blob: 456fadb3d8193a2238bf30881506cd66dbe11957 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
Jesper Juhl02c30a82005-05-05 16:16:16 -07007 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
Stephen Hemminger89bddce2006-09-01 00:19:31 -070045 * Tigran Aivazian : Made listen(2) backlog sanity checks
Linus Torvalds1da177e2005-04-16 15:20:36 -070046 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -070056 * paradigm.
Linus Torvalds1da177e2005-04-16 15:20:36 -070057 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070066#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070067#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080071#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030073#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
Daniel Borkmann408eccc2014-04-01 16:20:23 +020075#include <linux/ptp_classify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070081#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010086#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010087#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070088#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070089#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090090#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000091#include <linux/xattr.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
93#include <asm/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070097#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070098#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
Arnd Bergmann6b960182009-11-06 23:10:54 -0800103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800106#include <linux/sockios.h>
107#include <linux/atalk.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300108#include <net/busy_poll.h>
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400109#include <linux/errqueue.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300110
Cong Wange0d10952013-08-01 11:10:25 +0800111#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300112unsigned int sysctl_net_busy_read __read_mostly;
113unsigned int sysctl_net_busy_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300114#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800115
Al Viro8ae5e032014-11-28 19:40:50 -0500116static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
120static int sock_close(struct inode *inode, struct file *file);
121static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700123static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800124#ifdef CONFIG_COMPAT
125static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700126 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800127#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800131static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700132 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800133 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135/*
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
138 */
139
Arjan van de Venda7071d2007-02-12 00:55:36 -0800140static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 .owner = THIS_MODULE,
142 .llseek = no_llseek,
Al Viro8ae5e032014-11-28 19:40:50 -0500143 .read_iter = sock_read_iter,
144 .write_iter = sock_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145 .poll = sock_poll,
146 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800147#ifdef CONFIG_COMPAT
148 .compat_ioctl = compat_sock_ioctl,
149#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 .mmap = sock_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 .release = sock_close,
152 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200153 .sendpage = sock_sendpage,
154 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800155 .splice_read = sock_splice_read,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156};
157
158/*
159 * The protocol list. Each protocol is registered in here.
160 */
161
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000163static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165/*
166 * Statistics counters of the socket lists
167 */
168
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700169static DEFINE_PER_CPU(int, sockets_in_use);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170
171/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700172 * Support routines.
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 */
176
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177/**
178 * move_addr_to_kernel - copy a socket address into kernel space
179 * @uaddr: Address in user space
180 * @kaddr: Address in kernel space
181 * @ulen: Length in user space
182 *
183 * The address is copied into kernel space. If the provided address is
184 * too long an error code of -EINVAL is returned. If the copy gives
185 * invalid addresses -EFAULT is returned. On a success 0 is returned.
186 */
187
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000188int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700190 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700192 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700194 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100196 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197}
198
199/**
200 * move_addr_to_user - copy an address to user space
201 * @kaddr: kernel space address
202 * @klen: length of address in kernel
203 * @uaddr: user space address
204 * @ulen: pointer to user length field
205 *
206 * The value pointed to by ulen on entry is the buffer length available.
207 * This is overwritten with the buffer space used. -EINVAL is returned
208 * if an overlong buffer is specified or a negative buffer size. -EFAULT
209 * is returned if either the buffer or the length field are not
210 * accessible.
211 * After copying the data up to the limit the user specifies, the true
212 * length of the data is written over the length limit the user
213 * specified. Zero is returned for a success.
214 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700215
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000216static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000217 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218{
219 int err;
220 int len;
221
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100222 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700223 err = get_user(len, ulen);
224 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700226 if (len > klen)
227 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100228 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700230 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500231 if (audit_sockaddr(klen, kaddr))
232 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700233 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 return -EFAULT;
235 }
236 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700237 * "fromlen shall refer to the value before truncation.."
238 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 */
240 return __put_user(klen, ulen);
241}
242
Christoph Lametere18b8902006-12-06 20:33:20 -0800243static struct kmem_cache *sock_inode_cachep __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
245static struct inode *sock_alloc_inode(struct super_block *sb)
246{
247 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000248 struct socket_wq *wq;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700249
Christoph Lametere94b1762006-12-06 20:33:17 -0800250 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 if (!ei)
252 return NULL;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000253 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
254 if (!wq) {
Eric Dumazet43815482010-04-29 11:01:49 +0000255 kmem_cache_free(sock_inode_cachep, ei);
256 return NULL;
257 }
Eric Dumazeteaefd112011-02-18 03:26:36 +0000258 init_waitqueue_head(&wq->wait);
259 wq->fasync_list = NULL;
260 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700261
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 ei->socket.state = SS_UNCONNECTED;
263 ei->socket.flags = 0;
264 ei->socket.ops = NULL;
265 ei->socket.sk = NULL;
266 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
268 return &ei->vfs_inode;
269}
270
271static void sock_destroy_inode(struct inode *inode)
272{
Eric Dumazet43815482010-04-29 11:01:49 +0000273 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000274 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000275
276 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000277 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800278 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000279 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280}
281
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700282static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700284 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
Christoph Lametera35afb82007-05-16 22:10:57 -0700286 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700288
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289static int init_inodecache(void)
290{
291 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700292 sizeof(struct socket_alloc),
293 0,
294 (SLAB_HWCACHE_ALIGN |
295 SLAB_RECLAIM_ACCOUNT |
296 SLAB_MEM_SPREAD),
Paul Mundt20c2df82007-07-20 10:11:58 +0900297 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 if (sock_inode_cachep == NULL)
299 return -ENOMEM;
300 return 0;
301}
302
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700303static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700304 .alloc_inode = sock_alloc_inode,
305 .destroy_inode = sock_destroy_inode,
306 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307};
308
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700309/*
310 * sockfs_dname() is called from d_path().
311 */
312static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
313{
314 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000315 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700316}
317
Al Viro3ba13d12009-02-20 06:02:22 +0000318static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700319 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320};
321
Al Viroc74a1cb2011-01-12 16:59:34 -0500322static struct dentry *sockfs_mount(struct file_system_type *fs_type,
323 int flags, const char *dev_name, void *data)
324{
325 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
326 &sockfs_dentry_operations, SOCKFS_MAGIC);
327}
328
329static struct vfsmount *sock_mnt __read_mostly;
330
331static struct file_system_type sock_fs_type = {
332 .name = "sockfs",
333 .mount = sockfs_mount,
334 .kill_sb = kill_anon_super,
335};
336
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337/*
338 * Obtains the first available file descriptor and sets it up for use.
339 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
347 *
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
352 */
353
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700354struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355{
Al Viro7cbe66b2009-08-05 19:59:08 +0400356 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400357 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400358 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800359
Masatake YAMATO600e1772012-08-29 10:44:29 +0000360 if (dname) {
361 name.name = dname;
362 name.len = strlen(name.name);
363 } else if (sock->sk) {
364 name.name = sock->sk->sk_prot_creator->name;
365 name.len = strlen(name.name);
366 }
Nick Piggin4b936882011-01-07 17:50:07 +1100367 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400368 if (unlikely(!path.dentry))
369 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400370 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
Al Viro2c48b9c2009-08-09 00:52:35 +0400372 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400373
Al Viro2c48b9c2009-08-09 00:52:35 +0400374 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400375 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530376 if (IS_ERR(file)) {
Al Virocc3808f2009-08-06 09:43:59 +0400377 /* drop dentry, keep inode */
David Howellsc5ef6032015-03-17 22:26:16 +0000378 ihold(d_inode(path.dentry));
Al Viro2c48b9c2009-08-09 00:52:35 +0400379 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700380 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400381 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800382
383 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700384 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800385 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400386 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800387}
Al Viro56b31d12012-08-18 00:25:51 -0400388EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800389
Al Viro56b31d12012-08-18 00:25:51 -0400390static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800391{
392 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400393 int fd = get_unused_fd_flags(flags);
394 if (unlikely(fd < 0))
395 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800396
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700397 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400398 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800399 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400400 return fd;
401 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400402
Al Viro28407632012-08-17 23:54:15 -0400403 put_unused_fd(fd);
404 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405}
406
John Fastabend406a3c62012-07-20 10:39:25 +0000407struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800408{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800409 if (file->f_op == &socket_file_ops)
410 return file->private_data; /* set in sock_map_fd */
411
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800412 *err = -ENOTSOCK;
413 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800414}
John Fastabend406a3c62012-07-20 10:39:25 +0000415EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800416
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700418 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 * @fd: file handle
420 * @err: pointer to an error code return
421 *
422 * The file handle passed in is locked and the socket it is bound
423 * too is returned. If an error occurs the err pointer is overwritten
424 * with a negative errno code and NULL is returned. The function checks
425 * for both invalid handles and passing a handle which is not a socket.
426 *
427 * On a success the socket object pointer is returned.
