blob: c2564eb25c6b8faf5c99504ef1d3c90f9bb57c73 [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
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080093#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#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
yuan linyu1e911632017-01-07 17:18:31 +0800290static void init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291{
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);
yuan linyu1e911632017-01-07 17:18:31 +0800299 BUG_ON(sock_inode_cachep == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300}
301
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700302static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306};
307
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700308/*
309 * sockfs_dname() is called from d_path().
310 */
311static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
312{
313 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000314 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700315}
316
Al Viro3ba13d12009-02-20 06:02:22 +0000317static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700318 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319};
320
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200321static int sockfs_xattr_get(const struct xattr_handler *handler,
322 struct dentry *dentry, struct inode *inode,
323 const char *suffix, void *value, size_t size)
324{
325 if (value) {
326 if (dentry->d_name.len + 1 > size)
327 return -ERANGE;
328 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
329 }
330 return dentry->d_name.len + 1;
331}
332
333#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
336
337static const struct xattr_handler sockfs_xattr_handler = {
338 .name = XATTR_NAME_SOCKPROTONAME,
339 .get = sockfs_xattr_get,
340};
341
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100342static int sockfs_security_xattr_set(const struct xattr_handler *handler,
343 struct dentry *dentry, struct inode *inode,
344 const char *suffix, const void *value,
345 size_t size, int flags)
346{
347 /* Handled by LSM. */
348 return -EAGAIN;
349}
350
351static const struct xattr_handler sockfs_security_xattr_handler = {
352 .prefix = XATTR_SECURITY_PREFIX,
353 .set = sockfs_security_xattr_set,
354};
355
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200356static const struct xattr_handler *sockfs_xattr_handlers[] = {
357 &sockfs_xattr_handler,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100358 &sockfs_security_xattr_handler,
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200359 NULL
360};
361
Al Viroc74a1cb2011-01-12 16:59:34 -0500362static struct dentry *sockfs_mount(struct file_system_type *fs_type,
363 int flags, const char *dev_name, void *data)
364{
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200365 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
366 sockfs_xattr_handlers,
367 &sockfs_dentry_operations, SOCKFS_MAGIC);
Al Viroc74a1cb2011-01-12 16:59:34 -0500368}
369
370static struct vfsmount *sock_mnt __read_mostly;
371
372static struct file_system_type sock_fs_type = {
373 .name = "sockfs",
374 .mount = sockfs_mount,
375 .kill_sb = kill_anon_super,
376};
377
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378/*
379 * Obtains the first available file descriptor and sets it up for use.
380 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800381 * These functions create file structures and maps them to fd space
382 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 * and file struct implicitly stored in sock->file.
384 * Note that another thread may close file descriptor before we return
385 * from this function. We use the fact that now we do not refer
386 * to socket after mapping. If one day we will need it, this
387 * function will increment ref. count on file by 1.
388 *
389 * In any case returned fd MAY BE not valid!
390 * This race condition is unavoidable
391 * with shared fd spaces, we cannot solve it inside kernel,
392 * but we take care of internal coherence yet.
393 */
394
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700395struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396{
Al Viro7cbe66b2009-08-05 19:59:08 +0400397 struct qstr name = { .name = "" };
Al Viro2c48b9c2009-08-09 00:52:35 +0400398 struct path path;
Al Viro7cbe66b2009-08-05 19:59:08 +0400399 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800400
Masatake YAMATO600e1772012-08-29 10:44:29 +0000401 if (dname) {
402 name.name = dname;
403 name.len = strlen(name.name);
404 } else if (sock->sk) {
405 name.name = sock->sk->sk_prot_creator->name;
406 name.len = strlen(name.name);
407 }
Nick Piggin4b936882011-01-07 17:50:07 +1100408 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
Al Viro28407632012-08-17 23:54:15 -0400409 if (unlikely(!path.dentry))
410 return ERR_PTR(-ENOMEM);
Al Viro2c48b9c2009-08-09 00:52:35 +0400411 path.mnt = mntget(sock_mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412
Al Viro2c48b9c2009-08-09 00:52:35 +0400413 d_instantiate(path.dentry, SOCK_INODE(sock));
Al Virocc3808f2009-08-06 09:43:59 +0400414
Al Viro2c48b9c2009-08-09 00:52:35 +0400415 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
Al Virocc3808f2009-08-06 09:43:59 +0400416 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530417 if (IS_ERR(file)) {
Al Virocc3808f2009-08-06 09:43:59 +0400418 /* drop dentry, keep inode */
David Howellsc5ef6032015-03-17 22:26:16 +0000419 ihold(d_inode(path.dentry));
Al Viro2c48b9c2009-08-09 00:52:35 +0400420 path_put(&path);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700421 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400422 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800423
424 sock->file = file;
Ulrich Drepper77d27202008-07-23 21:29:35 -0700425 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800426 file->private_data = sock;
Al Viro28407632012-08-17 23:54:15 -0400427 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800428}
Al Viro56b31d12012-08-18 00:25:51 -0400429EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800430
Al Viro56b31d12012-08-18 00:25:51 -0400431static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800432{
433 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400434 int fd = get_unused_fd_flags(flags);
435 if (unlikely(fd < 0))
436 return fd;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800437
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700438 newfile = sock_alloc_file(sock, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -0400439 if (likely(!IS_ERR(newfile))) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800440 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400441 return fd;
442 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400443
Al Viro28407632012-08-17 23:54:15 -0400444 put_unused_fd(fd);
445 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446}
447
John Fastabend406a3c62012-07-20 10:39:25 +0000448struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800449{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800450 if (file->f_op == &socket_file_ops)
451 return file->private_data; /* set in sock_map_fd */
452
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800453 *err = -ENOTSOCK;
454 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800455}
John Fastabend406a3c62012-07-20 10:39:25 +0000456EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800457
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700459 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 * @fd: file handle
461 * @err: pointer to an error code return
462 *
463 * The file handle passed in is locked and the socket it is bound
464 * too is returned. If an error occurs the err pointer is overwritten
465 * with a negative errno code and NULL is returned. The function checks
466 * for both invalid handles and passing a handle which is not a socket.
467 *
468 * On a success the socket object pointer is returned.
469 */
470
471struct socket *sockfd_lookup(int fd, int *err)
472{
473 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 struct socket *sock;
475
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700476 file = fget(fd);
477 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 *err = -EBADF;
479 return NULL;
480 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700481
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800482 sock = sock_from_file(file, err);
483 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 return sock;
486}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700487EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800489static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
490{
Al Viro00e188e2014-03-03 23:48:18 -0500491 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800492 struct socket *sock;
493
Hua Zhong36725582006-04-19 15:25:02 -0700494 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500495 if (f.file) {
496 sock = sock_from_file(f.file, err);
497 if (likely(sock)) {
498 *fput_needed = f.flags;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800499 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500500 }
501 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800502 }
503 return NULL;
504}
505
Masatake YAMATO600e1772012-08-29 10:44:29 +0000506static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
507 size_t size)
508{
509 ssize_t len;
510 ssize_t used = 0;
511
David Howellsc5ef6032015-03-17 22:26:16 +0000512 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000513 if (len < 0)
514 return len;
515 used += len;
516 if (buffer) {
517 if (size < used)
518 return -ERANGE;
519 buffer += len;
520 }
521
522 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
523 used += len;
524 if (buffer) {
525 if (size < used)
526 return -ERANGE;
527 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
528 buffer += len;
529 }
530
531 return used;
532}
533
Tobias Klauserdc647ec2017-01-10 09:30:51 +0100534static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900535{
536 int err = simple_setattr(dentry, iattr);
537
Eric Biggerse1a3a602016-12-30 17:42:32 -0600538 if (!err && (iattr->ia_valid & ATTR_UID)) {
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900539 struct socket *sock = SOCKET_I(d_inode(dentry));
540
541 sock->sk->sk_uid = iattr->ia_uid;
542 }
543
544 return err;
545}
546
Masatake YAMATO600e1772012-08-29 10:44:29 +0000547static const struct inode_operations sockfs_inode_ops = {
Masatake YAMATO600e1772012-08-29 10:44:29 +0000548 .listxattr = sockfs_listxattr,
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900549 .setattr = sockfs_setattr,
Masatake YAMATO600e1772012-08-29 10:44:29 +0000550};
551
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552/**
553 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700554 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 * Allocate a new inode and socket object. The two are bound together
556 * and initialised. The socket is then returned. If we are out of inodes
557 * NULL is returned.
558 */
559
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800560struct socket *sock_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700562 struct inode *inode;
563 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200565 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 if (!inode)
567 return NULL;
568
569 sock = SOCKET_I(inode);
570
Eric Dumazet29a020d2009-09-15 02:39:20 -0700571 kmemcheck_annotate_bitfield(sock, type);
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400572 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700573 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100574 inode->i_uid = current_fsuid();
575 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000576 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577
Alex Shi19e8d692012-05-14 14:15:31 -0700578 this_cpu_add(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 return sock;
580}
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800581EXPORT_SYMBOL(sock_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583/**
584 * sock_release - close a socket
585 * @sock: socket to close
586 *
587 * The socket is released from the protocol stack if it has a release
588 * callback, and the inode is then released if the socket is bound to
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700589 * an inode not a file.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700591
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592void sock_release(struct socket *sock)
593{
594 if (sock->ops) {
595 struct module *owner = sock->ops->owner;
596
597 sock->ops->release(sock);
598 sock->ops = NULL;
599 module_put(owner);
600 }
601
Eric Dumazeteaefd112011-02-18 03:26:36 +0000602 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800603 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
Alex Shi19e8d692012-05-14 14:15:31 -0700605 this_cpu_sub(sockets_in_use, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 if (!sock->file) {
607 iput(SOCK_INODE(sock));
608 return;
609 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700610 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700612EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400614void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000615{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200616 u8 flags = *tx_flags;
617
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400618 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200619 flags |= SKBTX_HW_TSTAMP;
620
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400621 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200622 flags |= SKBTX_SW_TSTAMP;
623
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400624 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200625 flags |= SKBTX_SCHED_TSTAMP;
626
Eric Dumazet140c55d2014-08-06 11:49:29 +0200627 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000628}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400629EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000630
Al Virod8725c82014-12-11 00:02:50 -0500631static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632{
Al Viro01e97e62014-12-15 21:39:31 -0500633 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500634 BUG_ON(ret == -EIOCBQUEUED);
635 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800637
Al Virod8725c82014-12-11 00:02:50 -0500638int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800639{
Al Virod8725c82014-12-11 00:02:50 -0500640 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500641 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800642
Al Virod8725c82014-12-11 00:02:50 -0500643 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800644}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700645EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
647int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648 struct kvec *vec, size_t num, size_t size)
649{
Al Viro6aa24812015-03-21 19:56:16 -0400650 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500651 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700653EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400655static bool skb_is_err_queue(const struct sk_buff *skb)
656{
657 /* pkt_type of skbs enqueued on the error queue are set to
658 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
659 * in recvmsg, since skbs received on a local socket will never
660 * have a pkt_type of PACKET_OUTGOING.
