blob: e44b8ac47f6f46295af11c59f2ea3451cb7e0556 [file] [log] [blame]
Thomas Gleixner2874c5f2019-05-27 08:55:01 +02001// SPDX-License-Identifier: GPL-2.0-or-later
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * NET An implementation of the SOCKET network access protocol.
4 *
5 * Version: @(#)socket.c 1.1.93 18/02/95
6 *
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
Jesper Juhl02c30a82005-05-05 16:16:16 -07008 * Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -07009 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 *
11 * Fixes:
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * shutdown()
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
18 * top level.
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * tty drivers).
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
26 * configurable.
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
35 * stuff.
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
41 * moment.
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
Stephen Hemminger89bddce2006-09-01 00:19:31 -070046 * Tigran Aivazian : Made listen(2) backlog sanity checks
Linus Torvalds1da177e2005-04-16 15:20:36 -070047 * protocol-independent
48 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070049 * This module is effectively the top level interface to the BSD socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -070050 * paradigm.
Linus Torvalds1da177e2005-04-16 15:20:36 -070051 *
52 * Based upon Swansea University Computer Society NET3.039
53 */
54
Linus Torvalds1da177e2005-04-16 15:20:36 -070055#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056#include <linux/socket.h>
57#include <linux/file.h>
58#include <linux/net.h>
59#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070060#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070061#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070062#include <linux/netdevice.h>
63#include <linux/proc_fs.h>
64#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080065#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070066#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030067#include <linux/if_frad.h>
68#include <linux/if_vlan.h>
Daniel Borkmann408eccc2014-04-01 16:20:23 +020069#include <linux/ptp_classify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070070#include <linux/init.h>
71#include <linux/poll.h>
72#include <linux/cache.h>
73#include <linux/module.h>
74#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070075#include <linux/mount.h>
David Howellsfba9be42019-03-25 16:38:24 +000076#include <linux/pseudo_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#include <linux/security.h>
78#include <linux/syscalls.h>
79#include <linux/compat.h>
80#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010081#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010082#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070083#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070084#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090085#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000086#include <linux/xattr.h>
Jeremy Clinec8e8cd52018-07-27 22:43:01 +000087#include <linux/nospec.h>
Paolo Abeni8c3c4472019-05-03 17:01:39 +020088#include <linux/indirect_call_wrapper.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070089
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080090#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070091#include <asm/unistd.h>
92
93#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070094#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070095#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070096
97#include <net/sock.h>
98#include <linux/netfilter.h>
99
Arnd Bergmann6b960182009-11-06 23:10:54 -0800100#include <linux/if_tun.h>
101#include <linux/ipv6_route.h>
102#include <linux/route.h>
Arnd Bergmannc7dc5042019-06-03 23:07:12 +0200103#include <linux/termios.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800104#include <linux/sockios.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300105#include <net/busy_poll.h>
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400106#include <linux/errqueue.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300107
Cong Wange0d10952013-08-01 11:10:25 +0800108#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300109unsigned int sysctl_net_busy_read __read_mostly;
110unsigned int sysctl_net_busy_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300111#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800112
Al Viro8ae5e0302014-11-28 19:40:50 -0500113static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
114static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700115static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
117static int sock_close(struct inode *inode, struct file *file);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -0700118static __poll_t sock_poll(struct file *file,
119 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700120static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800121#ifdef CONFIG_COMPAT
122static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700123 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800124#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126static ssize_t sock_sendpage(struct file *file, struct page *page,
127 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800128static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700129 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800130 unsigned int flags);
Arnd Bergmann542d3062020-01-08 22:44:43 +0100131
132#ifdef CONFIG_PROC_FS
133static void sock_show_fdinfo(struct seq_file *m, struct file *f)
134{
135 struct socket *sock = f->private_data;
136
137 if (sock->ops->show_fdinfo)
138 sock->ops->show_fdinfo(m, sock);
139}
140#else
141#define sock_show_fdinfo NULL
142#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144/*
145 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
146 * in the operation structures but are done directly via the socketcall() multiplexor.
147 */
148
Arjan van de Venda7071d2007-02-12 00:55:36 -0800149static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150 .owner = THIS_MODULE,
151 .llseek = no_llseek,
Al Viro8ae5e0302014-11-28 19:40:50 -0500152 .read_iter = sock_read_iter,
153 .write_iter = sock_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 .poll = sock_poll,
155 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800156#ifdef CONFIG_COMPAT
157 .compat_ioctl = compat_sock_ioctl,
158#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 .mmap = sock_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 .release = sock_close,
161 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200162 .sendpage = sock_sendpage,
163 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800164 .splice_read = sock_splice_read,
Kirill Tkhaib4653342019-12-09 13:03:40 +0300165 .show_fdinfo = sock_show_fdinfo,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166};
167
168/*
169 * The protocol list. Each protocol is registered in here.
170 */
171
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000173static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700176 * Support routines.
177 * Move socket addresses back and forth across the kernel/user
178 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700179 */
180
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181/**
182 * move_addr_to_kernel - copy a socket address into kernel space
183 * @uaddr: Address in user space
184 * @kaddr: Address in kernel space
185 * @ulen: Length in user space
186 *
187 * The address is copied into kernel space. If the provided address is
188 * too long an error code of -EINVAL is returned. If the copy gives
189 * invalid addresses -EFAULT is returned. On a success 0 is returned.
190 */
191
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000192int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700194 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700196 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700198 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100200 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700201}
202
203/**
204 * move_addr_to_user - copy an address to user space
205 * @kaddr: kernel space address
206 * @klen: length of address in kernel
207 * @uaddr: user space address
208 * @ulen: pointer to user length field
209 *
210 * The value pointed to by ulen on entry is the buffer length available.
211 * This is overwritten with the buffer space used. -EINVAL is returned
212 * if an overlong buffer is specified or a negative buffer size. -EFAULT
213 * is returned if either the buffer or the length field are not
214 * accessible.
215 * After copying the data up to the limit the user specifies, the true
216 * length of the data is written over the length limit the user
217 * specified. Zero is returned for a success.
218 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700219
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000220static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000221 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222{
223 int err;
224 int len;
225
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100226 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700227 err = get_user(len, ulen);
228 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700230 if (len > klen)
231 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100232 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700234 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500235 if (audit_sockaddr(klen, kaddr))
236 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700237 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 return -EFAULT;
239 }
240 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700241 * "fromlen shall refer to the value before truncation.."
242 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 */
244 return __put_user(klen, ulen);
245}
246
Alexey Dobriyan08009a72018-02-24 21:20:33 +0300247static struct kmem_cache *sock_inode_cachep __ro_after_init;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
249static struct inode *sock_alloc_inode(struct super_block *sb)
250{
251 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700252
Christoph Lametere94b1762006-12-06 20:33:17 -0800253 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 if (!ei)
255 return NULL;
Al Viro333f7902019-07-05 20:14:16 +0100256 init_waitqueue_head(&ei->socket.wq.wait);
257 ei->socket.wq.fasync_list = NULL;
258 ei->socket.wq.flags = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700259
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 ei->socket.state = SS_UNCONNECTED;
261 ei->socket.flags = 0;
262 ei->socket.ops = NULL;
263 ei->socket.sk = NULL;
264 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
266 return &ei->vfs_inode;
267}
268
Al Viro6d7855c2019-07-05 20:13:22 +0100269static void sock_free_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270{
Eric Dumazet43815482010-04-29 11:01:49 +0000271 struct socket_alloc *ei;
272
273 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazet43815482010-04-29 11:01:49 +0000274 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275}
276
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700277static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700279 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280
Christoph Lametera35afb82007-05-16 22:10:57 -0700281 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700283
yuan linyu1e911632017-01-07 17:18:31 +0800284static void init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285{
286 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700287 sizeof(struct socket_alloc),
288 0,
289 (SLAB_HWCACHE_ALIGN |
290 SLAB_RECLAIM_ACCOUNT |
Vladimir Davydov5d097052016-01-14 15:18:21 -0800291 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
Paul Mundt20c2df82007-07-20 10:11:58 +0900292 init_once);
yuan linyu1e911632017-01-07 17:18:31 +0800293 BUG_ON(sock_inode_cachep == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294}
295
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700296static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700297 .alloc_inode = sock_alloc_inode,
Al Viro6d7855c2019-07-05 20:13:22 +0100298 .free_inode = sock_free_inode,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700299 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300};
301
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700302/*
303 * sockfs_dname() is called from d_path().
304 */
305static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
306{
307 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000308 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700309}
310
Al Viro3ba13d12009-02-20 06:02:22 +0000311static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700312 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313};
314
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200315static int sockfs_xattr_get(const struct xattr_handler *handler,
316 struct dentry *dentry, struct inode *inode,
317 const char *suffix, void *value, size_t size)
318{
319 if (value) {
320 if (dentry->d_name.len + 1 > size)
321 return -ERANGE;
322 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
323 }
324 return dentry->d_name.len + 1;
325}
326
327#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
328#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
329#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
330
331static const struct xattr_handler sockfs_xattr_handler = {
332 .name = XATTR_NAME_SOCKPROTONAME,
333 .get = sockfs_xattr_get,
334};
335
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100336static int sockfs_security_xattr_set(const struct xattr_handler *handler,
337 struct dentry *dentry, struct inode *inode,
338 const char *suffix, const void *value,
339 size_t size, int flags)
340{
341 /* Handled by LSM. */
342 return -EAGAIN;
343}
344
345static const struct xattr_handler sockfs_security_xattr_handler = {
346 .prefix = XATTR_SECURITY_PREFIX,
347 .set = sockfs_security_xattr_set,
348};
349
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200350static const struct xattr_handler *sockfs_xattr_handlers[] = {
351 &sockfs_xattr_handler,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100352 &sockfs_security_xattr_handler,
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200353 NULL
354};
355
David Howellsfba9be42019-03-25 16:38:24 +0000356static int sockfs_init_fs_context(struct fs_context *fc)
Al Viroc74a1cb2011-01-12 16:59:34 -0500357{
David Howellsfba9be42019-03-25 16:38:24 +0000358 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
359 if (!ctx)
360 return -ENOMEM;
361 ctx->ops = &sockfs_ops;
362 ctx->dops = &sockfs_dentry_operations;
363 ctx->xattr = sockfs_xattr_handlers;
364 return 0;
Al Viroc74a1cb2011-01-12 16:59:34 -0500365}
366
367static struct vfsmount *sock_mnt __read_mostly;
368
369static struct file_system_type sock_fs_type = {
370 .name = "sockfs",
David Howellsfba9be42019-03-25 16:38:24 +0000371 .init_fs_context = sockfs_init_fs_context,
Al Viroc74a1cb2011-01-12 16:59:34 -0500372 .kill_sb = kill_anon_super,
373};
374
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375/*
376 * Obtains the first available file descriptor and sets it up for use.
377 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800378 * These functions create file structures and maps them to fd space
379 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 * and file struct implicitly stored in sock->file.
381 * Note that another thread may close file descriptor before we return
382 * from this function. We use the fact that now we do not refer
383 * to socket after mapping. If one day we will need it, this
384 * function will increment ref. count on file by 1.
385 *
386 * In any case returned fd MAY BE not valid!
387 * This race condition is unavoidable
388 * with shared fd spaces, we cannot solve it inside kernel,
389 * but we take care of internal coherence yet.
390 */
391
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300392/**
393 * sock_alloc_file - Bind a &socket to a &file
394 * @sock: socket
395 * @flags: file status flags
396 * @dname: protocol name
397 *
398 * Returns the &file bound with @sock, implicitly storing it
399 * in sock->file. If dname is %NULL, sets to "".
400 * On failure the return is a ERR pointer (see linux/err.h).
401 * This function uses GFP_KERNEL internally.
402 */
403
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700404struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405{
Al Viro7cbe66b2009-08-05 19:59:08 +0400406 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800407
Al Virod93aa9d2018-06-09 09:40:05 -0400408 if (!dname)
409 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410
Al Virod93aa9d2018-06-09 09:40:05 -0400411 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
412 O_RDWR | (flags & O_NONBLOCK),
413 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530414 if (IS_ERR(file)) {
Al Viro8e1611e2017-12-05 23:29:09 +0000415 sock_release(sock);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700416 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400417 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800418
419 sock->file = file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800420 file->private_data = sock;
Linus Torvaldsd8e464e2019-11-17 11:20:48 -0800421 stream_open(SOCK_INODE(sock), file);
Al Viro28407632012-08-17 23:54:15 -0400422 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800423}
Al Viro56b31d12012-08-18 00:25:51 -0400424EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800425
Al Viro56b31d12012-08-18 00:25:51 -0400426static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800427{
428 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400429 int fd = get_unused_fd_flags(flags);
Al Viroce4bb042018-01-10 18:47:05 -0500430 if (unlikely(fd < 0)) {
431 sock_release(sock);
Al Viro28407632012-08-17 23:54:15 -0400432 return fd;
Al Viroce4bb042018-01-10 18:47:05 -0500433 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800434
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700435 newfile = sock_alloc_file(sock, flags, NULL);
Enrico Weigelt4546e442019-06-05 22:58:50 +0200436 if (!IS_ERR(newfile)) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800437 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400438 return fd;
439 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400440
Al Viro28407632012-08-17 23:54:15 -0400441 put_unused_fd(fd);
442 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443}
444
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300445/**
446 * sock_from_file - Return the &socket bounded to @file.
447 * @file: file
448 * @err: pointer to an error code return
449 *
450 * On failure returns %NULL and assigns -ENOTSOCK to @err.
451 */
452
John Fastabend406a3c62012-07-20 10:39:25 +0000453struct socket *sock_from_file(struct file *file, int *err)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800454{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800455 if (file->f_op == &socket_file_ops)
456 return file->private_data; /* set in sock_map_fd */
457
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800458 *err = -ENOTSOCK;
459 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800460}
John Fastabend406a3c62012-07-20 10:39:25 +0000461EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700464 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 * @fd: file handle
466 * @err: pointer to an error code return
467 *
468 * The file handle passed in is locked and the socket it is bound
Rosen, Rami241c4662017-05-21 22:12:38 +0300469 * to is returned. If an error occurs the err pointer is overwritten
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 * with a negative errno code and NULL is returned. The function checks
471 * for both invalid handles and passing a handle which is not a socket.
