blob: 8b8182cdb4bf260b35f251a61bf091253f5c2975 [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
Eric Dumazet767dd032010-04-28 19:14:43 +0000932void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
Neil Horman3b885782009-10-12 13:26:31 -0700933 struct sk_buff *skb)
934{
935 sock_recv_timestamp(msg, sk, skb);
936 sock_recv_drops(msg, sk, skb);
937}
Eric Dumazet767dd032010-04-28 19:14:43 +0000938EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
Neil Horman3b885782009-10-12 13:26:31 -0700939
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200940INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
Paolo Abenia648a592019-07-03 16:06:54 +0200941 size_t, int));
942INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
943 size_t, int));
Ying Xue1b784142015-03-02 15:37:48 +0800944static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -0400945 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946{
Paolo Abenia648a592019-07-03 16:06:54 +0200947 return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
948 inet_recvmsg, sock, msg, msg_data_left(msg),
949 flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950}
951
Randy Dunlap85806af2019-05-18 21:23:07 -0700952/**
953 * sock_recvmsg - receive a message from @sock
954 * @sock: socket
955 * @msg: message to receive
956 * @flags: message flags
957 *
958 * Receives @msg from @sock, passing through LSM. Returns the total number
959 * of bytes received, or an error.
960 */
Al Viro2da62902015-03-14 21:13:46 -0400961int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700962{
Al Viro2da62902015-03-14 21:13:46 -0400963 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -0700964
Al Viro2da62902015-03-14 21:13:46 -0400965 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700967EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000969/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300970 * kernel_recvmsg - Receive a message from a socket (kernel space)
971 * @sock: The socket to receive the message from
972 * @msg: Received message
973 * @vec: Input s/g array for message data
974 * @num: Size of input s/g array
975 * @size: Number of bytes to read
976 * @flags: Message flags (MSG_DONTWAIT, etc...)
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000977 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300978 * On return the msg structure contains the scatter/gather array passed in the
979 * vec argument. The array is modified so that it consists of the unfilled
980 * portion of the original array.
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000981 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300982 * The returned value is the total number of bytes received, or an error.
Martin Lucinac1249c0a2010-12-10 00:04:05 +0000983 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300984
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700985int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
986 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987{
Christoph Hellwig1f466e12020-05-11 13:59:13 +0200988 msg->msg_control_is_user = false;
David Howellsaa563d72018-10-20 00:57:56 +0100989 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
Christoph Hellwig1f466e12020-05-11 13:59:13 +0200990 return sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700992EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -0300994static ssize_t sock_sendpage(struct file *file, struct page *page,
995 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996{
997 struct socket *sock;
998 int flags;
999
Eric Dumazetb69aee02005-09-06 14:42:45 -07001000 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001
Eric Dumazet35f9c092012-04-05 03:05:35 +00001002 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1003 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1004 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Linus Torvaldse6949582009-08-13 08:28:36 -07001006 return kernel_sendpage(sock, page, offset, size, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007}
1008
Jens Axboe9c55e012007-11-06 23:30:13 -08001009static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001010 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -08001011 unsigned int flags)
1012{
1013 struct socket *sock = file->private_data;
1014
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001015 if (unlikely(!sock->ops->splice_read))
Slavomir Kaslev95506582018-11-16 11:27:53 +02001016 return generic_file_splice_read(file, ppos, pipe, len, flags);
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001017
Jens Axboe9c55e012007-11-06 23:30:13 -08001018 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1019}
1020
Al Viro8ae5e0302014-11-28 19:40:50 -05001021static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001022{
Al Viro6d652332015-01-30 16:12:56 -05001023 struct file *file = iocb->ki_filp;
1024 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001025 struct msghdr msg = {.msg_iter = *to,
1026 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001027 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001028
Jens Axboeebfcd892019-12-09 20:58:56 -07001029 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001030 msg.msg_flags = MSG_DONTWAIT;
1031
1032 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001033 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001034
Christoph Hellwig66ee59a2015-02-11 19:56:46 +01001035 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001036 return 0;
1037
Al Viro2da62902015-03-14 21:13:46 -04001038 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e0302014-11-28 19:40:50 -05001039 *to = msg.msg_iter;
1040 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001041}
1042
Al Viro8ae5e0302014-11-28 19:40:50 -05001043static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044{
Al Viro6d652332015-01-30 16:12:56 -05001045 struct file *file = iocb->ki_filp;
1046 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001047 struct msghdr msg = {.msg_iter = *from,
1048 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001049 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001050
Al Viro8ae5e0302014-11-28 19:40:50 -05001051 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001052 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001053
Jens Axboeebfcd892019-12-09 20:58:56 -07001054 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001055 msg.msg_flags = MSG_DONTWAIT;
1056
Al Viro6d652332015-01-30 16:12:56 -05001057 if (sock->type == SOCK_SEQPACKET)
1058 msg.msg_flags |= MSG_EOR;
1059
Al Virod8725c82014-12-11 00:02:50 -05001060 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e0302014-11-28 19:40:50 -05001061 *from = msg.msg_iter;
1062 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063}
1064
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065/*
1066 * Atomic setting of ioctl hooks to avoid race
1067 * with module unload.