428 */
429
430struct socket *sockfd_lookup(int fd, int *err)
431{
432 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 struct socket *sock;
434
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700435 file = fget(fd);
436 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 *err = -EBADF;
438 return NULL;
439 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700440
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800441 sock = sock_from_file(file, err);
442 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 return sock;
445}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700446EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800448static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
449{
Al Viro00e188e2014-03-03 23:48:18 -0500450 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800451 struct socket *sock;
452
Hua Zhong36725582006-04-19 15:25:02 -0700453 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500454 if (f.file) {
455 sock = sock_from_file(f.file, err);
456 if (likely(sock)) {
457 *fput_needed = f.flags;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800458 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500459 }
460 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800461 }
462 return NULL;
463}
464
Masatake YAMATO600e1772012-08-29 10:44:29 +0000465#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
466#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
467#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
468static ssize_t sockfs_getxattr(struct dentry *dentry,
469 const char *name, void *value, size_t size)
470{
471 const char *proto_name;
472 size_t proto_size;
473 int error;
474
475 error = -ENODATA;
476 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
477 proto_name = dentry->d_name.name;
478 proto_size = strlen(proto_name);
479
480 if (value) {
481 error = -ERANGE;
482 if (proto_size + 1 > size)
483 goto out;
484
485 strncpy(value, proto_name, proto_size + 1);
486 }
487 error = proto_size + 1;
488 }
489
490out:
491 return error;
492}
493
494static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
495 size_t size)
496{
497 ssize_t len;
498 ssize_t used = 0;
499
David Howellsc5ef6032015-03-17 22:26:16 +0000500 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000501 if (len < 0)
502 return len;
503 used += len;
504 if (buffer) {
505 if (size < used)
506 return -ERANGE;
507 buffer += len;
508 }
509
510 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
511 used += len;
512 if (buffer) {
513 if (size < used)
514 return -ERANGE;
515 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
516 buffer += len;
517 }
518
519 return used;
520}
521
522static const struct inode_operations sockfs_inode_ops = {
523 .getxattr = sockfs_getxattr,
524 .listxattr = sockfs_listxattr,
525};
526
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527/**
528 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700529 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 * Allocate a new inode and socket object. The two are bound together
531 * and initialised. The socket is then returned. If we are out of inodes
532 * NULL is returned.
533 */
534
535static struct socket *sock_alloc(void)
536{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700537 struct inode *inode;
538 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200540 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 if (!inode)
542 return NULL;
543
544 sock = SOCKET_I(inode);
545
Eric Dumazet29a020d2009-09-15 02:39:20 -0700546 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400547 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700548 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100549 inode->i_uid = current_fsuid();
550 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000551 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Alex Shi19e8d692012-05-14 14:15:31 -0700553 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 return sock;
555}
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557/**
558 * sock_release - close a socket
559 * @sock: socket to close
560 *
561 * The socket is released from the protocol stack if it has a release
562 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700563 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700565
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566void sock_release(struct socket *sock)
567{
568 if (sock->ops) {
569 struct module *owner = sock->ops->owner;
570
571 sock->ops->release(sock);
572 sock->ops = NULL;
573 module_put(owner);
574 }
575
Eric Dumazeteaefd112011-02-18 03:26:36 +0000576 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800577 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
Alex Shi19e8d692012-05-14 14:15:31 -0700579 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 if (!sock->file) {
581 iput(SOCK_INODE(sock));
582 return;
583 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700584 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700586EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400588void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000589{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200590 u8 flags = *tx_flags;
591
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400592 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200593 flags |= SKBTX_HW_TSTAMP;
594
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400595 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200596 flags |= SKBTX_SW_TSTAMP;
597
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400598 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200599 flags |= SKBTX_SCHED_TSTAMP;
600
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400601 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200602 flags |= SKBTX_ACK_TSTAMP;
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400603
Eric Dumazet140c55d2014-08-06 11:49:29 +0200604 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000605}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400606EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000607
Al Virod8725c82014-12-11 00:02:50 -0500608static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609{
Al Viro01e97e62014-12-15 21:39:31 -0500610 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500611 BUG_ON(ret == -EIOCBQUEUED);
612 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800614
Al Virod8725c82014-12-11 00:02:50 -0500615int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800616{
Al Virod8725c82014-12-11 00:02:50 -0500617 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500618 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800619
Al Virod8725c82014-12-11 00:02:50 -0500620 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800621}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700622EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
624int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
625 struct kvec *vec, size_t num, size_t size)
626{
Al Viro6aa24812015-03-21 19:56:16 -0400627 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500628 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700630EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700632/*
633 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
634 */
635void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
636 struct sk_buff *skb)
637{
Patrick Ohly20d49472009-02-12 05:03:38 +0000638 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400639 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000640 int empty = 1;
641 struct skb_shared_hwtstamps *shhwtstamps =
642 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700643
Patrick Ohly20d49472009-02-12 05:03:38 +0000644 /* Race occurred between timestamp enabling and packet
645 receiving. Fill in the current time for now. */
646 if (need_software_tstamp && skb->tstamp.tv64 == 0)
647 __net_timestamp(skb);
648
649 if (need_software_tstamp) {
650 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
651 struct timeval tv;
652 skb_get_timestamp(skb, &tv);
653 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
654 sizeof(tv), &tv);
655 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400656 struct timespec ts;
657 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000658 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400659 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000660 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700661 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000662
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400663 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400664 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400665 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000666 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400667 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400668 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400669 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400670 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000671 if (!empty)
672 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400673 SCM_TIMESTAMPING, sizeof(tss), &tss);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700674}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300675EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
676
Johannes Berg6e3e9392011-11-09 10:15:42 +0100677void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
678 struct sk_buff *skb)
679{
680 int ack;
681
682 if (!sock_flag(sk, SOCK_WIFI_STATUS))
683 return;
684 if (!skb->wifi_acked_valid)
685 return;
686
687 ack = skb->wifi_acked;
688
689 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
690}
691EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
692
stephen hemminger11165f12010-10-18 14:27:29 +0000693static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
694 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700695{
Eyal Birger744d5a32015-03-01 14:58:31 +0200696 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700697 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200698 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700699}
700
Eric Dumazet767dd032010-04-28 19:14:43 +0000701void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700702 struct sk_buff *skb)
703{
704 sock_recv_timestamp(msg, sk, skb);
705 sock_recv_drops(msg, sk, skb);
706}
Eric Dumazet767dd032010-04-28 19:14:43 +0000707EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700708
Ying Xue1b784142015-03-02 15:37:48 +0800709static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
710 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711{
Ying Xue1b784142015-03-02 15:37:48 +0800712 return sock->ops->recvmsg(sock, msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713}
714
Ying Xue1b784142015-03-02 15:37:48 +0800715int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
716 int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700717{
718 int err = security_socket_recvmsg(sock, msg, size, flags);
719
Ying Xue1b784142015-03-02 15:37:48 +0800720 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700722EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
Martin Lucinac1249c02010-12-10 00:04:05 +0000724/**
725 * kernel_recvmsg - Receive a message from a socket (kernel space)
726 * @sock: The socket to receive the message from
727 * @msg: Received message
728 * @vec: Input s/g array for message data
729 * @num: Size of input s/g array
730 * @size: Number of bytes to read
731 * @flags: Message flags (MSG_DONTWAIT, etc...)
732 *
733 * On return the msg structure contains the scatter/gather array passed in the
734 * vec argument. The array is modified so that it consists of the unfilled
735 * portion of the original array.
736 *
737 * The returned value is the total number of bytes received, or an error.
738 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700739int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
740 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741{
742 mm_segment_t oldfs = get_fs();
743 int result;
744
Al Viro6aa24812015-03-21 19:56:16 -0400745 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 set_fs(KERNEL_DS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 result = sock_recvmsg(sock, msg, size, flags);
748 set_fs(oldfs);
749 return result;
750}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700751EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300753static ssize_t sock_sendpage(struct file *file, struct page *page,
754 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
756 struct socket *sock;
757 int flags;
758
Eric Dumazetb69aee02005-09-06 14:42:45 -0700759 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
Eric Dumazet35f9c092012-04-05 03:05:35 +0000761 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
762 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
763 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
Linus Torvaldse6949582009-08-13 08:28:36 -0700765 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766}
767
Jens Axboe9c55e012007-11-06 23:30:13 -0800768static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700769 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800770 unsigned int flags)
771{
772 struct socket *sock = file->private_data;
773
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800774 if (unlikely(!sock->ops->splice_read))
775 return -EINVAL;
776
Jens Axboe9c55e012007-11-06 23:30:13 -0800777 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
778}
779
Al Viro8ae5e032014-11-28 19:40:50 -0500780static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800781{
Al Viro6d652332015-01-30 16:12:56 -0500782 struct file *file = iocb->ki_filp;
783 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700784 struct msghdr msg = {.msg_iter = *to,
785 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500786 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800787
Al Viro8ae5e032014-11-28 19:40:50 -0500788 if (file->f_flags & O_NONBLOCK)
789 msg.msg_flags = MSG_DONTWAIT;
790
791 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800792 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700793
Christoph Hellwig66ee59a2015-02-11 19:56:46 +0100794 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800795 return 0;
796
Al Viro237dae82015-04-09 00:00:30 -0400797 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
Al Viro8ae5e032014-11-28 19:40:50 -0500798 *to = msg.msg_iter;
799 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800800}
801
Al Viro8ae5e032014-11-28 19:40:50 -0500802static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803{
Al Viro6d652332015-01-30 16:12:56 -0500804 struct file *file = iocb->ki_filp;
805 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700806 struct msghdr msg = {.msg_iter = *from,
807 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500808 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800809
Al Viro8ae5e032014-11-28 19:40:50 -0500810 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800811 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700812
Al Viro8ae5e032014-11-28 19:40:50 -0500813 if (file->f_flags & O_NONBLOCK)
814 msg.msg_flags = MSG_DONTWAIT;
815
Al Viro6d652332015-01-30 16:12:56 -0500816 if (sock->type == SOCK_SEQPACKET)
817 msg.msg_flags |= MSG_EOR;
818
Al Virod8725c82014-12-11 00:02:50 -0500819 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e032014-11-28 19:40:50 -0500820 *from = msg.msg_iter;
821 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822}
823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824/*
825 * Atomic setting of ioctl hooks to avoid race
826 * with module unload.
