blob: 5f77a8e93830bd30cef60e68354bda683c9acc43 [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;
Nicolai Stange574aab12015-12-29 13:29:55 +0100260 wq->flags = 0;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000261 RCU_INIT_POINTER(ei->socket.wq, wq);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700262
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
269 return &ei->vfs_inode;
270}
271
272static void sock_destroy_inode(struct inode *inode)
273{
Eric Dumazet43815482010-04-29 11:01:49 +0000274 struct socket_alloc *ei;
Eric Dumazeteaefd112011-02-18 03:26:36 +0000275 struct socket_wq *wq;
Eric Dumazet43815482010-04-29 11:01:49 +0000276
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazeteaefd112011-02-18 03:26:36 +0000278 wq = rcu_dereference_protected(ei->socket.wq, 1);
Lai Jiangshan61845222011-03-18 12:10:25 +0800279 kfree_rcu(wq, rcu);
Eric Dumazet43815482010-04-29 11:01:49 +0000280 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281}
282
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700283static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700285 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286
Christoph Lametera35afb82007-05-16 22:10:57 -0700287 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700289
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290static int init_inodecache(void)
291{
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700293 sizeof(struct socket_alloc),
294 0,
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
Vladimir Davydov5d097052016-01-14 15:18:21 -0800297 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
Paul Mundt20c2df82007-07-20 10:11:58 +0900298 init_once);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 if (sock_inode_cachep == NULL)
300 return -ENOMEM;
301 return 0;
302}
303
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700304static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700305 .alloc_inode = sock_alloc_inode,
306 .destroy_inode = sock_destroy_inode,
307 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308};
309
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700310/*
311 * sockfs_dname() is called from d_path().
312 */
313static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314{
315 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000316 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700317}
318
Al Viro3ba13d12009-02-20 06:02:22 +0000319static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700320 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321};
322
Al Viroc74a1cb2011-01-12 16:59:34 -0500323static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324 int flags, const char *dev_name, void *data)
325{
326 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327 &sockfs_dentry_operations, SOCKFS_MAGIC);
328}
329
330static struct vfsmount *sock_mnt __read_mostly;
331
332static struct file_system_type sock_fs_type = {
333 .name = "sockfs",
334 .mount = sockfs_mount,
335 .kill_sb = kill_anon_super,
336};
337
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338/*
339 * Obtains the first available file descriptor and sets it up for use.
340 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
348 *
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
353 */
354
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700355struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356{
Al Viro7cbe66b2009-08-05 19:59:08 +0400357 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400358 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400359 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800360
Masatake YAMATO600e1772012-08-29 10:44:29 +0000361 if (dname) {
362 name.name = dname;
363 name.len = strlen(name.name);
364 } else if (sock->sk) {
365 name.name = sock->sk->sk_prot_creator->name;
366 name.len = strlen(name.name);
367 }
Nick Piggin4b936882011-01-07 17:50:07 +1100368 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400369 if (unlikely(!path.dentry))
370 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400371 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372
Al Viro2c48b9c2009-08-09 00:52:35 +0400373 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400374
Al Viro2c48b9c2009-08-09 00:52:35 +0400375 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400376 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530377 if (IS_ERR(file)) {
Al Virocc3808f2009-08-06 09:43:59 +0400378 /* drop dentry, keep inode */
David Howellsc5ef6032015-03-17 22:26:16 +0000379 ihold(d_inode(path.dentry));
Al Viro2c48b9c2009-08-09 00:52:35 +0400380 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700381 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400382 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800383
384 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700385 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800386 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400387 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800388}
Al Viro56b31d12012-08-18 00:25:51 -0400389EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800390
Al Viro56b31d12012-08-18 00:25:51 -0400391static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800392{
393 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400394 int fd = get_unused_fd_flags(flags);
395 if (unlikely(fd < 0))
396 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800397
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700398 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400399 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800400 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400401 return fd;
402 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400403
Al Viro28407632012-08-17 23:54:15 -0400404 put_unused_fd(fd);
405 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406}
407
John Fastabend406a3c62012-07-20 10:39:25 +0000408struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800409{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
412
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800413 *err = -ENOTSOCK;
414 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800415}
John Fastabend406a3c62012-07-20 10:39:25 +0000416EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800417
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700419 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420 * @fd: file handle
421 * @err: pointer to an error code return
422 *
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
427 *
428 * On a success the socket object pointer is returned.
429 */
430
431struct socket *sockfd_lookup(int fd, int *err)
432{
433 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 struct socket *sock;
435
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700436 file = fget(fd);
437 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 *err = -EBADF;
439 return NULL;
440 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700441
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800442 sock = sock_from_file(file, err);
443 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 return sock;
446}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700447EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800449static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
450{
Al Viro00e188e2014-03-03 23:48:18 -0500451 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800452 struct socket *sock;
453
Hua Zhong36725582006-04-19 15:25:02 -0700454 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500455 if (f.file) {
456 sock = sock_from_file(f.file, err);
457 if (likely(sock)) {
458 *fput_needed = f.flags;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800459 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500460 }
461 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800462 }
463 return NULL;
464}
465
Masatake YAMATO600e1772012-08-29 10:44:29 +0000466#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469static ssize_t sockfs_getxattr(struct dentry *dentry,
470 const char *name, void *value, size_t size)
471{
472 const char *proto_name;
473 size_t proto_size;
474 int error;
475
476 error = -ENODATA;
477 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478 proto_name = dentry->d_name.name;
479 proto_size = strlen(proto_name);
480
481 if (value) {
482 error = -ERANGE;
483 if (proto_size + 1 > size)
484 goto out;
485
486 strncpy(value, proto_name, proto_size + 1);
487 }
488 error = proto_size + 1;
489 }
490
491out:
492 return error;
493}
494
495static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496 size_t size)
497{
498 ssize_t len;
499 ssize_t used = 0;
500
David Howellsc5ef6032015-03-17 22:26:16 +0000501 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000502 if (len < 0)
503 return len;
504 used += len;
505 if (buffer) {
506 if (size < used)
507 return -ERANGE;
508 buffer += len;
509 }
510
511 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512 used += len;
513 if (buffer) {
514 if (size < used)
515 return -ERANGE;
516 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517 buffer += len;
518 }
519
520 return used;
521}
522
523static const struct inode_operations sockfs_inode_ops = {
524 .getxattr = sockfs_getxattr,
525 .listxattr = sockfs_listxattr,
526};
527
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528/**
529 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700530 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 * Allocate a new inode and socket object. The two are bound together
532 * and initialised. The socket is then returned. If we are out of inodes
533 * NULL is returned.