1068 */
1069
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001070static DEFINE_MUTEX(br_ioctl_mutex);
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001071static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1072 unsigned int cmd, struct ifreq *ifr,
1073 void __user *uarg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001075void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1076 unsigned int cmd, struct ifreq *ifr,
1077 void __user *uarg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001079 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001081 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082}
1083EXPORT_SYMBOL(brioctl_set);
1084
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001085int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1086 struct ifreq *ifr, void __user *uarg)
1087{
1088 int err = -ENOPKG;
1089
1090 if (!br_ioctl_hook)
1091 request_module("bridge");
1092
1093 mutex_lock(&br_ioctl_mutex);
1094 if (br_ioctl_hook)
1095 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1096 mutex_unlock(&br_ioctl_mutex);
1097
1098 return err;
1099}
1100
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001101static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001102static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103
Eric W. Biederman881d9662007-09-17 11:56:21 -07001104void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001106 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001108 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109}
1110EXPORT_SYMBOL(vlan_ioctl_set);
1111
Arnd Bergmann6b960182009-11-06 23:10:54 -08001112static long sock_do_ioctl(struct net *net, struct socket *sock,
Johannes Berg63ff03a2019-01-25 22:43:17 +01001113 unsigned int cmd, unsigned long arg)
Arnd Bergmann6b960182009-11-06 23:10:54 -08001114{
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001115 struct ifreq ifr;
1116 bool need_copyout;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001117 int err;
1118 void __user *argp = (void __user *)arg;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001119 void __user *data;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001120
1121 err = sock->ops->ioctl(sock, cmd, arg);
1122
1123 /*
1124 * If this ioctl is unknown try to hand it down
1125 * to the NIC driver.
1126 */
Al Viro36fd6332017-06-26 13:19:16 -04001127 if (err != -ENOIOCTLCMD)
1128 return err;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001129
Jakub Kicinski29ce8f92021-08-31 09:06:04 -07001130 if (!is_socket_ioctl_cmd(cmd))
1131 return -ENOTTY;
1132
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001133 if (get_user_ifreq(&ifr, &data, argp))
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001134 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001135 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001136 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001137 if (put_user_ifreq(&ifr, argp))
Al Viro36fd6332017-06-26 13:19:16 -04001138 return -EFAULT;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001139
Arnd Bergmann6b960182009-11-06 23:10:54 -08001140 return err;
1141}
1142
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143/*
1144 * With an ioctl, arg may well be a user mode pointer, but we don't know
1145 * what to do with it - that's up to the protocol still.
1146 */
1147
1148static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1149{
1150 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001151 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 void __user *argp = (void __user *)arg;
1153 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001154 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
Eric Dumazetb69aee02005-09-06 14:42:45 -07001156 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001157 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001158 net = sock_net(sk);
Al Viro44c02a22017-10-05 12:59:44 -04001159 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1160 struct ifreq ifr;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001161 void __user *data;
Al Viro44c02a22017-10-05 12:59:44 -04001162 bool need_copyout;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001163 if (get_user_ifreq(&ifr, &data, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001164 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001165 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Al Viro44c02a22017-10-05 12:59:44 -04001166 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001167 if (put_user_ifreq(&ifr, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001168 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001170#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Al Virob1b0c242017-10-01 20:13:08 -04001172 err = wext_handle_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001174#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001175 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 case FIOSETOWN:
1177 case SIOCSPGRP:
1178 err = -EFAULT;
1179 if (get_user(pid, (int __user *)argp))
1180 break;
Jiri Slaby393cc3f2017-06-13 13:35:50 +02001181 err = f_setown(sock->file, pid, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182 break;
1183 case FIOGETOWN:
1184 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001185 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001186 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 break;
1188 case SIOCGIFBR:
1189 case SIOCSIFBR:
1190 case SIOCBRADDBR:
1191 case SIOCBRDELBR:
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001192 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 break;
1194 case SIOCGIFVLAN:
1195 case SIOCSIFVLAN:
1196 err = -ENOPKG;
1197 if (!vlan_ioctl_hook)
1198 request_module("8021q");
1199
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001200 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001202 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001203 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 break;
Andrey Vaginc62cce22016-10-24 18:29:13 -07001205 case SIOCGSKNS:
1206 err = -EPERM;
1207 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1208 break;
1209
1210 err = open_related_ns(&net->ns, get_net_ns);
1211 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001212 case SIOCGSTAMP_OLD:
1213 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001214 if (!sock->ops->gettstamp) {
1215 err = -ENOIOCTLCMD;
1216 break;
1217 }
1218 err = sock->ops->gettstamp(sock, argp,
Arnd Bergmann0768e172019-04-17 22:56:11 +02001219 cmd == SIOCGSTAMP_OLD,
1220 !IS_ENABLED(CONFIG_64BIT));
Gustavo A. R. Silva60747822019-04-24 10:31:24 -05001221 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001222 case SIOCGSTAMP_NEW:
1223 case SIOCGSTAMPNS_NEW:
1224 if (!sock->ops->gettstamp) {
1225 err = -ENOIOCTLCMD;
1226 break;
1227 }
1228 err = sock->ops->gettstamp(sock, argp,
1229 cmd == SIOCGSTAMP_NEW,
1230 false);
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001231 break;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001232
1233 case SIOCGIFCONF:
1234 err = dev_ifconf(net, argp);
1235 break;
1236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 default:
Johannes Berg63ff03a2019-01-25 22:43:17 +01001238 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001240 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 return err;
1242}
1243
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001244/**
1245 * sock_create_lite - creates a socket
1246 * @family: protocol family (AF_INET, ...)
