blob: 418795caa8979fd0eb5d19e4d1fe3376fc0e9b20 [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
Badari Pulavarty027445c2006-09-30 23:28:46 -0700116static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
117 unsigned long nr_segs, loff_t pos);
118static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
119 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700120static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
122static int sock_close(struct inode *inode, struct file *file);
123static unsigned int sock_poll(struct file *file,
124 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700125static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800126#ifdef CONFIG_COMPAT
127static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700128 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800129#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131static ssize_t sock_sendpage(struct file *file, struct page *page,
132 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800133static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700134 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800135 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137/*
138 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
139 * in the operation structures but are done directly via the socketcall() multiplexor.
140 */
141
Arjan van de Venda7071d2007-02-12 00:55:36 -0800142static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 .owner = THIS_MODULE,
144 .llseek = no_llseek,
145 .aio_read = sock_aio_read,
146 .aio_write = sock_aio_write,
147 .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{
616 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
618 si->sock = sock;
619 si->scm = NULL;
620 si->msg = msg;
621 si->size = size;
622
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 return sock->ops->sendmsg(iocb, sock, msg, size);
624}
625
Anton Blanchard228e5482011-05-02 20:21:35 +0000626static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
627 struct msghdr *msg, size_t size)
628{
629 int err = security_socket_sendmsg(sock, msg, size);
630
631 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
632}
633
Gu Zheng0cf00c62014-12-05 15:14:23 +0800634static int do_sock_sendmsg(struct socket *sock, struct msghdr *msg,
635 size_t size, bool nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636{
637 struct kiocb iocb;
638 struct sock_iocb siocb;
639 int ret;
640
641 init_sync_kiocb(&iocb, NULL);
642 iocb.private = &siocb;
Gu Zheng0cf00c62014-12-05 15:14:23 +0800643 ret = nosec ? __sock_sendmsg_nosec(&iocb, sock, msg, size) :
644 __sock_sendmsg(&iocb, sock, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 if (-EIOCBQUEUED == ret)
646 ret = wait_on_sync_kiocb(&iocb);
647 return ret;
648}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800649
650int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
651{
652 return do_sock_sendmsg(sock, msg, size, false);
653}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700654EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
Eric Dumazet894dc242011-07-26 02:39:41 +0000656static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000657{
Gu Zheng0cf00c62014-12-05 15:14:23 +0800658 return do_sock_sendmsg(sock, msg, size, true);
Anton Blanchard228e5482011-05-02 20:21:35 +0000659}
660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
662 struct kvec *vec, size_t num, size_t size)
663{
664 mm_segment_t oldfs = get_fs();
665 int result;
666
667 set_fs(KERNEL_DS);
668 /*
669 * the following is safe, since for compiler definitions of kvec and
670 * iovec are identical, yielding the same in-core layout and alignment
671 */
Al Viroc0371da2014-11-24 10:42:55 -0500672 iov_iter_init(&msg->msg_iter, WRITE, (struct iovec *)vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 result = sock_sendmsg(sock, msg, size);
674 set_fs(oldfs);
675 return result;
676}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700677EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700679/*
680 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
681 */
682void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
683 struct sk_buff *skb)
684{
Patrick Ohly20d49472009-02-12 05:03:38 +0000685 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400686 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000687 int empty = 1;
688 struct skb_shared_hwtstamps *shhwtstamps =
689 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700690
Patrick Ohly20d49472009-02-12 05:03:38 +0000691 /* Race occurred between timestamp enabling and packet
692 receiving. Fill in the current time for now. */
693 if (need_software_tstamp && skb->tstamp.tv64 == 0)
694 __net_timestamp(skb);
695
696 if (need_software_tstamp) {
697 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
698 struct timeval tv;
699 skb_get_timestamp(skb, &tv);
700 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
701 sizeof(tv), &tv);
702 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400703 struct timespec ts;
704 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000705 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400706 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000707 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700708 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000709
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400710 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400711 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400712 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000713 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400714 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400715 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400716 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400717 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000718 if (!empty)
719 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400720 SCM_TIMESTAMPING, sizeof(tss), &tss);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700721}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300722EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
723
Johannes Berg6e3e9392011-11-09 10:15:42 +0100724void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
725 struct sk_buff *skb)
726{
727 int ack;
728
729 if (!sock_flag(sk, SOCK_WIFI_STATUS))
730 return;
731 if (!skb->wifi_acked_valid)
732 return;
733
734 ack = skb->wifi_acked;
735
736 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
737}
738EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
739
stephen hemminger11165f12010-10-18 14:27:29 +0000740static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
741 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700742{
743 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
744 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
745 sizeof(__u32), &skb->dropcount);
746}
747
Eric Dumazet767dd032010-04-28 19:14:43 +0000748void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700749 struct sk_buff *skb)
750{
751 sock_recv_timestamp(msg, sk, skb);
752 sock_recv_drops(msg, sk, skb);
753}
Eric Dumazet767dd032010-04-28 19:14:43 +0000754EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700755
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700756static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
757 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 struct sock_iocb *si = kiocb_to_siocb(iocb);
760
761 si->sock = sock;
762 si->scm = NULL;
763 si->msg = msg;
764 si->size = size;
765 si->flags = flags;
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
768}
769
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700770static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
771 struct msghdr *msg, size_t size, int flags)
772{
773 int err = security_socket_recvmsg(sock, msg, size, flags);
774
775 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
776}
777
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700778int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 size_t size, int flags)
780{
781 struct kiocb iocb;
782 struct sock_iocb siocb;
783 int ret;
784
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700785 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 iocb.private = &siocb;
787 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
788 if (-EIOCBQUEUED == ret)
789 ret = wait_on_sync_kiocb(&iocb);
790 return ret;
791}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700792EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700794static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
795 size_t size, int flags)
796{
797 struct kiocb iocb;
798 struct sock_iocb siocb;
799 int ret;
800
801 init_sync_kiocb(&iocb, NULL);
802 iocb.private = &siocb;
803 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
804 if (-EIOCBQUEUED == ret)
805 ret = wait_on_sync_kiocb(&iocb);
806 return ret;
807}
808
Martin Lucinac1249c02010-12-10 00:04:05 +0000809/**
810 * kernel_recvmsg - Receive a message from a socket (kernel space)
811 * @sock: The socket to receive the message from
812 * @msg: Received message
813 * @vec: Input s/g array for message data
814 * @num: Size of input s/g array
815 * @size: Number of bytes to read
816 * @flags: Message flags (MSG_DONTWAIT, etc...)
817 *
818 * On return the msg structure contains the scatter/gather array passed in the
819 * vec argument. The array is modified so that it consists of the unfilled
820 * portion of the original array.
821 *
822 * The returned value is the total number of bytes received, or an error.
823 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700824int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
825 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826{
827 mm_segment_t oldfs = get_fs();
828 int result;
829
830 set_fs(KERNEL_DS);
831 /*
832 * the following is safe, since for compiler definitions of kvec and
833 * iovec are identical, yielding the same in-core layout and alignment
834 */
Al Viroc0371da2014-11-24 10:42:55 -0500835 iov_iter_init(&msg->msg_iter, READ, (struct iovec *)vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 result = sock_recvmsg(sock, msg, size, flags);
837 set_fs(oldfs);
838 return result;
839}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700840EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300842static ssize_t sock_sendpage(struct file *file, struct page *page,
843 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844{
845 struct socket *sock;
846 int flags;
847
Eric Dumazetb69aee02005-09-06 14:42:45 -0700848 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Eric Dumazet35f9c092012-04-05 03:05:35 +0000850 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
851 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
852 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Linus Torvaldse6949582009-08-13 08:28:36 -0700854 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855}
856
Jens Axboe9c55e012007-11-06 23:30:13 -0800857static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700858 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800859 unsigned int flags)
860{
861 struct socket *sock = file->private_data;
862
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800863 if (unlikely(!sock->ops->splice_read))
864 return -EINVAL;
865
Jens Axboe9c55e012007-11-06 23:30:13 -0800866 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
867}
868
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800869static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700870 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800871{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800872 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800873 iocb->private = siocb;
874 return siocb;
875}
876
877static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700878 struct file *file, const struct iovec *iov,
879 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800880{
881 struct socket *sock = file->private_data;
882 size_t size = 0;
883 int i;
884
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700885 for (i = 0; i < nr_segs; i++)
886 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800887
888 msg->msg_name = NULL;
889 msg->msg_namelen = 0;
890 msg->msg_control = NULL;
891 msg->msg_controllen = 0;
Al Viroc0371da2014-11-24 10:42:55 -0500892 iov_iter_init(&msg->msg_iter, READ, iov, nr_segs, size);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800893 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
894
895 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
896}
897
Badari Pulavarty027445c2006-09-30 23:28:46 -0700898static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
899 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800900{
901 struct sock_iocb siocb, *x;
902
903 if (pos != 0)
904 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700905
Kent Overstreet73a70752013-05-09 15:03:42 -0700906 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800907 return 0;
908
Badari Pulavarty027445c2006-09-30 23:28:46 -0700909
910 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800911 if (!x)
912 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700913 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800914}
915
916static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700917 struct file *file, const struct iovec *iov,
918 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800919{
920 struct socket *sock = file->private_data;
921 size_t size = 0;
922 int i;
923
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700924 for (i = 0; i < nr_segs; i++)
925 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800926
927 msg->msg_name = NULL;
928 msg->msg_namelen = 0;
929 msg->msg_control = NULL;
930 msg->msg_controllen = 0;
Al Viroc0371da2014-11-24 10:42:55 -0500931 iov_iter_init(&msg->msg_iter, WRITE, iov, nr_segs, size);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800932 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
933 if (sock->type == SOCK_SEQPACKET)
934 msg->msg_flags |= MSG_EOR;
935
936 return __sock_sendmsg(iocb, sock, msg, size);
937}
938
Badari Pulavarty027445c2006-09-30 23:28:46 -0700939static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
940 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800942 struct sock_iocb siocb, *x;
943
944 if (pos != 0)
945 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700946
Badari Pulavarty027445c2006-09-30 23:28:46 -0700947 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800948 if (!x)
949 return -ENOMEM;
950
Badari Pulavarty027445c2006-09-30 23:28:46 -0700951 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952}
953
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954/*
955 * Atomic setting of ioctl hooks to avoid race
956 * with module unload.