534 */
535
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800536struct socket *sock_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700538 struct inode *inode;
539 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200541 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 if (!inode)
543 return NULL;
544
545 sock = SOCKET_I(inode);
546
Eric Dumazet29a020d2009-09-15 02:39:20 -0700547 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400548 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700549 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100550 inode->i_uid = current_fsuid();
551 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000552 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
Alex Shi19e8d692012-05-14 14:15:31 -0700554 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 return sock;
556}
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800557EXPORT_SYMBOL(sock_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559/**
560 * sock_release - close a socket
561 * @sock: socket to close
562 *
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700565 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568void sock_release(struct socket *sock)
569{
570 if (sock->ops) {
571 struct module *owner = sock->ops->owner;
572
573 sock->ops->release(sock);
574 sock->ops = NULL;
575 module_put(owner);
576 }
577
Eric Dumazeteaefd112011-02-18 03:26:36 +0000578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800579 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
Alex Shi19e8d692012-05-14 14:15:31 -0700581 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 if (!sock->file) {
583 iput(SOCK_INODE(sock));
584 return;
585 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700586 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700588EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400590void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000591{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200592 u8 flags = *tx_flags;
593
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400594 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200595 flags |= SKBTX_HW_TSTAMP;
596
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200598 flags |= SKBTX_SW_TSTAMP;
599
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200601 flags |= SKBTX_SCHED_TSTAMP;
602
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200604 flags |= SKBTX_ACK_TSTAMP;
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400605
Eric Dumazet140c55d2014-08-06 11:49:29 +0200606 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000607}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400608EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000609
Al Virod8725c82014-12-11 00:02:50 -0500610static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611{
Al Viro01e97e62014-12-15 21:39:31 -0500612 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500613 BUG_ON(ret == -EIOCBQUEUED);
614 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800616
Al Virod8725c82014-12-11 00:02:50 -0500617int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800618{
Al Virod8725c82014-12-11 00:02:50 -0500619 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500620 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800621
Al Virod8725c82014-12-11 00:02:50 -0500622 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800623}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700624EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
626int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
627 struct kvec *vec, size_t num, size_t size)
628{
Al Viro6aa24812015-03-21 19:56:16 -0400629 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500630 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700632EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700634/*
635 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
636 */
637void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
638 struct sk_buff *skb)
639{
Patrick Ohly20d49472009-02-12 05:03:38 +0000640 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400641 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000642 int empty = 1;
643 struct skb_shared_hwtstamps *shhwtstamps =
644 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700645
Patrick Ohly20d49472009-02-12 05:03:38 +0000646 /* Race occurred between timestamp enabling and packet
647 receiving. Fill in the current time for now. */
648 if (need_software_tstamp && skb->tstamp.tv64 == 0)
649 __net_timestamp(skb);
650
651 if (need_software_tstamp) {
652 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
653 struct timeval tv;
654 skb_get_timestamp(skb, &tv);
655 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
656 sizeof(tv), &tv);
657 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400658 struct timespec ts;
659 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000660 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400661 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000662 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700663 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000664
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400665 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400666 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400667 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000668 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400669 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400670 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400671 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400672 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000673 if (!empty)
674 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400675 SCM_TIMESTAMPING, sizeof(tss), &tss);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700676}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300677EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
678
Johannes Berg6e3e9392011-11-09 10:15:42 +0100679void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
680 struct sk_buff *skb)
681{
682 int ack;
683
684 if (!sock_flag(sk, SOCK_WIFI_STATUS))
685 return;
686 if (!skb->wifi_acked_valid)
687 return;
688
689 ack = skb->wifi_acked;
690
691 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
692}
693EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
694
stephen hemminger11165f12010-10-18 14:27:29 +0000695static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
696 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700697{
Eyal Birger744d5a32015-03-01 14:58:31 +0200698 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700699 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200700 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700701}
702
Eric Dumazet767dd032010-04-28 19:14:43 +0000703void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700704 struct sk_buff *skb)
705{
706 sock_recv_timestamp(msg, sk, skb);
707 sock_recv_drops(msg, sk, skb);
708}
Eric Dumazet767dd032010-04-28 19:14:43 +0000709EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700710
Ying Xue1b784142015-03-02 15:37:48 +0800711static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
712 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713{
Ying Xue1b784142015-03-02 15:37:48 +0800714 return sock->ops->recvmsg(sock, msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715}
716
Ying Xue1b784142015-03-02 15:37:48 +0800717int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
718 int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700719{
720 int err = security_socket_recvmsg(sock, msg, size, flags);
721
Ying Xue1b784142015-03-02 15:37:48 +0800722 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700724EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725
Martin Lucinac1249c02010-12-10 00:04:05 +0000726/**
727 * kernel_recvmsg - Receive a message from a socket (kernel space)
728 * @sock: The socket to receive the message from
729 * @msg: Received message
730 * @vec: Input s/g array for message data
731 * @num: Size of input s/g array
732 * @size: Number of bytes to read
733 * @flags: Message flags (MSG_DONTWAIT, etc...)
734 *
735 * On return the msg structure contains the scatter/gather array passed in the
736 * vec argument. The array is modified so that it consists of the unfilled
737 * portion of the original array.
738 *
739 * The returned value is the total number of bytes received, or an error.
740 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700741int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
742 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743{
744 mm_segment_t oldfs = get_fs();
745 int result;
746
Al Viro6aa24812015-03-21 19:56:16 -0400747 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748 set_fs(KERNEL_DS);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 result = sock_recvmsg(sock, msg, size, flags);
750 set_fs(oldfs);
751 return result;
752}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700753EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300755static ssize_t sock_sendpage(struct file *file, struct page *page,
756 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757{
758 struct socket *sock;
759 int flags;
760
Eric Dumazetb69aee02005-09-06 14:42:45 -0700761 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
Eric Dumazet35f9c092012-04-05 03:05:35 +0000763 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
764 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
765 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
Linus Torvaldse6949582009-08-13 08:28:36 -0700767 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768}
769
Jens Axboe9c55e012007-11-06 23:30:13 -0800770static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700771 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800772 unsigned int flags)
773{
774 struct socket *sock = file->private_data;
775
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800776 if (unlikely(!sock->ops->splice_read))
777 return -EINVAL;
778
Jens Axboe9c55e012007-11-06 23:30:13 -0800779 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
780}
781
Al Viro8ae5e032014-11-28 19:40:50 -0500782static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800783{
Al Viro6d652332015-01-30 16:12:56 -0500784 struct file *file = iocb->ki_filp;
785 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700786 struct msghdr msg = {.msg_iter = *to,
787 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500788 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800789
Al Viro8ae5e032014-11-28 19:40:50 -0500790 if (file->f_flags & O_NONBLOCK)
791 msg.msg_flags = MSG_DONTWAIT;
792
793 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800794 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700795
Christoph Hellwig66ee59a2015-02-11 19:56:46 +0100796 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800797 return 0;
798
Al Viro237dae82015-04-09 00:00:30 -0400799 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
Al Viro8ae5e032014-11-28 19:40:50 -0500800 *to = msg.msg_iter;
801 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800802}
803
Al Viro8ae5e032014-11-28 19:40:50 -0500804static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
Al Viro6d652332015-01-30 16:12:56 -0500806 struct file *file = iocb->ki_filp;
807 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700808 struct msghdr msg = {.msg_iter = *from,
809 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500810 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800811
Al Viro8ae5e032014-11-28 19:40:50 -0500812 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800813 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700814
Al Viro8ae5e032014-11-28 19:40:50 -0500815 if (file->f_flags & O_NONBLOCK)
816 msg.msg_flags = MSG_DONTWAIT;
817
Al Viro6d652332015-01-30 16:12:56 -0500818 if (sock->type == SOCK_SEQPACKET)
819 msg.msg_flags |= MSG_EOR;
820
Al Virod8725c82014-12-11 00:02:50 -0500821 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e032014-11-28 19:40:50 -0500822 *from = msg.msg_iter;
823 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824}
825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826/*
827 * Atomic setting of ioctl hooks to avoid race
828 * with module unload.
829 */
830
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800831static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700832static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
Eric W. Biederman881d9662007-09-17 11:56:21 -0700834void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800836 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800838 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839}
840EXPORT_SYMBOL(brioctl_set);
841
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800842static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700843static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844
Eric W. Biederman881d9662007-09-17 11:56:21 -0700845void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800847 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800849 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850}
851EXPORT_SYMBOL(vlan_ioctl_set);
852
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800853static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700854static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700856void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800858 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800860 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861}
862EXPORT_SYMBOL(dlci_ioctl_set);
863
Arnd Bergmann6b960182009-11-06 23:10:54 -0800864static long sock_do_ioctl(struct net *net, struct socket *sock,
865 unsigned int cmd, unsigned long arg)
866{
867 int err;
868 void __user *argp = (void __user *)arg;
869
870 err = sock->ops->ioctl(sock, cmd, arg);
871
872 /*
873 * If this ioctl is unknown try to hand it down
874 * to the NIC driver.
875 */
876 if (err == -ENOIOCTLCMD)
877 err = dev_ioctl(net, cmd, argp);
878
879 return err;
880}
881
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882/*
883 * With an ioctl, arg may well be a user mode pointer, but we don't know
884 * what to do with it - that's up to the protocol still.
885 */
886
887static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
888{
889 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700890 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 void __user *argp = (void __user *)arg;
892 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700893 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Eric Dumazetb69aee02005-09-06 14:42:45 -0700895 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700896 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900897 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700899 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200901#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700903 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200905#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700906 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 case FIOSETOWN:
908 case SIOCSPGRP:
909 err = -EFAULT;
910 if (get_user(pid, (int __user *)argp))
911 break;
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400912 f_setown(sock->file, pid, 1);
913 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 break;
915 case FIOGETOWN:
916 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700917 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700918 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 break;
920 case SIOCGIFBR:
921 case SIOCSIFBR:
922 case SIOCBRADDBR:
923 case SIOCBRDELBR:
924 err = -ENOPKG;
925 if (!br_ioctl_hook)
926 request_module("bridge");
927
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800928 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700929 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700930 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800931 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 break;
933 case SIOCGIFVLAN:
934 case SIOCSIFVLAN:
935 err = -ENOPKG;
936 if (!vlan_ioctl_hook)
937 request_module("8021q");
938
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800939 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700941 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800942 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 case SIOCADDDLCI:
945 case SIOCDELDLCI:
946 err = -ENOPKG;
947 if (!dlci_ioctl_hook)
948 request_module("dlci");
949
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700950 mutex_lock(&dlci_ioctl_mutex);
951 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700953 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954 break;
955 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -0800956 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700958 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 return err;
960}
961
962int sock_create_lite(int family, int type, int protocol, struct socket **res)
963{
964 int err;
965 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700966
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 err = security_socket_create(family, type, protocol, 1);
968 if (err)
969 goto out;
970
971 sock = sock_alloc();
972 if (!sock) {
973 err = -ENOMEM;
974 goto out;
975 }
976
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700978 err = security_socket_post_create(sock, family, type, protocol, 1);
979 if (err)
980 goto out_release;
981
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982out:
983 *res = sock;
984 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700985out_release:
986 sock_release(sock);
987 sock = NULL;
988 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700990EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991
992/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700993static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994{
Eliezer Tamircbf55002013-07-08 16:20:34 +0300995 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 struct socket *sock;
997
998 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700999 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001001 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001002
Eliezer Tamircbf55002013-07-08 16:20:34 +03001003 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001004 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001005 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001006
1007 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001008 if (wait && (wait->_key & POLL_BUSY_LOOP))
1009 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001010 }
1011
Eliezer Tamircbf55002013-07-08 16:20:34 +03001012 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013}
1014
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001015static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001017 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018
1019 return sock->ops->mmap(file, sock, vma);
1020}
1021
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001022static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 sock_release(SOCKET_I(inode));
1025 return 0;
1026}
1027
1028/*
1029 * Update the socket async list
1030 *
1031 * Fasync_list locking strategy.