1247 * @type: communication type (SOCK_STREAM, ...)
1248 * @protocol: protocol (0, ...)
1249 * @res: new socket
1250 *
1251 * Creates a new socket and assigns it to @res, passing through LSM.
1252 * The new socket initialization is not complete, see kernel_accept().
1253 * Returns 0 or an error. On failure @res is set to %NULL.
1254 * This function internally uses GFP_KERNEL.
1255 */
1256
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257int sock_create_lite(int family, int type, int protocol, struct socket **res)
1258{
1259 int err;
1260 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001261
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 err = security_socket_create(family, type, protocol, 1);
1263 if (err)
1264 goto out;
1265
1266 sock = sock_alloc();
1267 if (!sock) {
1268 err = -ENOMEM;
1269 goto out;
1270 }
1271
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001273 err = security_socket_post_create(sock, family, type, protocol, 1);
1274 if (err)
1275 goto out_release;
1276
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277out:
1278 *res = sock;
1279 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001280out_release:
1281 sock_release(sock);
1282 sock = NULL;
1283 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001285EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286
1287/* No kernel lock held - perfect */
Al Viroade994f2017-07-03 00:01:49 -04001288static __poll_t sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289{
Christoph Hellwig3cafb372018-01-09 16:07:34 +01001290 struct socket *sock = file->private_data;
Christoph Hellwiga331de32018-07-30 09:42:13 +02001291 __poll_t events = poll_requested_events(wait), flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
Christoph Hellwige88958e2018-06-29 15:37:24 +02001293 if (!sock->ops->poll)
1294 return 0;
Christoph Hellwigf641f13b2018-07-30 09:42:12 +02001295
Christoph Hellwiga331de32018-07-30 09:42:13 +02001296 if (sk_can_busy_loop(sock->sk)) {
1297 /* poll once if requested by the syscall */
1298 if (events & POLL_BUSY_LOOP)
1299 sk_busy_loop(sock->sk, 1);
1300
1301 /* if this socket can poll_ll, tell the system call */
1302 flag = POLL_BUSY_LOOP;
1303 }
1304
1305 return sock->ops->poll(file, sock, wait) | flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306}
1307
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001308static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001310 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311
1312 return sock->ops->mmap(file, sock, vma);
1313}
1314
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001315static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316{
Cong Wang6d8c50d2018-06-07 13:39:49 -07001317 __sock_release(SOCKET_I(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318 return 0;
1319}
1320
1321/*
1322 * Update the socket async list
1323 *
1324 * Fasync_list locking strategy.
1325 *
1326 * 1. fasync_list is modified only under process context socket lock
1327 * i.e. under semaphore.
1328 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001329 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 */
1331
1332static int sock_fasync(int fd, struct file *filp, int on)
1333{
Eric Dumazet989a2972010-04-14 09:55:35 +00001334 struct socket *sock = filp->private_data;
1335 struct sock *sk = sock->sk;
Al Viro333f7902019-07-05 20:14:16 +01001336 struct socket_wq *wq = &sock->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337
Eric Dumazet989a2972010-04-14 09:55:35 +00001338 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340
1341 lock_sock(sk);
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001342 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001344 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001345 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001346 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001347 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348
Eric Dumazet989a2972010-04-14 09:55:35 +00001349 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 return 0;
1351}
1352
Eric Dumazetceb5d582015-11-29 20:03:11 -08001353/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354
Eric Dumazetceb5d582015-11-29 20:03:11 -08001355int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001357 if (!wq || !wq->fasync_list)
1358 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001359
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001360 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001361 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001362 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 break;
1364 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001365 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001366 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 break;
Miaohe Lin7c7ab582020-08-08 16:23:10 +08001368 fallthrough;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001369 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001370call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001371 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001373 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001374 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001376
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377 return 0;
1378}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001379EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001381/**
1382 * __sock_create - creates a socket
1383 * @net: net namespace
1384 * @family: protocol family (AF_INET, ...)
1385 * @type: communication type (SOCK_STREAM, ...)
1386 * @protocol: protocol (0, ...)
1387 * @res: new socket
1388 * @kern: boolean for kernel space sockets
1389 *
1390 * Creates a new socket and assigns it to @res, passing through LSM.
1391 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1392 * be set to true if the socket resides in kernel space.
1393 * This function internally uses GFP_KERNEL.
1394 */
1395
Pavel Emelyanov721db932010-09-29 16:06:32 +04001396int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001397 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398{
1399 int err;
1400 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001401 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402
1403 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001404 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 */
1406 if (family < 0 || family >= NPROTO)
1407 return -EAFNOSUPPORT;
1408 if (type < 0 || type >= SOCK_MAX)
1409 return -EINVAL;
1410
1411 /* Compatibility.
1412
1413 This uglymoron is moved from INET layer to here to avoid
1414 deadlock in module load.
1415 */
1416 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001417 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1418 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 family = PF_PACKET;
1420 }
1421
1422 err = security_socket_create(family, type, protocol, kern);
1423 if (err)
1424 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001425
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001426 /*
1427 * Allocate the socket and allow the family to set things up. if
1428 * the protocol is 0, the family is instructed to select an appropriate
1429 * default.
