blob: a1bd16106625a14ef95f1da0e560a5638ffa4774 [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)
Al Viroce23e642016-04-11 00:48:00 -0400469static ssize_t sockfs_getxattr(struct dentry *dentry, struct inode *inode,
Masatake YAMATO600e1772012-08-29 10:44:29 +0000470 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
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400590void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000591{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200592 u8 flags = *tx_flags;
593
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400594 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200595 flags |= SKBTX_HW_TSTAMP;
596
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400597 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200598 flags |= SKBTX_SW_TSTAMP;
599
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400600 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200601 flags |= SKBTX_SCHED_TSTAMP;
602
Eric Dumazet140c55d2014-08-06 11:49:29 +0200603 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000604}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400605EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000606
Al Virod8725c82014-12-11 00:02:50 -0500607static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608{
Al Viro01e97e62014-12-15 21:39:31 -0500609 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500610 BUG_ON(ret == -EIOCBQUEUED);
611 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800613
Al Virod8725c82014-12-11 00:02:50 -0500614int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800615{
Al Virod8725c82014-12-11 00:02:50 -0500616 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500617 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800618
Al Virod8725c82014-12-11 00:02:50 -0500619 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800620}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700621EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
623int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
624 struct kvec *vec, size_t num, size_t size)
625{
Al Viro6aa24812015-03-21 19:56:16 -0400626 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500627 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700629EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700631/*
632 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
633 */
634void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
635 struct sk_buff *skb)
636{
Patrick Ohly20d49472009-02-12 05:03:38 +0000637 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400638 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000639 int empty = 1;
640 struct skb_shared_hwtstamps *shhwtstamps =
641 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700642
Patrick Ohly20d49472009-02-12 05:03:38 +0000643 /* Race occurred between timestamp enabling and packet
644 receiving. Fill in the current time for now. */
645 if (need_software_tstamp && skb->tstamp.tv64 == 0)
646 __net_timestamp(skb);
647
648 if (need_software_tstamp) {
649 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
650 struct timeval tv;
651 skb_get_timestamp(skb, &tv);
652 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
653 sizeof(tv), &tv);
654 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400655 struct timespec ts;
656 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000657 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400658 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000659 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700660 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000661
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400662 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400663 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400664 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000665 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400666 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400667 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400668 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400669 empty = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000670 if (!empty)
671 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400672 SCM_TIMESTAMPING, sizeof(tss), &tss);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700673}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300674EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
675
Johannes Berg6e3e9392011-11-09 10:15:42 +0100676void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
677 struct sk_buff *skb)
678{
679 int ack;
680
681 if (!sock_flag(sk, SOCK_WIFI_STATUS))
682 return;
683 if (!skb->wifi_acked_valid)
684 return;
685
686 ack = skb->wifi_acked;
687
688 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
689}
690EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
691
stephen hemminger11165f12010-10-18 14:27:29 +0000692static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
693 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700694{
Eyal Birger744d5a32015-03-01 14:58:31 +0200695 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700696 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200697 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700698}
699
Eric Dumazet767dd032010-04-28 19:14:43 +0000700void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700701 struct sk_buff *skb)
702{
703 sock_recv_timestamp(msg, sk, skb);
704 sock_recv_drops(msg, sk, skb);
705}
Eric Dumazet767dd032010-04-28 19:14:43 +0000706EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700707
Ying Xue1b784142015-03-02 15:37:48 +0800708static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -0400709 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710{
Al Viro2da62902015-03-14 21:13:46 -0400711 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700712}
713
Al Viro2da62902015-03-14 21:13:46 -0400714int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700715{
Al Viro2da62902015-03-14 21:13:46 -0400716 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700717
Al Viro2da62902015-03-14 21:13:46 -0400718 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700720EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
Martin Lucinac1249c02010-12-10 00:04:05 +0000722/**
723 * kernel_recvmsg - Receive a message from a socket (kernel space)
724 * @sock: The socket to receive the message from
725 * @msg: Received message
726 * @vec: Input s/g array for message data
727 * @num: Size of input s/g array
728 * @size: Number of bytes to read
729 * @flags: Message flags (MSG_DONTWAIT, etc...)
730 *
731 * On return the msg structure contains the scatter/gather array passed in the
732 * vec argument. The array is modified so that it consists of the unfilled
733 * portion of the original array.
734 *
735 * The returned value is the total number of bytes received, or an error.
736 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700737int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
738 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739{
740 mm_segment_t oldfs = get_fs();
741 int result;
742
Al Viro6aa24812015-03-21 19:56:16 -0400743 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 set_fs(KERNEL_DS);
Al Viro2da62902015-03-14 21:13:46 -0400745 result = sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 set_fs(oldfs);
747 return result;
748}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700749EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300751static ssize_t sock_sendpage(struct file *file, struct page *page,
752 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753{
754 struct socket *sock;
755 int flags;
756
Eric Dumazetb69aee02005-09-06 14:42:45 -0700757 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Eric Dumazet35f9c092012-04-05 03:05:35 +0000759 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
760 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
761 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
Linus Torvaldse6949582009-08-13 08:28:36 -0700763 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764}
765
Jens Axboe9c55e012007-11-06 23:30:13 -0800766static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700767 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800768 unsigned int flags)
769{
770 struct socket *sock = file->private_data;
771
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800772 if (unlikely(!sock->ops->splice_read))
773 return -EINVAL;
774
Jens Axboe9c55e012007-11-06 23:30:13 -0800775 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
776}
777
Al Viro8ae5e032014-11-28 19:40:50 -0500778static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800779{
Al Viro6d652332015-01-30 16:12:56 -0500780 struct file *file = iocb->ki_filp;
781 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700782 struct msghdr msg = {.msg_iter = *to,
783 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500784 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800785
Al Viro8ae5e032014-11-28 19:40:50 -0500786 if (file->f_flags & O_NONBLOCK)
787 msg.msg_flags = MSG_DONTWAIT;
788
789 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800790 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700791
Christoph Hellwig66ee59a2015-02-11 19:56:46 +0100792 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800793 return 0;
794
Al Viro2da62902015-03-14 21:13:46 -0400795 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e032014-11-28 19:40:50 -0500796 *to = msg.msg_iter;
797 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800798}
799
Al Viro8ae5e032014-11-28 19:40:50 -0500800static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801{
Al Viro6d652332015-01-30 16:12:56 -0500802 struct file *file = iocb->ki_filp;
803 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700804 struct msghdr msg = {.msg_iter = *from,
805 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500806 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800807
Al Viro8ae5e032014-11-28 19:40:50 -0500808 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800809 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700810
Al Viro8ae5e032014-11-28 19:40:50 -0500811 if (file->f_flags & O_NONBLOCK)
812 msg.msg_flags = MSG_DONTWAIT;
813
Al Viro6d652332015-01-30 16:12:56 -0500814 if (sock->type == SOCK_SEQPACKET)
815 msg.msg_flags |= MSG_EOR;
816
Al Virod8725c82014-12-11 00:02:50 -0500817 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e032014-11-28 19:40:50 -0500818 *from = msg.msg_iter;
819 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820}
821
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822/*
823 * Atomic setting of ioctl hooks to avoid race
824 * with module unload.
825 */
826
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800827static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700828static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
Eric W. Biederman881d9662007-09-17 11:56:21 -0700830void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800832 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800834 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835}
836EXPORT_SYMBOL(brioctl_set);
837
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800838static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700839static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840
Eric W. Biederman881d9662007-09-17 11:56:21 -0700841void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800843 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800845 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846}
847EXPORT_SYMBOL(vlan_ioctl_set);
848
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800849static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700850static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700852void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800854 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800856 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857}
858EXPORT_SYMBOL(dlci_ioctl_set);
859
Arnd Bergmann6b960182009-11-06 23:10:54 -0800860static long sock_do_ioctl(struct net *net, struct socket *sock,
861 unsigned int cmd, unsigned long arg)
862{
863 int err;
864 void __user *argp = (void __user *)arg;
865
866 err = sock->ops->ioctl(sock, cmd, arg);
867
868 /*
869 * If this ioctl is unknown try to hand it down
870 * to the NIC driver.