1032 *
1033 * 1. fasync_list is modified only under process context socket lock
1034 * i.e. under semaphore.
1035 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001036 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 */
1038
1039static int sock_fasync(int fd, struct file *filp, int on)
1040{
Eric Dumazet989a2972010-04-14 09:55:35 +00001041 struct socket *sock = filp->private_data;
1042 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001043 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044
Eric Dumazet989a2972010-04-14 09:55:35 +00001045 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
1048 lock_sock(sk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001049 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1050 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
Eric Dumazeteaefd112011-02-18 03:26:36 +00001052 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001053 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001054 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001055 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
Eric Dumazet989a2972010-04-14 09:55:35 +00001057 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058 return 0;
1059}
1060
Eric Dumazetceb5d582015-11-29 20:03:11 -08001061/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062
Eric Dumazetceb5d582015-11-29 20:03:11 -08001063int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001065 if (!wq || !wq->fasync_list)
1066 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001067
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001068 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001069 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001070 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 break;
1072 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001073 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001074 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 break;
1076 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001077 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001078call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001079 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001081 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001082 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001084
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 return 0;
1086}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001087EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088
Pavel Emelyanov721db932010-09-29 16:06:32 +04001089int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001090 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091{
1092 int err;
1093 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001094 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
1096 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001097 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 */
1099 if (family < 0 || family >= NPROTO)
1100 return -EAFNOSUPPORT;
1101 if (type < 0 || type >= SOCK_MAX)
1102 return -EINVAL;
1103
1104 /* Compatibility.
1105
1106 This uglymoron is moved from INET layer to here to avoid
1107 deadlock in module load.
1108 */
1109 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001110 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1111 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 family = PF_PACKET;
1113 }
1114
1115 err = security_socket_create(family, type, protocol, kern);
1116 if (err)
1117 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001118
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001119 /*
1120 * Allocate the socket and allow the family to set things up. if
1121 * the protocol is 0, the family is instructed to select an appropriate
1122 * default.
1123 */
1124 sock = sock_alloc();
1125 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001126 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001127 return -ENFILE; /* Not exactly a match, but its the
1128 closest posix thing */
1129 }
1130
1131 sock->type = type;
1132
Johannes Berg95a5afc2008-10-16 15:24:51 -07001133#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001134 /* Attempt to load a protocol module if the find failed.
1135 *
1136 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137 * requested real, full-featured networking support upon configuration.
1138 * Otherwise module support will break!
1139 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001140 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001141 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142#endif
1143
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001144 rcu_read_lock();
1145 pf = rcu_dereference(net_families[family]);
1146 err = -EAFNOSUPPORT;
1147 if (!pf)
1148 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
1150 /*
1151 * We will call the ->create function, that possibly is in a loadable
1152 * module, so we have to bump that loadable module refcnt first.
1153 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001154 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 goto out_release;
1156
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001157 /* Now protected by module ref count */
1158 rcu_read_unlock();
1159
Eric Paris3f378b62009-11-05 22:18:14 -08001160 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001161 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001163
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 /*
1165 * Now to bump the refcnt of the [loadable] module that owns this
1166 * socket at sock_release time we decrement its refcnt.
1167 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001168 if (!try_module_get(sock->ops->owner))
1169 goto out_module_busy;
1170
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 /*
1172 * Now that we're done with the ->create function, the [loadable]
1173 * module can have its refcnt decremented
1174 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001175 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001176 err = security_socket_post_create(sock, family, type, protocol, kern);
1177 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001178 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001179 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001181 return 0;
1182
1183out_module_busy:
1184 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001186 sock->ops = NULL;
1187 module_put(pf->owner);
1188out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001190 return err;
1191
1192out_release:
1193 rcu_read_unlock();
1194 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001196EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197
1198int sock_create(int family, int type, int protocol, struct socket **res)
1199{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001200 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001202EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001204int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001206 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001208EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001210SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211{
1212 int retval;
1213 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001214 int flags;
1215
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001216 /* Check the SOCK_* constants for consistency. */
1217 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1218 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1219 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1220 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1221
Ulrich Dreppera677a032008-07-23 21:29:17 -07001222 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001223 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001224 return -EINVAL;
1225 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001227 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1228 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 retval = sock_create(family, type, protocol, &sock);
1231 if (retval < 0)
1232 goto out;
1233
Ulrich Drepper77d27202008-07-23 21:29:35 -07001234 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 if (retval < 0)
1236 goto out_release;
1237
1238out:
1239 /* It may be already another descriptor 8) Not kernel problem. */
1240 return retval;
1241
1242out_release:
1243 sock_release(sock);
1244 return retval;
1245}
1246
1247/*
1248 * Create a pair of connected sockets.
1249 */
1250
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001251SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1252 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253{
1254 struct socket *sock1, *sock2;
1255 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001256 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001257 int flags;
1258
1259 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001260 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001261 return -EINVAL;
1262 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001264 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1265 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1266
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 /*
1268 * Obtain the first socket and check if the underlying protocol
1269 * supports the socketpair call.
1270 */
1271
1272 err = sock_create(family, type, protocol, &sock1);
1273 if (err < 0)
1274 goto out;
1275
1276 err = sock_create(family, type, protocol, &sock2);
1277 if (err < 0)
1278 goto out_release_1;
1279
1280 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001281 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 goto out_release_both;
1283
Al Viro28407632012-08-17 23:54:15 -04001284 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001285 if (unlikely(fd1 < 0)) {
1286 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001288 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001289
Al Viro28407632012-08-17 23:54:15 -04001290 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001291 if (unlikely(fd2 < 0)) {
1292 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001293 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001294 }
1295
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001296 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301297 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001298 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001299 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001300 }
1301
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001302 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001303 if (IS_ERR(newfile2)) {
1304 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001305 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001306 }
1307
Yann Droneaudd73aa282013-12-09 22:42:20 +01001308 err = put_user(fd1, &usockvec[0]);
1309 if (err)
1310 goto out_fput_both;
1311
1312 err = put_user(fd2, &usockvec[1]);
1313 if (err)
1314 goto out_fput_both;
1315
Al Viro157cf642008-12-14 04:57:47 -05001316 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001317
Al Virodb349502007-02-07 01:48:00 -05001318 fd_install(fd1, newfile1);
1319 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 /* fd1 and fd2 may be already another descriptors.
1321 * Not kernel problem.
1322 */
1323
Yann Droneaudd73aa282013-12-09 22:42:20 +01001324 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325
Yann Droneaudd73aa282013-12-09 22:42:20 +01001326out_fput_both:
1327 fput(newfile2);
1328 fput(newfile1);
1329 put_unused_fd(fd2);
1330 put_unused_fd(fd1);
1331 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332
Yann Droneaudd73aa282013-12-09 22:42:20 +01001333out_fput_1:
1334 fput(newfile1);
1335 put_unused_fd(fd2);
1336 put_unused_fd(fd1);
1337 sock_release(sock2);
1338 goto out;
1339
1340out_put_unused_both:
1341 put_unused_fd(fd2);
1342out_put_unused_1:
1343 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001345 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001347 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348out:
1349 return err;
1350}
1351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352/*
1353 * Bind a name to a socket. Nothing much to do here since it's
1354 * the protocol's responsibility to handle the local address.
1355 *
1356 * We move the socket address to kernel space before we call
1357 * the protocol layer (having also checked the address is ok).
1358 */
1359
Heiko Carstens20f37032009-01-14 14:14:23 +01001360SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361{
1362 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001363 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001364 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001366 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001367 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001368 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001369 if (err >= 0) {
1370 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001371 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001372 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001373 if (!err)
1374 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001375 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001376 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001378 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001379 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 return err;
1381}
1382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383/*
1384 * Perform a listen. Basically, we allow the protocol to do anything
1385 * necessary for a listen, and if that works, we mark the socket as
1386 * ready for listening.
1387 */
1388
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001389SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390{
1391 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001392 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001393 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001394
1395 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1396 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001397 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001398 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001399 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400
1401 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001402 if (!err)
1403 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001405 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 }
1407 return err;
1408}
1409
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410/*
1411 * For accept, we attempt to create a new socket, set up the link
1412 * with the client, wake up the client, then return the new
1413 * connected fd. We collect the address of the connector in kernel
1414 * space and move it to user at the very end. This is unclean because
1415 * we open the socket then return an error.