472 *
473 * On a success the socket object pointer is returned.
474 */
475
476struct socket *sockfd_lookup(int fd, int *err)
477{
478 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 struct socket *sock;
480
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700481 file = fget(fd);
482 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 *err = -EBADF;
484 return NULL;
485 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700486
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800487 sock = sock_from_file(file, err);
488 if (!sock)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 fput(file);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 return sock;
491}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700492EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800494static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
495{
Al Viro00e188e2014-03-03 23:48:18 -0500496 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800497 struct socket *sock;
498
Hua Zhong36725582006-04-19 15:25:02 -0700499 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500500 if (f.file) {
501 sock = sock_from_file(f.file, err);
502 if (likely(sock)) {
503 *fput_needed = f.flags;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800504 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500505 }
506 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800507 }
508 return NULL;
509}
510
Masatake YAMATO600e1772012-08-29 10:44:29 +0000511static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
512 size_t size)
513{
514 ssize_t len;
515 ssize_t used = 0;
516
David Howellsc5ef6032015-03-17 22:26:16 +0000517 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000518 if (len < 0)
519 return len;
520 used += len;
521 if (buffer) {
522 if (size < used)
523 return -ERANGE;
524 buffer += len;
525 }
526
527 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
528 used += len;
529 if (buffer) {
530 if (size < used)
531 return -ERANGE;
532 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
533 buffer += len;
534 }
535
536 return used;
537}
538
Tobias Klauserdc647ec2017-01-10 09:30:51 +0100539static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900540{
541 int err = simple_setattr(dentry, iattr);
542
Eric Biggerse1a3a602016-12-30 17:42:32 -0600543 if (!err && (iattr->ia_valid & ATTR_UID)) {
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900544 struct socket *sock = SOCKET_I(d_inode(dentry));
545
Cong Wang6d8c50d2018-06-07 13:39:49 -0700546 if (sock->sk)
547 sock->sk->sk_uid = iattr->ia_uid;
548 else
549 err = -ENOENT;
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900550 }
551
552 return err;
553}
554
Masatake YAMATO600e1772012-08-29 10:44:29 +0000555static const struct inode_operations sockfs_inode_ops = {
Masatake YAMATO600e1772012-08-29 10:44:29 +0000556 .listxattr = sockfs_listxattr,
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900557 .setattr = sockfs_setattr,
Masatake YAMATO600e1772012-08-29 10:44:29 +0000558};
559
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300561 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700562 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 * Allocate a new inode and socket object. The two are bound together
564 * and initialised. The socket is then returned. If we are out of inodes
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300565 * NULL is returned. This functions uses GFP_KERNEL internally.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 */
567
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800568struct socket *sock_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700570 struct inode *inode;
571 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200573 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 if (!inode)
575 return NULL;
576
577 sock = SOCKET_I(inode);
578
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400579 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700580 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100581 inode->i_uid = current_fsuid();
582 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000583 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 return sock;
586}
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800587EXPORT_SYMBOL(sock_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
Cong Wang6d8c50d2018-06-07 13:39:49 -0700589static void __sock_release(struct socket *sock, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590{
591 if (sock->ops) {
592 struct module *owner = sock->ops->owner;
593
Cong Wang6d8c50d2018-06-07 13:39:49 -0700594 if (inode)
595 inode_lock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 sock->ops->release(sock);
Eric Biggersff7b11a2019-02-21 14:13:56 -0800597 sock->sk = NULL;
Cong Wang6d8c50d2018-06-07 13:39:49 -0700598 if (inode)
599 inode_unlock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 sock->ops = NULL;
601 module_put(owner);
602 }
603
Al Viro333f7902019-07-05 20:14:16 +0100604 if (sock->wq.fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800605 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 if (!sock->file) {
608 iput(SOCK_INODE(sock));
609 return;
610 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700611 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612}
Cong Wang6d8c50d2018-06-07 13:39:49 -0700613
Andrew Lunn9a8ad9a2020-07-13 01:15:12 +0200614/**
615 * sock_release - close a socket
616 * @sock: socket to close
617 *
618 * The socket is released from the protocol stack if it has a release
619 * callback, and the inode is then released if the socket is bound to
620 * an inode not a file.
621 */
Cong Wang6d8c50d2018-06-07 13:39:49 -0700622void sock_release(struct socket *sock)
623{
624 __sock_release(sock, NULL);
625}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700626EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400628void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000629{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200630 u8 flags = *tx_flags;
631
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400632 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200633 flags |= SKBTX_HW_TSTAMP;
634
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400635 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200636 flags |= SKBTX_SW_TSTAMP;
637
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400638 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200639 flags |= SKBTX_SCHED_TSTAMP;
640
Eric Dumazet140c55d2014-08-06 11:49:29 +0200641 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000642}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400643EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000644
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200645INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
646 size_t));
Paolo Abenia648a592019-07-03 16:06:54 +0200647INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
648 size_t));
Al Virod8725c82014-12-11 00:02:50 -0500649static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650{
Paolo Abenia648a592019-07-03 16:06:54 +0200651 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
652 inet_sendmsg, sock, msg,
653 msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500654 BUG_ON(ret == -EIOCBQUEUED);
655 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800657
Randy Dunlap85806af2019-05-18 21:23:07 -0700658/**
659 * sock_sendmsg - send a message through @sock
660 * @sock: socket
661 * @msg: message to send
662 *
663 * Sends @msg through @sock, passing through LSM.
664 * Returns the number of bytes sent, or an error code.
665 */
Al Virod8725c82014-12-11 00:02:50 -0500666int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800667{
Al Virod8725c82014-12-11 00:02:50 -0500668 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500669 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800670
Al Virod8725c82014-12-11 00:02:50 -0500671 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800672}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700673EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300675/**
676 * kernel_sendmsg - send a message through @sock (kernel-space)
677 * @sock: socket
678 * @msg: message header
679 * @vec: kernel vec
680 * @num: vec array length
681 * @size: total message data size
682 *
683 * Builds the message data with @vec and sends it through @sock.
684 * Returns the number of bytes sent, or an error code.
685 */
686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
688 struct kvec *vec, size_t num, size_t size)
689{
David Howellsaa563d72018-10-20 00:57:56 +0100690 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500691 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700693EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300695/**
696 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
697 * @sk: sock
698 * @msg: message header
699 * @vec: output s/g array
700 * @num: output s/g array length
701 * @size: total message data size
702 *
703 * Builds the message data with @vec and sends it through @sock.
704 * Returns the number of bytes sent, or an error code.
705 * Caller must hold @sk.
706 */
707
Tom Herbert306b13e2017-07-28 16:22:41 -0700708int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
709 struct kvec *vec, size_t num, size_t size)
710{
711 struct socket *sock = sk->sk_socket;
712
713 if (!sock->ops->sendmsg_locked)
John Fastabenddb5980d2017-08-15 22:31:34 -0700714 return sock_no_sendmsg_locked(sk, msg, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700715
David Howellsaa563d72018-10-20 00:57:56 +0100716 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700717
718 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
719}
720EXPORT_SYMBOL(kernel_sendmsg_locked);
721
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400722static bool skb_is_err_queue(const struct sk_buff *skb)
723{
724 /* pkt_type of skbs enqueued on the error queue are set to
725 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
726 * in recvmsg, since skbs received on a local socket will never
727 * have a pkt_type of PACKET_OUTGOING.
728 */
729 return skb->pkt_type == PACKET_OUTGOING;
730}
731
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200732/* On transmit, software and hardware timestamps are returned independently.
733 * As the two skb clones share the hardware timestamp, which may be updated
734 * before the software timestamp is received, a hardware TX timestamp may be
735 * returned only if there is no software TX timestamp. Ignore false software
736 * timestamps, which may be made in the __sock_recv_timestamp() call when the
Deepa Dinamani7f1bc6e2019-02-02 07:34:46 -0800737 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200738 * hardware timestamp.
739 */
740static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
741{
742 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
743}
744
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200745static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
746{
747 struct scm_ts_pktinfo ts_pktinfo;
748 struct net_device *orig_dev;
749
750 if (!skb_mac_header_was_set(skb))
751 return;
752
753 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
754
755 rcu_read_lock();
756 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
757 if (orig_dev)
758 ts_pktinfo.if_index = orig_dev->ifindex;
759 rcu_read_unlock();
760
761 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
762 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
763 sizeof(ts_pktinfo), &ts_pktinfo);
764}
765
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700766/*
767 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
768 */
769void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
770 struct sk_buff *skb)
771{
Patrick Ohly20d49472009-02-12 05:03:38 +0000772 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Deepa Dinamani887feae2019-02-02 07:34:50 -0800773 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
Deepa Dinamani97184752019-02-02 07:34:51 -0800774 struct scm_timestamping_internal tss;
775
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200776 int empty = 1, false_tstamp = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000777 struct skb_shared_hwtstamps *shhwtstamps =
778 skb_hwtstamps(skb);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700779
Patrick Ohly20d49472009-02-12 05:03:38 +0000780 /* Race occurred between timestamp enabling and packet
781 receiving. Fill in the current time for now. */
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200782 if (need_software_tstamp && skb->tstamp == 0) {
Patrick Ohly20d49472009-02-12 05:03:38 +0000783 __net_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200784 false_tstamp = 1;
785 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000786
787 if (need_software_tstamp) {
788 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800789 if (new_tstamp) {
790 struct __kernel_sock_timeval tv;
791
792 skb_get_new_timestamp(skb, &tv);
793 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
794 sizeof(tv), &tv);
795 } else {
796 struct __kernel_old_timeval tv;
797
798 skb_get_timestamp(skb, &tv);
799 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
800 sizeof(tv), &tv);
801 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000802 } else {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800803 if (new_tstamp) {
804 struct __kernel_timespec ts;
805
806 skb_get_new_timestampns(skb, &ts);
807 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
808 sizeof(ts), &ts);
809 } else {
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +0200810 struct __kernel_old_timespec ts;
Deepa Dinamani887feae2019-02-02 07:34:50 -0800811
812 skb_get_timestampns(skb, &ts);
813 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
814 sizeof(ts), &ts);
815 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000816 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700817 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000818
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400819 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400820 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Deepa Dinamani97184752019-02-02 07:34:51 -0800821 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000822 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400823 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400824 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200825 !skb_is_swtx_tstamp(skb, false_tstamp) &&
Deepa Dinamani97184752019-02-02 07:34:51 -0800826 ktime_to_timespec64_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400827 empty = 0;
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200828 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
829 !skb_is_err_queue(skb))
830 put_ts_pktinfo(msg, skb);
831 }
Francis Yan1c885802016-11-27 23:07:18 -0800832 if (!empty) {
Deepa Dinamani97184752019-02-02 07:34:51 -0800833 if (sock_flag(sk, SOCK_TSTAMP_NEW))
834 put_cmsg_scm_timestamping64(msg, &tss);
835 else
836 put_cmsg_scm_timestamping(msg, &tss);
Francis Yan1c885802016-11-27 23:07:18 -0800837
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400838 if (skb_is_err_queue(skb) && skb->len &&
Soheil Hassas Yeganeh4ef1b282017-03-18 17:03:00 -0400839 SKB_EXT_ERR(skb)->opt_stats)
Francis Yan1c885802016-11-27 23:07:18 -0800840 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
841 skb->len, skb->data);
842 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700843}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300844EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
845
Johannes Berg6e3e9392011-11-09 10:15:42 +0100846void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
847 struct sk_buff *skb)
848{
849 int ack;
850
851 if (!sock_flag(sk, SOCK_WIFI_STATUS))
852 return;
853 if (!skb->wifi_acked_valid)
854 return;
855
856 ack = skb->wifi_acked;
857
858 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
859}
860EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
861
stephen hemminger11165f12010-10-18 14:27:29 +0000862static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
863 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700864{
Eyal Birger744d5a32015-03-01 14:58:31 +0200865 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700866 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200867 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700868}
869
Eric Dumazet767dd032010-04-28 19:14:43 +0000870void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700871 struct sk_buff *skb)
872{
873 sock_recv_timestamp(msg, sk, skb);
874 sock_recv_drops(msg, sk, skb);
875}
Eric Dumazet767dd032010-04-28 19:14:43 +0000876EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700877
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200878INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
Paolo Abenia648a592019-07-03 16:06:54 +0200879 size_t, int));
880INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
881 size_t, int));
Ying Xue1b784142015-03-02 15:37:48 +0800882static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -0400883 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884{
Paolo Abenia648a592019-07-03 16:06:54 +0200885 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
886 inet_recvmsg, sock, msg, msg_data_left(msg),
887 flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888}
889
Randy Dunlap85806af2019-05-18 21:23:07 -0700890/**
891 * sock_recvmsg - receive a message from @sock
892 * @sock: socket
893 * @msg: message to receive
894 * @flags: message flags
895 *
896 * Receives @msg from @sock, passing through LSM. Returns the total number
897 * of bytes received, or an error.
898 */
Al Viro2da62902015-03-14 21:13:46 -0400899int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700900{
Al Viro2da62902015-03-14 21:13:46 -0400901 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700902
Al Viro2da62902015-03-14 21:13:46 -0400903 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700905EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000907/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300908 * kernel_recvmsg - Receive a message from a socket (kernel space)
909 * @sock: The socket to receive the message from
910 * @msg: Received message
911 * @vec: Input s/g array for message data
912 * @num: Size of input s/g array
913 * @size: Number of bytes to read
914 * @flags: Message flags (MSG_DONTWAIT, etc...)
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000915 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300916 * On return the msg structure contains the scatter/gather array passed in the
917 * vec argument. The array is modified so that it consists of the unfilled
918 * portion of the original array.
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000919 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300920 * The returned value is the total number of bytes received, or an error.
