blob: 0b87fcf97830807328aebeecc5caa334d567c2c7 [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
Jakub Kicinskicc698372020-11-20 14:50:52 -080055#include <linux/ethtool.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070056#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/socket.h>
58#include <linux/file.h>
59#include <linux/net.h>
60#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070061#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070062#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070063#include <linux/netdevice.h>
64#include <linux/proc_fs.h>
65#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080066#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030068#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>
Yangbo Lud7c08822021-06-30 16:12:00 +0800107#include <linux/ptp_clock_kernel.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300108
Cong Wange0d10952013-08-01 11:10:25 +0800109#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300110unsigned int sysctl_net_busy_read __read_mostly;
111unsigned int sysctl_net_busy_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300112#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800113
Al Viro8ae5e0302014-11-28 19:40:50 -0500114static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
115static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700116static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118static int sock_close(struct inode *inode, struct file *file);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -0700119static __poll_t sock_poll(struct file *file,
120 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700121static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800122#ifdef CONFIG_COMPAT
123static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700124 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800125#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127static ssize_t sock_sendpage(struct file *file, struct page *page,
128 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800129static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700130 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800131 unsigned int flags);
Arnd Bergmann542d3062020-01-08 22:44:43 +0100132
133#ifdef CONFIG_PROC_FS
134static void sock_show_fdinfo(struct seq_file *m, struct file *f)
135{
136 struct socket *sock = f->private_data;
137
138 if (sock->ops->show_fdinfo)
139 sock->ops->show_fdinfo(m, sock);
140}
141#else
142#define sock_show_fdinfo NULL
143#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145/*
146 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
147 * in the operation structures but are done directly via the socketcall() multiplexor.
148 */
149
Arjan van de Venda7071d2007-02-12 00:55:36 -0800150static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151 .owner = THIS_MODULE,
152 .llseek = no_llseek,
Al Viro8ae5e0302014-11-28 19:40:50 -0500153 .read_iter = sock_read_iter,
154 .write_iter = sock_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 .poll = sock_poll,
156 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800157#ifdef CONFIG_COMPAT
158 .compat_ioctl = compat_sock_ioctl,
159#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 .mmap = sock_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 .release = sock_close,
162 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200163 .sendpage = sock_sendpage,
164 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800165 .splice_read = sock_splice_read,
Kirill Tkhaib4653342019-12-09 13:03:40 +0300166 .show_fdinfo = sock_show_fdinfo,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167};
168
Yejune Dengfe0bdbd2021-06-21 13:12:25 +0800169static const char * const pf_family_names[] = {
170 [PF_UNSPEC] = "PF_UNSPEC",
171 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
172 [PF_INET] = "PF_INET",
173 [PF_AX25] = "PF_AX25",
174 [PF_IPX] = "PF_IPX",
175 [PF_APPLETALK] = "PF_APPLETALK",
176 [PF_NETROM] = "PF_NETROM",
177 [PF_BRIDGE] = "PF_BRIDGE",
178 [PF_ATMPVC] = "PF_ATMPVC",
179 [PF_X25] = "PF_X25",
180 [PF_INET6] = "PF_INET6",
181 [PF_ROSE] = "PF_ROSE",
182 [PF_DECnet] = "PF_DECnet",
183 [PF_NETBEUI] = "PF_NETBEUI",
184 [PF_SECURITY] = "PF_SECURITY",
185 [PF_KEY] = "PF_KEY",
186 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
187 [PF_PACKET] = "PF_PACKET",
188 [PF_ASH] = "PF_ASH",
189 [PF_ECONET] = "PF_ECONET",
190 [PF_ATMSVC] = "PF_ATMSVC",
191 [PF_RDS] = "PF_RDS",
192 [PF_SNA] = "PF_SNA",
193 [PF_IRDA] = "PF_IRDA",
194 [PF_PPPOX] = "PF_PPPOX",
195 [PF_WANPIPE] = "PF_WANPIPE",
196 [PF_LLC] = "PF_LLC",
197 [PF_IB] = "PF_IB",
198 [PF_MPLS] = "PF_MPLS",
199 [PF_CAN] = "PF_CAN",
200 [PF_TIPC] = "PF_TIPC",
201 [PF_BLUETOOTH] = "PF_BLUETOOTH",
202 [PF_IUCV] = "PF_IUCV",
203 [PF_RXRPC] = "PF_RXRPC",
204 [PF_ISDN] = "PF_ISDN",
205 [PF_PHONET] = "PF_PHONET",
206 [PF_IEEE802154] = "PF_IEEE802154",
207 [PF_CAIF] = "PF_CAIF",
208 [PF_ALG] = "PF_ALG",
209 [PF_NFC] = "PF_NFC",
210 [PF_VSOCK] = "PF_VSOCK",
211 [PF_KCM] = "PF_KCM",
212 [PF_QIPCRTR] = "PF_QIPCRTR",
213 [PF_SMC] = "PF_SMC",
214 [PF_XDP] = "PF_XDP",
Jeremy Kerrbc49d812021-07-29 10:20:39 +0800215 [PF_MCTP] = "PF_MCTP",
Yejune Dengfe0bdbd2021-06-21 13:12:25 +0800216};
217
Linus Torvalds1da177e2005-04-16 15:20:36 -0700218/*
219 * The protocol list. Each protocol is registered in here.
220 */
221
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000223static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700226 * Support routines.
227 * Move socket addresses back and forth across the kernel/user
228 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229 */
230
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231/**
232 * move_addr_to_kernel - copy a socket address into kernel space
233 * @uaddr: Address in user space
234 * @kaddr: Address in kernel space
235 * @ulen: Length in user space
236 *
237 * The address is copied into kernel space. If the provided address is
238 * too long an error code of -EINVAL is returned. If the copy gives
239 * invalid addresses -EFAULT is returned. On a success 0 is returned.
240 */
241
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000242int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700244 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700246 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700248 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100250 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251}
252
253/**
254 * move_addr_to_user - copy an address to user space
255 * @kaddr: kernel space address
256 * @klen: length of address in kernel
257 * @uaddr: user space address
258 * @ulen: pointer to user length field
259 *
260 * The value pointed to by ulen on entry is the buffer length available.
261 * This is overwritten with the buffer space used. -EINVAL is returned
262 * if an overlong buffer is specified or a negative buffer size. -EFAULT
263 * is returned if either the buffer or the length field are not
264 * accessible.
265 * After copying the data up to the limit the user specifies, the true
266 * length of the data is written over the length limit the user
267 * specified. Zero is returned for a success.
268 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700269
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000270static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000271 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272{
273 int err;
274 int len;
275
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100276 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700277 err = get_user(len, ulen);
278 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700280 if (len > klen)
281 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100282 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700284 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500285 if (audit_sockaddr(klen, kaddr))
286 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700287 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 return -EFAULT;
289 }
290 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700291 * "fromlen shall refer to the value before truncation.."
292 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 */
294 return __put_user(klen, ulen);
295}
296
Alexey Dobriyan08009a72018-02-24 21:20:33 +0300297static struct kmem_cache *sock_inode_cachep __ro_after_init;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298
299static struct inode *sock_alloc_inode(struct super_block *sb)
300{
301 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700302
Christoph Lametere94b1762006-12-06 20:33:17 -0800303 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 if (!ei)
305 return NULL;
Al Viro333f7902019-07-05 20:14:16 +0100306 init_waitqueue_head(&ei->socket.wq.wait);
307 ei->socket.wq.fasync_list = NULL;
308 ei->socket.wq.flags = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 ei->socket.state = SS_UNCONNECTED;
311 ei->socket.flags = 0;
312 ei->socket.ops = NULL;
313 ei->socket.sk = NULL;
314 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315
316 return &ei->vfs_inode;
317}
318
Al Viro6d7855c2019-07-05 20:13:22 +0100319static void sock_free_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320{
Eric Dumazet43815482010-04-29 11:01:49 +0000321 struct socket_alloc *ei;
322
323 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazet43815482010-04-29 11:01:49 +0000324 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325}
326
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700327static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700329 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330
Christoph Lametera35afb82007-05-16 22:10:57 -0700331 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700333
yuan linyu1e911632017-01-07 17:18:31 +0800334static void init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335{
336 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700337 sizeof(struct socket_alloc),
338 0,
339 (SLAB_HWCACHE_ALIGN |
340 SLAB_RECLAIM_ACCOUNT |
Vladimir Davydov5d097052016-01-14 15:18:21 -0800341 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
Paul Mundt20c2df82007-07-20 10:11:58 +0900342 init_once);
yuan linyu1e911632017-01-07 17:18:31 +0800343 BUG_ON(sock_inode_cachep == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344}
345
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700346static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700347 .alloc_inode = sock_alloc_inode,
Al Viro6d7855c2019-07-05 20:13:22 +0100348 .free_inode = sock_free_inode,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700349 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350};
351
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700352/*
353 * sockfs_dname() is called from d_path().
354 */
355static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
356{
357 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000358 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700359}
360
Al Viro3ba13d12009-02-20 06:02:22 +0000361static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700362 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363};
364
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200365static int sockfs_xattr_get(const struct xattr_handler *handler,
366 struct dentry *dentry, struct inode *inode,
367 const char *suffix, void *value, size_t size)
368{
369 if (value) {
370 if (dentry->d_name.len + 1 > size)
371 return -ERANGE;
372 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
373 }
374 return dentry->d_name.len + 1;
375}
376
377#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
378#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
379#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
380
381static const struct xattr_handler sockfs_xattr_handler = {
382 .name = XATTR_NAME_SOCKPROTONAME,
383 .get = sockfs_xattr_get,
384};
385
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100386static int sockfs_security_xattr_set(const struct xattr_handler *handler,
Christian Braunere65ce2a2021-01-21 14:19:27 +0100387 struct user_namespace *mnt_userns,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100388 struct dentry *dentry, struct inode *inode,
389 const char *suffix, const void *value,
390 size_t size, int flags)
391{
392 /* Handled by LSM. */
393 return -EAGAIN;
394}
395
396static const struct xattr_handler sockfs_security_xattr_handler = {
397 .prefix = XATTR_SECURITY_PREFIX,
398 .set = sockfs_security_xattr_set,
399};
400
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200401static const struct xattr_handler *sockfs_xattr_handlers[] = {
402 &sockfs_xattr_handler,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100403 &sockfs_security_xattr_handler,
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200404 NULL
405};
406
David Howellsfba9be42019-03-25 16:38:24 +0000407static int sockfs_init_fs_context(struct fs_context *fc)
Al Viroc74a1cb2011-01-12 16:59:34 -0500408{
David Howellsfba9be42019-03-25 16:38:24 +0000409 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
410 if (!ctx)
411 return -ENOMEM;
412 ctx->ops = &sockfs_ops;
413 ctx->dops = &sockfs_dentry_operations;
414 ctx->xattr = sockfs_xattr_handlers;
415 return 0;
Al Viroc74a1cb2011-01-12 16:59:34 -0500416}
417
418static struct vfsmount *sock_mnt __read_mostly;
419
420static struct file_system_type sock_fs_type = {
421 .name = "sockfs",
David Howellsfba9be42019-03-25 16:38:24 +0000422 .init_fs_context = sockfs_init_fs_context,
Al Viroc74a1cb2011-01-12 16:59:34 -0500423 .kill_sb = kill_anon_super,
424};
425
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426/*
427 * Obtains the first available file descriptor and sets it up for use.
428 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800429 * These functions create file structures and maps them to fd space
430 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 * and file struct implicitly stored in sock->file.
432 * Note that another thread may close file descriptor before we return
433 * from this function. We use the fact that now we do not refer
434 * to socket after mapping. If one day we will need it, this
435 * function will increment ref. count on file by 1.
436 *
437 * In any case returned fd MAY BE not valid!
438 * This race condition is unavoidable
439 * with shared fd spaces, we cannot solve it inside kernel,
440 * but we take care of internal coherence yet.
441 */
442
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300443/**
444 * sock_alloc_file - Bind a &socket to a &file
445 * @sock: socket
446 * @flags: file status flags
447 * @dname: protocol name
448 *
449 * Returns the &file bound with @sock, implicitly storing it
450 * in sock->file. If dname is %NULL, sets to "".
451 * On failure the return is a ERR pointer (see linux/err.h).
452 * This function uses GFP_KERNEL internally.
453 */
454
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700455struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456{
Al Viro7cbe66b2009-08-05 19:59:08 +0400457 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800458
Al Virod93aa9d2018-06-09 09:40:05 -0400459 if (!dname)
460 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461
Al Virod93aa9d2018-06-09 09:40:05 -0400462 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
463 O_RDWR | (flags & O_NONBLOCK),
464 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530465 if (IS_ERR(file)) {
Al Viro8e1611e2017-12-05 23:29:09 +0000466 sock_release(sock);
Anatol Pomozov39b65252012-09-12 20:11:55 -0700467 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400468 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800469
470 sock->file = file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800471 file->private_data = sock;
Linus Torvaldsd8e464e2019-11-17 11:20:48 -0800472 stream_open(SOCK_INODE(sock), file);
Al Viro28407632012-08-17 23:54:15 -0400473 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800474}
Al Viro56b31d12012-08-18 00:25:51 -0400475EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800476
Al Viro56b31d12012-08-18 00:25:51 -0400477static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800478{
479 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400480 int fd = get_unused_fd_flags(flags);
Al Viroce4bb042018-01-10 18:47:05 -0500481 if (unlikely(fd < 0)) {
482 sock_release(sock);
Al Viro28407632012-08-17 23:54:15 -0400483 return fd;
Al Viroce4bb042018-01-10 18:47:05 -0500484 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800485
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700486 newfile = sock_alloc_file(sock, flags, NULL);
Enrico Weigelt4546e442019-06-05 22:58:50 +0200487 if (!IS_ERR(newfile)) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800488 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400489 return fd;
490 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400491
Al Viro28407632012-08-17 23:54:15 -0400492 put_unused_fd(fd);
493 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494}
495
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300496/**
497 * sock_from_file - Return the &socket bounded to @file.