1416 *
1417 * 1003.1g adds the ability to recvmsg() to query connection pending
1418 * status to recvmsg. We need to add that support in a way thats
1419 * clean when we restucture accept also.
1420 */
1421
Heiko Carstens20f37032009-01-14 14:14:23 +01001422SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1423 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424{
1425 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001426 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001427 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001428 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Ulrich Drepper77d27202008-07-23 21:29:35 -07001430 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001431 return -EINVAL;
1432
1433 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1434 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1435
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001436 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 if (!sock)
1438 goto out;
1439
1440 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001441 newsock = sock_alloc();
1442 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 goto out_put;
1444
1445 newsock->type = sock->type;
1446 newsock->ops = sock->ops;
1447
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 /*
1449 * We don't need try_module_get here, as the listening socket (sock)
1450 * has the protocol module (sock->ops->owner) held.
1451 */
1452 __module_get(newsock->ops->owner);
1453
Al Viro28407632012-08-17 23:54:15 -04001454 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001455 if (unlikely(newfd < 0)) {
1456 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001457 sock_release(newsock);
1458 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001459 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001460 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301461 if (IS_ERR(newfile)) {
Al Viro28407632012-08-17 23:54:15 -04001462 err = PTR_ERR(newfile);
1463 put_unused_fd(newfd);
1464 sock_release(newsock);
1465 goto out_put;
1466 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001467
Frank Filza79af592005-09-27 15:23:38 -07001468 err = security_socket_accept(sock, newsock);
1469 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001470 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001471
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1473 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001474 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475
1476 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001477 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001478 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001480 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001482 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001483 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001485 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 }
1487
1488 /* File flags are not inherited via accept() unlike another OSes. */
1489
David S. Miller39d8c1b2006-03-20 17:13:49 -08001490 fd_install(newfd, newfile);
1491 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001494 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495out:
1496 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001497out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001498 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001499 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 goto out_put;
1501}
1502
Heiko Carstens20f37032009-01-14 14:14:23 +01001503SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1504 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001505{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001506 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001507}
1508
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509/*
1510 * Attempt to connect to a socket with the server address. The address
1511 * is in user space so we verify it is OK and move it to kernel space.
1512 *
1513 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1514 * break bindings
1515 *
1516 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1517 * other SEQPACKET protocols that take time to connect() as it doesn't
1518 * include the -EINPROGRESS status for such sockets.
1519 */
1520
Heiko Carstens20f37032009-01-14 14:14:23 +01001521SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1522 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523{
1524 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001525 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001526 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001528 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 if (!sock)
1530 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001531 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 if (err < 0)
1533 goto out_put;
1534
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001535 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001536 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 if (err)
1538 goto out_put;
1539
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001540 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 sock->file->f_flags);
1542out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001543 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544out:
1545 return err;
1546}
1547
1548/*
1549 * Get the local address ('name') of a socket object. Move the obtained
1550 * name to user space.
1551 */
1552
Heiko Carstens20f37032009-01-14 14:14:23 +01001553SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1554 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555{
1556 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001557 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001558 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001559
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001560 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 if (!sock)
1562 goto out;
1563
1564 err = security_socket_getsockname(sock);
1565 if (err)
1566 goto out_put;
1567
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001568 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 if (err)
1570 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001571 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
1573out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001574 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575out:
1576 return err;
1577}
1578
1579/*
1580 * Get the remote address ('name') of a socket object. Move the obtained
1581 * name to user space.
1582 */
1583
Heiko Carstens20f37032009-01-14 14:14:23 +01001584SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1585 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586{
1587 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001588 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001589 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001591 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1592 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 err = security_socket_getpeername(sock);
1594 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001595 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596 return err;
1597 }
1598
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001599 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001600 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001601 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001603 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001604 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001605 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 }
1607 return err;
1608}
1609
1610/*
1611 * Send a datagram to a given address. We move the address into kernel
1612 * space and check the user space data area is readable before invoking
1613 * the protocol.
1614 */
1615
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001616SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001617 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001618 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619{
1620 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001621 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 int err;
1623 struct msghdr msg;
1624 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001625 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001626
Al Viro602bd0e2015-03-21 19:12:32 -04001627 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1628 if (unlikely(err))
1629 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001630 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1631 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001632 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001633
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001634 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001635 msg.msg_control = NULL;
1636 msg.msg_controllen = 0;
1637 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001638 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001639 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 if (err < 0)
1641 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001642 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001643 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 }
1645 if (sock->file->f_flags & O_NONBLOCK)
1646 flags |= MSG_DONTWAIT;
1647 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05001648 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001650out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001651 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001652out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 return err;
1654}
1655
1656/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001657 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 */
1659
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001660SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001661 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662{
1663 return sys_sendto(fd, buff, len, flags, NULL, 0);
1664}
1665
1666/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001667 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 * sender. We verify the buffers are writable and if needed move the
1669 * sender address from kernel to user space.
1670 */
1671
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001672SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001673 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001674 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675{
1676 struct socket *sock;
1677 struct iovec iov;
1678 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001679 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001680 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001681 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682
Al Viro602bd0e2015-03-21 19:12:32 -04001683 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1684 if (unlikely(err))
1685 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001686 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001688 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001690 msg.msg_control = NULL;
1691 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001692 /* Save some cycles and don't copy the address if not needed */
1693 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1694 /* We assume all kernel code knows the size of sockaddr_storage */
1695 msg.msg_namelen = 0;
tadeusz.struk@intel.com130ed5d2015-12-15 10:46:17 -08001696 msg.msg_iocb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697 if (sock->file->f_flags & O_NONBLOCK)
1698 flags |= MSG_DONTWAIT;
Al Viro602bd0e2015-03-21 19:12:32 -04001699 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001701 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001702 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001703 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001704 if (err2 < 0)
1705 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001707
1708 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001709out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 return err;
1711}
1712
1713/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001714 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715 */
1716
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001717SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1718 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719{
1720 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1721}
1722
1723/*
1724 * Set a socket option. Because we don't know the option lengths we have
1725 * to pass the user mode parameter for the protocols to sort out.
1726 */
1727
Heiko Carstens20f37032009-01-14 14:14:23 +01001728SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1729 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001731 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732 struct socket *sock;
1733
1734 if (optlen < 0)
1735 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001736
1737 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1738 if (sock != NULL) {
1739 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001740 if (err)
1741 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742
1743 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001744 err =
1745 sock_setsockopt(sock, level, optname, optval,
1746 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001748 err =
1749 sock->ops->setsockopt(sock, level, optname, optval,
1750 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001751out_put:
1752 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 }
1754 return err;
1755}
1756
1757/*
1758 * Get a socket option. Because we don't know the option lengths we have
1759 * to pass a user mode parameter for the protocols to sort out.
1760 */
1761
Heiko Carstens20f37032009-01-14 14:14:23 +01001762SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1763 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001765 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 struct socket *sock;
1767
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001768 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001770 err = security_socket_getsockopt(sock, level, optname);
1771 if (err)
1772 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773
1774 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001775 err =
1776 sock_getsockopt(sock, level, optname, optval,
1777 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001779 err =
1780 sock->ops->getsockopt(sock, level, optname, optval,
1781 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001782out_put:
1783 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784 }
1785 return err;
1786}
1787
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788/*
1789 * Shutdown a socket.
1790 */
1791
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001792SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001794 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 struct socket *sock;
1796
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1798 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001800 if (!err)
1801 err = sock->ops->shutdown(sock, how);
1802 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803 }
1804 return err;
1805}
1806
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001807/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 * fields which are the same type (int / unsigned) on our platforms.