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000921 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300922
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700923int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
924 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925{
Christoph Hellwig1f466e12020-05-11 13:59:13 +0200926 msg->msg_control_is_user = false;
David Howellsaa563d72018-10-20 00:57:56 +0100927 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
Christoph Hellwig1f466e12020-05-11 13:59:13 +0200928 return sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700930EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300932static ssize_t sock_sendpage(struct file *file, struct page *page,
933 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934{
935 struct socket *sock;
936 int flags;
937
Eric Dumazetb69aee02005-09-06 14:42:45 -0700938 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939
Eric Dumazet35f9c092012-04-05 03:05:35 +0000940 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
941 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
942 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943
Linus Torvaldse6949582009-08-13 08:28:36 -0700944 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945}
946
Jens Axboe9c55e012007-11-06 23:30:13 -0800947static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700948 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800949 unsigned int flags)
950{
951 struct socket *sock = file->private_data;
952
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800953 if (unlikely(!sock->ops->splice_read))
Slavomir Kaslev95506582018-11-16 11:27:53 +0200954 return generic_file_splice_read(file, ppos, pipe, len, flags);
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -0800955
Jens Axboe9c55e012007-11-06 23:30:13 -0800956 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
957}
958
Al Viro8ae5e0302014-11-28 19:40:50 -0500959static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800960{
Al Viro6d652332015-01-30 16:12:56 -0500961 struct file *file = iocb->ki_filp;
962 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700963 struct msghdr msg = {.msg_iter = *to,
964 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -0500965 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800966
Jens Axboeebfcd892019-12-09 20:58:56 -0700967 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -0500968 msg.msg_flags = MSG_DONTWAIT;
969
970 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800971 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700972
Christoph Hellwig66ee59a2015-02-11 19:56:46 +0100973 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800974 return 0;
975
Al Viro2da62902015-03-14 21:13:46 -0400976 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e0302014-11-28 19:40:50 -0500977 *to = msg.msg_iter;
978 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800979}
980
Al Viro8ae5e0302014-11-28 19:40:50 -0500981static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982{
Al Viro6d652332015-01-30 16:12:56 -0500983 struct file *file = iocb->ki_filp;
984 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -0700985 struct msghdr msg = {.msg_iter = *from,
986 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -0500987 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800988
Al Viro8ae5e0302014-11-28 19:40:50 -0500989 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -0800990 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -0700991
Jens Axboeebfcd892019-12-09 20:58:56 -0700992 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -0500993 msg.msg_flags = MSG_DONTWAIT;
994
Al Viro6d652332015-01-30 16:12:56 -0500995 if (sock->type == SOCK_SEQPACKET)
996 msg.msg_flags |= MSG_EOR;
997
Al Virod8725c82014-12-11 00:02:50 -0500998 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e0302014-11-28 19:40:50 -0500999 *from = msg.msg_iter;
1000 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001}
1002
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003/*
1004 * Atomic setting of ioctl hooks to avoid race
1005 * with module unload.
1006 */
1007
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001008static DEFINE_MUTEX(br_ioctl_mutex);
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001009static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Eric W. Biederman881d9662007-09-17 11:56:21 -07001011void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001013 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001015 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016}
1017EXPORT_SYMBOL(brioctl_set);
1018
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001019static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001020static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
Eric W. Biederman881d9662007-09-17 11:56:21 -07001022void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001024 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001026 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027}
1028EXPORT_SYMBOL(vlan_ioctl_set);
1029
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001030static DEFINE_MUTEX(dlci_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001031static int (*dlci_ioctl_hook) (unsigned int, void __user *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001033void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001035 mutex_lock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 dlci_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001037 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038}
1039EXPORT_SYMBOL(dlci_ioctl_set);
1040
Arnd Bergmann6b960182009-11-06 23:10:54 -08001041static long sock_do_ioctl(struct net *net, struct socket *sock,
Johannes Berg63ff03a2019-01-25 22:43:17 +01001042 unsigned int cmd, unsigned long arg)
Arnd Bergmann6b960182009-11-06 23:10:54 -08001043{
1044 int err;
1045 void __user *argp = (void __user *)arg;
1046
1047 err = sock->ops->ioctl(sock, cmd, arg);
1048
1049 /*
1050 * If this ioctl is unknown try to hand it down
1051 * to the NIC driver.
1052 */
Al Viro36fd6332017-06-26 13:19:16 -04001053 if (err != -ENOIOCTLCMD)
1054 return err;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001055
Al Viro36fd6332017-06-26 13:19:16 -04001056 if (cmd == SIOCGIFCONF) {
1057 struct ifconf ifc;
1058 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
1059 return -EFAULT;
1060 rtnl_lock();
1061 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
1062 rtnl_unlock();
1063 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
1064 err = -EFAULT;
Al Viro44c02a22017-10-05 12:59:44 -04001065 } else {
1066 struct ifreq ifr;
1067 bool need_copyout;
Johannes Berg63ff03a2019-01-25 22:43:17 +01001068 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
Al Viro44c02a22017-10-05 12:59:44 -04001069 return -EFAULT;
1070 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1071 if (!err && need_copyout)
Johannes Berg63ff03a2019-01-25 22:43:17 +01001072 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
Al Viro44c02a22017-10-05 12:59:44 -04001073 return -EFAULT;
Al Viro36fd6332017-06-26 13:19:16 -04001074 }
Arnd Bergmann6b960182009-11-06 23:10:54 -08001075 return err;
1076}
1077
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078/*
1079 * With an ioctl, arg may well be a user mode pointer, but we don't know
1080 * what to do with it - that's up to the protocol still.
1081 */
1082
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001083/**
1084 * get_net_ns - increment the refcount of the network namespace
1085 * @ns: common namespace (net)
1086 *
1087 * Returns the net's common namespace.
1088 */
1089
Kirill Tkhaid8d211a2018-02-14 16:39:56 +03001090struct ns_common *get_net_ns(struct ns_common *ns)
Andrey Vaginc62cce22016-10-24 18:29:13 -07001091{
1092 return &get_net(container_of(ns, struct net, ns))->ns;
1093}
Kirill Tkhaid8d211a2018-02-14 16:39:56 +03001094EXPORT_SYMBOL_GPL(get_net_ns);
Andrey Vaginc62cce22016-10-24 18:29:13 -07001095
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1097{
1098 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001099 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 void __user *argp = (void __user *)arg;
1101 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001102 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103
Eric Dumazetb69aee02005-09-06 14:42:45 -07001104 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001105 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001106 net = sock_net(sk);
Al Viro44c02a22017-10-05 12:59:44 -04001107 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1108 struct ifreq ifr;
1109 bool need_copyout;
1110 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1111 return -EFAULT;
1112 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1113 if (!err && need_copyout)
1114 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1115 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001117#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Al Virob1b0c242017-10-01 20:13:08 -04001119 err = wext_handle_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001121#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001122 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123 case FIOSETOWN:
1124 case SIOCSPGRP:
1125 err = -EFAULT;
1126 if (get_user(pid, (int __user *)argp))
1127 break;
Jiri Slaby393cc3f2017-06-13 13:35:50 +02001128 err = f_setown(sock->file, pid, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 break;
1130 case FIOGETOWN:
1131 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001132 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001133 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 break;
1135 case SIOCGIFBR:
1136 case SIOCSIFBR:
1137 case SIOCBRADDBR:
1138 case SIOCBRDELBR:
1139 err = -ENOPKG;
1140 if (!br_ioctl_hook)
1141 request_module("bridge");
1142
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001143 mutex_lock(&br_ioctl_mutex);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001144 if (br_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001145 err = br_ioctl_hook(net, cmd, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001146 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 break;
1148 case SIOCGIFVLAN:
1149 case SIOCSIFVLAN:
1150 err = -ENOPKG;
1151 if (!vlan_ioctl_hook)
1152 request_module("8021q");
1153
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001154 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001156 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001157 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 case SIOCADDDLCI:
1160 case SIOCDELDLCI:
1161 err = -ENOPKG;
1162 if (!dlci_ioctl_hook)
1163 request_module("dlci");
1164
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001165 mutex_lock(&dlci_ioctl_mutex);
1166 if (dlci_ioctl_hook)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167 err = dlci_ioctl_hook(cmd, argp);
Pavel Emelyanov7512cbf2008-03-21 15:58:52 -07001168 mutex_unlock(&dlci_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 break;
Andrey Vaginc62cce22016-10-24 18:29:13 -07001170 case SIOCGSKNS:
1171 err = -EPERM;
1172 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1173 break;
1174
1175 err = open_related_ns(&net->ns, get_net_ns);
1176 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001177 case SIOCGSTAMP_OLD:
1178 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001179 if (!sock->ops->gettstamp) {
1180 err = -ENOIOCTLCMD;
1181 break;
1182 }
1183 err = sock->ops->gettstamp(sock, argp,
Arnd Bergmann0768e172019-04-17 22:56:11 +02001184 cmd == SIOCGSTAMP_OLD,
1185 !IS_ENABLED(CONFIG_64BIT));
Gustavo A. R. Silva60747822019-04-24 10:31:24 -05001186 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001187 case SIOCGSTAMP_NEW:
1188 case SIOCGSTAMPNS_NEW:
1189 if (!sock->ops->gettstamp) {
1190 err = -ENOIOCTLCMD;
1191 break;
1192 }
1193 err = sock->ops->gettstamp(sock, argp,
1194 cmd == SIOCGSTAMP_NEW,
1195 false);
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001196 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 default:
Johannes Berg63ff03a2019-01-25 22:43:17 +01001198 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001200 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 return err;
1202}
1203
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001204/**
1205 * sock_create_lite - creates a socket
1206 * @family: protocol family (AF_INET, ...)
1207 * @type: communication type (SOCK_STREAM, ...)
1208 * @protocol: protocol (0, ...)
1209 * @res: new socket
1210 *
1211 * Creates a new socket and assigns it to @res, passing through LSM.
1212 * The new socket initialization is not complete, see kernel_accept().
1213 * Returns 0 or an error. On failure @res is set to %NULL.
1214 * This function internally uses GFP_KERNEL.
1215 */
1216
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217int sock_create_lite(int family, int type, int protocol, struct socket **res)
1218{
1219 int err;
1220 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001221
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 err = security_socket_create(family, type, protocol, 1);
1223 if (err)
1224 goto out;
1225
1226 sock = sock_alloc();
1227 if (!sock) {
1228 err = -ENOMEM;
1229 goto out;
1230 }
1231
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001233 err = security_socket_post_create(sock, family, type, protocol, 1);
1234 if (err)
1235 goto out_release;
1236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237out:
1238 *res = sock;
1239 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001240out_release:
1241 sock_release(sock);
1242 sock = NULL;
1243 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001245EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
1247/* No kernel lock held - perfect */
Al Viroade994f2017-07-03 00:01:49 -04001248static __poll_t sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249{
Christoph Hellwig3cafb372018-01-09 16:07:34 +01001250 struct socket *sock = file->private_data;
Christoph Hellwiga331de32018-07-30 09:42:13 +02001251 __poll_t events = poll_requested_events(wait), flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252
Christoph Hellwige88958e2018-06-29 15:37:24 +02001253 if (!sock->ops->poll)
1254 return 0;
Christoph Hellwigf641f13b2018-07-30 09:42:12 +02001255
Christoph Hellwiga331de32018-07-30 09:42:13 +02001256 if (sk_can_busy_loop(sock->sk)) {
1257 /* poll once if requested by the syscall */
1258 if (events & POLL_BUSY_LOOP)
1259 sk_busy_loop(sock->sk, 1);
1260
1261 /* if this socket can poll_ll, tell the system call */
1262 flag = POLL_BUSY_LOOP;
1263 }
1264
1265 return sock->ops->poll(file, sock, wait) | flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266}
1267
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001268static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001270 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271
1272 return sock->ops->mmap(file, sock, vma);
1273}
1274
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001275static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276{
Cong Wang6d8c50d2018-06-07 13:39:49 -07001277 __sock_release(SOCKET_I(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 return 0;
1279}
1280
1281/*
1282 * Update the socket async list
1283 *
1284 * Fasync_list locking strategy.
1285 *
1286 * 1. fasync_list is modified only under process context socket lock
1287 * i.e. under semaphore.
1288 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001289 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 */
1291
1292static int sock_fasync(int fd, struct file *filp, int on)
1293{
Eric Dumazet989a2972010-04-14 09:55:35 +00001294 struct socket *sock = filp->private_data;
1295 struct sock *sk = sock->sk;
Al Viro333f7902019-07-05 20:14:16 +01001296 struct socket_wq *wq = &sock->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
Eric Dumazet989a2972010-04-14 09:55:35 +00001298 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
1301 lock_sock(sk);
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001302 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001304 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001305 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001306 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001307 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
Eric Dumazet989a2972010-04-14 09:55:35 +00001309 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310 return 0;
1311}
1312
Eric Dumazetceb5d582015-11-29 20:03:11 -08001313/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314
Eric Dumazetceb5d582015-11-29 20:03:11 -08001315int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001317 if (!wq || !wq->fasync_list)
1318 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001319
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001320 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001321 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001322 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 break;
1324 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001325 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001326 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 break;
1328 /* fall through */
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001329 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001330call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001331 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001333 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001334 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001336
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 return 0;
1338}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001339EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001341/**
1342 * __sock_create - creates a socket
1343 * @net: net namespace
1344 * @family: protocol family (AF_INET, ...)
1345 * @type: communication type (SOCK_STREAM, ...)
1346 * @protocol: protocol (0, ...)
1347 * @res: new socket
1348 * @kern: boolean for kernel space sockets
1349 *
1350 * Creates a new socket and assigns it to @res, passing through LSM.
1351 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1352 * be set to true if the socket resides in kernel space.
1353 * This function internally uses GFP_KERNEL.
1354 */
1355
Pavel Emelyanov721db932010-09-29 16:06:32 +04001356int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001357 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358{
1359 int err;
1360 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001361 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362
1363 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001364 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 */
1366 if (family < 0 || family >= NPROTO)
1367 return -EAFNOSUPPORT;
1368 if (type < 0 || type >= SOCK_MAX)
1369 return -EINVAL;
1370
1371 /* Compatibility.
1372
1373 This uglymoron is moved from INET layer to here to avoid
1374 deadlock in module load.
1375 */
1376 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001377 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1378 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 family = PF_PACKET;
1380 }
1381
1382 err = security_socket_create(family, type, protocol, kern);
1383 if (err)
1384 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001385
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001386 /*
1387 * Allocate the socket and allow the family to set things up. if
1388 * the protocol is 0, the family is instructed to select an appropriate
1389 * default.