957 */
958
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800959static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700960static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961
Eric W. Biederman881d9662007-09-17 11:56:21 -0700962void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800964 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800966 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967}
968EXPORT_SYMBOL(brioctl_set);
969
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800970static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700971static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
Eric W. Biederman881d9662007-09-17 11:56:21 -0700973void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800975 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800977 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978}
979EXPORT_SYMBOL(vlan_ioctl_set);
980
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800981static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700982static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700984void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800986 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800988 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989}
990EXPORT_SYMBOL(dlci_ioctl_set);
991
Arnd Bergmann6b960182009-11-06 23:10:54 -0800992static long sock_do_ioctl(struct net *net, struct socket *sock,
993 unsigned int cmd, unsigned long arg)
994{
995 int err;
996 void __user *argp = (void __user *)arg;
997
998 err = sock->ops->ioctl(sock, cmd, arg);
999
1000 /*
1001 * If this ioctl is unknown try to hand it down
1002 * to the NIC driver.
1003 */
1004 if (err == -ENOIOCTLCMD)
1005 err = dev_ioctl(net, cmd, argp);
1006
1007 return err;
1008}
1009
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010/*
1011 * With an ioctl, arg may well be a user mode pointer, but we don't know
1012 * what to do with it - that's up to the protocol still.
1013 */
1014
1015static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1016{
1017 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001018 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 void __user *argp = (void __user *)arg;
1020 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001021 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022
Eric Dumazetb69aee02005-09-06 14:42:45 -07001023 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001024 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001025 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001027 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001029#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001031 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001033#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001034 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 case FIOSETOWN:
1036 case SIOCSPGRP:
1037 err = -EFAULT;
1038 if (get_user(pid, (int __user *)argp))
1039 break;
Jeff Laytone0b93ed2014-08-22 11:27:32 -04001040 f_setown(sock->file, pid, 1);
1041 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 break;
1043 case FIOGETOWN:
1044 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001045 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001046 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 break;
1048 case SIOCGIFBR:
1049 case SIOCSIFBR:
1050 case SIOCBRADDBR:
1051 case SIOCBRDELBR:
1052 err = -ENOPKG;
1053 if (!br_ioctl_hook)
1054 request_module("bridge");
1055
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001056 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001057 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001058 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001059 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 break;
1061 case SIOCGIFVLAN:
1062 case SIOCSIFVLAN:
1063 err = -ENOPKG;
1064 if (!vlan_ioctl_hook)
1065 request_module("8021q");
1066
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001067 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001069 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001070 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 case SIOCADDDLCI:
1073 case SIOCDELDLCI:
1074 err = -ENOPKG;
1075 if (!dlci_ioctl_hook)
1076 request_module("dlci");
1077
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001078 mutex_lock(&dlci_ioctl_mutex);
1079 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001081 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 break;
1083 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001084 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001086 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087 return err;
1088}
1089
1090int sock_create_lite(int family, int type, int protocol, struct socket **res)
1091{
1092 int err;
1093 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001094
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 err = security_socket_create(family, type, protocol, 1);
1096 if (err)
1097 goto out;
1098
1099 sock = sock_alloc();
1100 if (!sock) {
1101 err = -ENOMEM;
1102 goto out;
1103 }
1104
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001106 err = security_socket_post_create(sock, family, type, protocol, 1);
1107 if (err)
1108 goto out_release;
1109
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110out:
1111 *res = sock;
1112 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001113out_release:
1114 sock_release(sock);
1115 sock = NULL;
1116 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001118EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119
1120/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001121static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122{
Eliezer Tamircbf55002013-07-08 16:20:34 +03001123 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 struct socket *sock;
1125
1126 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001127 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001129 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001130
Eliezer Tamircbf55002013-07-08 16:20:34 +03001131 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001132 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001133 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001134
1135 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001136 if (wait && (wait->_key & POLL_BUSY_LOOP))
1137 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001138 }
1139
Eliezer Tamircbf55002013-07-08 16:20:34 +03001140 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141}
1142
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001143static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001145 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
1147 return sock->ops->mmap(file, sock, vma);
1148}
1149
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001150static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 sock_release(SOCKET_I(inode));
1153 return 0;
1154}
1155
1156/*
1157 * Update the socket async list
1158 *
1159 * Fasync_list locking strategy.
1160 *
1161 * 1. fasync_list is modified only under process context socket lock
1162 * i.e. under semaphore.
1163 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001164 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 */
1166
1167static int sock_fasync(int fd, struct file *filp, int on)
1168{
Eric Dumazet989a2972010-04-14 09:55:35 +00001169 struct socket *sock = filp->private_data;
1170 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001171 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
Eric Dumazet989a2972010-04-14 09:55:35 +00001173 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175
1176 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001177 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1178 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179
Eric Dumazeteaefd112011-02-18 03:26:36 +00001180 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001181 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001182 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001183 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184
Eric Dumazet989a2972010-04-14 09:55:35 +00001185 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 return 0;
1187}
1188
Eric Dumazet43815482010-04-29 11:01:49 +00001189/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190
1191int sock_wake_async(struct socket *sock, int how, int band)
1192{
Eric Dumazet43815482010-04-29 11:01:49 +00001193 struct socket_wq *wq;
1194
1195 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001197 rcu_read_lock();
1198 wq = rcu_dereference(sock->wq);
1199 if (!wq || !wq->fasync_list) {
1200 rcu_read_unlock();
1201 return -1;
1202 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001203 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001204 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1206 break;
1207 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001208 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1210 break;
1211 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001212 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001213call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001214 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001216 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001217 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 }
Eric Dumazet43815482010-04-29 11:01:49 +00001219 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 return 0;
1221}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001222EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223
Pavel Emelyanov721db932010-09-29 16:06:32 +04001224int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001225 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226{
1227 int err;
1228 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001229 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230
1231 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001232 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 */
1234 if (family < 0 || family >= NPROTO)
1235 return -EAFNOSUPPORT;
1236 if (type < 0 || type >= SOCK_MAX)
1237 return -EINVAL;
1238
1239 /* Compatibility.
1240
1241 This uglymoron is moved from INET layer to here to avoid
1242 deadlock in module load.
1243 */
1244 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001245 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 if (!warned) {
1247 warned = 1;
Yang Yingliang3410f222014-02-12 17:09:55 +08001248 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1249 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 }
1251 family = PF_PACKET;
1252 }
1253
1254 err = security_socket_create(family, type, protocol, kern);
1255 if (err)
1256 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001257
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001258 /*
1259 * Allocate the socket and allow the family to set things up. if
1260 * the protocol is 0, the family is instructed to select an appropriate
1261 * default.
1262 */
1263 sock = sock_alloc();
1264 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001265 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001266 return -ENFILE; /* Not exactly a match, but its the
1267 closest posix thing */
1268 }
1269
1270 sock->type = type;
1271
Johannes Berg95a5afc2008-10-16 15:24:51 -07001272#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001273 /* Attempt to load a protocol module if the find failed.