661 */
662 return skb->pkt_type == PACKET_OUTGOING;
663}
664
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700665/*
666 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
667 */
668void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
669 struct sk_buff *skb)
670{
Patrick Ohly20d49472009-02-12 05:03:38 +0000671 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400672 struct scm_timestamping tss;
Patrick Ohly20d49472009-02-12 05:03:38 +0000673 int empty = 1;
674 struct skb_shared_hwtstamps *shhwtstamps =
675 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700676
Patrick Ohly20d49472009-02-12 05:03:38 +0000677 /* Race occurred between timestamp enabling and packet
678 receiving. Fill in the current time for now. */
Thomas Gleixner2456e852016-12-25 11:38:40 +0100679 if (need_software_tstamp && skb->tstamp == 0)
Patrick Ohly20d49472009-02-12 05:03:38 +0000680 __net_timestamp(skb);
681
682 if (need_software_tstamp) {
683 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
684 struct timeval tv;
685 skb_get_timestamp(skb, &tv);
686 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
687 sizeof(tv), &tv);
688 } else {
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400689 struct timespec ts;
690 skb_get_timestampns(skb, &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000691 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400692 sizeof(ts), &ts);
Patrick Ohly20d49472009-02-12 05:03:38 +0000693 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700694 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000695
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400696 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400697 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400698 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000699 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400700 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400701 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400702 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400703 empty = 0;
Francis Yan1c885802016-11-27 23:07:18 -0800704 if (!empty) {
Patrick Ohly20d49472009-02-12 05:03:38 +0000705 put_cmsg(msg, SOL_SOCKET,
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400706 SCM_TIMESTAMPING, sizeof(tss), &tss);
Francis Yan1c885802016-11-27 23:07:18 -0800707
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400708 if (skb_is_err_queue(skb) && skb->len &&
Soheil Hassas Yeganeh4ef1b282017-03-18 17:03:00 -0400709 SKB_EXT_ERR(skb)->opt_stats)
Francis Yan1c885802016-11-27 23:07:18 -0800710 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
711 skb->len, skb->data);
712 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700713}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300714EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
715
Johannes Berg6e3e9392011-11-09 10:15:42 +0100716void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
717 struct sk_buff *skb)
718{
719 int ack;
720
721 if (!sock_flag(sk, SOCK_WIFI_STATUS))
722 return;
723 if (!skb->wifi_acked_valid)
724 return;
725
726 ack = skb->wifi_acked;
727
728 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
729}
730EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
731
stephen hemminger11165f12010-10-18 14:27:29 +0000732static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
733 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700734{
Eyal Birger744d5a32015-03-01 14:58:31 +0200735 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700736 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200737 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700738}
739
Eric Dumazet767dd032010-04-28 19:14:43 +0000740void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700741 struct sk_buff *skb)
742{
743 sock_recv_timestamp(msg, sk, skb);
744 sock_recv_drops(msg, sk, skb);
745}
Eric Dumazet767dd032010-04-28 19:14:43 +0000746EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700747
Ying Xue1b784142015-03-02 15:37:48 +0800748static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -0400749 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750{
Al Viro2da62902015-03-14 21:13:46 -0400751 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752}
753
Al Viro2da62902015-03-14 21:13:46 -0400754int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700755{
Al Viro2da62902015-03-14 21:13:46 -0400756 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700757
Al Viro2da62902015-03-14 21:13:46 -0400758 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700760EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Martin Lucinac1249c02010-12-10 00:04:05 +0000762/**
763 * kernel_recvmsg - Receive a message from a socket (kernel space)
764 * @sock: The socket to receive the message from
765 * @msg: Received message
766 * @vec: Input s/g array for message data
767 * @num: Size of input s/g array
768 * @size: Number of bytes to read
769 * @flags: Message flags (MSG_DONTWAIT, etc...)
770 *
771 * On return the msg structure contains the scatter/gather array passed in the
772 * vec argument. The array is modified so that it consists of the unfilled
773 * portion of the original array.
774 *
775 * The returned value is the total number of bytes received, or an error.
776 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700777int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
778 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779{
780 mm_segment_t oldfs = get_fs();
781 int result;
782
Al Viro6aa24812015-03-21 19:56:16 -0400783 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 set_fs(KERNEL_DS);
Al Viro2da62902015-03-14 21:13:46 -0400785 result = sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 set_fs(oldfs);
787 return result;
788}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700789EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300791static ssize_t sock_sendpage(struct file *file, struct page *page,
792 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793{
794 struct socket *sock;
795 int flags;
796
Eric Dumazetb69aee02005-09-06 14:42:45 -0700797 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798
Eric Dumazet35f9c092012-04-05 03:05:35 +0000799 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
800 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
801 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
Linus Torvaldse6949582009-08-13 08:28:36 -0700803 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804}
805
Jens Axboe9c55e012007-11-06 23:30:13 -0800806static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700807 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800808 unsigned int flags)
809{
810 struct socket *sock = file->private_data;
811
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800812 if (unlikely(!sock->ops->splice_read))
813 return -EINVAL;
814
Jens Axboe9c55e012007-11-06 23:30:13 -0800815 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
816}
817
Al Viro8ae5e032014-11-28 19:40:50 -0500818static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800819{
Al Viro6d652332015-01-30 16:12:56 -0500820 struct file *file = iocb->ki_filp;
821 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700822 struct msghdr msg = {.msg_iter = *to,
823 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500824 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800825
Al Viro8ae5e032014-11-28 19:40:50 -0500826 if (file->f_flags & O_NONBLOCK)
827 msg.msg_flags = MSG_DONTWAIT;
828
829 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800830 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700831
Christoph Hellwig66ee59a2015-02-11 19:56:46 +0100832 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800833 return 0;
834
Al Viro2da62902015-03-14 21:13:46 -0400835 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e032014-11-28 19:40:50 -0500836 *to = msg.msg_iter;
837 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800838}
839
Al Viro8ae5e032014-11-28 19:40:50 -0500840static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841{
Al Viro6d652332015-01-30 16:12:56 -0500842 struct file *file = iocb->ki_filp;
843 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700844 struct msghdr msg = {.msg_iter = *from,
845 .msg_iocb = iocb};
Al Viro8ae5e032014-11-28 19:40:50 -0500846 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800847
Al Viro8ae5e032014-11-28 19:40:50 -0500848 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800849 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700850
Al Viro8ae5e032014-11-28 19:40:50 -0500851 if (file->f_flags & O_NONBLOCK)
852 msg.msg_flags = MSG_DONTWAIT;
853
Al Viro6d652332015-01-30 16:12:56 -0500854 if (sock->type == SOCK_SEQPACKET)
855 msg.msg_flags |= MSG_EOR;
856
Al Virod8725c82014-12-11 00:02:50 -0500857 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e032014-11-28 19:40:50 -0500858 *from = msg.msg_iter;
859 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860}
861
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862/*
863 * Atomic setting of ioctl hooks to avoid race
864 * with module unload.
865 */
866
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800867static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700868static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
Eric W. Biederman881d9662007-09-17 11:56:21 -0700870void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800872 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800874 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875}
876EXPORT_SYMBOL(brioctl_set);
877
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800878static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -0700879static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880
Eric W. Biederman881d9662007-09-17 11:56:21 -0700881void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800883 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800885 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886}
887EXPORT_SYMBOL(vlan_ioctl_set);
888
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800889static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700890static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700892void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800894 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800896 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897}
898EXPORT_SYMBOL(dlci_ioctl_set);
899
Arnd Bergmann6b960182009-11-06 23:10:54 -0800900static long sock_do_ioctl(struct net *net, struct socket *sock,
901 unsigned int cmd, unsigned long arg)
902{
903 int err;
904 void __user *argp = (void __user *)arg;
905
906 err = sock->ops->ioctl(sock, cmd, arg);
907
908 /*
909 * If this ioctl is unknown try to hand it down
910 * to the NIC driver.
911 */
912 if (err == -ENOIOCTLCMD)
913 err = dev_ioctl(net, cmd, argp);
914
915 return err;
916}
917
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918/*
919 * With an ioctl, arg may well be a user mode pointer, but we don't know
920 * what to do with it - that's up to the protocol still.
921 */
922
Andrey Vaginc62cce22016-10-24 18:29:13 -0700923static struct ns_common *get_net_ns(struct ns_common *ns)
924{
925 return &get_net(container_of(ns, struct net, ns))->ns;
926}
927
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
929{
930 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700931 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 void __user *argp = (void __user *)arg;
933 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700934 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935
Eric Dumazetb69aee02005-09-06 14:42:45 -0700936 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -0700937 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900938 net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700940 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200942#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700944 err = dev_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 } else
Johannes Berg3d23e342009-09-29 23:27:28 +0200946#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700947 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 case FIOSETOWN:
949 case SIOCSPGRP:
950 err = -EFAULT;
951 if (get_user(pid, (int __user *)argp))
952 break;
Jeff Laytone0b93ed2014-08-22 11:27:32 -0400953 f_setown(sock->file, pid, 1);
954 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955 break;
956 case FIOGETOWN:
957 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -0700958 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700959 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700960 break;
961 case SIOCGIFBR:
962 case SIOCSIFBR:
963 case SIOCBRADDBR:
964 case SIOCBRDELBR:
965 err = -ENOPKG;
966 if (!br_ioctl_hook)
967 request_module("bridge");
968
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800969 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700970 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700971 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800972 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 break;
974 case SIOCGIFVLAN:
975 case SIOCSIFVLAN:
976 err = -ENOPKG;
977 if (!vlan_ioctl_hook)
978 request_module("8021q");
979
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800980 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -0700982 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -0800983 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985 case SIOCADDDLCI:
986 case SIOCDELDLCI:
987 err = -ENOPKG;
988 if (!dlci_ioctl_hook)
989 request_module("dlci");
990
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700991 mutex_lock(&dlci_ioctl_mutex);
992 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -0700994 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 break;
Andrey Vaginc62cce22016-10-24 18:29:13 -0700996 case SIOCGSKNS:
997 err = -EPERM;
998 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
999 break;
1000
1001 err = open_related_ns(&net->ns, get_net_ns);
1002 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 default:
Arnd Bergmann6b960182009-11-06 23:10:54 -08001004 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001006 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 return err;
1008}
1009
1010int sock_create_lite(int family, int type, int protocol, struct socket **res)
1011{
1012 int err;
1013 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 err = security_socket_create(family, type, protocol, 1);
1016 if (err)
1017 goto out;
1018
1019 sock = sock_alloc();
1020 if (!sock) {
1021 err = -ENOMEM;
1022 goto out;
1023 }
1024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001026 err = security_socket_post_create(sock, family, type, protocol, 1);
1027 if (err)
1028 goto out_release;
1029
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030out:
1031 *res = sock;
1032 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001033out_release:
1034 sock_release(sock);
1035 sock = NULL;
1036 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001038EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
1040/* No kernel lock held - perfect */
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001041static unsigned int sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042{
Eliezer Tamircbf55002013-07-08 16:20:34 +03001043 unsigned int busy_flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 struct socket *sock;
1045
1046 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001047 * We can't return errors to poll, so it's either yes or no.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 */
Eric Dumazetb69aee02005-09-06 14:42:45 -07001049 sock = file->private_data;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001050
Eliezer Tamircbf55002013-07-08 16:20:34 +03001051 if (sk_can_busy_loop(sock->sk)) {
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001052 /* this socket can poll_ll so tell the system call */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001053 busy_flag = POLL_BUSY_LOOP;
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001054
1055 /* once, only if requested by syscall */
Eliezer Tamircbf55002013-07-08 16:20:34 +03001056 if (wait && (wait->_key & POLL_BUSY_LOOP))
1057 sk_busy_loop(sock->sk, 1);
Eliezer Tamir2d48d672013-06-24 10:28:03 +03001058 }
1059
Eliezer Tamircbf55002013-07-08 16:20:34 +03001060 return busy_flag | sock->ops->poll(file, sock, wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061}
1062
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001063static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001065 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066
1067 return sock->ops->mmap(file, sock, vma);
1068}
1069
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001070static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 sock_release(SOCKET_I(inode));
1073 return 0;
1074}
1075
1076/*
1077 * Update the socket async list
1078 *
1079 * Fasync_list locking strategy.