1390 */
1391 sock = sock_alloc();
1392 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001393 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001394 return -ENFILE; /* Not exactly a match, but its the
1395 closest posix thing */
1396 }
1397
1398 sock->type = type;
1399
Johannes Berg95a5afc2008-10-16 15:24:51 -07001400#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001401 /* Attempt to load a protocol module if the find failed.
1402 *
1403 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 * requested real, full-featured networking support upon configuration.
1405 * Otherwise module support will break!
1406 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001407 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001408 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409#endif
1410
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001411 rcu_read_lock();
1412 pf = rcu_dereference(net_families[family]);
1413 err = -EAFNOSUPPORT;
1414 if (!pf)
1415 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416
1417 /*
1418 * We will call the ->create function, that possibly is in a loadable
1419 * module, so we have to bump that loadable module refcnt first.
1420 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001421 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 goto out_release;
1423
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001424 /* Now protected by module ref count */
1425 rcu_read_unlock();
1426
Eric Paris3f378b62009-11-05 22:18:14 -08001427 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001428 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001430
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 /*
1432 * Now to bump the refcnt of the [loadable] module that owns this
1433 * socket at sock_release time we decrement its refcnt.
1434 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001435 if (!try_module_get(sock->ops->owner))
1436 goto out_module_busy;
1437
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 /*
1439 * Now that we're done with the ->create function, the [loadable]
1440 * module can have its refcnt decremented
1441 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001442 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001443 err = security_socket_post_create(sock, family, type, protocol, kern);
1444 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001445 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001446 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001448 return 0;
1449
1450out_module_busy:
1451 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001453 sock->ops = NULL;
1454 module_put(pf->owner);
1455out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001457 return err;
1458
1459out_release:
1460 rcu_read_unlock();
1461 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001463EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001465/**
1466 * sock_create - creates a socket
1467 * @family: protocol family (AF_INET, ...)
1468 * @type: communication type (SOCK_STREAM, ...)
1469 * @protocol: protocol (0, ...)
1470 * @res: new socket
1471 *
1472 * A wrapper around __sock_create().
1473 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1474 */
1475
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476int sock_create(int family, int type, int protocol, struct socket **res)
1477{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001478 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001480EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001482/**
1483 * sock_create_kern - creates a socket (kernel space)
1484 * @net: net namespace
1485 * @family: protocol family (AF_INET, ...)
1486 * @type: communication type (SOCK_STREAM, ...)
1487 * @protocol: protocol (0, ...)
1488 * @res: new socket
1489 *
1490 * A wrapper around __sock_create().
1491 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1492 */
1493
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001494int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001496 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001498EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001500int __sys_socket(int family, int type, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501{
1502 int retval;
1503 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001504 int flags;
1505
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001506 /* Check the SOCK_* constants for consistency. */
1507 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1508 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1509 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1510 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1511
Ulrich Dreppera677a032008-07-23 21:29:17 -07001512 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001513 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001514 return -EINVAL;
1515 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001517 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1518 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1519
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 retval = sock_create(family, type, protocol, &sock);
1521 if (retval < 0)
Al Viro8e1611e2017-12-05 23:29:09 +00001522 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523
Al Viro8e1611e2017-12-05 23:29:09 +00001524 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525}
1526
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001527SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1528{
1529 return __sys_socket(family, type, protocol);
1530}
1531
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532/*
1533 * Create a pair of connected sockets.
1534 */
1535
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001536int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537{
1538 struct socket *sock1, *sock2;
1539 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001540 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001541 int flags;
1542
1543 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001544 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001545 return -EINVAL;
1546 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001548 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1549 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1550
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 /*
Al Viro016a2662017-12-05 23:28:38 +00001552 * reserve descriptors and make sure we won't fail
1553 * to return them to userland.
1554 */
1555 fd1 = get_unused_fd_flags(flags);
1556 if (unlikely(fd1 < 0))
1557 return fd1;
1558
1559 fd2 = get_unused_fd_flags(flags);
1560 if (unlikely(fd2 < 0)) {
1561 put_unused_fd(fd1);
1562 return fd2;
1563 }
1564
1565 err = put_user(fd1, &usockvec[0]);
1566 if (err)
1567 goto out;
1568
1569 err = put_user(fd2, &usockvec[1]);
1570 if (err)
1571 goto out;
1572
1573 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 * Obtain the first socket and check if the underlying protocol
1575 * supports the socketpair call.
1576 */
1577
1578 err = sock_create(family, type, protocol, &sock1);
Al Viro016a2662017-12-05 23:28:38 +00001579 if (unlikely(err < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 goto out;
1581
1582 err = sock_create(family, type, protocol, &sock2);
Al Viro016a2662017-12-05 23:28:38 +00001583 if (unlikely(err < 0)) {
1584 sock_release(sock1);
1585 goto out;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001586 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001587
David Herrmannd47cd942018-05-04 16:28:20 +02001588 err = security_socket_socketpair(sock1, sock2);
1589 if (unlikely(err)) {
1590 sock_release(sock2);
1591 sock_release(sock1);
1592 goto out;
1593 }
1594
Al Viro016a2662017-12-05 23:28:38 +00001595 err = sock1->ops->socketpair(sock1, sock2);
1596 if (unlikely(err < 0)) {
1597 sock_release(sock2);
1598 sock_release(sock1);
1599 goto out;
Al Viro28407632012-08-17 23:54:15 -04001600 }
1601
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001602 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301603 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001604 err = PTR_ERR(newfile1);
Al Viro016a2662017-12-05 23:28:38 +00001605 sock_release(sock2);
1606 goto out;
Al Viro28407632012-08-17 23:54:15 -04001607 }
1608
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001609 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001610 if (IS_ERR(newfile2)) {
1611 err = PTR_ERR(newfile2);
Al Viro016a2662017-12-05 23:28:38 +00001612 fput(newfile1);
1613 goto out;
Al Virodb349502007-02-07 01:48:00 -05001614 }
1615
Al Viro157cf642008-12-14 04:57:47 -05001616 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001617
Al Virodb349502007-02-07 01:48:00 -05001618 fd_install(fd1, newfile1);
1619 fd_install(fd2, newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001620 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622out:
Al Viro016a2662017-12-05 23:28:38 +00001623 put_unused_fd(fd2);
1624 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 return err;
1626}
1627
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001628SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1629 int __user *, usockvec)
1630{
1631 return __sys_socketpair(family, type, protocol, usockvec);
1632}
1633
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634/*
1635 * Bind a name to a socket. Nothing much to do here since it's
1636 * the protocol's responsibility to handle the local address.
1637 *
1638 * We move the socket address to kernel space before we call
1639 * the protocol layer (having also checked the address is ok).
1640 */
1641
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001642int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643{
1644 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001645 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001646 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001648 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001649 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001650 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Jakub Sitnicki068b88c2018-10-04 11:09:40 +02001651 if (!err) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001652 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001653 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001654 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001655 if (!err)
1656 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001657 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001658 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001660 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001661 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 return err;
1663}
1664
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001665SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1666{
1667 return __sys_bind(fd, umyaddr, addrlen);
1668}
1669
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670/*
1671 * Perform a listen. Basically, we allow the protocol to do anything
1672 * necessary for a listen, and if that works, we mark the socket as
1673 * ready for listening.
1674 */
1675
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001676int __sys_listen(int fd, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677{
1678 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001679 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001680 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001681
1682 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1683 if (sock) {
Pavel Emelyanov8efa6e92008-03-31 19:41:14 -07001684 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
Eric Dumazet95c96172012-04-15 05:58:06 +00001685 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001686 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
1688 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001689 if (!err)
1690 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001692 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 }
1694 return err;
1695}
1696
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001697SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1698{
1699 return __sys_listen(fd, backlog);
1700}
1701
Jens Axboede2ea4b2019-10-17 14:41:29 -06001702int __sys_accept4_file(struct file *file, unsigned file_flags,
1703 struct sockaddr __user *upeer_sockaddr,
Jens Axboe09952e32020-03-19 20:16:56 -06001704 int __user *upeer_addrlen, int flags,
1705 unsigned long nofile)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706{
1707 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001708 struct file *newfile;
Jens Axboede2ea4b2019-10-17 14:41:29 -06001709 int err, len, newfd;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001710 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711
Ulrich Drepper77d27202008-07-23 21:29:35 -07001712 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001713 return -EINVAL;
1714
1715 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1716 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1717
Jens Axboede2ea4b2019-10-17 14:41:29 -06001718 sock = sock_from_file(file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 if (!sock)
1720 goto out;
1721
1722 err = -ENFILE;
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001723 newsock = sock_alloc();
1724 if (!newsock)
Jens Axboede2ea4b2019-10-17 14:41:29 -06001725 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
1727 newsock->type = sock->type;
1728 newsock->ops = sock->ops;
1729
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 /*
1731 * We don't need try_module_get here, as the listening socket (sock)
1732 * has the protocol module (sock->ops->owner) held.
1733 */
1734 __module_get(newsock->ops->owner);
1735
Jens Axboe09952e32020-03-19 20:16:56 -06001736 newfd = __get_unused_fd_flags(flags, nofile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001737 if (unlikely(newfd < 0)) {
1738 err = newfd;
David S. Miller9a1875e2006-04-01 12:48:36 -08001739 sock_release(newsock);
Jens Axboede2ea4b2019-10-17 14:41:29 -06001740 goto out;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001741 }
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001742 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301743 if (IS_ERR(newfile)) {
Al Viro28407632012-08-17 23:54:15 -04001744 err = PTR_ERR(newfile);
1745 put_unused_fd(newfd);
Jens Axboede2ea4b2019-10-17 14:41:29 -06001746 goto out;
Al Viro28407632012-08-17 23:54:15 -04001747 }
David S. Miller39d8c1b2006-03-20 17:13:49 -08001748
Frank Filza79af592005-09-27 15:23:38 -07001749 err = security_socket_accept(sock, newsock);
1750 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001751 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001752
Jens Axboede2ea4b2019-10-17 14:41:29 -06001753 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1754 false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001756 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
1758 if (upeer_sockaddr) {
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001759 len = newsock->ops->getname(newsock,
1760 (struct sockaddr *)&address, 2);
1761 if (len < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001763 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001765 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001766 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001768 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 }
1770
1771 /* File flags are not inherited via accept() unlike another OSes. */
1772
David S. Miller39d8c1b2006-03-20 17:13:49 -08001773 fd_install(newfd, newfile);
1774 err = newfd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775out:
1776 return err;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001777out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001778 fput(newfile);
David S. Miller39d8c1b2006-03-20 17:13:49 -08001779 put_unused_fd(newfd);
Jens Axboede2ea4b2019-10-17 14:41:29 -06001780 goto out;
1781
1782}
1783
1784/*
1785 * For accept, we attempt to create a new socket, set up the link
1786 * with the client, wake up the client, then return the new
1787 * connected fd. We collect the address of the connector in kernel
1788 * space and move it to user at the very end. This is unclean because
1789 * we open the socket then return an error.
1790 *
1791 * 1003.1g adds the ability to recvmsg() to query connection pending
1792 * status to recvmsg. We need to add that support in a way thats
1793 * clean when we restructure accept also.
1794 */
1795
1796int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1797 int __user *upeer_addrlen, int flags)
1798{
1799 int ret = -EBADF;
1800 struct fd f;
1801
1802 f = fdget(fd);
1803 if (f.file) {
1804 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
Jens Axboe09952e32020-03-19 20:16:56 -06001805 upeer_addrlen, flags,
1806 rlimit(RLIMIT_NOFILE));
Jens Axboede2ea4b2019-10-17 14:41:29 -06001807 if (f.flags)
1808 fput(f.file);
1809 }
1810
1811 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812}
1813
Dominik Brodowski4541e802018-03-13 19:24:23 +01001814SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1815 int __user *, upeer_addrlen, int, flags)
1816{
1817 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1818}
1819
Heiko Carstens20f37032009-01-14 14:14:23 +01001820SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1821 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001822{
Dominik Brodowski4541e802018-03-13 19:24:23 +01001823 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001824}
1825
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826/*
1827 * Attempt to connect to a socket with the server address. The address
1828 * is in user space so we verify it is OK and move it to kernel space.
1829 *
1830 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1831 * break bindings
1832 *
1833 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1834 * other SEQPACKET protocols that take time to connect() as it doesn't
1835 * include the -EINPROGRESS status for such sockets.
1836 */
1837
Jens Axboef499a022019-12-02 16:28:46 -07001838int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
Jens Axboebd3ded32019-11-23 14:17:16 -07001839 int addrlen, int file_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840{
1841 struct socket *sock;
Jens Axboebd3ded32019-11-23 14:17:16 -07001842 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843
Jens Axboebd3ded32019-11-23 14:17:16 -07001844 sock = sock_from_file(file, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845 if (!sock)
1846 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001848 err =
Jens Axboef499a022019-12-02 16:28:46 -07001849 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850 if (err)
Jens Axboebd3ded32019-11-23 14:17:16 -07001851 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
Jens Axboef499a022019-12-02 16:28:46 -07001853 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Jens Axboebd3ded32019-11-23 14:17:16 -07001854 sock->file->f_flags | file_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855out:
1856 return err;
1857}
1858
Jens Axboebd3ded32019-11-23 14:17:16 -07001859int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1860{
1861 int ret = -EBADF;
1862 struct fd f;
1863
1864 f = fdget(fd);
1865 if (f.file) {
Jens Axboef499a022019-12-02 16:28:46 -07001866 struct sockaddr_storage address;
1867
1868 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1869 if (!ret)
1870 ret = __sys_connect_file(f.file, &address, addrlen, 0);
Jens Axboebd3ded32019-11-23 14:17:16 -07001871 if (f.flags)
1872 fput(f.file);
1873 }
1874
1875 return ret;
1876}
1877
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01001878SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1879 int, addrlen)
1880{
1881 return __sys_connect(fd, uservaddr, addrlen);
1882}
1883
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884/*
1885 * Get the local address ('name') of a socket object. Move the obtained
1886 * name to user space.