1274 *
1275 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 * requested real, full-featured networking support upon configuration.
1277 * Otherwise module support will break!
1278 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001279 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001280 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281#endif
1282
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001283 rcu_read_lock();
1284 pf = rcu_dereference(net_families[family]);
1285 err = -EAFNOSUPPORT;
1286 if (!pf)
1287 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288
1289 /*
1290 * We will call the ->create function, that possibly is in a loadable
1291 * module, so we have to bump that loadable module refcnt first.
1292 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001293 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 goto out_release;
1295
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001296 /* Now protected by module ref count */
1297 rcu_read_unlock();
1298
Eric Paris3f378b62009-11-05 22:18:14 -08001299 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001300 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001302
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 /*
1304 * Now to bump the refcnt of the [loadable] module that owns this
1305 * socket at sock_release time we decrement its refcnt.
1306 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001307 if (!try_module_get(sock->ops->owner))
1308 goto out_module_busy;
1309
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 /*
1311 * Now that we're done with the ->create function, the [loadable]
1312 * module can have its refcnt decremented
1313 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001314 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001315 err = security_socket_post_create(sock, family, type, protocol, kern);
1316 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001317 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001318 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001320 return 0;
1321
1322out_module_busy:
1323 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001325 sock->ops = NULL;
1326 module_put(pf->owner);
1327out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001329 return err;
1330
1331out_release:
1332 rcu_read_unlock();
1333 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001335EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
1337int sock_create(int family, int type, int protocol, struct socket **res)
1338{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001339 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001341EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342
1343int sock_create_kern(int family, int type, int protocol, struct socket **res)
1344{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001345 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001347EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001349SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350{
1351 int retval;
1352 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001353 int flags;
1354
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001355 /* Check the SOCK_* constants for consistency. */
1356 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1357 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1358 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1359 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1360
Ulrich Dreppera677a032008-07-23 21:29:17 -07001361 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001362 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001363 return -EINVAL;
1364 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001366 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1367 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1368
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 retval = sock_create(family, type, protocol, &sock);
1370 if (retval < 0)
1371 goto out;
1372
Ulrich Drepper77d27202008-07-23 21:29:35 -07001373 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 if (retval < 0)
1375 goto out_release;
1376
1377out:
1378 /* It may be already another descriptor 8) Not kernel problem. */
1379 return retval;
1380
1381out_release:
1382 sock_release(sock);
1383 return retval;
1384}
1385
1386/*
1387 * Create a pair of connected sockets.
1388 */
1389
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001390SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1391 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392{
1393 struct socket *sock1, *sock2;
1394 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001395 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001396 int flags;
1397
1398 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001399 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001400 return -EINVAL;
1401 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001403 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1404 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1405
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 /*
1407 * Obtain the first socket and check if the underlying protocol
1408 * supports the socketpair call.
1409 */
1410
1411 err = sock_create(family, type, protocol, &sock1);
1412 if (err < 0)
1413 goto out;
1414
1415 err = sock_create(family, type, protocol, &sock2);
1416 if (err < 0)
1417 goto out_release_1;
1418
1419 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001420 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 goto out_release_both;
1422
Al Viro28407632012-08-17 23:54:15 -04001423 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001424 if (unlikely(fd1 < 0)) {
1425 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001427 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001428
Al Viro28407632012-08-17 23:54:15 -04001429 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001430 if (unlikely(fd2 < 0)) {
1431 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001432 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001433 }
1434
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001435 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001436 if (unlikely(IS_ERR(newfile1))) {
1437 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001438 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001439 }
1440
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001441 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001442 if (IS_ERR(newfile2)) {
1443 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001444 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001445 }
1446
Yann Droneaudd73aa282013-12-09 22:42:20 +01001447 err = put_user(fd1, &usockvec[0]);
1448 if (err)
1449 goto out_fput_both;
1450
1451 err = put_user(fd2, &usockvec[1]);
1452 if (err)
1453 goto out_fput_both;
1454
Al Viro157cf642008-12-14 04:57:47 -05001455 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001456
Al Virodb349502007-02-07 01:48:00 -05001457 fd_install(fd1, newfile1);
1458 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 /* fd1 and fd2 may be already another descriptors.
1460 * Not kernel problem.
1461 */
1462
Yann Droneaudd73aa282013-12-09 22:42:20 +01001463 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Yann Droneaudd73aa282013-12-09 22:42:20 +01001465out_fput_both:
1466 fput(newfile2);
1467 fput(newfile1);
1468 put_unused_fd(fd2);
1469 put_unused_fd(fd1);
1470 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
Yann Droneaudd73aa282013-12-09 22:42:20 +01001472out_fput_1:
1473 fput(newfile1);
1474 put_unused_fd(fd2);
1475 put_unused_fd(fd1);
1476 sock_release(sock2);
1477 goto out;
1478
1479out_put_unused_both:
1480 put_unused_fd(fd2);
1481out_put_unused_1:
1482 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001484 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001486 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487out:
1488 return err;
1489}
1490
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491/*
1492 * Bind a name to a socket. Nothing much to do here since it's
1493 * the protocol's responsibility to handle the local address.
1494 *
1495 * We move the socket address to kernel space before we call
1496 * the protocol layer (having also checked the address is ok).
1497 */
1498
Heiko Carstens20f37032009-01-14 14:14:23 +01001499SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500{
1501 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001502 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001503 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001505 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001506 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001507 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001508 if (err >= 0) {
1509 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001510 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001511 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001512 if (!err)
1513 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001514 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001515 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001517 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001518 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 return err;
1520}
1521
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522/*
1523 * Perform a listen. Basically, we allow the protocol to do anything
1524 * necessary for a listen, and if that works, we mark the socket as
1525 * ready for listening.
1526 */
1527
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001528SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529{
1530 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001531 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001532 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001533
1534 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1535 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001536 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001537 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001538 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539
1540 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001541 if (!err)
1542 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001544 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 }
1546 return err;
1547}
1548
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549/*
1550 * For accept, we attempt to create a new socket, set up the link
1551 * with the client, wake up the client, then return the new
1552 * connected fd. We collect the address of the connector in kernel
1553 * space and move it to user at the very end. This is unclean because
1554 * we open the socket then return an error.
1555 *
1556 * 1003.1g adds the ability to recvmsg() to query connection pending
1557 * status to recvmsg. We need to add that support in a way thats
1558 * clean when we restucture accept also.
1559 */
1560
Heiko Carstens20f37032009-01-14 14:14:23 +01001561SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1562 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563{
1564 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001565 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001566 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001567 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
Ulrich Drepper77d27202008-07-23 21:29:35 -07001569 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001570 return -EINVAL;
1571
1572 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1573 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1574
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001575 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 if (!sock)
1577 goto out;
1578
1579 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001580 newsock = sock_alloc();
1581 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 goto out_put;
1583
1584 newsock->type = sock->type;
1585 newsock->ops = sock->ops;
1586
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587 /*
1588 * We don't need try_module_get here, as the listening socket (sock)
1589 * has the protocol module (sock->ops->owner) held.
1590 */
1591 __module_get(newsock->ops->owner);
1592
Al Viro28407632012-08-17 23:54:15 -04001593 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001594 if (unlikely(newfd < 0)) {
1595 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001596 sock_release(newsock);
1597 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001598 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001599 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001600 if (unlikely(IS_ERR(newfile))) {
1601 err = PTR_ERR(newfile);
1602 put_unused_fd(newfd);
1603 sock_release(newsock);
1604 goto out_put;
1605 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001606
Frank Filza79af592005-09-27 15:23:38 -07001607 err = security_socket_accept(sock, newsock);
1608 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001609 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001610
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1612 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001613 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614
1615 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001616 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001617 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001619 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001621 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001622 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001624 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 }
1626
1627 /* File flags are not inherited via accept() unlike another OSes. */
1628
David S. Miller39d8c1b2006-03-20 17:13:49 -08001629 fd_install(newfd, newfile);
1630 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001633 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634out:
1635 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001636out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001637 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001638 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639 goto out_put;
1640}
1641
Heiko Carstens20f37032009-01-14 14:14:23 +01001642SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1643 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001644{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001645 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001646}
1647
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648/*
1649 * Attempt to connect to a socket with the server address. The address
1650 * is in user space so we verify it is OK and move it to kernel space.
1651 *
1652 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1653 * break bindings
1654 *
1655 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1656 * other SEQPACKET protocols that take time to connect() as it doesn't
1657 * include the -EINPROGRESS status for such sockets.