1809 */
1810#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1811#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1812#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1813
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001814struct used_address {
1815 struct sockaddr_storage name;
1816 unsigned int name_len;
1817};
1818
Al Viroda184282015-03-21 19:29:06 -04001819static int copy_msghdr_from_user(struct msghdr *kmsg,
1820 struct user_msghdr __user *umsg,
1821 struct sockaddr __user **save_addr,
1822 struct iovec **iov)
Dan Carpenter1661bf32013-10-03 00:27:20 +03001823{
Al Viro08adb7d2014-11-10 20:23:13 -05001824 struct sockaddr __user *uaddr;
1825 struct iovec __user *uiov;
Al Viroc0371da2014-11-24 10:42:55 -05001826 size_t nr_segs;
Al Viro08adb7d2014-11-10 20:23:13 -05001827 ssize_t err;
1828
1829 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1830 __get_user(uaddr, &umsg->msg_name) ||
1831 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1832 __get_user(uiov, &umsg->msg_iov) ||
Al Viroc0371da2014-11-24 10:42:55 -05001833 __get_user(nr_segs, &umsg->msg_iovlen) ||
Al Viro08adb7d2014-11-10 20:23:13 -05001834 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1835 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1836 __get_user(kmsg->msg_flags, &umsg->msg_flags))
Dan Carpenter1661bf32013-10-03 00:27:20 +03001837 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001838
Al Viro08adb7d2014-11-10 20:23:13 -05001839 if (!uaddr)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07001840 kmsg->msg_namelen = 0;
1841
Matthew Leachdbb490b2014-03-11 11:58:27 +00001842 if (kmsg->msg_namelen < 0)
1843 return -EINVAL;
1844
Dan Carpenter1661bf32013-10-03 00:27:20 +03001845 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001846 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05001847
1848 if (save_addr)
1849 *save_addr = uaddr;
1850
1851 if (uaddr && kmsg->msg_namelen) {
1852 if (!save_addr) {
1853 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1854 kmsg->msg_name);
1855 if (err < 0)
1856 return err;
1857 }
1858 } else {
1859 kmsg->msg_name = NULL;
1860 kmsg->msg_namelen = 0;
1861 }
1862
Al Viroc0371da2014-11-24 10:42:55 -05001863 if (nr_segs > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05001864 return -EMSGSIZE;
1865
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001866 kmsg->msg_iocb = NULL;
1867
Al Viroda184282015-03-21 19:29:06 -04001868 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1869 UIO_FASTIOV, iov, &kmsg->msg_iter);
Dan Carpenter1661bf32013-10-03 00:27:20 +03001870}
1871
Al Viro666547f2014-04-06 14:03:05 -04001872static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001873 struct msghdr *msg_sys, unsigned int flags,
Tom Herbert28a94d82016-03-07 14:11:02 -08001874 struct used_address *used_address,
1875 unsigned int allowed_msghdr_flags)
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;
Tom Herbert28a94d82016-03-07 14:11:02 -08001901 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00001902 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001904 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001905 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001906 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 if (err)
1908 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001909 ctl_buf = msg_sys->msg_control;
1910 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001912 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001914 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915 goto out_freeiov;
1916 }
1917 err = -EFAULT;
1918 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00001919 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1921 * checking falls down on this.
1922 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001923 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00001924 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001925 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00001927 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928 }
Anton Blanchard228e5482011-05-02 20:21:35 +00001929 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930
1931 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00001932 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001933 /*
1934 * If this is sendmmsg() and current destination address is same as
1935 * previously succeeded address, omit asking LSM's decision.
1936 * used_address->name_len is initialized to UINT_MAX so that the first
1937 * destination address never matches.
1938 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001939 if (used_address && msg_sys->msg_name &&
1940 used_address->name_len == msg_sys->msg_namelen &&
1941 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001942 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05001943 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001944 goto out_freectl;
1945 }
Al Virod8725c82014-12-11 00:02:50 -05001946 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001947 /*
1948 * If this is sendmmsg() and sending to current destination address was
1949 * successful, remember it.
1950 */
1951 if (used_address && err >= 0) {
1952 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001953 if (msg_sys->msg_name)
1954 memcpy(&used_address->name, msg_sys->msg_name,
1955 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001956 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
1958out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001959 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1961out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04001962 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00001963 return err;
1964}
1965
1966/*
1967 * BSD sendmsg interface
1968 */
1969
Al Viro666547f2014-04-06 14:03:05 -04001970long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00001971{
1972 int fput_needed, err;
1973 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07001974 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00001975
Andy Lutomirski1be374a2013-05-22 14:07:44 -07001976 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00001977 if (!sock)
1978 goto out;
1979
Tom Herbert28a94d82016-03-07 14:11:02 -08001980 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00001981
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001982 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001983out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 return err;
1985}
1986
Al Viro666547f2014-04-06 14:03:05 -04001987SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001988{
1989 if (flags & MSG_CMSG_COMPAT)
1990 return -EINVAL;
1991 return __sys_sendmsg(fd, msg, flags);
1992}
1993
Anton Blanchard228e5482011-05-02 20:21:35 +00001994/*
1995 * Linux sendmmsg interface
1996 */
1997
1998int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1999 unsigned int flags)
2000{
2001 int fput_needed, err, datagrams;
2002 struct socket *sock;
2003 struct mmsghdr __user *entry;
2004 struct compat_mmsghdr __user *compat_entry;
2005 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002006 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002007 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002008
Anton Blanchard98382f42011-08-04 14:07:39 +00002009 if (vlen > UIO_MAXIOV)
2010 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002011
2012 datagrams = 0;
2013
2014 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2015 if (!sock)
2016 return err;
2017
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002018 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002019 entry = mmsg;
2020 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002021 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002022 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002023
2024 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002025 if (datagrams == vlen - 1)
2026 flags = oflags;
2027
Anton Blanchard228e5482011-05-02 20:21:35 +00002028 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002029 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002030 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002031 if (err < 0)
2032 break;
2033 err = __put_user(err, &compat_entry->msg_len);
2034 ++compat_entry;
2035 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002036 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002037 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002038 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002039 if (err < 0)
2040 break;
2041 err = put_user(err, &entry->msg_len);
2042 ++entry;
2043 }
2044
2045 if (err)
2046 break;
2047 ++datagrams;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002048 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002049 }
2050
Anton Blanchard228e5482011-05-02 20:21:35 +00002051 fput_light(sock->file, fput_needed);
2052
Anton Blanchard728ffb82011-08-04 14:07:38 +00002053 /* We only return an error if no datagrams were able to be sent */
2054 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002055 return datagrams;
2056
Anton Blanchard228e5482011-05-02 20:21:35 +00002057 return err;
2058}
2059
2060SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2061 unsigned int, vlen, unsigned int, flags)
2062{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002063 if (flags & MSG_CMSG_COMPAT)
2064 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002065 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2066}
2067
Al Viro666547f2014-04-06 14:03:05 -04002068static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002069 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002071 struct compat_msghdr __user *msg_compat =
2072 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002074 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 unsigned long cmsg_ptr;
Al Viro08adb7d2014-11-10 20:23:13 -05002076 int total_len, len;
2077 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078
2079 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002080 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081
2082 /* user mode address pointers */
2083 struct sockaddr __user *uaddr;
Al Viro08adb7d2014-11-10 20:23:13 -05002084 int __user *uaddr_len = COMPAT_NAMELEN(msg);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002085
Al Viro08adb7d2014-11-10 20:23:13 -05002086 msg_sys->msg_name = &addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002088 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002089 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002090 else
Al Viro08adb7d2014-11-10 20:23:13 -05002091 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04002093 return err;
2094 total_len = iov_iter_count(&msg_sys->msg_iter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002096 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2097 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002098
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002099 /* We assume all kernel code knows the size of sockaddr_storage */
2100 msg_sys->msg_namelen = 0;
2101
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102 if (sock->file->f_flags & O_NONBLOCK)
2103 flags |= MSG_DONTWAIT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002104 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2105 total_len, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106 if (err < 0)
2107 goto out_freeiov;
2108 len = err;
2109
2110 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002111 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002112 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002113 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114 if (err < 0)
2115 goto out_freeiov;
2116 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002117 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002118 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 if (err)
2120 goto out_freeiov;
2121 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002122 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 &msg_compat->msg_controllen);
2124 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002125 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 &msg->msg_controllen);
2127 if (err)
2128 goto out_freeiov;
2129 err = len;
2130
2131out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04002132 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002133 return err;
2134}
2135
2136/*
2137 * BSD recvmsg interface
2138 */
2139
Al Viro666547f2014-04-06 14:03:05 -04002140long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002141{
2142 int fput_needed, err;
2143 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002144 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002145
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002146 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002147 if (!sock)
2148 goto out;
2149
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002150 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002151
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002152 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153out:
2154 return err;
2155}
2156
Al Viro666547f2014-04-06 14:03:05 -04002157SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002158 unsigned int, flags)
2159{
2160 if (flags & MSG_CMSG_COMPAT)
2161 return -EINVAL;
2162 return __sys_recvmsg(fd, msg, flags);
2163}
2164
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002165/*
2166 * Linux recvmmsg interface
2167 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002169int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2170 unsigned int flags, struct timespec *timeout)
2171{
2172 int fput_needed, err, datagrams;
2173 struct socket *sock;
2174 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002175 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002176 struct msghdr msg_sys;
2177 struct timespec end_time;
2178
2179 if (timeout &&
2180 poll_select_set_timeout(&end_time, timeout->tv_sec,
2181 timeout->tv_nsec))
2182 return -EINVAL;
2183
2184 datagrams = 0;
2185
2186 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2187 if (!sock)
2188 return err;
2189
2190 err = sock_error(sock->sk);
2191 if (err)
2192 goto out_put;
2193
2194 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002195 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002196
2197 while (datagrams < vlen) {
2198 /*
2199 * No need to ask LSM for more than the first datagram.
