULE,
142 .llseek = no_llseek,
Al Viro8ae5e032014-11-28 19:40:50 -0500143 .read = new_sync_read,
144 .write = new_sync_write,
145 .read_iter = sock_read_iter,
146 .write_iter = sock_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 .poll = sock_poll,
148 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800149#ifdef CONFIG_COMPAT
150 .compat_ioctl = compat_sock_ioctl,
151#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 .mmap = sock_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 .release = sock_close,
154 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200155 .sendpage = sock_sendpage,
156 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800157 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158};
159
160/*
161 * The protocol list. Each protocol is registered in here.
162 */
163
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000165static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167/*
168 * Statistics counters of the socket lists
169 */
170
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700171static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700174 * Support routines.
175 * Move socket addresses back and forth across the kernel/user
176 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 */
178
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179/**
180 * move_addr_to_kernel - copy a socket address into kernel space
181 * @uaddr: Address in user space
182 * @kaddr: Address in kernel space
183 * @ulen: Length in user space
184 *
185 * The address is copied into kernel space. If the provided address is
186 * too long an error code of -EINVAL is returned. If the copy gives
187 * invalid addresses -EFAULT is returned. On a success 0 is returned.
188 */
189
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000190int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700192 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700194 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700196 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100198 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199}
200
201/**
202 * move_addr_to_user - copy an address to user space
203 * @kaddr: kernel space address
204 * @klen: length of address in kernel
205 * @uaddr: user space address
206 * @ulen: pointer to user length field
207 *
208 * The value pointed to by ulen on entry is the buffer length available.
209 * This is overwritten with the buffer space used. -EINVAL is returned
210 * if an overlong buffer is specified or a negative buffer size. -EFAULT
211 * is returned if either the buffer or the length field are not
212 * accessible.
213 * After copying the data up to the limit the user specifies, the true
214 * length of the data is written over the length limit the user
215 * specified. Zero is returned for a success.
216 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700217
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000218static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000219 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220{
221 int err;
222 int len;
223
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100224 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700225 err = get_user(len, ulen);
226 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 if (len > klen)
229 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100230 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700232 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500233 if (audit_sockaddr(klen, kaddr))
234 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700235 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 return -EFAULT;
237 }
238 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700239 * "fromlen shall refer to the value before truncation.."
240 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241 */
242 return __put_user(klen, ulen);
243}
244
Christoph Lametere18b8902006-12-06 20:33:20 -0800245static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
247static struct inode *sock_alloc_inode(struct super_block *sb)
248{
249 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000250 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700251
Christoph Lametere94b1762006-12-06 20:33:17 -0800252 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 if (!ei)
254 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000255 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
256 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000257 kmem_cache_free(sock_inode_cachep, ei);
258 return NULL;
259 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000260 init_waitqueue_head(&wq->wait);
261 wq->fasync_list = NULL;
262 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700263
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 ei->socket.state = SS_UNCONNECTED;
265 ei->socket.flags = 0;
266 ei->socket.ops = NULL;
267 ei->socket.sk = NULL;
268 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
270 return &ei->vfs_inode;
271}
272
273static void sock_destroy_inode(struct inode *inode)
274{
Eric Dumazet43815482010-04-29 11:01:49 +0000275 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000276 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000277
278 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000279 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800280 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000281 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282}
283
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700284static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700286 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287
Christoph Lametera35afb82007-05-16 22:10:57 -0700288 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700290
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291static int init_inodecache(void)
292{
293 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700294 sizeof(struct socket_alloc),
295 0,
296 (SLAB_HWCACHE_ALIGN |
297 SLAB_RECLAIM_ACCOUNT |
298 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900299 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300 if (sock_inode_cachep == NULL)
301 return -ENOMEM;
302 return 0;
303}
304
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700305static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700306 .alloc_inode = sock_alloc_inode,
307 .destroy_inode = sock_destroy_inode,
308 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309};
310
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700311/*
312 * sockfs_dname() is called from d_path().
313 */
314static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
315{
316 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
317 dentry->d_inode->i_ino);
318}
319
Al Viro3ba13d12009-02-20 06:02:22 +0000320static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700321 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322};
323
Al Viroc74a1cb2011-01-12 16:59:34 -0500324static struct dentry *sockfs_mount(struct file_system_type *fs_type,
325 int flags, const char *dev_name, void *data)
326{
327 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
328 &sockfs_dentry_operations, SOCKFS_MAGIC);
329}
330
331static struct vfsmount *sock_mnt __read_mostly;
332
333static struct file_system_type sock_fs_type = {
334 .name = "sockfs",
335 .mount = sockfs_mount,
336 .kill_sb = kill_anon_super,
337};
338
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339/*
340 * Obtains the first available file descriptor and sets it up for use.
341 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800342 * These functions create file structures and maps them to fd space
343 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 * and file struct implicitly stored in sock->file.
345 * Note that another thread may close file descriptor before we return
346 * from this function. We use the fact that now we do not refer
347 * to socket after mapping. If one day we will need it, this
348 * function will increment ref. count on file by 1.
349 *
350 * In any case returned fd MAY BE not valid!
351 * This race condition is unavoidable
352 * with shared fd spaces, we cannot solve it inside kernel,
353 * but we take care of internal coherence yet.
354 */
355
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700356struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357{
Al Viro7cbe66b2009-08-05 19:59:08 +0400358 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400359 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400360 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800361
Masatake YAMATO600e1772012-08-29 10:44:29 +0000362 if (dname) {
363 name.name = dname;
364 name.len = strlen(name.name);
365 } else if (sock->sk) {
366 name.name = sock->sk->sk_prot_creator->name;
367 name.len = strlen(name.name);
368 }
Nick Piggin4b936882011-01-07 17:50:07 +1100369 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400370 if (unlikely(!path.dentry))
371 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400372 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373
Al Viro2c48b9c2009-08-09 00:52:35 +0400374 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400375
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{
Al Viro00e188e2014-03-03 23:48:18 -0500452 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800453 struct socket *sock;
454
Hua Zhong36725582006-04-19 15:25:02 -0700455 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500456 if (f.file) {
457 sock = sock_from_file(f.file, err);
458 if (likely(sock)) {
459 *fput_needed = f.flags;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800460 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500461 }
462 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800463 }
464 return NULL;
465}
466
Masatake YAMATO600e1772012-08-29 10:44:29 +0000467#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
468#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
469#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
470static ssize_t sockfs_getxattr(struct dentry *dentry,
471 const char *name, void *value, size_t size)
472{
473 const char *proto_name;
474 size_t proto_size;
475 int error;
476
477 error = -ENODATA;
478 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
479 proto_name = dentry->d_name.name;
480 proto_size = strlen(proto_name);
481
482 if (value) {
483 error = -ERANGE;
484 if (proto_size + 1 > size)
485 goto out;
486
487 strncpy(value, proto_name, proto_size + 1);
488 }
489 error = proto_size + 1;
490 }
491
492out:
493 return error;
494}
495
496static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
497 size_t size)
498{
499 ssize_t len;
500 ssize_t used = 0;
501
502 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
503 if (len < 0)
504 return len;
505 used += len;
506 if (buffer) {
507 if (size < used)
508 return -ERANGE;
509 buffer += len;
510 }
511
512 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
513 used += len;
514 if (buffer) {
515 if (size < used)
516 return -ERANGE;
517 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
518 buffer += len;
519 }
520
521 return used;
522}
523
524static const struct inode_operations sockfs_inode_ops = {
525 .getxattr = sockfs_getxattr,
526 .listxattr = sockfs_listxattr,
527};
528
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529/**
530 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700531 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 * Allocate a new inode and socket object. The two are bound together
533 * and initialised. The socket is then returned. If we are out of inodes
534 * NULL is returned.