1658 */
1659
Heiko Carstens20f37032009-01-14 14:14:23 +01001660SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1661 int, addrlen)
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 err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001667 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 if (!sock)
1669 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001670 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 if (err < 0)
1672 goto out_put;
1673
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001674 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001675 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 if (err)
1677 goto out_put;
1678
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001679 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 sock->file->f_flags);
1681out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001682 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683out:
1684 return err;
1685}
1686
1687/*
1688 * Get the local address ('name') of a socket object. Move the obtained
1689 * name to user space.
1690 */
1691
Heiko Carstens20f37032009-01-14 14:14:23 +01001692SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1693 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694{
1695 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001696 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001697 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001698
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001699 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 if (!sock)
1701 goto out;
1702
1703 err = security_socket_getsockname(sock);
1704 if (err)
1705 goto out_put;
1706
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001707 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 if (err)
1709 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001710 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711
1712out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001713 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714out:
1715 return err;
1716}
1717
1718/*
1719 * Get the remote address ('name') of a socket object. Move the obtained
1720 * name to user space.
1721 */
1722
Heiko Carstens20f37032009-01-14 14:14:23 +01001723SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1724 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725{
1726 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001727 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001728 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001730 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1731 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 err = security_socket_getpeername(sock);
1733 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001734 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 return err;
1736 }
1737
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001738 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001739 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001740 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001742 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001743 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001744 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 }
1746 return err;
1747}
1748
1749/*
1750 * Send a datagram to a given address. We move the address into kernel
1751 * space and check the user space data area is readable before invoking
1752 * the protocol.
1753 */
1754
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001755SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001756 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001757 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758{
1759 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001760 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 int err;
1762 struct msghdr msg;
1763 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001764 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001765
Linus Torvalds253eacc2010-10-30 16:43:10 -07001766 if (len > INT_MAX)
1767 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001768 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001770 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001771
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001772 iov.iov_base = buff;
1773 iov.iov_len = len;
1774 msg.msg_name = NULL;
Al Viroc0371da2014-11-24 10:42:55 -05001775 iov_iter_init(&msg.msg_iter, WRITE, &iov, 1, len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001776 msg.msg_control = NULL;
1777 msg.msg_controllen = 0;
1778 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001779 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001780 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 if (err < 0)
1782 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001783 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001784 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 }
1786 if (sock->file->f_flags & O_NONBLOCK)
1787 flags |= MSG_DONTWAIT;
1788 msg.msg_flags = flags;
1789 err = sock_sendmsg(sock, &msg, len);
1790
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001791out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001792 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001793out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 return err;
1795}
1796
1797/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001798 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 */
1800
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001801SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001802 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803{
1804 return sys_sendto(fd, buff, len, flags, NULL, 0);
1805}
1806
1807/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001808 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 * sender. We verify the buffers are writable and if needed move the
1810 * sender address from kernel to user space.
1811 */
1812
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001813SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001814 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001815 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816{
1817 struct socket *sock;
1818 struct iovec iov;
1819 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001820 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001821 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001822 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823
Linus Torvalds253eacc2010-10-30 16:43:10 -07001824 if (size > INT_MAX)
1825 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001826 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001828 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001830 msg.msg_control = NULL;
1831 msg.msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001832 iov.iov_len = size;
1833 iov.iov_base = ubuf;
Al Viroc0371da2014-11-24 10:42:55 -05001834 iov_iter_init(&msg.msg_iter, READ, &iov, 1, size);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001835 /* Save some cycles and don't copy the address if not needed */
1836 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1837 /* We assume all kernel code knows the size of sockaddr_storage */
1838 msg.msg_namelen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 if (sock->file->f_flags & O_NONBLOCK)
1840 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001841 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001843 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001844 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001845 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001846 if (err2 < 0)
1847 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001849
1850 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001851out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 return err;
1853}
1854
1855/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001856 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 */
1858
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001859SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1860 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861{
1862 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1863}
1864
1865/*
1866 * Set a socket option. Because we don't know the option lengths we have
1867 * to pass the user mode parameter for the protocols to sort out.
1868 */
1869
Heiko Carstens20f37032009-01-14 14:14:23 +01001870SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1871 char __user *, optval, int, optlen)
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
1876 if (optlen < 0)
1877 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001878
1879 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1880 if (sock != NULL) {
1881 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001882 if (err)
1883 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884
1885 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001886 err =
1887 sock_setsockopt(sock, level, optname, optval,
1888 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890 err =
1891 sock->ops->setsockopt(sock, level, optname, optval,
1892 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001893out_put:
1894 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 }
1896 return err;
1897}
1898
1899/*
1900 * Get a socket option. Because we don't know the option lengths we have
1901 * to pass a user mode parameter for the protocols to sort out.
1902 */
1903
Heiko Carstens20f37032009-01-14 14:14:23 +01001904SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1905 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001907 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908 struct socket *sock;
1909
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1911 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001912 err = security_socket_getsockopt(sock, level, optname);
1913 if (err)
1914 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915
1916 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001917 err =
1918 sock_getsockopt(sock, level, optname, optval,
1919 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001921 err =
1922 sock->ops->getsockopt(sock, level, optname, optval,
1923 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001924out_put:
1925 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 }
1927 return err;
1928}
1929
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930/*
1931 * Shutdown a socket.
1932 */
1933
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001934SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001936 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 struct socket *sock;
1938
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001939 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1940 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001942 if (!err)
1943 err = sock->ops->shutdown(sock, how);
1944 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 }
1946 return err;
1947}
1948
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001949/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 * fields which are the same type (int / unsigned) on our platforms.
1951 */
1952#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1953#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1954#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1955
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001956struct used_address {
1957 struct sockaddr_storage name;
1958 unsigned int name_len;
1959};
1960
Al Viro08adb7d2014-11-10 20:23:13 -05001961static ssize_t copy_msghdr_from_user(struct msghdr *kmsg,
1962 struct user_msghdr __user *umsg,
1963 struct sockaddr __user **save_addr,
1964 struct iovec **iov)
Dan Carpenter1661bf32013-10-03 00:27:20 +03001965{
Al Viro08adb7d2014-11-10 20:23:13 -05001966 struct sockaddr __user *uaddr;
1967 struct iovec __user *uiov;
Al Viroc0371da2014-11-24 10:42:55 -05001968 size_t nr_segs;
Al Viro08adb7d2014-11-10 20:23:13 -05001969 ssize_t err;
1970
1971 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1972 __get_user(uaddr, &umsg->msg_name) ||
1973 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1974 __get_user(uiov, &umsg->msg_iov) ||
Al Viroc0371da2014-11-24 10:42:55 -05001975 __get_user(nr_segs, &umsg->msg_iovlen) ||
Al Viro08adb7d2014-11-10 20:23:13 -05001976 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1977 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1978 __get_user(kmsg->msg_flags, &umsg->msg_flags))
Dan Carpenter1661bf32013-10-03 00:27:20 +03001979 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001980
Al Viro08adb7d2014-11-10 20:23:13 -05001981 if (!uaddr)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07001982 kmsg->msg_namelen = 0;
1983
Matthew Leachdbb490b2014-03-11 11:58:27 +00001984 if (kmsg->msg_namelen < 0)
1985 return -EINVAL;
1986
Dan Carpenter1661bf32013-10-03 00:27:20 +03001987 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001988 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05001989
1990 if (save_addr)
1991 *save_addr = uaddr;
1992
1993 if (uaddr && kmsg->msg_namelen) {
1994 if (!save_addr) {
1995 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1996 kmsg->msg_name);
1997 if (err < 0)
1998 return err;
1999 }
2000 } else {
2001 kmsg->msg_name = NULL;
2002 kmsg->msg_namelen = 0;
2003 }
2004
Al Viroc0371da2014-11-24 10:42:55 -05002005 if (nr_segs > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05002006 return -EMSGSIZE;
2007
2008 err = rw_copy_check_uvector(save_addr ? READ : WRITE,
Al Viroc0371da2014-11-24 10:42:55 -05002009 uiov, nr_segs,
Al Viro08adb7d2014-11-10 20:23:13 -05002010 UIO_FASTIOV, *iov, iov);
2011 if (err >= 0)
Al Viroc0371da2014-11-24 10:42:55 -05002012 iov_iter_init(&kmsg->msg_iter, save_addr ? READ : WRITE,
2013 *iov, nr_segs, err);
Al Viro08adb7d2014-11-10 20:23:13 -05002014 return err;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002015}
2016
Al Viro666547f2014-04-06 14:03:05 -04002017static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002018 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002019 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002021 struct compat_msghdr __user *msg_compat =
2022 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002023 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002025 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002026 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2027 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 unsigned char *ctl_buf = ctl;
Al Viro08adb7d2014-11-10 20:23:13 -05002029 int ctl_len, total_len;
2030 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002031
Al Viro08adb7d2014-11-10 20:23:13 -05002032 msg_sys->msg_name = &address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Al Viro08449322014-11-09 22:33:45 -05002034 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002035 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
Al Viro08449322014-11-09 22:33:45 -05002036 else
Al Viro08adb7d2014-11-10 20:23:13 -05002037 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002038 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039 goto out_freeiov;
2040 total_len = err;
2041
2042 err = -ENOBUFS;
2043
Anton Blanchard228e5482011-05-02 20:21:35 +00002044 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002045 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002046 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002048 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002049 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002050 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 if (err)
2052 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002053 ctl_buf = msg_sys->msg_control;
2054 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002056 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002058 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 goto out_freeiov;
2060 }
2061 err = -EFAULT;
2062 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002063 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2065 * checking falls down on this.