1430 */
1431 sock = sock_alloc();
1432 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001433 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001434 return -ENFILE; /* Not exactly a match, but its the
1435 closest posix thing */
1436 }
1437
1438 sock->type = type;
1439
Johannes Berg95a5afc2008-10-16 15:24:51 -07001440#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001441 /* Attempt to load a protocol module if the find failed.
1442 *
1443 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 * requested real, full-featured networking support upon configuration.
1445 * Otherwise module support will break!
1446 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001447 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001448 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449#endif
1450
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001451 rcu_read_lock();
1452 pf = rcu_dereference(net_families[family]);
1453 err = -EAFNOSUPPORT;
1454 if (!pf)
1455 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456
1457 /*
1458 * We will call the ->create function, that possibly is in a loadable
1459 * module, so we have to bump that loadable module refcnt first.
1460 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001461 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 goto out_release;
1463
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001464 /* Now protected by module ref count */
1465 rcu_read_unlock();
1466
Eric Paris3f378b62009-11-05 22:18:14 -08001467 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001468 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001470
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 /*
1472 * Now to bump the refcnt of the [loadable] module that owns this
1473 * socket at sock_release time we decrement its refcnt.
1474 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001475 if (!try_module_get(sock->ops->owner))
1476 goto out_module_busy;
1477
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 /*
1479 * Now that we're done with the ->create function, the [loadable]
1480 * module can have its refcnt decremented
1481 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001482 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001483 err = security_socket_post_create(sock, family, type, protocol, kern);
1484 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001485 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001486 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001488 return 0;
1489
1490out_module_busy:
1491 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001493 sock->ops = NULL;
1494 module_put(pf->owner);
1495out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001497 return err;
1498
1499out_release:
1500 rcu_read_unlock();
1501 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001503EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001505/**
1506 * sock_create - creates a socket
1507 * @family: protocol family (AF_INET, ...)
1508 * @type: communication type (SOCK_STREAM, ...)
1509 * @protocol: protocol (0, ...)
1510 * @res: new socket
1511 *
1512 * A wrapper around __sock_create().
1513 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1514 */
1515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516int sock_create(int family, int type, int protocol, struct socket **res)
1517{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001518 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001520EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001522/**
1523 * sock_create_kern - creates a socket (kernel space)
1524 * @net: net namespace
1525 * @family: protocol family (AF_INET, ...)
1526 * @type: communication type (SOCK_STREAM, ...)
1527 * @protocol: protocol (0, ...)
1528 * @res: new socket
1529 *
1530 * A wrapper around __sock_create().
1531 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1532 */
1533
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001534int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001536 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001538EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001540int __sys_socket(int family, int type, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541{
1542 int retval;
1543 struct socket *sock;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001544 int flags;
1545
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001546 /* Check the SOCK_* constants for consistency. */
1547 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1548 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1549 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1550 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1551
Ulrich Dreppera677a032008-07-23 21:29:17 -07001552 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001553 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001554 return -EINVAL;
1555 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001557 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1558 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1559
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 retval = sock_create(family, type, protocol, &sock);
1561 if (retval < 0)
Al Viro8e1611e2017-12-05 23:29:09 +00001562 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
Al Viro8e1611e2017-12-05 23:29:09 +00001564 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565}
1566
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001567SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1568{
1569 return __sys_socket(family, type, protocol);
1570}
1571
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572/*
1573 * Create a pair of connected sockets.
1574 */
1575
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001576int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577{
1578 struct socket *sock1, *sock2;
1579 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001580 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001581 int flags;
1582
1583 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001584 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001585 return -EINVAL;
1586 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001588 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1589 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1590
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591 /*
Al Viro016a2662017-12-05 23:28:38 +00001592 * reserve descriptors and make sure we won't fail
1593 * to return them to userland.
1594 */
1595 fd1 = get_unused_fd_flags(flags);
1596 if (unlikely(fd1 < 0))
1597 return fd1;
1598
1599 fd2 = get_unused_fd_flags(flags);
1600 if (unlikely(fd2 < 0)) {
1601 put_unused_fd(fd1);
1602 return fd2;
1603 }
1604
1605 err = put_user(fd1, &usockvec[0]);
1606 if (err)
1607 goto out;
1608
1609 err = put_user(fd2, &usockvec[1]);
1610 if (err)
1611 goto out;
1612
1613 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 * Obtain the first socket and check if the underlying protocol
1615 * supports the socketpair call.
1616 */
1617
1618 err = sock_create(family, type, protocol, &sock1);
Al Viro016a2662017-12-05 23:28:38 +00001619 if (unlikely(err < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 goto out;
1621
1622 err = sock_create(family, type, protocol, &sock2);
Al Viro016a2662017-12-05 23:28:38 +00001623 if (unlikely(err < 0)) {
1624 sock_release(sock1);
1625 goto out;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001626 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001627
David Herrmannd47cd942018-05-04 16:28:20 +02001628 err = security_socket_socketpair(sock1, sock2);
1629 if (unlikely(err)) {
1630 sock_release(sock2);
1631 sock_release(sock1);
1632 goto out;
1633 }
1634
Al Viro016a2662017-12-05 23:28:38 +00001635 err = sock1->ops->socketpair(sock1, sock2);
1636 if (unlikely(err < 0)) {
1637 sock_release(sock2);
1638 sock_release(sock1);
1639 goto out;
Al Viro28407632012-08-17 23:54:15 -04001640 }
1641
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001642 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301643 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001644 err = PTR_ERR(newfile1);
Al Viro016a2662017-12-05 23:28:38 +00001645 sock_release(sock2);
1646 goto out;
Al Viro28407632012-08-17 23:54:15 -04001647 }
1648
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001649 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001650 if (IS_ERR(newfile2)) {
1651 err = PTR_ERR(newfile2);
Al Viro016a2662017-12-05 23:28:38 +00001652 fput(newfile1);
1653 goto out;
Al Virodb349502007-02-07 01:48:00 -05001654 }
1655
Al Viro157cf642008-12-14 04:57:47 -05001656 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001657
Al Virodb349502007-02-07 01:48:00 -05001658 fd_install(fd1, newfile1);
1659 fd_install(fd2, newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001660 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662out:
Al Viro016a2662017-12-05 23:28:38 +00001663 put_unused_fd(fd2);
1664 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 return err;
1666}
1667
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001668SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1669 int __user *, usockvec)
1670{
1671 return __sys_socketpair(family, type, protocol, usockvec);
1672}
1673
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674/*
1675 * Bind a name to a socket. Nothing much to do here since it's
1676 * the protocol's responsibility to handle the local address.