498 * @file: file
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300499 *
Florent Revestdba4a922020-12-04 12:36:04 +0100500 * On failure returns %NULL.
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300501 */
502
Florent Revestdba4a922020-12-04 12:36:04 +0100503struct socket *sock_from_file(struct file *file)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800504{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800505 if (file->f_op == &socket_file_ops)
506 return file->private_data; /* set in sock_map_fd */
507
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800508 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800509}
John Fastabend406a3c62012-07-20 10:39:25 +0000510EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800511
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700513 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 * @fd: file handle
515 * @err: pointer to an error code return
516 *
517 * The file handle passed in is locked and the socket it is bound
Rosen, Rami241c4662017-05-21 22:12:38 +0300518 * to is returned. If an error occurs the err pointer is overwritten
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 * with a negative errno code and NULL is returned. The function checks
520 * for both invalid handles and passing a handle which is not a socket.
521 *
522 * On a success the socket object pointer is returned.
523 */
524
525struct socket *sockfd_lookup(int fd, int *err)
526{
527 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 struct socket *sock;
529
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700530 file = fget(fd);
531 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 *err = -EBADF;
533 return NULL;
534 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700535
Florent Revestdba4a922020-12-04 12:36:04 +0100536 sock = sock_from_file(file);
537 if (!sock) {
538 *err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 fput(file);
Florent Revestdba4a922020-12-04 12:36:04 +0100540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 return sock;
542}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700543EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800545static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
546{
Al Viro00e188e2014-03-03 23:48:18 -0500547 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800548 struct socket *sock;
549
Hua Zhong36725582006-04-19 15:25:02 -0700550 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500551 if (f.file) {
Florent Revestdba4a922020-12-04 12:36:04 +0100552 sock = sock_from_file(f.file);
Al Viro00e188e2014-03-03 23:48:18 -0500553 if (likely(sock)) {
Miaohe Lince787a52020-08-06 19:53:16 +0800554 *fput_needed = f.flags & FDPUT_FPUT;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800555 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500556 }
Florent Revestdba4a922020-12-04 12:36:04 +0100557 *err = -ENOTSOCK;
Al Viro00e188e2014-03-03 23:48:18 -0500558 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800559 }
560 return NULL;
561}
562
Masatake YAMATO600e1772012-08-29 10:44:29 +0000563static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
564 size_t size)
565{
566 ssize_t len;
567 ssize_t used = 0;
568
David Howellsc5ef6032015-03-17 22:26:16 +0000569 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000570 if (len < 0)
571 return len;
572 used += len;
573 if (buffer) {
574 if (size < used)
575 return -ERANGE;
576 buffer += len;
577 }
578
579 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
580 used += len;
581 if (buffer) {
582 if (size < used)
583 return -ERANGE;
584 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
585 buffer += len;
586 }
587
588 return used;
589}
590
Christian Brauner549c7292021-01-21 14:19:43 +0100591static int sockfs_setattr(struct user_namespace *mnt_userns,
592 struct dentry *dentry, struct iattr *iattr)
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900593{
Christian Brauner549c7292021-01-21 14:19:43 +0100594 int err = simple_setattr(&init_user_ns, dentry, iattr);
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900595
Eric Biggerse1a3a602016-12-30 17:42:32 -0600596 if (!err && (iattr->ia_valid & ATTR_UID)) {
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900597 struct socket *sock = SOCKET_I(d_inode(dentry));
598
Cong Wang6d8c50d2018-06-07 13:39:49 -0700599 if (sock->sk)
600 sock->sk->sk_uid = iattr->ia_uid;
601 else
602 err = -ENOENT;
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900603 }
604
605 return err;
606}
607
Masatake YAMATO600e1772012-08-29 10:44:29 +0000608static const struct inode_operations sockfs_inode_ops = {
Masatake YAMATO600e1772012-08-29 10:44:29 +0000609 .listxattr = sockfs_listxattr,
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900610 .setattr = sockfs_setattr,
Masatake YAMATO600e1772012-08-29 10:44:29 +0000611};
612
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300614 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700615 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 * Allocate a new inode and socket object. The two are bound together
617 * and initialised. The socket is then returned. If we are out of inodes
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300618 * NULL is returned. This functions uses GFP_KERNEL internally.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 */
620
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800621struct socket *sock_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700623 struct inode *inode;
624 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200626 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 if (!inode)
628 return NULL;
629
630 sock = SOCKET_I(inode);
631
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400632 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700633 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100634 inode->i_uid = current_fsuid();
635 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000636 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 return sock;
639}
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800640EXPORT_SYMBOL(sock_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
Cong Wang6d8c50d2018-06-07 13:39:49 -0700642static void __sock_release(struct socket *sock, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643{
644 if (sock->ops) {
645 struct module *owner = sock->ops->owner;
646
Cong Wang6d8c50d2018-06-07 13:39:49 -0700647 if (inode)
648 inode_lock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649 sock->ops->release(sock);
Eric Biggersff7b11a2019-02-21 14:13:56 -0800650 sock->sk = NULL;
Cong Wang6d8c50d2018-06-07 13:39:49 -0700651 if (inode)
652 inode_unlock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 sock->ops = NULL;
654 module_put(owner);
655 }
656
Al Viro333f7902019-07-05 20:14:16 +0100657 if (sock->wq.fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800658 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 if (!sock->file) {
661 iput(SOCK_INODE(sock));
662 return;
663 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700664 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665}
Cong Wang6d8c50d2018-06-07 13:39:49 -0700666
Andrew Lunn9a8ad9a2020-07-13 01:15:12 +0200667/**
668 * sock_release - close a socket
669 * @sock: socket to close
670 *
671 * The socket is released from the protocol stack if it has a release
672 * callback, and the inode is then released if the socket is bound to
673 * an inode not a file.
674 */
Cong Wang6d8c50d2018-06-07 13:39:49 -0700675void sock_release(struct socket *sock)
676{
677 __sock_release(sock, NULL);
678}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700679EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400681void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000682{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200683 u8 flags = *tx_flags;
684
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400685 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200686 flags |= SKBTX_HW_TSTAMP;
687
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400688 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200689 flags |= SKBTX_SW_TSTAMP;
690
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400691 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200692 flags |= SKBTX_SCHED_TSTAMP;
693
Eric Dumazet140c55d2014-08-06 11:49:29 +0200694 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000695}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400696EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000697
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200698INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
699 size_t));
Paolo Abenia648a592019-07-03 16:06:54 +0200700INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
701 size_t));
Al Virod8725c82014-12-11 00:02:50 -0500702static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703{
Paolo Abenia648a592019-07-03 16:06:54 +0200704 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
705 inet_sendmsg, sock, msg,
706 msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500707 BUG_ON(ret == -EIOCBQUEUED);
708 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800710
Randy Dunlap85806af2019-05-18 21:23:07 -0700711/**
712 * sock_sendmsg - send a message through @sock
713 * @sock: socket
714 * @msg: message to send
715 *
716 * Sends @msg through @sock, passing through LSM.
717 * Returns the number of bytes sent, or an error code.
718 */
Al Virod8725c82014-12-11 00:02:50 -0500719int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800720{
Al Virod8725c82014-12-11 00:02:50 -0500721 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500722 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800723
Al Virod8725c82014-12-11 00:02:50 -0500724 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800725}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700726EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300728/**
729 * kernel_sendmsg - send a message through @sock (kernel-space)
730 * @sock: socket
731 * @msg: message header
732 * @vec: kernel vec
733 * @num: vec array length
734 * @size: total message data size
735 *
736 * Builds the message data with @vec and sends it through @sock.
737 * Returns the number of bytes sent, or an error code.
738 */
739
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
741 struct kvec *vec, size_t num, size_t size)
742{
David Howellsaa563d72018-10-20 00:57:56 +0100743 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500744 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700746EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300748/**
749 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
750 * @sk: sock
751 * @msg: message header
752 * @vec: output s/g array
753 * @num: output s/g array length
754 * @size: total message data size
755 *
756 * Builds the message data with @vec and sends it through @sock.
757 * Returns the number of bytes sent, or an error code.
758 * Caller must hold @sk.
759 */
760
Tom Herbert306b13e2017-07-28 16:22:41 -0700761int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
762 struct kvec *vec, size_t num, size_t size)
763{
764 struct socket *sock = sk->sk_socket;
765
766 if (!sock->ops->sendmsg_locked)
John Fastabenddb5980d2017-08-15 22:31:34 -0700767 return sock_no_sendmsg_locked(sk, msg, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700768
David Howellsaa563d72018-10-20 00:57:56 +0100769 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700770
771 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
772}
773EXPORT_SYMBOL(kernel_sendmsg_locked);
774
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400775static bool skb_is_err_queue(const struct sk_buff *skb)
776{
777 /* pkt_type of skbs enqueued on the error queue are set to
778 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
779 * in recvmsg, since skbs received on a local socket will never
780 * have a pkt_type of PACKET_OUTGOING.
781 */
782 return skb->pkt_type == PACKET_OUTGOING;
783}
784
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200785/* On transmit, software and hardware timestamps are returned independently.
786 * As the two skb clones share the hardware timestamp, which may be updated
787 * before the software timestamp is received, a hardware TX timestamp may be
788 * returned only if there is no software TX timestamp. Ignore false software
789 * timestamps, which may be made in the __sock_recv_timestamp() call when the
Deepa Dinamani7f1bc6e2019-02-02 07:34:46 -0800790 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200791 * hardware timestamp.