2066 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002067 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002068 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002069 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002071 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002073 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074
2075 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002076 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002077 /*
2078 * If this is sendmmsg() and current destination address is same as
2079 * previously succeeded address, omit asking LSM's decision.
2080 * used_address->name_len is initialized to UINT_MAX so that the first
2081 * destination address never matches.
2082 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002083 if (used_address && msg_sys->msg_name &&
2084 used_address->name_len == msg_sys->msg_namelen &&
2085 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002086 used_address->name_len)) {
2087 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2088 goto out_freectl;
2089 }
2090 err = sock_sendmsg(sock, msg_sys, total_len);
2091 /*
2092 * If this is sendmmsg() and sending to current destination address was
2093 * successful, remember it.
2094 */
2095 if (used_address && err >= 0) {
2096 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002097 if (msg_sys->msg_name)
2098 memcpy(&used_address->name, msg_sys->msg_name,
2099 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101
2102out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002103 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2105out_freeiov:
2106 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002107 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002108 return err;
2109}
2110
2111/*
2112 * BSD sendmsg interface
2113 */
2114
Al Viro666547f2014-04-06 14:03:05 -04002115long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002116{
2117 int fput_needed, err;
2118 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002119 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002120
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002121 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002122 if (!sock)
2123 goto out;
2124
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002125 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002126
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002127 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002128out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129 return err;
2130}
2131
Al Viro666547f2014-04-06 14:03:05 -04002132SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002133{
2134 if (flags & MSG_CMSG_COMPAT)
2135 return -EINVAL;
2136 return __sys_sendmsg(fd, msg, flags);
2137}
2138
Anton Blanchard228e5482011-05-02 20:21:35 +00002139/*
2140 * Linux sendmmsg interface
2141 */
2142
2143int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2144 unsigned int flags)
2145{
2146 int fput_needed, err, datagrams;
2147 struct socket *sock;
2148 struct mmsghdr __user *entry;
2149 struct compat_mmsghdr __user *compat_entry;
2150 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002151 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002152
Anton Blanchard98382f42011-08-04 14:07:39 +00002153 if (vlen > UIO_MAXIOV)
2154 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002155
2156 datagrams = 0;
2157
2158 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2159 if (!sock)
2160 return err;
2161
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002162 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002163 entry = mmsg;
2164 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002165 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002166
2167 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002168 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002169 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002170 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002171 if (err < 0)
2172 break;
2173 err = __put_user(err, &compat_entry->msg_len);
2174 ++compat_entry;
2175 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002176 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002177 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002178 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002179 if (err < 0)
2180 break;
2181 err = put_user(err, &entry->msg_len);
2182 ++entry;
2183 }
2184
2185 if (err)
2186 break;
2187 ++datagrams;
2188 }
2189
Anton Blanchard228e5482011-05-02 20:21:35 +00002190 fput_light(sock->file, fput_needed);
2191
Anton Blanchard728ffb82011-08-04 14:07:38 +00002192 /* We only return an error if no datagrams were able to be sent */
2193 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002194 return datagrams;
2195
Anton Blanchard228e5482011-05-02 20:21:35 +00002196 return err;
2197}
2198
2199SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2200 unsigned int, vlen, unsigned int, flags)
2201{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002202 if (flags & MSG_CMSG_COMPAT)
2203 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002204 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2205}
2206
Al Viro666547f2014-04-06 14:03:05 -04002207static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002208 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002210 struct compat_msghdr __user *msg_compat =
2211 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002213 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 unsigned long cmsg_ptr;
Al Viro08adb7d2014-11-10 20:23:13 -05002215 int total_len, len;
2216 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217
2218 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002219 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220
2221 /* user mode address pointers */
2222 struct sockaddr __user *uaddr;
Al Viro08adb7d2014-11-10 20:23:13 -05002223 int __user *uaddr_len = COMPAT_NAMELEN(msg);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002224
Al Viro08adb7d2014-11-10 20:23:13 -05002225 msg_sys->msg_name = &addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002227 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002228 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002229 else
Al Viro08adb7d2014-11-10 20:23:13 -05002230 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 if (err < 0)
2232 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002233 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002235 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2236 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002237
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002238 /* We assume all kernel code knows the size of sockaddr_storage */
2239 msg_sys->msg_namelen = 0;
2240
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 if (sock->file->f_flags & O_NONBLOCK)
2242 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002243 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2244 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245 if (err < 0)
2246 goto out_freeiov;
2247 len = err;
2248
2249 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002250 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002251 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002252 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 if (err < 0)
2254 goto out_freeiov;
2255 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002256 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002257 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258 if (err)
2259 goto out_freeiov;
2260 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002261 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262 &msg_compat->msg_controllen);
2263 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002264 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265 &msg->msg_controllen);
2266 if (err)
2267 goto out_freeiov;
2268 err = len;
2269
2270out_freeiov:
2271 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002272 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002273 return err;
2274}
2275
2276/*
2277 * BSD recvmsg interface
2278 */
2279
Al Viro666547f2014-04-06 14:03:05 -04002280long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002281{
2282 int fput_needed, err;
2283 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002284 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002285
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002286 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002287 if (!sock)
2288 goto out;
2289
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002290 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002291
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002292 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293out:
2294 return err;
2295}
2296
Al Viro666547f2014-04-06 14:03:05 -04002297SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002298 unsigned int, flags)
2299{
2300 if (flags & MSG_CMSG_COMPAT)
2301 return -EINVAL;
2302 return __sys_recvmsg(fd, msg, flags);
2303}
2304
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002305/*
2306 * Linux recvmmsg interface
2307 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002309int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2310 unsigned int flags, struct timespec *timeout)
2311{
2312 int fput_needed, err, datagrams;
2313 struct socket *sock;
2314 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002315 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002316 struct msghdr msg_sys;
2317 struct timespec end_time;
2318
2319 if (timeout &&
2320 poll_select_set_timeout(&end_time, timeout->tv_sec,
2321 timeout->tv_nsec))
2322 return -EINVAL;
2323
2324 datagrams = 0;
2325
2326 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2327 if (!sock)
2328 return err;
2329
2330 err = sock_error(sock->sk);
2331 if (err)
2332 goto out_put;
2333
2334 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002335 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002336
2337 while (datagrams < vlen) {
2338 /*
2339 * No need to ask LSM for more than the first datagram.
2340 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002341 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002342 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002343 &msg_sys, flags & ~MSG_WAITFORONE,
2344 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002345 if (err < 0)
2346 break;
2347 err = __put_user(err, &compat_entry->msg_len);
2348 ++compat_entry;
2349 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002350 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002351 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002352 &msg_sys, flags & ~MSG_WAITFORONE,
2353 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002354 if (err < 0)
2355 break;
2356 err = put_user(err, &entry->msg_len);
2357 ++entry;
2358 }
2359
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002360 if (err)
2361 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002362 ++datagrams;
2363
Brandon L Black71c5c152010-03-26 16:18:03 +00002364 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2365 if (flags & MSG_WAITFORONE)
2366 flags |= MSG_DONTWAIT;
2367
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002368 if (timeout) {
2369 ktime_get_ts(timeout);
2370 *timeout = timespec_sub(end_time, *timeout);
2371 if (timeout->tv_sec < 0) {
2372 timeout->tv_sec = timeout->tv_nsec = 0;
2373 break;
2374 }
2375
2376 /* Timeout, return less than vlen datagrams */
2377 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2378 break;
2379 }
2380
2381 /* Out of band data, return right away */
2382 if (msg_sys.msg_flags & MSG_OOB)
2383 break;
2384 }
2385
2386out_put:
2387 fput_light(sock->file, fput_needed);
2388
2389 if (err == 0)
2390 return datagrams;
2391
2392 if (datagrams != 0) {
2393 /*
2394 * We may return less entries than requested (vlen) if the
2395 * sock is non block and there aren't enough datagrams...