535 */
536
537static struct socket *sock_alloc(void)
538{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700539 struct inode *inode;
540 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200542 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 if (!inode)
544 return NULL;
545
546 sock = SOCKET_I(inode);
547
Eric Dumazet29a020d2009-09-15 02:39:20 -0700548 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400549 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700550 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100551 inode->i_uid = current_fsuid();
552 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000553 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554
Alex Shi19e8d692012-05-14 14:15:31 -0700555 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 return sock;
557}
558
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559/**
560 * sock_release - close a socket
561 * @sock: socket to close
562 *
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700565 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568void sock_release(struct socket *sock)
569{
570 if (sock->ops) {
571 struct module *owner = sock->ops->owner;
572
573 sock->ops->release(sock);
574 sock->ops = NULL;
575 module_put(owner);
576 }
577
Eric Dumazeteaefd112011-02-18 03:26:36 +0000578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800579 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
Mikulas Patockab09e7862012-07-19 06:13:36 +0000581 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
582 return;
583
Alex Shi19e8d692012-05-14 14:15:31 -0700584 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 if (!sock->file) {
586 iput(SOCK_INODE(sock));
587 return;
588 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700589 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700591EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400593void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000594{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200595 u8 flags = *tx_flags;
596
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200598 flags |= SKBTX_HW_TSTAMP;
599
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200601 flags |= SKBTX_SW_TSTAMP;
602
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200604 flags |= SKBTX_SCHED_TSTAMP;
605
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400606 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200607 flags |= SKBTX_ACK_TSTAMP;
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400608
Eric Dumazet140c55d2014-08-06 11:49:29 +0200609 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000610}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400611EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000612
Anton Blanchard228e5482011-05-02 20:21:35 +0000613static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
614 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 return sock->ops->sendmsg(iocb, sock, msg, size);
617}
618
Anton Blanchard228e5482011-05-02 20:21:35 +0000619static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
620 struct msghdr *msg, size_t size)
621{
622 int err = security_socket_sendmsg(sock, msg, size);
623
624 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
625}
626
Gu Zheng0cf00c62014-12-05 15:14:23 +0800627static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
628 size_t size, bool nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
630 struct kiocb iocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 int ret;
632
633 init_sync_kiocb(&iocb, NULL);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800634 ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
635 __sock_sendmsg(&iocb, sock, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 if (-EIOCBQUEUED == ret)
637 ret = wait_on_sync_kiocb(&iocb);
638 return ret;
639}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800640
641int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
642{
643 return do_sock_sendmsg(sock, msg, size, false);
644}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700645EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
Eric Dumazet894dc242011-07-26 02:39:41 +0000647static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000648{
Gu Zheng0cf00c62014-12-05 15:14:23 +0800649 return do_sock_sendmsg(sock, msg, size, true);
Anton Blanchard228e5482011-05-02 20:21:35 +0000650}
651
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
653 struct kvec *vec, size_t num, size_t size)
654{
655 mm_segment_t oldfs = get_fs();
656 int result;
657
658 set_fs(KERNEL_DS);
659 /*
660 * the following is safe, since for compiler definitions of kvec and
661 * iovec are identical, yielding the same in-core layout and alignment
662 */
Al Viroc0371da2014-11-24 10:42:55 -0500663 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 result = sock_sendmsg(sock, msg, size);
665 set_fs(oldfs);
666 return result;
667}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700668EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700670/*
671 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
672 */
673void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
674 struct sk_buff *skb)
675{
Patrick Ohly20d49472009-02-12 05:03:38 +0000676 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400677 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000678 int empty = 1;
679 struct skb_shared_hwtstamps *shhwtstamps =
680 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700681
Patrick Ohly20d49472009-02-12 05:03:38 +0000682 /* Race occurred between timestamp enabling and packet
683 receiving. Fill in the current time for now. */
684 if (need_software_tstamp && skb->tstamp.tv64 == 0)
685 __net_timestamp(skb);
686
687 if (need_software_tstamp) {
688 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
689 struct timeval tv;
690 skb_get_timestamp(skb, &tv);
691 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
692 sizeof(tv), &tv);
693 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400694 struct timespec ts;
695 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000696 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400697 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000698 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700699 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000700
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400701 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400702 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400703 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000704 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400705 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400706 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400707 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400708 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000709 if (!empty)
710 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400711 SCM_TIMESTAMPING, sizeof(tss), &tss);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700712}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300713EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
714
Johannes Berg6e3e9392011-11-09 10:15:42 +0100715void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
716 struct sk_buff *skb)
717{
718 int ack;
719
720 if (!sock_flag(sk, SOCK_WIFI_STATUS))
721 return;
722 if (!skb->wifi_acked_valid)
723 return;
724
725 ack = skb->wifi_acked;
726
727 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
728}
729EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
730
stephen hemminger11165f12010-10-18 14:27:29 +0000731static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
732 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700733{
734 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
735 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
736 sizeof(__u32), &skb->dropcount);
737}
738
Eric Dumazet767dd032010-04-28 19:14:43 +0000739void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700740 struct sk_buff *skb)
741{
742 sock_recv_timestamp(msg, sk, skb);
743 sock_recv_drops(msg, sk, skb);
744}
Eric Dumazet767dd032010-04-28 19:14:43 +0000745EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700746
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700747static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
748 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
751}
752
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700753static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
754 struct msghdr *msg, size_t size, int flags)
755{
756 int err = security_socket_recvmsg(sock, msg, size, flags);
757
758 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
759}
760
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700761int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 size_t size, int flags)
763{
764 struct kiocb iocb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 int ret;
766
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700767 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
769 if (-EIOCBQUEUED == ret)
770 ret = wait_on_sync_kiocb(&iocb);
771 return ret;
772}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700773EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700775static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
776 size_t size, int flags)
777{
778 struct kiocb iocb;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700779 int ret;
780
781 init_sync_kiocb(&iocb, NULL);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700782 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
783 if (-EIOCBQUEUED == ret)
784 ret = wait_on_sync_kiocb(&iocb);
785 return ret;
786}
787
Martin Lucinac1249c02010-12-10 00:04:05 +0000788/**
789 * kernel_recvmsg - Receive a message from a socket (kernel space)
790 * @sock: The socket to receive the message from
791 * @msg: Received message
792 * @vec: Input s/g array for message data
793 * @num: Size of input s/g array
794 * @size: Number of bytes to read
795 * @flags: Message flags (MSG_DONTWAIT, etc...)
796 *
797 * On return the msg structure contains the scatter/gather array passed in the
798 * vec argument. The array is modified so that it consists of the unfilled
799 * portion of the original array.
800 *
801 * The returned value is the total number of bytes received, or an error.
802 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700803int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
804 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
806 mm_segment_t oldfs = get_fs();
807 int result;
808
809 set_fs(KERNEL_DS);
810 /*
811 * the following is safe, since for compiler definitions of kvec and
812 * iovec are identical, yielding the same in-core layout and alignment
813 */
Al Viroc0371da2014-11-24 10:42:55 -0500814 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 result = sock_recvmsg(sock, msg, size, flags);
816 set_fs(oldfs);
817 return result;
818}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700819EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300821static ssize_t sock_sendpage(struct file *file, struct page *page,
822 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823{
824 struct socket *sock;
825 int flags;
826
Eric Dumazetb69aee02005-09-06 14:42:45 -0700827 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828
Eric Dumazet35f9c092012-04-05 03:05:35 +0000829 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
830 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
831 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Linus Torvaldse6949582009-08-13 08:28:36 -0700833 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834}
835
Jens Axboe9c55e012007-11-06 23:30:13 -0800836static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700837 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800838 unsigned int flags)
839{
840 struct socket *sock = file->private_data;
841
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800842 if (unlikely(!sock->ops->splice_read))
843 return -EINVAL;
844
Jens Axboe9c55e012007-11-06 23:30:13 -0800845 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
846}
847
Al Viro8ae5e032014-11-28 19:40:50 -0500848static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800849{
Al Viro6d652332015-01-30 16:12:56 -0500850 struct file *file = iocb->ki_filp;
851 struct socket *sock = file->private_data;
Al Viro8ae5e032014-11-28 19:40:50 -0500852 struct msghdr msg = {.msg_iter = *to};
853 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800854
Al Viro8ae5e032014-11-28 19:40:50 -0500855 if (file->f_flags & O_NONBLOCK)
856 msg.msg_flags = MSG_DONTWAIT;
857
858 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800859 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700860
Kent Overstreet73a70752013-05-09 15:03:42 -0700861 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800862 return 0;
863
Al Viro8ae5e032014-11-28 19:40:50 -0500864 res = __sock_recvmsg(iocb, sock, &msg,
865 iocb->ki_nbytes, msg.msg_flags);
866 *to = msg.msg_iter;
867 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800868}
869
Al Viro8ae5e032014-11-28 19:40:50 -0500870static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871{
Al Viro6d652332015-01-30 16:12:56 -0500872 struct file *file = iocb->ki_filp;
873 struct socket *sock = file->private_data;
Al Viro8ae5e032014-11-28 19:40:50 -0500874 struct msghdr msg = {.msg_iter = *from};
875 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800876
Al Viro8ae5e032014-11-28 19:40:50 -0500877 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800878 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700879
Al Viro8ae5e032014-11-28 19:40:50 -0500880 if (file->f_flags & O_NONBLOCK)
881 msg.msg_flags = MSG_DONTWAIT;
882
Al Viro6d652332015-01-30 16:12:56 -0500883 if (sock->type == SOCK_SEQPACKET)
884 msg.msg_flags |= MSG_EOR;
885
Al Viro8ae5e032014-11-28 19:40:50 -0500886 res = __sock_sendmsg(iocb, sock, &msg, iocb->ki_nbytes);
887 *from = msg.msg_iter;
888 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889}
890
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891/*
892 * Atomic setting of ioctl hooks to avoid race
893 * with module unload.
894 */
895
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800896static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700897static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
Eric W. Biederman881d9662007-09-17 11:56:21 -0700899void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800901 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800903 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904}
905EXPORT_SYMBOL(brioctl_set);
906
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800907static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700908static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909
Eric W. Biederman881d9662007-09-17 11:56:21 -0700910void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800912 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800914 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916EXPORT_SYMBOL(vlan_ioctl_set);
917
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800918static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700919static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700921void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800923 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800925 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926}
927EXPORT_SYMBOL(dlci_ioctl_set);
928
Arnd Bergmann6b960182009-11-06 23:10:54 -0800929static long sock_do_ioctl(struct net *net, struct socket *sock,
930 unsigned int cmd, unsigned long arg)
931{
932 int err;
933 void __user *argp = (void __user *)arg;
934
935 err = sock->ops->ioctl(sock, cmd, arg);
936
937 /*
938 * If this ioctl is unknown try to hand it down
939 * to the NIC driver.
940 */
941 if (err == -ENOIOCTLCMD)
942 err = dev_ioctl(net, cmd, argp);
943
944 return err;
945}
946
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947/*
948 * With an ioctl, arg may well be a user mode pointer, but we don't know
949 * what to do with it - that's up to the protocol still.