827 */
828
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800829static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700830static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
Eric W. Biederman881d9662007-09-17 11:56:21 -0700832void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800834 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800836 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837}
838EXPORT_SYMBOL(brioctl_set);
839
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800840static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700841static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842
Eric W. Biederman881d9662007-09-17 11:56:21 -0700843void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800845 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800847 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848}
849EXPORT_SYMBOL(vlan_ioctl_set);
850
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800851static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700852static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700854void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800856 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800858 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859}
860EXPORT_SYMBOL(dlci_ioctl_set);
861
Arnd Bergmann6b960182009-11-06 23:10:54 -0800862static long sock_do_ioctl(struct net *net, struct socket *sock,
863 unsigned int cmd, unsigned long arg)
864{
865 int err;
866 void __user *argp = (void __user *)arg;
867
868 err = sock->ops->ioctl(sock, cmd, arg);
869
870 /*
871 * If this ioctl is unknown try to hand it down
872 * to the NIC driver.
873 */
874 if (err == -ENOIOCTLCMD)
875 err = dev_ioctl(net, cmd, argp);
876
877 return err;
878}
879
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880/*
881 * With an ioctl, arg may well be a user mode pointer, but we don't know
882 * what to do with it - that's up to the protocol still.
883 */
884
885static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
886{
887 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700888 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 void __user *argp = (void __user *)arg;
890 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700891 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Eric Dumazetb69aee02005-09-06 14:42:45 -0700893 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700894 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900895 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700897 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200899#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700901 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200903#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700904 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 case FIOSETOWN:
906 case SIOCSPGRP:
907 err = -EFAULT;
908 if (get_user(pid, (int __user *)argp))
909 break;
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400910 f_setown(sock->file, pid, 1);
911 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 break;
913 case FIOGETOWN:
914 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700915 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700916 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917 break;
918 case SIOCGIFBR:
919 case SIOCSIFBR:
920 case SIOCBRADDBR:
921 case SIOCBRDELBR:
922 err = -ENOPKG;
923 if (!br_ioctl_hook)
924 request_module("bridge");
925
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800926 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700927 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700928 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800929 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930 break;
931 case SIOCGIFVLAN:
932 case SIOCSIFVLAN:
933 err = -ENOPKG;
934 if (!vlan_ioctl_hook)
935 request_module("8021q");
936
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800937 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700939 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800940 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 case SIOCADDDLCI:
943 case SIOCDELDLCI:
944 err = -ENOPKG;
945 if (!dlci_ioctl_hook)
946 request_module("dlci");
947
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700948 mutex_lock(&dlci_ioctl_mutex);
949 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700951 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 break;
953 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -0800954 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700956 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 return err;
958}
959
960int sock_create_lite(int family, int type, int protocol, struct socket **res)
961{
962 int err;
963 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 err = security_socket_create(family, type, protocol, 1);
966 if (err)
967 goto out;
968
969 sock = sock_alloc();
970 if (!sock) {
971 err = -ENOMEM;
972 goto out;
973 }
974
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700976 err = security_socket_post_create(sock, family, type, protocol, 1);
977 if (err)
978 goto out_release;
979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980out:
981 *res = sock;
982 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700983out_release:
984 sock_release(sock);
985 sock = NULL;
986 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700988EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
990/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700991static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992{
Eliezer Tamircbf55002013-07-08 16:20:34 +0300993 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 struct socket *sock;
995
996 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700997 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700999 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001000
Eliezer Tamircbf55002013-07-08 16:20:34 +03001001 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001002 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001003 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001004
1005 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001006 if (wait && (wait->_key & POLL_BUSY_LOOP))
1007 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001008 }
1009
Eliezer Tamircbf55002013-07-08 16:20:34 +03001010 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011}
1012
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001013static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001015 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016
1017 return sock->ops->mmap(file, sock, vma);
1018}
1019
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001020static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 sock_release(SOCKET_I(inode));
1023 return 0;
1024}
1025
1026/*
1027 * Update the socket async list
1028 *
1029 * Fasync_list locking strategy.
1030 *
1031 * 1. fasync_list is modified only under process context socket lock
1032 * i.e. under semaphore.
1033 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001034 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 */
1036
1037static int sock_fasync(int fd, struct file *filp, int on)
1038{
Eric Dumazet989a2972010-04-14 09:55:35 +00001039 struct socket *sock = filp->private_data;
1040 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001041 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042
Eric Dumazet989a2972010-04-14 09:55:35 +00001043 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045
1046 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001047 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1048 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001049
Eric Dumazeteaefd112011-02-18 03:26:36 +00001050 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001051 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001052 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001053 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054
Eric Dumazet989a2972010-04-14 09:55:35 +00001055 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056 return 0;
1057}
1058
Eric Dumazetceb5d582015-11-29 20:03:11 -08001059/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060
Eric Dumazetceb5d582015-11-29 20:03:11 -08001061int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001063 if (!wq || !wq->fasync_list)
1064 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001065
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001066 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001067 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001068 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 break;
1070 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001071 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001072 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 break;
1074 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001075 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001076call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001077 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001079 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001080 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001082
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 return 0;
1084}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001085EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086
Pavel Emelyanov721db932010-09-29 16:06:32 +04001087int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001088 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089{
1090 int err;
1091 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001092 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
1094 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001095 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 */
1097 if (family < 0 || family >= NPROTO)
1098 return -EAFNOSUPPORT;
1099 if (type < 0 || type >= SOCK_MAX)
1100 return -EINVAL;
1101
1102 /* Compatibility.
1103
1104 This uglymoron is moved from INET layer to here to avoid
1105 deadlock in module load.
1106 */
1107 if (family == PF_INET && type == SOCK_PACKET) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001108 static int warned;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 if (!warned) {
1110 warned = 1;
Yang Yingliang3410f222014-02-12 17:09:55 +08001111 pr_info("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1112 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 }
1114 family = PF_PACKET;
1115 }
1116
1117 err = security_socket_create(family, type, protocol, kern);
1118 if (err)
1119 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001120
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001121 /*
1122 * Allocate the socket and allow the family to set things up. if
1123 * the protocol is 0, the family is instructed to select an appropriate
1124 * default.
1125 */
1126 sock = sock_alloc();
1127 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001128 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001129 return -ENFILE; /* Not exactly a match, but its the
1130 closest posix thing */
1131 }
1132
1133 sock->type = type;
1134
Johannes Berg95a5afc2008-10-16 15:24:51 -07001135#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001136 /* Attempt to load a protocol module if the find failed.
1137 *
1138 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139 * requested real, full-featured networking support upon configuration.
1140 * Otherwise module support will break!
1141 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001142 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001143 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144#endif
1145
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001146 rcu_read_lock();
1147 pf = rcu_dereference(net_families[family]);
1148 err = -EAFNOSUPPORT;
1149 if (!pf)
1150 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
1152 /*
1153 * We will call the ->create function, that possibly is in a loadable
1154 * module, so we have to bump that loadable module refcnt first.
1155 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001156 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 goto out_release;
1158
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001159 /* Now protected by module ref count */
1160 rcu_read_unlock();
1161
Eric Paris3f378b62009-11-05 22:18:14 -08001162 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001163 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001165
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 /*
1167 * Now to bump the refcnt of the [loadable] module that owns this
1168 * socket at sock_release time we decrement its refcnt.
1169 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001170 if (!try_module_get(sock->ops->owner))
1171 goto out_module_busy;
1172
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 /*
1174 * Now that we're done with the ->create function, the [loadable]
1175 * module can have its refcnt decremented
1176 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001177 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001178 err = security_socket_post_create(sock, family, type, protocol, kern);
1179 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001180 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001181 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001183 return 0;
1184
1185out_module_busy:
1186 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001188 sock->ops = NULL;
1189 module_put(pf->owner);
1190out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001192 return err;
1193
1194out_release:
1195 rcu_read_unlock();
1196 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001198EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
1200int sock_create(int family, int type, int protocol, struct socket **res)
1201{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001202 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001204EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001206int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001208 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001210EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001212SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213{
1214 int retval;
1215 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001216 int flags;
1217
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001218 /* Check the SOCK_* constants for consistency. */
1219 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1220 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1221 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1222 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1223
Ulrich Dreppera677a032008-07-23 21:29:17 -07001224 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001225 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001226 return -EINVAL;
1227 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001229 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1230 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1231
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 retval = sock_create(family, type, protocol, &sock);
1233 if (retval < 0)
1234 goto out;
1235
Ulrich Drepper77d27202008-07-23 21:29:35 -07001236 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 if (retval < 0)
1238 goto out_release;
1239
1240out:
1241 /* It may be already another descriptor 8) Not kernel problem. */
1242 return retval;
1243
1244out_release:
1245 sock_release(sock);
1246 return retval;
1247}
1248
1249/*
1250 * Create a pair of connected sockets.
1251 */
1252
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001253SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1254 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255{
1256 struct socket *sock1, *sock2;
1257 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001258 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001259 int flags;
1260
1261 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001262 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001263 return -EINVAL;
1264 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001266 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1267 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1268
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 /*
1270 * Obtain the first socket and check if the underlying protocol
1271 * supports the socketpair call.
