blob: ae89569a2db5fe24cafec01570acef55836492db [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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
Badari Pulavarty027445c2006-09-30 23:28:46 -0700117static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
118 unsigned long nr_segs, loff_t pos);
119static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
120 unsigned long nr_segs, loff_t pos);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700121static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
123static int sock_close(struct inode *inode, struct file *file);
124static unsigned int sock_poll(struct file *file,
125 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700126static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800127#ifdef CONFIG_COMPAT
128static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700129 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800130#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132static ssize_t sock_sendpage(struct file *file, struct page *page,
133 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800134static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700135 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800136 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138/*
139 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
140 * in the operation structures but are done directly via the socketcall() multiplexor.
141 */
142
Arjan van de Venda7071d2007-02-12 00:55:36 -0800143static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 .owner = THIS_MODULE,
145 .llseek = no_llseek,
146 .aio_read = sock_aio_read,
147 .aio_write = sock_aio_write,
148 .poll = sock_poll,
149 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800150#ifdef CONFIG_COMPAT
151 .compat_ioctl = compat_sock_ioctl,
152#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 .mmap = sock_mmap,
154 .open = sock_no_open, /* special open code to disallow open via /proc */
155 .release = sock_close,
156 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200157 .sendpage = sock_sendpage,
158 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800159 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160};
161
162/*
163 * The protocol list. Each protocol is registered in here.
164 */
165
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000167static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169/*
170 * Statistics counters of the socket lists
171 */
172
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700173static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
175/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700176 * Support routines.
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 */
180
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181/**
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
186 *
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 */
191
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000192int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700196 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700198 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100200 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201}
202
203/**
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
209 *
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
214 * accessible.
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
218 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700219
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000220static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000221 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
223 int err;
224 int len;
225
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100226 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700227 err = get_user(len, ulen);
228 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700230 if (len > klen)
231 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100232 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700234 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500235 if (audit_sockaddr(klen, kaddr))
236 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700237 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 return -EFAULT;
239 }
240 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700241 * "fromlen shall refer to the value before truncation.."
242 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 */
244 return __put_user(klen, ulen);
245}
246
Christoph Lametere18b8902006-12-06 20:33:20 -0800247static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
249static struct inode *sock_alloc_inode(struct super_block *sb)
250{
251 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000252 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700253
Christoph Lametere94b1762006-12-06 20:33:17 -0800254 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255 if (!ei)
256 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000257 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
258 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000259 kmem_cache_free(sock_inode_cachep, ei);
260 return NULL;
261 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000262 init_waitqueue_head(&wq->wait);
263 wq->fasync_list = NULL;
264 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700265
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 ei->socket.state = SS_UNCONNECTED;
267 ei->socket.flags = 0;
268 ei->socket.ops = NULL;
269 ei->socket.sk = NULL;
270 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
272 return &ei->vfs_inode;
273}
274
275static void sock_destroy_inode(struct inode *inode)
276{
Eric Dumazet43815482010-04-29 11:01:49 +0000277 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000278 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000279
280 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000281 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800282 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000283 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284}
285
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700286static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700288 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
Christoph Lametera35afb82007-05-16 22:10:57 -0700290 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700292
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293static int init_inodecache(void)
294{
295 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700296 sizeof(struct socket_alloc),
297 0,
298 (SLAB_HWCACHE_ALIGN |
299 SLAB_RECLAIM_ACCOUNT |
300 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900301 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 if (sock_inode_cachep == NULL)
303 return -ENOMEM;
304 return 0;
305}
306
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700307static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700308 .alloc_inode = sock_alloc_inode,
309 .destroy_inode = sock_destroy_inode,
310 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311};
312
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700313/*
314 * sockfs_dname() is called from d_path().
315 */
316static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
317{
318 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
319 dentry->d_inode->i_ino);
320}
321
Al Viro3ba13d12009-02-20 06:02:22 +0000322static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700323 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324};
325
Al Viroc74a1cb2011-01-12 16:59:34 -0500326static struct dentry *sockfs_mount(struct file_system_type *fs_type,
327 int flags, const char *dev_name, void *data)
328{
329 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
330 &sockfs_dentry_operations, SOCKFS_MAGIC);
331}
332
333static struct vfsmount *sock_mnt __read_mostly;
334
335static struct file_system_type sock_fs_type = {
336 .name = "sockfs",
337 .mount = sockfs_mount,
338 .kill_sb = kill_anon_super,
339};
340
Linus Torvalds1da177e2005-04-16 15:20:36 -0700341/*
342 * Obtains the first available file descriptor and sets it up for use.
343 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800344 * These functions create file structures and maps them to fd space
345 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * and file struct implicitly stored in sock->file.
347 * Note that another thread may close file descriptor before we return
348 * from this function. We use the fact that now we do not refer
349 * to socket after mapping. If one day we will need it, this
350 * function will increment ref. count on file by 1.
351 *
352 * In any case returned fd MAY BE not valid!
353 * This race condition is unavoidable
354 * with shared fd spaces, we cannot solve it inside kernel,
355 * but we take care of internal coherence yet.
356 */
357
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700358struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359{
Al Viro7cbe66b2009-08-05 19:59:08 +0400360 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400361 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400362 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800363
Masatake YAMATO600e1772012-08-29 10:44:29 +0000364 if (dname) {
365 name.name = dname;
366 name.len = strlen(name.name);
367 } else if (sock->sk) {
368 name.name = sock->sk->sk_prot_creator->name;
369 name.len = strlen(name.name);
370 }
Nick Piggin4b936882011-01-07 17:50:07 +1100371 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400372 if (unlikely(!path.dentry))
373 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400374 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375
Al Viro2c48b9c2009-08-09 00:52:35 +0400376 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400377 SOCK_INODE(sock)->i_fop = &socket_file_ops;
378
Al Viro2c48b9c2009-08-09 00:52:35 +0400379 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400380 &socket_file_ops);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700381 if (unlikely(IS_ERR(file))) {
Al Virocc3808f2009-08-06 09:43:59 +0400382 /* drop dentry, keep inode */
Al Viro7de9c6ee2010-10-23 11:11:40 -0400383 ihold(path.dentry->d_inode);
Al Viro2c48b9c2009-08-09 00:52:35 +0400384 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700385 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400386 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387
388 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700389 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800390 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400391 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392}
Al Viro56b31d12012-08-18 00:25:51 -0400393EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800394
Al Viro56b31d12012-08-18 00:25:51 -0400395static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800396{
397 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400398 int fd = get_unused_fd_flags(flags);
399 if (unlikely(fd < 0))
400 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800401
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700402 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400403 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800404 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400405 return fd;
406 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400407
Al Viro28407632012-08-17 23:54:15 -0400408 put_unused_fd(fd);
409 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410}
411
John Fastabend406a3c62012-07-20 10:39:25 +0000412struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800413{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800414 if (file->f_op == &socket_file_ops)
415 return file->private_data; /* set in sock_map_fd */
416
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800417 *err = -ENOTSOCK;
418 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800419}
John Fastabend406a3c62012-07-20 10:39:25 +0000420EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800421
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700423 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 * @fd: file handle
425 * @err: pointer to an error code return
426 *
427 * The file handle passed in is locked and the socket it is bound
428 * too is returned. If an error occurs the err pointer is overwritten
429 * with a negative errno code and NULL is returned. The function checks
430 * for both invalid handles and passing a handle which is not a socket.
431 *
432 * On a success the socket object pointer is returned.
433 */
434
435struct socket *sockfd_lookup(int fd, int *err)
436{
437 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 struct socket *sock;
439
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700440 file = fget(fd);
441 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 *err = -EBADF;
443 return NULL;
444 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700445
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800446 sock = sock_from_file(file, err);
447 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 return sock;
450}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700451EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800453static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
454{
Al Viro00e188e2014-03-03 23:48:18 -0500455 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800456 struct socket *sock;
457
Hua Zhong36725582006-04-19 15:25:02 -0700458 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500459 if (f.file) {
460 sock = sock_from_file(f.file, err);
461 if (likely(sock)) {
462 *fput_needed = f.flags;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800463 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500464 }
465 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800466 }
467 return NULL;
468}
469
Masatake YAMATO600e1772012-08-29 10:44:29 +0000470#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
471#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
472#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
473static ssize_t sockfs_getxattr(struct dentry *dentry,
474 const char *name, void *value, size_t size)
475{
476 const char *proto_name;
477 size_t proto_size;
478 int error;
479
480 error = -ENODATA;
481 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
482 proto_name = dentry->d_name.name;
483 proto_size = strlen(proto_name);
484
485 if (value) {
486 error = -ERANGE;
487 if (proto_size + 1 > size)
488 goto out;
489
490 strncpy(value, proto_name, proto_size + 1);
491 }
492 error = proto_size + 1;
493 }
494
495out:
496 return error;
497}
498
499static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
500 size_t size)
501{
502 ssize_t len;
503 ssize_t used = 0;
504
505 len = security_inode_listsecurity(dentry->d_inode, buffer, size);
506 if (len < 0)
507 return len;
508 used += len;
509 if (buffer) {
510 if (size < used)
511 return -ERANGE;
512 buffer += len;
513 }
514
515 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
516 used += len;
517 if (buffer) {
518 if (size < used)
519 return -ERANGE;
520 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
521 buffer += len;
522 }
523
524 return used;
525}
526
527static const struct inode_operations sockfs_inode_ops = {
528 .getxattr = sockfs_getxattr,
529 .listxattr = sockfs_listxattr,
530};
531
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532/**
533 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700534 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 * Allocate a new inode and socket object. The two are bound together
536 * and initialised. The socket is then returned. If we are out of inodes
537 * NULL is returned.
538 */
539
540static struct socket *sock_alloc(void)
541{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700542 struct inode *inode;
543 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200545 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 if (!inode)
547 return NULL;
548
549 sock = SOCKET_I(inode);
550
Eric Dumazet29a020d2009-09-15 02:39:20 -0700551 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400552 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700553 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100554 inode->i_uid = current_fsuid();
555 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000556 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557
Alex Shi19e8d692012-05-14 14:15:31 -0700558 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 return sock;
560}
561
562/*
563 * In theory you can't get an open on this inode, but /proc provides
564 * a back door. Remember to keep it shut otherwise you'll let the
565 * creepy crawlies in.