1677 *
1678 * We move the socket address to kernel space before we call
1679 * the protocol layer (having also checked the address is ok).
1680 */
1681
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001682int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683{
1684 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001685 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001686 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001688 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001689 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001690 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Jakub Sitnicki068b88c2018-10-04 11:09:40 +02001691 if (!err) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001692 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001693 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001694 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001695 if (!err)
1696 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001697 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001698 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001700 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001701 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 return err;
1703}
1704
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001705SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1706{
1707 return __sys_bind(fd, umyaddr, addrlen);
1708}
1709
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710/*
1711 * Perform a listen. Basically, we allow the protocol to do anything
1712 * necessary for a listen, and if that works, we mark the socket as
1713 * ready for listening.
1714 */
1715
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001716int __sys_listen(int fd, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717{
1718 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001719 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001720 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001721
1722 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1723 if (sock) {
Kuniyuki Iwashimacb9eaed2022-08-23 10:47:00 -07001724 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
Eric Dumazet95c96172012-04-15 05:58:06 +00001725 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001726 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727
1728 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001729 if (!err)
1730 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001732 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 }
1734 return err;
1735}
1736
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001737SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1738{
1739 return __sys_listen(fd, backlog);
1740}
1741
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001742struct file *do_accept(struct file *file, unsigned file_flags,
Jens Axboede2ea4b2019-10-17 14:41:29 -06001743 struct sockaddr __user *upeer_sockaddr,
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001744 int __user *upeer_addrlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745{
1746 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001747 struct file *newfile;
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001748 int err, len;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001749 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
Florent Revestdba4a922020-12-04 12:36:04 +01001751 sock = sock_from_file(file);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001752 if (!sock)
1753 return ERR_PTR(-ENOTSOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001755 newsock = sock_alloc();
1756 if (!newsock)
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001757 return ERR_PTR(-ENFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
1759 newsock->type = sock->type;
1760 newsock->ops = sock->ops;
1761
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 /*
1763 * We don't need try_module_get here, as the listening socket (sock)
1764 * has the protocol module (sock->ops->owner) held.
1765 */
1766 __module_get(newsock->ops->owner);
1767
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001768 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001769 if (IS_ERR(newfile))
1770 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001771
Frank Filza79af592005-09-27 15:23:38 -07001772 err = security_socket_accept(sock, newsock);
1773 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001774 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001775
Jens Axboede2ea4b2019-10-17 14:41:29 -06001776 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1777 false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001779 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780
1781 if (upeer_sockaddr) {
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001782 len = newsock->ops->getname(newsock,
1783 (struct sockaddr *)&address, 2);
1784 if (len < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001786 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001788 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001789 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001791 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 }
1793
1794 /* File flags are not inherited via accept() unlike another OSes. */
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001795 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001796out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001797 fput(newfile);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001798 return ERR_PTR(err);
1799}
Jens Axboede2ea4b2019-10-17 14:41:29 -06001800
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001801int __sys_accept4_file(struct file *file, unsigned file_flags,
1802 struct sockaddr __user *upeer_sockaddr,
1803 int __user *upeer_addrlen, int flags,
1804 unsigned long nofile)
1805{
1806 struct file *newfile;
1807 int newfd;
1808
1809 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1810 return -EINVAL;
1811
1812 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1813 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1814
1815 newfd = __get_unused_fd_flags(flags, nofile);
1816 if (unlikely(newfd < 0))
1817 return newfd;
1818
1819 newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen,
1820 flags);
1821 if (IS_ERR(newfile)) {
1822 put_unused_fd(newfd);
1823 return PTR_ERR(newfile);
1824 }
1825 fd_install(newfd, newfile);
1826 return newfd;
Jens Axboede2ea4b2019-10-17 14:41:29 -06001827}
1828
1829/*
1830 * For accept, we attempt to create a new socket, set up the link
1831 * with the client, wake up the client, then return the new
1832 * connected fd. We collect the address of the connector in kernel
1833 * space and move it to user at the very end. This is unclean because
1834 * we open the socket then return an error.
1835 *
1836 * 1003.1g adds the ability to recvmsg() to query connection pending
1837 * status to recvmsg. We need to add that support in a way thats
1838 * clean when we restructure accept also.