1272 */
1273
1274 err = sock_create(family, type, protocol, &sock1);
1275 if (err < 0)
1276 goto out;
1277
1278 err = sock_create(family, type, protocol, &sock2);
1279 if (err < 0)
1280 goto out_release_1;
1281
1282 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001283 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 goto out_release_both;
1285
Al Viro28407632012-08-17 23:54:15 -04001286 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001287 if (unlikely(fd1 < 0)) {
1288 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001290 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001291
Al Viro28407632012-08-17 23:54:15 -04001292 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001293 if (unlikely(fd2 < 0)) {
1294 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001295 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001296 }
1297
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001298 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301299 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001300 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001301 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001302 }
1303
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001304 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001305 if (IS_ERR(newfile2)) {
1306 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001307 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001308 }
1309
Yann Droneaudd73aa282013-12-09 22:42:20 +01001310 err = put_user(fd1, &usockvec[0]);
1311 if (err)
1312 goto out_fput_both;
1313
1314 err = put_user(fd2, &usockvec[1]);
1315 if (err)
1316 goto out_fput_both;
1317
Al Viro157cf642008-12-14 04:57:47 -05001318 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001319
Al Virodb349502007-02-07 01:48:00 -05001320 fd_install(fd1, newfile1);
1321 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 /* fd1 and fd2 may be already another descriptors.
1323 * Not kernel problem.
1324 */
1325
Yann Droneaudd73aa282013-12-09 22:42:20 +01001326 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
Yann Droneaudd73aa282013-12-09 22:42:20 +01001328out_fput_both:
1329 fput(newfile2);
1330 fput(newfile1);
1331 put_unused_fd(fd2);
1332 put_unused_fd(fd1);
1333 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334
Yann Droneaudd73aa282013-12-09 22:42:20 +01001335out_fput_1:
1336 fput(newfile1);
1337 put_unused_fd(fd2);
1338 put_unused_fd(fd1);
1339 sock_release(sock2);
1340 goto out;
1341
1342out_put_unused_both:
1343 put_unused_fd(fd2);
1344out_put_unused_1:
1345 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001347 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001349 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350out:
1351 return err;
1352}
1353
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354/*
1355 * Bind a name to a socket. Nothing much to do here since it's
1356 * the protocol's responsibility to handle the local address.
1357 *
1358 * We move the socket address to kernel space before we call
1359 * the protocol layer (having also checked the address is ok).
1360 */
1361
Heiko Carstens20f37032009-01-14 14:14:23 +01001362SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363{
1364 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001365 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001366 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001368 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001369 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001370 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001371 if (err >= 0) {
1372 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001373 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001374 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001375 if (!err)
1376 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001377 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001378 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001380 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001381 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 return err;
1383}
1384
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385/*
1386 * Perform a listen. Basically, we allow the protocol to do anything
1387 * necessary for a listen, and if that works, we mark the socket as
1388 * ready for listening.
1389 */
1390
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001391SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392{
1393 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001394 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001395 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001396
1397 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1398 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001399 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001400 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001401 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
1403 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001404 if (!err)
1405 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001407 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 }
1409 return err;
1410}
1411
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412/*
1413 * For accept, we attempt to create a new socket, set up the link
1414 * with the client, wake up the client, then return the new
1415 * connected fd. We collect the address of the connector in kernel
1416 * space and move it to user at the very end. This is unclean because
1417 * we open the socket then return an error.
1418 *
1419 * 1003.1g adds the ability to recvmsg() to query connection pending
1420 * status to recvmsg. We need to add that support in a way thats
1421 * clean when we restucture accept also.
1422 */
1423
Heiko Carstens20f37032009-01-14 14:14:23 +01001424SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1425 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426{
1427 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001428 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001429 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001430 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431
Ulrich Drepper77d27202008-07-23 21:29:35 -07001432 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001433 return -EINVAL;
1434
1435 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1436 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1437
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001438 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 if (!sock)
1440 goto out;
1441
1442 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001443 newsock = sock_alloc();
1444 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 goto out_put;
1446
1447 newsock->type = sock->type;
1448 newsock->ops = sock->ops;
1449
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 /*
1451 * We don't need try_module_get here, as the listening socket (sock)
1452 * has the protocol module (sock->ops->owner) held.
1453 */
1454 __module_get(newsock->ops->owner);
1455
Al Viro28407632012-08-17 23:54:15 -04001456 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001457 if (unlikely(newfd < 0)) {
1458 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001459 sock_release(newsock);
1460 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001461 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001462 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301463 if (IS_ERR(newfile)) {
Al Viro28407632012-08-17 23:54:15 -04001464 err = PTR_ERR(newfile);
1465 put_unused_fd(newfd);
1466 sock_release(newsock);
1467 goto out_put;
1468 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001469
Frank Filza79af592005-09-27 15:23:38 -07001470 err = security_socket_accept(sock, newsock);
1471 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001472 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001473
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1475 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001476 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
1478 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001479 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001480 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001482 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001484 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001485 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001487 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 }
1489
1490 /* File flags are not inherited via accept() unlike another OSes. */
1491
David S. Miller39d8c1b2006-03-20 17:13:49 -08001492 fd_install(newfd, newfile);
1493 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001496 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497out:
1498 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001499out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001500 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001501 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 goto out_put;
1503}
1504
Heiko Carstens20f37032009-01-14 14:14:23 +01001505SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1506 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001507{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001508 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001509}
1510
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511/*
1512 * Attempt to connect to a socket with the server address. The address
1513 * is in user space so we verify it is OK and move it to kernel space.
1514 *
1515 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1516 * break bindings
1517 *
1518 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1519 * other SEQPACKET protocols that take time to connect() as it doesn't
1520 * include the -EINPROGRESS status for such sockets.
1521 */
1522
Heiko Carstens20f37032009-01-14 14:14:23 +01001523SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1524 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
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001530 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 if (!sock)
1532 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001533 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 if (err < 0)
1535 goto out_put;
1536
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001537 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001538 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 if (err)
1540 goto out_put;
1541
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001542 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 sock->file->f_flags);
1544out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001545 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546out:
1547 return err;
1548}
1549
1550/*
1551 * Get the local address ('name') of a socket object. Move the obtained
1552 * name to user space.
1553 */
1554
Heiko Carstens20f37032009-01-14 14:14:23 +01001555SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1556 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557{
1558 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001559 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001560 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001561
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001562 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563 if (!sock)
1564 goto out;
1565
1566 err = security_socket_getsockname(sock);
1567 if (err)
1568 goto out_put;
1569
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001570 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 if (err)
1572 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001573 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574
1575out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001576 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577out:
1578 return err;
1579}
1580
1581/*
1582 * Get the remote address ('name') of a socket object. Move the obtained
1583 * name to user space.
1584 */
1585
Heiko Carstens20f37032009-01-14 14:14:23 +01001586SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1587 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588{
1589 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001590 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001591 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001593 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1594 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 err = security_socket_getpeername(sock);
1596 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001597 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 return err;
1599 }
1600
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001601 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001602 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001603 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001605 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001606 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001607 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 }
1609 return err;
1610}
1611
1612/*
1613 * Send a datagram to a given address. We move the address into kernel
1614 * space and check the user space data area is readable before invoking
1615 * the protocol.
1616 */
1617
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001618SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001619 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001620 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621{
1622 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001623 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 int err;
1625 struct msghdr msg;
1626 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001627 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001628
Al Viro602bd0e2015-03-21 19:12:32 -04001629 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1630 if (unlikely(err))
1631 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001632 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1633 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001634 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001635
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001636 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001637 msg.msg_control = NULL;
1638 msg.msg_controllen = 0;
1639 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001640 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001641 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 if (err < 0)
1643 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001644 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001645 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 }
1647 if (sock->file->f_flags & O_NONBLOCK)
1648 flags |= MSG_DONTWAIT;
1649 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05001650 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001652out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001653 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001654out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 return err;
1656}
1657
1658/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001659 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 */
1661
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001662SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001663 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664{
1665 return sys_sendto(fd, buff, len, flags, NULL, 0);
1666}
1667
1668/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001669 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 * sender. We verify the buffers are writable and if needed move the
1671 * sender address from kernel to user space.
1672 */
1673
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001674SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001675 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001676 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677{
1678 struct socket *sock;
1679 struct iovec iov;
1680 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001681 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001682 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001683 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684
Al Viro602bd0e2015-03-21 19:12:32 -04001685 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1686 if (unlikely(err))
1687 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001688 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001690 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001692 msg.msg_control = NULL;
1693 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001694 /* Save some cycles and don't copy the address if not needed */
1695 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1696 /* We assume all kernel code knows the size of sockaddr_storage */
1697 msg.msg_namelen = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 if (sock->file->f_flags & O_NONBLOCK)
1699 flags |= MSG_DONTWAIT;
Al Viro602bd0e2015-03-21 19:12:32 -04001700 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001702 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001703 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001704 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001705 if (err2 < 0)
1706 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001708
1709 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001710out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 return err;
1712}
1713
1714/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001715 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 */
1717
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001718SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1719 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720{
1721 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1722}
1723
1724/*
1725 * Set a socket option. Because we don't know the option lengths we have
1726 * to pass the user mode parameter for the protocols to sort out.
1727 */
1728
Heiko Carstens20f37032009-01-14 14:14:23 +01001729SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1730 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001732 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 struct socket *sock;
1734
1735 if (optlen < 0)
1736 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001737
1738 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1739 if (sock != NULL) {
1740 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001741 if (err)
1742 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
1744 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001745 err =
1746 sock_setsockopt(sock, level, optname, optval,
1747 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001749 err =
1750 sock->ops->setsockopt(sock, level, optname, optval,
1751 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001752out_put:
1753 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 }
1755 return err;
1756}
1757
1758/*
1759 * Get a socket option. Because we don't know the option lengths we have
1760 * to pass a user mode parameter for the protocols to sort out.
1761 */
1762
Heiko Carstens20f37032009-01-14 14:14:23 +01001763SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1764 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001766 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 struct socket *sock;
1768
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001769 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1770 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001771 err = security_socket_getsockopt(sock, level, optname);
1772 if (err)
1773 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774
1775 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001776 err =
1777 sock_getsockopt(sock, level, optname, optval,
1778 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001780 err =
1781 sock->ops->getsockopt(sock, level, optname, optval,
1782 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001783out_put:
1784 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 }
1786 return err;
1787}
1788
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789/*
1790 * Shutdown a socket.