1080 *
1081 * 1. fasync_list is modified only under process context socket lock
1082 * i.e. under semaphore.
1083 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001084 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 */
1086
1087static int sock_fasync(int fd, struct file *filp, int on)
1088{
Eric Dumazet989a2972010-04-14 09:55:35 +00001089 struct socket *sock = filp->private_data;
1090 struct sock *sk = sock->sk;
Eric Dumazeteaefd112011-02-18 03:26:36 +00001091 struct socket_wq *wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092
Eric Dumazet989a2972010-04-14 09:55:35 +00001093 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095
1096 lock_sock(sk);
Hannes Frederic Sowa1e1d04e2016-04-05 17:10:15 +02001097 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
Eric Dumazeteaefd112011-02-18 03:26:36 +00001098 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099
Eric Dumazeteaefd112011-02-18 03:26:36 +00001100 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001101 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001102 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001103 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104
Eric Dumazet989a2972010-04-14 09:55:35 +00001105 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 return 0;
1107}
1108
Eric Dumazetceb5d582015-11-29 20:03:11 -08001109/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
Eric Dumazetceb5d582015-11-29 20:03:11 -08001111int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001113 if (!wq || !wq->fasync_list)
1114 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001115
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001116 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001117 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001118 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 break;
1120 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001121 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001122 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 break;
1124 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001125 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001126call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001127 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001129 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001130 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 return 0;
1134}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001135EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136
Pavel Emelyanov721db932010-09-29 16:06:32 +04001137int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001138 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139{
1140 int err;
1141 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001142 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143
1144 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001145 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 */
1147 if (family < 0 || family >= NPROTO)
1148 return -EAFNOSUPPORT;
1149 if (type < 0 || type >= SOCK_MAX)
1150 return -EINVAL;
1151
1152 /* Compatibility.
1153
1154 This uglymoron is moved from INET layer to here to avoid
1155 deadlock in module load.
1156 */
1157 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001158 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1159 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 family = PF_PACKET;
1161 }
1162
1163 err = security_socket_create(family, type, protocol, kern);
1164 if (err)
1165 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001166
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001167 /*
1168 * Allocate the socket and allow the family to set things up. if
1169 * the protocol is 0, the family is instructed to select an appropriate
1170 * default.
1171 */
1172 sock = sock_alloc();
1173 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001174 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001175 return -ENFILE; /* Not exactly a match, but its the
1176 closest posix thing */
1177 }
1178
1179 sock->type = type;
1180
Johannes Berg95a5afc2008-10-16 15:24:51 -07001181#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001182 /* Attempt to load a protocol module if the find failed.
1183 *
1184 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 * requested real, full-featured networking support upon configuration.
1186 * Otherwise module support will break!
1187 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001188 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001189 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001190#endif
1191
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001192 rcu_read_lock();
1193 pf = rcu_dereference(net_families[family]);
1194 err = -EAFNOSUPPORT;
1195 if (!pf)
1196 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197
1198 /*
1199 * We will call the ->create function, that possibly is in a loadable
1200 * module, so we have to bump that loadable module refcnt first.
1201 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001202 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203 goto out_release;
1204
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001205 /* Now protected by module ref count */
1206 rcu_read_unlock();
1207
Eric Paris3f378b62009-11-05 22:18:14 -08001208 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001209 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 /*
1213 * Now to bump the refcnt of the [loadable] module that owns this
1214 * socket at sock_release time we decrement its refcnt.
1215 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001216 if (!try_module_get(sock->ops->owner))
1217 goto out_module_busy;
1218
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 /*
1220 * Now that we're done with the ->create function, the [loadable]
1221 * module can have its refcnt decremented
1222 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001223 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001224 err = security_socket_post_create(sock, family, type, protocol, kern);
1225 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001226 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001227 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001229 return 0;
1230
1231out_module_busy:
1232 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001234 sock->ops = NULL;
1235 module_put(pf->owner);
1236out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001238 return err;
1239
1240out_release:
1241 rcu_read_unlock();
1242 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001244EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245
1246int sock_create(int family, int type, int protocol, struct socket **res)
1247{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001248 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001250EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001252int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001254 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001256EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001258SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259{
1260 int retval;
1261 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001262 int flags;
1263
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001264 /* Check the SOCK_* constants for consistency. */
1265 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1266 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1267 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1268 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1269
Ulrich Dreppera677a032008-07-23 21:29:17 -07001270 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001271 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001272 return -EINVAL;
1273 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001275 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1276 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1277
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 retval = sock_create(family, type, protocol, &sock);
1279 if (retval < 0)
1280 goto out;
1281
Ulrich Drepper77d27202008-07-23 21:29:35 -07001282 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 if (retval < 0)
1284 goto out_release;
1285
1286out:
1287 /* It may be already another descriptor 8) Not kernel problem. */
1288 return retval;
1289
1290out_release:
1291 sock_release(sock);
1292 return retval;
1293}
1294
1295/*
1296 * Create a pair of connected sockets.
1297 */
1298
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001299SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1300 int __user *, usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301{
1302 struct socket *sock1, *sock2;
1303 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001304 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001305 int flags;
1306
1307 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001308 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001309 return -EINVAL;
1310 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001312 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1313 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1314
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 /*
1316 * Obtain the first socket and check if the underlying protocol
1317 * supports the socketpair call.
1318 */
1319
1320 err = sock_create(family, type, protocol, &sock1);
1321 if (err < 0)
1322 goto out;
1323
1324 err = sock_create(family, type, protocol, &sock2);
1325 if (err < 0)
1326 goto out_release_1;
1327
1328 err = sock1->ops->socketpair(sock1, sock2);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001329 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 goto out_release_both;
1331
Al Viro28407632012-08-17 23:54:15 -04001332 fd1 = get_unused_fd_flags(flags);
David S. Millerbf3c23d2007-10-29 21:54:02 -07001333 if (unlikely(fd1 < 0)) {
1334 err = fd1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 goto out_release_both;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001336 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001337
Al Viro28407632012-08-17 23:54:15 -04001338 fd2 = get_unused_fd_flags(flags);
Al Viro198de4d2009-08-05 19:29:23 +04001339 if (unlikely(fd2 < 0)) {
1340 err = fd2;
Yann Droneaudd73aa282013-12-09 22:42:20 +01001341 goto out_put_unused_1;
Al Viro28407632012-08-17 23:54:15 -04001342 }
1343
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001344 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301345 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001346 err = PTR_ERR(newfile1);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001347 goto out_put_unused_both;
Al Viro28407632012-08-17 23:54:15 -04001348 }
1349
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001350 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001351 if (IS_ERR(newfile2)) {
1352 err = PTR_ERR(newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001353 goto out_fput_1;
Al Virodb349502007-02-07 01:48:00 -05001354 }
1355
Yann Droneaudd73aa282013-12-09 22:42:20 +01001356 err = put_user(fd1, &usockvec[0]);
1357 if (err)
1358 goto out_fput_both;
1359
1360 err = put_user(fd2, &usockvec[1]);
1361 if (err)
1362 goto out_fput_both;
1363
Al Viro157cf642008-12-14 04:57:47 -05001364 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001365
Al Virodb349502007-02-07 01:48:00 -05001366 fd_install(fd1, newfile1);
1367 fd_install(fd2, newfile2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 /* fd1 and fd2 may be already another descriptors.
1369 * Not kernel problem.
1370 */
1371
Yann Droneaudd73aa282013-12-09 22:42:20 +01001372 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373
Yann Droneaudd73aa282013-12-09 22:42:20 +01001374out_fput_both:
1375 fput(newfile2);
1376 fput(newfile1);
1377 put_unused_fd(fd2);
1378 put_unused_fd(fd1);
1379 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
Yann Droneaudd73aa282013-12-09 22:42:20 +01001381out_fput_1:
1382 fput(newfile1);
1383 put_unused_fd(fd2);
1384 put_unused_fd(fd1);
1385 sock_release(sock2);
1386 goto out;
1387
1388out_put_unused_both:
1389 put_unused_fd(fd2);
1390out_put_unused_1:
1391 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392out_release_both:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001393 sock_release(sock2);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394out_release_1:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001395 sock_release(sock1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396out:
1397 return err;
1398}
1399
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400/*
1401 * Bind a name to a socket. Nothing much to do here since it's
1402 * the protocol's responsibility to handle the local address.
1403 *
1404 * We move the socket address to kernel space before we call
1405 * the protocol layer (having also checked the address is ok).
1406 */
1407
Heiko Carstens20f37032009-01-14 14:14:23 +01001408SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409{
1410 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001411 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001412 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001414 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001415 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001416 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001417 if (err >= 0) {
1418 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001419 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001420 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001421 if (!err)
1422 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001423 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001424 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001426 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001427 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 return err;
1429}
1430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431/*
1432 * Perform a listen. Basically, we allow the protocol to do anything
1433 * necessary for a listen, and if that works, we mark the socket as
1434 * ready for listening.
1435 */
1436
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001437SYSCALL_DEFINE2(listen, int, fd, int, backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438{
1439 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001440 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001441 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001442
1443 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1444 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001445 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001446 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001447 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448
1449 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001450 if (!err)
1451 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001453 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 }
1455 return err;
1456}
1457
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458/*
1459 * For accept, we attempt to create a new socket, set up the link
1460 * with the client, wake up the client, then return the new
1461 * connected fd. We collect the address of the connector in kernel
1462 * space and move it to user at the very end. This is unclean because
1463 * we open the socket then return an error.
1464 *
1465 * 1003.1g adds the ability to recvmsg() to query connection pending
1466 * status to recvmsg. We need to add that support in a way thats
1467 * clean when we restucture accept also.