1887 */
1888
Dominik Brodowski8882a102018-03-13 19:43:14 +01001889int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1890 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891{
1892 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001893 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001894 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001895
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001896 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 if (!sock)
1898 goto out;
1899
1900 err = security_socket_getsockname(sock);
1901 if (err)
1902 goto out_put;
1903
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001904 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1905 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 goto out_put;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001907 /* "err" is actually length in this case */
1908 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909
1910out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001911 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912out:
1913 return err;
1914}
1915
Dominik Brodowski8882a102018-03-13 19:43:14 +01001916SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1917 int __user *, usockaddr_len)
1918{
1919 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1920}
1921
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922/*
1923 * Get the remote address ('name') of a socket object. Move the obtained
1924 * name to user space.
1925 */
1926
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01001927int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1928 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929{
1930 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001931 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001932 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001934 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1935 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 err = security_socket_getpeername(sock);
1937 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001938 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 return err;
1940 }
1941
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001942 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1943 if (err >= 0)
1944 /* "err" is actually length in this case */
1945 err = move_addr_to_user(&address, err, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001946 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001947 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 }
1949 return err;
1950}
1951
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01001952SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1953 int __user *, usockaddr_len)
1954{
1955 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1956}
1957
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958/*
1959 * Send a datagram to a given address. We move the address into kernel
1960 * space and check the user space data area is readable before invoking
1961 * the protocol.
1962 */
Dominik Brodowski211b6342018-03-13 19:18:52 +01001963int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1964 struct sockaddr __user *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965{
1966 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001967 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 int err;
1969 struct msghdr msg;
1970 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001971 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001972
Al Viro602bd0e2015-03-21 19:12:32 -04001973 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1974 if (unlikely(err))
1975 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001976 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1977 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08001978 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001979
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001980 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001981 msg.msg_control = NULL;
1982 msg.msg_controllen = 0;
1983 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001984 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001985 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 if (err < 0)
1987 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001988 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001989 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 }
1991 if (sock->file->f_flags & O_NONBLOCK)
1992 flags |= MSG_DONTWAIT;
1993 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05001994 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001996out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08001997 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08001998out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 return err;
2000}
2001
Dominik Brodowski211b6342018-03-13 19:18:52 +01002002SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2003 unsigned int, flags, struct sockaddr __user *, addr,
2004 int, addr_len)
2005{
2006 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2007}
2008
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002010 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011 */
2012
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002013SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00002014 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015{
Dominik Brodowski211b6342018-03-13 19:18:52 +01002016 return __sys_sendto(fd, buff, len, flags, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017}
2018
2019/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002020 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 * sender. We verify the buffers are writable and if needed move the
2022 * sender address from kernel to user space.
2023 */
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002024int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2025 struct sockaddr __user *addr, int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026{
2027 struct socket *sock;
2028 struct iovec iov;
2029 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002030 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002031 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002032 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Al Viro602bd0e2015-03-21 19:12:32 -04002034 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2035 if (unlikely(err))
2036 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002037 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002039 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002041 msg.msg_control = NULL;
2042 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002043 /* Save some cycles and don't copy the address if not needed */
2044 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2045 /* We assume all kernel code knows the size of sockaddr_storage */
2046 msg.msg_namelen = 0;
tadeusz.struk@intel.com130ed5d2015-12-15 10:46:17 -08002047 msg.msg_iocb = NULL;
Alexander Potapenko9f138fa2017-03-08 18:08:16 +01002048 msg.msg_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 if (sock->file->f_flags & O_NONBLOCK)
2050 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002051 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002053 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002054 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002055 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002056 if (err2 < 0)
2057 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002059
2060 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08002061out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062 return err;
2063}
2064
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002065SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2066 unsigned int, flags, struct sockaddr __user *, addr,
2067 int __user *, addr_len)
2068{
2069 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2070}
2071
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002073 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 */
2075
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02002076SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2077 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078{
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002079 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080}
2081
Florian Westphal83f0c102020-07-05 01:30:15 +02002082static bool sock_use_custom_sol_socket(const struct socket *sock)
2083{
2084 const struct sock *sk = sock->sk;
2085
2086 /* Use sock->ops->setsockopt() for MPTCP */
2087 return IS_ENABLED(CONFIG_MPTCP) &&
2088 sk->sk_protocol == IPPROTO_MPTCP &&
2089 sk->sk_type == SOCK_STREAM &&
2090 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2091}
2092
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093/*
2094 * Set a socket option. Because we don't know the option lengths we have
2095 * to pass the user mode parameter for the protocols to sort out.
2096 */
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002097int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002098 int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099{
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002100 sockptr_t optval = USER_SOCKPTR(user_optval);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002101 char *kernel_optval = NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002102 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 struct socket *sock;
2104
2105 if (optlen < 0)
2106 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002107
2108 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002109 if (!sock)
2110 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111
Christoph Hellwig4a367292020-07-17 08:23:11 +02002112 err = security_socket_setsockopt(sock, level, optname);
2113 if (err)
2114 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002115
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002116 if (!in_compat_syscall())
2117 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002118 user_optval, &optlen,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002119 &kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002120 if (err < 0)
2121 goto out_put;
2122 if (err > 0) {
2123 err = 0;
2124 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 }
Christoph Hellwig4a367292020-07-17 08:23:11 +02002126
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002127 if (kernel_optval)
2128 optval = KERNEL_SOCKPTR(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002129 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002130 err = sock_setsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002131 else if (unlikely(!sock->ops->setsockopt))
2132 err = -EOPNOTSUPP;
Christoph Hellwig4a367292020-07-17 08:23:11 +02002133 else
2134 err = sock->ops->setsockopt(sock, level, optname, optval,
2135 optlen);
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002136 kfree(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002137out_put:
2138 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139 return err;
2140}
2141
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002142SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2143 char __user *, optval, int, optlen)
2144{
2145 return __sys_setsockopt(fd, level, optname, optval, optlen);
2146}
2147
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148/*
2149 * Get a socket option. Because we don't know the option lengths we have
2150 * to pass a user mode parameter for the protocols to sort out.
2151 */
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002152int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2153 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002155 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 struct socket *sock;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002157 int max_optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002159 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002160 if (!sock)
2161 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002163 err = security_socket_getsockopt(sock, level, optname);
2164 if (err)
2165 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002166
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002167 if (!in_compat_syscall())
2168 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002169
2170 if (level == SOL_SOCKET)
2171 err = sock_getsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002172 else if (unlikely(!sock->ops->getsockopt))
2173 err = -EOPNOTSUPP;
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002174 else
2175 err = sock->ops->getsockopt(sock, level, optname, optval,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002176 optlen);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002177
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002178 if (!in_compat_syscall())
2179 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2180 optval, optlen, max_optlen,
2181 err);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002182out_put:
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002183 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 return err;
2185}
2186
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002187SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2188 char __user *, optval, int __user *, optlen)
2189{
2190 return __sys_getsockopt(fd, level, optname, optval, optlen);
2191}
2192
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193/*
2194 * Shutdown a socket.
2195 */
2196
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002197int __sys_shutdown(int fd, int how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002199 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200 struct socket *sock;
2201
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002202 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2203 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204 err = security_socket_shutdown(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002205 if (!err)
2206 err = sock->ops->shutdown(sock, how);
2207 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208 }
2209 return err;
2210}
2211
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002212SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2213{
2214 return __sys_shutdown(fd, how);
2215}
2216
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002217/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218 * fields which are the same type (int / unsigned) on our platforms.
2219 */
2220#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2221#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2222#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2223
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002224struct used_address {
2225 struct sockaddr_storage name;
2226 unsigned int name_len;
2227};
2228
Jens Axboe0a384ab2020-02-27 08:11:20 -07002229int __copy_msghdr_from_user(struct msghdr *kmsg,
2230 struct user_msghdr __user *umsg,
2231 struct sockaddr __user **save_addr,
2232 struct iovec __user **uiov, size_t *nsegs)
Dan Carpenter1661bf32013-10-03 00:27:20 +03002233{
Al Viroffb07552017-06-27 19:32:04 -04002234 struct user_msghdr msg;
Al Viro08adb7d2014-11-10 20:23:13 -05002235 ssize_t err;
2236
Al Viroffb07552017-06-27 19:32:04 -04002237 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
Dan Carpenter1661bf32013-10-03 00:27:20 +03002238 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00002239
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002240 kmsg->msg_control_is_user = true;
2241 kmsg->msg_control_user = msg.msg_control;
Al Viroffb07552017-06-27 19:32:04 -04002242 kmsg->msg_controllen = msg.msg_controllen;
2243 kmsg->msg_flags = msg.msg_flags;
2244
2245 kmsg->msg_namelen = msg.msg_namelen;
2246 if (!msg.msg_name)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07002247 kmsg->msg_namelen = 0;
2248
Matthew Leachdbb490b2014-03-11 11:58:27 +00002249 if (kmsg->msg_namelen < 0)
2250 return -EINVAL;
2251
Dan Carpenter1661bf32013-10-03 00:27:20 +03002252 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03002253 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05002254
2255 if (save_addr)
Al Viroffb07552017-06-27 19:32:04 -04002256 *save_addr = msg.msg_name;
Al Viro08adb7d2014-11-10 20:23:13 -05002257
Al Viroffb07552017-06-27 19:32:04 -04002258 if (msg.msg_name && kmsg->msg_namelen) {
Al Viro08adb7d2014-11-10 20:23:13 -05002259 if (!save_addr) {
Paolo Abeni864d9662017-07-21 18:49:45 +02002260 err = move_addr_to_kernel(msg.msg_name,
2261 kmsg->msg_namelen,
Al Viro08adb7d2014-11-10 20:23:13 -05002262 kmsg->msg_name);
2263 if (err < 0)
2264 return err;
2265 }
2266 } else {
2267 kmsg->msg_name = NULL;
2268 kmsg->msg_namelen = 0;
2269 }
2270
Al Viroffb07552017-06-27 19:32:04 -04002271 if (msg.msg_iovlen > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05002272 return -EMSGSIZE;
2273
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002274 kmsg->msg_iocb = NULL;
Jens Axboe0a384ab2020-02-27 08:11:20 -07002275 *uiov = msg.msg_iov;
2276 *nsegs = msg.msg_iovlen;
2277 return 0;
2278}
2279
2280static int copy_msghdr_from_user(struct msghdr *kmsg,
2281 struct user_msghdr __user *umsg,
2282 struct sockaddr __user **save_addr,
2283 struct iovec **iov)
2284{
2285 struct user_msghdr msg;
2286 ssize_t err;
2287
2288 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2289 &msg.msg_iovlen);
2290 if (err)
2291 return err;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002292
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002293 err = import_iovec(save_addr ? READ : WRITE,
Al Viroffb07552017-06-27 19:32:04 -04002294 msg.msg_iov, msg.msg_iovlen,
Al Viroda184282015-03-21 19:29:06 -04002295 UIO_FASTIOV, iov, &kmsg->msg_iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002296 return err < 0 ? err : 0;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002297}
2298
Jens Axboe4257c8c2019-11-25 14:27:34 -07002299static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2300 unsigned int flags, struct used_address *used_address,
2301 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302{
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002303 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Amit Kushwaha846cc122016-12-08 18:21:53 +05302304 __aligned(sizeof(__kernel_size_t));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002305 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05002307 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05002308 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002309
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310 err = -ENOBUFS;
2311
Anton Blanchard228e5482011-05-02 20:21:35 +00002312 if (msg_sys->msg_controllen > INT_MAX)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002313 goto out;
Tom Herbert28a94d82016-03-07 14:11:02 -08002314 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00002315 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002317 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002318 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002319 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002321 goto out;
Anton Blanchard228e5482011-05-02 20:21:35 +00002322 ctl_buf = msg_sys->msg_control;
2323 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324 } else if (ctl_len) {
David S. Millerac4340f2017-01-04 13:24:19 -05002325 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2326 CMSG_ALIGN(sizeof(struct cmsghdr)));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002327 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002329 if (ctl_buf == NULL)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002330 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002331 }
2332 err = -EFAULT;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002333 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002335 msg_sys->msg_control = ctl_buf;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002336 msg_sys->msg_control_is_user = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002338 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339
2340 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002341 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002342 /*
2343 * If this is sendmmsg() and current destination address is same as
2344 * previously succeeded address, omit asking LSM's decision.
2345 * used_address->name_len is initialized to UINT_MAX so that the first
2346 * destination address never matches.
2347 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002348 if (used_address && msg_sys->msg_name &&
2349 used_address->name_len == msg_sys->msg_namelen &&
2350 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002351 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05002352 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002353 goto out_freectl;
2354 }
Al Virod8725c82014-12-11 00:02:50 -05002355 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002356 /*
2357 * If this is sendmmsg() and sending to current destination address was
2358 * successful, remember it.