950 */
951
952static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
953{
954 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700955 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 void __user *argp = (void __user *)arg;
957 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700958 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959
Eric Dumazetb69aee02005-09-06 14:42:45 -0700960 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700961 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900962 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700964 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200966#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700968 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200970#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700971 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 case FIOSETOWN:
973 case SIOCSPGRP:
974 err = -EFAULT;
975 if (get_user(pid, (int __user *)argp))
976 break;
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400977 f_setown(sock->file, pid, 1);
978 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 break;
980 case FIOGETOWN:
981 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700982 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700983 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 break;
985 case SIOCGIFBR:
986 case SIOCSIFBR:
987 case SIOCBRADDBR:
988 case SIOCBRDELBR:
989 err = -ENOPKG;
990 if (!br_ioctl_hook)
991 request_module("bridge");
992
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800993 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700994 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700995 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800996 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 break;
998 case SIOCGIFVLAN:
999 case SIOCSIFVLAN:
1000 err = -ENOPKG;
1001 if (!vlan_ioctl_hook)
1002 request_module("8021q");
1003
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001004 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001006 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001007 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 case SIOCADDDLCI:
1010 case SIOCDELDLCI:
1011 err = -ENOPKG;
1012 if (!dlci_ioctl_hook)
1013 request_module("dlci");
1014
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001015 mutex_lock(&dlci_ioctl_mutex);
1016 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001018 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 break;
1020 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001021 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001023 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 return err;
1025}
1026
1027int sock_create_lite(int family, int type, int protocol, struct socket **res)
1028{
1029 int err;
1030 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001031
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 err = security_socket_create(family, type, protocol, 1);
1033 if (err)
1034 goto out;
1035
1036 sock = sock_alloc();
1037 if (!sock) {
1038 err = -ENOMEM;
1039 goto out;
1040 }
1041
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001043 err = security_socket_post_create(sock, family, type, protocol, 1);
1044 if (err)
1045 goto out_release;
1046
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047out:
1048 *res = sock;
1049 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001050out_release:
1051 sock_release(sock);
1052 sock = NULL;
1053 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001055EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
1057/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001058static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059{
Eliezer Tamircbf55002013-07-08 16:20:34 +03001060 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 struct socket *sock;
1062
1063 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001064 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001066 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001067
Eliezer Tamircbf55002013-07-08 16:20:34 +03001068 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001069 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001070 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001071
1072 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001073 if (wait && (wait->_key & POLL_BUSY_LOOP))
1074 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001075 }
1076
Eliezer Tamircbf55002013-07-08 16:20:34 +03001077 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078}
1079
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001080static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001082 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
1084 return sock->ops->mmap(file, sock, vma);
1085}
1086
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001087static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 sock_release(SOCKET_I(inode));
1090 return 0;
1091}
1092
1093/*
1094 * Update the socket async list
1095 *
1096 * Fasync_list locking strategy.
1097 *
1098 * 1. fasync_list is modified only under process context socket lock
1099 * i.e. under semaphore.
1100 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001101 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 */
1103
1104static int sock_fasync(int fd, struct file *filp, int on)
1105{
Eric Dumazet989a2972010-04-14 09:55:35 +00001106 struct socket *sock = filp->private_data;
1107 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001108 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109
Eric Dumazet989a2972010-04-14 09:55:35 +00001110 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
1113 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001114 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1115 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
Eric Dumazeteaefd112011-02-18 03:26:36 +00001117 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001118 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001119 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001120 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121
Eric Dumazet989a2972010-04-14 09:55:35 +00001122 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 return 0;
1124}
1125
Eric Dumazet43815482010-04-29 11:01:49 +00001126/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127
1128int sock_wake_async(struct socket *sock, int how, int band)
1129{
Eric Dumazet43815482010-04-29 11:01:49 +00001130 struct socket_wq *wq;
1131
1132 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001134 rcu_read_lock();
1135 wq = rcu_dereference(sock->wq);
1136 if (!wq || !wq->fasync_list) {
1137 rcu_read_unlock();
1138 return -1;
1139 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001140 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001141 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1143 break;
1144 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001145 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1147 break;
1148 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001149 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001150call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001151 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001153 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001154 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 }
Eric Dumazet43815482010-04-29 11:01:49 +00001156 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 return 0;
1158}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001159EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160
Pavel Emelyanov721db932010-09-29 16:06:32 +04001161int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001162 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163{
1164 int err;
1165 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001166 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
1168 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001169 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 */
1171 if (family < 0 || family >= NPROTO)
1172 return -EAFNOSUPPORT;
1173 if (type < 0 || type >= SOCK_MAX)
1174 return -EINVAL;
1175
1176 /* Compatibility.
1177
1178 This uglymoron is moved from INET layer to here to avoid
1179 deadlock in module load.
1180 */
1181 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001182 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 if (!warned) {
1184 warned = 1;
Yang Yingliang3410f222014-02-12 17:09:55 +08001185 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1186 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 }
1188 family = PF_PACKET;
1189 }
1190
1191 err = security_socket_create(family, type, protocol, kern);
1192 if (err)
1193 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001194
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001195 /*
1196 * Allocate the socket and allow the family to set things up. if
1197 * the protocol is 0, the family is instructed to select an appropriate
1198 * default.
1199 */
1200 sock = sock_alloc();
1201 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001202 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001203 return -ENFILE; /* Not exactly a match, but its the
1204 closest posix thing */
1205 }
1206
1207 sock->type = type;
1208
Johannes Berg95a5afc2008-10-16 15:24:51 -07001209#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001210 /* Attempt to load a protocol module if the find failed.
1211 *
1212 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 * requested real, full-featured networking support upon configuration.
1214 * Otherwise module support will break!
1215 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001216 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001217 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218#endif
1219
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001220 rcu_read_lock();
1221 pf = rcu_dereference(net_families[family]);
1222 err = -EAFNOSUPPORT;
1223 if (!pf)
1224 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225
1226 /*
1227 * We will call the ->create function, that possibly is in a loadable
1228 * module, so we have to bump that loadable module refcnt first.
1229 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001230 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 goto out_release;
1232
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001233 /* Now protected by module ref count */
1234 rcu_read_unlock();
1235
Eric Paris3f378b62009-11-05 22:18:14 -08001236 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001237 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001239
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 /*
1241 * Now to bump the refcnt of the [loadable] module that owns this
1242 * socket at sock_release time we decrement its refcnt.
1243 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001244 if (!try_module_get(sock->ops->owner))
1245 goto out_module_busy;
1246
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247 /*
1248 * Now that we're done with the ->create function, the [loadable]
1249 * module can have its refcnt decremented
1250 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001251 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001252 err = security_socket_post_create(sock, family, type, protocol, kern);
1253 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001254 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001255 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001257 return 0;
1258
1259out_module_busy:
1260 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001262 sock->ops = NULL;
1263 module_put(pf->owner);
1264out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001266 return err;
1267
1268out_release:
1269 rcu_read_unlock();
1270 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001272EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273
1274int sock_create(int family, int type, int protocol, struct socket **res)
1275{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001276 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001278EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279
1280int sock_create_kern(int family, int type, int protocol, struct socket **res)
1281{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001282 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001284EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001286SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287{
1288 int retval;
1289 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001290 int flags;
1291
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001292 /* Check the SOCK_* constants for consistency. */
1293 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1294 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1295 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1296 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1297
Ulrich Dreppera677a032008-07-23 21:29:17 -07001298 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001299 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001300 return -EINVAL;
1301 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001303 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1304 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1305
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 retval = sock_create(family, type, protocol, &sock);
1307 if (retval < 0)
1308 goto out;
1309
Ulrich Drepper77d27202008-07-23 21:29:35 -07001310 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 if (retval < 0)
1312 goto out_release;
1313
1314out:
1315 /* It may be already another descriptor 8) Not kernel problem. */
1316 return retval;
1317
1318out_release:
1319 sock_release(sock);
1320 return retval;
1321}
1322
1323/*
1324 * Create a pair of connected sockets.
1325 */
1326
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001327SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1328 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329{
1330 struct socket *sock1, *sock2;
1331 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001332 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001333 int flags;
1334
1335 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001336 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001337 return -EINVAL;
1338 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001340 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1341 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1342
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 /*
1344 * Obtain the first socket and check if the underlying protocol
1345 * supports the socketpair call.
1346 */
1347
1348 err = sock_create(family, type, protocol, &sock1);
1349 if (err < 0)
1350 goto out;
1351
1352 err = sock_create(family, type, protocol, &sock2);
1353 if (err < 0)
1354 goto out_release_1;
1355
1356 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001357 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 goto out_release_both;
1359
Al Viro28407632012-08-17 23:54:15 -04001360 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001361 if (unlikely(fd1 < 0)) {
1362 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001364 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001365
Al Viro28407632012-08-17 23:54:15 -04001366 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001367 if (unlikely(fd2 < 0)) {
1368 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001369 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001370 }
1371
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001372 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001373 if (unlikely(IS_ERR(newfile1))) {
1374 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001375 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001376 }
1377
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001378 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001379 if (IS_ERR(newfile2)) {
1380 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001381 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001382 }
1383
Yann Droneaudd73aa282013-12-09 22:42:20 +01001384 err = put_user(fd1, &usockvec[0]);
1385 if (err)
1386 goto out_fput_both;
1387
1388 err = put_user(fd2, &usockvec[1]);
1389 if (err)
1390 goto out_fput_both;
1391
Al Viro157cf642008-12-14 04:57:47 -05001392 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001393
Al Virodb349502007-02-07 01:48:00 -05001394 fd_install(fd1, newfile1);
1395 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396 /* fd1 and fd2 may be already another descriptors.