1468 */
1469
Heiko Carstens20f37032009-01-14 14:14:23 +01001470SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1471 int __user *, upeer_addrlen, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472{
1473 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001474 struct file *newfile;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001475 int err, len, newfd, fput_needed;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001476 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
Ulrich Drepper77d27202008-07-23 21:29:35 -07001478 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001479 return -EINVAL;
1480
1481 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1482 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1483
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 if (!sock)
1486 goto out;
1487
1488 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001489 newsock = sock_alloc();
1490 if (!newsock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 goto out_put;
1492
1493 newsock->type = sock->type;
1494 newsock->ops = sock->ops;
1495
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 /*
1497 * We don't need try_module_get here, as the listening socket (sock)
1498 * has the protocol module (sock->ops->owner) held.
1499 */
1500 __module_get(newsock->ops->owner);
1501
Al Viro28407632012-08-17 23:54:15 -04001502 newfd = get_unused_fd_flags(flags);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001503 if (unlikely(newfd < 0)) {
1504 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001505 sock_release(newsock);
1506 goto out_put;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001507 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001508 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301509 if (IS_ERR(newfile)) {
Al Viro28407632012-08-17 23:54:15 -04001510 err = PTR_ERR(newfile);
1511 put_unused_fd(newfd);
1512 sock_release(newsock);
1513 goto out_put;
1514 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001515
Frank Filza79af592005-09-27 15:23:38 -07001516 err = security_socket_accept(sock, newsock);
1517 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001518 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001519
David Howellscdfbabf2017-03-09 08:09:05 +00001520 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001522 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
1524 if (upeer_sockaddr) {
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001525 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001526 &len, 2) < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001528 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001530 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001531 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001533 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 }
1535
1536 /* File flags are not inherited via accept() unlike another OSes. */
1537
David S. Miller39d8c1b2006-03-20 17:13:49 -08001538 fd_install(newfd, newfile);
1539 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001542 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543out:
1544 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001545out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001546 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001547 put_unused_fd(newfd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 goto out_put;
1549}
1550
Heiko Carstens20f37032009-01-14 14:14:23 +01001551SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1552 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001553{
Ulrich Drepperde11def2008-11-19 15:36:14 -08001554 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001555}
1556
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557/*
1558 * Attempt to connect to a socket with the server address. The address
1559 * is in user space so we verify it is OK and move it to kernel space.
1560 *
1561 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1562 * break bindings
1563 *
1564 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1565 * other SEQPACKET protocols that take time to connect() as it doesn't
1566 * include the -EINPROGRESS status for such sockets.
1567 */
1568
Heiko Carstens20f37032009-01-14 14:14:23 +01001569SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1570 int, addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571{
1572 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001573 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001574 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001576 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 if (!sock)
1578 goto out;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001579 err = move_addr_to_kernel(uservaddr, addrlen, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 if (err < 0)
1581 goto out_put;
1582
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001583 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001584 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 if (err)
1586 goto out_put;
1587
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001588 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 sock->file->f_flags);
1590out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001591 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592out:
1593 return err;
1594}
1595
1596/*
1597 * Get the local address ('name') of a socket object. Move the obtained
1598 * name to user space.
1599 */
1600
Heiko Carstens20f37032009-01-14 14:14:23 +01001601SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1602 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603{
1604 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001605 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001606 int len, err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001607
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001608 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 if (!sock)
1610 goto out;
1611
1612 err = security_socket_getsockname(sock);
1613 if (err)
1614 goto out_put;
1615
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001616 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617 if (err)
1618 goto out_put;
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001619 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620
1621out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001622 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623out:
1624 return err;
1625}
1626
1627/*
1628 * Get the remote address ('name') of a socket object. Move the obtained
1629 * name to user space.
1630 */
1631
Heiko Carstens20f37032009-01-14 14:14:23 +01001632SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1633 int __user *, usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634{
1635 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001636 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001637 int len, err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001639 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1640 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 err = security_socket_getpeername(sock);
1642 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001643 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644 return err;
1645 }
1646
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001647 err =
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001648 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001649 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 if (!err)
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001651 err = move_addr_to_user(&address, len, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001652 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001653 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654 }
1655 return err;
1656}
1657
1658/*
1659 * Send a datagram to a given address. We move the address into kernel
1660 * space and check the user space data area is readable before invoking
1661 * the protocol.
1662 */
1663
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001664SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001665 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001666 int, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667{
1668 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001669 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 int err;
1671 struct msghdr msg;
1672 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001673 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001674
Al Viro602bd0e2015-03-21 19:12:32 -04001675 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1676 if (unlikely(err))
1677 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001678 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1679 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001680 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001681
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001682 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001683 msg.msg_control = NULL;
1684 msg.msg_controllen = 0;
1685 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001686 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001687 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001688 if (err < 0)
1689 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001690 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001691 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 }
1693 if (sock->file->f_flags & O_NONBLOCK)
1694 flags |= MSG_DONTWAIT;
1695 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05001696 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001698out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001699 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001700out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701 return err;
1702}
1703
1704/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001705 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 */
1707
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001708SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00001709 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710{
1711 return sys_sendto(fd, buff, len, flags, NULL, 0);
1712}
1713
1714/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001715 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716 * sender. We verify the buffers are writable and if needed move the
1717 * sender address from kernel to user space.
1718 */
1719
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001720SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
Eric Dumazet95c96172012-04-15 05:58:06 +00001721 unsigned int, flags, struct sockaddr __user *, addr,
Heiko Carstens3e0fa652009-01-14 14:14:24 +01001722 int __user *, addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723{
1724 struct socket *sock;
1725 struct iovec iov;
1726 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001727 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001728 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001729 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
Al Viro602bd0e2015-03-21 19:12:32 -04001731 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1732 if (unlikely(err))
1733 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001734 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001736 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001738 msg.msg_control = NULL;
1739 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01001740 /* Save some cycles and don't copy the address if not needed */
1741 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1742 /* We assume all kernel code knows the size of sockaddr_storage */
1743 msg.msg_namelen = 0;
tadeusz.struk@intel.com130ed5d2015-12-15 10:46:17 -08001744 msg.msg_iocb = NULL;
Alexander Potapenko9f138fa2017-03-08 18:08:16 +01001745 msg.msg_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (sock->file->f_flags & O_NONBLOCK)
1747 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04001748 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001750 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001751 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001752 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001753 if (err2 < 0)
1754 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001756
1757 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001758out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 return err;
1760}
1761
1762/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001763 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 */
1765
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02001766SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1767 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768{
1769 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1770}
1771
1772/*
1773 * Set a socket option. Because we don't know the option lengths we have
1774 * to pass the user mode parameter for the protocols to sort out.
1775 */
1776
Heiko Carstens20f37032009-01-14 14:14:23 +01001777SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1778 char __user *, optval, int, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001780 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 struct socket *sock;
1782
1783 if (optlen < 0)
1784 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001785
1786 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1787 if (sock != NULL) {
1788 err = security_socket_setsockopt(sock, level, optname);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001789 if (err)
1790 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791
1792 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001793 err =
1794 sock_setsockopt(sock, level, optname, optval,
1795 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001797 err =
1798 sock->ops->setsockopt(sock, level, optname, optval,
1799 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001800out_put:
1801 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 }
1803 return err;
1804}
1805
1806/*
1807 * Get a socket option. Because we don't know the option lengths we have
1808 * to pass a user mode parameter for the protocols to sort out.
1809 */
1810
Heiko Carstens20f37032009-01-14 14:14:23 +01001811SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1812 char __user *, optval, int __user *, optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001814 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 struct socket *sock;
1816
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001817 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1818 if (sock != NULL) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001819 err = security_socket_getsockopt(sock, level, optname);
1820 if (err)
1821 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
1823 if (level == SOL_SOCKET)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001824 err =
1825 sock_getsockopt(sock, level, optname, optval,
1826 optlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827 else
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828 err =
1829 sock->ops->getsockopt(sock, level, optname, optval,
1830 optlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001831out_put:
1832 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 }
1834 return err;
1835}
1836
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837/*
1838 * Shutdown a socket.
1839 */
1840
Heiko Carstens754fe8d2009-01-14 14:14:09 +01001841SYSCALL_DEFINE2(shutdown, int, fd, int, how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001843 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844 struct socket *sock;
1845
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001846 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1847 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001849 if (!err)
1850 err = sock->ops->shutdown(sock, how);
1851 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 }
1853 return err;
1854}
1855
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001856/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 * fields which are the same type (int / unsigned) on our platforms.
1858 */
1859#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1860#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1861#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1862
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001863struct used_address {
1864 struct sockaddr_storage name;
1865 unsigned int name_len;
1866};
1867
Al Viroda184282015-03-21 19:29:06 -04001868static int copy_msghdr_from_user(struct msghdr *kmsg,
1869 struct user_msghdr __user *umsg,
1870 struct sockaddr __user **save_addr,
1871 struct iovec **iov)
Dan Carpenter1661bf32013-10-03 00:27:20 +03001872{
Al Viro08adb7d2014-11-10 20:23:13 -05001873 struct sockaddr __user *uaddr;
1874 struct iovec __user *uiov;
Al Viroc0371da2014-11-24 10:42:55 -05001875 size_t nr_segs;
Al Viro08adb7d2014-11-10 20:23:13 -05001876 ssize_t err;
1877
1878 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1879 __get_user(uaddr, &umsg->msg_name) ||
1880 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1881 __get_user(uiov, &umsg->msg_iov) ||
Al Viroc0371da2014-11-24 10:42:55 -05001882 __get_user(nr_segs, &umsg->msg_iovlen) ||
Al Viro08adb7d2014-11-10 20:23:13 -05001883 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1884 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1885 __get_user(kmsg->msg_flags, &umsg->msg_flags))
Dan Carpenter1661bf32013-10-03 00:27:20 +03001886 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00001887
Al Viro08adb7d2014-11-10 20:23:13 -05001888 if (!uaddr)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07001889 kmsg->msg_namelen = 0;
1890
Matthew Leachdbb490b2014-03-11 11:58:27 +00001891 if (kmsg->msg_namelen < 0)
1892 return -EINVAL;
1893
Dan Carpenter1661bf32013-10-03 00:27:20 +03001894 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03001895 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05001896
1897 if (save_addr)
1898 *save_addr = uaddr;
1899
1900 if (uaddr && kmsg->msg_namelen) {
1901 if (!save_addr) {
1902 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1903 kmsg->msg_name);
1904 if (err < 0)
1905 return err;
1906 }
1907 } else {
1908 kmsg->msg_name = NULL;
1909 kmsg->msg_namelen = 0;
1910 }
1911
Al Viroc0371da2014-11-24 10:42:55 -05001912 if (nr_segs > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05001913 return -EMSGSIZE;
1914
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001915 kmsg->msg_iocb = NULL;
1916
Al Viroda184282015-03-21 19:29:06 -04001917 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1918 UIO_FASTIOV, iov, &kmsg->msg_iter);
Dan Carpenter1661bf32013-10-03 00:27:20 +03001919}
1920
Al Viro666547f2014-04-06 14:03:05 -04001921static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00001922 struct msghdr *msg_sys, unsigned int flags,
Tom Herbert28a94d82016-03-07 14:11:02 -08001923 struct used_address *used_address,
1924 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001926 struct compat_msghdr __user *msg_compat =
1927 (struct compat_msghdr __user *)msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001928 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
Alex Williamsonb9d717a2005-09-26 14:28:02 -07001930 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Amit Kushwaha846cc122016-12-08 18:21:53 +05301931 __aligned(sizeof(__kernel_size_t));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001932 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05001934 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05001935 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001936
Al Viro08adb7d2014-11-10 20:23:13 -05001937 msg_sys->msg_name = &address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938
Al Viro08449322014-11-09 22:33:45 -05001939 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05001940 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
Al Viro08449322014-11-09 22:33:45 -05001941 else
Al Viro08adb7d2014-11-10 20:23:13 -05001942 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001943 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04001944 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945
1946 err = -ENOBUFS;
1947
Anton Blanchard228e5482011-05-02 20:21:35 +00001948 if (msg_sys->msg_controllen > INT_MAX)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 goto out_freeiov;
Tom Herbert28a94d82016-03-07 14:11:02 -08001950 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00001951 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001953 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00001954 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001955 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 if (err)
1957 goto out_freeiov;
Anton Blanchard228e5482011-05-02 20:21:35 +00001958 ctl_buf = msg_sys->msg_control;
1959 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 } else if (ctl_len) {
David S. Millerac4340f2017-01-04 13:24:19 -05001961 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
1962 CMSG_ALIGN(sizeof(struct cmsghdr)));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001963 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001965 if (ctl_buf == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 goto out_freeiov;
1967 }
1968 err = -EFAULT;
1969 /*
Anton Blanchard228e5482011-05-02 20:21:35 +00001970 * Careful! Before this, msg_sys->msg_control contains a user pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1972 * checking falls down on this.