2200 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002201 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002202 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002203 &msg_sys, flags & ~MSG_WAITFORONE,
2204 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002205 if (err < 0)
2206 break;
2207 err = __put_user(err, &compat_entry->msg_len);
2208 ++compat_entry;
2209 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002210 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002211 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002212 &msg_sys, flags & ~MSG_WAITFORONE,
2213 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002214 if (err < 0)
2215 break;
2216 err = put_user(err, &entry->msg_len);
2217 ++entry;
2218 }
2219
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002220 if (err)
2221 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002222 ++datagrams;
2223
Brandon L Black71c5c152010-03-26 16:18:03 +00002224 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2225 if (flags & MSG_WAITFORONE)
2226 flags |= MSG_DONTWAIT;
2227
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002228 if (timeout) {
2229 ktime_get_ts(timeout);
2230 *timeout = timespec_sub(end_time, *timeout);
2231 if (timeout->tv_sec < 0) {
2232 timeout->tv_sec = timeout->tv_nsec = 0;
2233 break;
2234 }
2235
2236 /* Timeout, return less than vlen datagrams */
2237 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2238 break;
2239 }
2240
2241 /* Out of band data, return right away */
2242 if (msg_sys.msg_flags & MSG_OOB)
2243 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002244 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002245 }
2246
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002247 if (err == 0)
2248 goto out_put;
2249
2250 if (datagrams == 0) {
2251 datagrams = err;
2252 goto out_put;
2253 }
2254
2255 /*
2256 * We may return less entries than requested (vlen) if the
2257 * sock is non block and there aren't enough datagrams...
2258 */
2259 if (err != -EAGAIN) {
2260 /*
2261 * ... or if recvmsg returns an error after we
2262 * received some datagrams, where we record the
2263 * error to return on the next call or if the
2264 * app asks about it using getsockopt(SO_ERROR).
2265 */
2266 sock->sk->sk_err = -err;
2267 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002268out_put:
2269 fput_light(sock->file, fput_needed);
2270
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002271 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002272}
2273
2274SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2275 unsigned int, vlen, unsigned int, flags,
2276 struct timespec __user *, timeout)
2277{
2278 int datagrams;
2279 struct timespec timeout_sys;
2280
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002281 if (flags & MSG_CMSG_COMPAT)
2282 return -EINVAL;
2283
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002284 if (!timeout)
2285 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2286
2287 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2288 return -EFAULT;
2289
2290 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2291
2292 if (datagrams > 0 &&
2293 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2294 datagrams = -EFAULT;
2295
2296 return datagrams;
2297}
2298
2299#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002300/* Argument list sizes for sys_socketcall */
2301#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002302static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002303 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2304 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2305 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002306 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002307};
2308
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309#undef AL
2310
2311/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002312 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313 *
2314 * Argument checking cleaned up. Saved 20% in size.
2315 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002316 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 */
2318
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002319SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320{
Chen Gang2950fa92013-04-07 16:55:23 +08002321 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002322 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002324 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325
Anton Blanchard228e5482011-05-02 20:21:35 +00002326 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327 return -EINVAL;
2328
Arjan van de Ven47379052009-09-28 12:57:44 -07002329 len = nargs[call];
2330 if (len > sizeof(a))
2331 return -EINVAL;
2332
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002334 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002336
Chen Gang2950fa92013-04-07 16:55:23 +08002337 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2338 if (err)
2339 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002340
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002341 a0 = a[0];
2342 a1 = a[1];
2343
2344 switch (call) {
2345 case SYS_SOCKET:
2346 err = sys_socket(a0, a1, a[2]);
2347 break;
2348 case SYS_BIND:
2349 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2350 break;
2351 case SYS_CONNECT:
2352 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2353 break;
2354 case SYS_LISTEN:
2355 err = sys_listen(a0, a1);
2356 break;
2357 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002358 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2359 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002360 break;
2361 case SYS_GETSOCKNAME:
2362 err =
2363 sys_getsockname(a0, (struct sockaddr __user *)a1,
2364 (int __user *)a[2]);
2365 break;
2366 case SYS_GETPEERNAME:
2367 err =
2368 sys_getpeername(a0, (struct sockaddr __user *)a1,
2369 (int __user *)a[2]);
2370 break;
2371 case SYS_SOCKETPAIR:
2372 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2373 break;
2374 case SYS_SEND:
2375 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2376 break;
2377 case SYS_SENDTO:
2378 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2379 (struct sockaddr __user *)a[4], a[5]);
2380 break;
2381 case SYS_RECV:
2382 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2383 break;
2384 case SYS_RECVFROM:
2385 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2386 (struct sockaddr __user *)a[4],
2387 (int __user *)a[5]);
2388 break;
2389 case SYS_SHUTDOWN:
2390 err = sys_shutdown(a0, a1);
2391 break;
2392 case SYS_SETSOCKOPT:
2393 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2394 break;
2395 case SYS_GETSOCKOPT:
2396 err =
2397 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2398 (int __user *)a[4]);
2399 break;
2400 case SYS_SENDMSG:
Al Viro666547f2014-04-06 14:03:05 -04002401 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002402 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002403 case SYS_SENDMMSG:
2404 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2405 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002406 case SYS_RECVMSG:
Al Viro666547f2014-04-06 14:03:05 -04002407 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002408 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002409 case SYS_RECVMMSG:
2410 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2411 (struct timespec __user *)a[4]);
2412 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002413 case SYS_ACCEPT4:
2414 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2415 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002416 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002417 default:
2418 err = -EINVAL;
2419 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002420 }
2421 return err;
2422}
2423
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002424#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002426/**
2427 * sock_register - add a socket protocol handler
2428 * @ops: description of protocol
2429 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430 * This function is called by a protocol handler that wants to
2431 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002432 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002433 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002435int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002436{
2437 int err;
2438
2439 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002440 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 return -ENOBUFS;
2442 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002443
2444 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002445 if (rcu_dereference_protected(net_families[ops->family],
2446 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002447 err = -EEXIST;
2448 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002449 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 err = 0;
2451 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002452 spin_unlock(&net_family_lock);
2453
Yang Yingliang3410f222014-02-12 17:09:55 +08002454 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455 return err;
2456}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002457EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002459/**
2460 * sock_unregister - remove a protocol handler
2461 * @family: protocol family to remove
2462 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 * This function is called by a protocol handler that wants to
2464 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002465 * new socket creation.
2466 *
2467 * If protocol handler is a module, then it can use module reference
2468 * counts to protect against new references. If protocol handler is not
2469 * a module then it needs to provide its own protection in
2470 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002472void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002474 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002476 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002477 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002478 spin_unlock(&net_family_lock);
2479
2480 synchronize_rcu();
2481
Yang Yingliang3410f222014-02-12 17:09:55 +08002482 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002484EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002485
Andi Kleen77d76ea2005-12-22 12:43:42 -08002486static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002488 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002489 /*
2490 * Initialize the network sysctl infrastructure.
2491 */
2492 err = net_sysctl_init();
2493 if (err)
2494 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002495
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002497 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002498 */
2499 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002500
2501 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002502 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002503 */
2504
2505 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002506
2507 err = register_filesystem(&sock_fs_type);
2508 if (err)
2509 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002511 if (IS_ERR(sock_mnt)) {
2512 err = PTR_ERR(sock_mnt);
2513 goto out_mount;
2514 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002515
2516 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002517 */
2518
2519#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002520 err = netfilter_init();
2521 if (err)
2522 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002524
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002525 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002526
Nick Pigginb3e19d92011-01-07 17:50:11 +11002527out:
2528 return err;
2529
2530out_mount:
2531 unregister_filesystem(&sock_fs_type);
2532out_fs:
2533 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534}
2535
Andi Kleen77d76ea2005-12-22 12:43:42 -08002536core_initcall(sock_init); /* early initcall */
2537
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538#ifdef CONFIG_PROC_FS
2539void socket_seq_show(struct seq_file *seq)
2540{
2541 int cpu;
2542 int counter = 0;
2543
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002544 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002545 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546
2547 /* It can be negative, by the way. 8) */
2548 if (counter < 0)
2549 counter = 0;
2550
2551 seq_printf(seq, "sockets: used %d\n", counter);
2552}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002553#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002555#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002556static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002557 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002558{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002559 mm_segment_t old_fs = get_fs();
2560 struct timeval ktv;
2561 int err;
2562
2563 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002564 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002565 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002566 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002567 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002568
Arnd Bergmann7a229382009-11-06 23:00:29 -08002569 return err;
2570}
2571
Arnd Bergmann6b960182009-11-06 23:10:54 -08002572static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002573 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002574{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002575 mm_segment_t old_fs = get_fs();
2576 struct timespec kts;
2577 int err;
2578
2579 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002580 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002581 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002582 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002583 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002584
Arnd Bergmann7a229382009-11-06 23:00:29 -08002585 return err;
2586}
2587
Arnd Bergmann6b960182009-11-06 23:10:54 -08002588static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002589{
2590 struct ifreq __user *uifr;
2591 int err;
2592
2593 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002594 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002595 return -EFAULT;
2596
Arnd Bergmann6b960182009-11-06 23:10:54 -08002597 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002598 if (err)
2599 return err;
2600
Arnd Bergmann6b960182009-11-06 23:10:54 -08002601 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002602 return -EFAULT;
2603
2604 return 0;
2605}
2606
Arnd Bergmann6b960182009-11-06 23:10:54 -08002607static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002608{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002609 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002610 struct ifconf ifc;
2611 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002612 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002613 struct ifreq __user *ifr;
2614 unsigned int i, j;
2615 int err;
2616
Arnd Bergmann6b960182009-11-06 23:10:54 -08002617 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002618 return -EFAULT;
2619
Mathias Krause43da5f22012-08-15 11:31:57 +00002620 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002621 if (ifc32.ifcbuf == 0) {
2622 ifc32.ifc_len = 0;
2623 ifc.ifc_len = 0;
2624 ifc.ifc_req = NULL;
2625 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2626 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002627 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2628 sizeof(struct ifreq);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002629 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2630 ifc.ifc_len = len;
2631 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2632 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002633 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002634 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002635 return -EFAULT;
2636 ifr++;
2637 ifr32++;
2638 }
2639 }
2640 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2641 return -EFAULT;
2642
Arnd Bergmann6b960182009-11-06 23:10:54 -08002643 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002644 if (err)
2645 return err;
2646
2647 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2648 return -EFAULT;
2649
2650 ifr = ifc.ifc_req;
2651 ifr32 = compat_ptr(ifc32.ifcbuf);
2652 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002653 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2654 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2655 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002656 return -EFAULT;
2657 ifr32++;
2658 ifr++;
2659 }
2660
2661 if (ifc32.ifcbuf == 0) {
2662 /* Translate from 64-bit structure multiple to
2663 * a 32-bit one.