1397 * Not kernel problem.
1398 */
1399
Yann Droneaudd73aa282013-12-09 22:42:20 +01001400 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401
Yann Droneaudd73aa282013-12-09 22:42:20 +01001402out_fput_both:
1403 fput(newfile2);
1404 fput(newfile1);
1405 put_unused_fd(fd2);
1406 put_unused_fd(fd1);
1407 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
Yann Droneaudd73aa282013-12-09 22:42:20 +01001409out_fput_1:
1410 fput(newfile1);
1411 put_unused_fd(fd2);
1412 put_unused_fd(fd1);
1413 sock_release(sock2);
1414 goto out;
1415
1416out_put_unused_both:
1417 put_unused_fd(fd2);
1418out_put_unused_1:
1419 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001421 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001423 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424out:
1425 return err;
1426}
1427
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428/*
1429 * Bind a name to a socket. Nothing much to do here since it's
1430 * the protocol's responsibility to handle the local address.
1431 *
1432 * We move the socket address to kernel space before we call
1433 * the protocol layer (having also checked the address is ok).
1434 */
1435
Heiko Carstens20f37032009-01-14 14:14:23 +01001436SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437{
1438 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001439 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001440 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001442 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001443 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001444 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001445 if (err >= 0) {
1446 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001447 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001448 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001449 if (!err)
1450 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001451 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001452 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001454 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001455 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 return err;
1457}
1458
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459/*
1460 * Perform a listen. Basically, we allow the protocol to do anything
1461 * necessary for a listen, and if that works, we mark the socket as
1462 * ready for listening.
1463 */
1464
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001465SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466{
1467 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001468 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001469 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001470
1471 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1472 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001473 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001474 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001475 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476
1477 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001478 if (!err)
1479 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001481 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 }
1483 return err;
1484}
1485
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486/*
1487 * For accept, we attempt to create a new socket, set up the link
1488 * with the client, wake up the client, then return the new
1489 * connected fd. We collect the address of the connector in kernel
1490 * space and move it to user at the very end. This is unclean because
1491 * we open the socket then return an error.
1492 *
1493 * 1003.1g adds the ability to recvmsg() to query connection pending
1494 * status to recvmsg. We need to add that support in a way thats
1495 * clean when we restucture accept also.
1496 */
1497
Heiko Carstens20f37032009-01-14 14:14:23 +01001498SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1499 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500{
1501 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001502 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001503 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001504 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505
Ulrich Drepper77d27202008-07-23 21:29:35 -07001506 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001507 return -EINVAL;
1508
1509 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1510 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1511
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001512 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 if (!sock)
1514 goto out;
1515
1516 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001517 newsock = sock_alloc();
1518 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 goto out_put;
1520
1521 newsock->type = sock->type;
1522 newsock->ops = sock->ops;
1523
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 /*
1525 * We don't need try_module_get here, as the listening socket (sock)
1526 * has the protocol module (sock->ops->owner) held.
1527 */
1528 __module_get(newsock->ops->owner);
1529
Al Viro28407632012-08-17 23:54:15 -04001530 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001531 if (unlikely(newfd < 0)) {
1532 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001533 sock_release(newsock);
1534 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001535 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001536 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001537 if (unlikely(IS_ERR(newfile))) {
1538 err = PTR_ERR(newfile);
1539 put_unused_fd(newfd);
1540 sock_release(newsock);
1541 goto out_put;
1542 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001543
Frank Filza79af592005-09-27 15:23:38 -07001544 err = security_socket_accept(sock, newsock);
1545 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001546 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001547
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1549 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001550 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551
1552 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001553 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001554 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001556 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001558 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001559 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001561 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 }
1563
1564 /* File flags are not inherited via accept() unlike another OSes. */
1565
David S. Miller39d8c1b2006-03-20 17:13:49 -08001566 fd_install(newfd, newfile);
1567 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001570 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571out:
1572 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001573out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001574 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001575 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 goto out_put;
1577}
1578
Heiko Carstens20f37032009-01-14 14:14:23 +01001579SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1580 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001581{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001582 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001583}
1584
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585/*
1586 * Attempt to connect to a socket with the server address. The address
1587 * is in user space so we verify it is OK and move it to kernel space.
1588 *
1589 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1590 * break bindings
1591 *
1592 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1593 * other SEQPACKET protocols that take time to connect() as it doesn't
1594 * include the -EINPROGRESS status for such sockets.
1595 */
1596
Heiko Carstens20f37032009-01-14 14:14:23 +01001597SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1598 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599{
1600 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001601 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001602 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001604 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 if (!sock)
1606 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001607 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 if (err < 0)
1609 goto out_put;
1610
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001611 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001612 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 if (err)
1614 goto out_put;
1615
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001616 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617 sock->file->f_flags);
1618out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001619 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620out:
1621 return err;
1622}
1623
1624/*
1625 * Get the local address ('name') of a socket object. Move the obtained
1626 * name to user space.
1627 */
1628
Heiko Carstens20f37032009-01-14 14:14:23 +01001629SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1630 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631{
1632 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001633 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001634 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001635
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001636 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637 if (!sock)
1638 goto out;
1639
1640 err = security_socket_getsockname(sock);
1641 if (err)
1642 goto out_put;
1643
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001644 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 if (err)
1646 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001647 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648
1649out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001650 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651out:
1652 return err;
1653}
1654
1655/*
1656 * Get the remote address ('name') of a socket object. Move the obtained
1657 * name to user space.
1658 */
1659
Heiko Carstens20f37032009-01-14 14:14:23 +01001660SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1661 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662{
1663 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001664 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001665 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001667 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1668 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 err = security_socket_getpeername(sock);
1670 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001671 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 return err;
1673 }
1674
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001675 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001676 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001677 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001679 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001680 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001681 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682 }
1683 return err;
1684}
1685
1686/*
1687 * Send a datagram to a given address. We move the address into kernel
1688 * space and check the user space data area is readable before invoking
1689 * the protocol.
1690 */
1691
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001692SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001693 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001694 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695{
1696 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001697 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 int err;
1699 struct msghdr msg;
1700 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001701 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001702
Linus Torvalds253eacc2010-10-30 16:43:10 -07001703 if (len > INT_MAX)
1704 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001705 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1706 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001707 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001708
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001709 iov.iov_base = buff;
1710 iov.iov_len = len;
1711 msg.msg_name = NULL;
Al Viroc0371da2014-11-24 10:42:55 -05001712 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001713 msg.msg_control = NULL;
1714 msg.msg_controllen = 0;
1715 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001716 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001717 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718 if (err < 0)
1719 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001720 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001721 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 }
1723 if (sock->file->f_flags & O_NONBLOCK)
1724 flags |= MSG_DONTWAIT;
1725 msg.msg_flags = flags;
1726 err = sock_sendmsg(sock, &msg, len);
1727
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001728out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001729 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001730out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 return err;
1732}
1733
1734/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001735 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 */
1737
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001738SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001739 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740{
1741 return sys_sendto(fd, buff, len, flags, NULL, 0);
1742}
1743
1744/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001745 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 * sender. We verify the buffers are writable and if needed move the
1747 * sender address from kernel to user space.
1748 */
1749
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001750SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001751 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001752 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753{
1754 struct socket *sock;
1755 struct iovec iov;
1756 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001757 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001758 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001759 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760
Linus Torvalds253eacc2010-10-30 16:43:10 -07001761 if (size > INT_MAX)
1762 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001763 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001765 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001767 msg.msg_control = NULL;
1768 msg.msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001769 iov.iov_len = size;
1770 iov.iov_base = ubuf;
Al Viroc0371da2014-11-24 10:42:55 -05001771 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001772 /* Save some cycles and don't copy the address if not needed */
1773 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1774 /* We assume all kernel code knows the size of sockaddr_storage */
1775 msg.msg_namelen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776 if (sock->file->f_flags & O_NONBLOCK)
1777 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001778 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001780 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001781 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001782 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001783 if (err2 < 0)
1784 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001786
1787 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001788out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789 return err;
1790}
1791
1792/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001793 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 */
1795
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001796SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1797 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798{
1799 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1800}
1801
1802/*
1803 * Set a socket option. Because we don't know the option lengths we have
1804 * to pass the user mode parameter for the protocols to sort out.
1805 */
1806
Heiko Carstens20f37032009-01-14 14:14:23 +01001807SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1808 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001810 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 struct socket *sock;
1812
1813 if (optlen < 0)
1814 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001815
1816 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1817 if (sock != NULL) {
1818 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001819 if (err)
1820 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821
1822 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001823 err =
1824 sock_setsockopt(sock, level, optname, optval,
1825 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001827 err =
1828 sock->ops->setsockopt(sock, level, optname, optval,
1829 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001830out_put:
1831 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832 }
1833 return err;
1834}
1835
1836/*
1837 * Get a socket option. Because we don't know the option lengths we have
1838 * to pass a user mode parameter for the protocols to sort out.
1839 */
1840
Heiko Carstens20f37032009-01-14 14:14:23 +01001841SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1842 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001844 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 struct socket *sock;
1846
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001847 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1848 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001849 err = security_socket_getsockopt(sock, level, optname);
1850 if (err)
1851 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
1853 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001854 err =
1855 sock_getsockopt(sock, level, optname, optval,
1856 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001858 err =
1859 sock->ops->getsockopt(sock, level, optname, optval,
1860 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001861out_put:
1862 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 }
1864 return err;
1865}
1866
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867/*
1868 * Shutdown a socket.