566 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
569{
570 return -ENXIO;
571}
572
Arjan van de Ven4b6f5d22006-03-28 01:56:42 -0800573const struct file_operations bad_sock_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 .owner = THIS_MODULE,
575 .open = sock_no_open,
Arnd Bergmann6038f372010-08-15 18:52:59 +0200576 .llseek = noop_llseek,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577};
578
579/**
580 * sock_release - close a socket
581 * @sock: socket to close
582 *
583 * The socket is released from the protocol stack if it has a release
584 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700585 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588void sock_release(struct socket *sock)
589{
590 if (sock->ops) {
591 struct module *owner = sock->ops->owner;
592
593 sock->ops->release(sock);
594 sock->ops = NULL;
595 module_put(owner);
596 }
597
Eric Dumazeteaefd112011-02-18 03:26:36 +0000598 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800599 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
Mikulas Patockab09e7862012-07-19 06:13:36 +0000601 if (test_bit(SOCK_EXTERNALLY_ALLOCATED, &sock->flags))
602 return;
603
Alex Shi19e8d692012-05-14 14:15:31 -0700604 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 if (!sock->file) {
606 iput(SOCK_INODE(sock));
607 return;
608 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700609 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700611EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
Daniel Borkmannbf84a012013-04-14 08:08:13 +0000613void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000614{
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000615 *tx_flags = 0;
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400616 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000617 *tx_flags |= SKBTX_HW_TSTAMP;
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400618 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000619 *tx_flags |= SKBTX_SW_TSTAMP;
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400620 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
621 *tx_flags |= SKBTX_SCHED_TSTAMP;
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400622 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
623 *tx_flags |= SKBTX_ACK_TSTAMP;
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400624
Johannes Berg6e3e9392011-11-09 10:15:42 +0100625 if (sock_flag(sk, SOCK_WIFI_STATUS))
626 *tx_flags |= SKBTX_WIFI_STATUS;
Patrick Ohly20d49472009-02-12 05:03:38 +0000627}
628EXPORT_SYMBOL(sock_tx_timestamp);
629
Anton Blanchard228e5482011-05-02 20:21:35 +0000630static inline int __sock_sendmsg_nosec(struct kiocb *iocb, struct socket *sock,
631 struct msghdr *msg, size_t size)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632{
633 struct sock_iocb *si = kiocb_to_siocb(iocb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
635 si->sock = sock;
636 si->scm = NULL;
637 si->msg = msg;
638 si->size = size;
639
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 return sock->ops->sendmsg(iocb, sock, msg, size);
641}
642
Anton Blanchard228e5482011-05-02 20:21:35 +0000643static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
644 struct msghdr *msg, size_t size)
645{
646 int err = security_socket_sendmsg(sock, msg, size);
647
648 return err ?: __sock_sendmsg_nosec(iocb, sock, msg, size);
649}
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
652{
653 struct kiocb iocb;
654 struct sock_iocb siocb;
655 int ret;
656
657 init_sync_kiocb(&iocb, NULL);
658 iocb.private = &siocb;
659 ret = __sock_sendmsg(&iocb, sock, msg, size);
660 if (-EIOCBQUEUED == ret)
661 ret = wait_on_sync_kiocb(&iocb);
662 return ret;
663}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700664EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
Eric Dumazet894dc242011-07-26 02:39:41 +0000666static int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg, size_t size)
Anton Blanchard228e5482011-05-02 20:21:35 +0000667{
668 struct kiocb iocb;
669 struct sock_iocb siocb;
670 int ret;
671
672 init_sync_kiocb(&iocb, NULL);
673 iocb.private = &siocb;
674 ret = __sock_sendmsg_nosec(&iocb, sock, msg, size);
675 if (-EIOCBQUEUED == ret)
676 ret = wait_on_sync_kiocb(&iocb);
677 return ret;
678}
679
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
681 struct kvec *vec, size_t num, size_t size)
682{
683 mm_segment_t oldfs = get_fs();
684 int result;
685
686 set_fs(KERNEL_DS);
687 /*
688 * the following is safe, since for compiler definitions of kvec and
689 * iovec are identical, yielding the same in-core layout and alignment
690 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700691 msg->msg_iov = (struct iovec *)vec;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 msg->msg_iovlen = num;
693 result = sock_sendmsg(sock, msg, size);
694 set_fs(oldfs);
695 return result;
696}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700697EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700699/*
700 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
701 */
702void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
703 struct sk_buff *skb)
704{
Patrick Ohly20d49472009-02-12 05:03:38 +0000705 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400706 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000707 int empty = 1;
708 struct skb_shared_hwtstamps *shhwtstamps =
709 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700710
Patrick Ohly20d49472009-02-12 05:03:38 +0000711 /* Race occurred between timestamp enabling and packet
712 receiving. Fill in the current time for now. */
713 if (need_software_tstamp && skb->tstamp.tv64 == 0)
714 __net_timestamp(skb);
715
716 if (need_software_tstamp) {
717 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
718 struct timeval tv;
719 skb_get_timestamp(skb, &tv);
720 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
721 sizeof(tv), &tv);
722 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400723 struct timespec ts;
724 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000725 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400726 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000727 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700728 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000729
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400730 memset(&tss, 0, sizeof(tss));
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400731 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE ||
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400732 skb_shinfo(skb)->tx_flags & SKBTX_ANY_SW_TSTAMP) &&
733 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000734 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400735 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400736 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400737 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400738 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000739 if (!empty)
740 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400741 SCM_TIMESTAMPING, sizeof(tss), &tss);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700742}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300743EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
744
Johannes Berg6e3e9392011-11-09 10:15:42 +0100745void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
746 struct sk_buff *skb)
747{
748 int ack;
749
750 if (!sock_flag(sk, SOCK_WIFI_STATUS))
751 return;
752 if (!skb->wifi_acked_valid)
753 return;
754
755 ack = skb->wifi_acked;
756
757 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
758}
759EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
760
stephen hemminger11165f12010-10-18 14:27:29 +0000761static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
762 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700763{
764 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
765 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
766 sizeof(__u32), &skb->dropcount);
767}
768
Eric Dumazet767dd032010-04-28 19:14:43 +0000769void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700770 struct sk_buff *skb)
771{
772 sock_recv_timestamp(msg, sk, skb);
773 sock_recv_drops(msg, sk, skb);
774}
Eric Dumazet767dd032010-04-28 19:14:43 +0000775EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700776
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700777static inline int __sock_recvmsg_nosec(struct kiocb *iocb, struct socket *sock,
778 struct msghdr *msg, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 struct sock_iocb *si = kiocb_to_siocb(iocb);
781
782 si->sock = sock;
783 si->scm = NULL;
784 si->msg = msg;
785 si->size = size;
786 si->flags = flags;
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
789}
790
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700791static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
792 struct msghdr *msg, size_t size, int flags)
793{
794 int err = security_socket_recvmsg(sock, msg, size, flags);
795
796 return err ?: __sock_recvmsg_nosec(iocb, sock, msg, size, flags);
797}
798
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700799int sock_recvmsg(struct socket *sock, struct msghdr *msg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 size_t size, int flags)
801{
802 struct kiocb iocb;
803 struct sock_iocb siocb;
804 int ret;
805
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700806 init_sync_kiocb(&iocb, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 iocb.private = &siocb;
808 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
809 if (-EIOCBQUEUED == ret)
810 ret = wait_on_sync_kiocb(&iocb);
811 return ret;
812}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700813EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700815static int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
816 size_t size, int flags)
817{
818 struct kiocb iocb;
819 struct sock_iocb siocb;
820 int ret;
821
822 init_sync_kiocb(&iocb, NULL);
823 iocb.private = &siocb;
824 ret = __sock_recvmsg_nosec(&iocb, sock, msg, size, flags);
825 if (-EIOCBQUEUED == ret)
826 ret = wait_on_sync_kiocb(&iocb);
827 return ret;
828}
829
Martin Lucinac1249c02010-12-10 00:04:05 +0000830/**
831 * kernel_recvmsg - Receive a message from a socket (kernel space)
832 * @sock: The socket to receive the message from
833 * @msg: Received message
834 * @vec: Input s/g array for message data
835 * @num: Size of input s/g array
836 * @size: Number of bytes to read
837 * @flags: Message flags (MSG_DONTWAIT, etc...)
838 *
839 * On return the msg structure contains the scatter/gather array passed in the
840 * vec argument. The array is modified so that it consists of the unfilled
841 * portion of the original array.
842 *
843 * The returned value is the total number of bytes received, or an error.