2359 */
2360 if (used_address && err >= 0) {
2361 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002362 if (msg_sys->msg_name)
2363 memcpy(&used_address->name, msg_sys->msg_name,
2364 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002365 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366
2367out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002368 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
Jens Axboe4257c8c2019-11-25 14:27:34 -07002370out:
2371 return err;
2372}
2373
Jens Axboe03b12302019-12-02 18:50:25 -07002374int sendmsg_copy_msghdr(struct msghdr *msg,
2375 struct user_msghdr __user *umsg, unsigned flags,
2376 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002377{
2378 int err;
2379
2380 if (flags & MSG_CMSG_COMPAT) {
2381 struct compat_msghdr __user *msg_compat;
2382
2383 msg_compat = (struct compat_msghdr __user *) umsg;
2384 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2385 } else {
2386 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2387 }
2388 if (err < 0)
2389 return err;
2390
2391 return 0;
2392}
2393
2394static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2395 struct msghdr *msg_sys, unsigned int flags,
2396 struct used_address *used_address,
2397 unsigned int allowed_msghdr_flags)
2398{
2399 struct sockaddr_storage address;
2400 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2401 ssize_t err;
2402
2403 msg_sys->msg_name = &address;
2404
2405 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2406 if (err < 0)
2407 return err;
2408
2409 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2410 allowed_msghdr_flags);
Al Viroda184282015-03-21 19:29:06 -04002411 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002412 return err;
2413}
2414
2415/*
2416 * BSD sendmsg interface
2417 */
Jens Axboe03b12302019-12-02 18:50:25 -07002418long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
Jens Axboe0fa03c62019-04-19 13:34:07 -06002419 unsigned int flags)
2420{
Jens Axboed69e0772019-11-25 17:04:13 -07002421 /* disallow ancillary data requests from this path */
Jens Axboe03b12302019-12-02 18:50:25 -07002422 if (msg->msg_control || msg->msg_controllen)
2423 return -EINVAL;
Jens Axboed69e0772019-11-25 17:04:13 -07002424
Jens Axboe03b12302019-12-02 18:50:25 -07002425 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
Jens Axboe0fa03c62019-04-19 13:34:07 -06002426}
Anton Blanchard228e5482011-05-02 20:21:35 +00002427
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002428long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2429 bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002430{
2431 int fput_needed, err;
2432 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002433 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002434
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002435 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2436 return -EINVAL;
2437
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002438 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002439 if (!sock)
2440 goto out;
2441
Tom Herbert28a94d82016-03-07 14:11:02 -08002442 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00002443
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002444 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002445out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446 return err;
2447}
2448
Al Viro666547f2014-04-06 14:03:05 -04002449SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002450{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002451 return __sys_sendmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002452}
2453
Anton Blanchard228e5482011-05-02 20:21:35 +00002454/*
2455 * Linux sendmmsg interface
2456 */
2457
2458int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002459 unsigned int flags, bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002460{
2461 int fput_needed, err, datagrams;
2462 struct socket *sock;
2463 struct mmsghdr __user *entry;
2464 struct compat_mmsghdr __user *compat_entry;
2465 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002466 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002467 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002468
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002469 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2470 return -EINVAL;
2471
Anton Blanchard98382f42011-08-04 14:07:39 +00002472 if (vlen > UIO_MAXIOV)
2473 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002474
2475 datagrams = 0;
2476
2477 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2478 if (!sock)
2479 return err;
2480
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002481 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002482 entry = mmsg;
2483 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002484 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002485 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002486
2487 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002488 if (datagrams == vlen - 1)
2489 flags = oflags;
2490
Anton Blanchard228e5482011-05-02 20:21:35 +00002491 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002492 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002493 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002494 if (err < 0)
2495 break;
2496 err = __put_user(err, &compat_entry->msg_len);
2497 ++compat_entry;
2498 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002499 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002500 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002501 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002502 if (err < 0)
2503 break;
2504 err = put_user(err, &entry->msg_len);
2505 ++entry;
2506 }
2507
2508 if (err)
2509 break;
2510 ++datagrams;
Soheil Hassas Yeganeh30238982016-11-04 15:36:49 -04002511 if (msg_data_left(&msg_sys))
2512 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002513 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002514 }
2515
Anton Blanchard228e5482011-05-02 20:21:35 +00002516 fput_light(sock->file, fput_needed);
2517
Anton Blanchard728ffb82011-08-04 14:07:38 +00002518 /* We only return an error if no datagrams were able to be sent */
2519 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002520 return datagrams;
2521
Anton Blanchard228e5482011-05-02 20:21:35 +00002522 return err;
2523}
2524
2525SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2526 unsigned int, vlen, unsigned int, flags)
2527{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002528 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002529}
2530
Jens Axboe03b12302019-12-02 18:50:25 -07002531int recvmsg_copy_msghdr(struct msghdr *msg,
2532 struct user_msghdr __user *umsg, unsigned flags,
2533 struct sockaddr __user **uaddr,
2534 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002535{
2536 ssize_t err;
2537
2538 if (MSG_CMSG_COMPAT & flags) {
2539 struct compat_msghdr __user *msg_compat;
2540
2541 msg_compat = (struct compat_msghdr __user *) umsg;
2542 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2543 } else {
2544 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2545 }
2546 if (err < 0)
2547 return err;
2548
2549 return 0;
2550}
2551
2552static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2553 struct user_msghdr __user *msg,
2554 struct sockaddr __user *uaddr,
2555 unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002557 struct compat_msghdr __user *msg_compat =
Jens Axboe4257c8c2019-11-25 14:27:34 -07002558 (struct compat_msghdr __user *) msg;
2559 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2560 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002562 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002563 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002564
Al Viro08adb7d2014-11-10 20:23:13 -05002565 msg_sys->msg_name = &addr;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002566 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2567 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002568
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002569 /* We assume all kernel code knows the size of sockaddr_storage */
2570 msg_sys->msg_namelen = 0;
2571
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572 if (sock->file->f_flags & O_NONBLOCK)
2573 flags |= MSG_DONTWAIT;
Eric Dumazet1af66222019-12-06 09:38:36 -08002574
2575 if (unlikely(nosec))
2576 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2577 else
2578 err = sock_recvmsg(sock, msg_sys, flags);
2579
Linus Torvalds1da177e2005-04-16 15:20:36 -07002580 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002581 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002582 len = err;
2583
2584 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002585 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002586 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002587 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002588 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002589 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002591 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002592 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002594 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002595 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002596 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597 &msg_compat->msg_controllen);
2598 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002599 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002600 &msg->msg_controllen);
2601 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002602 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603 err = len;
Jens Axboe4257c8c2019-11-25 14:27:34 -07002604out:
2605 return err;
2606}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002607
Jens Axboe4257c8c2019-11-25 14:27:34 -07002608static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2609 struct msghdr *msg_sys, unsigned int flags, int nosec)
2610{
2611 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2612 /* user mode address pointers */
2613 struct sockaddr __user *uaddr;
2614 ssize_t err;
2615
2616 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2617 if (err < 0)
2618 return err;
2619
2620 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
Al Viroda184282015-03-21 19:29:06 -04002621 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002622 return err;
2623}
2624
2625/*
2626 * BSD recvmsg interface
2627 */
2628
Jens Axboe03b12302019-12-02 18:50:25 -07002629long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2630 struct user_msghdr __user *umsg,
2631 struct sockaddr __user *uaddr, unsigned int flags)
Jens Axboeaa1fa282019-04-19 13:38:09 -06002632{
Jens Axboed69e0772019-11-25 17:04:13 -07002633 /* disallow ancillary data requests from this path */
Jens Axboe03b12302019-12-02 18:50:25 -07002634 if (msg->msg_control || msg->msg_controllen)
2635 return -EINVAL;
Jens Axboed69e0772019-11-25 17:04:13 -07002636
Jens Axboe03b12302019-12-02 18:50:25 -07002637 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
Jens Axboeaa1fa282019-04-19 13:38:09 -06002638}
2639
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002640long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2641 bool forbid_cmsg_compat)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002642{
2643 int fput_needed, err;
2644 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002645 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002646
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002647 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2648 return -EINVAL;
2649
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002650 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002651 if (!sock)
2652 goto out;
2653
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002654 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002655
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002656 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657out:
2658 return err;
2659}
2660
Al Viro666547f2014-04-06 14:03:05 -04002661SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002662 unsigned int, flags)
2663{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002664 return __sys_recvmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002665}
2666
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002667/*
2668 * Linux recvmmsg interface
2669 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002671static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2672 unsigned int vlen, unsigned int flags,
2673 struct timespec64 *timeout)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002674{
2675 int fput_needed, err, datagrams;
2676 struct socket *sock;
2677 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002678 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002679 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002680 struct timespec64 end_time;
2681 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002682
2683 if (timeout &&
2684 poll_select_set_timeout(&end_time, timeout->tv_sec,
2685 timeout->tv_nsec))
2686 return -EINVAL;
2687
2688 datagrams = 0;
2689
2690 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2691 if (!sock)
2692 return err;
2693
Soheil Hassas Yeganeh7797dc42018-02-27 18:22:40 -05002694 if (likely(!(flags & MSG_ERRQUEUE))) {
2695 err = sock_error(sock->sk);
2696 if (err) {
2697 datagrams = err;
2698 goto out_put;
2699 }
Maxime Jayate623a9e2017-02-21 18:35:51 +01002700 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002701
2702 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002703 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002704
2705 while (datagrams < vlen) {
2706 /*
2707 * No need to ask LSM for more than the first datagram.
2708 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002709 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002710 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002711 &msg_sys, flags & ~MSG_WAITFORONE,
2712 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002713 if (err < 0)
2714 break;
2715 err = __put_user(err, &compat_entry->msg_len);
2716 ++compat_entry;
2717 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002718 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002719 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002720 &msg_sys, flags & ~MSG_WAITFORONE,
2721 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002722 if (err < 0)
2723 break;
2724 err = put_user(err, &entry->msg_len);
2725 ++entry;
2726 }
2727
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002728 if (err)
2729 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002730 ++datagrams;
2731
Brandon L Black71c5c1592010-03-26 16:18:03 +00002732 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2733 if (flags & MSG_WAITFORONE)
2734 flags |= MSG_DONTWAIT;
2735
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002736 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002737 ktime_get_ts64(&timeout64);
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002738 *timeout = timespec64_sub(end_time, timeout64);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002739 if (timeout->tv_sec < 0) {
2740 timeout->tv_sec = timeout->tv_nsec = 0;
2741 break;
2742 }
2743
2744 /* Timeout, return less than vlen datagrams */
2745 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2746 break;
2747 }
2748
2749 /* Out of band data, return right away */
2750 if (msg_sys.msg_flags & MSG_OOB)
2751 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002752 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002753 }
2754
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002755 if (err == 0)
2756 goto out_put;
2757
2758 if (datagrams == 0) {
2759 datagrams = err;
2760 goto out_put;
2761 }
2762
2763 /*
2764 * We may return less entries than requested (vlen) if the
2765 * sock is non block and there aren't enough datagrams...
2766 */
2767 if (err != -EAGAIN) {
2768 /*
2769 * ... or if recvmsg returns an error after we
2770 * received some datagrams, where we record the
2771 * error to return on the next call or if the
2772 * app asks about it using getsockopt(SO_ERROR).
2773 */
2774 sock->sk->sk_err = -err;
2775 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002776out_put:
2777 fput_light(sock->file, fput_needed);
2778
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002779 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002780}
2781
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002782int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2783 unsigned int vlen, unsigned int flags,
2784 struct __kernel_timespec __user *timeout,
2785 struct old_timespec32 __user *timeout32)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002786{
2787 int datagrams;
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002788 struct timespec64 timeout_sys;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002789
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002790 if (timeout && get_timespec64(&timeout_sys, timeout))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002791 return -EFAULT;
2792
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002793 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2794 return -EFAULT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002795
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002796 if (!timeout && !timeout32)
2797 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2798
2799 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2800
2801 if (datagrams <= 0)
2802 return datagrams;
2803
2804 if (timeout && put_timespec64(&timeout_sys, timeout))
2805 datagrams = -EFAULT;
2806
2807 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002808 datagrams = -EFAULT;
2809
2810 return datagrams;
2811}
2812
Dominik Brodowski1255e262018-03-13 20:44:21 +01002813SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2814 unsigned int, vlen, unsigned int, flags,
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002815 struct __kernel_timespec __user *, timeout)
Dominik Brodowski1255e262018-03-13 20:44:21 +01002816{
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002817 if (flags & MSG_CMSG_COMPAT)
2818 return -EINVAL;
2819
2820 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
Dominik Brodowski1255e262018-03-13 20:44:21 +01002821}
2822
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002823#ifdef CONFIG_COMPAT_32BIT_TIME
2824SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2825 unsigned int, vlen, unsigned int, flags,
2826 struct old_timespec32 __user *, timeout)
2827{
2828 if (flags & MSG_CMSG_COMPAT)
2829 return -EINVAL;
2830
2831 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2832}
2833#endif
2834
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002835#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836/* Argument list sizes for sys_socketcall */
2837#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002838static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002839 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2840 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2841 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002842 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002843};
2844
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845#undef AL
2846
2847/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002848 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002849 *
2850 * Argument checking cleaned up. Saved 20% in size.
2851 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002852 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002853 */
2854
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002855SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856{
Chen Gang2950fa92013-04-07 16:55:23 +08002857 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002858 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002860 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002861
Anton Blanchard228e5482011-05-02 20:21:35 +00002862 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863 return -EINVAL;
Jeremy Clinec8e8cd52018-07-27 22:43:01 +00002864 call = array_index_nospec(call, SYS_SENDMMSG + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865
Arjan van de Ven47379052009-09-28 12:57:44 -07002866 len = nargs[call];
2867 if (len > sizeof(a))
2868 return -EINVAL;
2869
Linus Torvalds1da177e2005-04-16 15:20:36 -07002870 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002871 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002872 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002873
Chen Gang2950fa92013-04-07 16:55:23 +08002874 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2875 if (err)
2876 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002877
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002878 a0 = a[0];
2879 a1 = a[1];
2880
2881 switch (call) {
2882 case SYS_SOCKET:
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01002883 err = __sys_socket(a0, a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002884 break;
2885 case SYS_BIND:
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01002886 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002887 break;
2888 case SYS_CONNECT:
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01002889 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002890 break;
2891 case SYS_LISTEN:
Dominik Brodowski25e290e2018-03-13 19:36:54 +01002892 err = __sys_listen(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002893 break;
2894 case SYS_ACCEPT:
Dominik Brodowski4541e802018-03-13 19:24:23 +01002895 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2896 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002897 break;
2898 case SYS_GETSOCKNAME:
2899 err =
Dominik Brodowski8882a102018-03-13 19:43:14 +01002900 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2901 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002902 break;
2903 case SYS_GETPEERNAME:
2904 err =
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002905 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2906 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002907 break;
2908 case SYS_SOCKETPAIR:
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01002909 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002910 break;
2911 case SYS_SEND:
Dominik Brodowskif3bf8962018-03-13 19:52:00 +01002912 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2913 NULL, 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002914 break;
2915 case SYS_SENDTO:
Dominik Brodowski211b6342018-03-13 19:18:52 +01002916 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2917 (struct sockaddr __user *)a[4], a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002918 break;
2919 case SYS_RECV:
Dominik Brodowskid27e9af2018-03-13 19:54:17 +01002920 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2921 NULL, NULL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002922 break;
2923 case SYS_RECVFROM:
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002924 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2925 (struct sockaddr __user *)a[4],
2926 (int __user *)a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002927 break;
2928 case SYS_SHUTDOWN:
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002929 err = __sys_shutdown(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002930 break;
2931 case SYS_SETSOCKOPT:
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002932 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2933 a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002934 break;
2935 case SYS_GETSOCKOPT:
2936 err =
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002937 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2938 (int __user *)a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002939 break;
2940 case SYS_SENDMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002941 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2942 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002943 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002944 case SYS_SENDMMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002945 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2946 a[3], true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002947 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002948 case SYS_RECVMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002949 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2950 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002951 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002952 case SYS_RECVMMSG:
Arnd Bergmann3ca47e92019-04-23 17:43:50 +02002953 if (IS_ENABLED(CONFIG_64BIT))
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002954 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2955 a[2], a[3],
2956 (struct __kernel_timespec __user *)a[4],
2957 NULL);
2958 else
2959 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
2960 a[2], a[3], NULL,
2961 (struct old_timespec32 __user *)a[4]);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002962 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08002963 case SYS_ACCEPT4:
Dominik Brodowski4541e802018-03-13 19:24:23 +01002964 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2965 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002966 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002967 default:
2968 err = -EINVAL;
2969 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002970 }
2971 return err;
2972}
2973
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002974#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002975
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002976/**
2977 * sock_register - add a socket protocol handler
2978 * @ops: description of protocol
2979 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002980 * This function is called by a protocol handler that wants to
2981 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09002982 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002983 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002984 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07002985int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002986{
2987 int err;
2988
2989 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08002990 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002991 return -ENOBUFS;
2992 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002993
2994 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00002995 if (rcu_dereference_protected(net_families[ops->family],
2996 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07002997 err = -EEXIST;
2998 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00002999 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003000 err = 0;
3001 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003002 spin_unlock(&net_family_lock);
3003
Yang Yingliang3410f222014-02-12 17:09:55 +08003004 pr_info("NET: Registered protocol family %d\n", ops->family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003005 return err;
3006}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003007EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003008
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003009/**
3010 * sock_unregister - remove a protocol handler
3011 * @family: protocol family to remove
3012 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003013 * This function is called by a protocol handler that wants to
3014 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003015 * new socket creation.