1791 */
1792
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001793SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001795 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 struct socket *sock;
1797
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001798 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1799 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001801 if (!err)
1802 err = sock->ops->shutdown(sock, how);
1803 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 }
1805 return err;
1806}
1807
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001808/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 * fields which are the same type (int / unsigned) on our platforms.
1810 */
1811#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1812#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1813#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1814
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001815struct used_address {
1816 struct sockaddr_storage name;
1817 unsigned int name_len;
1818};
1819
Al Viroda184282015-03-21 19:29:06 -04001820static int copy_msghdr_from_user(struct msghdr *kmsg,
1821 struct user_msghdr __user *umsg,
1822 struct sockaddr __user **save_addr,
1823 struct iovec **iov)
Dan Carpenter1661bf32013-10-03 00:27:20 +03001824{
Al Viro08adb7d2014-11-10 20:23:13 -05001825 struct sockaddr __user *uaddr;
1826 struct iovec __user *uiov;
Al Viroc0371da2014-11-24 10:42:55 -05001827 size_t nr_segs;
Al Viro08adb7d2014-11-10 20:23:13 -05001828 ssize_t err;
1829
1830 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1831 __get_user(uaddr, &umsg->msg_name) ||
1832 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1833 __get_user(uiov, &umsg->msg_iov) ||
Al Viroc0371da2014-11-24 10:42:55 -05001834 __get_user(nr_segs, &umsg->msg_iovlen) ||
Al Viro08adb7d2014-11-10 20:23:13 -05001835 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1836 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1837 __get_user(kmsg->msg_flags, &umsg->msg_flags))
Dan Carpenter1661bf32013-10-03 00:27:20 +03001838 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001839
Al Viro08adb7d2014-11-10 20:23:13 -05001840 if (!uaddr)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07001841 kmsg->msg_namelen = 0;
1842
Matthew Leachdbb490b2014-03-11 11:58:27 +00001843 if (kmsg->msg_namelen < 0)
1844 return -EINVAL;
1845
Dan Carpenter1661bf32013-10-03 00:27:20 +03001846 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001847 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05001848
1849 if (save_addr)
1850 *save_addr = uaddr;
1851
1852 if (uaddr && kmsg->msg_namelen) {
1853 if (!save_addr) {
1854 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1855 kmsg->msg_name);
1856 if (err < 0)
1857 return err;
1858 }
1859 } else {
1860 kmsg->msg_name = NULL;
1861 kmsg->msg_namelen = 0;
1862 }
1863
Al Viroc0371da2014-11-24 10:42:55 -05001864 if (nr_segs > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05001865 return -EMSGSIZE;
1866
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001867 kmsg->msg_iocb = NULL;
1868
Al Viroda184282015-03-21 19:29:06 -04001869 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1870 UIO_FASTIOV, iov, &kmsg->msg_iter);
Dan Carpenter1661bf32013-10-03 00:27:20 +03001871}
1872
Al Viro666547f2014-04-06 14:03:05 -04001873static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001874 struct msghdr *msg_sys, unsigned int flags,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001875 struct used_address *used_address)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001877 struct compat_msghdr __user *msg_compat =
1878 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001879 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001881 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001882 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1883 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05001885 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05001886 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001887
Al Viro08adb7d2014-11-10 20:23:13 -05001888 msg_sys->msg_name = &address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889
Al Viro08449322014-11-09 22:33:45 -05001890 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05001891 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
Al Viro08449322014-11-09 22:33:45 -05001892 else
Al Viro08adb7d2014-11-10 20:23:13 -05001893 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001894 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04001895 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896
1897 err = -ENOBUFS;
1898
Anton Blanchard228e5482011-05-02 20:21:35 +00001899 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001901 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001903 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001904 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001905 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 if (err)
1907 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001908 ctl_buf = msg_sys->msg_control;
1909 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001911 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001913 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914 goto out_freeiov;
1915 }
1916 err = -EFAULT;
1917 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00001918 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1920 * checking falls down on this.
1921 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001922 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00001923 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001924 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00001926 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 }
Anton Blanchard228e5482011-05-02 20:21:35 +00001928 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929
1930 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00001931 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001932 /*
1933 * If this is sendmmsg() and current destination address is same as
1934 * previously succeeded address, omit asking LSM's decision.
1935 * used_address->name_len is initialized to UINT_MAX so that the first
1936 * destination address never matches.
1937 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001938 if (used_address && msg_sys->msg_name &&
1939 used_address->name_len == msg_sys->msg_namelen &&
1940 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001941 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05001942 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001943 goto out_freectl;
1944 }
Al Virod8725c82014-12-11 00:02:50 -05001945 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001946 /*
1947 * If this is sendmmsg() and sending to current destination address was
1948 * successful, remember it.
1949 */
1950 if (used_address && err >= 0) {
1951 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001952 if (msg_sys->msg_name)
1953 memcpy(&used_address->name, msg_sys->msg_name,
1954 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001955 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956
1957out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001958 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1960out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04001961 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00001962 return err;
1963}
1964
1965/*
1966 * BSD sendmsg interface
1967 */
1968
Al Viro666547f2014-04-06 14:03:05 -04001969long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00001970{
1971 int fput_needed, err;
1972 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07001973 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00001974
Andy Lutomirski1be374a2013-05-22 14:07:44 -07001975 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00001976 if (!sock)
1977 goto out;
1978
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001979 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL);
Anton Blanchard228e5482011-05-02 20:21:35 +00001980
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001981 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001982out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 return err;
1984}
1985
Al Viro666547f2014-04-06 14:03:05 -04001986SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001987{
1988 if (flags & MSG_CMSG_COMPAT)
1989 return -EINVAL;
1990 return __sys_sendmsg(fd, msg, flags);
1991}
1992
Anton Blanchard228e5482011-05-02 20:21:35 +00001993/*
1994 * Linux sendmmsg interface
1995 */
1996
1997int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1998 unsigned int flags)
1999{
2000 int fput_needed, err, datagrams;
2001 struct socket *sock;
2002 struct mmsghdr __user *entry;
2003 struct compat_mmsghdr __user *compat_entry;
2004 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002005 struct used_address used_address;
Anton Blanchard228e5482011-05-02 20:21:35 +00002006
Anton Blanchard98382f42011-08-04 14:07:39 +00002007 if (vlen > UIO_MAXIOV)
2008 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002009
2010 datagrams = 0;
2011
2012 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2013 if (!sock)
2014 return err;
2015
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002016 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002017 entry = mmsg;
2018 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002019 err = 0;
Anton Blanchard228e5482011-05-02 20:21:35 +00002020
2021 while (datagrams < vlen) {
Anton Blanchard228e5482011-05-02 20:21:35 +00002022 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002023 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002024 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002025 if (err < 0)
2026 break;
2027 err = __put_user(err, &compat_entry->msg_len);
2028 ++compat_entry;
2029 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002030 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002031 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002032 &msg_sys, flags, &used_address);
Anton Blanchard228e5482011-05-02 20:21:35 +00002033 if (err < 0)
2034 break;
2035 err = put_user(err, &entry->msg_len);
2036 ++entry;
2037 }
2038
2039 if (err)
2040 break;
2041 ++datagrams;
2042 }
2043
Anton Blanchard228e5482011-05-02 20:21:35 +00002044 fput_light(sock->file, fput_needed);
2045
Anton Blanchard728ffb82011-08-04 14:07:38 +00002046 /* We only return an error if no datagrams were able to be sent */
2047 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002048 return datagrams;
2049
Anton Blanchard228e5482011-05-02 20:21:35 +00002050 return err;
2051}
2052
2053SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2054 unsigned int, vlen, unsigned int, flags)
2055{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002056 if (flags & MSG_CMSG_COMPAT)
2057 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002058 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2059}
2060
Al Viro666547f2014-04-06 14:03:05 -04002061static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002062 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002064 struct compat_msghdr __user *msg_compat =
2065 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002067 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068 unsigned long cmsg_ptr;
Al Viro08adb7d2014-11-10 20:23:13 -05002069 int total_len, len;
2070 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071
2072 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002073 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074
2075 /* user mode address pointers */
2076 struct sockaddr __user *uaddr;
Al Viro08adb7d2014-11-10 20:23:13 -05002077 int __user *uaddr_len = COMPAT_NAMELEN(msg);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002078
Al Viro08adb7d2014-11-10 20:23:13 -05002079 msg_sys->msg_name = &addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002081 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002082 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002083 else
Al Viro08adb7d2014-11-10 20:23:13 -05002084 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04002086 return err;
2087 total_len = iov_iter_count(&msg_sys->msg_iter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002089 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2090 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002091
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002092 /* We assume all kernel code knows the size of sockaddr_storage */
2093 msg_sys->msg_namelen = 0;
2094
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 if (sock->file->f_flags & O_NONBLOCK)
2096 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002097 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2098 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 if (err < 0)
2100 goto out_freeiov;
2101 len = err;
2102
2103 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002104 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002105 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002106 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 if (err < 0)
2108 goto out_freeiov;
2109 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002110 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002111 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 if (err)
2113 goto out_freeiov;
2114 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002115 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 &msg_compat->msg_controllen);
2117 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002118 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 &msg->msg_controllen);
2120 if (err)
2121 goto out_freeiov;
2122 err = len;
2123
2124out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04002125 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002126 return err;
2127}
2128
2129/*
2130 * BSD recvmsg interface
2131 */
2132
Al Viro666547f2014-04-06 14:03:05 -04002133long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002134{
2135 int fput_needed, err;
2136 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002137 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002138
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002139 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002140 if (!sock)
2141 goto out;
2142
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002143 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002144
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002145 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146out:
2147 return err;
2148}
2149
Al Viro666547f2014-04-06 14:03:05 -04002150SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002151 unsigned int, flags)
2152{
2153 if (flags & MSG_CMSG_COMPAT)
2154 return -EINVAL;
2155 return __sys_recvmsg(fd, msg, flags);
2156}
2157
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002158/*
2159 * Linux recvmmsg interface
2160 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002162int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2163 unsigned int flags, struct timespec *timeout)
2164{
2165 int fput_needed, err, datagrams;
2166 struct socket *sock;
2167 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002168 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002169 struct msghdr msg_sys;
2170 struct timespec end_time;
2171
2172 if (timeout &&
2173 poll_select_set_timeout(&end_time, timeout->tv_sec,
2174 timeout->tv_nsec))
2175 return -EINVAL;
2176
2177 datagrams = 0;
2178
2179 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2180 if (!sock)
2181 return err;
2182
2183 err = sock_error(sock->sk);
2184 if (err)
2185 goto out_put;
2186
2187 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002188 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002189
2190 while (datagrams < vlen) {
2191 /*
2192 * No need to ask LSM for more than the first datagram.