1973 */
Namhyung Kimfb8621b2010-09-07 03:55:00 +00001974 if (copy_from_user(ctl_buf,
Anton Blanchard228e5482011-05-02 20:21:35 +00001975 (void __user __force *)msg_sys->msg_control,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001976 ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00001978 msg_sys->msg_control = ctl_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 }
Anton Blanchard228e5482011-05-02 20:21:35 +00001980 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981
1982 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00001983 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001984 /*
1985 * If this is sendmmsg() and current destination address is same as
1986 * previously succeeded address, omit asking LSM's decision.
1987 * used_address->name_len is initialized to UINT_MAX so that the first
1988 * destination address never matches.
1989 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07001990 if (used_address && msg_sys->msg_name &&
1991 used_address->name_len == msg_sys->msg_namelen &&
1992 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001993 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05001994 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001995 goto out_freectl;
1996 }
Al Virod8725c82014-12-11 00:02:50 -05001997 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00001998 /*
1999 * If this is sendmmsg() and sending to current destination address was
2000 * successful, remember it.
2001 */
2002 if (used_address && err >= 0) {
2003 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002004 if (msg_sys->msg_name)
2005 memcpy(&used_address->name, msg_sys->msg_name,
2006 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002007 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008
2009out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002010 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2012out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04002013 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002014 return err;
2015}
2016
2017/*
2018 * BSD sendmsg interface
2019 */
2020
Al Viro666547f2014-04-06 14:03:05 -04002021long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Anton Blanchard228e5482011-05-02 20:21:35 +00002022{
2023 int fput_needed, err;
2024 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002025 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002026
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002027 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002028 if (!sock)
2029 goto out;
2030
Tom Herbert28a94d82016-03-07 14:11:02 -08002031 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00002032
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002033 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002034out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035 return err;
2036}
2037
Al Viro666547f2014-04-06 14:03:05 -04002038SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002039{
2040 if (flags & MSG_CMSG_COMPAT)
2041 return -EINVAL;
2042 return __sys_sendmsg(fd, msg, flags);
2043}
2044
Anton Blanchard228e5482011-05-02 20:21:35 +00002045/*
2046 * Linux sendmmsg interface
2047 */
2048
2049int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2050 unsigned int flags)
2051{
2052 int fput_needed, err, datagrams;
2053 struct socket *sock;
2054 struct mmsghdr __user *entry;
2055 struct compat_mmsghdr __user *compat_entry;
2056 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002057 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002058 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002059
Anton Blanchard98382f42011-08-04 14:07:39 +00002060 if (vlen > UIO_MAXIOV)
2061 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002062
2063 datagrams = 0;
2064
2065 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2066 if (!sock)
2067 return err;
2068
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002069 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002070 entry = mmsg;
2071 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002072 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002073 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002074
2075 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002076 if (datagrams == vlen - 1)
2077 flags = oflags;
2078
Anton Blanchard228e5482011-05-02 20:21:35 +00002079 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002080 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002081 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002082 if (err < 0)
2083 break;
2084 err = __put_user(err, &compat_entry->msg_len);
2085 ++compat_entry;
2086 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002087 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002088 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002089 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002090 if (err < 0)
2091 break;
2092 err = put_user(err, &entry->msg_len);
2093 ++entry;
2094 }
2095
2096 if (err)
2097 break;
2098 ++datagrams;
Soheil Hassas Yeganeh30238982016-11-04 15:36:49 -04002099 if (msg_data_left(&msg_sys))
2100 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002101 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002102 }
2103
Anton Blanchard228e5482011-05-02 20:21:35 +00002104 fput_light(sock->file, fput_needed);
2105
Anton Blanchard728ffb82011-08-04 14:07:38 +00002106 /* We only return an error if no datagrams were able to be sent */
2107 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002108 return datagrams;
2109
Anton Blanchard228e5482011-05-02 20:21:35 +00002110 return err;
2111}
2112
2113SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2114 unsigned int, vlen, unsigned int, flags)
2115{
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002116 if (flags & MSG_CMSG_COMPAT)
2117 return -EINVAL;
Anton Blanchard228e5482011-05-02 20:21:35 +00002118 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2119}
2120
Al Viro666547f2014-04-06 14:03:05 -04002121static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
Eric Dumazet95c96172012-04-15 05:58:06 +00002122 struct msghdr *msg_sys, unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002124 struct compat_msghdr __user *msg_compat =
2125 (struct compat_msghdr __user *)msg;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126 struct iovec iovstack[UIO_FASTIOV];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002127 struct iovec *iov = iovstack;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002129 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002130 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131
2132 /* kernel mode address */
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002133 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134
2135 /* user mode address pointers */
2136 struct sockaddr __user *uaddr;
Al Viro08adb7d2014-11-10 20:23:13 -05002137 int __user *uaddr_len = COMPAT_NAMELEN(msg);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002138
Al Viro08adb7d2014-11-10 20:23:13 -05002139 msg_sys->msg_name = &addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002141 if (MSG_CMSG_COMPAT & flags)
Al Viro08adb7d2014-11-10 20:23:13 -05002142 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002143 else
Al Viro08adb7d2014-11-10 20:23:13 -05002144 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 if (err < 0)
Al Viroda184282015-03-21 19:29:06 -04002146 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002148 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2149 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002150
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002151 /* We assume all kernel code knows the size of sockaddr_storage */
2152 msg_sys->msg_namelen = 0;
2153
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 if (sock->file->f_flags & O_NONBLOCK)
2155 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002156 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 if (err < 0)
2158 goto out_freeiov;
2159 len = err;
2160
2161 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002162 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002163 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002164 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165 if (err < 0)
2166 goto out_freeiov;
2167 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002168 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002169 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170 if (err)
2171 goto out_freeiov;
2172 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002173 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174 &msg_compat->msg_controllen);
2175 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002176 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 &msg->msg_controllen);
2178 if (err)
2179 goto out_freeiov;
2180 err = len;
2181
2182out_freeiov:
Al Viroda184282015-03-21 19:29:06 -04002183 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002184 return err;
2185}
2186
2187/*
2188 * BSD recvmsg interface
2189 */
2190
Al Viro666547f2014-04-06 14:03:05 -04002191long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002192{
2193 int fput_needed, err;
2194 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002195 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002196
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002197 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002198 if (!sock)
2199 goto out;
2200
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002201 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002202
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002203 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204out:
2205 return err;
2206}
2207
Al Viro666547f2014-04-06 14:03:05 -04002208SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002209 unsigned int, flags)
2210{
2211 if (flags & MSG_CMSG_COMPAT)
2212 return -EINVAL;
2213 return __sys_recvmsg(fd, msg, flags);
2214}
2215
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002216/*
2217 * Linux recvmmsg interface
2218 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002220int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2221 unsigned int flags, struct timespec *timeout)
2222{
2223 int fput_needed, err, datagrams;
2224 struct socket *sock;
2225 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002226 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002227 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002228 struct timespec64 end_time;
2229 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002230
2231 if (timeout &&
2232 poll_select_set_timeout(&end_time, timeout->tv_sec,
2233 timeout->tv_nsec))
2234 return -EINVAL;
2235
2236 datagrams = 0;
2237
2238 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2239 if (!sock)
2240 return err;
2241
2242 err = sock_error(sock->sk);
Maxime Jayate623a9e2017-02-21 18:35:51 +01002243 if (err) {
2244 datagrams = err;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002245 goto out_put;
Maxime Jayate623a9e2017-02-21 18:35:51 +01002246 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002247
2248 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002249 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002250
2251 while (datagrams < vlen) {
2252 /*
2253 * No need to ask LSM for more than the first datagram.
2254 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002255 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002256 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002257 &msg_sys, flags & ~MSG_WAITFORONE,
2258 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002259 if (err < 0)
2260 break;
2261 err = __put_user(err, &compat_entry->msg_len);
2262 ++compat_entry;
2263 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002264 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002265 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002266 &msg_sys, flags & ~MSG_WAITFORONE,
2267 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002268 if (err < 0)
2269 break;
2270 err = put_user(err, &entry->msg_len);
2271 ++entry;
2272 }
2273
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002274 if (err)
2275 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002276 ++datagrams;
2277
Brandon L Black71c5c152010-03-26 16:18:03 +00002278 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2279 if (flags & MSG_WAITFORONE)
2280 flags |= MSG_DONTWAIT;
2281
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002282 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002283 ktime_get_ts64(&timeout64);
2284 *timeout = timespec64_to_timespec(
2285 timespec64_sub(end_time, timeout64));
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002286 if (timeout->tv_sec < 0) {
2287 timeout->tv_sec = timeout->tv_nsec = 0;
2288 break;
2289 }
2290
2291 /* Timeout, return less than vlen datagrams */
2292 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2293 break;
2294 }
2295
2296 /* Out of band data, return right away */
2297 if (msg_sys.msg_flags & MSG_OOB)
2298 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002299 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002300 }
2301
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002302 if (err == 0)
2303 goto out_put;
2304
2305 if (datagrams == 0) {
2306 datagrams = err;
2307 goto out_put;
2308 }
2309
2310 /*
2311 * We may return less entries than requested (vlen) if the
2312 * sock is non block and there aren't enough datagrams...