871 */
872 if (err == -ENOIOCTLCMD)
873 err = dev_ioctl(net, cmd, argp);
874
875 return err;
876}
877
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878/*
879 * With an ioctl, arg may well be a user mode pointer, but we don't know
880 * what to do with it - that's up to the protocol still.
881 */
882
883static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
884{
885 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700886 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 void __user *argp = (void __user *)arg;
888 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700889 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
Eric Dumazetb69aee02005-09-06 14:42:45 -0700891 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700892 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900893 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700895 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200897#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
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#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700902 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 case FIOSETOWN:
904 case SIOCSPGRP:
905 err = -EFAULT;
906 if (get_user(pid, (int __user *)argp))
907 break;
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400908 f_setown(sock->file, pid, 1);
909 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 break;
911 case FIOGETOWN:
912 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700913 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700914 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915 break;
916 case SIOCGIFBR:
917 case SIOCSIFBR:
918 case SIOCBRADDBR:
919 case SIOCBRDELBR:
920 err = -ENOPKG;
921 if (!br_ioctl_hook)
922 request_module("bridge");
923
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800924 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700925 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700926 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800927 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 break;
929 case SIOCGIFVLAN:
930 case SIOCSIFVLAN:
931 err = -ENOPKG;
932 if (!vlan_ioctl_hook)
933 request_module("8021q");
934
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800935 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700937 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800938 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 case SIOCADDDLCI:
941 case SIOCDELDLCI:
942 err = -ENOPKG;
943 if (!dlci_ioctl_hook)
944 request_module("dlci");
945
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700946 mutex_lock(&dlci_ioctl_mutex);
947 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700949 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950 break;
951 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -0800952 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700954 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 return err;
956}
957
958int sock_create_lite(int family, int type, int protocol, struct socket **res)
959{
960 int err;
961 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700962
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 err = security_socket_create(family, type, protocol, 1);
964 if (err)
965 goto out;
966
967 sock = sock_alloc();
968 if (!sock) {
969 err = -ENOMEM;
970 goto out;
971 }
972
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700974 err = security_socket_post_create(sock, family, type, protocol, 1);
975 if (err)
976 goto out_release;
977
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978out:
979 *res = sock;
980 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -0700981out_release:
982 sock_release(sock);
983 sock = NULL;
984 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700986EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
988/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700989static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990{
Eliezer Tamircbf55002013-07-08 16:20:34 +0300991 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 struct socket *sock;
993
994 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700995 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 */
Eric Dumazetb69aee02005-09-06 14:42:45 -0700997 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +0300998
Eliezer Tamircbf55002013-07-08 16:20:34 +0300999 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001000 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001001 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001002
1003 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001004 if (wait && (wait->_key & POLL_BUSY_LOOP))
1005 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001006 }
1007
Eliezer Tamircbf55002013-07-08 16:20:34 +03001008 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
1010
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001011static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001013 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
1015 return sock->ops->mmap(file, sock, vma);
1016}
1017
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001018static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 sock_release(SOCKET_I(inode));
1021 return 0;
1022}
1023
1024/*
1025 * Update the socket async list
1026 *
1027 * Fasync_list locking strategy.
1028 *
1029 * 1. fasync_list is modified only under process context socket lock
1030 * i.e. under semaphore.
1031 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001032 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 */
1034
1035static int sock_fasync(int fd, struct file *filp, int on)
1036{
Eric Dumazet989a2972010-04-14 09:55:35 +00001037 struct socket *sock = filp->private_data;
1038 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001039 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040
Eric Dumazet989a2972010-04-14 09:55:35 +00001041 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
1044 lock_sock(sk);
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001045 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
Eric Dumazeteaefd112011-02-18 03:26:36 +00001046 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
Eric Dumazeteaefd112011-02-18 03:26:36 +00001048 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001049 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001050 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001051 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052
Eric Dumazet989a2972010-04-14 09:55:35 +00001053 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054 return 0;
1055}
1056
Eric Dumazetceb5d582015-11-29 20:03:11 -08001057/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001058
Eric Dumazetceb5d582015-11-29 20:03:11 -08001059int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001061 if (!wq || !wq->fasync_list)
1062 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001063
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001064 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001065 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001066 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 break;
1068 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001069 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001070 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 break;
1072 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001073 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001074call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001075 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001077 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001078 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001080
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 return 0;
1082}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001083EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084
Pavel Emelyanov721db932010-09-29 16:06:32 +04001085int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001086 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087{
1088 int err;
1089 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001090 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091
1092 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001093 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 */
1095 if (family < 0 || family >= NPROTO)
1096 return -EAFNOSUPPORT;
1097 if (type < 0 || type >= SOCK_MAX)
1098 return -EINVAL;
1099
1100 /* Compatibility.
1101
1102 This uglymoron is moved from INET layer to here to avoid
1103 deadlock in module load.
1104 */
1105 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001106 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1107 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 family = PF_PACKET;
1109 }
1110
1111 err = security_socket_create(family, type, protocol, kern);
1112 if (err)
1113 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001114
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001115 /*
1116 * Allocate the socket and allow the family to set things up. if
1117 * the protocol is 0, the family is instructed to select an appropriate
1118 * default.
1119 */
1120 sock = sock_alloc();
1121 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001122 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001123 return -ENFILE; /* Not exactly a match, but its the
1124 closest posix thing */
1125 }
1126
1127 sock->type = type;
1128
Johannes Berg95a5afc2008-10-16 15:24:51 -07001129#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001130 /* Attempt to load a protocol module if the find failed.
1131 *
1132 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 * requested real, full-featured networking support upon configuration.
1134 * Otherwise module support will break!
1135 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001136 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001137 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138#endif
1139
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001140 rcu_read_lock();
1141 pf = rcu_dereference(net_families[family]);
1142 err = -EAFNOSUPPORT;
1143 if (!pf)
1144 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
1146 /*
1147 * We will call the ->create function, that possibly is in a loadable
1148 * module, so we have to bump that loadable module refcnt first.
1149 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001150 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 goto out_release;
1152
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001153 /* Now protected by module ref count */
1154 rcu_read_unlock();
1155
Eric Paris3f378b62009-11-05 22:18:14 -08001156 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001157 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001159
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 /*
1161 * Now to bump the refcnt of the [loadable] module that owns this
1162 * socket at sock_release time we decrement its refcnt.
1163 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001164 if (!try_module_get(sock->ops->owner))
1165 goto out_module_busy;
1166
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 /*
1168 * Now that we're done with the ->create function, the [loadable]
1169 * module can have its refcnt decremented
1170 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001171 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001172 err = security_socket_post_create(sock, family, type, protocol, kern);
1173 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001174 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001175 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001177 return 0;
1178
1179out_module_busy:
1180 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001182 sock->ops = NULL;
1183 module_put(pf->owner);
1184out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001186 return err;
1187
1188out_release:
1189 rcu_read_unlock();
1190 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001192EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
1194int sock_create(int family, int type, int protocol, struct socket **res)
1195{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001196 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001198EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001200int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001202 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001204EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001206SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207{
1208 int retval;
1209 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001210 int flags;
1211
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001212 /* Check the SOCK_* constants for consistency. */
1213 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1214 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1215 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1216 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1217
Ulrich Dreppera677a032008-07-23 21:29:17 -07001218 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001219 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001220 return -EINVAL;
1221 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001223 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1224 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1225
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226 retval = sock_create(family, type, protocol, &sock);
1227 if (retval < 0)
1228 goto out;
1229
Ulrich Drepper77d27202008-07-23 21:29:35 -07001230 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 if (retval < 0)
1232 goto out_release;
1233
1234out:
1235 /* It may be already another descriptor 8) Not kernel problem. */
1236 return retval;
1237
1238out_release:
1239 sock_release(sock);
1240 return retval;
1241}
1242
1243/*
1244 * Create a pair of connected sockets.
1245 */
1246
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001247SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1248 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249{
1250 struct socket *sock1, *sock2;
1251 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001252 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001253 int flags;
1254
1255 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001256 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001257 return -EINVAL;
1258 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001260 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1261 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1262
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 /*
1264 * Obtain the first socket and check if the underlying protocol
1265 * supports the socketpair call.