1839 */
1840
1841int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1842 int __user *upeer_addrlen, int flags)
1843{
1844 int ret = -EBADF;
1845 struct fd f;
1846
1847 f = fdget(fd);
1848 if (f.file) {
1849 ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
Jens Axboe09952e32020-03-19 20:16:56 -06001850 upeer_addrlen, flags,
1851 rlimit(RLIMIT_NOFILE));
Miaohe Lin6b07ede2020-08-06 19:52:24 +08001852 fdput(f);
Jens Axboede2ea4b2019-10-17 14:41:29 -06001853 }
1854
1855 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856}
1857
Dominik Brodowski4541e802018-03-13 19:24:23 +01001858SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1859 int __user *, upeer_addrlen, int, flags)
1860{
1861 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1862}
1863
Heiko Carstens20f37032009-01-14 14:14:23 +01001864SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1865 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001866{
Dominik Brodowski4541e802018-03-13 19:24:23 +01001867 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001868}
1869
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870/*
1871 * Attempt to connect to a socket with the server address. The address
1872 * is in user space so we verify it is OK and move it to kernel space.
1873 *
1874 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1875 * break bindings
1876 *
1877 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1878 * other SEQPACKET protocols that take time to connect() as it doesn't
1879 * include the -EINPROGRESS status for such sockets.
1880 */
1881
Jens Axboef499a022019-12-02 16:28:46 -07001882int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
Jens Axboebd3ded32019-11-23 14:17:16 -07001883 int addrlen, int file_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884{
1885 struct socket *sock;
Jens Axboebd3ded32019-11-23 14:17:16 -07001886 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887
Florent Revestdba4a922020-12-04 12:36:04 +01001888 sock = sock_from_file(file);
1889 if (!sock) {
1890 err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 goto out;
Florent Revestdba4a922020-12-04 12:36:04 +01001892 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001894 err =
Jens Axboef499a022019-12-02 16:28:46 -07001895 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 if (err)
Jens Axboebd3ded32019-11-23 14:17:16 -07001897 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898
Jens Axboef499a022019-12-02 16:28:46 -07001899 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Jens Axboebd3ded32019-11-23 14:17:16 -07001900 sock->file->f_flags | file_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901out:
1902 return err;
1903}
1904
Jens Axboebd3ded32019-11-23 14:17:16 -07001905int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1906{
1907 int ret = -EBADF;
1908 struct fd f;
1909
1910 f = fdget(fd);
1911 if (f.file) {
Jens Axboef499a022019-12-02 16:28:46 -07001912 struct sockaddr_storage address;
1913
1914 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1915 if (!ret)
1916 ret = __sys_connect_file(f.file, &address, addrlen, 0);
Miaohe Lin6b07ede2020-08-06 19:52:24 +08001917 fdput(f);
Jens Axboebd3ded32019-11-23 14:17:16 -07001918 }
1919
1920 return ret;
1921}
1922
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01001923SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1924 int, addrlen)
1925{
1926 return __sys_connect(fd, uservaddr, addrlen);
1927}
1928
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929/*
1930 * Get the local address ('name') of a socket object. Move the obtained
1931 * name to user space.
1932 */
1933
Dominik Brodowski8882a102018-03-13 19:43:14 +01001934int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1935 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936{
1937 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001938 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001939 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001940
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001941 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 if (!sock)
1943 goto out;
1944
1945 err = security_socket_getsockname(sock);
1946 if (err)
1947 goto out_put;
1948
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001949 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1950 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 goto out_put;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001952 /* "err" is actually length in this case */
1953 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954
1955out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001956 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957out:
1958 return err;
1959}
1960
Dominik Brodowski8882a102018-03-13 19:43:14 +01001961SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1962 int __user *, usockaddr_len)
1963{
1964 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1965}
1966
Linus Torvalds1da177e2005-04-16 15:20:36 -07001967/*
1968 * Get the remote address ('name') of a socket object. Move the obtained
1969 * name to user space.
1970 */
1971
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01001972int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1973 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974{
1975 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001976 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001977 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001979 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1980 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 err = security_socket_getpeername(sock);
1982 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001983 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 return err;
1985 }
1986
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001987 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1988 if (err >= 0)
1989 /* "err" is actually length in this case */
1990 err = move_addr_to_user(&address, err, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001991 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001992 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 }
1994 return err;
1995}
1996
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01001997SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1998 int __user *, usockaddr_len)
1999{
2000 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2001}
2002
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003/*
2004 * Send a datagram to a given address. We move the address into kernel
2005 * space and check the user space data area is readable before invoking
2006 * the protocol.
2007 */
Dominik Brodowski211b6342018-03-13 19:18:52 +01002008int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2009 struct sockaddr __user *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010{
2011 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002012 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 int err;
2014 struct msghdr msg;
2015 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002016 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002017
Al Viro602bd0e2015-03-21 19:12:32 -04002018 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
2019 if (unlikely(err))
2020 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002021 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2022 if (!sock)
David S. Miller4387ff72007-02-08 15:06:08 -08002023 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002024
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002025 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002026 msg.msg_control = NULL;
2027 msg.msg_controllen = 0;
2028 msg.msg_namelen = 0;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002029 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002030 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 if (err < 0)
2032 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002033 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002034 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035 }
2036 if (sock->file->f_flags & O_NONBLOCK)
2037 flags |= MSG_DONTWAIT;
2038 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05002039 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002041out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002042 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08002043out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044 return err;
2045}
2046
Dominik Brodowski211b6342018-03-13 19:18:52 +01002047SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2048 unsigned int, flags, struct sockaddr __user *, addr,
2049 int, addr_len)
2050{
2051 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2052}
2053
Linus Torvalds1da177e2005-04-16 15:20:36 -07002054/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002055 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 */
2057
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002058SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00002059 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060{
Dominik Brodowski211b6342018-03-13 19:18:52 +01002061 return __sys_sendto(fd, buff, len, flags, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002062}
2063
2064/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002065 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066 * sender. We verify the buffers are writable and if needed move the
2067 * sender address from kernel to user space.