792 */
793static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
794{
795 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
796}
797
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200798static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
799{
800 struct scm_ts_pktinfo ts_pktinfo;
801 struct net_device *orig_dev;
802
803 if (!skb_mac_header_was_set(skb))
804 return;
805
806 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
807
808 rcu_read_lock();
809 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
810 if (orig_dev)
811 ts_pktinfo.if_index = orig_dev->ifindex;
812 rcu_read_unlock();
813
814 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
815 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
816 sizeof(ts_pktinfo), &ts_pktinfo);
817}
818
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700819/*
820 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
821 */
822void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
823 struct sk_buff *skb)
824{
Patrick Ohly20d49472009-02-12 05:03:38 +0000825 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Deepa Dinamani887feae2019-02-02 07:34:50 -0800826 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
Deepa Dinamani97184752019-02-02 07:34:51 -0800827 struct scm_timestamping_internal tss;
828
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200829 int empty = 1, false_tstamp = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000830 struct skb_shared_hwtstamps *shhwtstamps =
831 skb_hwtstamps(skb);
Miroslav Lichvarff53c862022-01-05 11:33:26 +0100832 ktime_t hwtstamp;
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700833
Patrick Ohly20d49472009-02-12 05:03:38 +0000834 /* Race occurred between timestamp enabling and packet
835 receiving. Fill in the current time for now. */
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200836 if (need_software_tstamp && skb->tstamp == 0) {
Patrick Ohly20d49472009-02-12 05:03:38 +0000837 __net_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200838 false_tstamp = 1;
839 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000840
841 if (need_software_tstamp) {
842 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800843 if (new_tstamp) {
844 struct __kernel_sock_timeval tv;
845
846 skb_get_new_timestamp(skb, &tv);
847 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
848 sizeof(tv), &tv);
849 } else {
850 struct __kernel_old_timeval tv;
851
852 skb_get_timestamp(skb, &tv);
853 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
854 sizeof(tv), &tv);
855 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000856 } else {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800857 if (new_tstamp) {
858 struct __kernel_timespec ts;
859
860 skb_get_new_timestampns(skb, &ts);
861 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
862 sizeof(ts), &ts);
863 } else {
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +0200864 struct __kernel_old_timespec ts;
Deepa Dinamani887feae2019-02-02 07:34:50 -0800865
866 skb_get_timestampns(skb, &ts);
867 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
868 sizeof(ts), &ts);
869 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000870 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700871 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000872
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400873 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400874 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Deepa Dinamani97184752019-02-02 07:34:51 -0800875 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000876 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400877 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400878 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Yangbo Lud7c08822021-06-30 16:12:00 +0800879 !skb_is_swtx_tstamp(skb, false_tstamp)) {
880 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
Miroslav Lichvarff53c862022-01-05 11:33:26 +0100881 hwtstamp = ptp_convert_timestamp(shhwtstamps,
882 sk->sk_bind_phc);
883 else
884 hwtstamp = shhwtstamps->hwtstamp;
Yangbo Lud7c08822021-06-30 16:12:00 +0800885
Miroslav Lichvarff53c862022-01-05 11:33:26 +0100886 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
Yangbo Lud7c08822021-06-30 16:12:00 +0800887 empty = 0;
888
889 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
890 !skb_is_err_queue(skb))
891 put_ts_pktinfo(msg, skb);
892 }
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200893 }
Francis Yan1c885802016-11-27 23:07:18 -0800894 if (!empty) {
Deepa Dinamani97184752019-02-02 07:34:51 -0800895 if (sock_flag(sk, SOCK_TSTAMP_NEW))
896 put_cmsg_scm_timestamping64(msg, &tss);
897 else
898 put_cmsg_scm_timestamping(msg, &tss);
Francis Yan1c885802016-11-27 23:07:18 -0800899
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400900 if (skb_is_err_queue(skb) && skb->len &&
Soheil Hassas Yeganeh4ef1b282017-03-18 17:03:00 -0400901 SKB_EXT_ERR(skb)->opt_stats)
Francis Yan1c885802016-11-27 23:07:18 -0800902 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
903 skb->len, skb->data);
904 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700905}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300906EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
907
Johannes Berg6e3e9392011-11-09 10:15:42 +0100908void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
909 struct sk_buff *skb)
910{
911 int ack;
912
913 if (!sock_flag(sk, SOCK_WIFI_STATUS))
914 return;
915 if (!skb->wifi_acked_valid)
916 return;
917
918 ack = skb->wifi_acked;
919
920 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
921}
922EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
923
stephen hemminger11165f12010-10-18 14:27:29 +0000924static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
925 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700926{
Eyal Birger744d5a32015-03-01 14:58:31 +0200927 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700928 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200929 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700930}
931
Erin MacNeil645bc812022-04-27 16:02:37 -0400932static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
933 struct sk_buff *skb)
934{
935 if (sock_flag(sk, SOCK_RCVMARK) && skb)
936 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32),
937 &skb->mark);
938}
939
940void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
941 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700942{
943 sock_recv_timestamp(msg, sk, skb);
944 sock_recv_drops(msg, sk, skb);
Erin MacNeil645bc812022-04-27 16:02:37 -0400945 sock_recv_mark(msg, sk, skb);
Neil Horman3b885782009-10-12 13:26:31 -0700946}
Erin MacNeil645bc812022-04-27 16:02:37 -0400947EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
Neil Horman3b885782009-10-12 13:26:31 -0700948
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200949INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
Paolo Abenia648a592019-07-03 16:06:54 +0200950 size_t, int));
951INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
952 size_t, int));
Ying Xue1b784142015-03-02 15:37:48 +0800953static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -0400954 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955{
Paolo Abenia648a592019-07-03 16:06:54 +0200956 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
957 inet_recvmsg, sock, msg, msg_data_left(msg),
958 flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959}
960
Randy Dunlap85806af2019-05-18 21:23:07 -0700961/**
962 * sock_recvmsg - receive a message from @sock
963 * @sock: socket
964 * @msg: message to receive
965 * @flags: message flags
966 *
967 * Receives @msg from @sock, passing through LSM. Returns the total number
968 * of bytes received, or an error.
969 */
Al Viro2da62902015-03-14 21:13:46 -0400970int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700971{
Al Viro2da62902015-03-14 21:13:46 -0400972 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700973
Al Viro2da62902015-03-14 21:13:46 -0400974 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700976EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000978/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300979 * kernel_recvmsg - Receive a message from a socket (kernel space)
980 * @sock: The socket to receive the message from
981 * @msg: Received message
982 * @vec: Input s/g array for message data
983 * @num: Size of input s/g array
984 * @size: Number of bytes to read
985 * @flags: Message flags (MSG_DONTWAIT, etc...)
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000986 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300987 * On return the msg structure contains the scatter/gather array passed in the
988 * vec argument. The array is modified so that it consists of the unfilled
989 * portion of the original array.
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000990 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300991 * The returned value is the total number of bytes received, or an error.
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000992 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300993
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700994int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
995 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996{
Christoph Hellwig1f466e12020-05-11 13:59:13 +0200997 msg->msg_control_is_user = false;
David Howellsaa563d72018-10-20 00:57:56 +0100998 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
Christoph Hellwig1f466e12020-05-11 13:59:13 +0200999 return sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001001EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001003static ssize_t sock_sendpage(struct file *file, struct page *page,
1004 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005{
1006 struct socket *sock;
1007 int flags;
1008
Eric Dumazetb69aee02005-09-06 14:42:45 -07001009 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Eric Dumazet35f9c092012-04-05 03:05:35 +00001011 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1012 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1013 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
Linus Torvaldse6949582009-08-13 08:28:36 -07001015 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016}
1017
Jens Axboe9c55e012007-11-06 23:30:13 -08001018static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001019 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -08001020 unsigned int flags)
1021{
1022 struct socket *sock = file->private_data;
1023
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001024 if (unlikely(!sock->ops->splice_read))
Slavomir Kaslev95506582018-11-16 11:27:53 +02001025 return generic_file_splice_read(file, ppos, pipe, len, flags);
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001026
Jens Axboe9c55e012007-11-06 23:30:13 -08001027 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1028}
1029
Al Viro8ae5e0302014-11-28 19:40:50 -05001030static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001031{
Al Viro6d652332015-01-30 16:12:56 -05001032 struct file *file = iocb->ki_filp;
1033 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001034 struct msghdr msg = {.msg_iter = *to,
1035 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001036 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001037
Jens Axboeebfcd892019-12-09 20:58:56 -07001038 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001039 msg.msg_flags = MSG_DONTWAIT;
1040
1041 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001042 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001043
Christoph Hellwig66ee59a2015-02-11 19:56:46 +01001044 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001045 return 0;
1046
Al Viro2da62902015-03-14 21:13:46 -04001047 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e0302014-11-28 19:40:50 -05001048 *to = msg.msg_iter;
1049 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001050}
1051
Al Viro8ae5e0302014-11-28 19:40:50 -05001052static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
Al Viro6d652332015-01-30 16:12:56 -05001054 struct file *file = iocb->ki_filp;
1055 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001056 struct msghdr msg = {.msg_iter = *from,
1057 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001058 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001059
Al Viro8ae5e0302014-11-28 19:40:50 -05001060 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001061 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001062
Jens Axboeebfcd892019-12-09 20:58:56 -07001063 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001064 msg.msg_flags = MSG_DONTWAIT;
1065
Al Viro6d652332015-01-30 16:12:56 -05001066 if (sock->type == SOCK_SEQPACKET)
1067 msg.msg_flags |= MSG_EOR;
1068
Al Virod8725c82014-12-11 00:02:50 -05001069 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e0302014-11-28 19:40:50 -05001070 *from = msg.msg_iter;
1071 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072}
1073
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074/*
1075 * Atomic setting of ioctl hooks to avoid race
1076 * with module unload.
1077 */
1078
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001079static DEFINE_MUTEX(br_ioctl_mutex);
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001080static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1081 unsigned int cmd, struct ifreq *ifr,
1082 void __user *uarg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001084void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1085 unsigned int cmd, struct ifreq *ifr,
1086 void __user *uarg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001088 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001090 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091}
1092EXPORT_SYMBOL(brioctl_set);
1093
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001094int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1095 struct ifreq *ifr, void __user *uarg)
1096{
1097 int err = -ENOPKG;
1098
1099 if (!br_ioctl_hook)
1100 request_module("bridge");
1101
1102 mutex_lock(&br_ioctl_mutex);
1103 if (br_ioctl_hook)
1104 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1105 mutex_unlock(&br_ioctl_mutex);
1106
1107 return err;
1108}
1109
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001110static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001111static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Eric W. Biederman881d9662007-09-17 11:56:21 -07001113void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001115 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001117 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118}
1119EXPORT_SYMBOL(vlan_ioctl_set);
1120
Arnd Bergmann6b960182009-11-06 23:10:54 -08001121static long sock_do_ioctl(struct net *net, struct socket *sock,
Johannes Berg63ff03a2019-01-25 22:43:17 +01001122 unsigned int cmd, unsigned long arg)
Arnd Bergmann6b960182009-11-06 23:10:54 -08001123{
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001124 struct ifreq ifr;
1125 bool need_copyout;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001126 int err;
1127 void __user *argp = (void __user *)arg;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001128 void __user *data;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001129
1130 err = sock->ops->ioctl(sock, cmd, arg);
1131
1132 /*
1133 * If this ioctl is unknown try to hand it down
1134 * to the NIC driver.
1135 */
Al Viro36fd6332017-06-26 13:19:16 -04001136 if (err != -ENOIOCTLCMD)
1137 return err;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001138
Jakub Kicinski29ce8f92021-08-31 09:06:04 -07001139 if (!is_socket_ioctl_cmd(cmd))
1140 return -ENOTTY;
1141
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001142 if (get_user_ifreq(&ifr, &data, argp))
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001143 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001144 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001145 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001146 if (put_user_ifreq(&ifr, argp))
Al Viro36fd6332017-06-26 13:19:16 -04001147 return -EFAULT;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001148
Arnd Bergmann6b960182009-11-06 23:10:54 -08001149 return err;
1150}
1151
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152/*
1153 * With an ioctl, arg may well be a user mode pointer, but we don't know
1154 * what to do with it - that's up to the protocol still.
1155 */
1156
1157static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1158{
1159 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001160 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 void __user *argp = (void __user *)arg;
1162 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001163 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
Eric Dumazetb69aee02005-09-06 14:42:45 -07001165 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001166 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001167 net = sock_net(sk);
Al Viro44c02a22017-10-05 12:59:44 -04001168 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1169 struct ifreq ifr;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001170 void __user *data;
Al Viro44c02a22017-10-05 12:59:44 -04001171 bool need_copyout;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001172 if (get_user_ifreq(&ifr, &data, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001173 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001174 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Al Viro44c02a22017-10-05 12:59:44 -04001175 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001176 if (put_user_ifreq(&ifr, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001177 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001179#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Al Virob1b0c242017-10-01 20:13:08 -04001181 err = wext_handle_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001183#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001184 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 case FIOSETOWN:
1186 case SIOCSPGRP:
1187 err = -EFAULT;
1188 if (get_user(pid, (int __user *)argp))
1189 break;
Jiri Slaby393cc3f2017-06-13 13:35:50 +02001190 err = f_setown(sock->file, pid, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 break;
1192 case FIOGETOWN:
1193 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001194 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001195 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 break;
1197 case SIOCGIFBR:
1198 case SIOCSIFBR:
1199 case SIOCBRADDBR:
1200 case SIOCBRDELBR:
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001201 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202 break;
1203 case SIOCGIFVLAN:
1204 case SIOCSIFVLAN:
1205 err = -ENOPKG;
1206 if (!vlan_ioctl_hook)
1207 request_module("8021q");
1208
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001209 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001211 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001212 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 break;
Andrey Vaginc62cce22016-10-24 18:29:13 -07001214 case SIOCGSKNS:
1215 err = -EPERM;
1216 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1217 break;
1218
1219 err = open_related_ns(&net->ns, get_net_ns);
1220 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001221 case SIOCGSTAMP_OLD:
1222 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001223 if (!sock->ops->gettstamp) {
1224 err = -ENOIOCTLCMD;
1225 break;
1226 }
1227 err = sock->ops->gettstamp(sock, argp,
Arnd Bergmann0768e172019-04-17 22:56:11 +02001228 cmd == SIOCGSTAMP_OLD,
1229 !IS_ENABLED(CONFIG_64BIT));
Gustavo A. R. Silva60747822019-04-24 10:31:24 -05001230 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001231 case SIOCGSTAMP_NEW:
1232 case SIOCGSTAMPNS_NEW:
1233 if (!sock->ops->gettstamp) {
1234 err = -ENOIOCTLCMD;
1235 break;
1236 }
1237 err = sock->ops->gettstamp(sock, argp,
1238 cmd == SIOCGSTAMP_NEW,
1239 false);
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001240 break;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001241
1242 case SIOCGIFCONF:
1243 err = dev_ifconf(net, argp);
1244 break;
1245
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 default:
Johannes Berg63ff03a2019-01-25 22:43:17 +01001247 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001249 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 return err;
1251}
1252
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001253/**
1254 * sock_create_lite - creates a socket
1255 * @family: protocol family (AF_INET, ...)
1256 * @type: communication type (SOCK_STREAM, ...)
1257 * @protocol: protocol (0, ...)
1258 * @res: new socket
1259 *
1260 * Creates a new socket and assigns it to @res, passing through LSM.
1261 * The new socket initialization is not complete, see kernel_accept().
1262 * Returns 0 or an error. On failure @res is set to %NULL.
1263 * This function internally uses GFP_KERNEL.