2313 */
2314 if (err != -EAGAIN) {
2315 /*
2316 * ... or if recvmsg returns an error after we
2317 * received some datagrams, where we record the
2318 * error to return on the next call or if the
2319 * app asks about it using getsockopt(SO_ERROR).
2320 */
2321 sock->sk->sk_err = -err;
2322 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002323out_put:
2324 fput_light(sock->file, fput_needed);
2325
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002326 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002327}
2328
2329SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2330 unsigned int, vlen, unsigned int, flags,
2331 struct timespec __user *, timeout)
2332{
2333 int datagrams;
2334 struct timespec timeout_sys;
2335
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002336 if (flags & MSG_CMSG_COMPAT)
2337 return -EINVAL;
2338
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002339 if (!timeout)
2340 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2341
2342 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2343 return -EFAULT;
2344
2345 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2346
2347 if (datagrams > 0 &&
2348 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2349 datagrams = -EFAULT;
2350
2351 return datagrams;
2352}
2353
2354#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355/* Argument list sizes for sys_socketcall */
2356#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002357static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002358 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2359 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2360 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002361 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002362};
2363
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364#undef AL
2365
2366/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002367 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368 *
2369 * Argument checking cleaned up. Saved 20% in size.
2370 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002371 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372 */
2373
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002374SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375{
Chen Gang2950fa92013-04-07 16:55:23 +08002376 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002377 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002379 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380
Anton Blanchard228e5482011-05-02 20:21:35 +00002381 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382 return -EINVAL;
2383
Arjan van de Ven47379052009-09-28 12:57:44 -07002384 len = nargs[call];
2385 if (len > sizeof(a))
2386 return -EINVAL;
2387
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002389 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002391
Chen Gang2950fa92013-04-07 16:55:23 +08002392 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2393 if (err)
2394 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002395
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002396 a0 = a[0];
2397 a1 = a[1];
2398
2399 switch (call) {
2400 case SYS_SOCKET:
2401 err = sys_socket(a0, a1, a[2]);
2402 break;
2403 case SYS_BIND:
2404 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2405 break;
2406 case SYS_CONNECT:
2407 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2408 break;
2409 case SYS_LISTEN:
2410 err = sys_listen(a0, a1);
2411 break;
2412 case SYS_ACCEPT:
Ulrich Drepperde11def2008-11-19 15:36:14 -08002413 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2414 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002415 break;
2416 case SYS_GETSOCKNAME:
2417 err =
2418 sys_getsockname(a0, (struct sockaddr __user *)a1,
2419 (int __user *)a[2]);
2420 break;
2421 case SYS_GETPEERNAME:
2422 err =
2423 sys_getpeername(a0, (struct sockaddr __user *)a1,
2424 (int __user *)a[2]);
2425 break;
2426 case SYS_SOCKETPAIR:
2427 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2428 break;
2429 case SYS_SEND:
2430 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2431 break;
2432 case SYS_SENDTO:
2433 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2434 (struct sockaddr __user *)a[4], a[5]);
2435 break;
2436 case SYS_RECV:
2437 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2438 break;
2439 case SYS_RECVFROM:
2440 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2441 (struct sockaddr __user *)a[4],
2442 (int __user *)a[5]);
2443 break;
2444 case SYS_SHUTDOWN:
2445 err = sys_shutdown(a0, a1);
2446 break;
2447 case SYS_SETSOCKOPT:
2448 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2449 break;
2450 case SYS_GETSOCKOPT:
2451 err =
2452 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2453 (int __user *)a[4]);
2454 break;
2455 case SYS_SENDMSG:
Al Viro666547f2014-04-06 14:03:05 -04002456 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002457 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002458 case SYS_SENDMMSG:
2459 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2460 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002461 case SYS_RECVMSG:
Al Viro666547f2014-04-06 14:03:05 -04002462 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002463 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002464 case SYS_RECVMMSG:
2465 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2466 (struct timespec __user *)a[4]);
2467 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002468 case SYS_ACCEPT4:
2469 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2470 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002471 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002472 default:
2473 err = -EINVAL;
2474 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475 }
2476 return err;
2477}
2478
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002479#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002481/**
2482 * sock_register - add a socket protocol handler
2483 * @ops: description of protocol
2484 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002485 * This function is called by a protocol handler that wants to
2486 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002487 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002488 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002489 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002490int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491{
2492 int err;
2493
2494 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002495 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496 return -ENOBUFS;
2497 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002498
2499 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002500 if (rcu_dereference_protected(net_families[ops->family],
2501 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002502 err = -EEXIST;
2503 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002504 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002505 err = 0;
2506 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002507 spin_unlock(&net_family_lock);
2508
Yang Yingliang3410f222014-02-12 17:09:55 +08002509 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002510 return err;
2511}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002512EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002514/**
2515 * sock_unregister - remove a protocol handler
2516 * @family: protocol family to remove
2517 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518 * This function is called by a protocol handler that wants to
2519 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002520 * new socket creation.
2521 *
2522 * If protocol handler is a module, then it can use module reference
2523 * counts to protect against new references. If protocol handler is not
2524 * a module then it needs to provide its own protection in
2525 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002526 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002527void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002529 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002531 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00002532 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002533 spin_unlock(&net_family_lock);
2534
2535 synchronize_rcu();
2536
Yang Yingliang3410f222014-02-12 17:09:55 +08002537 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002539EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540
Andi Kleen77d76ea2005-12-22 12:43:42 -08002541static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542{
Nick Pigginb3e19d92011-01-07 17:50:11 +11002543 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00002544 /*
2545 * Initialize the network sysctl infrastructure.
2546 */
2547 err = net_sysctl_init();
2548 if (err)
2549 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002550
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002552 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553 */
2554 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555
2556 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002557 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558 */
2559
2560 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11002561
2562 err = register_filesystem(&sock_fs_type);
2563 if (err)
2564 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002566 if (IS_ERR(sock_mnt)) {
2567 err = PTR_ERR(sock_mnt);
2568 goto out_mount;
2569 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08002570
2571 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572 */
2573
2574#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00002575 err = netfilter_init();
2576 if (err)
2577 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578#endif
David S. Millercbeb3212005-12-22 12:58:55 -08002579
Daniel Borkmann408eccc2014-04-01 16:20:23 +02002580 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00002581
Nick Pigginb3e19d92011-01-07 17:50:11 +11002582out:
2583 return err;
2584
2585out_mount:
2586 unregister_filesystem(&sock_fs_type);
2587out_fs:
2588 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002589}
2590
Andi Kleen77d76ea2005-12-22 12:43:42 -08002591core_initcall(sock_init); /* early initcall */
2592
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593#ifdef CONFIG_PROC_FS
2594void socket_seq_show(struct seq_file *seq)
2595{
2596 int cpu;
2597 int counter = 0;
2598
KAMEZAWA Hiroyuki6f912042006-04-10 22:52:50 -07002599 for_each_possible_cpu(cpu)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002600 counter += per_cpu(sockets_in_use, cpu);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601
2602 /* It can be negative, by the way. 8) */
2603 if (counter < 0)
2604 counter = 0;
2605
2606 seq_printf(seq, "sockets: used %d\n", counter);
2607}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002608#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609
Shaun Pereira89bbfc92006-03-21 23:58:08 -08002610#ifdef CONFIG_COMPAT
Arnd Bergmann6b960182009-11-06 23:10:54 -08002611static int do_siocgstamp(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002612 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002613{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002614 mm_segment_t old_fs = get_fs();
2615 struct timeval ktv;
2616 int err;
2617
2618 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002619 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002620 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002621 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002622 err = compat_put_timeval(&ktv, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002623
Arnd Bergmann7a229382009-11-06 23:00:29 -08002624 return err;
2625}
2626
Arnd Bergmann6b960182009-11-06 23:10:54 -08002627static int do_siocgstampns(struct net *net, struct socket *sock,
H. Peter Anvin644595f2012-02-19 17:51:59 -08002628 unsigned int cmd, void __user *up)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002629{
Arnd Bergmann7a229382009-11-06 23:00:29 -08002630 mm_segment_t old_fs = get_fs();
2631 struct timespec kts;
2632 int err;
2633
2634 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08002635 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002636 set_fs(old_fs);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002637 if (!err)
Mikulas Patockaed6fe9d2012-09-01 12:34:07 -04002638 err = compat_put_timespec(&kts, up);
H. Peter Anvin644595f2012-02-19 17:51:59 -08002639
Arnd Bergmann7a229382009-11-06 23:00:29 -08002640 return err;
2641}
2642
Arnd Bergmann6b960182009-11-06 23:10:54 -08002643static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002644{
2645 struct ifreq __user *uifr;
2646 int err;
2647
2648 uifr = compat_alloc_user_space(sizeof(struct ifreq));
Arnd Bergmann6b960182009-11-06 23:10:54 -08002649 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002650 return -EFAULT;
2651
Arnd Bergmann6b960182009-11-06 23:10:54 -08002652 err = dev_ioctl(net, SIOCGIFNAME, uifr);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002653 if (err)
2654 return err;
2655
Arnd Bergmann6b960182009-11-06 23:10:54 -08002656 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002657 return -EFAULT;
2658
2659 return 0;
2660}
2661
Arnd Bergmann6b960182009-11-06 23:10:54 -08002662static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002663{
Arnd Bergmann6b960182009-11-06 23:10:54 -08002664 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002665 struct ifconf ifc;
2666 struct ifconf __user *uifc;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002667 struct compat_ifreq __user *ifr32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002668 struct ifreq __user *ifr;
2669 unsigned int i, j;
2670 int err;
2671
Arnd Bergmann6b960182009-11-06 23:10:54 -08002672 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002673 return -EFAULT;
2674
Mathias Krause43da5f22012-08-15 11:31:57 +00002675 memset(&ifc, 0, sizeof(ifc));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002676 if (ifc32.ifcbuf == 0) {
2677 ifc32.ifc_len = 0;
2678 ifc.ifc_len = 0;
2679 ifc.ifc_req = NULL;
2680 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2681 } else {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002682 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2683 sizeof(struct ifreq);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002684 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2685 ifc.ifc_len = len;
2686 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2687 ifr32 = compat_ptr(ifc32.ifcbuf);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002688 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
Arnd Bergmann6b960182009-11-06 23:10:54 -08002689 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002690 return -EFAULT;
2691 ifr++;
2692 ifr32++;
2693 }
2694 }
2695 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2696 return -EFAULT;
2697
Arnd Bergmann6b960182009-11-06 23:10:54 -08002698 err = dev_ioctl(net, SIOCGIFCONF, uifc);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002699 if (err)
2700 return err;
2701
2702 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2703 return -EFAULT;
2704
2705 ifr = ifc.ifc_req;
2706 ifr32 = compat_ptr(ifc32.ifcbuf);
2707 for (i = 0, j = 0;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002708 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2709 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2710 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002711 return -EFAULT;
2712 ifr32++;
2713 ifr++;
2714 }
2715
2716 if (ifc32.ifcbuf == 0) {
2717 /* Translate from 64-bit structure multiple to
2718 * a 32-bit one.