844 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700845int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
846 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847{
848 mm_segment_t oldfs = get_fs();
849 int result;
850
851 set_fs(KERNEL_DS);
852 /*
853 * the following is safe, since for compiler definitions of kvec and
854 * iovec are identical, yielding the same in-core layout and alignment
855 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700856 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 result = sock_recvmsg(sock, msg, size, flags);
858 set_fs(oldfs);
859 return result;
860}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700861EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300863static ssize_t sock_sendpage(struct file *file, struct page *page,
864 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865{
866 struct socket *sock;
867 int flags;
868
Eric Dumazetb69aee02005-09-06 14:42:45 -0700869 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
Eric Dumazet35f9c092012-04-05 03:05:35 +0000871 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
872 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
873 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
Linus Torvaldse6949582009-08-13 08:28:36 -0700875 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876}
877
Jens Axboe9c55e012007-11-06 23:30:13 -0800878static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700879 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800880 unsigned int flags)
881{
882 struct socket *sock = file->private_data;
883
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800884 if (unlikely(!sock->ops->splice_read))
885 return -EINVAL;
886
Jens Axboe9c55e012007-11-06 23:30:13 -0800887 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
888}
889
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800890static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700891 struct sock_iocb *siocb)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800892{
Kent Overstreetd29c4452013-03-18 11:09:26 -0700893 if (!is_sync_kiocb(iocb))
894 BUG();
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800895
896 siocb->kiocb = iocb;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800897 iocb->private = siocb;
898 return siocb;
899}
900
901static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700902 struct file *file, const struct iovec *iov,
903 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800904{
905 struct socket *sock = file->private_data;
906 size_t size = 0;
907 int i;
908
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700909 for (i = 0; i < nr_segs; i++)
910 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800911
912 msg->msg_name = NULL;
913 msg->msg_namelen = 0;
914 msg->msg_control = NULL;
915 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700916 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800917 msg->msg_iovlen = nr_segs;
918 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
919
920 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
921}
922
Badari Pulavarty027445c2006-09-30 23:28:46 -0700923static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
924 unsigned long nr_segs, loff_t pos)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800925{
926 struct sock_iocb siocb, *x;
927
928 if (pos != 0)
929 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700930
Kent Overstreet73a70752013-05-09 15:03:42 -0700931 if (iocb->ki_nbytes == 0) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800932 return 0;
933
Badari Pulavarty027445c2006-09-30 23:28:46 -0700934
935 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800936 if (!x)
937 return -ENOMEM;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700938 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800939}
940
941static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
Badari Pulavarty027445c2006-09-30 23:28:46 -0700942 struct file *file, const struct iovec *iov,
943 unsigned long nr_segs)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800944{
945 struct socket *sock = file->private_data;
946 size_t size = 0;
947 int i;
948
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700949 for (i = 0; i < nr_segs; i++)
950 size += iov[i].iov_len;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800951
952 msg->msg_name = NULL;
953 msg->msg_namelen = 0;
954 msg->msg_control = NULL;
955 msg->msg_controllen = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700956 msg->msg_iov = (struct iovec *)iov;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800957 msg->msg_iovlen = nr_segs;
958 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
959 if (sock->type == SOCK_SEQPACKET)
960 msg->msg_flags |= MSG_EOR;
961
962 return __sock_sendmsg(iocb, sock, msg, size);
963}
964
Badari Pulavarty027445c2006-09-30 23:28:46 -0700965static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
966 unsigned long nr_segs, loff_t pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967{
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800968 struct sock_iocb siocb, *x;
969
970 if (pos != 0)
971 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700972
Badari Pulavarty027445c2006-09-30 23:28:46 -0700973 x = alloc_sock_iocb(iocb, &siocb);
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800974 if (!x)
975 return -ENOMEM;
976
Badari Pulavarty027445c2006-09-30 23:28:46 -0700977 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978}
979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980/*
981 * Atomic setting of ioctl hooks to avoid race
982 * with module unload.
983 */
984
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800985static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700986static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
Eric W. Biederman881d9662007-09-17 11:56:21 -0700988void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800990 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800992 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993}
994EXPORT_SYMBOL(brioctl_set);
995
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800996static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700997static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998
Eric W. Biederman881d9662007-09-17 11:56:21 -0700999void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001001 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001003 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004}
1005EXPORT_SYMBOL(vlan_ioctl_set);
1006
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001007static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001008static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001010void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001012 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001014 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015}
1016EXPORT_SYMBOL(dlci_ioctl_set);
1017
Arnd Bergmann6b960182009-11-06 23:10:54 -08001018static long sock_do_ioctl(struct net *net, struct socket *sock,
1019 unsigned int cmd, unsigned long arg)
1020{
1021 int err;
1022 void __user *argp = (void __user *)arg;
1023
1024 err = sock->ops->ioctl(sock, cmd, arg);
1025
1026 /*
1027 * If this ioctl is unknown try to hand it down
1028 * to the NIC driver.
1029 */
1030 if (err == -ENOIOCTLCMD)
1031 err = dev_ioctl(net, cmd, argp);
1032
1033 return err;
1034}
1035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036/*
1037 * With an ioctl, arg may well be a user mode pointer, but we don't know
1038 * what to do with it - that's up to the protocol still.
1039 */
1040
1041static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1042{
1043 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001044 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 void __user *argp = (void __user *)arg;
1046 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001047 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048
Eric Dumazetb69aee02005-09-06 14:42:45 -07001049 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001050 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001051 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001053 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001055#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -07001057 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001059#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001060 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061 case FIOSETOWN:
1062 case SIOCSPGRP:
1063 err = -EFAULT;
1064 if (get_user(pid, (int __user *)argp))
1065 break;
1066 err = f_setown(sock->file, pid, 1);
1067 break;
1068 case FIOGETOWN:
1069 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001070 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001071 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 break;
1073 case SIOCGIFBR:
1074 case SIOCSIFBR:
1075 case SIOCBRADDBR:
1076 case SIOCBRDELBR:
1077 err = -ENOPKG;
1078 if (!br_ioctl_hook)
1079 request_module("bridge");
1080
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001081 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001082 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001083 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001084 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 break;
1086 case SIOCGIFVLAN:
1087 case SIOCSIFVLAN:
1088 err = -ENOPKG;
1089 if (!vlan_ioctl_hook)
1090 request_module("8021q");
1091
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001092 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001094 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001095 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 case SIOCADDDLCI:
1098 case SIOCDELDLCI:
1099 err = -ENOPKG;
1100 if (!dlci_ioctl_hook)
1101 request_module("dlci");
1102
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001103 mutex_lock(&dlci_ioctl_mutex);
1104 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001106 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 break;
1108 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001109 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001111 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 return err;
1113}
1114
1115int sock_create_lite(int family, int type, int protocol, struct socket **res)
1116{
1117 int err;
1118 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001119
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 err = security_socket_create(family, type, protocol, 1);
1121 if (err)
1122 goto out;
1123
1124 sock = sock_alloc();
1125 if (!sock) {
1126 err = -ENOMEM;
1127 goto out;
1128 }
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001131 err = security_socket_post_create(sock, family, type, protocol, 1);
1132 if (err)
1133 goto out_release;
1134
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135out:
1136 *res = sock;
1137 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001138out_release:
1139 sock_release(sock);
1140 sock = NULL;
1141 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001143EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144
1145/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001146static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147{
Eliezer Tamircbf55002013-07-08 16:20:34 +03001148 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 struct socket *sock;
1150
1151 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001152 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001154 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001155
Eliezer Tamircbf55002013-07-08 16:20:34 +03001156 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001157 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001158 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001159
1160 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001161 if (wait && (wait->_key & POLL_BUSY_LOOP))
1162 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001163 }
1164
Eliezer Tamircbf55002013-07-08 16:20:34 +03001165 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166}
1167
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001168static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001170 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171
1172 return sock->ops->mmap(file, sock, vma);
1173}
1174
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001175static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 sock_release(SOCKET_I(inode));
1178 return 0;
1179}
1180
1181/*
1182 * Update the socket async list
1183 *
1184 * Fasync_list locking strategy.
1185 *
1186 * 1. fasync_list is modified only under process context socket lock
1187 * i.e. under semaphore.
1188 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001189 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190 */
1191
1192static int sock_fasync(int fd, struct file *filp, int on)
1193{
Eric Dumazet989a2972010-04-14 09:55:35 +00001194 struct socket *sock = filp->private_data;
1195 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001196 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197
Eric Dumazet989a2972010-04-14 09:55:35 +00001198 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200
1201 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001202 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1203 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204
Eric Dumazeteaefd112011-02-18 03:26:36 +00001205 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001206 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001207 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001208 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
Eric Dumazet989a2972010-04-14 09:55:35 +00001210 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 return 0;
1212}
1213
Eric Dumazet43815482010-04-29 11:01:49 +00001214/* This function may be called only under socket lock or callback_lock or rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215
1216int sock_wake_async(struct socket *sock, int how, int band)
1217{
Eric Dumazet43815482010-04-29 11:01:49 +00001218 struct socket_wq *wq;
1219
1220 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001222 rcu_read_lock();
1223 wq = rcu_dereference(sock->wq);
1224 if (!wq || !wq->fasync_list) {
1225 rcu_read_unlock();
1226 return -1;
1227 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001228 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001229 case SOCK_WAKE_WAITD:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1231 break;
1232 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001233 case SOCK_WAKE_SPACE:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1235 break;
1236 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001237 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001238call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001239 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001241 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001242 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 }
Eric Dumazet43815482010-04-29 11:01:49 +00001244 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 return 0;
1246}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001247EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248
Pavel Emelyanov721db932010-09-29 16:06:32 +04001249int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001250 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251{
1252 int err;
1253 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001254 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255
1256 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001257 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 */
1259 if (family < 0 || family >= NPROTO)
1260 return -EAFNOSUPPORT;
1261 if (type < 0 || type >= SOCK_MAX)
1262 return -EINVAL;
1263
1264 /* Compatibility.
1265
1266 This uglymoron is moved from INET layer to here to avoid
1267 deadlock in module load.
1268 */
1269 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001270 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 if (!warned) {
1272 warned = 1;
Yang Yingliang3410f222014-02-12 17:09:55 +08001273 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1274 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 }
1276 family = PF_PACKET;
1277 }
1278
1279 err = security_socket_create(family, type, protocol, kern);
1280 if (err)
1281 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001282
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001283 /*
1284 * Allocate the socket and allow the family to set things up. if
1285 * the protocol is 0, the family is instructed to select an appropriate
1286 * default.
1287 */
1288 sock = sock_alloc();
1289 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001290 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001291 return -ENFILE; /* Not exactly a match, but its the
1292 closest posix thing */
1293 }
1294
1295 sock->type = type;
1296
Johannes Berg95a5afc2008-10-16 15:24:51 -07001297#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001298 /* Attempt to load a protocol module if the find failed.
1299 *
1300 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 * requested real, full-featured networking support upon configuration.
1302 * Otherwise module support will break!
1303 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001304 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001305 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306#endif
1307
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001308 rcu_read_lock();
1309 pf = rcu_dereference(net_families[family]);
1310 err = -EAFNOSUPPORT;
1311 if (!pf)
1312 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
1314 /*
1315 * We will call the ->create function, that possibly is in a loadable
1316 * module, so we have to bump that loadable module refcnt first.
1317 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001318 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 goto out_release;
1320
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001321 /* Now protected by module ref count */
1322 rcu_read_unlock();
1323
Eric Paris3f378b62009-11-05 22:18:14 -08001324 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001325 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001327
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 /*
1329 * Now to bump the refcnt of the [loadable] module that owns this
1330 * socket at sock_release time we decrement its refcnt.