1266 */
1267
1268 err = sock_create(family, type, protocol, &sock1);
1269 if (err < 0)
1270 goto out;
1271
1272 err = sock_create(family, type, protocol, &sock2);
1273 if (err < 0)
1274 goto out_release_1;
1275
1276 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001277 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 goto out_release_both;
1279
Al Viro28407632012-08-17 23:54:15 -04001280 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001281 if (unlikely(fd1 < 0)) {
1282 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001284 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001285
Al Viro28407632012-08-17 23:54:15 -04001286 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001287 if (unlikely(fd2 < 0)) {
1288 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001289 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001290 }
1291
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001292 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301293 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001294 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001295 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001296 }
1297
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001298 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001299 if (IS_ERR(newfile2)) {
1300 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001301 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001302 }
1303
Yann Droneaudd73aa282013-12-09 22:42:20 +01001304 err = put_user(fd1, &usockvec[0]);
1305 if (err)
1306 goto out_fput_both;
1307
1308 err = put_user(fd2, &usockvec[1]);
1309 if (err)
1310 goto out_fput_both;
1311
Al Viro157cf642008-12-14 04:57:47 -05001312 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001313
Al Virodb349502007-02-07 01:48:00 -05001314 fd_install(fd1, newfile1);
1315 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 /* fd1 and fd2 may be already another descriptors.
1317 * Not kernel problem.
1318 */
1319
Yann Droneaudd73aa282013-12-09 22:42:20 +01001320 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
Yann Droneaudd73aa282013-12-09 22:42:20 +01001322out_fput_both:
1323 fput(newfile2);
1324 fput(newfile1);
1325 put_unused_fd(fd2);
1326 put_unused_fd(fd1);
1327 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
Yann Droneaudd73aa282013-12-09 22:42:20 +01001329out_fput_1:
1330 fput(newfile1);
1331 put_unused_fd(fd2);
1332 put_unused_fd(fd1);
1333 sock_release(sock2);
1334 goto out;
1335
1336out_put_unused_both:
1337 put_unused_fd(fd2);
1338out_put_unused_1:
1339 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001341 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001343 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344out:
1345 return err;
1346}
1347
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348/*
1349 * Bind a name to a socket. Nothing much to do here since it's
1350 * the protocol's responsibility to handle the local address.
1351 *
1352 * We move the socket address to kernel space before we call
1353 * the protocol layer (having also checked the address is ok).
1354 */
1355
Heiko Carstens20f37032009-01-14 14:14:23 +01001356SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357{
1358 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001359 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001360 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001362 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001363 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001364 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001365 if (err >= 0) {
1366 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001367 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001368 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001369 if (!err)
1370 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001371 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001372 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001374 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001375 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 return err;
1377}
1378
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379/*
1380 * Perform a listen. Basically, we allow the protocol to do anything
1381 * necessary for a listen, and if that works, we mark the socket as
1382 * ready for listening.
1383 */
1384
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001385SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386{
1387 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001388 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001389 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001390
1391 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1392 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001393 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001394 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001395 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396
1397 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001398 if (!err)
1399 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001401 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 }
1403 return err;
1404}
1405
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406/*
1407 * For accept, we attempt to create a new socket, set up the link
1408 * with the client, wake up the client, then return the new
1409 * connected fd. We collect the address of the connector in kernel
1410 * space and move it to user at the very end. This is unclean because
1411 * we open the socket then return an error.
1412 *
1413 * 1003.1g adds the ability to recvmsg() to query connection pending
1414 * status to recvmsg. We need to add that support in a way thats
1415 * clean when we restucture accept also.
1416 */
1417
Heiko Carstens20f37032009-01-14 14:14:23 +01001418SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1419 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420{
1421 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001422 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001423 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001424 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
Ulrich Drepper77d27202008-07-23 21:29:35 -07001426 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001427 return -EINVAL;
1428
1429 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1430 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1431
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001432 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 if (!sock)
1434 goto out;
1435
1436 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001437 newsock = sock_alloc();
1438 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 goto out_put;
1440
1441 newsock->type = sock->type;
1442 newsock->ops = sock->ops;
1443
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 /*
1445 * We don't need try_module_get here, as the listening socket (sock)
1446 * has the protocol module (sock->ops->owner) held.
1447 */
1448 __module_get(newsock->ops->owner);
1449
Al Viro28407632012-08-17 23:54:15 -04001450 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001451 if (unlikely(newfd < 0)) {
1452 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001453 sock_release(newsock);
1454 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001455 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001456 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301457 if (IS_ERR(newfile)) {
Al Viro28407632012-08-17 23:54:15 -04001458 err = PTR_ERR(newfile);
1459 put_unused_fd(newfd);
1460 sock_release(newsock);
1461 goto out_put;
1462 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001463
Frank Filza79af592005-09-27 15:23:38 -07001464 err = security_socket_accept(sock, newsock);
1465 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001466 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001467
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1469 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001470 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471
1472 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001473 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001474 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001476 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001478 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001479 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001481 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 }
1483
1484 /* File flags are not inherited via accept() unlike another OSes. */
1485
David S. Miller39d8c1b2006-03-20 17:13:49 -08001486 fd_install(newfd, newfile);
1487 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001490 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491out:
1492 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001493out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001494 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001495 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 goto out_put;
1497}
1498
Heiko Carstens20f37032009-01-14 14:14:23 +01001499SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1500 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001501{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001502 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001503}
1504
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505/*
1506 * Attempt to connect to a socket with the server address. The address
1507 * is in user space so we verify it is OK and move it to kernel space.
1508 *
1509 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1510 * break bindings
1511 *
1512 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1513 * other SEQPACKET protocols that take time to connect() as it doesn't
1514 * include the -EINPROGRESS status for such sockets.
1515 */
1516
Heiko Carstens20f37032009-01-14 14:14:23 +01001517SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1518 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519{
1520 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001521 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001522 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001524 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 if (!sock)
1526 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001527 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 if (err < 0)
1529 goto out_put;
1530
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001531 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001532 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 if (err)
1534 goto out_put;
1535
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001536 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 sock->file->f_flags);
1538out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001539 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540out:
1541 return err;
1542}
1543
1544/*
1545 * Get the local address ('name') of a socket object. Move the obtained
1546 * name to user space.
1547 */
1548
Heiko Carstens20f37032009-01-14 14:14:23 +01001549SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1550 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551{
1552 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001553 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001554 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001555
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001556 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 if (!sock)
1558 goto out;
1559
1560 err = security_socket_getsockname(sock);
1561 if (err)
1562 goto out_put;
1563
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001564 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 if (err)
1566 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001567 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
1569out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001570 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571out:
1572 return err;
1573}
1574
1575/*
1576 * Get the remote address ('name') of a socket object. Move the obtained
1577 * name to user space.
1578 */
1579
Heiko Carstens20f37032009-01-14 14:14:23 +01001580SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1581 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582{
1583 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001584 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001585 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001587 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1588 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 err = security_socket_getpeername(sock);
1590 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001591 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592 return err;
1593 }
1594
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001595 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001596 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001597 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001599 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001600 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001601 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 }
1603 return err;
1604}
1605
1606/*
1607 * Send a datagram to a given address. We move the address into kernel
1608 * space and check the user space data area is readable before invoking
1609 * the protocol.
1610 */
1611
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001612SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001613 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001614 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615{
1616 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001617 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 int err;
1619 struct msghdr msg;
1620 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001621 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001622
Al Viro602bd0e2015-03-21 19:12:32 -04001623 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1624 if (unlikely(err))
1625 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001626 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1627 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001628 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001629
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001630 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001631 msg.msg_control = NULL;
1632 msg.msg_controllen = 0;
1633 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001634 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001635 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 if (err < 0)
1637 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001638 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001639 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 }
1641 if (sock->file->f_flags & O_NONBLOCK)
1642 flags |= MSG_DONTWAIT;
1643 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05001644 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001646out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001647 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001648out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 return err;
1650}
1651
1652/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001653 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 */
1655
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001656SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001657 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658{
1659 return sys_sendto(fd, buff, len, flags, NULL, 0);
1660}
1661
1662/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001663 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 * sender. We verify the buffers are writable and if needed move the
1665 * sender address from kernel to user space.