2719 */
2720 i = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08002721 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
Arnd Bergmann7a229382009-11-06 23:00:29 -08002722 ifc32.ifc_len = i;
2723 } else {
2724 ifc32.ifc_len = i;
2725 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08002726 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002727 return -EFAULT;
2728
2729 return 0;
2730}
2731
Arnd Bergmann6b960182009-11-06 23:10:54 -08002732static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002733{
Ben Hutchings3a7da392011-03-17 07:34:32 +00002734 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2735 bool convert_in = false, convert_out = false;
2736 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2737 struct ethtool_rxnfc __user *rxnfc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002738 struct ifreq __user *ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002739 u32 rule_cnt = 0, actual_rule_cnt;
2740 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002741 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00002742 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002743
2744 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2745 return -EFAULT;
2746
Ben Hutchings3a7da392011-03-17 07:34:32 +00002747 compat_rxnfc = compat_ptr(data);
2748
2749 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002750 return -EFAULT;
2751
Ben Hutchings3a7da392011-03-17 07:34:32 +00002752 /* Most ethtool structures are defined without padding.
2753 * Unfortunately struct ethtool_rxnfc is an exception.
2754 */
2755 switch (ethcmd) {
2756 default:
2757 break;
2758 case ETHTOOL_GRXCLSRLALL:
2759 /* Buffer size is variable */
2760 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2761 return -EFAULT;
2762 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2763 return -ENOMEM;
2764 buf_size += rule_cnt * sizeof(u32);
2765 /* fall through */
2766 case ETHTOOL_GRXRINGS:
2767 case ETHTOOL_GRXCLSRLCNT:
2768 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00002769 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002770 convert_out = true;
2771 /* fall through */
2772 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00002773 buf_size += sizeof(struct ethtool_rxnfc);
2774 convert_in = true;
2775 break;
2776 }
2777
2778 ifr = compat_alloc_user_space(buf_size);
Stephen Hemminger954b1242013-02-11 06:22:28 +00002779 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
Ben Hutchings3a7da392011-03-17 07:34:32 +00002780
2781 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2782 return -EFAULT;
2783
2784 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2785 &ifr->ifr_ifru.ifru_data))
2786 return -EFAULT;
2787
2788 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00002789 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00002790 * fs.ring_cookie and at the end of fs, but nowhere else.
2791 */
Alexander Duyck127fe532011-04-08 18:01:59 +00002792 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2793 sizeof(compat_rxnfc->fs.m_ext) !=
2794 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2795 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00002796 BUILD_BUG_ON(
2797 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2798 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2799 offsetof(struct ethtool_rxnfc, fs.location) -
2800 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2801
2802 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002803 (void __user *)(&rxnfc->fs.m_ext + 1) -
2804 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002805 copy_in_user(&rxnfc->fs.ring_cookie,
2806 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002807 (void __user *)(&rxnfc->fs.location + 1) -
2808 (void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002809 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2810 sizeof(rxnfc->rule_cnt)))
2811 return -EFAULT;
2812 }
2813
2814 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2815 if (ret)
2816 return ret;
2817
2818 if (convert_out) {
2819 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002820 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2821 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002822 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2823 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00002824 (const void __user *)(&rxnfc->fs.location + 1) -
2825 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00002826 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2827 sizeof(rxnfc->rule_cnt)))
2828 return -EFAULT;
2829
2830 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2831 /* As an optimisation, we only copy the actual
2832 * number of rules that the underlying
2833 * function returned. Since Mallory might
2834 * change the rule count in user memory, we
2835 * check that it is less than the rule count
2836 * originally given (as the user buffer size),
2837 * which has been range-checked.
2838 */
2839 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2840 return -EFAULT;
2841 if (actual_rule_cnt < rule_cnt)
2842 rule_cnt = actual_rule_cnt;
2843 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2844 &rxnfc->rule_locs[0],
2845 rule_cnt * sizeof(u32)))
2846 return -EFAULT;
2847 }
2848 }
2849
2850 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002851}
2852
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002853static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2854{
2855 void __user *uptr;
2856 compat_uptr_t uptr32;
2857 struct ifreq __user *uifr;
2858
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002859 uifr = compat_alloc_user_space(sizeof(*uifr));
Arnd Bergmann7a50a242009-11-08 20:57:03 -08002860 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2861 return -EFAULT;
2862
2863 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2864 return -EFAULT;
2865
2866 uptr = compat_ptr(uptr32);
2867
2868 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2869 return -EFAULT;
2870
2871 return dev_ioctl(net, SIOCWANDEV, uifr);
2872}
2873
Arnd Bergmann6b960182009-11-06 23:10:54 -08002874static int bond_ioctl(struct net *net, unsigned int cmd,
2875 struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002876{
2877 struct ifreq kifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002878 mm_segment_t old_fs;
2879 int err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002880
2881 switch (cmd) {
2882 case SIOCBONDENSLAVE:
2883 case SIOCBONDRELEASE:
2884 case SIOCBONDSETHWADDR:
2885 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08002886 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002887 return -EFAULT;
2888
2889 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002890 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002891 err = dev_ioctl(net, cmd,
2892 (struct ifreq __user __force *) &kifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002893 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002894
2895 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002896 default:
Linus Torvalds07d106d2012-01-05 15:40:12 -08002897 return -ENOIOCTLCMD;
Joe Perchesccbd6a52010-05-14 10:58:26 +00002898 }
Arnd Bergmann7a229382009-11-06 23:00:29 -08002899}
2900
Ben Hutchings590d4692013-11-18 17:04:13 +00002901/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2902static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08002903 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002904{
2905 struct ifreq __user *u_ifreq64;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002906 char tmp_buf[IFNAMSIZ];
2907 void __user *data64;
2908 u32 data32;
2909
2910 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2911 IFNAMSIZ))
2912 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002913 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002914 return -EFAULT;
2915 data64 = compat_ptr(data32);
2916
2917 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2918
Arnd Bergmann7a229382009-11-06 23:00:29 -08002919 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2920 IFNAMSIZ))
2921 return -EFAULT;
Ben Hutchings417c3522013-11-18 17:04:58 +00002922 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002923 return -EFAULT;
2924
Arnd Bergmann6b960182009-11-06 23:10:54 -08002925 return dev_ioctl(net, cmd, u_ifreq64);
Arnd Bergmann7a229382009-11-06 23:00:29 -08002926}
2927
Arnd Bergmann6b960182009-11-06 23:10:54 -08002928static int dev_ifsioc(struct net *net, struct socket *sock,
2929 unsigned int cmd, struct compat_ifreq __user *uifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08002930{
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002931 struct ifreq __user *uifr;
Arnd Bergmann7a229382009-11-06 23:00:29 -08002932 int err;
2933
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002934 uifr = compat_alloc_user_space(sizeof(*uifr));
2935 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2936 return -EFAULT;
2937
2938 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2939
Arnd Bergmann7a229382009-11-06 23:00:29 -08002940 if (!err) {
2941 switch (cmd) {
2942 case SIOCGIFFLAGS:
2943 case SIOCGIFMETRIC:
2944 case SIOCGIFMTU:
2945 case SIOCGIFMEM:
2946 case SIOCGIFHWADDR:
2947 case SIOCGIFINDEX:
2948 case SIOCGIFADDR:
2949 case SIOCGIFBRDADDR:
2950 case SIOCGIFDSTADDR:
2951 case SIOCGIFNETMASK:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002952 case SIOCGIFPFLAGS:
Arnd Bergmann7a229382009-11-06 23:00:29 -08002953 case SIOCGIFTXQLEN:
Arnd Bergmannfab25322009-11-08 20:56:21 -08002954 case SIOCGMIIPHY:
2955 case SIOCGMIIREG:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002956 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08002957 err = -EFAULT;
2958 break;
2959 }
2960 }
2961 return err;
2962}
2963
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002964static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2965 struct compat_ifreq __user *uifr32)
2966{
2967 struct ifreq ifr;
2968 struct compat_ifmap __user *uifmap32;
2969 mm_segment_t old_fs;
2970 int err;
2971
2972 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2973 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002974 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2975 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2976 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2977 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2978 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2979 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002980 if (err)
2981 return -EFAULT;
2982
2983 old_fs = get_fs();
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002984 set_fs(KERNEL_DS);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00002985 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002986 set_fs(old_fs);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002987
2988 if (cmd == SIOCGIFMAP && !err) {
2989 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07002990 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2991 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2992 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2993 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2994 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2995 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00002996 if (err)
2997 err = -EFAULT;
2998 }
2999 return err;
3000}
3001
Arnd Bergmann7a229382009-11-06 23:00:29 -08003002struct rtentry32 {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003003 u32 rt_pad1;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003004 struct sockaddr rt_dst; /* target address */
3005 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3006 struct sockaddr rt_genmask; /* target network mask (IP) */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003007 unsigned short rt_flags;
3008 short rt_pad2;
3009 u32 rt_pad3;
3010 unsigned char rt_tos;
3011 unsigned char rt_class;
3012 short rt_pad4;
3013 short rt_metric; /* +1 for binary compatibility! */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003014 /* char * */ u32 rt_dev; /* forcing the device at add */
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003015 u32 rt_mtu; /* per route MTU/Window */
3016 u32 rt_window; /* Window clamping */
Arnd Bergmann7a229382009-11-06 23:00:29 -08003017 unsigned short rt_irtt; /* Initial RTT */
3018};
3019
3020struct in6_rtmsg32 {
3021 struct in6_addr rtmsg_dst;
3022 struct in6_addr rtmsg_src;
3023 struct in6_addr rtmsg_gateway;
3024 u32 rtmsg_type;
3025 u16 rtmsg_dst_len;
3026 u16 rtmsg_src_len;
3027 u32 rtmsg_metric;
3028 u32 rtmsg_info;
3029 u32 rtmsg_flags;
3030 s32 rtmsg_ifindex;
3031};
3032
Arnd Bergmann6b960182009-11-06 23:10:54 -08003033static int routing_ioctl(struct net *net, struct socket *sock,
3034 unsigned int cmd, void __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003035{
3036 int ret;
3037 void *r = NULL;
3038 struct in6_rtmsg r6;
3039 struct rtentry r4;
3040 char devname[16];
3041 u32 rtdev;
3042 mm_segment_t old_fs = get_fs();
3043
Arnd Bergmann6b960182009-11-06 23:10:54 -08003044 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3045 struct in6_rtmsg32 __user *ur6 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003046 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003047 3 * sizeof(struct in6_addr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003048 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3049 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3050 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3051 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3052 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3053 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3054 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003055
3056 r = (void *) &r6;