1264 */
1265
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266int sock_create_lite(int family, int type, int protocol, struct socket **res)
1267{
1268 int err;
1269 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001270
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 err = security_socket_create(family, type, protocol, 1);
1272 if (err)
1273 goto out;
1274
1275 sock = sock_alloc();
1276 if (!sock) {
1277 err = -ENOMEM;
1278 goto out;
1279 }
1280
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001282 err = security_socket_post_create(sock, family, type, protocol, 1);
1283 if (err)
1284 goto out_release;
1285
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286out:
1287 *res = sock;
1288 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001289out_release:
1290 sock_release(sock);
1291 sock = NULL;
1292 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001294EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
1296/* No kernel lock held - perfect */
Al Viroade994f2017-07-03 00:01:49 -04001297static __poll_t sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298{
Christoph Hellwig3cafb372018-01-09 16:07:34 +01001299 struct socket *sock = file->private_data;
Christoph Hellwiga331de32018-07-30 09:42:13 +02001300 __poll_t events = poll_requested_events(wait), flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301
Christoph Hellwige88958e2018-06-29 15:37:24 +02001302 if (!sock->ops->poll)
1303 return 0;
Christoph Hellwigf641f13b2018-07-30 09:42:12 +02001304
Christoph Hellwiga331de32018-07-30 09:42:13 +02001305 if (sk_can_busy_loop(sock->sk)) {
1306 /* poll once if requested by the syscall */
1307 if (events & POLL_BUSY_LOOP)
1308 sk_busy_loop(sock->sk, 1);
1309
1310 /* if this socket can poll_ll, tell the system call */
1311 flag = POLL_BUSY_LOOP;
1312 }
1313
1314 return sock->ops->poll(file, sock, wait) | flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315}
1316
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001317static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001319 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320
1321 return sock->ops->mmap(file, sock, vma);
1322}
1323
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001324static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325{
Cong Wang6d8c50d2018-06-07 13:39:49 -07001326 __sock_release(SOCKET_I(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 return 0;
1328}
1329
1330/*
1331 * Update the socket async list
1332 *
1333 * Fasync_list locking strategy.
1334 *
1335 * 1. fasync_list is modified only under process context socket lock
1336 * i.e. under semaphore.
1337 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001338 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 */
1340
1341static int sock_fasync(int fd, struct file *filp, int on)
1342{
Eric Dumazet989a2972010-04-14 09:55:35 +00001343 struct socket *sock = filp->private_data;
1344 struct sock *sk = sock->sk;
Al Viro333f7902019-07-05 20:14:16 +01001345 struct socket_wq *wq = &sock->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346
Eric Dumazet989a2972010-04-14 09:55:35 +00001347 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
1350 lock_sock(sk);
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001351 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001353 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001354 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001355 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001356 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357
Eric Dumazet989a2972010-04-14 09:55:35 +00001358 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 return 0;
1360}
1361
Eric Dumazetceb5d582015-11-29 20:03:11 -08001362/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
Eric Dumazetceb5d582015-11-29 20:03:11 -08001364int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001366 if (!wq || !wq->fasync_list)
1367 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001368
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001369 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001370 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001371 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 break;
1373 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001374 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001375 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 break;
Miaohe Lin7c7ab582020-08-08 16:23:10 +08001377 fallthrough;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001378 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001379call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001380 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001382 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001383 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001385
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 return 0;
1387}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001388EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001390/**
1391 * __sock_create - creates a socket
1392 * @net: net namespace
1393 * @family: protocol family (AF_INET, ...)
1394 * @type: communication type (SOCK_STREAM, ...)
1395 * @protocol: protocol (0, ...)
1396 * @res: new socket
1397 * @kern: boolean for kernel space sockets
1398 *
1399 * Creates a new socket and assigns it to @res, passing through LSM.
1400 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1401 * be set to true if the socket resides in kernel space.
1402 * This function internally uses GFP_KERNEL.
1403 */
1404
Pavel Emelyanov721db932010-09-29 16:06:32 +04001405int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001406 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407{
1408 int err;
1409 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001410 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411
1412 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001413 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 */
1415 if (family < 0 || family >= NPROTO)
1416 return -EAFNOSUPPORT;
1417 if (type < 0 || type >= SOCK_MAX)
1418 return -EINVAL;
1419
1420 /* Compatibility.
1421
1422 This uglymoron is moved from INET layer to here to avoid
1423 deadlock in module load.
1424 */
1425 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001426 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1427 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 family = PF_PACKET;
1429 }
1430
1431 err = security_socket_create(family, type, protocol, kern);
1432 if (err)
1433 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001434
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001435 /*
1436 * Allocate the socket and allow the family to set things up. if
1437 * the protocol is 0, the family is instructed to select an appropriate
1438 * default.
1439 */
1440 sock = sock_alloc();
1441 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001442 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001443 return -ENFILE; /* Not exactly a match, but its the
1444 closest posix thing */
1445 }
1446
1447 sock->type = type;
1448
Johannes Berg95a5afc2008-10-16 15:24:51 -07001449#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001450 /* Attempt to load a protocol module if the find failed.
1451 *
1452 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 * requested real, full-featured networking support upon configuration.
1454 * Otherwise module support will break!
1455 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001456 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001457 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458#endif
1459
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001460 rcu_read_lock();
1461 pf = rcu_dereference(net_families[family]);
1462 err = -EAFNOSUPPORT;
1463 if (!pf)
1464 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
1466 /*
1467 * We will call the ->create function, that possibly is in a loadable
1468 * module, so we have to bump that loadable module refcnt first.
1469 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001470 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 goto out_release;
1472
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001473 /* Now protected by module ref count */
1474 rcu_read_unlock();
1475
Eric Paris3f378b62009-11-05 22:18:14 -08001476 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001477 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001479
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 /*
1481 * Now to bump the refcnt of the [loadable] module that owns this
1482 * socket at sock_release time we decrement its refcnt.
1483 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001484 if (!try_module_get(sock->ops->owner))
1485 goto out_module_busy;
1486
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 /*
1488 * Now that we're done with the ->create function, the [loadable]
1489 * module can have its refcnt decremented
1490 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001491 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001492 err = security_socket_post_create(sock, family, type, protocol, kern);
1493 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001494 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001495 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001497 return 0;
1498
1499out_module_busy:
1500 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001502 sock->ops = NULL;
1503 module_put(pf->owner);
1504out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001506 return err;
1507
1508out_release:
1509 rcu_read_unlock();
1510 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001512EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001514/**
1515 * sock_create - creates a socket
1516 * @family: protocol family (AF_INET, ...)
1517 * @type: communication type (SOCK_STREAM, ...)
1518 * @protocol: protocol (0, ...)
1519 * @res: new socket
1520 *
1521 * A wrapper around __sock_create().
1522 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1523 */
1524
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525int sock_create(int family, int type, int protocol, struct socket **res)
1526{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001527 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001529EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001531/**
1532 * sock_create_kern - creates a socket (kernel space)
1533 * @net: net namespace
1534 * @family: protocol family (AF_INET, ...)
1535 * @type: communication type (SOCK_STREAM, ...)
1536 * @protocol: protocol (0, ...)
1537 * @res: new socket
1538 *
1539 * A wrapper around __sock_create().
1540 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1541 */
1542
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001543int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001545 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001547EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001549int __sys_socket(int family, int type, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550{
1551 int retval;
1552 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001553 int flags;
1554
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001555 /* Check the SOCK_* constants for consistency. */
1556 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1557 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1558 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1559 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1560
Ulrich Dreppera677a032008-07-23 21:29:17 -07001561 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001562 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001563 return -EINVAL;
1564 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001566 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1567 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1568
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 retval = sock_create(family, type, protocol, &sock);
1570 if (retval < 0)
Al Viro8e1611e2017-12-05 23:29:09 +00001571 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
Al Viro8e1611e2017-12-05 23:29:09 +00001573 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574}
1575
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001576SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1577{
1578 return __sys_socket(family, type, protocol);
1579}
1580
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581/*
1582 * Create a pair of connected sockets.
1583 */
1584
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001585int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586{
1587 struct socket *sock1, *sock2;
1588 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001589 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001590 int flags;
1591
1592 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001593 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001594 return -EINVAL;
1595 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001597 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1598 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1599
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 /*
Al Viro016a2662017-12-05 23:28:38 +00001601 * reserve descriptors and make sure we won't fail
1602 * to return them to userland.
1603 */
1604 fd1 = get_unused_fd_flags(flags);
1605 if (unlikely(fd1 < 0))
1606 return fd1;
1607
1608 fd2 = get_unused_fd_flags(flags);
1609 if (unlikely(fd2 < 0)) {
1610 put_unused_fd(fd1);
1611 return fd2;
1612 }
1613
1614 err = put_user(fd1, &usockvec[0]);
1615 if (err)
1616 goto out;
1617
1618 err = put_user(fd2, &usockvec[1]);
1619 if (err)
1620 goto out;
1621
1622 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 * Obtain the first socket and check if the underlying protocol
1624 * supports the socketpair call.
1625 */
1626
1627 err = sock_create(family, type, protocol, &sock1);
Al Viro016a2662017-12-05 23:28:38 +00001628 if (unlikely(err < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 goto out;
1630
1631 err = sock_create(family, type, protocol, &sock2);
Al Viro016a2662017-12-05 23:28:38 +00001632 if (unlikely(err < 0)) {
1633 sock_release(sock1);
1634 goto out;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001635 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001636
David Herrmannd47cd942018-05-04 16:28:20 +02001637 err = security_socket_socketpair(sock1, sock2);
1638 if (unlikely(err)) {
1639 sock_release(sock2);
1640 sock_release(sock1);
1641 goto out;
1642 }
1643
Al Viro016a2662017-12-05 23:28:38 +00001644 err = sock1->ops->socketpair(sock1, sock2);
1645 if (unlikely(err < 0)) {
1646 sock_release(sock2);
1647 sock_release(sock1);
1648 goto out;
Al Viro28407632012-08-17 23:54:15 -04001649 }
1650
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001651 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301652 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001653 err = PTR_ERR(newfile1);
Al Viro016a2662017-12-05 23:28:38 +00001654 sock_release(sock2);
1655 goto out;
Al Viro28407632012-08-17 23:54:15 -04001656 }
1657
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001658 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001659 if (IS_ERR(newfile2)) {
1660 err = PTR_ERR(newfile2);
Al Viro016a2662017-12-05 23:28:38 +00001661 fput(newfile1);
1662 goto out;
Al Virodb349502007-02-07 01:48:00 -05001663 }
1664
Al Viro157cf642008-12-14 04:57:47 -05001665 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001666
Al Virodb349502007-02-07 01:48:00 -05001667 fd_install(fd1, newfile1);
1668 fd_install(fd2, newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001669 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671out:
Al Viro016a2662017-12-05 23:28:38 +00001672 put_unused_fd(fd2);
1673 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674 return err;
1675}
1676
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001677SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1678 int __user *, usockvec)
1679{
1680 return __sys_socketpair(family, type, protocol, usockvec);
1681}
1682
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683/*
1684 * Bind a name to a socket. Nothing much to do here since it's
1685 * the protocol's responsibility to handle the local address.
1686 *
1687 * We move the socket address to kernel space before we call
1688 * the protocol layer (having also checked the address is ok).
1689 */
1690
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001691int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692{
1693 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001694 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001695 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001697 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001698 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001699 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Jakub Sitnicki068b88c2018-10-04 11:09:40 +02001700 if (!err) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001701 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001702 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001703 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001704 if (!err)
1705 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001706 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001707 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001709 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001710 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 return err;
1712}
1713
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001714SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1715{
1716 return __sys_bind(fd, umyaddr, addrlen);
1717}
1718
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719/*
1720 * Perform a listen. Basically, we allow the protocol to do anything
1721 * necessary for a listen, and if that works, we mark the socket as
1722 * ready for listening.
1723 */
1724
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001725int __sys_listen(int fd, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726{
1727 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001728 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001729 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001730
1731 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1732 if (sock) {
Kuniyuki Iwashimacb9eaed2022-08-23 10:47:00 -07001733 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
Eric Dumazet95c96172012-04-15 05:58:06 +00001734 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001735 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
1737 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001738 if (!err)
1739 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001741 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 }
1743 return err;
1744}
1745
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001746SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1747{
1748 return __sys_listen(fd, backlog);
1749}
1750
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001751struct file *do_accept(struct file *file, unsigned file_flags,
Jens Axboede2ea4b2019-10-17 14:41:29 -06001752 struct sockaddr __user *upeer_sockaddr,
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001753 int __user *upeer_addrlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754{
1755 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001756 struct file *newfile;
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001757 int err, len;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001758 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
Florent Revestdba4a922020-12-04 12:36:04 +01001760 sock = sock_from_file(file);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001761 if (!sock)
1762 return ERR_PTR(-ENOTSOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001764 newsock = sock_alloc();
1765 if (!newsock)
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001766 return ERR_PTR(-ENFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767
1768 newsock->type = sock->type;
1769 newsock->ops = sock->ops;
1770
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 /*
1772 * We don't need try_module_get here, as the listening socket (sock)
1773 * has the protocol module (sock->ops->owner) held.