3016 *
3017 * If protocol handler is a module, then it can use module reference
3018 * counts to protect against new references. If protocol handler is not
3019 * a module then it needs to provide its own protection in
3020 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003021 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003022void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003023{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003024 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003025
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003026 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00003027 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003028 spin_unlock(&net_family_lock);
3029
3030 synchronize_rcu();
3031
Yang Yingliang3410f222014-02-12 17:09:55 +08003032 pr_info("NET: Unregistered protocol family %d\n", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003034EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003035
Xin Longbf2ae2e2018-03-10 18:57:50 +08003036bool sock_is_registered(int family)
3037{
Jeremy Cline66b51b02018-08-13 22:23:13 +00003038 return family < NPROTO && rcu_access_pointer(net_families[family]);
Xin Longbf2ae2e2018-03-10 18:57:50 +08003039}
3040
Andi Kleen77d76ea2005-12-22 12:43:42 -08003041static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003042{
Nick Pigginb3e19d92011-01-07 17:50:11 +11003043 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00003044 /*
3045 * Initialize the network sysctl infrastructure.
3046 */
3047 err = net_sysctl_init();
3048 if (err)
3049 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11003050
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003052 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07003053 */
3054 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003055
3056 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003057 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003058 */
3059
3060 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11003061
3062 err = register_filesystem(&sock_fs_type);
3063 if (err)
3064 goto out_fs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003065 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003066 if (IS_ERR(sock_mnt)) {
3067 err = PTR_ERR(sock_mnt);
3068 goto out_mount;
3069 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08003070
3071 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003072 */
3073
3074#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00003075 err = netfilter_init();
3076 if (err)
3077 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003078#endif
David S. Millercbeb3212005-12-22 12:58:55 -08003079
Daniel Borkmann408eccc2014-04-01 16:20:23 +02003080 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00003081
Nick Pigginb3e19d92011-01-07 17:50:11 +11003082out:
3083 return err;
3084
3085out_mount:
3086 unregister_filesystem(&sock_fs_type);
3087out_fs:
3088 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003089}
3090
Andi Kleen77d76ea2005-12-22 12:43:42 -08003091core_initcall(sock_init); /* early initcall */
3092
Linus Torvalds1da177e2005-04-16 15:20:36 -07003093#ifdef CONFIG_PROC_FS
3094void socket_seq_show(struct seq_file *seq)
3095{
Tonghao Zhang648845a2017-12-14 05:51:58 -08003096 seq_printf(seq, "sockets: used %d\n",
3097 sock_inuse_get(seq->private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003098}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003099#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003100
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003101#ifdef CONFIG_COMPAT
Al Viro36fd6332017-06-26 13:19:16 -04003102static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003103{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003104 struct compat_ifconf ifc32;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003105 struct ifconf ifc;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003106 int err;
3107
Arnd Bergmann6b960182009-11-06 23:10:54 -08003108 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003109 return -EFAULT;
3110
Al Viro36fd6332017-06-26 13:19:16 -04003111 ifc.ifc_len = ifc32.ifc_len;
3112 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003113
Al Viro36fd6332017-06-26 13:19:16 -04003114 rtnl_lock();
3115 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
3116 rtnl_unlock();
Arnd Bergmann7a229382009-11-06 23:00:29 -08003117 if (err)
3118 return err;
3119
Al Viro36fd6332017-06-26 13:19:16 -04003120 ifc32.ifc_len = ifc.ifc_len;
Arnd Bergmann6b960182009-11-06 23:10:54 -08003121 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003122 return -EFAULT;
3123
3124 return 0;
3125}
3126
Arnd Bergmann6b960182009-11-06 23:10:54 -08003127static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003128{
Ben Hutchings3a7da392011-03-17 07:34:32 +00003129 struct compat_ethtool_rxnfc __user *compat_rxnfc;
3130 bool convert_in = false, convert_out = false;
Al Viro44c02a22017-10-05 12:59:44 -04003131 size_t buf_size = 0;
3132 struct ethtool_rxnfc __user *rxnfc = NULL;
3133 struct ifreq ifr;
Ben Hutchings3a7da392011-03-17 07:34:32 +00003134 u32 rule_cnt = 0, actual_rule_cnt;
3135 u32 ethcmd;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003136 u32 data;
Ben Hutchings3a7da392011-03-17 07:34:32 +00003137 int ret;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003138
3139 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
3140 return -EFAULT;
3141
Ben Hutchings3a7da392011-03-17 07:34:32 +00003142 compat_rxnfc = compat_ptr(data);
3143
3144 if (get_user(ethcmd, &compat_rxnfc->cmd))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003145 return -EFAULT;
3146
Ben Hutchings3a7da392011-03-17 07:34:32 +00003147 /* Most ethtool structures are defined without padding.
3148 * Unfortunately struct ethtool_rxnfc is an exception.
3149 */
3150 switch (ethcmd) {
3151 default:
3152 break;
3153 case ETHTOOL_GRXCLSRLALL:
3154 /* Buffer size is variable */
3155 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
3156 return -EFAULT;
3157 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
3158 return -ENOMEM;
3159 buf_size += rule_cnt * sizeof(u32);
3160 /* fall through */
3161 case ETHTOOL_GRXRINGS:
3162 case ETHTOOL_GRXCLSRLCNT:
3163 case ETHTOOL_GRXCLSRULE:
Ben Hutchings55664f32012-01-03 12:04:51 +00003164 case ETHTOOL_SRXCLSRLINS:
Ben Hutchings3a7da392011-03-17 07:34:32 +00003165 convert_out = true;
3166 /* fall through */
3167 case ETHTOOL_SRXCLSRLDEL:
Ben Hutchings3a7da392011-03-17 07:34:32 +00003168 buf_size += sizeof(struct ethtool_rxnfc);
3169 convert_in = true;
Al Viro44c02a22017-10-05 12:59:44 -04003170 rxnfc = compat_alloc_user_space(buf_size);
Ben Hutchings3a7da392011-03-17 07:34:32 +00003171 break;
3172 }
3173
Al Viro44c02a22017-10-05 12:59:44 -04003174 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
Ben Hutchings3a7da392011-03-17 07:34:32 +00003175 return -EFAULT;
3176
Al Viro44c02a22017-10-05 12:59:44 -04003177 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
Ben Hutchings3a7da392011-03-17 07:34:32 +00003178
3179 if (convert_in) {
Alexander Duyck127fe532011-04-08 18:01:59 +00003180 /* We expect there to be holes between fs.m_ext and
Ben Hutchings3a7da392011-03-17 07:34:32 +00003181 * fs.ring_cookie and at the end of fs, but nowhere else.
3182 */
Alexander Duyck127fe532011-04-08 18:01:59 +00003183 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
3184 sizeof(compat_rxnfc->fs.m_ext) !=
3185 offsetof(struct ethtool_rxnfc, fs.m_ext) +
3186 sizeof(rxnfc->fs.m_ext));
Ben Hutchings3a7da392011-03-17 07:34:32 +00003187 BUILD_BUG_ON(
3188 offsetof(struct compat_ethtool_rxnfc, fs.location) -
3189 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
3190 offsetof(struct ethtool_rxnfc, fs.location) -
3191 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
3192
3193 if (copy_in_user(rxnfc, compat_rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00003194 (void __user *)(&rxnfc->fs.m_ext + 1) -
3195 (void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00003196 copy_in_user(&rxnfc->fs.ring_cookie,
3197 &compat_rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00003198 (void __user *)(&rxnfc->fs.location + 1) -
Wenwen Wangb6168562018-10-18 09:36:46 -05003199 (void __user *)&rxnfc->fs.ring_cookie))
3200 return -EFAULT;
3201 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3202 if (put_user(rule_cnt, &rxnfc->rule_cnt))
3203 return -EFAULT;
3204 } else if (copy_in_user(&rxnfc->rule_cnt,
3205 &compat_rxnfc->rule_cnt,
3206 sizeof(rxnfc->rule_cnt)))
Ben Hutchings3a7da392011-03-17 07:34:32 +00003207 return -EFAULT;
3208 }
3209
Al Viro44c02a22017-10-05 12:59:44 -04003210 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
Ben Hutchings3a7da392011-03-17 07:34:32 +00003211 if (ret)
3212 return ret;
3213
3214 if (convert_out) {
3215 if (copy_in_user(compat_rxnfc, rxnfc,
Stephen Hemminger954b1242013-02-11 06:22:28 +00003216 (const void __user *)(&rxnfc->fs.m_ext + 1) -
3217 (const void __user *)rxnfc) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00003218 copy_in_user(&compat_rxnfc->fs.ring_cookie,
3219 &rxnfc->fs.ring_cookie,
Stephen Hemminger954b1242013-02-11 06:22:28 +00003220 (const void __user *)(&rxnfc->fs.location + 1) -
3221 (const void __user *)&rxnfc->fs.ring_cookie) ||
Ben Hutchings3a7da392011-03-17 07:34:32 +00003222 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
3223 sizeof(rxnfc->rule_cnt)))
3224 return -EFAULT;
3225
3226 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
3227 /* As an optimisation, we only copy the actual
3228 * number of rules that the underlying
3229 * function returned. Since Mallory might
3230 * change the rule count in user memory, we
3231 * check that it is less than the rule count
3232 * originally given (as the user buffer size),
3233 * which has been range-checked.
3234 */
3235 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
3236 return -EFAULT;
3237 if (actual_rule_cnt < rule_cnt)
3238 rule_cnt = actual_rule_cnt;
3239 if (copy_in_user(&compat_rxnfc->rule_locs[0],
3240 &rxnfc->rule_locs[0],
3241 rule_cnt * sizeof(u32)))
3242 return -EFAULT;
3243 }
3244 }
3245
3246 return 0;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003247}
3248
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003249static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3250{
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003251 compat_uptr_t uptr32;
Al Viro44c02a22017-10-05 12:59:44 -04003252 struct ifreq ifr;
3253 void __user *saved;
3254 int err;
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003255
Al Viro44c02a22017-10-05 12:59:44 -04003256 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003257 return -EFAULT;
3258
3259 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3260 return -EFAULT;
3261
Al Viro44c02a22017-10-05 12:59:44 -04003262 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3263 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003264
Al Viro44c02a22017-10-05 12:59:44 -04003265 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
3266 if (!err) {
3267 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3268 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
3269 err = -EFAULT;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003270 }
Al Viro44c02a22017-10-05 12:59:44 -04003271 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003272}
3273
Ben Hutchings590d4692013-11-18 17:04:13 +00003274/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3275static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08003276 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003277{
Al Viro44c02a22017-10-05 12:59:44 -04003278 struct ifreq ifreq;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003279 u32 data32;
3280
Al Viro44c02a22017-10-05 12:59:44 -04003281 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003282 return -EFAULT;
Al Viro44c02a22017-10-05 12:59:44 -04003283 if (get_user(data32, &u_ifreq32->ifr_data))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003284 return -EFAULT;
Al Viro44c02a22017-10-05 12:59:44 -04003285 ifreq.ifr_data = compat_ptr(data32);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003286
Al Viro44c02a22017-10-05 12:59:44 -04003287 return dev_ioctl(net, cmd, &ifreq, NULL);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003288}
3289
Johannes Berg37ac39b2019-01-25 22:43:18 +01003290static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
3291 unsigned int cmd,
3292 struct compat_ifreq __user *uifr32)
3293{
3294 struct ifreq __user *uifr;
3295 int err;
3296
3297 /* Handle the fact that while struct ifreq has the same *layout* on
3298 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3299 * which are handled elsewhere, it still has different *size* due to
3300 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3301 * resulting in struct ifreq being 32 and 40 bytes respectively).
3302 * As a result, if the struct happens to be at the end of a page and
3303 * the next page isn't readable/writable, we get a fault. To prevent
3304 * that, copy back and forth to the full size.