1666 */
1667
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001668SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001669 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001670 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671{
1672 struct socket *sock;
1673 struct iovec iov;
1674 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001675 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001676 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001677 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678
Al Viro602bd0e2015-03-21 19:12:32 -04001679 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1680 if (unlikely(err))
1681 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001682 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001684 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001686 msg.msg_control = NULL;
1687 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001688 /* Save some cycles and don't copy the address if not needed */
1689 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1690 /* We assume all kernel code knows the size of sockaddr_storage */
1691 msg.msg_namelen = 0;
tadeusz.struk@intel.com130ed5d2015-12-15 10:46:17 -08001692 msg.msg_iocb = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 if (sock->file->f_flags & O_NONBLOCK)
1694 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04001695 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001697 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001698 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001699 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001700 if (err2 < 0)
1701 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001703
1704 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001705out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 return err;
1707}
1708
1709/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001710 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 */
1712
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001713SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1714 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715{
1716 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1717}
1718
1719/*
1720 * Set a socket option. Because we don't know the option lengths we have
1721 * to pass the user mode parameter for the protocols to sort out.
1722 */
1723
Heiko Carstens20f37032009-01-14 14:14:23 +01001724SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1725 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001727 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 struct socket *sock;
1729
1730 if (optlen < 0)
1731 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001732
1733 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1734 if (sock != NULL) {
1735 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001736 if (err)
1737 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
1739 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001740 err =
1741 sock_setsockopt(sock, level, optname, optval,
1742 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001744 err =
1745 sock->ops->setsockopt(sock, level, optname, optval,
1746 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001747out_put:
1748 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 }
1750 return err;
1751}
1752
1753/*
1754 * Get a socket option. Because we don't know the option lengths we have
1755 * to pass a user mode parameter for the protocols to sort out.
1756 */
1757
Heiko Carstens20f37032009-01-14 14:14:23 +01001758SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1759 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001761 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 struct socket *sock;
1763
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001764 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1765 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001766 err = security_socket_getsockopt(sock, level, optname);
1767 if (err)
1768 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769
1770 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001771 err =
1772 sock_getsockopt(sock, level, optname, optval,
1773 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001775 err =
1776 sock->ops->getsockopt(sock, level, optname, optval,
1777 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001778out_put:
1779 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 }
1781 return err;
1782}
1783
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784/*
1785 * Shutdown a socket.
1786 */
1787
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001788SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001790 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 struct socket *sock;
1792
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001793 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1794 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001796 if (!err)
1797 err = sock->ops->shutdown(sock, how);
1798 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 }
1800 return err;
1801}
1802
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001803/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 * fields which are the same type (int / unsigned) on our platforms.
1805 */
1806#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1807#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1808#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1809
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001810struct used_address {
1811 struct sockaddr_storage name;
1812 unsigned int name_len;
1813};
1814
Al Viroda184282015-03-21 19:29:06 -04001815static int copy_msghdr_from_user(struct msghdr *kmsg,
1816 struct user_msghdr __user *umsg,
1817 struct sockaddr __user **save_addr,
1818 struct iovec **iov)
Dan Carpenter1661bf32013-10-03 00:27:20 +03001819{
Al Viro08adb7d2014-11-10 20:23:13 -05001820 struct sockaddr __user *uaddr;
1821 struct iovec __user *uiov;
Al Viroc0371da2014-11-24 10:42:55 -05001822 size_t nr_segs;
Al Viro08adb7d2014-11-10 20:23:13 -05001823 ssize_t err;
1824
1825 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1826 __get_user(uaddr, &umsg->msg_name) ||
1827 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1828 __get_user(uiov, &umsg->msg_iov) ||
Al Viroc0371da2014-11-24 10:42:55 -05001829 __get_user(nr_segs, &umsg->msg_iovlen) ||
Al Viro08adb7d2014-11-10 20:23:13 -05001830 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1831 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1832 __get_user(kmsg->msg_flags, &umsg->msg_flags))
Dan Carpenter1661bf32013-10-03 00:27:20 +03001833 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001834
Al Viro08adb7d2014-11-10 20:23:13 -05001835 if (!uaddr)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07001836 kmsg->msg_namelen = 0;
1837
Matthew Leachdbb490b2014-03-11 11:58:27 +00001838 if (kmsg->msg_namelen < 0)
1839 return -EINVAL;
1840
Dan Carpenter1661bf32013-10-03 00:27:20 +03001841 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001842 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05001843
1844 if (save_addr)
1845 *save_addr = uaddr;
1846
1847 if (uaddr && kmsg->msg_namelen) {
1848 if (!save_addr) {
1849 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1850 kmsg->msg_name);
1851 if (err < 0)
1852 return err;
1853 }
1854 } else {
1855 kmsg->msg_name = NULL;
1856 kmsg->msg_namelen = 0;
1857 }
1858
Al Viroc0371da2014-11-24 10:42:55 -05001859 if (nr_segs > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05001860 return -EMSGSIZE;
1861
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001862 kmsg->msg_iocb = NULL;
1863
Al Viroda184282015-03-21 19:29:06 -04001864 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1865 UIO_FASTIOV, iov, &kmsg->msg_iter);
Dan Carpenter1661bf32013-10-03 00:27:20 +03001866}
1867
Al Viro666547f2014-04-06 14:03:05 -04001868static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001869 struct msghdr *msg_sys, unsigned int flags,
Tom Herbert28a94d82016-03-07 14:11:02 -08001870 struct used_address *used_address,
1871 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001873 struct compat_msghdr __user *msg_compat =
1874 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001875 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001877 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001878 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1879 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05001881 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05001882 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001883
Al Viro08adb7d2014-11-10 20:23:13 -05001884 msg_sys->msg_name = &address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885
Al Viro08449322014-11-09 22:33:45 -05001886 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05001887 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
Al Viro08449322014-11-09 22:33:45 -05001888 else
Al Viro08adb7d2014-11-10 20:23:13 -05001889 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001890 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04001891 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892
1893 err = -ENOBUFS;
1894
Anton Blanchard228e5482011-05-02 20:21:35 +00001895 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 goto out_freeiov;
Tom Herbert28a94d82016-03-07 14:11:02 -08001897 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00001898 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001900 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001901 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001902 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 if (err)
1904 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001905 ctl_buf = msg_sys->msg_control;
1906 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 } else if (ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001908 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001910 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911 goto out_freeiov;
1912 }
1913 err = -EFAULT;
1914 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00001915 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1917 * checking falls down on this.
1918 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001919 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00001920 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001921 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00001923 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924 }
Anton Blanchard228e5482011-05-02 20:21:35 +00001925 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926
1927 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00001928 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001929 /*
1930 * If this is sendmmsg() and current destination address is same as
1931 * previously succeeded address, omit asking LSM's decision.
1932 * used_address->name_len is initialized to UINT_MAX so that the first
1933 * destination address never matches.
1934 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001935 if (used_address && msg_sys->msg_name &&
1936 used_address->name_len == msg_sys->msg_namelen &&
1937 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001938 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05001939 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001940 goto out_freectl;
1941 }
Al Virod8725c82014-12-11 00:02:50 -05001942 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001943 /*
1944 * If this is sendmmsg() and sending to current destination address was
1945 * successful, remember it.