2664 */
2665 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002666 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002667 ifc32.ifc_len = i;
2668 } else {
2669 ifc32.ifc_len = i;
2670 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002671 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002672 return -EFAULT;
2673
2674 return 0;
2675}
2676
Arnd Bergmann6b960182009-11-06 23:10:54 -08002677static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002678{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002679 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2680 bool convert_in = false, convert_out = false;
2681 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2682 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002683 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002684 u32 rule_cnt = 0, actual_rule_cnt;
2685 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002686 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002687 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002688
2689 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2690 return -EFAULT;
2691
Ben Hutchings3a7da392011-03-17 07:34:32 +00002692 compat_rxnfc = compat_ptr(data);
2693
2694 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002695 return -EFAULT;
2696
Ben Hutchings3a7da392011-03-17 07:34:32 +00002697 /* Most ethtool structures are defined without padding.
2698 * Unfortunately struct ethtool_rxnfc is an exception.
2699 */
2700 switch (ethcmd) {
2701 default:
2702 break;
2703 case ETHTOOL_GRXCLSRLALL:
2704 /* Buffer size is variable */
2705 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2706 return -EFAULT;
2707 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2708 return -ENOMEM;
2709 buf_size += rule_cnt * sizeof(u32);
2710 /* fall through */
2711 case ETHTOOL_GRXRINGS:
2712 case ETHTOOL_GRXCLSRLCNT:
2713 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002714 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002715 convert_out = true;
2716 /* fall through */
2717 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002718 buf_size += sizeof(struct ethtool_rxnfc);
2719 convert_in = true;
2720 break;
2721 }
2722
2723 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002724 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002725
2726 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2727 return -EFAULT;
2728
2729 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2730 &ifr->ifr_ifru.ifru_data))
2731 return -EFAULT;
2732
2733 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002734 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002735 * fs.ring_cookie and at the end of fs, but nowhere else.
2736 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002737 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2738 sizeof(compat_rxnfc->fs.m_ext) !=
2739 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2740 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002741 BUILD_BUG_ON(
2742 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2743 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2744 offsetof(struct ethtool_rxnfc, fs.location) -
2745 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2746
2747 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002748 (void __user *)(&rxnfc->fs.m_ext + 1) -
2749 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002750 copy_in_user(&rxnfc->fs.ring_cookie,
2751 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002752 (void __user *)(&rxnfc->fs.location + 1) -
2753 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002754 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2755 sizeof(rxnfc->rule_cnt)))
2756 return -EFAULT;
2757 }
2758
2759 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2760 if (ret)
2761 return ret;
2762
2763 if (convert_out) {
2764 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002765 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2766 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002767 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2768 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002769 (const void __user *)(&rxnfc->fs.location + 1) -
2770 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002771 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2772 sizeof(rxnfc->rule_cnt)))
2773 return -EFAULT;
2774
2775 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2776 /* As an optimisation, we only copy the actual
2777 * number of rules that the underlying
2778 * function returned. Since Mallory might
2779 * change the rule count in user memory, we
2780 * check that it is less than the rule count
2781 * originally given (as the user buffer size),
2782 * which has been range-checked.
2783 */
2784 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2785 return -EFAULT;
2786 if (actual_rule_cnt < rule_cnt)
2787 rule_cnt = actual_rule_cnt;
2788 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2789 &rxnfc->rule_locs[0],
2790 rule_cnt * sizeof(u32)))
2791 return -EFAULT;
2792 }
2793 }
2794
2795 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002796}
2797
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002798static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2799{
2800 void __user *uptr;
2801 compat_uptr_t uptr32;
2802 struct ifreq __user *uifr;
2803
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002804 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002805 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2806 return -EFAULT;
2807
2808 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2809 return -EFAULT;
2810
2811 uptr = compat_ptr(uptr32);
2812
2813 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2814 return -EFAULT;
2815
2816 return dev_ioctl(net, SIOCWANDEV, uifr);
2817}
2818
Arnd Bergmann6b960182009-11-06 23:10:54 -08002819static int bond_ioctl(struct net *net, unsigned int cmd,
2820 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002821{
2822 struct ifreq kifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002823 mm_segment_t old_fs;
2824 int err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002825
2826 switch (cmd) {
2827 case SIOCBONDENSLAVE:
2828 case SIOCBONDRELEASE:
2829 case SIOCBONDSETHWADDR:
2830 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002831 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002832 return -EFAULT;
2833
2834 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002835 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002836 err = dev_ioctl(net, cmd,
2837 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002838 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002839
2840 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002841 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002842 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002843 }
Arnd Bergmann7a229382009-11-06 23:00:29 -08002844}
2845
Ben Hutchings590d4692013-11-18 17:04:13 +00002846/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2847static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08002848 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002849{
2850 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002851 char tmp_buf[IFNAMSIZ];
2852 void __user *data64;
2853 u32 data32;
2854
2855 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2856 IFNAMSIZ))
2857 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002858 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002859 return -EFAULT;
2860 data64 = compat_ptr(data32);
2861
2862 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2863
Arnd Bergmann7a229382009-11-06 23:00:29 -08002864 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2865 IFNAMSIZ))
2866 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002867 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002868 return -EFAULT;
2869
Arnd Bergmann6b960182009-11-06 23:10:54 -08002870 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002871}
2872
Arnd Bergmann6b960182009-11-06 23:10:54 -08002873static int dev_ifsioc(struct net *net, struct socket *sock,
2874 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002875{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002876 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002877 int err;
2878
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002879 uifr = compat_alloc_user_space(sizeof(*uifr));
2880 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2881 return -EFAULT;
2882
2883 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2884
Arnd Bergmann7a229382009-11-06 23:00:29 -08002885 if (!err) {
2886 switch (cmd) {
2887 case SIOCGIFFLAGS:
2888 case SIOCGIFMETRIC:
2889 case SIOCGIFMTU:
2890 case SIOCGIFMEM:
2891 case SIOCGIFHWADDR:
2892 case SIOCGIFINDEX:
2893 case SIOCGIFADDR:
2894 case SIOCGIFBRDADDR:
2895 case SIOCGIFDSTADDR:
2896 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002897 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002898 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002899 case SIOCGMIIPHY:
2900 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002901 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002902 err = -EFAULT;
2903 break;
2904 }
2905 }
2906 return err;
2907}
2908
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002909static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2910 struct compat_ifreq __user *uifr32)
2911{
2912 struct ifreq ifr;
2913 struct compat_ifmap __user *uifmap32;
2914 mm_segment_t old_fs;
2915 int err;
2916
2917 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2918 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002919 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2920 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2921 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2922 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2923 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2924 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002925 if (err)
2926 return -EFAULT;
2927
2928 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002929 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002930 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002931 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002932
2933 if (cmd == SIOCGIFMAP && !err) {
2934 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002935 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2936 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2937 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2938 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2939 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2940 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002941 if (err)
2942 err = -EFAULT;
2943 }
2944 return err;
2945}
2946
Arnd Bergmann7a229382009-11-06 23:00:29 -08002947struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002948 u32 rt_pad1;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002949 struct sockaddr rt_dst; /* target address */
2950 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2951 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002952 unsigned short rt_flags;
2953 short rt_pad2;
2954 u32 rt_pad3;
2955 unsigned char rt_tos;
2956 unsigned char rt_class;
2957 short rt_pad4;
2958 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a229382009-11-06 23:00:29 -08002959 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002960 u32 rt_mtu; /* per route MTU/Window */
2961 u32 rt_window; /* Window clamping */
Arnd Bergmann7a229382009-11-06 23:00:29 -08002962 unsigned short rt_irtt; /* Initial RTT */
2963};
2964
2965struct in6_rtmsg32 {
2966 struct in6_addr rtmsg_dst;
2967 struct in6_addr rtmsg_src;
2968 struct in6_addr rtmsg_gateway;
2969 u32 rtmsg_type;
2970 u16 rtmsg_dst_len;
2971 u16 rtmsg_src_len;
2972 u32 rtmsg_metric;
2973 u32 rtmsg_info;
2974 u32 rtmsg_flags;
2975 s32 rtmsg_ifindex;
2976};
2977
Arnd Bergmann6b960182009-11-06 23:10:54 -08002978static int routing_ioctl(struct net *net, struct socket *sock,
2979 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002980{
2981 int ret;
2982 void *r = NULL;