2193 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002194 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002195 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002196 &msg_sys, flags & ~MSG_WAITFORONE,
2197 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002198 if (err < 0)
2199 break;
2200 err = __put_user(err, &compat_entry->msg_len);
2201 ++compat_entry;
2202 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002203 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002204 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002205 &msg_sys, flags & ~MSG_WAITFORONE,
2206 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002207 if (err < 0)
2208 break;
2209 err = put_user(err, &entry->msg_len);
2210 ++entry;
2211 }
2212
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002213 if (err)
2214 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002215 ++datagrams;
2216
Brandon L Black71c5c152010-03-26 16:18:03 +00002217 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2218 if (flags & MSG_WAITFORONE)
2219 flags |= MSG_DONTWAIT;
2220
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002221 if (timeout) {
2222 ktime_get_ts(timeout);
2223 *timeout = timespec_sub(end_time, *timeout);
2224 if (timeout->tv_sec < 0) {
2225 timeout->tv_sec = timeout->tv_nsec = 0;
2226 break;
2227 }
2228
2229 /* Timeout, return less than vlen datagrams */
2230 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2231 break;
2232 }
2233
2234 /* Out of band data, return right away */
2235 if (msg_sys.msg_flags & MSG_OOB)
2236 break;
2237 }
2238
2239out_put:
2240 fput_light(sock->file, fput_needed);
2241
2242 if (err == 0)
2243 return datagrams;
2244
2245 if (datagrams != 0) {
2246 /*
2247 * We may return less entries than requested (vlen) if the
2248 * sock is non block and there aren't enough datagrams...
2249 */
2250 if (err != -EAGAIN) {
2251 /*
2252 * ... or if recvmsg returns an error after we
2253 * received some datagrams, where we record the
2254 * error to return on the next call or if the
2255 * app asks about it using getsockopt(SO_ERROR).
2256 */
2257 sock->sk->sk_err = -err;
2258 }
2259
2260 return datagrams;
2261 }
2262
2263 return err;
2264}
2265
2266SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2267 unsigned int, vlen, unsigned int, flags,
2268 struct timespec __user *, timeout)
2269{
2270 int datagrams;
2271 struct timespec timeout_sys;
2272
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002273 if (flags & MSG_CMSG_COMPAT)
2274 return -EINVAL;
2275
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002276 if (!timeout)
2277 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2278
2279 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2280 return -EFAULT;
2281
2282 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2283
2284 if (datagrams > 0 &&
2285 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2286 datagrams = -EFAULT;
2287
2288 return datagrams;
2289}
2290
2291#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292/* Argument list sizes for sys_socketcall */
2293#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002294static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002295 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2296 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2297 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002298 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002299};
2300
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301#undef AL
2302
2303/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002304 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305 *
2306 * Argument checking cleaned up. Saved 20% in size.
2307 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002308 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 */
2310
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002311SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002312{
Chen Gang2950fa92013-04-07 16:55:23 +08002313 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002314 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002316 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317
Anton Blanchard228e5482011-05-02 20:21:35 +00002318 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319 return -EINVAL;
2320
Arjan van de Ven47379052009-09-28 12:57:44 -07002321 len = nargs[call];
2322 if (len > sizeof(a))
2323 return -EINVAL;
2324
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002326 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002328
Chen Gang2950fa92013-04-07 16:55:23 +08002329 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2330 if (err)
2331 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002332
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002333 a0 = a[0];
2334 a1 = a[1];
2335
2336 switch (call) {
2337 case SYS_SOCKET:
2338 err = sys_socket(a0, a1, a[2]);
2339 break;
2340 case SYS_BIND:
2341 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2342 break;
2343 case SYS_CONNECT:
2344 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2345 break;
2346 case SYS_LISTEN:
2347 err = sys_listen(a0, a1);
2348 break;
2349 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002350 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2351 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002352 break;
2353 case SYS_GETSOCKNAME:
2354 err =
2355 sys_getsockname(a0, (struct sockaddr __user *)a1,
2356 (int __user *)a[2]);
2357 break;
2358 case SYS_GETPEERNAME:
2359 err =
2360 sys_getpeername(a0, (struct sockaddr __user *)a1,
2361 (int __user *)a[2]);
2362 break;
2363 case SYS_SOCKETPAIR:
2364 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2365 break;
2366 case SYS_SEND:
2367 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2368 break;
2369 case SYS_SENDTO:
2370 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2371 (struct sockaddr __user *)a[4], a[5]);
2372 break;
2373 case SYS_RECV:
2374 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2375 break;
2376 case SYS_RECVFROM:
2377 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2378 (struct sockaddr __user *)a[4],
2379 (int __user *)a[5]);
2380 break;
2381 case SYS_SHUTDOWN:
2382 err = sys_shutdown(a0, a1);
2383 break;
2384 case SYS_SETSOCKOPT:
2385 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2386 break;
2387 case SYS_GETSOCKOPT:
2388 err =
2389 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2390 (int __user *)a[4]);
2391 break;
2392 case SYS_SENDMSG:
Al Viro666547f2014-04-06 14:03:05 -04002393 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002394 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002395 case SYS_SENDMMSG:
2396 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2397 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002398 case SYS_RECVMSG:
Al Viro666547f2014-04-06 14:03:05 -04002399 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002400 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002401 case SYS_RECVMMSG:
2402 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2403 (struct timespec __user *)a[4]);
2404 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002405 case SYS_ACCEPT4:
2406 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2407 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002408 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002409 default:
2410 err = -EINVAL;
2411 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412 }
2413 return err;
2414}
2415
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002416#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002417
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002418/**
2419 * sock_register - add a socket protocol handler
2420 * @ops: description of protocol
2421 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422 * This function is called by a protocol handler that wants to
2423 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002424 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002425 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002427int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428{
2429 int err;
2430
2431 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002432 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433 return -ENOBUFS;
2434 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002435
2436 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002437 if (rcu_dereference_protected(net_families[ops->family],
2438 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002439 err = -EEXIST;
2440 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002441 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 err = 0;
2443 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002444 spin_unlock(&net_family_lock);
2445
Yang Yingliang3410f222014-02-12 17:09:55 +08002446 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447 return err;
2448}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002449EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002451/**
2452 * sock_unregister - remove a protocol handler
2453 * @family: protocol family to remove
2454 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 * This function is called by a protocol handler that wants to
2456 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002457 * new socket creation.
2458 *
2459 * If protocol handler is a module, then it can use module reference
2460 * counts to protect against new references. If protocol handler is not
2461 * a module then it needs to provide its own protection in
2462 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002464void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002466 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002468 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002469 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002470 spin_unlock(&net_family_lock);
2471
2472 synchronize_rcu();
2473
Yang Yingliang3410f222014-02-12 17:09:55 +08002474 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002476EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477
Andi Kleen77d76ea2005-12-22 12:43:42 -08002478static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002480 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002481 /*
2482 * Initialize the network sysctl infrastructure.