2396 */
2397 if (err != -EAGAIN) {
2398 /*
2399 * ... or if recvmsg returns an error after we
2400 * received some datagrams, where we record the
2401 * error to return on the next call or if the
2402 * app asks about it using getsockopt(SO_ERROR).
2403 */
2404 sock->sk->sk_err = -err;
2405 }
2406
2407 return datagrams;
2408 }
2409
2410 return err;
2411}
2412
2413SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2414 unsigned int, vlen, unsigned int, flags,
2415 struct timespec __user *, timeout)
2416{
2417 int datagrams;
2418 struct timespec timeout_sys;
2419
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002420 if (flags & MSG_CMSG_COMPAT)
2421 return -EINVAL;
2422
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002423 if (!timeout)
2424 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2425
2426 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2427 return -EFAULT;
2428
2429 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2430
2431 if (datagrams > 0 &&
2432 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2433 datagrams = -EFAULT;
2434
2435 return datagrams;
2436}
2437
2438#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439/* Argument list sizes for sys_socketcall */
2440#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002441static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002442 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2443 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2444 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002445 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002446};
2447
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448#undef AL
2449
2450/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002451 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 *
2453 * Argument checking cleaned up. Saved 20% in size.
2454 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002455 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002456 */
2457
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002458SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459{
Chen Gang2950fa92013-04-07 16:55:23 +08002460 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002461 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002463 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002464
Anton Blanchard228e5482011-05-02 20:21:35 +00002465 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466 return -EINVAL;
2467
Arjan van de Ven47379052009-09-28 12:57:44 -07002468 len = nargs[call];
2469 if (len > sizeof(a))
2470 return -EINVAL;
2471
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002473 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002475
Chen Gang2950fa92013-04-07 16:55:23 +08002476 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2477 if (err)
2478 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002479
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002480 a0 = a[0];
2481 a1 = a[1];
2482
2483 switch (call) {
2484 case SYS_SOCKET:
2485 err = sys_socket(a0, a1, a[2]);
2486 break;
2487 case SYS_BIND:
2488 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2489 break;
2490 case SYS_CONNECT:
2491 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2492 break;
2493 case SYS_LISTEN:
2494 err = sys_listen(a0, a1);
2495 break;
2496 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002497 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2498 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002499 break;
2500 case SYS_GETSOCKNAME:
2501 err =
2502 sys_getsockname(a0, (struct sockaddr __user *)a1,
2503 (int __user *)a[2]);
2504 break;
2505 case SYS_GETPEERNAME:
2506 err =
2507 sys_getpeername(a0, (struct sockaddr __user *)a1,
2508 (int __user *)a[2]);
2509 break;
2510 case SYS_SOCKETPAIR:
2511 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2512 break;
2513 case SYS_SEND:
2514 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2515 break;
2516 case SYS_SENDTO:
2517 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2518 (struct sockaddr __user *)a[4], a[5]);
2519 break;
2520 case SYS_RECV:
2521 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2522 break;
2523 case SYS_RECVFROM:
2524 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2525 (struct sockaddr __user *)a[4],
2526 (int __user *)a[5]);
2527 break;
2528 case SYS_SHUTDOWN:
2529 err = sys_shutdown(a0, a1);
2530 break;
2531 case SYS_SETSOCKOPT:
2532 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2533 break;
2534 case SYS_GETSOCKOPT:
2535 err =
2536 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2537 (int __user *)a[4]);
2538 break;
2539 case SYS_SENDMSG:
Al Viro666547f2014-04-06 14:03:05 -04002540 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002541 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002542 case SYS_SENDMMSG:
2543 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2544 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002545 case SYS_RECVMSG:
Al Viro666547f2014-04-06 14:03:05 -04002546 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002547 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002548 case SYS_RECVMMSG:
2549 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2550 (struct timespec __user *)a[4]);
2551 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002552 case SYS_ACCEPT4:
2553 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2554 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002555 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002556 default:
2557 err = -EINVAL;
2558 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559 }
2560 return err;
2561}
2562
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002563#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002564
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002565/**
2566 * sock_register - add a socket protocol handler
2567 * @ops: description of protocol
2568 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002569 * This function is called by a protocol handler that wants to
2570 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002571 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002572 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002574int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575{
2576 int err;
2577
2578 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002579 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 return -ENOBUFS;
2581 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002582
2583 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002584 if (rcu_dereference_protected(net_families[ops->family],
2585 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002586 err = -EEXIST;
2587 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002588 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002589 err = 0;
2590 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002591 spin_unlock(&net_family_lock);
2592
Yang Yingliang3410f222014-02-12 17:09:55 +08002593 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002594 return err;
2595}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002596EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002598/**
2599 * sock_unregister - remove a protocol handler
2600 * @family: protocol family to remove
2601 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002602 * This function is called by a protocol handler that wants to
2603 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002604 * new socket creation.
2605 *
2606 * If protocol handler is a module, then it can use module reference
2607 * counts to protect against new references. If protocol handler is not
2608 * a module then it needs to provide its own protection in
2609 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002611void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002613 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002614
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002615 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002616 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002617 spin_unlock(&net_family_lock);
2618
2619 synchronize_rcu();
2620
Yang Yingliang3410f222014-02-12 17:09:55 +08002621 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002623EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624
Andi Kleen77d76ea2005-12-22 12:43:42 -08002625static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002626{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002627 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002628 /*
2629 * Initialize the network sysctl infrastructure.
2630 */
2631 err = net_sysctl_init();
2632 if (err)
2633 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002634
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002636 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002637 */
2638 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002639
2640 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002641 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642 */
2643
2644 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002645
2646 err = register_filesystem(&sock_fs_type);
2647 if (err)
2648 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002649 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002650 if (IS_ERR(sock_mnt)) {
2651 err = PTR_ERR(sock_mnt);
2652 goto out_mount;
2653 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002654
2655 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002656 */
2657
2658#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002659 err = netfilter_init();
2660 if (err)
2661 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002663
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002664 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002665
Nick Pigginb3e19d92011-01-07 17:50:11 +11002666out:
2667 return err;
2668
2669out_mount:
2670 unregister_filesystem(&sock_fs_type);
2671out_fs:
2672 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002673}
2674
Andi Kleen77d76ea2005-12-22 12:43:42 -08002675core_initcall(sock_init); /* early initcall */
2676
Linus Torvalds1da177e2005-04-16 15:20:36 -07002677#ifdef CONFIG_PROC_FS
2678void socket_seq_show(struct seq_file *seq)
2679{
2680 int cpu;
2681 int counter = 0;
2682
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002683 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002684 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002685
2686 /* It can be negative, by the way. 8) */
2687 if (counter < 0)
2688 counter = 0;
2689
2690 seq_printf(seq, "sockets: used %d\n", counter);
2691}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002692#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002693
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002694#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002695static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002696 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002697{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002698 mm_segment_t old_fs = get_fs();
2699 struct timeval ktv;
2700 int err;
2701
2702 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002703 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002704 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002705 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002706 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002707
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002708 return err;
2709}
2710
Arnd Bergmann6b960182009-11-06 23:10:54 -08002711static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002712 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002713{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002714 mm_segment_t old_fs = get_fs();
2715 struct timespec kts;
2716 int err;
2717
2718 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002719 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002720 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002721 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002722 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002723
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002724 return err;
2725}
2726
Arnd Bergmann6b960182009-11-06 23:10:54 -08002727static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002728{
2729 struct ifreq __user *uifr;
2730 int err;
2731
2732 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002733 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002734 return -EFAULT;
2735
Arnd Bergmann6b960182009-11-06 23:10:54 -08002736 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002737 if (err)
2738 return err;
2739
Arnd Bergmann6b960182009-11-06 23:10:54 -08002740 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002741 return -EFAULT;
2742
2743 return 0;
2744}
2745
Arnd Bergmann6b960182009-11-06 23:10:54 -08002746static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002747{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002748 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002749 struct ifconf ifc;
2750 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002751 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002752 struct ifreq __user *ifr;
2753 unsigned int i, j;
2754 int err;
2755
Arnd Bergmann6b960182009-11-06 23:10:54 -08002756 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002757 return -EFAULT;
2758
Mathias Krause43da5f22012-08-15 11:31:57 +00002759 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002760 if (ifc32.ifcbuf == 0) {
2761 ifc32.ifc_len = 0;
2762 ifc.ifc_len = 0;
2763 ifc.ifc_req = NULL;
2764 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2765 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002766 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2767 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002768 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2769 ifc.ifc_len = len;
2770 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2771 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002772 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002773 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002774 return -EFAULT;
2775 ifr++;
2776 ifr32++;
2777 }
2778 }
2779 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2780 return -EFAULT;
2781
Arnd Bergmann6b960182009-11-06 23:10:54 -08002782 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002783 if (err)
2784 return err;
2785
2786 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2787 return -EFAULT;
2788
2789 ifr = ifc.ifc_req;
2790 ifr32 = compat_ptr(ifc32.ifcbuf);
2791 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002792 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2793 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2794 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002795 return -EFAULT;
2796 ifr32++;
2797 ifr++;
2798 }
2799
2800 if (ifc32.ifcbuf == 0) {
2801 /* Translate from 64-bit structure multiple to
2802 * a 32-bit one.