1774 */
1775 __module_get(newsock->ops->owner);
1776
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001777 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001778 if (IS_ERR(newfile))
1779 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001780
Frank Filza79af592005-09-27 15:23:38 -07001781 err = security_socket_accept(sock, newsock);
1782 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001783 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001784
Jens Axboede2ea4b2019-10-17 14:41:29 -06001785 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1786 false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001788 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789
1790 if (upeer_sockaddr) {
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001791 len = newsock->ops->getname(newsock,
1792 (struct sockaddr *)&address, 2);
1793 if (len < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001795 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001797 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001798 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001800 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 }
1802
1803 /* File flags are not inherited via accept() unlike another OSes. */
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001804 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001805out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001806 fput(newfile);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001807 return ERR_PTR(err);
1808}
Jens Axboede2ea4b2019-10-17 14:41:29 -06001809
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001810int __sys_accept4_file(struct file *file, unsigned file_flags,
1811 struct sockaddr __user *upeer_sockaddr,
1812 int __user *upeer_addrlen, int flags,
1813 unsigned long nofile)
1814{
1815 struct file *newfile;
1816 int newfd;
1817
1818 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1819 return -EINVAL;
1820
1821 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1822 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1823
1824 newfd = __get_unused_fd_flags(flags, nofile);
1825 if (unlikely(newfd < 0))
1826 return newfd;
1827
1828 newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen,
1829 flags);
1830 if (IS_ERR(newfile)) {
1831 put_unused_fd(newfd);
1832 return PTR_ERR(newfile);
1833 }
1834 fd_install(newfd, newfile);
1835 return newfd;
Jens Axboede2ea4b2019-10-17 14:41:29 -06001836}
1837
1838/*
1839 * For accept, we attempt to create a new socket, set up the link
1840 * with the client, wake up the client, then return the new
1841 * connected fd. We collect the address of the connector in kernel
1842 * space and move it to user at the very end. This is unclean because
1843 * we open the socket then return an error.
1844 *
1845 * 1003.1g adds the ability to recvmsg() to query connection pending
1846 * status to recvmsg. We need to add that support in a way thats
1847 * clean when we restructure accept also.
1848 */
1849
1850int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1851 int __user *upeer_addrlen, int flags)
1852{
1853 int ret = -EBADF;
1854 struct fd f;
1855
1856 f = fdget(fd);
1857 if (f.file) {
1858 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
Jens Axboe09952e32020-03-19 20:16:56 -06001859 upeer_addrlen, flags,
1860 rlimit(RLIMIT_NOFILE));
Miaohe Lin6b07ede2020-08-06 19:52:24 +08001861 fdput(f);
Jens Axboede2ea4b2019-10-17 14:41:29 -06001862 }
1863
1864 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865}
1866
Dominik Brodowski4541e802018-03-13 19:24:23 +01001867SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1868 int __user *, upeer_addrlen, int, flags)
1869{
1870 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1871}
1872
Heiko Carstens20f37032009-01-14 14:14:23 +01001873SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1874 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001875{
Dominik Brodowski4541e802018-03-13 19:24:23 +01001876 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001877}
1878
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879/*
1880 * Attempt to connect to a socket with the server address. The address
1881 * is in user space so we verify it is OK and move it to kernel space.
1882 *
1883 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1884 * break bindings
1885 *
1886 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1887 * other SEQPACKET protocols that take time to connect() as it doesn't
1888 * include the -EINPROGRESS status for such sockets.
1889 */
1890
Jens Axboef499a022019-12-02 16:28:46 -07001891int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
Jens Axboebd3ded32019-11-23 14:17:16 -07001892 int addrlen, int file_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893{
1894 struct socket *sock;
Jens Axboebd3ded32019-11-23 14:17:16 -07001895 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896
Florent Revestdba4a922020-12-04 12:36:04 +01001897 sock = sock_from_file(file);
1898 if (!sock) {
1899 err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 goto out;
Florent Revestdba4a922020-12-04 12:36:04 +01001901 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001903 err =
Jens Axboef499a022019-12-02 16:28:46 -07001904 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905 if (err)
Jens Axboebd3ded32019-11-23 14:17:16 -07001906 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907
Jens Axboef499a022019-12-02 16:28:46 -07001908 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Jens Axboebd3ded32019-11-23 14:17:16 -07001909 sock->file->f_flags | file_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910out:
1911 return err;
1912}
1913
Jens Axboebd3ded32019-11-23 14:17:16 -07001914int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1915{
1916 int ret = -EBADF;
1917 struct fd f;
1918
1919 f = fdget(fd);
1920 if (f.file) {
Jens Axboef499a022019-12-02 16:28:46 -07001921 struct sockaddr_storage address;
1922
1923 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1924 if (!ret)
1925 ret = __sys_connect_file(f.file, &address, addrlen, 0);
Miaohe Lin6b07ede2020-08-06 19:52:24 +08001926 fdput(f);
Jens Axboebd3ded32019-11-23 14:17:16 -07001927 }
1928
1929 return ret;
1930}
1931
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01001932SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1933 int, addrlen)
1934{
1935 return __sys_connect(fd, uservaddr, addrlen);
1936}
1937
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938/*
1939 * Get the local address ('name') of a socket object. Move the obtained
1940 * name to user space.
1941 */
1942
Dominik Brodowski8882a102018-03-13 19:43:14 +01001943int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1944 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945{
1946 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001947 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001948 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001949
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001950 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 if (!sock)
1952 goto out;
1953
1954 err = security_socket_getsockname(sock);
1955 if (err)
1956 goto out_put;
1957
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001958 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1959 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 goto out_put;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001961 /* "err" is actually length in this case */
1962 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963
1964out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001965 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966out:
1967 return err;
1968}
1969
Dominik Brodowski8882a102018-03-13 19:43:14 +01001970SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1971 int __user *, usockaddr_len)
1972{
1973 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1974}
1975
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976/*
1977 * Get the remote address ('name') of a socket object. Move the obtained
1978 * name to user space.
1979 */
1980
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01001981int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1982 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983{
1984 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001985 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001986 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001988 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1989 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990 err = security_socket_getpeername(sock);
1991 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001992 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 return err;
1994 }
1995
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001996 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1997 if (err >= 0)
1998 /* "err" is actually length in this case */
1999 err = move_addr_to_user(&address, err, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002000 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002001 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 }
2003 return err;
2004}
2005
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002006SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2007 int __user *, usockaddr_len)
2008{
2009 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2010}
2011
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012/*
2013 * Send a datagram to a given address. We move the address into kernel
2014 * space and check the user space data area is readable before invoking
2015 * the protocol.
2016 */
Dominik Brodowski211b6342018-03-13 19:18:52 +01002017int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2018 struct sockaddr __user *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019{
2020 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002021 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 int err;
2023 struct msghdr msg;
2024 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002025 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002026
Al Viro602bd0e2015-03-21 19:12:32 -04002027 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
2028 if (unlikely(err))
2029 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002030 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2031 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08002032 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002033
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002034 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002035 msg.msg_control = NULL;
2036 msg.msg_controllen = 0;
2037 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002038 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002039 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 if (err < 0)
2041 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002042 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002043 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 }
2045 if (sock->file->f_flags & O_NONBLOCK)
2046 flags |= MSG_DONTWAIT;
2047 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05002048 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002050out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002051 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08002052out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 return err;
2054}
2055
Dominik Brodowski211b6342018-03-13 19:18:52 +01002056SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2057 unsigned int, flags, struct sockaddr __user *, addr,
2058 int, addr_len)
2059{
2060 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2061}
2062
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002064 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 */
2066
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002067SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00002068 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069{
Dominik Brodowski211b6342018-03-13 19:18:52 +01002070 return __sys_sendto(fd, buff, len, flags, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071}
2072
2073/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002074 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075 * sender. We verify the buffers are writable and if needed move the
2076 * sender address from kernel to user space.
2077 */
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002078int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2079 struct sockaddr __user *addr, int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080{
2081 struct socket *sock;
2082 struct iovec iov;
2083 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002084 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002085 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002086 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087
Al Viro602bd0e2015-03-21 19:12:32 -04002088 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2089 if (unlikely(err))
2090 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002091 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002093 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002095 msg.msg_control = NULL;
2096 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002097 /* Save some cycles and don't copy the address if not needed */
2098 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2099 /* We assume all kernel code knows the size of sockaddr_storage */
2100 msg.msg_namelen = 0;
tadeusz.struk@intel.com130ed5d2015-12-15 10:46:17 -08002101 msg.msg_iocb = NULL;
Alexander Potapenko9f138fa2017-03-08 18:08:16 +01002102 msg.msg_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 if (sock->file->f_flags & O_NONBLOCK)
2104 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002105 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002107 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002108 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002109 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002110 if (err2 < 0)
2111 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002113
2114 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08002115out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116 return err;
2117}
2118
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002119SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2120 unsigned int, flags, struct sockaddr __user *, addr,
2121 int __user *, addr_len)
2122{
2123 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2124}
2125
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002127 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 */
2129
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02002130SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2131 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132{
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002133 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134}
2135
Florian Westphal83f0c102020-07-05 01:30:15 +02002136static bool sock_use_custom_sol_socket(const struct socket *sock)
2137{
2138 const struct sock *sk = sock->sk;
2139
2140 /* Use sock->ops->setsockopt() for MPTCP */
2141 return IS_ENABLED(CONFIG_MPTCP) &&
2142 sk->sk_protocol == IPPROTO_MPTCP &&
2143 sk->sk_type == SOCK_STREAM &&
2144 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2145}
2146
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147/*
2148 * Set a socket option. Because we don't know the option lengths we have
2149 * to pass the user mode parameter for the protocols to sort out.
2150 */
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002151int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002152 int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153{
Christoph Hellwig519a8a62020-08-10 18:42:14 +02002154 sockptr_t optval = USER_SOCKPTR(user_optval);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002155 char *kernel_optval = NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002156 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 struct socket *sock;
2158
2159 if (optlen < 0)
2160 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002161
2162 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002163 if (!sock)
2164 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165
Christoph Hellwig4a367292020-07-17 08:23:11 +02002166 err = security_socket_setsockopt(sock, level, optname);
2167 if (err)
2168 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002169
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002170 if (!in_compat_syscall())
2171 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002172 user_optval, &optlen,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002173 &kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002174 if (err < 0)
2175 goto out_put;
2176 if (err > 0) {
2177 err = 0;
2178 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179 }
Christoph Hellwig4a367292020-07-17 08:23:11 +02002180
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002181 if (kernel_optval)
2182 optval = KERNEL_SOCKPTR(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002183 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002184 err = sock_setsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002185 else if (unlikely(!sock->ops->setsockopt))
2186 err = -EOPNOTSUPP;
Christoph Hellwig4a367292020-07-17 08:23:11 +02002187 else
2188 err = sock->ops->setsockopt(sock, level, optname, optval,
2189 optlen);
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002190 kfree(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002191out_put:
2192 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193 return err;
2194}
2195
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002196SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2197 char __user *, optval, int, optlen)
2198{
2199 return __sys_setsockopt(fd, level, optname, optval, optlen);
2200}
2201
Stanislav Fomichev9cacf812021-01-15 08:34:59 -08002202INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2203 int optname));
2204
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205/*
2206 * Get a socket option. Because we don't know the option lengths we have
2207 * to pass a user mode parameter for the protocols to sort out.
2208 */
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002209int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2210 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002212 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 struct socket *sock;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002214 int max_optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002216 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002217 if (!sock)
2218 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002220 err = security_socket_getsockopt(sock, level, optname);
2221 if (err)
2222 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002223
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002224 if (!in_compat_syscall())
2225 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002226
2227 if (level == SOL_SOCKET)
2228 err = sock_getsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002229 else if (unlikely(!sock->ops->getsockopt))
2230 err = -EOPNOTSUPP;
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002231 else
2232 err = sock->ops->getsockopt(sock, level, optname, optval,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002233 optlen);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002234
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002235 if (!in_compat_syscall())
2236 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2237 optval, optlen, max_optlen,
2238 err);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002239out_put:
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002240 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 return err;
2242}
2243
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002244SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2245 char __user *, optval, int __user *, optlen)
2246{
2247 return __sys_getsockopt(fd, level, optname, optval, optlen);
2248}
2249
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250/*
2251 * Shutdown a socket.
2252 */
2253
Jens Axboeb713c192020-09-05 11:13:35 -06002254int __sys_shutdown_sock(struct socket *sock, int how)
2255{
2256 int err;
2257
2258 err = security_socket_shutdown(sock, how);
2259 if (!err)
2260 err = sock->ops->shutdown(sock, how);
2261
2262 return err;
2263}
2264
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002265int __sys_shutdown(int fd, int how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002267 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268 struct socket *sock;
2269
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002270 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2271 if (sock != NULL) {
Jens Axboeb713c192020-09-05 11:13:35 -06002272 err = __sys_shutdown_sock(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002273 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 }
2275 return err;
2276}
2277
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002278SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2279{
2280 return __sys_shutdown(fd, how);
2281}
2282
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002283/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 * fields which are the same type (int / unsigned) on our platforms.