1331 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001332 if (!try_module_get(sock->ops->owner))
1333 goto out_module_busy;
1334
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 /*
1336 * Now that we're done with the ->create function, the [loadable]
1337 * module can have its refcnt decremented
1338 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001339 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001340 err = security_socket_post_create(sock, family, type, protocol, kern);
1341 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001342 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001343 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001345 return 0;
1346
1347out_module_busy:
1348 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001350 sock->ops = NULL;
1351 module_put(pf->owner);
1352out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001354 return err;
1355
1356out_release:
1357 rcu_read_unlock();
1358 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001360EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
1362int sock_create(int family, int type, int protocol, struct socket **res)
1363{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001364 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001366EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
1368int sock_create_kern(int family, int type, int protocol, struct socket **res)
1369{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001370 return __sock_create(&init_net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001372EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001374SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375{
1376 int retval;
1377 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001378 int flags;
1379
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001380 /* Check the SOCK_* constants for consistency. */
1381 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1382 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1383 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1384 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1385
Ulrich Dreppera677a032008-07-23 21:29:17 -07001386 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001387 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001388 return -EINVAL;
1389 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001391 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1392 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1393
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 retval = sock_create(family, type, protocol, &sock);
1395 if (retval < 0)
1396 goto out;
1397
Ulrich Drepper77d27202008-07-23 21:29:35 -07001398 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 if (retval < 0)
1400 goto out_release;
1401
1402out:
1403 /* It may be already another descriptor 8) Not kernel problem. */
1404 return retval;
1405
1406out_release:
1407 sock_release(sock);
1408 return retval;
1409}
1410
1411/*
1412 * Create a pair of connected sockets.
1413 */
1414
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001415SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1416 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
1418 struct socket *sock1, *sock2;
1419 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001420 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001421 int flags;
1422
1423 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001424 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001425 return -EINVAL;
1426 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001428 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1429 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 /*
1432 * Obtain the first socket and check if the underlying protocol
1433 * supports the socketpair call.
1434 */
1435
1436 err = sock_create(family, type, protocol, &sock1);
1437 if (err < 0)
1438 goto out;
1439
1440 err = sock_create(family, type, protocol, &sock2);
1441 if (err < 0)
1442 goto out_release_1;
1443
1444 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001445 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 goto out_release_both;
1447
Al Viro28407632012-08-17 23:54:15 -04001448 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001449 if (unlikely(fd1 < 0)) {
1450 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001452 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001453
Al Viro28407632012-08-17 23:54:15 -04001454 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001455 if (unlikely(fd2 < 0)) {
1456 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001457 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001458 }
1459
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001460 newfile1 = sock_alloc_file(sock1, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001461 if (unlikely(IS_ERR(newfile1))) {
1462 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001463 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001464 }
1465
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001466 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001467 if (IS_ERR(newfile2)) {
1468 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001469 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001470 }
1471
Yann Droneaudd73aa282013-12-09 22:42:20 +01001472 err = put_user(fd1, &usockvec[0]);
1473 if (err)
1474 goto out_fput_both;
1475
1476 err = put_user(fd2, &usockvec[1]);
1477 if (err)
1478 goto out_fput_both;
1479
Al Viro157cf642008-12-14 04:57:47 -05001480 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001481
Al Virodb349502007-02-07 01:48:00 -05001482 fd_install(fd1, newfile1);
1483 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 /* fd1 and fd2 may be already another descriptors.
1485 * Not kernel problem.
1486 */
1487
Yann Droneaudd73aa282013-12-09 22:42:20 +01001488 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489
Yann Droneaudd73aa282013-12-09 22:42:20 +01001490out_fput_both:
1491 fput(newfile2);
1492 fput(newfile1);
1493 put_unused_fd(fd2);
1494 put_unused_fd(fd1);
1495 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496
Yann Droneaudd73aa282013-12-09 22:42:20 +01001497out_fput_1:
1498 fput(newfile1);
1499 put_unused_fd(fd2);
1500 put_unused_fd(fd1);
1501 sock_release(sock2);
1502 goto out;
1503
1504out_put_unused_both:
1505 put_unused_fd(fd2);
1506out_put_unused_1:
1507 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001509 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001511 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512out:
1513 return err;
1514}
1515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516/*
1517 * Bind a name to a socket. Nothing much to do here since it's
1518 * the protocol's responsibility to handle the local address.
1519 *
1520 * We move the socket address to kernel space before we call
1521 * the protocol layer (having also checked the address is ok).
1522 */
1523
Heiko Carstens20f37032009-01-14 14:14:23 +01001524SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525{
1526 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001527 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001528 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001530 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001531 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001532 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001533 if (err >= 0) {
1534 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001535 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001536 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001537 if (!err)
1538 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001539 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001540 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001542 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001543 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 return err;
1545}
1546
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547/*
1548 * Perform a listen. Basically, we allow the protocol to do anything
1549 * necessary for a listen, and if that works, we mark the socket as
1550 * ready for listening.
1551 */
1552
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001553SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554{
1555 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001556 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001557 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001558
1559 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1560 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001561 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001562 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001563 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564
1565 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001566 if (!err)
1567 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001569 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 }
1571 return err;
1572}
1573
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574/*
1575 * For accept, we attempt to create a new socket, set up the link
1576 * with the client, wake up the client, then return the new
1577 * connected fd. We collect the address of the connector in kernel
1578 * space and move it to user at the very end. This is unclean because
1579 * we open the socket then return an error.
1580 *
1581 * 1003.1g adds the ability to recvmsg() to query connection pending
1582 * status to recvmsg. We need to add that support in a way thats
1583 * clean when we restucture accept also.
1584 */
1585
Heiko Carstens20f37032009-01-14 14:14:23 +01001586SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1587 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588{
1589 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001590 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001591 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001592 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593
Ulrich Drepper77d27202008-07-23 21:29:35 -07001594 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001595 return -EINVAL;
1596
1597 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1598 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1599
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001600 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 if (!sock)
1602 goto out;
1603
1604 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001605 newsock = sock_alloc();
1606 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 goto out_put;
1608
1609 newsock->type = sock->type;
1610 newsock->ops = sock->ops;
1611
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612 /*
1613 * We don't need try_module_get here, as the listening socket (sock)
1614 * has the protocol module (sock->ops->owner) held.
1615 */
1616 __module_get(newsock->ops->owner);
1617
Al Viro28407632012-08-17 23:54:15 -04001618 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001619 if (unlikely(newfd < 0)) {
1620 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001621 sock_release(newsock);
1622 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001623 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001624 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Al Viro28407632012-08-17 23:54:15 -04001625 if (unlikely(IS_ERR(newfile))) {
1626 err = PTR_ERR(newfile);
1627 put_unused_fd(newfd);
1628 sock_release(newsock);
1629 goto out_put;
1630 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001631
Frank Filza79af592005-09-27 15:23:38 -07001632 err = security_socket_accept(sock, newsock);
1633 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001634 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001635
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1637 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001638 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639
1640 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001641 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001642 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001644 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001646 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001647 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001649 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 }
1651
1652 /* File flags are not inherited via accept() unlike another OSes. */
1653
David S. Miller39d8c1b2006-03-20 17:13:49 -08001654 fd_install(newfd, newfile);
1655 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001658 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659out:
1660 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001661out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001662 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001663 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 goto out_put;
1665}
1666
Heiko Carstens20f37032009-01-14 14:14:23 +01001667SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1668 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001669{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001670 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001671}
1672
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673/*
1674 * Attempt to connect to a socket with the server address. The address
1675 * is in user space so we verify it is OK and move it to kernel space.
1676 *
1677 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1678 * break bindings
1679 *
1680 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1681 * other SEQPACKET protocols that take time to connect() as it doesn't
1682 * include the -EINPROGRESS status for such sockets.
1683 */
1684
Heiko Carstens20f37032009-01-14 14:14:23 +01001685SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1686 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687{
1688 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001689 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001690 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001692 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 if (!sock)
1694 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001695 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 if (err < 0)
1697 goto out_put;
1698
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001699 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001700 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 if (err)
1702 goto out_put;
1703
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001704 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705 sock->file->f_flags);
1706out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001707 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708out:
1709 return err;
1710}
1711
1712/*
1713 * Get the local address ('name') of a socket object. Move the obtained
1714 * name to user space.
1715 */
1716
Heiko Carstens20f37032009-01-14 14:14:23 +01001717SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1718 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719{
1720 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001721 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001722 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001723
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001724 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 if (!sock)
1726 goto out;
1727
1728 err = security_socket_getsockname(sock);
1729 if (err)
1730 goto out_put;
1731
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001732 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 if (err)
1734 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001735 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
1737out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001738 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739out:
1740 return err;
1741}
1742
1743/*
1744 * Get the remote address ('name') of a socket object. Move the obtained
1745 * name to user space.
1746 */
1747
Heiko Carstens20f37032009-01-14 14:14:23 +01001748SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1749 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
1751 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001752 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001753 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001755 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1756 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 err = security_socket_getpeername(sock);
1758 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001759 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 return err;
1761 }
1762
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001763 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001764 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001765 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001767 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001768 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001769 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 }
1771 return err;
1772}
1773
1774/*
1775 * Send a datagram to a given address. We move the address into kernel
1776 * space and check the user space data area is readable before invoking
1777 * the protocol.
1778 */
1779
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001780SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001781 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001782 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783{
1784 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001785 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 int err;
1787 struct msghdr msg;
1788 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001789 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001790
Linus Torvalds253eacc2010-10-30 16:43:10 -07001791 if (len > INT_MAX)
1792 len = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001793 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1794 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001795 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001796
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797 iov.iov_base = buff;
1798 iov.iov_len = len;
1799 msg.msg_name = NULL;
1800 msg.msg_iov = &iov;
1801 msg.msg_iovlen = 1;
1802 msg.msg_control = NULL;
1803 msg.msg_controllen = 0;
1804 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001805 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001806 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 if (err < 0)
1808 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001809 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001810 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 }
1812 if (sock->file->f_flags & O_NONBLOCK)
1813 flags |= MSG_DONTWAIT;
1814 msg.msg_flags = flags;
1815 err = sock_sendmsg(sock, &msg, len);
1816
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001817out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001818 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001819out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 return err;
1821}
1822
1823/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001824 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 */
1826
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001827SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001828 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829{
1830 return sys_sendto(fd, buff, len, flags, NULL, 0);
1831}
1832
1833/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001834 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835 * sender. We verify the buffers are writable and if needed move the
1836 * sender address from kernel to user space.