1946 */
1947 if (used_address && err >= 0) {
1948 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001949 if (msg_sys->msg_name)
1950 memcpy(&used_address->name, msg_sys->msg_name,
1951 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001952 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953
1954out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001955 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1957out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04001958 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00001959 return err;
1960}
1961
1962/*
1963 * BSD sendmsg interface
1964 */
1965
Al Viro666547f2014-04-06 14:03:05 -04001966long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00001967{
1968 int fput_needed, err;
1969 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07001970 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00001971
Andy Lutomirski1be374a2013-05-22 14:07:44 -07001972 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00001973 if (!sock)
1974 goto out;
1975
Tom Herbert28a94d82016-03-07 14:11:02 -08001976 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00001977
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001978 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001979out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 return err;
1981}
1982
Al Viro666547f2014-04-06 14:03:05 -04001983SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00001984{
1985 if (flags & MSG_CMSG_COMPAT)
1986 return -EINVAL;
1987 return __sys_sendmsg(fd, msg, flags);
1988}
1989
Anton Blanchard228e5482011-05-02 20:21:35 +00001990/*
1991 * Linux sendmmsg interface
1992 */
1993
1994int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1995 unsigned int flags)
1996{
1997 int fput_needed, err, datagrams;
1998 struct socket *sock;
1999 struct mmsghdr __user *entry;
2000 struct compat_mmsghdr __user *compat_entry;
2001 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002002 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002003 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002004
Anton Blanchard98382f42011-08-04 14:07:39 +00002005 if (vlen > UIO_MAXIOV)
2006 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002007
2008 datagrams = 0;
2009
2010 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2011 if (!sock)
2012 return err;
2013
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002014 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002015 entry = mmsg;
2016 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002017 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002018 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002019
2020 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002021 if (datagrams == vlen - 1)
2022 flags = oflags;
2023
Anton Blanchard228e5482011-05-02 20:21:35 +00002024 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002025 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002026 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002027 if (err < 0)
2028 break;
2029 err = __put_user(err, &compat_entry->msg_len);
2030 ++compat_entry;
2031 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002032 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002033 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002034 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002035 if (err < 0)
2036 break;
2037 err = put_user(err, &entry->msg_len);
2038 ++entry;
2039 }
2040
2041 if (err)
2042 break;
2043 ++datagrams;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002044 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002045 }
2046
Anton Blanchard228e5482011-05-02 20:21:35 +00002047 fput_light(sock->file, fput_needed);
2048
Anton Blanchard728ffb82011-08-04 14:07:38 +00002049 /* We only return an error if no datagrams were able to be sent */
2050 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002051 return datagrams;
2052
Anton Blanchard228e5482011-05-02 20:21:35 +00002053 return err;
2054}
2055
2056SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2057 unsigned int, vlen, unsigned int, flags)
2058{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002059 if (flags & MSG_CMSG_COMPAT)
2060 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002061 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2062}
2063
Al Viro666547f2014-04-06 14:03:05 -04002064static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002065 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002067 struct compat_msghdr __user *msg_compat =
2068 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002070 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002072 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002073 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074
2075 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002076 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077
2078 /* user mode address pointers */
2079 struct sockaddr __user *uaddr;
Al Viro08adb7d2014-11-10 20:23:13 -05002080 int __user *uaddr_len = COMPAT_NAMELEN(msg);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002081
Al Viro08adb7d2014-11-10 20:23:13 -05002082 msg_sys->msg_name = &addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002084 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002085 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002086 else
Al Viro08adb7d2014-11-10 20:23:13 -05002087 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04002089 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002091 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2092 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002093
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002094 /* We assume all kernel code knows the size of sockaddr_storage */
2095 msg_sys->msg_namelen = 0;
2096
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 if (sock->file->f_flags & O_NONBLOCK)
2098 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002099 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002100 if (err < 0)
2101 goto out_freeiov;
2102 len = err;
2103
2104 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002105 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002106 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002107 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 if (err < 0)
2109 goto out_freeiov;
2110 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002111 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002112 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 if (err)
2114 goto out_freeiov;
2115 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002116 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 &msg_compat->msg_controllen);
2118 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002119 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120 &msg->msg_controllen);
2121 if (err)
2122 goto out_freeiov;
2123 err = len;
2124
2125out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04002126 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002127 return err;
2128}
2129
2130/*
2131 * BSD recvmsg interface
2132 */
2133
Al Viro666547f2014-04-06 14:03:05 -04002134long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002135{
2136 int fput_needed, err;
2137 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002138 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002139
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002140 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002141 if (!sock)
2142 goto out;
2143
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002144 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002145
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002146 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147out:
2148 return err;
2149}
2150
Al Viro666547f2014-04-06 14:03:05 -04002151SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002152 unsigned int, flags)
2153{
2154 if (flags & MSG_CMSG_COMPAT)
2155 return -EINVAL;
2156 return __sys_recvmsg(fd, msg, flags);
2157}
2158
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002159/*
2160 * Linux recvmmsg interface
2161 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002163int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2164 unsigned int flags, struct timespec *timeout)
2165{
2166 int fput_needed, err, datagrams;
2167 struct socket *sock;
2168 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002169 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002170 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002171 struct timespec64 end_time;
2172 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002173
2174 if (timeout &&
2175 poll_select_set_timeout(&end_time, timeout->tv_sec,
2176 timeout->tv_nsec))
2177 return -EINVAL;
2178
2179 datagrams = 0;
2180
2181 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2182 if (!sock)
2183 return err;
2184
2185 err = sock_error(sock->sk);
2186 if (err)
2187 goto out_put;
2188
2189 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002190 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002191
2192 while (datagrams < vlen) {
2193 /*
2194 * No need to ask LSM for more than the first datagram.
2195 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002196 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002197 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002198 &msg_sys, flags & ~MSG_WAITFORONE,
2199 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002200 if (err < 0)
2201 break;
2202 err = __put_user(err, &compat_entry->msg_len);
2203 ++compat_entry;
2204 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002205 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002206 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002207 &msg_sys, flags & ~MSG_WAITFORONE,
2208 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002209 if (err < 0)
2210 break;
2211 err = put_user(err, &entry->msg_len);
2212 ++entry;
2213 }
2214
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002215 if (err)
2216 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002217 ++datagrams;
2218
Brandon L Black71c5c152010-03-26 16:18:03 +00002219 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2220 if (flags & MSG_WAITFORONE)
2221 flags |= MSG_DONTWAIT;
2222
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002223 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002224 ktime_get_ts64(&timeout64);
2225 *timeout = timespec64_to_timespec(
2226 timespec64_sub(end_time, timeout64));
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002227 if (timeout->tv_sec < 0) {
2228 timeout->tv_sec = timeout->tv_nsec = 0;
2229 break;
2230 }
2231
2232 /* Timeout, return less than vlen datagrams */
2233 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2234 break;
2235 }
2236
2237 /* Out of band data, return right away */
2238 if (msg_sys.msg_flags & MSG_OOB)
2239 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002240 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002241 }
2242
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002243 if (err == 0)
2244 goto out_put;
2245
2246 if (datagrams == 0) {
2247 datagrams = err;
2248 goto out_put;
2249 }
2250
2251 /*
2252 * We may return less entries than requested (vlen) if the
2253 * sock is non block and there aren't enough datagrams...
2254 */
2255 if (err != -EAGAIN) {
2256 /*
2257 * ... or if recvmsg returns an error after we
2258 * received some datagrams, where we record the
2259 * error to return on the next call or if the
2260 * app asks about it using getsockopt(SO_ERROR).
2261 */
2262 sock->sk->sk_err = -err;
2263 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002264out_put:
2265 fput_light(sock->file, fput_needed);
2266
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002267 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002268}
2269
2270SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2271 unsigned int, vlen, unsigned int, flags,
2272 struct timespec __user *, timeout)
2273{
2274 int datagrams;
2275 struct timespec timeout_sys;
2276
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002277 if (flags & MSG_CMSG_COMPAT)
2278 return -EINVAL;
2279
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002280 if (!timeout)
2281 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2282
2283 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2284 return -EFAULT;
2285
2286 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2287
2288 if (datagrams > 0 &&
2289 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2290 datagrams = -EFAULT;
2291
2292 return datagrams;
2293}
2294
2295#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296/* Argument list sizes for sys_socketcall */
2297#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002298static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002299 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2300 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2301 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002302 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002303};
2304
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305#undef AL
2306
2307/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002308 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 *
2310 * Argument checking cleaned up. Saved 20% in size.