2285 */
2286#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2287#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2288#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2289
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002290struct used_address {
2291 struct sockaddr_storage name;
2292 unsigned int name_len;
2293};
2294
Jens Axboe0a384ab2020-02-27 08:11:20 -07002295int __copy_msghdr_from_user(struct msghdr *kmsg,
2296 struct user_msghdr __user *umsg,
2297 struct sockaddr __user **save_addr,
2298 struct iovec __user **uiov, size_t *nsegs)
Dan Carpenter1661bf32013-10-03 00:27:20 +03002299{
Al Viroffb07552017-06-27 19:32:04 -04002300 struct user_msghdr msg;
Al Viro08adb7d2014-11-10 20:23:13 -05002301 ssize_t err;
2302
Al Viroffb07552017-06-27 19:32:04 -04002303 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
Dan Carpenter1661bf32013-10-03 00:27:20 +03002304 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00002305
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002306 kmsg->msg_control_is_user = true;
2307 kmsg->msg_control_user = msg.msg_control;
Al Viroffb07552017-06-27 19:32:04 -04002308 kmsg->msg_controllen = msg.msg_controllen;
2309 kmsg->msg_flags = msg.msg_flags;
2310
2311 kmsg->msg_namelen = msg.msg_namelen;
2312 if (!msg.msg_name)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07002313 kmsg->msg_namelen = 0;
2314
Matthew Leachdbb490b2014-03-11 11:58:27 +00002315 if (kmsg->msg_namelen < 0)
2316 return -EINVAL;
2317
Dan Carpenter1661bf32013-10-03 00:27:20 +03002318 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03002319 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05002320
2321 if (save_addr)
Al Viroffb07552017-06-27 19:32:04 -04002322 *save_addr = msg.msg_name;
Al Viro08adb7d2014-11-10 20:23:13 -05002323
Al Viroffb07552017-06-27 19:32:04 -04002324 if (msg.msg_name && kmsg->msg_namelen) {
Al Viro08adb7d2014-11-10 20:23:13 -05002325 if (!save_addr) {
Paolo Abeni864d9662017-07-21 18:49:45 +02002326 err = move_addr_to_kernel(msg.msg_name,
2327 kmsg->msg_namelen,
Al Viro08adb7d2014-11-10 20:23:13 -05002328 kmsg->msg_name);
2329 if (err < 0)
2330 return err;
2331 }
2332 } else {
2333 kmsg->msg_name = NULL;
2334 kmsg->msg_namelen = 0;
2335 }
2336
Al Viroffb07552017-06-27 19:32:04 -04002337 if (msg.msg_iovlen > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05002338 return -EMSGSIZE;
2339
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002340 kmsg->msg_iocb = NULL;
Jens Axboe0a384ab2020-02-27 08:11:20 -07002341 *uiov = msg.msg_iov;
2342 *nsegs = msg.msg_iovlen;
2343 return 0;
2344}
2345
2346static int copy_msghdr_from_user(struct msghdr *kmsg,
2347 struct user_msghdr __user *umsg,
2348 struct sockaddr __user **save_addr,
2349 struct iovec **iov)
2350{
2351 struct user_msghdr msg;
2352 ssize_t err;
2353
2354 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2355 &msg.msg_iovlen);
2356 if (err)
2357 return err;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002358
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002359 err = import_iovec(save_addr ? READ : WRITE,
Al Viroffb07552017-06-27 19:32:04 -04002360 msg.msg_iov, msg.msg_iovlen,
Al Viroda184282015-03-21 19:29:06 -04002361 UIO_FASTIOV, iov, &kmsg->msg_iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002362 return err < 0 ? err : 0;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002363}
2364
Jens Axboe4257c8c2019-11-25 14:27:34 -07002365static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2366 unsigned int flags, struct used_address *used_address,
2367 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368{
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002369 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Amit Kushwaha846cc122016-12-08 18:21:53 +05302370 __aligned(sizeof(__kernel_size_t));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002371 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05002373 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05002374 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002375
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 err = -ENOBUFS;
2377
Anton Blanchard228e5482011-05-02 20:21:35 +00002378 if (msg_sys->msg_controllen > INT_MAX)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002379 goto out;
Tom Herbert28a94d82016-03-07 14:11:02 -08002380 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00002381 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002383 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002384 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002385 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002387 goto out;
Anton Blanchard228e5482011-05-02 20:21:35 +00002388 ctl_buf = msg_sys->msg_control;
2389 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 } else if (ctl_len) {
David S. Millerac4340f2017-01-04 13:24:19 -05002391 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2392 CMSG_ALIGN(sizeof(struct cmsghdr)));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002393 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002395 if (ctl_buf == NULL)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002396 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397 }
2398 err = -EFAULT;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002399 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002401 msg_sys->msg_control = ctl_buf;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002402 msg_sys->msg_control_is_user = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002404 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002405
2406 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002407 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002408 /*
2409 * If this is sendmmsg() and current destination address is same as
2410 * previously succeeded address, omit asking LSM's decision.
2411 * used_address->name_len is initialized to UINT_MAX so that the first
2412 * destination address never matches.
2413 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002414 if (used_address && msg_sys->msg_name &&
2415 used_address->name_len == msg_sys->msg_namelen &&
2416 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002417 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05002418 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002419 goto out_freectl;
2420 }
Al Virod8725c82014-12-11 00:02:50 -05002421 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002422 /*
2423 * If this is sendmmsg() and sending to current destination address was
2424 * successful, remember it.
2425 */
2426 if (used_address && err >= 0) {
2427 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002428 if (msg_sys->msg_name)
2429 memcpy(&used_address->name, msg_sys->msg_name,
2430 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002431 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002432
2433out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002434 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
Jens Axboe4257c8c2019-11-25 14:27:34 -07002436out:
2437 return err;
2438}
2439
Jens Axboe03b12302019-12-02 18:50:25 -07002440int sendmsg_copy_msghdr(struct msghdr *msg,
2441 struct user_msghdr __user *umsg, unsigned flags,
2442 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002443{
2444 int err;
2445
2446 if (flags & MSG_CMSG_COMPAT) {
2447 struct compat_msghdr __user *msg_compat;
2448
2449 msg_compat = (struct compat_msghdr __user *) umsg;
2450 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2451 } else {
2452 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2453 }
2454 if (err < 0)
2455 return err;
2456
2457 return 0;
2458}
2459
2460static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2461 struct msghdr *msg_sys, unsigned int flags,
2462 struct used_address *used_address,
2463 unsigned int allowed_msghdr_flags)
2464{
2465 struct sockaddr_storage address;
2466 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2467 ssize_t err;
2468
2469 msg_sys->msg_name = &address;
2470
2471 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2472 if (err < 0)
2473 return err;
2474
2475 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2476 allowed_msghdr_flags);
Al Viroda184282015-03-21 19:29:06 -04002477 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002478 return err;
2479}
2480
2481/*
2482 * BSD sendmsg interface
2483 */
Jens Axboe03b12302019-12-02 18:50:25 -07002484long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
Jens Axboe0fa03c62019-04-19 13:34:07 -06002485 unsigned int flags)
2486{
Jens Axboe03b12302019-12-02 18:50:25 -07002487 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
Jens Axboe0fa03c62019-04-19 13:34:07 -06002488}
Anton Blanchard228e5482011-05-02 20:21:35 +00002489
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002490long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2491 bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002492{
2493 int fput_needed, err;
2494 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002495 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002496
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002497 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2498 return -EINVAL;
2499
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002500 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002501 if (!sock)
2502 goto out;
2503
Tom Herbert28a94d82016-03-07 14:11:02 -08002504 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00002505
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002506 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002507out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002508 return err;
2509}
2510
Al Viro666547f2014-04-06 14:03:05 -04002511SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002512{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002513 return __sys_sendmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002514}
2515
Anton Blanchard228e5482011-05-02 20:21:35 +00002516/*
2517 * Linux sendmmsg interface
2518 */
2519
2520int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002521 unsigned int flags, bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002522{
2523 int fput_needed, err, datagrams;
2524 struct socket *sock;
2525 struct mmsghdr __user *entry;
2526 struct compat_mmsghdr __user *compat_entry;
2527 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002528 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002529 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002530
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002531 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2532 return -EINVAL;
2533
Anton Blanchard98382f42011-08-04 14:07:39 +00002534 if (vlen > UIO_MAXIOV)
2535 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002536
2537 datagrams = 0;
2538
2539 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2540 if (!sock)
2541 return err;
2542
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002543 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002544 entry = mmsg;
2545 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002546 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002547 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002548
2549 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002550 if (datagrams == vlen - 1)
2551 flags = oflags;
2552
Anton Blanchard228e5482011-05-02 20:21:35 +00002553 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002554 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002555 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002556 if (err < 0)
2557 break;
2558 err = __put_user(err, &compat_entry->msg_len);
2559 ++compat_entry;
2560 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002561 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002562 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002563 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002564 if (err < 0)
2565 break;
2566 err = put_user(err, &entry->msg_len);
2567 ++entry;
2568 }
2569
2570 if (err)
2571 break;
2572 ++datagrams;
Soheil Hassas Yeganeh30238982016-11-04 15:36:49 -04002573 if (msg_data_left(&msg_sys))
2574 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002575 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002576 }
2577
Anton Blanchard228e5482011-05-02 20:21:35 +00002578 fput_light(sock->file, fput_needed);
2579
Anton Blanchard728ffb82011-08-04 14:07:38 +00002580 /* We only return an error if no datagrams were able to be sent */
2581 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002582 return datagrams;
2583
Anton Blanchard228e5482011-05-02 20:21:35 +00002584 return err;
2585}
2586
2587SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2588 unsigned int, vlen, unsigned int, flags)
2589{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002590 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002591}
2592
Jens Axboe03b12302019-12-02 18:50:25 -07002593int recvmsg_copy_msghdr(struct msghdr *msg,
2594 struct user_msghdr __user *umsg, unsigned flags,
2595 struct sockaddr __user **uaddr,
2596 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002597{
2598 ssize_t err;
2599
2600 if (MSG_CMSG_COMPAT & flags) {
2601 struct compat_msghdr __user *msg_compat;
2602
2603 msg_compat = (struct compat_msghdr __user *) umsg;
2604 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2605 } else {
2606 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2607 }
2608 if (err < 0)
2609 return err;
2610
2611 return 0;
2612}
2613
2614static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2615 struct user_msghdr __user *msg,
2616 struct sockaddr __user *uaddr,
2617 unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002619 struct compat_msghdr __user *msg_compat =
Jens Axboe4257c8c2019-11-25 14:27:34 -07002620 (struct compat_msghdr __user *) msg;
2621 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2622 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002623 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002624 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002625 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002626
Al Viro08adb7d2014-11-10 20:23:13 -05002627 msg_sys->msg_name = &addr;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002628 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2629 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002630
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002631 /* We assume all kernel code knows the size of sockaddr_storage */
2632 msg_sys->msg_namelen = 0;
2633
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634 if (sock->file->f_flags & O_NONBLOCK)
2635 flags |= MSG_DONTWAIT;
Eric Dumazet1af66222019-12-06 09:38:36 -08002636
2637 if (unlikely(nosec))
2638 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2639 else
2640 err = sock_recvmsg(sock, msg_sys, flags);
2641
Linus Torvalds1da177e2005-04-16 15:20:36 -07002642 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002643 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002644 len = err;
2645
2646 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002647 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002648 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002649 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002650 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002651 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002653 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002654 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002656 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002658 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659 &msg_compat->msg_controllen);
2660 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002661 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662 &msg->msg_controllen);
2663 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002664 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002665 err = len;
Jens Axboe4257c8c2019-11-25 14:27:34 -07002666out:
2667 return err;
2668}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002669
Jens Axboe4257c8c2019-11-25 14:27:34 -07002670static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2671 struct msghdr *msg_sys, unsigned int flags, int nosec)
2672{
2673 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2674 /* user mode address pointers */
2675 struct sockaddr __user *uaddr;
2676 ssize_t err;
2677
2678 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2679 if (err < 0)
2680 return err;
2681
2682 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
Al Viroda184282015-03-21 19:29:06 -04002683 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002684 return err;
2685}
2686
2687/*
2688 * BSD recvmsg interface
2689 */
2690
Jens Axboe03b12302019-12-02 18:50:25 -07002691long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2692 struct user_msghdr __user *umsg,
2693 struct sockaddr __user *uaddr, unsigned int flags)
Jens Axboeaa1fa282019-04-19 13:38:09 -06002694{
Jens Axboe03b12302019-12-02 18:50:25 -07002695 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
Jens Axboeaa1fa282019-04-19 13:38:09 -06002696}
2697
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002698long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2699 bool forbid_cmsg_compat)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002700{
2701 int fput_needed, err;
2702 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002703 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002704
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002705 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2706 return -EINVAL;
2707
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002708 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002709 if (!sock)
2710 goto out;
2711
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002712 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002713
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002714 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002715out:
2716 return err;
2717}
2718
Al Viro666547f2014-04-06 14:03:05 -04002719SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002720 unsigned int, flags)
2721{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002722 return __sys_recvmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002723}
2724
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002725/*
2726 * Linux recvmmsg interface
2727 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002729static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2730 unsigned int vlen, unsigned int flags,
2731 struct timespec64 *timeout)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002732{
2733 int fput_needed, err, datagrams;
2734 struct socket *sock;
2735 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002736 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002737 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002738 struct timespec64 end_time;
2739 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002740
2741 if (timeout &&
2742 poll_select_set_timeout(&end_time, timeout->tv_sec,
2743 timeout->tv_nsec))
2744 return -EINVAL;
2745
2746 datagrams = 0;
2747
2748 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2749 if (!sock)
2750 return err;
2751
Soheil Hassas Yeganeh7797dc42018-02-27 18:22:40 -05002752 if (likely(!(flags & MSG_ERRQUEUE))) {
2753 err = sock_error(sock->sk);
2754 if (err) {
2755 datagrams = err;
2756 goto out_put;
2757 }
Maxime Jayate623a9e2017-02-21 18:35:51 +01002758 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002759
2760 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002761 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002762
2763 while (datagrams < vlen) {
2764 /*
2765 * No need to ask LSM for more than the first datagram.