1837 */
1838
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001839SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001840 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001841 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842{
1843 struct socket *sock;
1844 struct iovec iov;
1845 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001846 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001847 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001848 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849
Linus Torvalds253eacc2010-10-30 16:43:10 -07001850 if (size > INT_MAX)
1851 size = INT_MAX;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001852 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001853 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001854 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001856 msg.msg_control = NULL;
1857 msg.msg_controllen = 0;
1858 msg.msg_iovlen = 1;
1859 msg.msg_iov = &iov;
1860 iov.iov_len = size;
1861 iov.iov_base = ubuf;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001862 /* Save some cycles and don't copy the address if not needed */
1863 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1864 /* We assume all kernel code knows the size of sockaddr_storage */
1865 msg.msg_namelen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 if (sock->file->f_flags & O_NONBLOCK)
1867 flags |= MSG_DONTWAIT;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001868 err = sock_recvmsg(sock, &msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001870 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001871 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001872 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001873 if (err2 < 0)
1874 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001876
1877 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001878out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 return err;
1880}
1881
1882/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001883 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 */
1885
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001886SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1887 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888{
1889 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1890}
1891
1892/*
1893 * Set a socket option. Because we don't know the option lengths we have
1894 * to pass the user mode parameter for the protocols to sort out.
1895 */
1896
Heiko Carstens20f37032009-01-14 14:14:23 +01001897SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1898 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001900 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 struct socket *sock;
1902
1903 if (optlen < 0)
1904 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001905
1906 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1907 if (sock != NULL) {
1908 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001909 if (err)
1910 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
1912 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001913 err =
1914 sock_setsockopt(sock, level, optname, optval,
1915 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001917 err =
1918 sock->ops->setsockopt(sock, level, optname, optval,
1919 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001920out_put:
1921 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 }
1923 return err;
1924}
1925
1926/*
1927 * Get a socket option. Because we don't know the option lengths we have
1928 * to pass a user mode parameter for the protocols to sort out.
1929 */
1930
Heiko Carstens20f37032009-01-14 14:14:23 +01001931SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1932 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001934 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 struct socket *sock;
1936
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001937 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1938 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001939 err = security_socket_getsockopt(sock, level, optname);
1940 if (err)
1941 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942
1943 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001944 err =
1945 sock_getsockopt(sock, level, optname, optval,
1946 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001948 err =
1949 sock->ops->getsockopt(sock, level, optname, optval,
1950 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001951out_put:
1952 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953 }
1954 return err;
1955}
1956
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957/*
1958 * Shutdown a socket.
1959 */
1960
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001961SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001963 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 struct socket *sock;
1965
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001966 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1967 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001969 if (!err)
1970 err = sock->ops->shutdown(sock, how);
1971 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 }
1973 return err;
1974}
1975
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001976/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 * fields which are the same type (int / unsigned) on our platforms.
1978 */
1979#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1980#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1981#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1982
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001983struct used_address {
1984 struct sockaddr_storage name;
1985 unsigned int name_len;
1986};
1987
Dan Carpenter1661bf32013-10-03 00:27:20 +03001988static int copy_msghdr_from_user(struct msghdr *kmsg,
1989 struct msghdr __user *umsg)
1990{
1991 if (copy_from_user(kmsg, umsg, sizeof(struct msghdr)))
1992 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001993
1994 if (kmsg->msg_namelen < 0)
1995 return -EINVAL;
1996
Dan Carpenter1661bf32013-10-03 00:27:20 +03001997 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001998 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Dan Carpenter1661bf32013-10-03 00:27:20 +03001999 return 0;
2000}
2001
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002002static int ___sys_sendmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002003 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002004 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002006 struct compat_msghdr __user *msg_compat =
2007 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002008 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002010 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002011 __attribute__ ((aligned(sizeof(__kernel_size_t))));
2012 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 unsigned char *ctl_buf = ctl;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002014 int err, ctl_len, total_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002015
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016 err = -EFAULT;
2017 if (MSG_CMSG_COMPAT & flags) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002018 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019 return -EFAULT;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002020 } else {
2021 err = copy_msghdr_from_user(msg_sys, msg);
2022 if (err)
2023 return err;
2024 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025
Anton Blanchard228e5482011-05-02 20:21:35 +00002026 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002027 err = -EMSGSIZE;
2028 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2029 goto out;
2030 err = -ENOMEM;
2031 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2032 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 if (!iov)
Anton Blanchard228e5482011-05-02 20:21:35 +00002034 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035 }
2036
2037 /* This will also move the address data into kernel space */
2038 if (MSG_CMSG_COMPAT & flags) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002039 err = verify_compat_iovec(msg_sys, iov, &address, VERIFY_READ);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 } else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002041 err = verify_iovec(msg_sys, iov, &address, VERIFY_READ);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002042 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043 goto out_freeiov;
2044 total_len = err;
2045
2046 err = -ENOBUFS;
2047
Anton Blanchard228e5482011-05-02 20:21:35 +00002048 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002050 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002051 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002052 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002053 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002054 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 if (err)
2056 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00002057 ctl_buf = msg_sys->msg_control;
2058 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002060 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002062 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063 goto out_freeiov;
2064 }
2065 err = -EFAULT;
2066 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00002067 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2069 * checking falls down on this.
2070 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00002071 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00002072 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002073 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002075 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002077 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078
2079 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002080 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002081 /*
2082 * If this is sendmmsg() and current destination address is same as
2083 * previously succeeded address, omit asking LSM's decision.
2084 * used_address->name_len is initialized to UINT_MAX so that the first
2085 * destination address never matches.
2086 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002087 if (used_address && msg_sys->msg_name &&
2088 used_address->name_len == msg_sys->msg_namelen &&
2089 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002090 used_address->name_len)) {
2091 err = sock_sendmsg_nosec(sock, msg_sys, total_len);
2092 goto out_freectl;
2093 }
2094 err = sock_sendmsg(sock, msg_sys, total_len);
2095 /*
2096 * If this is sendmmsg() and sending to current destination address was
2097 * successful, remember it.
2098 */
2099 if (used_address && err >= 0) {
2100 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002101 if (msg_sys->msg_name)
2102 memcpy(&used_address->name, msg_sys->msg_name,
2103 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002104 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105
2106out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002107 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2109out_freeiov:
2110 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002111 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002112out:
2113 return err;
2114}
2115
2116/*
2117 * BSD sendmsg interface
2118 */
2119
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002120long __sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002121{
2122 int fput_needed, err;
2123 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002124 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002125
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002126 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002127 if (!sock)
2128 goto out;
2129
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002130 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00002131
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002132 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002133out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134 return err;
2135}
2136
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002137SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned int, flags)
2138{
2139 if (flags & MSG_CMSG_COMPAT)
2140 return -EINVAL;
2141 return __sys_sendmsg(fd, msg, flags);
2142}
2143
Anton Blanchard228e5482011-05-02 20:21:35 +00002144/*
2145 * Linux sendmmsg interface
2146 */
2147
2148int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2149 unsigned int flags)
2150{
2151 int fput_needed, err, datagrams;
2152 struct socket *sock;
2153 struct mmsghdr __user *entry;
2154 struct compat_mmsghdr __user *compat_entry;
2155 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002156 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002157
Anton Blanchard98382f42011-08-04 14:07:39 +00002158 if (vlen > UIO_MAXIOV)
2159 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002160
2161 datagrams = 0;
2162
2163 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2164 if (!sock)
2165 return err;
2166
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002167 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002168 entry = mmsg;
2169 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002170 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002171
2172 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002173 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002174 err = ___sys_sendmsg(sock, (struct msghdr __user *)compat_entry,
2175 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002176 if (err < 0)
2177 break;
2178 err = __put_user(err, &compat_entry->msg_len);
2179 ++compat_entry;
2180 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002181 err = ___sys_sendmsg(sock,
2182 (struct msghdr __user *)entry,
2183 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002184 if (err < 0)
2185 break;
2186 err = put_user(err, &entry->msg_len);
2187 ++entry;
2188 }
2189
2190 if (err)
2191 break;
2192 ++datagrams;
2193 }
2194
Anton Blanchard228e5482011-05-02 20:21:35 +00002195 fput_light(sock->file, fput_needed);
2196
Anton Blanchard728ffb82011-08-04 14:07:38 +00002197 /* We only return an error if no datagrams were able to be sent */
2198 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002199 return datagrams;
2200
Anton Blanchard228e5482011-05-02 20:21:35 +00002201 return err;
2202}
2203
2204SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2205 unsigned int, vlen, unsigned int, flags)
2206{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002207 if (flags & MSG_CMSG_COMPAT)
2208 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002209 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2210}
2211
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002212static int ___sys_recvmsg(struct socket *sock, struct msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002213 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002215 struct compat_msghdr __user *msg_compat =
2216 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002218 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 unsigned long cmsg_ptr;
Eric Dumazeta74e9102012-04-20 20:04:01 +02002220 int err, total_len, len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221
2222 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002223 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002224
2225 /* user mode address pointers */
2226 struct sockaddr __user *uaddr;
2227 int __user *uaddr_len;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002228
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229 if (MSG_CMSG_COMPAT & flags) {
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002230 if (get_compat_msghdr(msg_sys, msg_compat))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 return -EFAULT;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002232 } else {