3057 } else { /* ipv4 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003058 struct rtentry32 __user *ur4 = argp;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003059 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
Arnd Bergmann7a229382009-11-06 23:00:29 -08003060 3 * sizeof(struct sockaddr));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003061 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3062 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3063 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3064 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3065 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3066 ret |= get_user(rtdev, &(ur4->rt_dev));
Arnd Bergmann7a229382009-11-06 23:00:29 -08003067 if (rtdev) {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003068 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
stephen hemmingerc3f52ae2011-02-23 09:06:48 +00003069 r4.rt_dev = (char __user __force *)devname;
3070 devname[15] = 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003071 } else
3072 r4.rt_dev = NULL;
3073
3074 r = (void *) &r4;
3075 }
3076
3077 if (ret) {
3078 ret = -EFAULT;
3079 goto out;
3080 }
3081
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003082 set_fs(KERNEL_DS);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003083 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003084 set_fs(old_fs);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003085
3086out:
Arnd Bergmann7a229382009-11-06 23:00:29 -08003087 return ret;
3088}
3089
3090/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3091 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003092 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003093 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003094static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003095{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003096 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003097
Arnd Bergmann6b960182009-11-06 23:10:54 -08003098 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003099 return -EFAULT;
3100 if (tmp == BRCTL_GET_VERSION)
3101 return BRCTL_VERSION + 1;
3102 return -EINVAL;
3103}
3104
Arnd Bergmann6b960182009-11-06 23:10:54 -08003105static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3106 unsigned int cmd, unsigned long arg)
3107{
3108 void __user *argp = compat_ptr(arg);
3109 struct sock *sk = sock->sk;
3110 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003111
Arnd Bergmann6b960182009-11-06 23:10:54 -08003112 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003113 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003114
Arnd Bergmann6b960182009-11-06 23:10:54 -08003115 switch (cmd) {
3116 case SIOCSIFBR:
3117 case SIOCGIFBR:
3118 return old_bridge_ioctl(argp);
3119 case SIOCGIFNAME:
3120 return dev_ifname32(net, argp);
3121 case SIOCGIFCONF:
3122 return dev_ifconf(net, argp);
3123 case SIOCETHTOOL:
3124 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003125 case SIOCWANDEV:
3126 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003127 case SIOCGIFMAP:
3128 case SIOCSIFMAP:
3129 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003130 case SIOCBONDENSLAVE:
3131 case SIOCBONDRELEASE:
3132 case SIOCBONDSETHWADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003133 case SIOCBONDCHANGEACTIVE:
3134 return bond_ioctl(net, cmd, argp);
3135 case SIOCADDRT:
3136 case SIOCDELRT:
3137 return routing_ioctl(net, sock, cmd, argp);
3138 case SIOCGSTAMP:
3139 return do_siocgstamp(net, sock, cmd, argp);
3140 case SIOCGSTAMPNS:
3141 return do_siocgstampns(net, sock, cmd, argp);
Ben Hutchings590d4692013-11-18 17:04:13 +00003142 case SIOCBONDSLAVEINFOQUERY:
3143 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003144 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003145 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003146 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003147
Arnd Bergmann6b960182009-11-06 23:10:54 -08003148 case FIOSETOWN:
3149 case SIOCSPGRP:
3150 case FIOGETOWN:
3151 case SIOCGPGRP:
3152 case SIOCBRADDBR:
3153 case SIOCBRDELBR:
3154 case SIOCGIFVLAN:
3155 case SIOCSIFVLAN:
3156 case SIOCADDDLCI:
3157 case SIOCDELDLCI:
Andrey Vaginc62cce22016-10-24 18:29:13 -07003158 case SIOCGSKNS:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003159 return sock_ioctl(file, cmd, arg);
3160
3161 case SIOCGIFFLAGS:
3162 case SIOCSIFFLAGS:
3163 case SIOCGIFMETRIC:
3164 case SIOCSIFMETRIC:
3165 case SIOCGIFMTU:
3166 case SIOCSIFMTU:
3167 case SIOCGIFMEM:
3168 case SIOCSIFMEM:
3169 case SIOCGIFHWADDR:
3170 case SIOCSIFHWADDR:
3171 case SIOCADDMULTI:
3172 case SIOCDELMULTI:
3173 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003174 case SIOCGIFADDR:
3175 case SIOCSIFADDR:
3176 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003177 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003178 case SIOCGIFBRDADDR:
3179 case SIOCSIFBRDADDR:
3180 case SIOCGIFDSTADDR:
3181 case SIOCSIFDSTADDR:
3182 case SIOCGIFNETMASK:
3183 case SIOCSIFNETMASK:
3184 case SIOCSIFPFLAGS:
3185 case SIOCGIFPFLAGS:
3186 case SIOCGIFTXQLEN:
3187 case SIOCSIFTXQLEN:
3188 case SIOCBRADDIF:
3189 case SIOCBRDELIF:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003190 case SIOCSIFNAME:
3191 case SIOCGMIIPHY:
3192 case SIOCGMIIREG:
3193 case SIOCSMIIREG:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003194 return dev_ifsioc(net, sock, cmd, argp);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003195
Arnd Bergmann6b960182009-11-06 23:10:54 -08003196 case SIOCSARP:
3197 case SIOCGARP:
3198 case SIOCDARP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003199 case SIOCATMARK:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003200 return sock_do_ioctl(net, sock, cmd, arg);
3201 }
3202
Arnd Bergmann6b960182009-11-06 23:10:54 -08003203 return -ENOIOCTLCMD;
3204}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003205
Eric Dumazet95c96172012-04-15 05:58:06 +00003206static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003207 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003208{
3209 struct socket *sock = file->private_data;
3210 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003211 struct sock *sk;
3212 struct net *net;
3213
3214 sk = sock->sk;
3215 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003216
3217 if (sock->ops->compat_ioctl)
3218 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3219
David S. Miller87de87d2008-06-03 09:14:03 -07003220 if (ret == -ENOIOCTLCMD &&
3221 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3222 ret = compat_wext_handle_ioctl(net, cmd, arg);
3223
Arnd Bergmann6b960182009-11-06 23:10:54 -08003224 if (ret == -ENOIOCTLCMD)
3225 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3226
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003227 return ret;
3228}
3229#endif
3230
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003231int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3232{
3233 return sock->ops->bind(sock, addr, addrlen);
3234}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003235EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003236
3237int kernel_listen(struct socket *sock, int backlog)
3238{
3239 return sock->ops->listen(sock, backlog);
3240}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003241EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003242
3243int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3244{
3245 struct sock *sk = sock->sk;
3246 int err;
3247
3248 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3249 newsock);
3250 if (err < 0)
3251 goto done;
3252
David Howellscdfbabf2017-03-09 08:09:05 +00003253 err = sock->ops->accept(sock, *newsock, flags, true);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003254 if (err < 0) {
3255 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003256 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003257 goto done;
3258 }
3259
3260 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003261 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003262
3263done:
3264 return err;
3265}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003266EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003267
3268int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003269 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003270{
3271 return sock->ops->connect(sock, addr, addrlen, flags);
3272}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003273EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003274
3275int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3276 int *addrlen)
3277{
3278 return sock->ops->getname(sock, addr, addrlen, 0);
3279}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003280EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003281
3282int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3283 int *addrlen)
3284{
3285 return sock->ops->getname(sock, addr, addrlen, 1);
3286}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003287EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003288
3289int kernel_getsockopt(struct socket *sock, int level, int optname,
3290 char *optval, int *optlen)
3291{
3292 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003293 char __user *uoptval;
3294 int __user *uoptlen;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003295 int err;
3296
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003297 uoptval = (char __user __force *) optval;
3298 uoptlen = (int __user __force *) optlen;
3299
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003300 set_fs(KERNEL_DS);
3301 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003302 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003303 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003304 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3305 uoptlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003306 set_fs(oldfs);
3307 return err;
3308}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003309EXPORT_SYMBOL(kernel_getsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003310
3311int kernel_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07003312 char *optval, unsigned int optlen)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003313{
3314 mm_segment_t oldfs = get_fs();
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003315 char __user *uoptval;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003316 int err;
3317
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003318 uoptval = (char __user __force *) optval;
3319
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003320 set_fs(KERNEL_DS);
3321 if (level == SOL_SOCKET)
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003322 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003323 else
Namhyung Kimfb8621b2010-09-07 03:55:00 +00003324 err = sock->ops->setsockopt(sock, level, optname, uoptval,
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003325 optlen);
3326 set_fs(oldfs);
3327 return err;
3328}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003329EXPORT_SYMBOL(kernel_setsockopt);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003330
3331int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3332 size_t size, int flags)
3333{
3334 if (sock->ops->sendpage)
3335 return sock->ops->sendpage(sock, page, offset, size, flags);
3336
3337 return sock_no_sendpage(sock, page, offset, size, flags);
3338}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003339EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003340
3341int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3342{
3343 mm_segment_t oldfs = get_fs();
3344 int err;
3345
3346 set_fs(KERNEL_DS);
3347 err = sock->ops->ioctl(sock, cmd, arg);
3348 set_fs(oldfs);
3349
3350 return err;
3351}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003352EXPORT_SYMBOL(kernel_sock_ioctl);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003353
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003354int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3355{
3356 return sock->ops->shutdown(sock, how);
3357}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003358EXPORT_SYMBOL(kernel_sock_shutdown);
R. Parameswaran113c3072017-04-05 16:50:35 -07003359
3360/* This routine returns the IP overhead imposed by a socket i.e.
3361 * the length of the underlying IP header, depending on whether
3362 * this is an IPv4 or IPv6 socket and the length from IP options turned
R. Parameswaran57240d02017-04-12 18:31:04 -07003363 * on at the socket. Assumes that the caller has a lock on the socket.
R. Parameswaran113c3072017-04-05 16:50:35 -07003364 */
3365u32 kernel_sock_ip_overhead(struct sock *sk)
3366{
3367 struct inet_sock *inet;
3368 struct ip_options_rcu *opt;
3369 u32 overhead = 0;
3370 bool owned_by_user;
3371#if IS_ENABLED(CONFIG_IPV6)
3372 struct ipv6_pinfo *np;
3373 struct ipv6_txoptions *optv6 = NULL;
3374#endif /* IS_ENABLED(CONFIG_IPV6) */
3375
3376 if (!sk)
3377 return overhead;
3378
3379 owned_by_user = sock_owned_by_user(sk);
3380 switch (sk->sk_family) {
3381 case AF_INET:
3382 inet = inet_sk(sk);
3383 overhead += sizeof(struct iphdr);
3384 opt = rcu_dereference_protected(inet->inet_opt,
3385 owned_by_user);
3386 if (opt)
3387 overhead += opt->opt.optlen;
3388 return overhead;
3389#if IS_ENABLED(CONFIG_IPV6)
3390 case AF_INET6:
3391 np = inet6_sk(sk);
3392 overhead += sizeof(struct ipv6hdr);
3393 if (np)
3394 optv6 = rcu_dereference_protected(np->opt,
3395 owned_by_user);
3396 if (optv6)
3397 overhead += (optv6->opt_flen + optv6->opt_nflen);
3398 return overhead;
3399#endif /* IS_ENABLED(CONFIG_IPV6) */
3400 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3401 return overhead;
3402 }
3403}
3404EXPORT_SYMBOL(kernel_sock_ip_overhead);