2803 */
2804 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002805 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002806 ifc32.ifc_len = i;
2807 } else {
2808 ifc32.ifc_len = i;
2809 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002810 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002811 return -EFAULT;
2812
2813 return 0;
2814}
2815
Arnd Bergmann6b960182009-11-06 23:10:54 -08002816static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002817{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002818 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2819 bool convert_in = false, convert_out = false;
2820 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2821 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002822 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002823 u32 rule_cnt = 0, actual_rule_cnt;
2824 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002825 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002826 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002827
2828 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2829 return -EFAULT;
2830
Ben Hutchings3a7da392011-03-17 07:34:32 +00002831 compat_rxnfc = compat_ptr(data);
2832
2833 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002834 return -EFAULT;
2835
Ben Hutchings3a7da392011-03-17 07:34:32 +00002836 /* Most ethtool structures are defined without padding.
2837 * Unfortunately struct ethtool_rxnfc is an exception.
2838 */
2839 switch (ethcmd) {
2840 default:
2841 break;
2842 case ETHTOOL_GRXCLSRLALL:
2843 /* Buffer size is variable */
2844 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2845 return -EFAULT;
2846 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2847 return -ENOMEM;
2848 buf_size += rule_cnt * sizeof(u32);
2849 /* fall through */
2850 case ETHTOOL_GRXRINGS:
2851 case ETHTOOL_GRXCLSRLCNT:
2852 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002853 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002854 convert_out = true;
2855 /* fall through */
2856 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002857 buf_size += sizeof(struct ethtool_rxnfc);
2858 convert_in = true;
2859 break;
2860 }
2861
2862 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002863 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002864
2865 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2866 return -EFAULT;
2867
2868 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2869 &ifr->ifr_ifru.ifru_data))
2870 return -EFAULT;
2871
2872 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002873 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002874 * fs.ring_cookie and at the end of fs, but nowhere else.
2875 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002876 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2877 sizeof(compat_rxnfc->fs.m_ext) !=
2878 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2879 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002880 BUILD_BUG_ON(
2881 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2882 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2883 offsetof(struct ethtool_rxnfc, fs.location) -
2884 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2885
2886 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002887 (void __user *)(&rxnfc->fs.m_ext + 1) -
2888 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002889 copy_in_user(&rxnfc->fs.ring_cookie,
2890 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002891 (void __user *)(&rxnfc->fs.location + 1) -
2892 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002893 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2894 sizeof(rxnfc->rule_cnt)))
2895 return -EFAULT;
2896 }
2897
2898 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2899 if (ret)
2900 return ret;
2901
2902 if (convert_out) {
2903 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002904 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2905 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002906 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2907 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002908 (const void __user *)(&rxnfc->fs.location + 1) -
2909 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002910 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2911 sizeof(rxnfc->rule_cnt)))
2912 return -EFAULT;
2913
2914 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2915 /* As an optimisation, we only copy the actual
2916 * number of rules that the underlying
2917 * function returned. Since Mallory might
2918 * change the rule count in user memory, we
2919 * check that it is less than the rule count
2920 * originally given (as the user buffer size),
2921 * which has been range-checked.
2922 */
2923 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2924 return -EFAULT;
2925 if (actual_rule_cnt < rule_cnt)
2926 rule_cnt = actual_rule_cnt;
2927 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2928 &rxnfc->rule_locs[0],
2929 rule_cnt * sizeof(u32)))
2930 return -EFAULT;
2931 }
2932 }
2933
2934 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002935}
2936
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002937static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2938{
2939 void __user *uptr;
2940 compat_uptr_t uptr32;
2941 struct ifreq __user *uifr;
2942
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002943 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002944 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2945 return -EFAULT;
2946
2947 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2948 return -EFAULT;
2949
2950 uptr = compat_ptr(uptr32);
2951
2952 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2953 return -EFAULT;
2954
2955 return dev_ioctl(net, SIOCWANDEV, uifr);
2956}
2957
Arnd Bergmann6b960182009-11-06 23:10:54 -08002958static int bond_ioctl(struct net *net, unsigned int cmd,
2959 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002960{
2961 struct ifreq kifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002962 mm_segment_t old_fs;
2963 int err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002964
2965 switch (cmd) {
2966 case SIOCBONDENSLAVE:
2967 case SIOCBONDRELEASE:
2968 case SIOCBONDSETHWADDR:
2969 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002970 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002971 return -EFAULT;
2972
2973 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002974 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002975 err = dev_ioctl(net, cmd,
2976 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002977 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002978
2979 return err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002980 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002981 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002982 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002983}
2984
Ben Hutchings590d4692013-11-18 17:04:13 +00002985/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2986static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08002987 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002988{
2989 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002990 char tmp_buf[IFNAMSIZ];
2991 void __user *data64;
2992 u32 data32;
2993
2994 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2995 IFNAMSIZ))
2996 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002997 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002998 return -EFAULT;
2999 data64 = compat_ptr(data32);
3000
3001 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3002
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003003 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3004 IFNAMSIZ))
3005 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00003006 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003007 return -EFAULT;
3008
Arnd Bergmann6b960182009-11-06 23:10:54 -08003009 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003010}
3011
Arnd Bergmann6b960182009-11-06 23:10:54 -08003012static int dev_ifsioc(struct net *net, struct socket *sock,
3013 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003014{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003015 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003016 int err;
3017
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003018 uifr = compat_alloc_user_space(sizeof(*uifr));
3019 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3020 return -EFAULT;
3021
3022 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3023
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003024 if (!err) {
3025 switch (cmd) {
3026 case SIOCGIFFLAGS:
3027 case SIOCGIFMETRIC:
3028 case SIOCGIFMTU:
3029 case SIOCGIFMEM:
3030 case SIOCGIFHWADDR:
3031 case SIOCGIFINDEX:
3032 case SIOCGIFADDR:
3033 case SIOCGIFBRDADDR:
3034 case SIOCGIFDSTADDR:
3035 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003036 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003037 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003038 case SIOCGMIIPHY:
3039 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003040 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003041 err = -EFAULT;
3042 break;
3043 }
3044 }
3045 return err;
3046}
3047
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003048static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3049 struct compat_ifreq __user *uifr32)
3050{
3051 struct ifreq ifr;
3052 struct compat_ifmap __user *uifmap32;
3053 mm_segment_t old_fs;
3054 int err;
3055
3056 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3057 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003058 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3059 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3060 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3061 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3062 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3063 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003064 if (err)
3065 return -EFAULT;
3066
3067 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003068 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003069 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003070 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003071
3072 if (cmd == SIOCGIFMAP && !err) {
3073 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003074 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3075 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3076 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3077 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3078 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3079 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003080 if (err)
3081 err = -EFAULT;
3082 }
3083 return err;
3084}
3085
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003086struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003087 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003088 struct sockaddr rt_dst; /* target address */
3089 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3090 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003091 unsigned short rt_flags;
3092 short rt_pad2;
3093 u32 rt_pad3;
3094 unsigned char rt_tos;
3095 unsigned char rt_class;
3096 short rt_pad4;
3097 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003098 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003099 u32 rt_mtu; /* per route MTU/Window */
3100 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003101 unsigned short rt_irtt; /* Initial RTT */
3102};
3103
3104struct in6_rtmsg32 {
3105 struct in6_addr rtmsg_dst;
3106 struct in6_addr rtmsg_src;
3107 struct in6_addr rtmsg_gateway;
3108 u32 rtmsg_type;
3109 u16 rtmsg_dst_len;
3110 u16 rtmsg_src_len;
3111 u32 rtmsg_metric;
3112 u32 rtmsg_info;
3113 u32 rtmsg_flags;
3114 s32 rtmsg_ifindex;
3115};
3116
Arnd Bergmann6b960182009-11-06 23:10:54 -08003117static int routing_ioctl(struct net *net, struct socket *sock,
3118 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003119{
3120 int ret;
3121 void *r = NULL;