2766 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002767 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002768 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002769 &msg_sys, flags & ~MSG_WAITFORONE,
2770 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002771 if (err < 0)
2772 break;
2773 err = __put_user(err, &compat_entry->msg_len);
2774 ++compat_entry;
2775 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002776 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002777 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002778 &msg_sys, flags & ~MSG_WAITFORONE,
2779 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002780 if (err < 0)
2781 break;
2782 err = put_user(err, &entry->msg_len);
2783 ++entry;
2784 }
2785
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002786 if (err)
2787 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002788 ++datagrams;
2789
Brandon L Black71c5c1592010-03-26 16:18:03 +00002790 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2791 if (flags & MSG_WAITFORONE)
2792 flags |= MSG_DONTWAIT;
2793
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002794 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002795 ktime_get_ts64(&timeout64);
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002796 *timeout = timespec64_sub(end_time, timeout64);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002797 if (timeout->tv_sec < 0) {
2798 timeout->tv_sec = timeout->tv_nsec = 0;
2799 break;
2800 }
2801
2802 /* Timeout, return less than vlen datagrams */
2803 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2804 break;
2805 }
2806
2807 /* Out of band data, return right away */
2808 if (msg_sys.msg_flags & MSG_OOB)
2809 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002810 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002811 }
2812
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002813 if (err == 0)
2814 goto out_put;
2815
2816 if (datagrams == 0) {
2817 datagrams = err;
2818 goto out_put;
2819 }
2820
2821 /*
2822 * We may return less entries than requested (vlen) if the
2823 * sock is non block and there aren't enough datagrams...
2824 */
2825 if (err != -EAGAIN) {
2826 /*
2827 * ... or if recvmsg returns an error after we
2828 * received some datagrams, where we record the
2829 * error to return on the next call or if the
2830 * app asks about it using getsockopt(SO_ERROR).
2831 */
2832 sock->sk->sk_err = -err;
2833 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002834out_put:
2835 fput_light(sock->file, fput_needed);
2836
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002837 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002838}
2839
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002840int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2841 unsigned int vlen, unsigned int flags,
2842 struct __kernel_timespec __user *timeout,
2843 struct old_timespec32 __user *timeout32)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002844{
2845 int datagrams;
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002846 struct timespec64 timeout_sys;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002847
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002848 if (timeout && get_timespec64(&timeout_sys, timeout))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002849 return -EFAULT;
2850
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002851 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2852 return -EFAULT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002853
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002854 if (!timeout && !timeout32)
2855 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2856
2857 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2858
2859 if (datagrams <= 0)
2860 return datagrams;
2861
2862 if (timeout && put_timespec64(&timeout_sys, timeout))
2863 datagrams = -EFAULT;
2864
2865 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002866 datagrams = -EFAULT;
2867
2868 return datagrams;
2869}
2870
Dominik Brodowski1255e262018-03-13 20:44:21 +01002871SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2872 unsigned int, vlen, unsigned int, flags,
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002873 struct __kernel_timespec __user *, timeout)
Dominik Brodowski1255e262018-03-13 20:44:21 +01002874{
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002875 if (flags & MSG_CMSG_COMPAT)
2876 return -EINVAL;
2877
2878 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
Dominik Brodowski1255e262018-03-13 20:44:21 +01002879}
2880
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002881#ifdef CONFIG_COMPAT_32BIT_TIME
2882SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2883 unsigned int, vlen, unsigned int, flags,
2884 struct old_timespec32 __user *, timeout)
2885{
2886 if (flags & MSG_CMSG_COMPAT)
2887 return -EINVAL;
2888
2889 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2890}
2891#endif
2892
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002893#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894/* Argument list sizes for sys_socketcall */
2895#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002896static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002897 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2898 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2899 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002900 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002901};
2902
Linus Torvalds1da177e2005-04-16 15:20:36 -07002903#undef AL
2904
2905/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002906 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002907 *
2908 * Argument checking cleaned up. Saved 20% in size.
2909 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002910 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911 */
2912
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002913SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002914{
Chen Gang2950fa92013-04-07 16:55:23 +08002915 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002916 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002917 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002918 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002919
Anton Blanchard228e5482011-05-02 20:21:35 +00002920 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002921 return -EINVAL;
Jeremy Clinec8e8cd52018-07-27 22:43:01 +00002922 call = array_index_nospec(call, SYS_SENDMMSG + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002923
Arjan van de Ven47379052009-09-28 12:57:44 -07002924 len = nargs[call];
2925 if (len > sizeof(a))
2926 return -EINVAL;
2927
Linus Torvalds1da177e2005-04-16 15:20:36 -07002928 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002929 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002930 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002931
Chen Gang2950fa92013-04-07 16:55:23 +08002932 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2933 if (err)
2934 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002935
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002936 a0 = a[0];
2937 a1 = a[1];
2938
2939 switch (call) {
2940 case SYS_SOCKET:
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01002941 err = __sys_socket(a0, a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002942 break;
2943 case SYS_BIND:
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01002944 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002945 break;
2946 case SYS_CONNECT:
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01002947 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002948 break;
2949 case SYS_LISTEN:
Dominik Brodowski25e290e2018-03-13 19:36:54 +01002950 err = __sys_listen(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002951 break;
2952 case SYS_ACCEPT:
Dominik Brodowski4541e802018-03-13 19:24:23 +01002953 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2954 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002955 break;
2956 case SYS_GETSOCKNAME:
2957 err =
Dominik Brodowski8882a102018-03-13 19:43:14 +01002958 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2959 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002960 break;
2961 case SYS_GETPEERNAME:
2962 err =
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002963 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2964 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002965 break;
2966 case SYS_SOCKETPAIR:
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01002967 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002968 break;
2969 case SYS_SEND:
Dominik Brodowskif3bf8962018-03-13 19:52:00 +01002970 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2971 NULL, 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002972 break;
2973 case SYS_SENDTO:
Dominik Brodowski211b6342018-03-13 19:18:52 +01002974 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2975 (struct sockaddr __user *)a[4], a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002976 break;
2977 case SYS_RECV:
Dominik Brodowskid27e9af2018-03-13 19:54:17 +01002978 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2979 NULL, NULL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002980 break;
2981 case SYS_RECVFROM:
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002982 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2983 (struct sockaddr __user *)a[4],
2984 (int __user *)a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002985 break;
2986 case SYS_SHUTDOWN:
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002987 err = __sys_shutdown(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002988 break;
2989 case SYS_SETSOCKOPT:
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002990 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2991 a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002992 break;
2993 case SYS_GETSOCKOPT:
2994 err =
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002995 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2996 (int __user *)a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002997 break;
2998 case SYS_SENDMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002999 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3000 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003001 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00003002 case SYS_SENDMMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003003 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3004 a[3], true);
Anton Blanchard228e5482011-05-02 20:21:35 +00003005 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003006 case SYS_RECVMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003007 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3008 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003009 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003010 case SYS_RECVMMSG:
Arnd Bergmann3ca47e92019-04-23 17:43:50 +02003011 if (IS_ENABLED(CONFIG_64BIT))
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003012 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3013 a[2], a[3],
3014 (struct __kernel_timespec __user *)a[4],
3015 NULL);
3016 else
3017 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3018 a[2], a[3], NULL,
3019 (struct old_timespec32 __user *)a[4]);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003020 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08003021 case SYS_ACCEPT4:
Dominik Brodowski4541e802018-03-13 19:24:23 +01003022 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3023 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07003024 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003025 default:
3026 err = -EINVAL;
3027 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003028 }
3029 return err;
3030}
3031
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003032#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003034/**
3035 * sock_register - add a socket protocol handler
3036 * @ops: description of protocol
3037 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003038 * This function is called by a protocol handler that wants to
3039 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09003040 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003041 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003042 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003043int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003044{
3045 int err;
3046
3047 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08003048 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049 return -ENOBUFS;
3050 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003051
3052 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00003053 if (rcu_dereference_protected(net_families[ops->family],
3054 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003055 err = -EEXIST;
3056 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00003057 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003058 err = 0;
3059 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003060 spin_unlock(&net_family_lock);
3061
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003062 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003063 return err;
3064}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003065EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003066
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003067/**
3068 * sock_unregister - remove a protocol handler
3069 * @family: protocol family to remove
3070 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003071 * This function is called by a protocol handler that wants to
3072 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003073 * new socket creation.
3074 *
3075 * If protocol handler is a module, then it can use module reference
3076 * counts to protect against new references. If protocol handler is not
3077 * a module then it needs to provide its own protection in
3078 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003079 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003080void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003081{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003082 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003083
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003084 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00003085 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003086 spin_unlock(&net_family_lock);
3087
3088 synchronize_rcu();
3089
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003090 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003091}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003092EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003093
Xin Longbf2ae2e2018-03-10 18:57:50 +08003094bool sock_is_registered(int family)
3095{
Jeremy Cline66b51b02018-08-13 22:23:13 +00003096 return family < NPROTO && rcu_access_pointer(net_families[family]);
Xin Longbf2ae2e2018-03-10 18:57:50 +08003097}
3098
Andi Kleen77d76ea2005-12-22 12:43:42 -08003099static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003100{
Nick Pigginb3e19d92011-01-07 17:50:11 +11003101 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00003102 /*
3103 * Initialize the network sysctl infrastructure.
3104 */
3105 err = net_sysctl_init();
3106 if (err)
3107 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11003108
Linus Torvalds1da177e2005-04-16 15:20:36 -07003109 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003110 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07003111 */
3112 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003113
3114 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003115 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003116 */
3117
3118 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11003119
3120 err = register_filesystem(&sock_fs_type);
3121 if (err)
Miaohe Lin47260ba2020-08-06 19:54:19 +08003122 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003123 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003124 if (IS_ERR(sock_mnt)) {
3125 err = PTR_ERR(sock_mnt);
3126 goto out_mount;
3127 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08003128
3129 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003130 */
3131
3132#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00003133 err = netfilter_init();
3134 if (err)
3135 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003136#endif
David S. Millercbeb3212005-12-22 12:58:55 -08003137
Daniel Borkmann408eccc2014-04-01 16:20:23 +02003138 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00003139
Nick Pigginb3e19d92011-01-07 17:50:11 +11003140out:
3141 return err;
3142
3143out_mount:
3144 unregister_filesystem(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003145 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003146}
3147
Andi Kleen77d76ea2005-12-22 12:43:42 -08003148core_initcall(sock_init); /* early initcall */
3149
Linus Torvalds1da177e2005-04-16 15:20:36 -07003150#ifdef CONFIG_PROC_FS
3151void socket_seq_show(struct seq_file *seq)
3152{
Tonghao Zhang648845a2017-12-14 05:51:58 -08003153 seq_printf(seq, "sockets: used %d\n",
3154 sock_inuse_get(seq->private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003155}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003156#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003157
Arnd Bergmann29c49642021-07-22 16:29:03 +02003158/* Handle the fact that while struct ifreq has the same *layout* on
3159 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3160 * which are handled elsewhere, it still has different *size* due to
3161 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3162 * resulting in struct ifreq being 32 and 40 bytes respectively).
3163 * As a result, if the struct happens to be at the end of a page and
3164 * the next page isn't readable/writable, we get a fault. To prevent
3165 * that, copy back and forth to the full size.