2311 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002312 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313 */
2314
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002315SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316{
Chen Gang2950fa92013-04-07 16:55:23 +08002317 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002318 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002320 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321
Anton Blanchard228e5482011-05-02 20:21:35 +00002322 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 return -EINVAL;
2324
Arjan van de Ven47379052009-09-28 12:57:44 -07002325 len = nargs[call];
2326 if (len > sizeof(a))
2327 return -EINVAL;
2328
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002330 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002332
Chen Gang2950fa92013-04-07 16:55:23 +08002333 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2334 if (err)
2335 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002336
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002337 a0 = a[0];
2338 a1 = a[1];
2339
2340 switch (call) {
2341 case SYS_SOCKET:
2342 err = sys_socket(a0, a1, a[2]);
2343 break;
2344 case SYS_BIND:
2345 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2346 break;
2347 case SYS_CONNECT:
2348 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2349 break;
2350 case SYS_LISTEN:
2351 err = sys_listen(a0, a1);
2352 break;
2353 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002354 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2355 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002356 break;
2357 case SYS_GETSOCKNAME:
2358 err =
2359 sys_getsockname(a0, (struct sockaddr __user *)a1,
2360 (int __user *)a[2]);
2361 break;
2362 case SYS_GETPEERNAME:
2363 err =
2364 sys_getpeername(a0, (struct sockaddr __user *)a1,
2365 (int __user *)a[2]);
2366 break;
2367 case SYS_SOCKETPAIR:
2368 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2369 break;
2370 case SYS_SEND:
2371 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2372 break;
2373 case SYS_SENDTO:
2374 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2375 (struct sockaddr __user *)a[4], a[5]);
2376 break;
2377 case SYS_RECV:
2378 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2379 break;
2380 case SYS_RECVFROM:
2381 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2382 (struct sockaddr __user *)a[4],
2383 (int __user *)a[5]);
2384 break;
2385 case SYS_SHUTDOWN:
2386 err = sys_shutdown(a0, a1);
2387 break;
2388 case SYS_SETSOCKOPT:
2389 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2390 break;
2391 case SYS_GETSOCKOPT:
2392 err =
2393 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2394 (int __user *)a[4]);
2395 break;
2396 case SYS_SENDMSG:
Al Viro666547f2014-04-06 14:03:05 -04002397 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002398 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002399 case SYS_SENDMMSG:
2400 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2401 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002402 case SYS_RECVMSG:
Al Viro666547f2014-04-06 14:03:05 -04002403 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002404 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002405 case SYS_RECVMMSG:
2406 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2407 (struct timespec __user *)a[4]);
2408 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002409 case SYS_ACCEPT4:
2410 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2411 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002412 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002413 default:
2414 err = -EINVAL;
2415 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416 }
2417 return err;
2418}
2419
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002420#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002422/**
2423 * sock_register - add a socket protocol handler
2424 * @ops: description of protocol
2425 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 * This function is called by a protocol handler that wants to
2427 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002428 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002429 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002431int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002432{
2433 int err;
2434
2435 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002436 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437 return -ENOBUFS;
2438 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002439
2440 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002441 if (rcu_dereference_protected(net_families[ops->family],
2442 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002443 err = -EEXIST;
2444 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002445 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446 err = 0;
2447 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002448 spin_unlock(&net_family_lock);
2449
Yang Yingliang3410f222014-02-12 17:09:55 +08002450 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 return err;
2452}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002453EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002455/**
2456 * sock_unregister - remove a protocol handler
2457 * @family: protocol family to remove
2458 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002459 * This function is called by a protocol handler that wants to
2460 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002461 * new socket creation.
2462 *
2463 * If protocol handler is a module, then it can use module reference
2464 * counts to protect against new references. If protocol handler is not
2465 * a module then it needs to provide its own protection in
2466 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002468void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002470 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002472 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002473 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002474 spin_unlock(&net_family_lock);
2475
2476 synchronize_rcu();
2477
Yang Yingliang3410f222014-02-12 17:09:55 +08002478 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002480EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002481
Andi Kleen77d76ea2005-12-22 12:43:42 -08002482static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002484 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002485 /*
2486 * Initialize the network sysctl infrastructure.
2487 */
2488 err = net_sysctl_init();
2489 if (err)
2490 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002491
Linus Torvalds1da177e2005-04-16 15:20:36 -07002492 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002493 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002494 */
2495 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496
2497 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002498 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002499 */
2500
2501 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002502
2503 err = register_filesystem(&sock_fs_type);
2504 if (err)
2505 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002506 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002507 if (IS_ERR(sock_mnt)) {
2508 err = PTR_ERR(sock_mnt);
2509 goto out_mount;
2510 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002511
2512 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513 */
2514
2515#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002516 err = netfilter_init();
2517 if (err)
2518 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002519#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002520
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002521 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002522
Nick Pigginb3e19d92011-01-07 17:50:11 +11002523out:
2524 return err;
2525
2526out_mount:
2527 unregister_filesystem(&sock_fs_type);
2528out_fs:
2529 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530}
2531
Andi Kleen77d76ea2005-12-22 12:43:42 -08002532core_initcall(sock_init); /* early initcall */
2533
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534#ifdef CONFIG_PROC_FS
2535void socket_seq_show(struct seq_file *seq)
2536{
2537 int cpu;
2538 int counter = 0;
2539
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002540 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002541 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542
2543 /* It can be negative, by the way. 8) */
2544 if (counter < 0)
2545 counter = 0;
2546
2547 seq_printf(seq, "sockets: used %d\n", counter);
2548}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002549#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002551#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002552static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002553 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002554{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002555 mm_segment_t old_fs = get_fs();
2556 struct timeval ktv;
2557 int err;
2558
2559 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002560 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002561 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002562 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002563 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002564
Arnd Bergmann7a229382009-11-06 23:00:29 -08002565 return err;
2566}
2567
Arnd Bergmann6b960182009-11-06 23:10:54 -08002568static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002569 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002570{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002571 mm_segment_t old_fs = get_fs();
2572 struct timespec kts;
2573 int err;
2574
2575 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002576 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002577 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002578 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002579 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002580
Arnd Bergmann7a229382009-11-06 23:00:29 -08002581 return err;
2582}
2583
Arnd Bergmann6b960182009-11-06 23:10:54 -08002584static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002585{
2586 struct ifreq __user *uifr;
2587 int err;
2588
2589 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002590 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002591 return -EFAULT;
2592
Arnd Bergmann6b960182009-11-06 23:10:54 -08002593 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002594 if (err)
2595 return err;
2596
Arnd Bergmann6b960182009-11-06 23:10:54 -08002597 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002598 return -EFAULT;
2599
2600 return 0;
2601}
2602
Arnd Bergmann6b960182009-11-06 23:10:54 -08002603static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002604{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002605 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002606 struct ifconf ifc;
2607 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002608 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002609 struct ifreq __user *ifr;
2610 unsigned int i, j;
2611 int err;
2612
Arnd Bergmann6b960182009-11-06 23:10:54 -08002613 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002614 return -EFAULT;
2615
Mathias Krause43da5f22012-08-15 11:31:57 +00002616 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002617 if (ifc32.ifcbuf == 0) {
2618 ifc32.ifc_len = 0;
2619 ifc.ifc_len = 0;
2620 ifc.ifc_req = NULL;
2621 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2622 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002623 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2624 sizeof(struct ifreq);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002625 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2626 ifc.ifc_len = len;
2627 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2628 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002629 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002630 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002631 return -EFAULT;
2632 ifr++;
2633 ifr32++;
2634 }
2635 }
2636 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2637 return -EFAULT;
2638
Arnd Bergmann6b960182009-11-06 23:10:54 -08002639 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002640 if (err)
2641 return err;
2642
2643 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2644 return -EFAULT;
2645
2646 ifr = ifc.ifc_req;
2647 ifr32 = compat_ptr(ifc32.ifcbuf);
2648 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002649 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2650 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2651 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002652 return -EFAULT;
2653 ifr32++;
2654 ifr++;
2655 }
2656
2657 if (ifc32.ifcbuf == 0) {
2658 /* Translate from 64-bit structure multiple to
2659 * a 32-bit one.