1869 */
1870
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001871SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001873 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 struct socket *sock;
1875
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001876 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1877 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001879 if (!err)
1880 err = sock->ops->shutdown(sock, how);
1881 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 }
1883 return err;
1884}
1885
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001886/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 * fields which are the same type (int / unsigned) on our platforms.
1888 */
1889#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1890#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1891#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1892
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001893struct used_address {
1894 struct sockaddr_storage name;
1895 unsigned int name_len;
1896};
1897
Al Viro08adb7d2014-11-10 20:23:13 -05001898static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1899 struct user_msghdr __user *umsg,
1900 struct sockaddr __user **save_addr,
1901 struct iovec **iov)
Dan Carpenter1661bf32013-10-03 00:27:20 +03001902{
Al Viro08adb7d2014-11-10 20:23:13 -05001903 struct sockaddr __user *uaddr;
1904 struct iovec __user *uiov;
Al Viroc0371da2014-11-24 10:42:55 -05001905 size_t nr_segs;
Al Viro08adb7d2014-11-10 20:23:13 -05001906 ssize_t err;
1907
1908 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1909 __get_user(uaddr, &umsg->msg_name) ||
1910 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1911 __get_user(uiov, &umsg->msg_iov) ||
Al Viroc0371da2014-11-24 10:42:55 -05001912 __get_user(nr_segs, &umsg->msg_iovlen) ||
Al Viro08adb7d2014-11-10 20:23:13 -05001913 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1914 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1915 __get_user(kmsg->msg_flags, &umsg->msg_flags))
Dan Carpenter1661bf32013-10-03 00:27:20 +03001916 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001917
Al Viro08adb7d2014-11-10 20:23:13 -05001918 if (!uaddr)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07001919 kmsg->msg_namelen = 0;
1920
Matthew Leachdbb490b2014-03-11 11:58:27 +00001921 if (kmsg->msg_namelen < 0)
1922 return -EINVAL;
1923
Dan Carpenter1661bf32013-10-03 00:27:20 +03001924 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001925 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05001926
1927 if (save_addr)
1928 *save_addr = uaddr;
1929
1930 if (uaddr && kmsg->msg_namelen) {
1931 if (!save_addr) {
1932 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1933 kmsg->msg_name);
1934 if (err < 0)
1935 return err;
1936 }
1937 } else {
1938 kmsg->msg_name = NULL;
1939 kmsg->msg_namelen = 0;
1940 }
1941
Al Viroc0371da2014-11-24 10:42:55 -05001942 if (nr_segs > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05001943 return -EMSGSIZE;
1944
1945 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
Al Viroc0371da2014-11-24 10:42:55 -05001946 uiov, nr_segs,
Al Viro08adb7d2014-11-10 20:23:13 -05001947 UIO_FASTIOV, *iov, iov);
1948 if (err >= 0)
Al Viroc0371da2014-11-24 10:42:55 -05001949 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
1950 *iov, nr_segs, err);
Al Viro08adb7d2014-11-10 20:23:13 -05001951 return err;
Dan Carpenter1661bf32013-10-03 00:27:20 +03001952}
1953
Al Viro666547f2014-04-06 14:03:05 -04001954static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001955 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001956 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001958 struct compat_msghdr __user *msg_compat =
1959 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001960 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001962 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001963 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1964 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965 unsigned char *ctl_buf = ctl;
Al Viro08adb7d2014-11-10 20:23:13 -05001966 int ctl_len, total_len;
1967 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001968
Al Viro08adb7d2014-11-10 20:23:13 -05001969 msg_sys->msg_name = &address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970
Al Viro08449322014-11-09 22:33:45 -05001971 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05001972 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
Al Viro08449322014-11-09 22:33:45 -05001973 else
Al Viro08adb7d2014-11-10 20:23:13 -05001974 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001975 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 goto out_freeiov;
1977 total_len = err;
1978
1979 err = -ENOBUFS;
1980
Anton Blanchard228e5482011-05-02 20:21:35 +00001981 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001983 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001985 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001986 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001987 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 if (err)
1989 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001990 ctl_buf = msg_sys->msg_control;
1991 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001993 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001995 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 goto out_freeiov;
1997 }
1998 err = -EFAULT;
1999 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002000 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2002 * checking falls down on this.
2003 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002004 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002005 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002006 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002008 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002010 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011
2012 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002013 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002014 /*
2015 * If this is sendmmsg() and current destination address is same as
2016 * previously succeeded address, omit asking LSM's decision.
2017 * used_address->name_len is initialized to UINT_MAX so that the first
2018 * destination address never matches.
2019 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002020 if (used_address && msg_sys->msg_name &&
2021 used_address->name_len == msg_sys->msg_namelen &&
2022 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002023 used_address->name_len)) {
2024 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2025 goto out_freectl;
2026 }
2027 err = sock_sendmsg(sock, msg_sys, total_len);
2028 /*
2029 * If this is sendmmsg() and sending to current destination address was
2030 * successful, remember it.
2031 */
2032 if (used_address && err >= 0) {
2033 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002034 if (msg_sys->msg_name)
2035 memcpy(&used_address->name, msg_sys->msg_name,
2036 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002037 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038
2039out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002040 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2042out_freeiov:
2043 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002044 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002045 return err;
2046}
2047
2048/*
2049 * BSD sendmsg interface
2050 */
2051
Al Viro666547f2014-04-06 14:03:05 -04002052long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002053{
2054 int fput_needed, err;
2055 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002056 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002057
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002058 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002059 if (!sock)
2060 goto out;
2061
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002062 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002063
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002064 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002065out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 return err;
2067}
2068
Al Viro666547f2014-04-06 14:03:05 -04002069SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002070{
2071 if (flags & MSG_CMSG_COMPAT)
2072 return -EINVAL;
2073 return __sys_sendmsg(fd, msg, flags);
2074}
2075
Anton Blanchard228e5482011-05-02 20:21:35 +00002076/*
2077 * Linux sendmmsg interface
2078 */
2079
2080int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2081 unsigned int flags)
2082{
2083 int fput_needed, err, datagrams;
2084 struct socket *sock;
2085 struct mmsghdr __user *entry;
2086 struct compat_mmsghdr __user *compat_entry;
2087 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002088 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002089
Anton Blanchard98382f42011-08-04 14:07:39 +00002090 if (vlen > UIO_MAXIOV)
2091 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002092
2093 datagrams = 0;
2094
2095 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2096 if (!sock)
2097 return err;
2098
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002099 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002100 entry = mmsg;
2101 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002102 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002103
2104 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002105 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002106 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002107 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002108 if (err < 0)
2109 break;
2110 err = __put_user(err, &compat_entry->msg_len);
2111 ++compat_entry;
2112 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002113 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002114 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002115 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002116 if (err < 0)
2117 break;
2118 err = put_user(err, &entry->msg_len);
2119 ++entry;
2120 }
2121
2122 if (err)
2123 break;
2124 ++datagrams;
2125 }
2126
Anton Blanchard228e5482011-05-02 20:21:35 +00002127 fput_light(sock->file, fput_needed);
2128
Anton Blanchard728ffb82011-08-04 14:07:38 +00002129 /* We only return an error if no datagrams were able to be sent */
2130 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002131 return datagrams;
2132
Anton Blanchard228e5482011-05-02 20:21:35 +00002133 return err;
2134}
2135
2136SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2137 unsigned int, vlen, unsigned int, flags)
2138{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002139 if (flags & MSG_CMSG_COMPAT)
2140 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002141 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2142}
2143
Al Viro666547f2014-04-06 14:03:05 -04002144static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002145 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002147 struct compat_msghdr __user *msg_compat =
2148 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002150 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 unsigned long cmsg_ptr;
Al Viro08adb7d2014-11-10 20:23:13 -05002152 int total_len, len;
2153 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154
2155 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002156 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157
2158 /* user mode address pointers */
2159 struct sockaddr __user *uaddr;
Al Viro08adb7d2014-11-10 20:23:13 -05002160 int __user *uaddr_len = COMPAT_NAMELEN(msg);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002161
Al Viro08adb7d2014-11-10 20:23:13 -05002162 msg_sys->msg_name = &addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002164 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002165 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002166 else
Al Viro08adb7d2014-11-10 20:23:13 -05002167 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168 if (err < 0)
2169 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002170 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002172 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2173 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002174
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002175 /* We assume all kernel code knows the size of sockaddr_storage */
2176 msg_sys->msg_namelen = 0;
2177
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178 if (sock->file->f_flags & O_NONBLOCK)
2179 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002180 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2181 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182 if (err < 0)
2183 goto out_freeiov;
2184 len = err;
2185
2186 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002187 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002188 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002189 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 if (err < 0)
2191 goto out_freeiov;
2192 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002193 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002194 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 if (err)
2196 goto out_freeiov;
2197 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002198 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199 &msg_compat->msg_controllen);
2200 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002201 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 &msg->msg_controllen);
2203 if (err)
2204 goto out_freeiov;
2205 err = len;
2206
2207out_freeiov:
2208 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002209 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002210 return err;
2211}
2212
2213/*
2214 * BSD recvmsg interface
2215 */
2216
Al Viro666547f2014-04-06 14:03:05 -04002217long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002218{
2219 int fput_needed, err;
2220 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002221 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002222
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002223 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002224 if (!sock)
2225 goto out;
2226
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002227 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002228
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002229 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230out:
2231 return err;
2232}
2233
Al Viro666547f2014-04-06 14:03:05 -04002234SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002235 unsigned int, flags)
2236{
2237 if (flags & MSG_CMSG_COMPAT)
2238 return -EINVAL;
2239 return __sys_recvmsg(fd, msg, flags);
2240}
2241
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002242/*
2243 * Linux recvmmsg interface
2244 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002246int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2247 unsigned int flags, struct timespec *timeout)
2248{
2249 int fput_needed, err, datagrams;
2250 struct socket *sock;
2251 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002252 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002253 struct msghdr msg_sys;
2254 struct timespec end_time;
2255
2256 if (timeout &&
2257 poll_select_set_timeout(&end_time, timeout->tv_sec,
2258 timeout->tv_nsec))
2259 return -EINVAL;
2260
2261 datagrams = 0;
2262
2263 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2264 if (!sock)
2265 return err;
2266
2267 err = sock_error(sock->sk);
2268 if (err)
2269 goto out_put;
2270
2271 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002272 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002273
2274 while (datagrams < vlen) {
2275 /*
2276 * No need to ask LSM for more than the first datagram.