2233 err = copy_msghdr_from_user(msg_sys, msg);
2234 if (err)
2235 return err;
2236 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002238 if (msg_sys->msg_iovlen > UIO_FASTIOV) {
Eric Dumazeta74e9102012-04-20 20:04:01 +02002239 err = -EMSGSIZE;
2240 if (msg_sys->msg_iovlen > UIO_MAXIOV)
2241 goto out;
2242 err = -ENOMEM;
2243 iov = kmalloc(msg_sys->msg_iovlen * sizeof(struct iovec),
2244 GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245 if (!iov)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002246 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247 }
2248
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002249 /* Save the user-mode address (verify_iovec will change the
2250 * kernel msghdr to use the kernel address space)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251 */
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002252 uaddr = (__force void __user *)msg_sys->msg_name;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 uaddr_len = COMPAT_NAMELEN(msg);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002254 if (MSG_CMSG_COMPAT & flags)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002255 err = verify_compat_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002256 else
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002257 err = verify_iovec(msg_sys, iov, &addr, VERIFY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258 if (err < 0)
2259 goto out_freeiov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002260 total_len = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002261
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002262 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2263 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002264
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002265 /* We assume all kernel code knows the size of sockaddr_storage */
2266 msg_sys->msg_namelen = 0;
2267
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268 if (sock->file->f_flags & O_NONBLOCK)
2269 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002270 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2271 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002272 if (err < 0)
2273 goto out_freeiov;
2274 len = err;
2275
2276 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002277 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002278 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002279 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280 if (err < 0)
2281 goto out_freeiov;
2282 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002283 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002284 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002285 if (err)
2286 goto out_freeiov;
2287 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002288 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 &msg_compat->msg_controllen);
2290 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002291 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292 &msg->msg_controllen);
2293 if (err)
2294 goto out_freeiov;
2295 err = len;
2296
2297out_freeiov:
2298 if (iov != iovstack)
Eric Dumazeta74e9102012-04-20 20:04:01 +02002299 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002300out:
2301 return err;
2302}
2303
2304/*
2305 * BSD recvmsg interface
2306 */
2307
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002308long __sys_recvmsg(int fd, struct msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002309{
2310 int fput_needed, err;
2311 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002312 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002313
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002314 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002315 if (!sock)
2316 goto out;
2317
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002318 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002319
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002320 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321out:
2322 return err;
2323}
2324
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002325SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
2326 unsigned int, flags)
2327{
2328 if (flags & MSG_CMSG_COMPAT)
2329 return -EINVAL;
2330 return __sys_recvmsg(fd, msg, flags);
2331}
2332
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002333/*
2334 * Linux recvmmsg interface
2335 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002337int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2338 unsigned int flags, struct timespec *timeout)
2339{
2340 int fput_needed, err, datagrams;
2341 struct socket *sock;
2342 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002343 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002344 struct msghdr msg_sys;
2345 struct timespec end_time;
2346
2347 if (timeout &&
2348 poll_select_set_timeout(&end_time, timeout->tv_sec,
2349 timeout->tv_nsec))
2350 return -EINVAL;
2351
2352 datagrams = 0;
2353
2354 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2355 if (!sock)
2356 return err;
2357
2358 err = sock_error(sock->sk);
2359 if (err)
2360 goto out_put;
2361
2362 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002363 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002364
2365 while (datagrams < vlen) {
2366 /*
2367 * No need to ask LSM for more than the first datagram.
2368 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002369 if (MSG_CMSG_COMPAT & flags) {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002370 err = ___sys_recvmsg(sock, (struct msghdr __user *)compat_entry,
2371 &msg_sys, flags & ~MSG_WAITFORONE,
2372 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002373 if (err < 0)
2374 break;
2375 err = __put_user(err, &compat_entry->msg_len);
2376 ++compat_entry;
2377 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002378 err = ___sys_recvmsg(sock,
2379 (struct msghdr __user *)entry,
2380 &msg_sys, flags & ~MSG_WAITFORONE,
2381 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002382 if (err < 0)
2383 break;
2384 err = put_user(err, &entry->msg_len);
2385 ++entry;
2386 }
2387
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002388 if (err)
2389 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002390 ++datagrams;
2391
Brandon L Black71c5c152010-03-26 16:18:03 +00002392 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2393 if (flags & MSG_WAITFORONE)
2394 flags |= MSG_DONTWAIT;
2395
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002396 if (timeout) {
2397 ktime_get_ts(timeout);
2398 *timeout = timespec_sub(end_time, *timeout);
2399 if (timeout->tv_sec < 0) {
2400 timeout->tv_sec = timeout->tv_nsec = 0;
2401 break;
2402 }
2403
2404 /* Timeout, return less than vlen datagrams */
2405 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2406 break;
2407 }
2408
2409 /* Out of band data, return right away */
2410 if (msg_sys.msg_flags & MSG_OOB)
2411 break;
2412 }
2413
2414out_put:
2415 fput_light(sock->file, fput_needed);
2416
2417 if (err == 0)
2418 return datagrams;
2419
2420 if (datagrams != 0) {
2421 /*
2422 * We may return less entries than requested (vlen) if the
2423 * sock is non block and there aren't enough datagrams...
2424 */
2425 if (err != -EAGAIN) {
2426 /*
2427 * ... or if recvmsg returns an error after we
2428 * received some datagrams, where we record the
2429 * error to return on the next call or if the
2430 * app asks about it using getsockopt(SO_ERROR).
2431 */
2432 sock->sk->sk_err = -err;
2433 }
2434
2435 return datagrams;
2436 }
2437
2438 return err;
2439}
2440
2441SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2442 unsigned int, vlen, unsigned int, flags,
2443 struct timespec __user *, timeout)
2444{
2445 int datagrams;
2446 struct timespec timeout_sys;
2447
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002448 if (flags & MSG_CMSG_COMPAT)
2449 return -EINVAL;
2450
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002451 if (!timeout)
2452 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2453
2454 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2455 return -EFAULT;
2456
2457 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2458
2459 if (datagrams > 0 &&
2460 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2461 datagrams = -EFAULT;
2462
2463 return datagrams;
2464}
2465
2466#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467/* Argument list sizes for sys_socketcall */
2468#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002469static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002470 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2471 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2472 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002473 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002474};
2475
Linus Torvalds1da177e2005-04-16 15:20:36 -07002476#undef AL
2477
2478/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002479 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480 *
2481 * Argument checking cleaned up. Saved 20% in size.
2482 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002483 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484 */
2485
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002486SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487{
Chen Gang2950fa92013-04-07 16:55:23 +08002488 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002489 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002491 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002492
Anton Blanchard228e5482011-05-02 20:21:35 +00002493 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002494 return -EINVAL;
2495
Arjan van de Ven47379052009-09-28 12:57:44 -07002496 len = nargs[call];
2497 if (len > sizeof(a))
2498 return -EINVAL;
2499
Linus Torvalds1da177e2005-04-16 15:20:36 -07002500 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002501 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002502 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002503
Chen Gang2950fa92013-04-07 16:55:23 +08002504 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2505 if (err)
2506 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002507
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002508 a0 = a[0];
2509 a1 = a[1];
2510
2511 switch (call) {
2512 case SYS_SOCKET:
2513 err = sys_socket(a0, a1, a[2]);
2514 break;
2515 case SYS_BIND:
2516 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2517 break;
2518 case SYS_CONNECT:
2519 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2520 break;
2521 case SYS_LISTEN:
2522 err = sys_listen(a0, a1);
2523 break;
2524 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002525 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2526 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002527 break;
2528 case SYS_GETSOCKNAME:
2529 err =
2530 sys_getsockname(a0, (struct sockaddr __user *)a1,
2531 (int __user *)a[2]);
2532 break;
2533 case SYS_GETPEERNAME:
2534 err =
2535 sys_getpeername(a0, (struct sockaddr __user *)a1,
2536 (int __user *)a[2]);
2537 break;
2538 case SYS_SOCKETPAIR:
2539 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2540 break;
2541 case SYS_SEND:
2542 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2543 break;
2544 case SYS_SENDTO:
2545 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2546 (struct sockaddr __user *)a[4], a[5]);
2547 break;
2548 case SYS_RECV:
2549 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2550 break;
2551 case SYS_RECVFROM:
2552 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2553 (struct sockaddr __user *)a[4],
2554 (int __user *)a[5]);
2555 break;
2556 case SYS_SHUTDOWN:
2557 err = sys_shutdown(a0, a1);
2558 break;
2559 case SYS_SETSOCKOPT:
2560 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2561 break;
2562 case SYS_GETSOCKOPT:
2563 err =
2564 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2565 (int __user *)a[4]);
2566 break;
2567 case SYS_SENDMSG:
2568 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2569 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002570 case SYS_SENDMMSG:
2571 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2572 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002573 case SYS_RECVMSG:
2574 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2575 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002576 case SYS_RECVMMSG:
2577 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2578 (struct timespec __user *)a[4]);
2579 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002580 case SYS_ACCEPT4:
2581 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2582 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002583 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002584 default:
2585 err = -EINVAL;
2586 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587 }
2588 return err;
2589}
2590
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002591#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002593/**
2594 * sock_register - add a socket protocol handler
2595 * @ops: description of protocol
2596 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597 * This function is called by a protocol handler that wants to
2598 * advertise its address family, and have it linked into the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002599 * socket interface. The value ops->family coresponds to the
2600 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002602int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603{
2604 int err;
2605
2606 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002607 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608 return -ENOBUFS;
2609 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002610
2611 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002612 if (rcu_dereference_protected(net_families[ops->family],
2613 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002614 err = -EEXIST;
2615 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002616 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002617 err = 0;
2618 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002619 spin_unlock(&net_family_lock);
2620
Yang Yingliang3410f222014-02-12 17:09:55 +08002621 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622 return err;
2623}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002624EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002625
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002626/**
2627 * sock_unregister - remove a protocol handler
2628 * @family: protocol family to remove
2629 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002630 * This function is called by a protocol handler that wants to
2631 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002632 * new socket creation.
2633 *
2634 * If protocol handler is a module, then it can use module reference
2635 * counts to protect against new references. If protocol handler is not
2636 * a module then it needs to provide its own protection in
2637 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002639void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002640{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002641 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002643 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002644 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002645 spin_unlock(&net_family_lock);
2646
2647 synchronize_rcu();
2648
Yang Yingliang3410f222014-02-12 17:09:55 +08002649 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002650}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002651EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652
Andi Kleen77d76ea2005-12-22 12:43:42 -08002653static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002654{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002655 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002656 /*
2657 * Initialize the network sysctl infrastructure.