2660 */
2661 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002662 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002663 ifc32.ifc_len = i;
2664 } else {
2665 ifc32.ifc_len = i;
2666 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002667 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002668 return -EFAULT;
2669
2670 return 0;
2671}
2672
Arnd Bergmann6b960182009-11-06 23:10:54 -08002673static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002674{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002675 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2676 bool convert_in = false, convert_out = false;
2677 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2678 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002679 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002680 u32 rule_cnt = 0, actual_rule_cnt;
2681 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002682 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002683 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002684
2685 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2686 return -EFAULT;
2687
Ben Hutchings3a7da392011-03-17 07:34:32 +00002688 compat_rxnfc = compat_ptr(data);
2689
2690 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002691 return -EFAULT;
2692
Ben Hutchings3a7da392011-03-17 07:34:32 +00002693 /* Most ethtool structures are defined without padding.
2694 * Unfortunately struct ethtool_rxnfc is an exception.
2695 */
2696 switch (ethcmd) {
2697 default:
2698 break;
2699 case ETHTOOL_GRXCLSRLALL:
2700 /* Buffer size is variable */
2701 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2702 return -EFAULT;
2703 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2704 return -ENOMEM;
2705 buf_size += rule_cnt * sizeof(u32);
2706 /* fall through */
2707 case ETHTOOL_GRXRINGS:
2708 case ETHTOOL_GRXCLSRLCNT:
2709 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002710 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002711 convert_out = true;
2712 /* fall through */
2713 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002714 buf_size += sizeof(struct ethtool_rxnfc);
2715 convert_in = true;
2716 break;
2717 }
2718
2719 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002720 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002721
2722 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2723 return -EFAULT;
2724
2725 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2726 &ifr->ifr_ifru.ifru_data))
2727 return -EFAULT;
2728
2729 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002730 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002731 * fs.ring_cookie and at the end of fs, but nowhere else.
2732 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002733 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2734 sizeof(compat_rxnfc->fs.m_ext) !=
2735 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2736 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002737 BUILD_BUG_ON(
2738 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2739 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2740 offsetof(struct ethtool_rxnfc, fs.location) -
2741 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2742
2743 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002744 (void __user *)(&rxnfc->fs.m_ext + 1) -
2745 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002746 copy_in_user(&rxnfc->fs.ring_cookie,
2747 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002748 (void __user *)(&rxnfc->fs.location + 1) -
2749 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002750 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2751 sizeof(rxnfc->rule_cnt)))
2752 return -EFAULT;
2753 }
2754
2755 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2756 if (ret)
2757 return ret;
2758
2759 if (convert_out) {
2760 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002761 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2762 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002763 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2764 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002765 (const void __user *)(&rxnfc->fs.location + 1) -
2766 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002767 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2768 sizeof(rxnfc->rule_cnt)))
2769 return -EFAULT;
2770
2771 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2772 /* As an optimisation, we only copy the actual
2773 * number of rules that the underlying
2774 * function returned. Since Mallory might
2775 * change the rule count in user memory, we
2776 * check that it is less than the rule count
2777 * originally given (as the user buffer size),
2778 * which has been range-checked.
2779 */
2780 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2781 return -EFAULT;
2782 if (actual_rule_cnt < rule_cnt)
2783 rule_cnt = actual_rule_cnt;
2784 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2785 &rxnfc->rule_locs[0],
2786 rule_cnt * sizeof(u32)))
2787 return -EFAULT;
2788 }
2789 }
2790
2791 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002792}
2793
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002794static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2795{
2796 void __user *uptr;
2797 compat_uptr_t uptr32;
2798 struct ifreq __user *uifr;
2799
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002800 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002801 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2802 return -EFAULT;
2803
2804 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2805 return -EFAULT;
2806
2807 uptr = compat_ptr(uptr32);
2808
2809 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2810 return -EFAULT;
2811
2812 return dev_ioctl(net, SIOCWANDEV, uifr);
2813}
2814
Arnd Bergmann6b960182009-11-06 23:10:54 -08002815static int bond_ioctl(struct net *net, unsigned int cmd,
2816 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002817{
2818 struct ifreq kifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002819 mm_segment_t old_fs;
2820 int err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002821
2822 switch (cmd) {
2823 case SIOCBONDENSLAVE:
2824 case SIOCBONDRELEASE:
2825 case SIOCBONDSETHWADDR:
2826 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002827 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002828 return -EFAULT;
2829
2830 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002831 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002832 err = dev_ioctl(net, cmd,
2833 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002834 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002835
2836 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002837 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002838 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002839 }
Arnd Bergmann7a229382009-11-06 23:00:29 -08002840}
2841
Ben Hutchings590d4692013-11-18 17:04:13 +00002842/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2843static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08002844 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002845{
2846 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002847 char tmp_buf[IFNAMSIZ];
2848 void __user *data64;
2849 u32 data32;
2850
2851 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2852 IFNAMSIZ))
2853 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002854 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002855 return -EFAULT;
2856 data64 = compat_ptr(data32);
2857
2858 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2859
Arnd Bergmann7a229382009-11-06 23:00:29 -08002860 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2861 IFNAMSIZ))
2862 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002863 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002864 return -EFAULT;
2865
Arnd Bergmann6b960182009-11-06 23:10:54 -08002866 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002867}
2868
Arnd Bergmann6b960182009-11-06 23:10:54 -08002869static int dev_ifsioc(struct net *net, struct socket *sock,
2870 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002871{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002872 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002873 int err;
2874
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002875 uifr = compat_alloc_user_space(sizeof(*uifr));
2876 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2877 return -EFAULT;
2878
2879 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2880
Arnd Bergmann7a229382009-11-06 23:00:29 -08002881 if (!err) {
2882 switch (cmd) {
2883 case SIOCGIFFLAGS:
2884 case SIOCGIFMETRIC:
2885 case SIOCGIFMTU:
2886 case SIOCGIFMEM:
2887 case SIOCGIFHWADDR:
2888 case SIOCGIFINDEX:
2889 case SIOCGIFADDR:
2890 case SIOCGIFBRDADDR:
2891 case SIOCGIFDSTADDR:
2892 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002893 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002894 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002895 case SIOCGMIIPHY:
2896 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002897 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002898 err = -EFAULT;
2899 break;
2900 }
2901 }
2902 return err;
2903}
2904
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002905static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2906 struct compat_ifreq __user *uifr32)
2907{
2908 struct ifreq ifr;
2909 struct compat_ifmap __user *uifmap32;
2910 mm_segment_t old_fs;
2911 int err;
2912
2913 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2914 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002915 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2916 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2917 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2918 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2919 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2920 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002921 if (err)
2922 return -EFAULT;
2923
2924 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002925 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002926 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002927 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002928
2929 if (cmd == SIOCGIFMAP && !err) {
2930 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002931 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2932 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2933 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2934 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2935 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2936 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002937 if (err)
2938 err = -EFAULT;
2939 }
2940 return err;
2941}
2942
Arnd Bergmann7a229382009-11-06 23:00:29 -08002943struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002944 u32 rt_pad1;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002945 struct sockaddr rt_dst; /* target address */
2946 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2947 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002948 unsigned short rt_flags;
2949 short rt_pad2;
2950 u32 rt_pad3;
2951 unsigned char rt_tos;
2952 unsigned char rt_class;
2953 short rt_pad4;
2954 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a229382009-11-06 23:00:29 -08002955 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002956 u32 rt_mtu; /* per route MTU/Window */
2957 u32 rt_window; /* Window clamping */
Arnd Bergmann7a229382009-11-06 23:00:29 -08002958 unsigned short rt_irtt; /* Initial RTT */
2959};
2960
2961struct in6_rtmsg32 {
2962 struct in6_addr rtmsg_dst;
2963 struct in6_addr rtmsg_src;
2964 struct in6_addr rtmsg_gateway;
2965 u32 rtmsg_type;
2966 u16 rtmsg_dst_len;
2967 u16 rtmsg_src_len;
2968 u32 rtmsg_metric;
2969 u32 rtmsg_info;
2970 u32 rtmsg_flags;
2971 s32 rtmsg_ifindex;
2972};
2973
Arnd Bergmann6b960182009-11-06 23:10:54 -08002974static int routing_ioctl(struct net *net, struct socket *sock,
2975 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002976{
2977 int ret;
2978 void *r = NULL;