2277 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002278 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002279 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002280 &msg_sys, flags & ~MSG_WAITFORONE,
2281 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002282 if (err < 0)
2283 break;
2284 err = __put_user(err, &compat_entry->msg_len);
2285 ++compat_entry;
2286 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002287 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002288 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002289 &msg_sys, flags & ~MSG_WAITFORONE,
2290 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002291 if (err < 0)
2292 break;
2293 err = put_user(err, &entry->msg_len);
2294 ++entry;
2295 }
2296
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002297 if (err)
2298 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002299 ++datagrams;
2300
Brandon L Black71c5c152010-03-26 16:18:03 +00002301 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2302 if (flags & MSG_WAITFORONE)
2303 flags |= MSG_DONTWAIT;
2304
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002305 if (timeout) {
2306 ktime_get_ts(timeout);
2307 *timeout = timespec_sub(end_time, *timeout);
2308 if (timeout->tv_sec < 0) {
2309 timeout->tv_sec = timeout->tv_nsec = 0;
2310 break;
2311 }
2312
2313 /* Timeout, return less than vlen datagrams */
2314 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2315 break;
2316 }
2317
2318 /* Out of band data, return right away */
2319 if (msg_sys.msg_flags & MSG_OOB)
2320 break;
2321 }
2322
2323out_put:
2324 fput_light(sock->file, fput_needed);
2325
2326 if (err == 0)
2327 return datagrams;
2328
2329 if (datagrams != 0) {
2330 /*
2331 * We may return less entries than requested (vlen) if the
2332 * sock is non block and there aren't enough datagrams...
2333 */
2334 if (err != -EAGAIN) {
2335 /*
2336 * ... or if recvmsg returns an error after we
2337 * received some datagrams, where we record the
2338 * error to return on the next call or if the
2339 * app asks about it using getsockopt(SO_ERROR).
2340 */
2341 sock->sk->sk_err = -err;
2342 }
2343
2344 return datagrams;
2345 }
2346
2347 return err;
2348}
2349
2350SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2351 unsigned int, vlen, unsigned int, flags,
2352 struct timespec __user *, timeout)
2353{
2354 int datagrams;
2355 struct timespec timeout_sys;
2356
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002357 if (flags & MSG_CMSG_COMPAT)
2358 return -EINVAL;
2359
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002360 if (!timeout)
2361 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2362
2363 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2364 return -EFAULT;
2365
2366 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2367
2368 if (datagrams > 0 &&
2369 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2370 datagrams = -EFAULT;
2371
2372 return datagrams;
2373}
2374
2375#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376/* Argument list sizes for sys_socketcall */
2377#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002378static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002379 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2380 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2381 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002382 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002383};
2384
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385#undef AL
2386
2387/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002388 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 *
2390 * Argument checking cleaned up. Saved 20% in size.
2391 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002392 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393 */
2394
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002395SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396{
Chen Gang2950fa92013-04-07 16:55:23 +08002397 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002398 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002400 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002401
Anton Blanchard228e5482011-05-02 20:21:35 +00002402 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403 return -EINVAL;
2404
Arjan van de Ven47379052009-09-28 12:57:44 -07002405 len = nargs[call];
2406 if (len > sizeof(a))
2407 return -EINVAL;
2408
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002410 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002412
Chen Gang2950fa92013-04-07 16:55:23 +08002413 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2414 if (err)
2415 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002416
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002417 a0 = a[0];
2418 a1 = a[1];
2419
2420 switch (call) {
2421 case SYS_SOCKET:
2422 err = sys_socket(a0, a1, a[2]);
2423 break;
2424 case SYS_BIND:
2425 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2426 break;
2427 case SYS_CONNECT:
2428 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2429 break;
2430 case SYS_LISTEN:
2431 err = sys_listen(a0, a1);
2432 break;
2433 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002434 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2435 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002436 break;
2437 case SYS_GETSOCKNAME:
2438 err =
2439 sys_getsockname(a0, (struct sockaddr __user *)a1,
2440 (int __user *)a[2]);
2441 break;
2442 case SYS_GETPEERNAME:
2443 err =
2444 sys_getpeername(a0, (struct sockaddr __user *)a1,
2445 (int __user *)a[2]);
2446 break;
2447 case SYS_SOCKETPAIR:
2448 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2449 break;
2450 case SYS_SEND:
2451 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2452 break;
2453 case SYS_SENDTO:
2454 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2455 (struct sockaddr __user *)a[4], a[5]);
2456 break;
2457 case SYS_RECV:
2458 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2459 break;
2460 case SYS_RECVFROM:
2461 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2462 (struct sockaddr __user *)a[4],
2463 (int __user *)a[5]);
2464 break;
2465 case SYS_SHUTDOWN:
2466 err = sys_shutdown(a0, a1);
2467 break;
2468 case SYS_SETSOCKOPT:
2469 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2470 break;
2471 case SYS_GETSOCKOPT:
2472 err =
2473 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2474 (int __user *)a[4]);
2475 break;
2476 case SYS_SENDMSG:
Al Viro666547f2014-04-06 14:03:05 -04002477 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002478 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002479 case SYS_SENDMMSG:
2480 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2481 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002482 case SYS_RECVMSG:
Al Viro666547f2014-04-06 14:03:05 -04002483 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002484 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002485 case SYS_RECVMMSG:
2486 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2487 (struct timespec __user *)a[4]);
2488 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002489 case SYS_ACCEPT4:
2490 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2491 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002492 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002493 default:
2494 err = -EINVAL;
2495 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 }
2497 return err;
2498}
2499
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002500#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002501
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002502/**
2503 * sock_register - add a socket protocol handler
2504 * @ops: description of protocol
2505 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002506 * This function is called by a protocol handler that wants to
2507 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002508 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002509 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002511int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002512{
2513 int err;
2514
2515 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002516 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002517 return -ENOBUFS;
2518 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002519
2520 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002521 if (rcu_dereference_protected(net_families[ops->family],
2522 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002523 err = -EEXIST;
2524 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002525 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002526 err = 0;
2527 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002528 spin_unlock(&net_family_lock);
2529
Yang Yingliang3410f222014-02-12 17:09:55 +08002530 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002531 return err;
2532}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002533EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002535/**
2536 * sock_unregister - remove a protocol handler
2537 * @family: protocol family to remove
2538 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002539 * This function is called by a protocol handler that wants to
2540 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002541 * new socket creation.
2542 *
2543 * If protocol handler is a module, then it can use module reference
2544 * counts to protect against new references. If protocol handler is not
2545 * a module then it needs to provide its own protection in
2546 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002547 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002548void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002550 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002552 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002553 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002554 spin_unlock(&net_family_lock);
2555
2556 synchronize_rcu();
2557
Yang Yingliang3410f222014-02-12 17:09:55 +08002558 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002560EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561
Andi Kleen77d76ea2005-12-22 12:43:42 -08002562static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002563{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002564 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002565 /*
2566 * Initialize the network sysctl infrastructure.