2658 */
2659 err = net_sysctl_init();
2660 if (err)
2661 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002662
Linus Torvalds1da177e2005-04-16 15:20:36 -07002663 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002664 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002665 */
2666 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002667
2668 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002669 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670 */
2671
2672 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002673
2674 err = register_filesystem(&sock_fs_type);
2675 if (err)
2676 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002677 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002678 if (IS_ERR(sock_mnt)) {
2679 err = PTR_ERR(sock_mnt);
2680 goto out_mount;
2681 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002682
2683 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002684 */
2685
2686#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002687 err = netfilter_init();
2688 if (err)
2689 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002690#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002691
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002692 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002693
Nick Pigginb3e19d92011-01-07 17:50:11 +11002694out:
2695 return err;
2696
2697out_mount:
2698 unregister_filesystem(&sock_fs_type);
2699out_fs:
2700 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002701}
2702
Andi Kleen77d76ea2005-12-22 12:43:42 -08002703core_initcall(sock_init); /* early initcall */
2704
Linus Torvalds1da177e2005-04-16 15:20:36 -07002705#ifdef CONFIG_PROC_FS
2706void socket_seq_show(struct seq_file *seq)
2707{
2708 int cpu;
2709 int counter = 0;
2710
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002711 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002712 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713
2714 /* It can be negative, by the way. 8) */
2715 if (counter < 0)
2716 counter = 0;
2717
2718 seq_printf(seq, "sockets: used %d\n", counter);
2719}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002720#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002721
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002722#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002723static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002724 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002725{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002726 mm_segment_t old_fs = get_fs();
2727 struct timeval ktv;
2728 int err;
2729
2730 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002731 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002732 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002733 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002734 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002735
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002736 return err;
2737}
2738
Arnd Bergmann6b960182009-11-06 23:10:54 -08002739static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002740 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002741{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002742 mm_segment_t old_fs = get_fs();
2743 struct timespec kts;
2744 int err;
2745
2746 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002747 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002748 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002749 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002750 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002751
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002752 return err;
2753}
2754
Arnd Bergmann6b960182009-11-06 23:10:54 -08002755static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002756{
2757 struct ifreq __user *uifr;
2758 int err;
2759
2760 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002761 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002762 return -EFAULT;
2763
Arnd Bergmann6b960182009-11-06 23:10:54 -08002764 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002765 if (err)
2766 return err;
2767
Arnd Bergmann6b960182009-11-06 23:10:54 -08002768 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002769 return -EFAULT;
2770
2771 return 0;
2772}
2773
Arnd Bergmann6b960182009-11-06 23:10:54 -08002774static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002775{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002776 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002777 struct ifconf ifc;
2778 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002779 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002780 struct ifreq __user *ifr;
2781 unsigned int i, j;
2782 int err;
2783
Arnd Bergmann6b960182009-11-06 23:10:54 -08002784 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002785 return -EFAULT;
2786
Mathias Krause43da5f22012-08-15 11:31:57 +00002787 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002788 if (ifc32.ifcbuf == 0) {
2789 ifc32.ifc_len = 0;
2790 ifc.ifc_len = 0;
2791 ifc.ifc_req = NULL;
2792 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2793 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002794 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2795 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002796 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2797 ifc.ifc_len = len;
2798 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2799 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002800 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002801 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002802 return -EFAULT;
2803 ifr++;
2804 ifr32++;
2805 }
2806 }
2807 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2808 return -EFAULT;
2809
Arnd Bergmann6b960182009-11-06 23:10:54 -08002810 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002811 if (err)
2812 return err;
2813
2814 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2815 return -EFAULT;
2816
2817 ifr = ifc.ifc_req;
2818 ifr32 = compat_ptr(ifc32.ifcbuf);
2819 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002820 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2821 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2822 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002823 return -EFAULT;
2824 ifr32++;
2825 ifr++;
2826 }
2827
2828 if (ifc32.ifcbuf == 0) {
2829 /* Translate from 64-bit structure multiple to
2830 * a 32-bit one.
2831 */
2832 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002833 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002834 ifc32.ifc_len = i;
2835 } else {
2836 ifc32.ifc_len = i;
2837 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002838 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002839 return -EFAULT;
2840
2841 return 0;
2842}
2843
Arnd Bergmann6b960182009-11-06 23:10:54 -08002844static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002845{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002846 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2847 bool convert_in = false, convert_out = false;
2848 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2849 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002850 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002851 u32 rule_cnt = 0, actual_rule_cnt;
2852 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002853 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002854 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002855
2856 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2857 return -EFAULT;
2858
Ben Hutchings3a7da392011-03-17 07:34:32 +00002859 compat_rxnfc = compat_ptr(data);
2860
2861 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002862 return -EFAULT;
2863
Ben Hutchings3a7da392011-03-17 07:34:32 +00002864 /* Most ethtool structures are defined without padding.
2865 * Unfortunately struct ethtool_rxnfc is an exception.
2866 */
2867 switch (ethcmd) {
2868 default:
2869 break;
2870 case ETHTOOL_GRXCLSRLALL:
2871 /* Buffer size is variable */
2872 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2873 return -EFAULT;
2874 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2875 return -ENOMEM;
2876 buf_size += rule_cnt * sizeof(u32);
2877 /* fall through */
2878 case ETHTOOL_GRXRINGS:
2879 case ETHTOOL_GRXCLSRLCNT:
2880 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002881 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002882 convert_out = true;
2883 /* fall through */
2884 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002885 buf_size += sizeof(struct ethtool_rxnfc);
2886 convert_in = true;
2887 break;
2888 }
2889
2890 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002891 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002892
2893 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2894 return -EFAULT;
2895
2896 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2897 &ifr->ifr_ifru.ifru_data))
2898 return -EFAULT;
2899
2900 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002901 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002902 * fs.ring_cookie and at the end of fs, but nowhere else.
2903 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002904 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2905 sizeof(compat_rxnfc->fs.m_ext) !=
2906 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2907 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002908 BUILD_BUG_ON(
2909 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2910 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2911 offsetof(struct ethtool_rxnfc, fs.location) -
2912 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2913
2914 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002915 (void __user *)(&rxnfc->fs.m_ext + 1) -
2916 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002917 copy_in_user(&rxnfc->fs.ring_cookie,
2918 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002919 (void __user *)(&rxnfc->fs.location + 1) -
2920 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002921 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2922 sizeof(rxnfc->rule_cnt)))
2923 return -EFAULT;
2924 }
2925
2926 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2927 if (ret)
2928 return ret;
2929
2930 if (convert_out) {
2931 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002932 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2933 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002934 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2935 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002936 (const void __user *)(&rxnfc->fs.location + 1) -
2937 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002938 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2939 sizeof(rxnfc->rule_cnt)))
2940 return -EFAULT;
2941
2942 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2943 /* As an optimisation, we only copy the actual
2944 * number of rules that the underlying
2945 * function returned. Since Mallory might
2946 * change the rule count in user memory, we
2947 * check that it is less than the rule count
2948 * originally given (as the user buffer size),
2949 * which has been range-checked.