2979 struct in6_rtmsg r6;
2980 struct rtentry r4;
2981 char devname[16];
2982 u32 rtdev;
2983 mm_segment_t old_fs = get_fs();
2984
Arnd Bergmann6b960182009-11-06 23:10:54 -08002985 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2986 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002987 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08002988 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002989 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2990 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2991 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2992 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2993 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2994 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2995 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002996
2997 r = (void *) &r6;
2998 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08002999 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003000 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003001 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003002 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3003 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3004 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3005 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3006 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3007 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003008 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003009 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003010 r4.rt_dev = (char __user __force *)devname;
3011 devname[15] = 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003012 } else
3013 r4.rt_dev = NULL;
3014
3015 r = (void *) &r4;
3016 }
3017
3018 if (ret) {
3019 ret = -EFAULT;
3020 goto out;
3021 }
3022
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003023 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003024 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003025 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003026
3027out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003028 return ret;
3029}
3030
3031/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3032 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003033 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003034 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003035static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003036{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003037 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003038
Arnd Bergmann6b960182009-11-06 23:10:54 -08003039 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003040 return -EFAULT;
3041 if (tmp == BRCTL_GET_VERSION)
3042 return BRCTL_VERSION + 1;
3043 return -EINVAL;
3044}
3045
Arnd Bergmann6b960182009-11-06 23:10:54 -08003046static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3047 unsigned int cmd, unsigned long arg)
3048{
3049 void __user *argp = compat_ptr(arg);
3050 struct sock *sk = sock->sk;
3051 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003052
Arnd Bergmann6b960182009-11-06 23:10:54 -08003053 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003054 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003055
Arnd Bergmann6b960182009-11-06 23:10:54 -08003056 switch (cmd) {
3057 case SIOCSIFBR:
3058 case SIOCGIFBR:
3059 return old_bridge_ioctl(argp);
3060 case SIOCGIFNAME:
3061 return dev_ifname32(net, argp);
3062 case SIOCGIFCONF:
3063 return dev_ifconf(net, argp);
3064 case SIOCETHTOOL:
3065 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003066 case SIOCWANDEV:
3067 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003068 case SIOCGIFMAP:
3069 case SIOCSIFMAP:
3070 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003071 case SIOCBONDENSLAVE:
3072 case SIOCBONDRELEASE:
3073 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003074 case SIOCBONDCHANGEACTIVE:
3075 return bond_ioctl(net, cmd, argp);
3076 case SIOCADDRT:
3077 case SIOCDELRT:
3078 return routing_ioctl(net, sock, cmd, argp);
3079 case SIOCGSTAMP:
3080 return do_siocgstamp(net, sock, cmd, argp);
3081 case SIOCGSTAMPNS:
3082 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003083 case SIOCBONDSLAVEINFOQUERY:
3084 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003085 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003086 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003087 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003088
Arnd Bergmann6b960182009-11-06 23:10:54 -08003089 case FIOSETOWN:
3090 case SIOCSPGRP:
3091 case FIOGETOWN:
3092 case SIOCGPGRP:
3093 case SIOCBRADDBR:
3094 case SIOCBRDELBR:
3095 case SIOCGIFVLAN:
3096 case SIOCSIFVLAN:
3097 case SIOCADDDLCI:
3098 case SIOCDELDLCI:
3099 return sock_ioctl(file, cmd, arg);
3100
3101 case SIOCGIFFLAGS:
3102 case SIOCSIFFLAGS:
3103 case SIOCGIFMETRIC:
3104 case SIOCSIFMETRIC:
3105 case SIOCGIFMTU:
3106 case SIOCSIFMTU:
3107 case SIOCGIFMEM:
3108 case SIOCSIFMEM:
3109 case SIOCGIFHWADDR:
3110 case SIOCSIFHWADDR:
3111 case SIOCADDMULTI:
3112 case SIOCDELMULTI:
3113 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003114 case SIOCGIFADDR:
3115 case SIOCSIFADDR:
3116 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003117 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003118 case SIOCGIFBRDADDR:
3119 case SIOCSIFBRDADDR:
3120 case SIOCGIFDSTADDR:
3121 case SIOCSIFDSTADDR:
3122 case SIOCGIFNETMASK:
3123 case SIOCSIFNETMASK:
3124 case SIOCSIFPFLAGS:
3125 case SIOCGIFPFLAGS:
3126 case SIOCGIFTXQLEN:
3127 case SIOCSIFTXQLEN:
3128 case SIOCBRADDIF:
3129 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003130 case SIOCSIFNAME:
3131 case SIOCGMIIPHY:
3132 case SIOCGMIIREG:
3133 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003134 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003135
Arnd Bergmann6b960182009-11-06 23:10:54 -08003136 case SIOCSARP:
3137 case SIOCGARP:
3138 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003139 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003140 return sock_do_ioctl(net, sock, cmd, arg);
3141 }
3142
Arnd Bergmann6b960182009-11-06 23:10:54 -08003143 return -ENOIOCTLCMD;
3144}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003145
Eric Dumazet95c96172012-04-15 05:58:06 +00003146static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003147 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003148{
3149 struct socket *sock = file->private_data;
3150 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003151 struct sock *sk;
3152 struct net *net;
3153
3154 sk = sock->sk;
3155 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003156
3157 if (sock->ops->compat_ioctl)
3158 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3159
David S. Miller87de87d2008-06-03 09:14:03 -07003160 if (ret == -ENOIOCTLCMD &&
3161 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3162 ret = compat_wext_handle_ioctl(net, cmd, arg);
3163
Arnd Bergmann6b960182009-11-06 23:10:54 -08003164 if (ret == -ENOIOCTLCMD)
3165 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3166
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003167 return ret;
3168}
3169#endif
3170
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003171int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3172{
3173 return sock->ops->bind(sock, addr, addrlen);
3174}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003175EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003176
3177int kernel_listen(struct socket *sock, int backlog)
3178{
3179 return sock->ops->listen(sock, backlog);
3180}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003181EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003182
3183int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3184{
3185 struct sock *sk = sock->sk;
3186 int err;
3187
3188 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3189 newsock);
3190 if (err < 0)
3191 goto done;
3192
3193 err = sock->ops->accept(sock, *newsock, flags);
3194 if (err < 0) {
3195 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003196 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003197 goto done;
3198 }
3199
3200 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003201 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003202
3203done:
3204 return err;
3205}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003206EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003207
3208int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003209 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003210{
3211 return sock->ops->connect(sock, addr, addrlen, flags);
3212}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003213EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003214
3215int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3216 int *addrlen)
3217{
3218 return sock->ops->getname(sock, addr, addrlen, 0);
3219}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003220EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003221
3222int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3223 int *addrlen)
3224{
3225 return sock->ops->getname(sock, addr, addrlen, 1);
3226}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003227EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003228
3229int kernel_getsockopt(struct socket *sock, int level, int optname,
3230 char *optval, int *optlen)
3231{
3232 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003233 char __user *uoptval;
3234 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003235 int err;
3236
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003237 uoptval = (char __user __force *) optval;
3238 uoptlen = (int __user __force *) optlen;
3239
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003240 set_fs(KERNEL_DS);
3241 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003242 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003243 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003244 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3245 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003246 set_fs(oldfs);
3247 return err;
3248}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003249EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003250
3251int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003252 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003253{
3254 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003255 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003256 int err;
3257
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003258 uoptval = (char __user __force *) optval;
3259
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003260 set_fs(KERNEL_DS);
3261 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003262 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003263 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003264 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003265 optlen);
3266 set_fs(oldfs);
3267 return err;
3268}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003269EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003270
3271int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3272 size_t size, int flags)
3273{
3274 if (sock->ops->sendpage)
3275 return sock->ops->sendpage(sock, page, offset, size, flags);
3276
3277 return sock_no_sendpage(sock, page, offset, size, flags);
3278}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003279EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003280
3281int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3282{
3283 mm_segment_t oldfs = get_fs();
3284 int err;
3285
3286 set_fs(KERNEL_DS);
3287 err = sock->ops->ioctl(sock, cmd, arg);
3288 set_fs(oldfs);
3289
3290 return err;
3291}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003292EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003293
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003294int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3295{
3296 return sock->ops->shutdown(sock, how);
3297}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003298EXPORT_SYMBOL(kernel_sock_shutdown);