2983 struct in6_rtmsg r6;
2984 struct rtentry r4;
2985 char devname[16];
2986 u32 rtdev;
2987 mm_segment_t old_fs = get_fs();
2988
Arnd Bergmann6b960182009-11-06 23:10:54 -08002989 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2990 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002991 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08002992 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002993 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2994 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2995 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2996 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2997 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2998 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2999 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003000
3001 r = (void *) &r6;
3002 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003003 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003004 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003005 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003006 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3007 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3008 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3009 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3010 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3011 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003012 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003013 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003014 r4.rt_dev = (char __user __force *)devname;
3015 devname[15] = 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003016 } else
3017 r4.rt_dev = NULL;
3018
3019 r = (void *) &r4;
3020 }
3021
3022 if (ret) {
3023 ret = -EFAULT;
3024 goto out;
3025 }
3026
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003027 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003028 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003029 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003030
3031out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003032 return ret;
3033}
3034
3035/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3036 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003037 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003038 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003039static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003040{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003041 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003042
Arnd Bergmann6b960182009-11-06 23:10:54 -08003043 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003044 return -EFAULT;
3045 if (tmp == BRCTL_GET_VERSION)
3046 return BRCTL_VERSION + 1;
3047 return -EINVAL;
3048}
3049
Arnd Bergmann6b960182009-11-06 23:10:54 -08003050static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3051 unsigned int cmd, unsigned long arg)
3052{
3053 void __user *argp = compat_ptr(arg);
3054 struct sock *sk = sock->sk;
3055 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003056
Arnd Bergmann6b960182009-11-06 23:10:54 -08003057 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003058 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003059
Arnd Bergmann6b960182009-11-06 23:10:54 -08003060 switch (cmd) {
3061 case SIOCSIFBR:
3062 case SIOCGIFBR:
3063 return old_bridge_ioctl(argp);
3064 case SIOCGIFNAME:
3065 return dev_ifname32(net, argp);
3066 case SIOCGIFCONF:
3067 return dev_ifconf(net, argp);
3068 case SIOCETHTOOL:
3069 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003070 case SIOCWANDEV:
3071 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003072 case SIOCGIFMAP:
3073 case SIOCSIFMAP:
3074 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003075 case SIOCBONDENSLAVE:
3076 case SIOCBONDRELEASE:
3077 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003078 case SIOCBONDCHANGEACTIVE:
3079 return bond_ioctl(net, cmd, argp);
3080 case SIOCADDRT:
3081 case SIOCDELRT:
3082 return routing_ioctl(net, sock, cmd, argp);
3083 case SIOCGSTAMP:
3084 return do_siocgstamp(net, sock, cmd, argp);
3085 case SIOCGSTAMPNS:
3086 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003087 case SIOCBONDSLAVEINFOQUERY:
3088 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003089 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003090 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003091 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003092
Arnd Bergmann6b960182009-11-06 23:10:54 -08003093 case FIOSETOWN:
3094 case SIOCSPGRP:
3095 case FIOGETOWN:
3096 case SIOCGPGRP:
3097 case SIOCBRADDBR:
3098 case SIOCBRDELBR:
3099 case SIOCGIFVLAN:
3100 case SIOCSIFVLAN:
3101 case SIOCADDDLCI:
3102 case SIOCDELDLCI:
3103 return sock_ioctl(file, cmd, arg);
3104
3105 case SIOCGIFFLAGS:
3106 case SIOCSIFFLAGS:
3107 case SIOCGIFMETRIC:
3108 case SIOCSIFMETRIC:
3109 case SIOCGIFMTU:
3110 case SIOCSIFMTU:
3111 case SIOCGIFMEM:
3112 case SIOCSIFMEM:
3113 case SIOCGIFHWADDR:
3114 case SIOCSIFHWADDR:
3115 case SIOCADDMULTI:
3116 case SIOCDELMULTI:
3117 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003118 case SIOCGIFADDR:
3119 case SIOCSIFADDR:
3120 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003121 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003122 case SIOCGIFBRDADDR:
3123 case SIOCSIFBRDADDR:
3124 case SIOCGIFDSTADDR:
3125 case SIOCSIFDSTADDR:
3126 case SIOCGIFNETMASK:
3127 case SIOCSIFNETMASK:
3128 case SIOCSIFPFLAGS:
3129 case SIOCGIFPFLAGS:
3130 case SIOCGIFTXQLEN:
3131 case SIOCSIFTXQLEN:
3132 case SIOCBRADDIF:
3133 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003134 case SIOCSIFNAME:
3135 case SIOCGMIIPHY:
3136 case SIOCGMIIREG:
3137 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003138 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003139
Arnd Bergmann6b960182009-11-06 23:10:54 -08003140 case SIOCSARP:
3141 case SIOCGARP:
3142 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003143 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003144 return sock_do_ioctl(net, sock, cmd, arg);
3145 }
3146
Arnd Bergmann6b960182009-11-06 23:10:54 -08003147 return -ENOIOCTLCMD;
3148}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003149
Eric Dumazet95c96172012-04-15 05:58:06 +00003150static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003151 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003152{
3153 struct socket *sock = file->private_data;
3154 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003155 struct sock *sk;
3156 struct net *net;
3157
3158 sk = sock->sk;
3159 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003160
3161 if (sock->ops->compat_ioctl)
3162 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3163
David S. Miller87de87d2008-06-03 09:14:03 -07003164 if (ret == -ENOIOCTLCMD &&
3165 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3166 ret = compat_wext_handle_ioctl(net, cmd, arg);
3167
Arnd Bergmann6b960182009-11-06 23:10:54 -08003168 if (ret == -ENOIOCTLCMD)
3169 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3170
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003171 return ret;
3172}
3173#endif
3174
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003175int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3176{
3177 return sock->ops->bind(sock, addr, addrlen);
3178}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003179EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003180
3181int kernel_listen(struct socket *sock, int backlog)
3182{
3183 return sock->ops->listen(sock, backlog);
3184}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003185EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003186
3187int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3188{
3189 struct sock *sk = sock->sk;
3190 int err;
3191
3192 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3193 newsock);
3194 if (err < 0)
3195 goto done;
3196
3197 err = sock->ops->accept(sock, *newsock, flags);
3198 if (err < 0) {
3199 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003200 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003201 goto done;
3202 }
3203
3204 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003205 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003206
3207done:
3208 return err;
3209}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003210EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003211
3212int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003213 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003214{
3215 return sock->ops->connect(sock, addr, addrlen, flags);
3216}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003217EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003218
3219int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3220 int *addrlen)
3221{
3222 return sock->ops->getname(sock, addr, addrlen, 0);
3223}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003224EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003225
3226int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3227 int *addrlen)
3228{
3229 return sock->ops->getname(sock, addr, addrlen, 1);
3230}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003231EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003232
3233int kernel_getsockopt(struct socket *sock, int level, int optname,
3234 char *optval, int *optlen)
3235{
3236 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003237 char __user *uoptval;
3238 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003239 int err;
3240
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003241 uoptval = (char __user __force *) optval;
3242 uoptlen = (int __user __force *) optlen;
3243
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003244 set_fs(KERNEL_DS);
3245 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003246 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003247 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003248 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3249 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003250 set_fs(oldfs);
3251 return err;
3252}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003253EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003254
3255int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003256 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003257{
3258 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003259 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003260 int err;
3261
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003262 uoptval = (char __user __force *) optval;
3263
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003264 set_fs(KERNEL_DS);
3265 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003266 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003267 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003268 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003269 optlen);
3270 set_fs(oldfs);
3271 return err;
3272}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003273EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003274
3275int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3276 size_t size, int flags)
3277{
3278 if (sock->ops->sendpage)
3279 return sock->ops->sendpage(sock, page, offset, size, flags);
3280
3281 return sock_no_sendpage(sock, page, offset, size, flags);
3282}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003283EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003284
3285int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3286{
3287 mm_segment_t oldfs = get_fs();
3288 int err;
3289
3290 set_fs(KERNEL_DS);
3291 err = sock->ops->ioctl(sock, cmd, arg);
3292 set_fs(oldfs);
3293
3294 return err;
3295}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003296EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003297
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003298int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3299{
3300 return sock->ops->shutdown(sock, how);
3301}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003302EXPORT_SYMBOL(kernel_sock_shutdown);