2068 */
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002069int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2070 struct sockaddr __user *addr, int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071{
2072 struct socket *sock;
2073 struct iovec iov;
2074 struct msghdr msg;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002075 struct sockaddr_storage address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002076 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002077 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078
Al Viro602bd0e2015-03-21 19:12:32 -04002079 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2080 if (unlikely(err))
2081 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002082 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002084 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002086 msg.msg_control = NULL;
2087 msg.msg_controllen = 0;
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002088 /* Save some cycles and don't copy the address if not needed */
2089 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
2090 /* We assume all kernel code knows the size of sockaddr_storage */
2091 msg.msg_namelen = 0;
tadeusz.struk@intel.com130ed5d2015-12-15 10:46:17 -08002092 msg.msg_iocb = NULL;
Alexander Potapenko9f138fa2017-03-08 18:08:16 +01002093 msg.msg_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002094 if (sock->file->f_flags & O_NONBLOCK)
2095 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002096 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002098 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002099 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002100 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002101 if (err2 < 0)
2102 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002104
2105 fput_light(sock->file, fput_needed);
David S. Miller4387ff72007-02-08 15:06:08 -08002106out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 return err;
2108}
2109
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002110SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2111 unsigned int, flags, struct sockaddr __user *, addr,
2112 int __user *, addr_len)
2113{
2114 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2115}
2116
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002118 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 */
2120
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02002121SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2122 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002124 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125}
2126
Florian Westphal83f0c102020-07-05 01:30:15 +02002127static bool sock_use_custom_sol_socket(const struct socket *sock)
2128{
2129 const struct sock *sk = sock->sk;
2130
2131 /* Use sock->ops->setsockopt() for MPTCP */
2132 return IS_ENABLED(CONFIG_MPTCP) &&
2133 sk->sk_protocol == IPPROTO_MPTCP &&
2134 sk->sk_type == SOCK_STREAM &&
2135 (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2136}
2137
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138/*
2139 * Set a socket option. Because we don't know the option lengths we have
2140 * to pass the user mode parameter for the protocols to sort out.
2141 */
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002142int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002143 int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144{
Christoph Hellwig519a8a62020-08-10 18:42:14 +02002145 sockptr_t optval = USER_SOCKPTR(user_optval);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002146 char *kernel_optval = NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002147 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148 struct socket *sock;
2149
2150 if (optlen < 0)
2151 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002152
2153 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002154 if (!sock)
2155 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156
Christoph Hellwig4a367292020-07-17 08:23:11 +02002157 err = security_socket_setsockopt(sock, level, optname);
2158 if (err)
2159 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002160
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002161 if (!in_compat_syscall())
2162 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002163 user_optval, &optlen,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002164 &kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002165 if (err < 0)
2166 goto out_put;
2167 if (err > 0) {
2168 err = 0;
2169 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170 }
Christoph Hellwig4a367292020-07-17 08:23:11 +02002171
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002172 if (kernel_optval)
2173 optval = KERNEL_SOCKPTR(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002174 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002175 err = sock_setsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002176 else if (unlikely(!sock->ops->setsockopt))
2177 err = -EOPNOTSUPP;
Christoph Hellwig4a367292020-07-17 08:23:11 +02002178 else
2179 err = sock->ops->setsockopt(sock, level, optname, optval,
2180 optlen);
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002181 kfree(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002182out_put:
2183 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 return err;
2185}
2186
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002187SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2188 char __user *, optval, int, optlen)
2189{
2190 return __sys_setsockopt(fd, level, optname, optval, optlen);
2191}
2192
Stanislav Fomichev9cacf812021-01-15 08:34:59 -08002193INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2194 int optname));
2195
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196/*
2197 * Get a socket option. Because we don't know the option lengths we have
2198 * to pass a user mode parameter for the protocols to sort out.
2199 */
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002200int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2201 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002203 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204 struct socket *sock;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002205 int max_optlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002207 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002208 if (!sock)
2209 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002211 err = security_socket_getsockopt(sock, level, optname);
2212 if (err)
2213 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002214
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002215 if (!in_compat_syscall())
2216 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002217
2218 if (level == SOL_SOCKET)
2219 err = sock_getsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002220 else if (unlikely(!sock->ops->getsockopt))
2221 err = -EOPNOTSUPP;
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002222 else
2223 err = sock->ops->getsockopt(sock, level, optname, optval,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002224 optlen);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002225
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002226 if (!in_compat_syscall())
2227 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2228 optval, optlen, max_optlen,
2229 err);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002230out_put:
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002231 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232 return err;
2233}
2234
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002235SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2236 char __user *, optval, int __user *, optlen)
2237{
2238 return __sys_getsockopt(fd, level, optname, optval, optlen);
2239}
2240
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241/*
2242 * Shutdown a socket.