3166 */
3167int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3168{
3169 if (in_compat_syscall()) {
3170 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3171
3172 memset(ifr, 0, sizeof(*ifr));
3173 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3174 return -EFAULT;
3175
3176 if (ifrdata)
3177 *ifrdata = compat_ptr(ifr32->ifr_data);
3178
3179 return 0;
3180 }
3181
3182 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3183 return -EFAULT;
3184
3185 if (ifrdata)
3186 *ifrdata = ifr->ifr_data;
3187
3188 return 0;
3189}
3190EXPORT_SYMBOL(get_user_ifreq);
3191
3192int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3193{
3194 size_t size = sizeof(*ifr);
3195
3196 if (in_compat_syscall())
3197 size = sizeof(struct compat_ifreq);
3198
3199 if (copy_to_user(arg, ifr, size))
3200 return -EFAULT;
3201
3202 return 0;
3203}
3204EXPORT_SYMBOL(put_user_ifreq);
3205
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003206#ifdef CONFIG_COMPAT
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003207static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3208{
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003209 compat_uptr_t uptr32;
Al Viro44c02a22017-10-05 12:59:44 -04003210 struct ifreq ifr;
3211 void __user *saved;
3212 int err;
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003213
Arnd Bergmann29c49642021-07-22 16:29:03 +02003214 if (get_user_ifreq(&ifr, NULL, uifr32))
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003215 return -EFAULT;
3216
3217 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3218 return -EFAULT;
3219
Al Viro44c02a22017-10-05 12:59:44 -04003220 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3221 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003222
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003223 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
Al Viro44c02a22017-10-05 12:59:44 -04003224 if (!err) {
3225 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
Arnd Bergmann29c49642021-07-22 16:29:03 +02003226 if (put_user_ifreq(&ifr, uifr32))
Al Viro44c02a22017-10-05 12:59:44 -04003227 err = -EFAULT;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003228 }
Al Viro44c02a22017-10-05 12:59:44 -04003229 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003230}
3231
Ben Hutchings590d4692013-11-18 17:04:13 +00003232/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3233static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08003234 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003235{
Al Viro44c02a22017-10-05 12:59:44 -04003236 struct ifreq ifreq;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003237 void __user *data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003238
Peter Collingbourned0efb162021-08-26 12:46:01 -07003239 if (!is_socket_ioctl_cmd(cmd))
3240 return -ENOTTY;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003241 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003242 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003243 ifreq.ifr_data = data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003244
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003245 return dev_ioctl(net, cmd, &ifreq, data, NULL);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003246}
3247
Arnd Bergmann7a229382009-11-06 23:00:29 -08003248/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3249 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003250 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003251 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003252static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003253{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003254 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003255
Arnd Bergmann6b960182009-11-06 23:10:54 -08003256 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003257 return -EFAULT;
3258 if (tmp == BRCTL_GET_VERSION)
3259 return BRCTL_VERSION + 1;
3260 return -EINVAL;
3261}
3262
Arnd Bergmann6b960182009-11-06 23:10:54 -08003263static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3264 unsigned int cmd, unsigned long arg)
3265{
3266 void __user *argp = compat_ptr(arg);
3267 struct sock *sk = sock->sk;
3268 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003269
Arnd Bergmann6b960182009-11-06 23:10:54 -08003270 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Arnd Bergmann88fc0232021-07-27 15:45:15 +02003271 return sock_ioctl(file, cmd, (unsigned long)argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003272
Arnd Bergmann6b960182009-11-06 23:10:54 -08003273 switch (cmd) {
3274 case SIOCSIFBR:
3275 case SIOCGIFBR:
3276 return old_bridge_ioctl(argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003277 case SIOCWANDEV:
3278 return compat_siocwandev(net, argp);
Arnd Bergmann0768e172019-04-17 22:56:11 +02003279 case SIOCGSTAMP_OLD:
3280 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003281 if (!sock->ops->gettstamp)
3282 return -ENOIOCTLCMD;
Arnd Bergmann0768e172019-04-17 22:56:11 +02003283 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003284 !COMPAT_USE_64BIT_TIME);
3285
Arnd Bergmanndd98d282021-07-22 16:28:59 +02003286 case SIOCETHTOOL:
Ben Hutchings590d4692013-11-18 17:04:13 +00003287 case SIOCBONDSLAVEINFOQUERY:
3288 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003289 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003290 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003291 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003292
Arnd Bergmann6b960182009-11-06 23:10:54 -08003293 case FIOSETOWN:
3294 case SIOCSPGRP:
3295 case FIOGETOWN:
3296 case SIOCGPGRP:
3297 case SIOCBRADDBR:
3298 case SIOCBRDELBR:
3299 case SIOCGIFVLAN:
3300 case SIOCSIFVLAN:
Andrey Vaginc62cce22016-10-24 18:29:13 -07003301 case SIOCGSKNS:
Arnd Bergmann0768e172019-04-17 22:56:11 +02003302 case SIOCGSTAMP_NEW:
3303 case SIOCGSTAMPNS_NEW:
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02003304 case SIOCGIFCONF:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003305 return sock_ioctl(file, cmd, arg);
3306
3307 case SIOCGIFFLAGS:
3308 case SIOCSIFFLAGS:
Arnd Bergmann709566d2021-07-22 16:29:00 +02003309 case SIOCGIFMAP:
3310 case SIOCSIFMAP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003311 case SIOCGIFMETRIC:
3312 case SIOCSIFMETRIC:
3313 case SIOCGIFMTU:
3314 case SIOCSIFMTU:
3315 case SIOCGIFMEM:
3316 case SIOCSIFMEM:
3317 case SIOCGIFHWADDR:
3318 case SIOCSIFHWADDR:
3319 case SIOCADDMULTI:
3320 case SIOCDELMULTI:
3321 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003322 case SIOCGIFADDR:
3323 case SIOCSIFADDR:
3324 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003325 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003326 case SIOCGIFBRDADDR:
3327 case SIOCSIFBRDADDR:
3328 case SIOCGIFDSTADDR:
3329 case SIOCSIFDSTADDR:
3330 case SIOCGIFNETMASK:
3331 case SIOCSIFNETMASK:
3332 case SIOCSIFPFLAGS:
3333 case SIOCGIFPFLAGS:
3334 case SIOCGIFTXQLEN:
3335 case SIOCSIFTXQLEN:
3336 case SIOCBRADDIF:
3337 case SIOCBRDELIF:
Johannes Bergc6c9fee2019-01-25 22:43:19 +01003338 case SIOCGIFNAME:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003339 case SIOCSIFNAME:
3340 case SIOCGMIIPHY:
3341 case SIOCGMIIREG:
3342 case SIOCSMIIREG:
Al Virof92d4fc2017-09-30 19:32:17 -04003343 case SIOCBONDENSLAVE:
3344 case SIOCBONDRELEASE:
3345 case SIOCBONDSETHWADDR:
3346 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003347 case SIOCSARP:
3348 case SIOCGARP:
3349 case SIOCDARP:
Arnd Bergmannc7dc5042019-06-03 23:07:12 +02003350 case SIOCOUTQ:
Arnd Bergmann9d7bf412019-06-03 23:06:00 +02003351 case SIOCOUTQNSD:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003352 case SIOCATMARK:
Johannes Berg63ff03a2019-01-25 22:43:17 +01003353 return sock_do_ioctl(net, sock, cmd, arg);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003354 }
3355
Arnd Bergmann6b960182009-11-06 23:10:54 -08003356 return -ENOIOCTLCMD;
3357}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003358
Eric Dumazet95c96172012-04-15 05:58:06 +00003359static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003360 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003361{
3362 struct socket *sock = file->private_data;
3363 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003364 struct sock *sk;
3365 struct net *net;
3366
3367 sk = sock->sk;
3368 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003369
3370 if (sock->ops->compat_ioctl)
3371 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3372
David S. Miller87de87d2008-06-03 09:14:03 -07003373 if (ret == -ENOIOCTLCMD &&
3374 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3375 ret = compat_wext_handle_ioctl(net, cmd, arg);
3376
Arnd Bergmann6b960182009-11-06 23:10:54 -08003377 if (ret == -ENOIOCTLCMD)
3378 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3379
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003380 return ret;
3381}
3382#endif
3383
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003384/**
3385 * kernel_bind - bind an address to a socket (kernel space)
3386 * @sock: socket
3387 * @addr: address
3388 * @addrlen: length of address
3389 *
3390 * Returns 0 or an error.
3391 */
3392
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003393int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3394{
3395 return sock->ops->bind(sock, addr, addrlen);
3396}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003397EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003398
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003399/**
3400 * kernel_listen - move socket to listening state (kernel space)
3401 * @sock: socket
3402 * @backlog: pending connections queue size
3403 *
3404 * Returns 0 or an error.
3405 */
3406
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003407int kernel_listen(struct socket *sock, int backlog)
3408{
3409 return sock->ops->listen(sock, backlog);
3410}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003411EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003412
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003413/**
3414 * kernel_accept - accept a connection (kernel space)
3415 * @sock: listening socket
3416 * @newsock: new connected socket
3417 * @flags: flags
3418 *
3419 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3420 * If it fails, @newsock is guaranteed to be %NULL.
3421 * Returns 0 or an error.
3422 */
3423
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003424int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3425{
3426 struct sock *sk = sock->sk;
3427 int err;
3428
3429 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3430 newsock);
3431 if (err < 0)
3432 goto done;
3433
David Howellscdfbabf2017-03-09 08:09:05 +00003434 err = sock->ops->accept(sock, *newsock, flags, true);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003435 if (err < 0) {
3436 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003437 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003438 goto done;
3439 }
3440
3441 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003442 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003443
3444done:
3445 return err;
3446}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003447EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003448
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003449/**
3450 * kernel_connect - connect a socket (kernel space)
3451 * @sock: socket
3452 * @addr: address
3453 * @addrlen: address length
3454 * @flags: flags (O_NONBLOCK, ...)
3455 *
Lu Weif1dcffc2021-03-25 11:01:55 +08003456 * For datagram sockets, @addr is the address to which datagrams are sent
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003457 * by default, and the only address from which datagrams are received.
3458 * For stream sockets, attempts to connect to @addr.
3459 * Returns 0 or an error code.
3460 */
3461
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003462int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003463 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003464{
3465 return sock->ops->connect(sock, addr, addrlen, flags);
3466}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003467EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003468
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003469/**
3470 * kernel_getsockname - get the address which the socket is bound (kernel space)
3471 * @sock: socket
3472 * @addr: address holder
3473 *
3474 * Fills the @addr pointer with the address which the socket is bound.
3475 * Returns 0 or an error code.
3476 */
3477
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003478int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003479{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003480 return sock->ops->getname(sock, addr, 0);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003481}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003482EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003483
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003484/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003485 * kernel_getpeername - get the address which the socket is connected (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003486 * @sock: socket
3487 * @addr: address holder
3488 *
3489 * Fills the @addr pointer with the address which the socket is connected.
3490 * Returns 0 or an error code.
3491 */
3492
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003493int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003494{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003495 return sock->ops->getname(sock, addr, 1);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003496}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003497EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003498
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003499/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003500 * kernel_sendpage - send a &page through a socket (kernel space)
3501 * @sock: socket
3502 * @page: page
3503 * @offset: page offset
3504 * @size: total size in bytes
3505 * @flags: flags (MSG_DONTWAIT, ...)
3506 *
3507 * Returns the total amount sent in bytes or an error.
3508 */
3509
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003510int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3511 size_t size, int flags)
3512{
Coly Li7b62d312020-10-02 16:27:29 +08003513 if (sock->ops->sendpage) {
3514 /* Warn in case the improper page to zero-copy send */
3515 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003516 return sock->ops->sendpage(sock, page, offset, size, flags);
Coly Li7b62d312020-10-02 16:27:29 +08003517 }
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003518 return sock_no_sendpage(sock, page, offset, size, flags);
3519}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003520EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003521
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003522/**
3523 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3524 * @sk: sock
3525 * @page: page
3526 * @offset: page offset
3527 * @size: total size in bytes
3528 * @flags: flags (MSG_DONTWAIT, ...)
3529 *
3530 * Returns the total amount sent in bytes or an error.
3531 * Caller must hold @sk.
3532 */
3533
Tom Herbert306b13e2017-07-28 16:22:41 -07003534int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3535 size_t size, int flags)
3536{
3537 struct socket *sock = sk->sk_socket;
3538
3539 if (sock->ops->sendpage_locked)
3540 return sock->ops->sendpage_locked(sk, page, offset, size,
3541 flags);
3542
3543 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3544}
3545EXPORT_SYMBOL(kernel_sendpage_locked);
3546
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003547/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003548 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003549 * @sock: socket
3550 * @how: connection part
3551 *
3552 * Returns 0 or an error.
3553 */
3554
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003555int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3556{
3557 return sock->ops->shutdown(sock, how);
3558}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003559EXPORT_SYMBOL(kernel_sock_shutdown);
R. Parameswaran113c3072017-04-05 16:50:35 -07003560
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003561/**
3562 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3563 * @sk: socket
3564 *
3565 * This routine returns the IP overhead imposed by a socket i.e.
3566 * the length of the underlying IP header, depending on whether
3567 * this is an IPv4 or IPv6 socket and the length from IP options turned
3568 * on at the socket. Assumes that the caller has a lock on the socket.
R. Parameswaran113c3072017-04-05 16:50:35 -07003569 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003570
R. Parameswaran113c3072017-04-05 16:50:35 -07003571u32 kernel_sock_ip_overhead(struct sock *sk)
3572{
3573 struct inet_sock *inet;
3574 struct ip_options_rcu *opt;
3575 u32 overhead = 0;
R. Parameswaran113c3072017-04-05 16:50:35 -07003576#if IS_ENABLED(CONFIG_IPV6)
3577 struct ipv6_pinfo *np;
3578 struct ipv6_txoptions *optv6 = NULL;
3579#endif /* IS_ENABLED(CONFIG_IPV6) */
3580
3581 if (!sk)
3582 return overhead;
3583
R. Parameswaran113c3072017-04-05 16:50:35 -07003584 switch (sk->sk_family) {
3585 case AF_INET:
3586 inet = inet_sk(sk);
3587 overhead += sizeof(struct iphdr);
3588 opt = rcu_dereference_protected(inet->inet_opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003589 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003590 if (opt)
3591 overhead += opt->opt.optlen;
3592 return overhead;
3593#if IS_ENABLED(CONFIG_IPV6)
3594 case AF_INET6:
3595 np = inet6_sk(sk);
3596 overhead += sizeof(struct ipv6hdr);
3597 if (np)
3598 optv6 = rcu_dereference_protected(np->opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003599 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003600 if (optv6)
3601 overhead += (optv6->opt_flen + optv6->opt_nflen);
3602 return overhead;
3603#endif /* IS_ENABLED(CONFIG_IPV6) */
3604 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3605 return overhead;
3606 }
3607}
3608EXPORT_SYMBOL(kernel_sock_ip_overhead);