2567 */
2568 err = net_sysctl_init();
2569 if (err)
2570 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002571
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002573 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002574 */
2575 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002576
2577 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002578 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002579 */
2580
2581 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002582
2583 err = register_filesystem(&sock_fs_type);
2584 if (err)
2585 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002586 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002587 if (IS_ERR(sock_mnt)) {
2588 err = PTR_ERR(sock_mnt);
2589 goto out_mount;
2590 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002591
2592 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593 */
2594
2595#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002596 err = netfilter_init();
2597 if (err)
2598 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002599#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002600
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002601 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002602
Nick Pigginb3e19d92011-01-07 17:50:11 +11002603out:
2604 return err;
2605
2606out_mount:
2607 unregister_filesystem(&sock_fs_type);
2608out_fs:
2609 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610}
2611
Andi Kleen77d76ea2005-12-22 12:43:42 -08002612core_initcall(sock_init); /* early initcall */
2613
Linus Torvalds1da177e2005-04-16 15:20:36 -07002614#ifdef CONFIG_PROC_FS
2615void socket_seq_show(struct seq_file *seq)
2616{
2617 int cpu;
2618 int counter = 0;
2619
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002620 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002621 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622
2623 /* It can be negative, by the way. 8) */
2624 if (counter < 0)
2625 counter = 0;
2626
2627 seq_printf(seq, "sockets: used %d\n", counter);
2628}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002629#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002630
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002631#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002632static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002633 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002634{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002635 mm_segment_t old_fs = get_fs();
2636 struct timeval ktv;
2637 int err;
2638
2639 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002640 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002641 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002642 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002643 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002644
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002645 return err;
2646}
2647
Arnd Bergmann6b960182009-11-06 23:10:54 -08002648static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002649 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002650{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002651 mm_segment_t old_fs = get_fs();
2652 struct timespec kts;
2653 int err;
2654
2655 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002656 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002657 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002658 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002659 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002660
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002661 return err;
2662}
2663
Arnd Bergmann6b960182009-11-06 23:10:54 -08002664static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002665{
2666 struct ifreq __user *uifr;
2667 int err;
2668
2669 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002670 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002671 return -EFAULT;
2672
Arnd Bergmann6b960182009-11-06 23:10:54 -08002673 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002674 if (err)
2675 return err;
2676
Arnd Bergmann6b960182009-11-06 23:10:54 -08002677 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002678 return -EFAULT;
2679
2680 return 0;
2681}
2682
Arnd Bergmann6b960182009-11-06 23:10:54 -08002683static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002684{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002685 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002686 struct ifconf ifc;
2687 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002688 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002689 struct ifreq __user *ifr;
2690 unsigned int i, j;
2691 int err;
2692
Arnd Bergmann6b960182009-11-06 23:10:54 -08002693 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002694 return -EFAULT;
2695
Mathias Krause43da5f22012-08-15 11:31:57 +00002696 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002697 if (ifc32.ifcbuf == 0) {
2698 ifc32.ifc_len = 0;
2699 ifc.ifc_len = 0;
2700 ifc.ifc_req = NULL;
2701 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2702 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002703 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2704 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002705 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2706 ifc.ifc_len = len;
2707 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2708 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002709 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002710 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002711 return -EFAULT;
2712 ifr++;
2713 ifr32++;
2714 }
2715 }
2716 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2717 return -EFAULT;
2718
Arnd Bergmann6b960182009-11-06 23:10:54 -08002719 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002720 if (err)
2721 return err;
2722
2723 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2724 return -EFAULT;
2725
2726 ifr = ifc.ifc_req;
2727 ifr32 = compat_ptr(ifc32.ifcbuf);
2728 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002729 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2730 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2731 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002732 return -EFAULT;
2733 ifr32++;
2734 ifr++;
2735 }
2736
2737 if (ifc32.ifcbuf == 0) {
2738 /* Translate from 64-bit structure multiple to
2739 * a 32-bit one.
2740 */
2741 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002742 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002743 ifc32.ifc_len = i;
2744 } else {
2745 ifc32.ifc_len = i;
2746 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002747 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002748 return -EFAULT;
2749
2750 return 0;
2751}
2752
Arnd Bergmann6b960182009-11-06 23:10:54 -08002753static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002754{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002755 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2756 bool convert_in = false, convert_out = false;
2757 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2758 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002759 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002760 u32 rule_cnt = 0, actual_rule_cnt;
2761 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002762 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002763 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002764
2765 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2766 return -EFAULT;
2767
Ben Hutchings3a7da392011-03-17 07:34:32 +00002768 compat_rxnfc = compat_ptr(data);
2769
2770 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002771 return -EFAULT;
2772
Ben Hutchings3a7da392011-03-17 07:34:32 +00002773 /* Most ethtool structures are defined without padding.
2774 * Unfortunately struct ethtool_rxnfc is an exception.
2775 */
2776 switch (ethcmd) {
2777 default:
2778 break;
2779 case ETHTOOL_GRXCLSRLALL:
2780 /* Buffer size is variable */
2781 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2782 return -EFAULT;
2783 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2784 return -ENOMEM;
2785 buf_size += rule_cnt * sizeof(u32);
2786 /* fall through */
2787 case ETHTOOL_GRXRINGS:
2788 case ETHTOOL_GRXCLSRLCNT:
2789 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002790 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002791 convert_out = true;
2792 /* fall through */
2793 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002794 buf_size += sizeof(struct ethtool_rxnfc);
2795 convert_in = true;
2796 break;
2797 }
2798
2799 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002800 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002801
2802 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2803 return -EFAULT;
2804
2805 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2806 &ifr->ifr_ifru.ifru_data))
2807 return -EFAULT;
2808
2809 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002810 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002811 * fs.ring_cookie and at the end of fs, but nowhere else.
2812 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002813 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2814 sizeof(compat_rxnfc->fs.m_ext) !=
2815 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2816 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002817 BUILD_BUG_ON(
2818 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2819 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2820 offsetof(struct ethtool_rxnfc, fs.location) -
2821 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2822
2823 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002824 (void __user *)(&rxnfc->fs.m_ext + 1) -
2825 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002826 copy_in_user(&rxnfc->fs.ring_cookie,
2827 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002828 (void __user *)(&rxnfc->fs.location + 1) -
2829 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002830 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2831 sizeof(rxnfc->rule_cnt)))
2832 return -EFAULT;
2833 }
2834
2835 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2836 if (ret)
2837 return ret;
2838
2839 if (convert_out) {
2840 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002841 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2842 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002843 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2844 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002845 (const void __user *)(&rxnfc->fs.location + 1) -
2846 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002847 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2848 sizeof(rxnfc->rule_cnt)))
2849 return -EFAULT;
2850
2851 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2852 /* As an optimisation, we only copy the actual
2853 * number of rules that the underlying
2854 * function returned. Since Mallory might
2855 * change the rule count in user memory, we
2856 * check that it is less than the rule count
2857 * originally given (as the user buffer size),
2858 * which has been range-checked.
2859 */
2860 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2861 return -EFAULT;
2862 if (actual_rule_cnt < rule_cnt)
2863 rule_cnt = actual_rule_cnt;
2864 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2865 &rxnfc->rule_locs[0],
2866 rule_cnt * sizeof(u32)))
2867 return -EFAULT;
2868 }
2869 }
2870
2871 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002872}
2873
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002874static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2875{
2876 void __user *uptr;
2877 compat_uptr_t uptr32;
2878 struct ifreq __user *uifr;
2879
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002880 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002881 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2882 return -EFAULT;
2883
2884 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2885 return -EFAULT;
2886
2887 uptr = compat_ptr(uptr32);
2888
2889 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2890 return -EFAULT;
2891
2892 return dev_ioctl(net, SIOCWANDEV, uifr);
2893}
2894
Arnd Bergmann6b960182009-11-06 23:10:54 -08002895static int bond_ioctl(struct net *net, unsigned int cmd,
2896 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002897{
2898 struct ifreq kifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002899 mm_segment_t old_fs;
2900 int err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002901
2902 switch (cmd) {
2903 case SIOCBONDENSLAVE:
2904 case SIOCBONDRELEASE:
2905 case SIOCBONDSETHWADDR:
2906 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002907 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002908 return -EFAULT;
2909
2910 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002911 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002912 err = dev_ioctl(net, cmd,
2913 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002914 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002915
2916 return err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002917 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002918 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002919 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002920}
2921
Ben Hutchings590d4692013-11-18 17:04:13 +00002922/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2923static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08002924 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002925{
2926 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002927 char tmp_buf[IFNAMSIZ];
2928 void __user *data64;
2929 u32 data32;
2930
2931 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2932 IFNAMSIZ))
2933 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002934 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002935 return -EFAULT;
2936 data64 = compat_ptr(data32);
2937
2938 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2939
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002940 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2941 IFNAMSIZ))
2942 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002943 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002944 return -EFAULT;
2945
Arnd Bergmann6b960182009-11-06 23:10:54 -08002946 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002947}
2948
Arnd Bergmann6b960182009-11-06 23:10:54 -08002949static int dev_ifsioc(struct net *net, struct socket *sock,
2950 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002951{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002952 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002953 int err;
2954
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002955 uifr = compat_alloc_user_space(sizeof(*uifr));
2956 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2957 return -EFAULT;
2958
2959 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2960
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002961 if (!err) {
2962 switch (cmd) {
2963 case SIOCGIFFLAGS:
2964 case SIOCGIFMETRIC:
2965 case SIOCGIFMTU:
2966 case SIOCGIFMEM:
2967 case SIOCGIFHWADDR:
2968 case SIOCGIFINDEX:
2969 case SIOCGIFADDR:
2970 case SIOCGIFBRDADDR:
2971 case SIOCGIFDSTADDR:
2972 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002973 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002974 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002975 case SIOCGMIIPHY:
2976 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002977 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002978 err = -EFAULT;
2979 break;
2980 }
2981 }
2982 return err;
2983}
2984
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002985static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2986 struct compat_ifreq __user *uifr32)
2987{
2988 struct ifreq ifr;
2989 struct compat_ifmap __user *uifmap32;
2990 mm_segment_t old_fs;
2991 int err;
2992
2993 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2994 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002995 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2996 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2997 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2998 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2999 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3000 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003001 if (err)
3002 return -EFAULT;
3003
3004 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003005 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003006 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003007 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003008
3009 if (cmd == SIOCGIFMAP && !err) {
3010 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003011 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3012 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3013 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3014 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3015 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3016 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003017 if (err)
3018 err = -EFAULT;
3019 }
3020 return err;
3021}
3022
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003023struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003024 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003025 struct sockaddr rt_dst; /* target address */
3026 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3027 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003028 unsigned short rt_flags;
3029 short rt_pad2;
3030 u32 rt_pad3;
3031 unsigned char rt_tos;
3032 unsigned char rt_class;
3033 short rt_pad4;
3034 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003035 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003036 u32 rt_mtu; /* per route MTU/Window */
3037 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003038 unsigned short rt_irtt; /* Initial RTT */
3039};
3040
3041struct in6_rtmsg32 {
3042 struct in6_addr rtmsg_dst;
3043 struct in6_addr rtmsg_src;
3044 struct in6_addr rtmsg_gateway;
3045 u32 rtmsg_type;
3046 u16 rtmsg_dst_len;
3047 u16 rtmsg_src_len;
3048 u32 rtmsg_metric;
3049 u32 rtmsg_info;
3050 u32 rtmsg_flags;
3051 s32 rtmsg_ifindex;
3052};
3053
Arnd Bergmann6b960182009-11-06 23:10:54 -08003054static int routing_ioctl(struct net *net, struct socket *sock,
3055 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003056{
3057 int ret;
3058 void *r = NULL;
3059 struct in6_rtmsg r6;
3060 struct rtentry r4;
3061 char devname[16];
3062 u32 rtdev;
3063 mm_segment_t old_fs = get_fs();
3064
Arnd Bergmann6b960182009-11-06 23:10:54 -08003065 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3066 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003067 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003068 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003069 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3070 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3071 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3072 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3073 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3074 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3075 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003076
3077 r = (void *) &r6;
3078 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003079 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003080 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003081 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003082 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3083 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3084 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3085 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3086 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3087 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003088 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003089 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003090 r4.rt_dev = (char __user __force *)devname;
3091 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003092 } else
3093 r4.rt_dev = NULL;
3094
3095 r = (void *) &r4;
3096 }
3097
3098 if (ret) {
3099 ret = -EFAULT;
3100 goto out;
3101 }
3102
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003103 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003104 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003105 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003106
3107out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003108 return ret;
3109}
3110
3111/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3112 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003113 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003114 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003115static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003116{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003117 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003118
Arnd Bergmann6b960182009-11-06 23:10:54 -08003119 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003120 return -EFAULT;
3121 if (tmp == BRCTL_GET_VERSION)
3122 return BRCTL_VERSION + 1;
3123 return -EINVAL;
3124}
3125
Arnd Bergmann6b960182009-11-06 23:10:54 -08003126static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3127 unsigned int cmd, unsigned long arg)
3128{
3129 void __user *argp = compat_ptr(arg);
3130 struct sock *sk = sock->sk;
3131 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003132
Arnd Bergmann6b960182009-11-06 23:10:54 -08003133 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003134 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003135
Arnd Bergmann6b960182009-11-06 23:10:54 -08003136 switch (cmd) {
3137 case SIOCSIFBR:
3138 case SIOCGIFBR:
3139 return old_bridge_ioctl(argp);
3140 case SIOCGIFNAME:
3141 return dev_ifname32(net, argp);
3142 case SIOCGIFCONF:
3143 return dev_ifconf(net, argp);
3144 case SIOCETHTOOL:
3145 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003146 case SIOCWANDEV:
3147 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003148 case SIOCGIFMAP:
3149 case SIOCSIFMAP:
3150 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003151 case SIOCBONDENSLAVE:
3152 case SIOCBONDRELEASE:
3153 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003154 case SIOCBONDCHANGEACTIVE:
3155 return bond_ioctl(net, cmd, argp);
3156 case SIOCADDRT:
3157 case SIOCDELRT:
3158 return routing_ioctl(net, sock, cmd, argp);
3159 case SIOCGSTAMP:
3160 return do_siocgstamp(net, sock, cmd, argp);
3161 case SIOCGSTAMPNS:
3162 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003163 case SIOCBONDSLAVEINFOQUERY:
3164 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003165 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003166 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003167 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003168
Arnd Bergmann6b960182009-11-06 23:10:54 -08003169 case FIOSETOWN:
3170 case SIOCSPGRP:
3171 case FIOGETOWN:
3172 case SIOCGPGRP:
3173 case SIOCBRADDBR:
3174 case SIOCBRDELBR:
3175 case SIOCGIFVLAN:
3176 case SIOCSIFVLAN:
3177 case SIOCADDDLCI:
3178 case SIOCDELDLCI:
3179 return sock_ioctl(file, cmd, arg);
3180
3181 case SIOCGIFFLAGS:
3182 case SIOCSIFFLAGS:
3183 case SIOCGIFMETRIC:
3184 case SIOCSIFMETRIC:
3185 case SIOCGIFMTU:
3186 case SIOCSIFMTU:
3187 case SIOCGIFMEM:
3188 case SIOCSIFMEM:
3189 case SIOCGIFHWADDR:
3190 case SIOCSIFHWADDR:
3191 case SIOCADDMULTI:
3192 case SIOCDELMULTI:
3193 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003194 case SIOCGIFADDR:
3195 case SIOCSIFADDR:
3196 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003197 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003198 case SIOCGIFBRDADDR:
3199 case SIOCSIFBRDADDR:
3200 case SIOCGIFDSTADDR:
3201 case SIOCSIFDSTADDR:
3202 case SIOCGIFNETMASK:
3203 case SIOCSIFNETMASK:
3204 case SIOCSIFPFLAGS:
3205 case SIOCGIFPFLAGS:
3206 case SIOCGIFTXQLEN:
3207 case SIOCSIFTXQLEN:
3208 case SIOCBRADDIF:
3209 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003210 case SIOCSIFNAME:
3211 case SIOCGMIIPHY:
3212 case SIOCGMIIREG:
3213 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003214 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003215
Arnd Bergmann6b960182009-11-06 23:10:54 -08003216 case SIOCSARP:
3217 case SIOCGARP:
3218 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003219 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003220 return sock_do_ioctl(net, sock, cmd, arg);
3221 }
3222
Arnd Bergmann6b960182009-11-06 23:10:54 -08003223 return -ENOIOCTLCMD;
3224}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003225
Eric Dumazet95c96172012-04-15 05:58:06 +00003226static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003227 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003228{
3229 struct socket *sock = file->private_data;
3230 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003231 struct sock *sk;
3232 struct net *net;
3233
3234 sk = sock->sk;
3235 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003236
3237 if (sock->ops->compat_ioctl)
3238 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3239
David S. Miller87de87d2008-06-03 09:14:03 -07003240 if (ret == -ENOIOCTLCMD &&
3241 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3242 ret = compat_wext_handle_ioctl(net, cmd, arg);
3243
Arnd Bergmann6b960182009-11-06 23:10:54 -08003244 if (ret == -ENOIOCTLCMD)
3245 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3246
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003247 return ret;
3248}
3249#endif
3250
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003251int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3252{
3253 return sock->ops->bind(sock, addr, addrlen);
3254}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003255EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003256
3257int kernel_listen(struct socket *sock, int backlog)
3258{
3259 return sock->ops->listen(sock, backlog);
3260}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003261EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003262
3263int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3264{
3265 struct sock *sk = sock->sk;
3266 int err;
3267
3268 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3269 newsock);
3270 if (err < 0)
3271 goto done;
3272
3273 err = sock->ops->accept(sock, *newsock, flags);
3274 if (err < 0) {
3275 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003276 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003277 goto done;
3278 }
3279
3280 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003281 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003282
3283done:
3284 return err;
3285}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003286EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003287
3288int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003289 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003290{
3291 return sock->ops->connect(sock, addr, addrlen, flags);
3292}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003293EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003294
3295int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3296 int *addrlen)
3297{
3298 return sock->ops->getname(sock, addr, addrlen, 0);
3299}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003300EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003301
3302int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3303 int *addrlen)
3304{
3305 return sock->ops->getname(sock, addr, addrlen, 1);
3306}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003307EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003308
3309int kernel_getsockopt(struct socket *sock, int level, int optname,
3310 char *optval, int *optlen)
3311{
3312 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003313 char __user *uoptval;
3314 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003315 int err;
3316
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003317 uoptval = (char __user __force *) optval;
3318 uoptlen = (int __user __force *) optlen;
3319
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003320 set_fs(KERNEL_DS);
3321 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003322 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003323 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003324 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3325 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003326 set_fs(oldfs);
3327 return err;
3328}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003329EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003330
3331int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003332 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003333{
3334 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003335 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003336 int err;
3337
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003338 uoptval = (char __user __force *) optval;
3339
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003340 set_fs(KERNEL_DS);
3341 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003342 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003343 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003344 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003345 optlen);
3346 set_fs(oldfs);
3347 return err;
3348}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003349EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003350
3351int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3352 size_t size, int flags)
3353{
3354 if (sock->ops->sendpage)
3355 return sock->ops->sendpage(sock, page, offset, size, flags);
3356
3357 return sock_no_sendpage(sock, page, offset, size, flags);
3358}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003359EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003360
3361int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3362{
3363 mm_segment_t oldfs = get_fs();
3364 int err;
3365
3366 set_fs(KERNEL_DS);
3367 err = sock->ops->ioctl(sock, cmd, arg);
3368 set_fs(oldfs);
3369
3370 return err;
3371}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003372EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003373
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003374int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3375{
3376 return sock->ops->shutdown(sock, how);
3377}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003378EXPORT_SYMBOL(kernel_sock_shutdown);