2950 */
2951 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2952 return -EFAULT;
2953 if (actual_rule_cnt < rule_cnt)
2954 rule_cnt = actual_rule_cnt;
2955 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2956 &rxnfc->rule_locs[0],
2957 rule_cnt * sizeof(u32)))
2958 return -EFAULT;
2959 }
2960 }
2961
2962 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002963}
2964
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002965static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2966{
2967 void __user *uptr;
2968 compat_uptr_t uptr32;
2969 struct ifreq __user *uifr;
2970
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002971 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002972 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2973 return -EFAULT;
2974
2975 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2976 return -EFAULT;
2977
2978 uptr = compat_ptr(uptr32);
2979
2980 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2981 return -EFAULT;
2982
2983 return dev_ioctl(net, SIOCWANDEV, uifr);
2984}
2985
Arnd Bergmann6b960182009-11-06 23:10:54 -08002986static int bond_ioctl(struct net *net, unsigned int cmd,
2987 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002988{
2989 struct ifreq kifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002990 mm_segment_t old_fs;
2991 int err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002992
2993 switch (cmd) {
2994 case SIOCBONDENSLAVE:
2995 case SIOCBONDRELEASE:
2996 case SIOCBONDSETHWADDR:
2997 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002998 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002999 return -EFAULT;
3000
3001 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003002 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003003 err = dev_ioctl(net, cmd,
3004 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003005 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003006
3007 return err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003008 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08003009 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003010 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003011}
3012
Ben Hutchings590d4692013-11-18 17:04:13 +00003013/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3014static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08003015 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003016{
3017 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003018 char tmp_buf[IFNAMSIZ];
3019 void __user *data64;
3020 u32 data32;
3021
3022 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
3023 IFNAMSIZ))
3024 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00003025 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003026 return -EFAULT;
3027 data64 = compat_ptr(data32);
3028
3029 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
3030
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003031 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
3032 IFNAMSIZ))
3033 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00003034 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003035 return -EFAULT;
3036
Arnd Bergmann6b960182009-11-06 23:10:54 -08003037 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003038}
3039
Arnd Bergmann6b960182009-11-06 23:10:54 -08003040static int dev_ifsioc(struct net *net, struct socket *sock,
3041 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003042{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003043 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003044 int err;
3045
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003046 uifr = compat_alloc_user_space(sizeof(*uifr));
3047 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3048 return -EFAULT;
3049
3050 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3051
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003052 if (!err) {
3053 switch (cmd) {
3054 case SIOCGIFFLAGS:
3055 case SIOCGIFMETRIC:
3056 case SIOCGIFMTU:
3057 case SIOCGIFMEM:
3058 case SIOCGIFHWADDR:
3059 case SIOCGIFINDEX:
3060 case SIOCGIFADDR:
3061 case SIOCGIFBRDADDR:
3062 case SIOCGIFDSTADDR:
3063 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003064 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003065 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08003066 case SIOCGMIIPHY:
3067 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003068 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003069 err = -EFAULT;
3070 break;
3071 }
3072 }
3073 return err;
3074}
3075
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003076static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3077 struct compat_ifreq __user *uifr32)
3078{
3079 struct ifreq ifr;
3080 struct compat_ifmap __user *uifmap32;
3081 mm_segment_t old_fs;
3082 int err;
3083
3084 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3085 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003086 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3087 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3088 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3089 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3090 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3091 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003092 if (err)
3093 return -EFAULT;
3094
3095 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003096 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003097 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003098 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003099
3100 if (cmd == SIOCGIFMAP && !err) {
3101 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003102 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3103 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3104 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3105 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3106 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3107 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003108 if (err)
3109 err = -EFAULT;
3110 }
3111 return err;
3112}
3113
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003114struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003115 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003116 struct sockaddr rt_dst; /* target address */
3117 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3118 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003119 unsigned short rt_flags;
3120 short rt_pad2;
3121 u32 rt_pad3;
3122 unsigned char rt_tos;
3123 unsigned char rt_class;
3124 short rt_pad4;
3125 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003126 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003127 u32 rt_mtu; /* per route MTU/Window */
3128 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003129 unsigned short rt_irtt; /* Initial RTT */
3130};
3131
3132struct in6_rtmsg32 {
3133 struct in6_addr rtmsg_dst;
3134 struct in6_addr rtmsg_src;
3135 struct in6_addr rtmsg_gateway;
3136 u32 rtmsg_type;
3137 u16 rtmsg_dst_len;
3138 u16 rtmsg_src_len;
3139 u32 rtmsg_metric;
3140 u32 rtmsg_info;
3141 u32 rtmsg_flags;
3142 s32 rtmsg_ifindex;
3143};
3144
Arnd Bergmann6b960182009-11-06 23:10:54 -08003145static int routing_ioctl(struct net *net, struct socket *sock,
3146 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003147{
3148 int ret;
3149 void *r = NULL;
3150 struct in6_rtmsg r6;
3151 struct rtentry r4;
3152 char devname[16];
3153 u32 rtdev;
3154 mm_segment_t old_fs = get_fs();
3155
Arnd Bergmann6b960182009-11-06 23:10:54 -08003156 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3157 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003158 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003159 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003160 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3161 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3162 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3163 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3164 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3165 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3166 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003167
3168 r = (void *) &r6;
3169 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003170 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003171 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003172 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003173 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3174 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3175 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3176 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3177 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3178 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003179 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003180 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003181 r4.rt_dev = (char __user __force *)devname;
3182 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003183 } else
3184 r4.rt_dev = NULL;
3185
3186 r = (void *) &r4;
3187 }
3188
3189 if (ret) {
3190 ret = -EFAULT;
3191 goto out;
3192 }
3193
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003194 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003195 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003196 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003197
3198out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003199 return ret;
3200}
3201
3202/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3203 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003204 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003205 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003206static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003207{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003208 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003209
Arnd Bergmann6b960182009-11-06 23:10:54 -08003210 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003211 return -EFAULT;
3212 if (tmp == BRCTL_GET_VERSION)
3213 return BRCTL_VERSION + 1;
3214 return -EINVAL;
3215}
3216
Arnd Bergmann6b960182009-11-06 23:10:54 -08003217static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3218 unsigned int cmd, unsigned long arg)
3219{
3220 void __user *argp = compat_ptr(arg);
3221 struct sock *sk = sock->sk;
3222 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003223
Arnd Bergmann6b960182009-11-06 23:10:54 -08003224 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003225 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003226
Arnd Bergmann6b960182009-11-06 23:10:54 -08003227 switch (cmd) {
3228 case SIOCSIFBR:
3229 case SIOCGIFBR:
3230 return old_bridge_ioctl(argp);
3231 case SIOCGIFNAME:
3232 return dev_ifname32(net, argp);
3233 case SIOCGIFCONF:
3234 return dev_ifconf(net, argp);
3235 case SIOCETHTOOL:
3236 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003237 case SIOCWANDEV:
3238 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003239 case SIOCGIFMAP:
3240 case SIOCSIFMAP:
3241 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003242 case SIOCBONDENSLAVE:
3243 case SIOCBONDRELEASE:
3244 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003245 case SIOCBONDCHANGEACTIVE:
3246 return bond_ioctl(net, cmd, argp);
3247 case SIOCADDRT:
3248 case SIOCDELRT:
3249 return routing_ioctl(net, sock, cmd, argp);
3250 case SIOCGSTAMP:
3251 return do_siocgstamp(net, sock, cmd, argp);
3252 case SIOCGSTAMPNS:
3253 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003254 case SIOCBONDSLAVEINFOQUERY:
3255 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003256 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003257 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003258 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003259
Arnd Bergmann6b960182009-11-06 23:10:54 -08003260 case FIOSETOWN:
3261 case SIOCSPGRP:
3262 case FIOGETOWN:
3263 case SIOCGPGRP:
3264 case SIOCBRADDBR:
3265 case SIOCBRDELBR:
3266 case SIOCGIFVLAN:
3267 case SIOCSIFVLAN:
3268 case SIOCADDDLCI:
3269 case SIOCDELDLCI:
3270 return sock_ioctl(file, cmd, arg);
3271
3272 case SIOCGIFFLAGS:
3273 case SIOCSIFFLAGS:
3274 case SIOCGIFMETRIC:
3275 case SIOCSIFMETRIC:
3276 case SIOCGIFMTU:
3277 case SIOCSIFMTU:
3278 case SIOCGIFMEM:
3279 case SIOCSIFMEM:
3280 case SIOCGIFHWADDR:
3281 case SIOCSIFHWADDR:
3282 case SIOCADDMULTI:
3283 case SIOCDELMULTI:
3284 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003285 case SIOCGIFADDR:
3286 case SIOCSIFADDR:
3287 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003288 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003289 case SIOCGIFBRDADDR:
3290 case SIOCSIFBRDADDR:
3291 case SIOCGIFDSTADDR:
3292 case SIOCSIFDSTADDR:
3293 case SIOCGIFNETMASK:
3294 case SIOCSIFNETMASK:
3295 case SIOCSIFPFLAGS:
3296 case SIOCGIFPFLAGS:
3297 case SIOCGIFTXQLEN:
3298 case SIOCSIFTXQLEN:
3299 case SIOCBRADDIF:
3300 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003301 case SIOCSIFNAME:
3302 case SIOCGMIIPHY:
3303 case SIOCGMIIREG:
3304 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003305 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003306
Arnd Bergmann6b960182009-11-06 23:10:54 -08003307 case SIOCSARP:
3308 case SIOCGARP:
3309 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003310 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003311 return sock_do_ioctl(net, sock, cmd, arg);
3312 }
3313
Arnd Bergmann6b960182009-11-06 23:10:54 -08003314 return -ENOIOCTLCMD;
3315}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003316
Eric Dumazet95c96172012-04-15 05:58:06 +00003317static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003318 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003319{
3320 struct socket *sock = file->private_data;
3321 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003322 struct sock *sk;
3323 struct net *net;
3324
3325 sk = sock->sk;
3326 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003327
3328 if (sock->ops->compat_ioctl)
3329 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3330
David S. Miller87de87d2008-06-03 09:14:03 -07003331 if (ret == -ENOIOCTLCMD &&
3332 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3333 ret = compat_wext_handle_ioctl(net, cmd, arg);
3334
Arnd Bergmann6b960182009-11-06 23:10:54 -08003335 if (ret == -ENOIOCTLCMD)
3336 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3337
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003338 return ret;
3339}
3340#endif
3341
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003342int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3343{
3344 return sock->ops->bind(sock, addr, addrlen);
3345}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003346EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003347
3348int kernel_listen(struct socket *sock, int backlog)
3349{
3350 return sock->ops->listen(sock, backlog);
3351}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003352EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003353
3354int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3355{
3356 struct sock *sk = sock->sk;
3357 int err;
3358
3359 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3360 newsock);
3361 if (err < 0)
3362 goto done;
3363
3364 err = sock->ops->accept(sock, *newsock, flags);
3365 if (err < 0) {
3366 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003367 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003368 goto done;
3369 }
3370
3371 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003372 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003373
3374done:
3375 return err;
3376}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003377EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003378
3379int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003380 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003381{
3382 return sock->ops->connect(sock, addr, addrlen, flags);
3383}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003384EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003385
3386int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3387 int *addrlen)
3388{
3389 return sock->ops->getname(sock, addr, addrlen, 0);
3390}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003391EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003392
3393int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3394 int *addrlen)
3395{
3396 return sock->ops->getname(sock, addr, addrlen, 1);
3397}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003398EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003399
3400int kernel_getsockopt(struct socket *sock, int level, int optname,
3401 char *optval, int *optlen)
3402{
3403 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003404 char __user *uoptval;
3405 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003406 int err;
3407
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003408 uoptval = (char __user __force *) optval;
3409 uoptlen = (int __user __force *) optlen;
3410
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003411 set_fs(KERNEL_DS);
3412 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003413 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003414 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003415 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3416 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003417 set_fs(oldfs);
3418 return err;
3419}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003420EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003421
3422int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003423 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003424{
3425 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003426 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003427 int err;
3428
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003429 uoptval = (char __user __force *) optval;
3430
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003431 set_fs(KERNEL_DS);
3432 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003433 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003434 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003435 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003436 optlen);
3437 set_fs(oldfs);
3438 return err;
3439}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003440EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003441
3442int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3443 size_t size, int flags)
3444{
3445 if (sock->ops->sendpage)
3446 return sock->ops->sendpage(sock, page, offset, size, flags);
3447
3448 return sock_no_sendpage(sock, page, offset, size, flags);
3449}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003450EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003451
3452int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3453{
3454 mm_segment_t oldfs = get_fs();
3455 int err;
3456
3457 set_fs(KERNEL_DS);
3458 err = sock->ops->ioctl(sock, cmd, arg);
3459 set_fs(oldfs);
3460
3461 return err;
3462}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003463EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003464
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003465int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3466{
3467 return sock->ops->shutdown(sock, how);
3468}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003469EXPORT_SYMBOL(kernel_sock_shutdown);