3305 */
3306
3307 uifr = compat_alloc_user_space(sizeof(*uifr));
3308 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
3309 return -EFAULT;
3310
3311 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
3312
3313 if (!err) {
3314 switch (cmd) {
3315 case SIOCGIFFLAGS:
3316 case SIOCGIFMETRIC:
3317 case SIOCGIFMTU:
3318 case SIOCGIFMEM:
3319 case SIOCGIFHWADDR:
3320 case SIOCGIFINDEX:
3321 case SIOCGIFADDR:
3322 case SIOCGIFBRDADDR:
3323 case SIOCGIFDSTADDR:
3324 case SIOCGIFNETMASK:
3325 case SIOCGIFPFLAGS:
3326 case SIOCGIFTXQLEN:
3327 case SIOCGMIIPHY:
3328 case SIOCGMIIREG:
Johannes Bergc6c9fee2019-01-25 22:43:19 +01003329 case SIOCGIFNAME:
Johannes Berg37ac39b2019-01-25 22:43:18 +01003330 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3331 err = -EFAULT;
3332 break;
3333 }
3334 }
3335 return err;
3336}
3337
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003338static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3339 struct compat_ifreq __user *uifr32)
3340{
3341 struct ifreq ifr;
3342 struct compat_ifmap __user *uifmap32;
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003343 int err;
3344
3345 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3346 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003347 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3348 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3349 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3350 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3351 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3352 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003353 if (err)
3354 return -EFAULT;
3355
Al Viro44c02a22017-10-05 12:59:44 -04003356 err = dev_ioctl(net, cmd, &ifr, NULL);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003357
3358 if (cmd == SIOCGIFMAP && !err) {
3359 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
Mathieu Desnoyers3ddc5b42013-09-11 14:23:18 -07003360 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3361 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3362 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3363 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3364 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3365 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003366 if (err)
3367 err = -EFAULT;
3368 }
3369 return err;
3370}
3371
Arnd Bergmann7a229382009-11-06 23:00:29 -08003372/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3373 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003374 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003375 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003376static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003377{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003378 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003379
Arnd Bergmann6b960182009-11-06 23:10:54 -08003380 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003381 return -EFAULT;
3382 if (tmp == BRCTL_GET_VERSION)
3383 return BRCTL_VERSION + 1;
3384 return -EINVAL;
3385}
3386
Arnd Bergmann6b960182009-11-06 23:10:54 -08003387static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3388 unsigned int cmd, unsigned long arg)
3389{
3390 void __user *argp = compat_ptr(arg);
3391 struct sock *sk = sock->sk;
3392 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003393
Arnd Bergmann6b960182009-11-06 23:10:54 -08003394 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Ben Hutchings590d4692013-11-18 17:04:13 +00003395 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003396
Arnd Bergmann6b960182009-11-06 23:10:54 -08003397 switch (cmd) {
3398 case SIOCSIFBR:
3399 case SIOCGIFBR:
3400 return old_bridge_ioctl(argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003401 case SIOCGIFCONF:
Al Viro36fd6332017-06-26 13:19:16 -04003402 return compat_dev_ifconf(net, argp);
Arnd Bergmann6b960182009-11-06 23:10:54 -08003403 case SIOCETHTOOL:
3404 return ethtool_ioctl(net, argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003405 case SIOCWANDEV:
3406 return compat_siocwandev(net, argp);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003407 case SIOCGIFMAP:
3408 case SIOCSIFMAP:
3409 return compat_sioc_ifmap(net, cmd, argp);
Arnd Bergmann0768e172019-04-17 22:56:11 +02003410 case SIOCGSTAMP_OLD:
3411 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003412 if (!sock->ops->gettstamp)
3413 return -ENOIOCTLCMD;
Arnd Bergmann0768e172019-04-17 22:56:11 +02003414 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003415 !COMPAT_USE_64BIT_TIME);
3416
Ben Hutchings590d4692013-11-18 17:04:13 +00003417 case SIOCBONDSLAVEINFOQUERY:
3418 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003419 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003420 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003421 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003422
Arnd Bergmann6b960182009-11-06 23:10:54 -08003423 case FIOSETOWN:
3424 case SIOCSPGRP:
3425 case FIOGETOWN:
3426 case SIOCGPGRP:
3427 case SIOCBRADDBR:
3428 case SIOCBRDELBR:
3429 case SIOCGIFVLAN:
3430 case SIOCSIFVLAN:
3431 case SIOCADDDLCI:
3432 case SIOCDELDLCI:
Andrey Vaginc62cce22016-10-24 18:29:13 -07003433 case SIOCGSKNS:
Arnd Bergmann0768e172019-04-17 22:56:11 +02003434 case SIOCGSTAMP_NEW:
3435 case SIOCGSTAMPNS_NEW:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003436 return sock_ioctl(file, cmd, arg);
3437
3438 case SIOCGIFFLAGS:
3439 case SIOCSIFFLAGS:
3440 case SIOCGIFMETRIC:
3441 case SIOCSIFMETRIC:
3442 case SIOCGIFMTU:
3443 case SIOCSIFMTU:
3444 case SIOCGIFMEM:
3445 case SIOCSIFMEM:
3446 case SIOCGIFHWADDR:
3447 case SIOCSIFHWADDR:
3448 case SIOCADDMULTI:
3449 case SIOCDELMULTI:
3450 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003451 case SIOCGIFADDR:
3452 case SIOCSIFADDR:
3453 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003454 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003455 case SIOCGIFBRDADDR:
3456 case SIOCSIFBRDADDR:
3457 case SIOCGIFDSTADDR:
3458 case SIOCSIFDSTADDR:
3459 case SIOCGIFNETMASK:
3460 case SIOCSIFNETMASK:
3461 case SIOCSIFPFLAGS:
3462 case SIOCGIFPFLAGS:
3463 case SIOCGIFTXQLEN:
3464 case SIOCSIFTXQLEN:
3465 case SIOCBRADDIF:
3466 case SIOCBRDELIF:
Johannes Bergc6c9fee2019-01-25 22:43:19 +01003467 case SIOCGIFNAME:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003468 case SIOCSIFNAME:
3469 case SIOCGMIIPHY:
3470 case SIOCGMIIREG:
3471 case SIOCSMIIREG:
Al Virof92d4fc2017-09-30 19:32:17 -04003472 case SIOCBONDENSLAVE:
3473 case SIOCBONDRELEASE:
3474 case SIOCBONDSETHWADDR:
3475 case SIOCBONDCHANGEACTIVE:
Johannes Berg37ac39b2019-01-25 22:43:18 +01003476 return compat_ifreq_ioctl(net, sock, cmd, argp);
3477
Arnd Bergmann6b960182009-11-06 23:10:54 -08003478 case SIOCSARP:
3479 case SIOCGARP:
3480 case SIOCDARP:
Arnd Bergmannc7dc5042019-06-03 23:07:12 +02003481 case SIOCOUTQ:
Arnd Bergmann9d7bf412019-06-03 23:06:00 +02003482 case SIOCOUTQNSD:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003483 case SIOCATMARK:
Johannes Berg63ff03a2019-01-25 22:43:17 +01003484 return sock_do_ioctl(net, sock, cmd, arg);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003485 }
3486
Arnd Bergmann6b960182009-11-06 23:10:54 -08003487 return -ENOIOCTLCMD;
3488}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003489
Eric Dumazet95c96172012-04-15 05:58:06 +00003490static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003491 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003492{
3493 struct socket *sock = file->private_data;
3494 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003495 struct sock *sk;
3496 struct net *net;
3497
3498 sk = sock->sk;
3499 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003500
3501 if (sock->ops->compat_ioctl)
3502 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3503
David S. Miller87de87d2008-06-03 09:14:03 -07003504 if (ret == -ENOIOCTLCMD &&
3505 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3506 ret = compat_wext_handle_ioctl(net, cmd, arg);
3507
Arnd Bergmann6b960182009-11-06 23:10:54 -08003508 if (ret == -ENOIOCTLCMD)
3509 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3510
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003511 return ret;
3512}
3513#endif
3514
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003515/**
3516 * kernel_bind - bind an address to a socket (kernel space)
3517 * @sock: socket
3518 * @addr: address
3519 * @addrlen: length of address
3520 *
3521 * Returns 0 or an error.
3522 */
3523
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003524int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3525{
3526 return sock->ops->bind(sock, addr, addrlen);
3527}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003528EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003529
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003530/**
3531 * kernel_listen - move socket to listening state (kernel space)
3532 * @sock: socket
3533 * @backlog: pending connections queue size
3534 *
3535 * Returns 0 or an error.
3536 */
3537
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003538int kernel_listen(struct socket *sock, int backlog)
3539{
3540 return sock->ops->listen(sock, backlog);
3541}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003542EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003543
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003544/**
3545 * kernel_accept - accept a connection (kernel space)
3546 * @sock: listening socket
3547 * @newsock: new connected socket
3548 * @flags: flags
3549 *
3550 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3551 * If it fails, @newsock is guaranteed to be %NULL.
3552 * Returns 0 or an error.
3553 */
3554
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003555int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3556{
3557 struct sock *sk = sock->sk;
3558 int err;
3559
3560 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3561 newsock);
3562 if (err < 0)
3563 goto done;
3564
David Howellscdfbabf2017-03-09 08:09:05 +00003565 err = sock->ops->accept(sock, *newsock, flags, true);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003566 if (err < 0) {
3567 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003568 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003569 goto done;
3570 }
3571
3572 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003573 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003574
3575done:
3576 return err;
3577}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003578EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003579
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003580/**
3581 * kernel_connect - connect a socket (kernel space)
3582 * @sock: socket
3583 * @addr: address
3584 * @addrlen: address length
3585 * @flags: flags (O_NONBLOCK, ...)
3586 *
3587 * For datagram sockets, @addr is the addres to which datagrams are sent
3588 * by default, and the only address from which datagrams are received.
3589 * For stream sockets, attempts to connect to @addr.
3590 * Returns 0 or an error code.
3591 */
3592
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003593int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003594 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003595{
3596 return sock->ops->connect(sock, addr, addrlen, flags);
3597}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003598EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003599
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003600/**
3601 * kernel_getsockname - get the address which the socket is bound (kernel space)
3602 * @sock: socket
3603 * @addr: address holder
3604 *
3605 * Fills the @addr pointer with the address which the socket is bound.
3606 * Returns 0 or an error code.
3607 */
3608
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003609int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003610{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003611 return sock->ops->getname(sock, addr, 0);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003612}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003613EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003614
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003615/**
3616 * kernel_peername - get the address which the socket is connected (kernel space)
3617 * @sock: socket
3618 * @addr: address holder
3619 *
3620 * Fills the @addr pointer with the address which the socket is connected.
3621 * Returns 0 or an error code.
3622 */
3623
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003624int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003625{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003626 return sock->ops->getname(sock, addr, 1);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003627}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003628EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003629
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003630/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003631 * kernel_sendpage - send a &page through a socket (kernel space)
3632 * @sock: socket
3633 * @page: page
3634 * @offset: page offset
3635 * @size: total size in bytes
3636 * @flags: flags (MSG_DONTWAIT, ...)
3637 *
3638 * Returns the total amount sent in bytes or an error.
3639 */
3640
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003641int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3642 size_t size, int flags)
3643{
3644 if (sock->ops->sendpage)
3645 return sock->ops->sendpage(sock, page, offset, size, flags);
3646
3647 return sock_no_sendpage(sock, page, offset, size, flags);
3648}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003649EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003650
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003651/**
3652 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3653 * @sk: sock
3654 * @page: page
3655 * @offset: page offset
3656 * @size: total size in bytes
3657 * @flags: flags (MSG_DONTWAIT, ...)
3658 *
3659 * Returns the total amount sent in bytes or an error.
3660 * Caller must hold @sk.
3661 */
3662
Tom Herbert306b13e2017-07-28 16:22:41 -07003663int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3664 size_t size, int flags)
3665{
3666 struct socket *sock = sk->sk_socket;
3667
3668 if (sock->ops->sendpage_locked)
3669 return sock->ops->sendpage_locked(sk, page, offset, size,
3670 flags);
3671
3672 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3673}
3674EXPORT_SYMBOL(kernel_sendpage_locked);
3675
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003676/**
3677 * kernel_shutdown - shut down part of a full-duplex connection (kernel space)
3678 * @sock: socket
3679 * @how: connection part
3680 *
3681 * Returns 0 or an error.
3682 */
3683
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003684int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3685{
3686 return sock->ops->shutdown(sock, how);
3687}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003688EXPORT_SYMBOL(kernel_sock_shutdown);
R. Parameswaran113c3072017-04-05 16:50:35 -07003689
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003690/**
3691 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3692 * @sk: socket
3693 *
3694 * This routine returns the IP overhead imposed by a socket i.e.
3695 * the length of the underlying IP header, depending on whether
3696 * this is an IPv4 or IPv6 socket and the length from IP options turned
3697 * on at the socket. Assumes that the caller has a lock on the socket.
R. Parameswaran113c3072017-04-05 16:50:35 -07003698 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003699
R. Parameswaran113c3072017-04-05 16:50:35 -07003700u32 kernel_sock_ip_overhead(struct sock *sk)
3701{
3702 struct inet_sock *inet;
3703 struct ip_options_rcu *opt;
3704 u32 overhead = 0;
R. Parameswaran113c3072017-04-05 16:50:35 -07003705#if IS_ENABLED(CONFIG_IPV6)
3706 struct ipv6_pinfo *np;
3707 struct ipv6_txoptions *optv6 = NULL;
3708#endif /* IS_ENABLED(CONFIG_IPV6) */
3709
3710 if (!sk)
3711 return overhead;
3712
R. Parameswaran113c3072017-04-05 16:50:35 -07003713 switch (sk->sk_family) {
3714 case AF_INET:
3715 inet = inet_sk(sk);
3716 overhead += sizeof(struct iphdr);
3717 opt = rcu_dereference_protected(inet->inet_opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003718 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003719 if (opt)
3720 overhead += opt->opt.optlen;
3721 return overhead;
3722#if IS_ENABLED(CONFIG_IPV6)
3723 case AF_INET6:
3724 np = inet6_sk(sk);
3725 overhead += sizeof(struct ipv6hdr);
3726 if (np)
3727 optv6 = rcu_dereference_protected(np->opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003728 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003729 if (optv6)
3730 overhead += (optv6->opt_flen + optv6->opt_nflen);
3731 return overhead;
3732#endif /* IS_ENABLED(CONFIG_IPV6) */
3733 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3734 return overhead;
3735 }
3736}
3737EXPORT_SYMBOL(kernel_sock_ip_overhead);