2483 */
2484 err = net_sysctl_init();
2485 if (err)
2486 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002487
Linus Torvalds1da177e2005-04-16 15:20:36 -07002488 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002489 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002490 */
2491 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002492
2493 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002494 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495 */
2496
2497 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002498
2499 err = register_filesystem(&sock_fs_type);
2500 if (err)
2501 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002502 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002503 if (IS_ERR(sock_mnt)) {
2504 err = PTR_ERR(sock_mnt);
2505 goto out_mount;
2506 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002507
2508 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509 */
2510
2511#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002512 err = netfilter_init();
2513 if (err)
2514 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002515#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002516
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002517 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002518
Nick Pigginb3e19d92011-01-07 17:50:11 +11002519out:
2520 return err;
2521
2522out_mount:
2523 unregister_filesystem(&sock_fs_type);
2524out_fs:
2525 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002526}
2527
Andi Kleen77d76ea2005-12-22 12:43:42 -08002528core_initcall(sock_init); /* early initcall */
2529
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530#ifdef CONFIG_PROC_FS
2531void socket_seq_show(struct seq_file *seq)
2532{
2533 int cpu;
2534 int counter = 0;
2535
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002536 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002537 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
2539 /* It can be negative, by the way. 8) */
2540 if (counter < 0)
2541 counter = 0;
2542
2543 seq_printf(seq, "sockets: used %d\n", counter);
2544}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002545#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002547#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002548static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002549 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002550{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002551 mm_segment_t old_fs = get_fs();
2552 struct timeval ktv;
2553 int err;
2554
2555 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002556 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002557 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002558 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002559 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002560
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002561 return err;
2562}
2563
Arnd Bergmann6b960182009-11-06 23:10:54 -08002564static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002565 unsigned int cmd, void __user *up)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002566{
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002567 mm_segment_t old_fs = get_fs();
2568 struct timespec kts;
2569 int err;
2570
2571 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002572 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002573 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002574 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002575 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002576
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002577 return err;
2578}
2579
Arnd Bergmann6b960182009-11-06 23:10:54 -08002580static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002581{
2582 struct ifreq __user *uifr;
2583 int err;
2584
2585 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002586 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002587 return -EFAULT;
2588
Arnd Bergmann6b960182009-11-06 23:10:54 -08002589 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002590 if (err)
2591 return err;
2592
Arnd Bergmann6b960182009-11-06 23:10:54 -08002593 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002594 return -EFAULT;
2595
2596 return 0;
2597}
2598
Arnd Bergmann6b960182009-11-06 23:10:54 -08002599static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002600{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002601 struct compat_ifconf ifc32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002602 struct ifconf ifc;
2603 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002604 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002605 struct ifreq __user *ifr;
2606 unsigned int i, j;
2607 int err;
2608
Arnd Bergmann6b960182009-11-06 23:10:54 -08002609 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002610 return -EFAULT;
2611
Mathias Krause43da5f22012-08-15 11:31:57 +00002612 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002613 if (ifc32.ifcbuf == 0) {
2614 ifc32.ifc_len = 0;
2615 ifc.ifc_len = 0;
2616 ifc.ifc_req = NULL;
2617 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2618 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002619 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2620 sizeof(struct ifreq);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002621 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2622 ifc.ifc_len = len;
2623 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2624 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002625 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002626 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002627 return -EFAULT;
2628 ifr++;
2629 ifr32++;
2630 }
2631 }
2632 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2633 return -EFAULT;
2634
Arnd Bergmann6b960182009-11-06 23:10:54 -08002635 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002636 if (err)
2637 return err;
2638
2639 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2640 return -EFAULT;
2641
2642 ifr = ifc.ifc_req;
2643 ifr32 = compat_ptr(ifc32.ifcbuf);
2644 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002645 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2646 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2647 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002648 return -EFAULT;
2649 ifr32++;
2650 ifr++;
2651 }
2652
2653 if (ifc32.ifcbuf == 0) {
2654 /* Translate from 64-bit structure multiple to
2655 * a 32-bit one.
2656 */
2657 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002658 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002659 ifc32.ifc_len = i;
2660 } else {
2661 ifc32.ifc_len = i;
2662 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002663 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002664 return -EFAULT;
2665
2666 return 0;
2667}
2668
Arnd Bergmann6b960182009-11-06 23:10:54 -08002669static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002670{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002671 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2672 bool convert_in = false, convert_out = false;
2673 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2674 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002675 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002676 u32 rule_cnt = 0, actual_rule_cnt;
2677 u32 ethcmd;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002678 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002679 int ret;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002680
2681 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2682 return -EFAULT;
2683
Ben Hutchings3a7da392011-03-17 07:34:32 +00002684 compat_rxnfc = compat_ptr(data);
2685
2686 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002687 return -EFAULT;
2688
Ben Hutchings3a7da392011-03-17 07:34:32 +00002689 /* Most ethtool structures are defined without padding.
2690 * Unfortunately struct ethtool_rxnfc is an exception.
2691 */
2692 switch (ethcmd) {
2693 default:
2694 break;
2695 case ETHTOOL_GRXCLSRLALL:
2696 /* Buffer size is variable */
2697 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2698 return -EFAULT;
2699 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2700 return -ENOMEM;
2701 buf_size += rule_cnt * sizeof(u32);
2702 /* fall through */
2703 case ETHTOOL_GRXRINGS:
2704 case ETHTOOL_GRXCLSRLCNT:
2705 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002706 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002707 convert_out = true;
2708 /* fall through */
2709 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002710 buf_size += sizeof(struct ethtool_rxnfc);
2711 convert_in = true;
2712 break;
2713 }
2714
2715 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002716 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002717
2718 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2719 return -EFAULT;
2720
2721 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2722 &ifr->ifr_ifru.ifru_data))
2723 return -EFAULT;
2724
2725 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002726 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002727 * fs.ring_cookie and at the end of fs, but nowhere else.
2728 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002729 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2730 sizeof(compat_rxnfc->fs.m_ext) !=
2731 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2732 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002733 BUILD_BUG_ON(
2734 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2735 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2736 offsetof(struct ethtool_rxnfc, fs.location) -
2737 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2738
2739 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002740 (void __user *)(&rxnfc->fs.m_ext + 1) -
2741 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002742 copy_in_user(&rxnfc->fs.ring_cookie,
2743 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002744 (void __user *)(&rxnfc->fs.location + 1) -
2745 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002746 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2747 sizeof(rxnfc->rule_cnt)))
2748 return -EFAULT;
2749 }
2750
2751 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2752 if (ret)
2753 return ret;
2754
2755 if (convert_out) {
2756 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002757 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2758 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002759 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2760 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002761 (const void __user *)(&rxnfc->fs.location + 1) -
2762 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002763 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2764 sizeof(rxnfc->rule_cnt)))
2765 return -EFAULT;
2766
2767 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2768 /* As an optimisation, we only copy the actual
2769 * number of rules that the underlying
2770 * function returned. Since Mallory might
2771 * change the rule count in user memory, we
2772 * check that it is less than the rule count
2773 * originally given (as the user buffer size),
2774 * which has been range-checked.
2775 */
2776 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2777 return -EFAULT;
2778 if (actual_rule_cnt < rule_cnt)
2779 rule_cnt = actual_rule_cnt;
2780 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2781 &rxnfc->rule_locs[0],
2782 rule_cnt * sizeof(u32)))
2783 return -EFAULT;
2784 }
2785 }
2786
2787 return 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002788}
2789
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002790static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2791{
2792 void __user *uptr;
2793 compat_uptr_t uptr32;
2794 struct ifreq __user *uifr;
2795
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002796 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002797 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2798 return -EFAULT;
2799
2800 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2801 return -EFAULT;
2802
2803 uptr = compat_ptr(uptr32);
2804
2805 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2806 return -EFAULT;
2807
2808 return dev_ioctl(net, SIOCWANDEV, uifr);
2809}
2810
Arnd Bergmann6b960182009-11-06 23:10:54 -08002811static int bond_ioctl(struct net *net, unsigned int cmd,
2812 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002813{
2814 struct ifreq kifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002815 mm_segment_t old_fs;
2816 int err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002817
2818 switch (cmd) {
2819 case SIOCBONDENSLAVE:
2820 case SIOCBONDRELEASE:
2821 case SIOCBONDSETHWADDR:
2822 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002823 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002824 return -EFAULT;
2825
2826 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002827 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002828 err = dev_ioctl(net, cmd,
2829 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002830 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002831
2832 return err;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002833 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002834 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002835 }
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002836}
2837
Ben Hutchings590d4692013-11-18 17:04:13 +00002838/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2839static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08002840 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002841{
2842 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002843 char tmp_buf[IFNAMSIZ];
2844 void __user *data64;
2845 u32 data32;
2846
2847 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2848 IFNAMSIZ))
2849 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002850 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002851 return -EFAULT;
2852 data64 = compat_ptr(data32);
2853
2854 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2855
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002856 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2857 IFNAMSIZ))
2858 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002859 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002860 return -EFAULT;
2861
Arnd Bergmann6b960182009-11-06 23:10:54 -08002862 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002863}
2864
Arnd Bergmann6b960182009-11-06 23:10:54 -08002865static int dev_ifsioc(struct net *net, struct socket *sock,
2866 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002867{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002868 struct ifreq __user *uifr;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002869 int err;
2870
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002871 uifr = compat_alloc_user_space(sizeof(*uifr));
2872 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2873 return -EFAULT;
2874
2875 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2876
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002877 if (!err) {
2878 switch (cmd) {
2879 case SIOCGIFFLAGS:
2880 case SIOCGIFMETRIC:
2881 case SIOCGIFMTU:
2882 case SIOCGIFMEM:
2883 case SIOCGIFHWADDR:
2884 case SIOCGIFINDEX:
2885 case SIOCGIFADDR:
2886 case SIOCGIFBRDADDR:
2887 case SIOCGIFDSTADDR:
2888 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002889 case SIOCGIFPFLAGS:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002890 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002891 case SIOCGMIIPHY:
2892 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002893 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002894 err = -EFAULT;
2895 break;
2896 }
2897 }
2898 return err;
2899}
2900
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002901static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2902 struct compat_ifreq __user *uifr32)
2903{
2904 struct ifreq ifr;
2905 struct compat_ifmap __user *uifmap32;
2906 mm_segment_t old_fs;
2907 int err;
2908
2909 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2910 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002911 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2912 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2913 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2914 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2915 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2916 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002917 if (err)
2918 return -EFAULT;
2919
2920 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002921 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002922 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002923 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002924
2925 if (cmd == SIOCGIFMAP && !err) {
2926 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002927 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2928 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2929 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2930 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2931 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2932 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002933 if (err)
2934 err = -EFAULT;
2935 }
2936 return err;
2937}
2938
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002939struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002940 u32 rt_pad1;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002941 struct sockaddr rt_dst; /* target address */
2942 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2943 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002944 unsigned short rt_flags;
2945 short rt_pad2;
2946 u32 rt_pad3;
2947 unsigned char rt_tos;
2948 unsigned char rt_class;
2949 short rt_pad4;
2950 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002951 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002952 u32 rt_mtu; /* per route MTU/Window */
2953 u32 rt_window; /* Window clamping */
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002954 unsigned short rt_irtt; /* Initial RTT */
2955};
2956
2957struct in6_rtmsg32 {
2958 struct in6_addr rtmsg_dst;
2959 struct in6_addr rtmsg_src;
2960 struct in6_addr rtmsg_gateway;
2961 u32 rtmsg_type;
2962 u16 rtmsg_dst_len;
2963 u16 rtmsg_src_len;
2964 u32 rtmsg_metric;
2965 u32 rtmsg_info;
2966 u32 rtmsg_flags;
2967 s32 rtmsg_ifindex;
2968};
2969
Arnd Bergmann6b960182009-11-06 23:10:54 -08002970static int routing_ioctl(struct net *net, struct socket *sock,
2971 unsigned int cmd, void __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002972{
2973 int ret;
2974 void *r = NULL;
2975 struct in6_rtmsg r6;
2976 struct rtentry r4;
2977 char devname[16];
2978 u32 rtdev;
2979 mm_segment_t old_fs = get_fs();
2980
Arnd Bergmann6b960182009-11-06 23:10:54 -08002981 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2982 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002983 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002984 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002985 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2986 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2987 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2988 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2989 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2990 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2991 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002992
2993 r = (void *) &r6;
2994 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002995 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002996 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a2293872009-11-06 23:00:29 -08002997 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002998 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
2999 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3000 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3001 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3002 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3003 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003004 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003005 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003006 r4.rt_dev = (char __user __force *)devname;
3007 devname[15] = 0;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003008 } else
3009 r4.rt_dev = NULL;
3010
3011 r = (void *) &r4;
3012 }
3013
3014 if (ret) {
3015 ret = -EFAULT;
3016 goto out;
3017 }
3018
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003019 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003020 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003021 set_fs(old_fs);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003022
3023out:
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003024 return ret;
3025}
3026
3027/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3028 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003029 * use compatible ioctls
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003030 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003031static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003032{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003033 compat_ulong_t tmp;
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003034
Arnd Bergmann6b960182009-11-06 23:10:54 -08003035 if (get_user(tmp, argp))
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003036 return -EFAULT;
3037 if (tmp == BRCTL_GET_VERSION)
3038 return BRCTL_VERSION + 1;
3039 return -EINVAL;
3040}
3041
Arnd Bergmann6b960182009-11-06 23:10:54 -08003042static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3043 unsigned int cmd, unsigned long arg)
3044{
3045 void __user *argp = compat_ptr(arg);
3046 struct sock *sk = sock->sk;
3047 struct net *net = sock_net(sk);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003048
Arnd Bergmann6b960182009-11-06 23:10:54 -08003049 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003050 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003051
Arnd Bergmann6b960182009-11-06 23:10:54 -08003052 switch (cmd) {
3053 case SIOCSIFBR:
3054 case SIOCGIFBR:
3055 return old_bridge_ioctl(argp);
3056 case SIOCGIFNAME:
3057 return dev_ifname32(net, argp);
3058 case SIOCGIFCONF:
3059 return dev_ifconf(net, argp);
3060 case SIOCETHTOOL:
3061 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003062 case SIOCWANDEV:
3063 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003064 case SIOCGIFMAP:
3065 case SIOCSIFMAP:
3066 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003067 case SIOCBONDENSLAVE:
3068 case SIOCBONDRELEASE:
3069 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003070 case SIOCBONDCHANGEACTIVE:
3071 return bond_ioctl(net, cmd, argp);
3072 case SIOCADDRT:
3073 case SIOCDELRT:
3074 return routing_ioctl(net, sock, cmd, argp);
3075 case SIOCGSTAMP:
3076 return do_siocgstamp(net, sock, cmd, argp);
3077 case SIOCGSTAMPNS:
3078 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003079 case SIOCBONDSLAVEINFOQUERY:
3080 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003081 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003082 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003083 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003084
Arnd Bergmann6b960182009-11-06 23:10:54 -08003085 case FIOSETOWN:
3086 case SIOCSPGRP:
3087 case FIOGETOWN:
3088 case SIOCGPGRP:
3089 case SIOCBRADDBR:
3090 case SIOCBRDELBR:
3091 case SIOCGIFVLAN:
3092 case SIOCSIFVLAN:
3093 case SIOCADDDLCI:
3094 case SIOCDELDLCI:
3095 return sock_ioctl(file, cmd, arg);
3096
3097 case SIOCGIFFLAGS:
3098 case SIOCSIFFLAGS:
3099 case SIOCGIFMETRIC:
3100 case SIOCSIFMETRIC:
3101 case SIOCGIFMTU:
3102 case SIOCSIFMTU:
3103 case SIOCGIFMEM:
3104 case SIOCSIFMEM:
3105 case SIOCGIFHWADDR:
3106 case SIOCSIFHWADDR:
3107 case SIOCADDMULTI:
3108 case SIOCDELMULTI:
3109 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003110 case SIOCGIFADDR:
3111 case SIOCSIFADDR:
3112 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003113 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003114 case SIOCGIFBRDADDR:
3115 case SIOCSIFBRDADDR:
3116 case SIOCGIFDSTADDR:
3117 case SIOCSIFDSTADDR:
3118 case SIOCGIFNETMASK:
3119 case SIOCSIFNETMASK:
3120 case SIOCSIFPFLAGS:
3121 case SIOCGIFPFLAGS:
3122 case SIOCGIFTXQLEN:
3123 case SIOCSIFTXQLEN:
3124 case SIOCBRADDIF:
3125 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003126 case SIOCSIFNAME:
3127 case SIOCGMIIPHY:
3128 case SIOCGMIIREG:
3129 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003130 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003131
Arnd Bergmann6b960182009-11-06 23:10:54 -08003132 case SIOCSARP:
3133 case SIOCGARP:
3134 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003135 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003136 return sock_do_ioctl(net, sock, cmd, arg);
3137 }
3138
Arnd Bergmann6b960182009-11-06 23:10:54 -08003139 return -ENOIOCTLCMD;
3140}
Arnd Bergmann7a2293872009-11-06 23:00:29 -08003141
Eric Dumazet95c96172012-04-15 05:58:06 +00003142static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003143 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003144{
3145 struct socket *sock = file->private_data;
3146 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003147 struct sock *sk;
3148 struct net *net;
3149
3150 sk = sock->sk;
3151 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003152
3153 if (sock->ops->compat_ioctl)
3154 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3155
David S. Miller87de87d2008-06-03 09:14:03 -07003156 if (ret == -ENOIOCTLCMD &&
3157 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3158 ret = compat_wext_handle_ioctl(net, cmd, arg);
3159
Arnd Bergmann6b960182009-11-06 23:10:54 -08003160 if (ret == -ENOIOCTLCMD)
3161 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3162
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003163 return ret;
3164}
3165#endif
3166
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003167int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3168{
3169 return sock->ops->bind(sock, addr, addrlen);
3170}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003171EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003172
3173int kernel_listen(struct socket *sock, int backlog)
3174{
3175 return sock->ops->listen(sock, backlog);
3176}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003177EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003178
3179int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3180{
3181 struct sock *sk = sock->sk;
3182 int err;
3183
3184 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3185 newsock);
3186 if (err < 0)
3187 goto done;
3188
3189 err = sock->ops->accept(sock, *newsock, flags);
3190 if (err < 0) {
3191 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003192 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003193 goto done;
3194 }
3195
3196 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003197 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003198
3199done:
3200 return err;
3201}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003202EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003203
3204int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003205 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003206{
3207 return sock->ops->connect(sock, addr, addrlen, flags);
3208}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003209EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003210
3211int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3212 int *addrlen)
3213{
3214 return sock->ops->getname(sock, addr, addrlen, 0);
3215}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003216EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003217
3218int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3219 int *addrlen)
3220{
3221 return sock->ops->getname(sock, addr, addrlen, 1);
3222}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003223EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003224
3225int kernel_getsockopt(struct socket *sock, int level, int optname,
3226 char *optval, int *optlen)
3227{
3228 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003229 char __user *uoptval;
3230 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003231 int err;
3232
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003233 uoptval = (char __user __force *) optval;
3234 uoptlen = (int __user __force *) optlen;
3235
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003236 set_fs(KERNEL_DS);
3237 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003238 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003239 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003240 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3241 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003242 set_fs(oldfs);
3243 return err;
3244}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003245EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003246
3247int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003248 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003249{
3250 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003251 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003252 int err;
3253
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003254 uoptval = (char __user __force *) optval;
3255
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003256 set_fs(KERNEL_DS);
3257 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003258 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003259 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003260 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003261 optlen);
3262 set_fs(oldfs);
3263 return err;
3264}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003265EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003266
3267int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3268 size_t size, int flags)
3269{
3270 if (sock->ops->sendpage)
3271 return sock->ops->sendpage(sock, page, offset, size, flags);
3272
3273 return sock_no_sendpage(sock, page, offset, size, flags);
3274}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003275EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003276
3277int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3278{
3279 mm_segment_t oldfs = get_fs();
3280 int err;
3281
3282 set_fs(KERNEL_DS);
3283 err = sock->ops->ioctl(sock, cmd, arg);
3284 set_fs(oldfs);
3285
3286 return err;
3287}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003288EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003289
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003290int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3291{
3292 return sock->ops->shutdown(sock, how);
3293}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003294EXPORT_SYMBOL(kernel_sock_shutdown);