3122 struct in6_rtmsg r6;
3123 struct rtentry r4;
3124 char devname[16];
3125 u32 rtdev;
3126 mm_segment_t old_fs = get_fs();
3127
Arnd Bergmann6b960182009-11-06 23:10:54 -08003128 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3129 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003130 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003131 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003132 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3133 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3134 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3135 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3136 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3137 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3138 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003139
3140 r = (void *) &r6;
3141 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003142 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003143 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003144 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003145 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3146 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3147 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3148 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3149 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3150 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003151 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003152 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003153 r4.rt_dev = (char __user __force *)devname;
3154 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003155 } else
3156 r4.rt_dev = NULL;
3157
3158 r = (void *) &r4;
3159 }
3160
3161 if (ret) {
3162 ret = -EFAULT;
3163 goto out;
3164 }
3165
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003166 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003167 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003168 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003169
3170out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003171 return ret;
3172}
3173
3174/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3175 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003176 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003177 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003178static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003179{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003180 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003181
Arnd Bergmann6b960182009-11-06 23:10:54 -08003182 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003183 return -EFAULT;
3184 if (tmp == BRCTL_GET_VERSION)
3185 return BRCTL_VERSION + 1;
3186 return -EINVAL;
3187}
3188
Arnd Bergmann6b960182009-11-06 23:10:54 -08003189static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3190 unsigned int cmd, unsigned long arg)
3191{
3192 void __user *argp = compat_ptr(arg);
3193 struct sock *sk = sock->sk;
3194 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003195
Arnd Bergmann6b960182009-11-06 23:10:54 -08003196 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003197 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003198
Arnd Bergmann6b960182009-11-06 23:10:54 -08003199 switch (cmd) {
3200 case SIOCSIFBR:
3201 case SIOCGIFBR:
3202 return old_bridge_ioctl(argp);
3203 case SIOCGIFNAME:
3204 return dev_ifname32(net, argp);
3205 case SIOCGIFCONF:
3206 return dev_ifconf(net, argp);
3207 case SIOCETHTOOL:
3208 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003209 case SIOCWANDEV:
3210 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003211 case SIOCGIFMAP:
3212 case SIOCSIFMAP:
3213 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003214 case SIOCBONDENSLAVE:
3215 case SIOCBONDRELEASE:
3216 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003217 case SIOCBONDCHANGEACTIVE:
3218 return bond_ioctl(net, cmd, argp);
3219 case SIOCADDRT:
3220 case SIOCDELRT:
3221 return routing_ioctl(net, sock, cmd, argp);
3222 case SIOCGSTAMP:
3223 return do_siocgstamp(net, sock, cmd, argp);
3224 case SIOCGSTAMPNS:
3225 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003226 case SIOCBONDSLAVEINFOQUERY:
3227 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003228 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003229 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003230 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003231
Arnd Bergmann6b960182009-11-06 23:10:54 -08003232 case FIOSETOWN:
3233 case SIOCSPGRP:
3234 case FIOGETOWN:
3235 case SIOCGPGRP:
3236 case SIOCBRADDBR:
3237 case SIOCBRDELBR:
3238 case SIOCGIFVLAN:
3239 case SIOCSIFVLAN:
3240 case SIOCADDDLCI:
3241 case SIOCDELDLCI:
3242 return sock_ioctl(file, cmd, arg);
3243
3244 case SIOCGIFFLAGS:
3245 case SIOCSIFFLAGS:
3246 case SIOCGIFMETRIC:
3247 case SIOCSIFMETRIC:
3248 case SIOCGIFMTU:
3249 case SIOCSIFMTU:
3250 case SIOCGIFMEM:
3251 case SIOCSIFMEM:
3252 case SIOCGIFHWADDR:
3253 case SIOCSIFHWADDR:
3254 case SIOCADDMULTI:
3255 case SIOCDELMULTI:
3256 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003257 case SIOCGIFADDR:
3258 case SIOCSIFADDR:
3259 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003260 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003261 case SIOCGIFBRDADDR:
3262 case SIOCSIFBRDADDR:
3263 case SIOCGIFDSTADDR:
3264 case SIOCSIFDSTADDR:
3265 case SIOCGIFNETMASK:
3266 case SIOCSIFNETMASK:
3267 case SIOCSIFPFLAGS:
3268 case SIOCGIFPFLAGS:
3269 case SIOCGIFTXQLEN:
3270 case SIOCSIFTXQLEN:
3271 case SIOCBRADDIF:
3272 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003273 case SIOCSIFNAME:
3274 case SIOCGMIIPHY:
3275 case SIOCGMIIREG:
3276 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003277 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003278
Arnd Bergmann6b960182009-11-06 23:10:54 -08003279 case SIOCSARP:
3280 case SIOCGARP:
3281 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003282 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003283 return sock_do_ioctl(net, sock, cmd, arg);
3284 }
3285
Arnd Bergmann6b960182009-11-06 23:10:54 -08003286 return -ENOIOCTLCMD;
3287}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003288
Eric Dumazet95c96172012-04-15 05:58:06 +00003289static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003290 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003291{
3292 struct socket *sock = file->private_data;
3293 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003294 struct sock *sk;
3295 struct net *net;
3296
3297 sk = sock->sk;
3298 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003299
3300 if (sock->ops->compat_ioctl)
3301 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3302
David S. Miller87de87d2008-06-03 09:14:03 -07003303 if (ret == -ENOIOCTLCMD &&
3304 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3305 ret = compat_wext_handle_ioctl(net, cmd, arg);
3306
Arnd Bergmann6b960182009-11-06 23:10:54 -08003307 if (ret == -ENOIOCTLCMD)
3308 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3309
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003310 return ret;
3311}
3312#endif
3313
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003314int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3315{
3316 return sock->ops->bind(sock, addr, addrlen);
3317}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003318EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003319
3320int kernel_listen(struct socket *sock, int backlog)
3321{
3322 return sock->ops->listen(sock, backlog);
3323}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003324EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003325
3326int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3327{
3328 struct sock *sk = sock->sk;
3329 int err;
3330
3331 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3332 newsock);
3333 if (err < 0)
3334 goto done;
3335
3336 err = sock->ops->accept(sock, *newsock, flags);
3337 if (err < 0) {
3338 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003339 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003340 goto done;
3341 }
3342
3343 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003344 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003345
3346done:
3347 return err;
3348}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003349EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003350
3351int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003352 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003353{
3354 return sock->ops->connect(sock, addr, addrlen, flags);
3355}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003356EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003357
3358int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3359 int *addrlen)
3360{
3361 return sock->ops->getname(sock, addr, addrlen, 0);
3362}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003363EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003364
3365int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3366 int *addrlen)
3367{
3368 return sock->ops->getname(sock, addr, addrlen, 1);
3369}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003370EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003371
3372int kernel_getsockopt(struct socket *sock, int level, int optname,
3373 char *optval, int *optlen)
3374{
3375 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003376 char __user *uoptval;
3377 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003378 int err;
3379
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003380 uoptval = (char __user __force *) optval;
3381 uoptlen = (int __user __force *) optlen;
3382
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003383 set_fs(KERNEL_DS);
3384 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003385 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003386 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003387 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3388 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003389 set_fs(oldfs);
3390 return err;
3391}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003392EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003393
3394int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003395 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003396{
3397 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003398 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003399 int err;
3400
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003401 uoptval = (char __user __force *) optval;
3402
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003403 set_fs(KERNEL_DS);
3404 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003405 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003406 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003407 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003408 optlen);
3409 set_fs(oldfs);
3410 return err;
3411}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003412EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003413
3414int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3415 size_t size, int flags)
3416{
3417 if (sock->ops->sendpage)
3418 return sock->ops->sendpage(sock, page, offset, size, flags);
3419
3420 return sock_no_sendpage(sock, page, offset, size, flags);
3421}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003422EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003423
3424int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3425{
3426 mm_segment_t oldfs = get_fs();
3427 int err;
3428
3429 set_fs(KERNEL_DS);
3430 err = sock->ops->ioctl(sock, cmd, arg);
3431 set_fs(oldfs);
3432
3433 return err;
3434}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003435EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003436
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003437int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3438{
3439 return sock->ops->shutdown(sock, how);
3440}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003441EXPORT_SYMBOL(kernel_sock_shutdown);