2243 */
2244
Jens Axboeb713c192020-09-05 11:13:35 -06002245int __sys_shutdown_sock(struct socket *sock, int how)
2246{
2247 int err;
2248
2249 err = security_socket_shutdown(sock, how);
2250 if (!err)
2251 err = sock->ops->shutdown(sock, how);
2252
2253 return err;
2254}
2255
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002256int __sys_shutdown(int fd, int how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002258 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 struct socket *sock;
2260
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002261 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2262 if (sock != NULL) {
Jens Axboeb713c192020-09-05 11:13:35 -06002263 err = __sys_shutdown_sock(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002264 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265 }
2266 return err;
2267}
2268
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002269SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2270{
2271 return __sys_shutdown(fd, how);
2272}
2273
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002274/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275 * fields which are the same type (int / unsigned) on our platforms.
2276 */
2277#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2278#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2279#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2280
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002281struct used_address {
2282 struct sockaddr_storage name;
2283 unsigned int name_len;
2284};
2285
Jens Axboe0a384ab2020-02-27 08:11:20 -07002286int __copy_msghdr_from_user(struct msghdr *kmsg,
2287 struct user_msghdr __user *umsg,
2288 struct sockaddr __user **save_addr,
2289 struct iovec __user **uiov, size_t *nsegs)
Dan Carpenter1661bf32013-10-03 00:27:20 +03002290{
Al Viroffb07552017-06-27 19:32:04 -04002291 struct user_msghdr msg;
Al Viro08adb7d2014-11-10 20:23:13 -05002292 ssize_t err;
2293
Al Viroffb07552017-06-27 19:32:04 -04002294 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
Dan Carpenter1661bf32013-10-03 00:27:20 +03002295 return -EFAULT;
Matthew Leachdbb490b2014-03-11 11:58:27 +00002296
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002297 kmsg->msg_control_is_user = true;
2298 kmsg->msg_control_user = msg.msg_control;
Al Viroffb07552017-06-27 19:32:04 -04002299 kmsg->msg_controllen = msg.msg_controllen;
2300 kmsg->msg_flags = msg.msg_flags;
2301
2302 kmsg->msg_namelen = msg.msg_namelen;
2303 if (!msg.msg_name)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07002304 kmsg->msg_namelen = 0;
2305
Matthew Leachdbb490b2014-03-11 11:58:27 +00002306 if (kmsg->msg_namelen < 0)
2307 return -EINVAL;
2308
Dan Carpenter1661bf32013-10-03 00:27:20 +03002309 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03002310 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05002311
2312 if (save_addr)
Al Viroffb07552017-06-27 19:32:04 -04002313 *save_addr = msg.msg_name;
Al Viro08adb7d2014-11-10 20:23:13 -05002314
Al Viroffb07552017-06-27 19:32:04 -04002315 if (msg.msg_name && kmsg->msg_namelen) {
Al Viro08adb7d2014-11-10 20:23:13 -05002316 if (!save_addr) {
Paolo Abeni864d9662017-07-21 18:49:45 +02002317 err = move_addr_to_kernel(msg.msg_name,
2318 kmsg->msg_namelen,
Al Viro08adb7d2014-11-10 20:23:13 -05002319 kmsg->msg_name);
2320 if (err < 0)
2321 return err;
2322 }
2323 } else {
2324 kmsg->msg_name = NULL;
2325 kmsg->msg_namelen = 0;
2326 }
2327
Al Viroffb07552017-06-27 19:32:04 -04002328 if (msg.msg_iovlen > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05002329 return -EMSGSIZE;
2330
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002331 kmsg->msg_iocb = NULL;
Jens Axboe0a384ab2020-02-27 08:11:20 -07002332 *uiov = msg.msg_iov;
2333 *nsegs = msg.msg_iovlen;
2334 return 0;
2335}
2336
2337static int copy_msghdr_from_user(struct msghdr *kmsg,
2338 struct user_msghdr __user *umsg,
2339 struct sockaddr __user **save_addr,
2340 struct iovec **iov)
2341{
2342 struct user_msghdr msg;
2343 ssize_t err;
2344
2345 err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2346 &msg.msg_iovlen);
2347 if (err)
2348 return err;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002349
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002350 err = import_iovec(save_addr ? READ : WRITE,
Al Viroffb07552017-06-27 19:32:04 -04002351 msg.msg_iov, msg.msg_iovlen,
Al Viroda184282015-03-21 19:29:06 -04002352 UIO_FASTIOV, iov, &kmsg->msg_iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002353 return err < 0 ? err : 0;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002354}
2355
Jens Axboe4257c8c2019-11-25 14:27:34 -07002356static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2357 unsigned int flags, struct used_address *used_address,
2358 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359{
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002360 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Amit Kushwaha846cc122016-12-08 18:21:53 +05302361 __aligned(sizeof(__kernel_size_t));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002362 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05002364 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05002365 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002366
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367 err = -ENOBUFS;
2368
Anton Blanchard228e5482011-05-02 20:21:35 +00002369 if (msg_sys->msg_controllen > INT_MAX)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002370 goto out;
Tom Herbert28a94d82016-03-07 14:11:02 -08002371 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00002372 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002374 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002375 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002376 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002378 goto out;
Anton Blanchard228e5482011-05-02 20:21:35 +00002379 ctl_buf = msg_sys->msg_control;
2380 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381 } else if (ctl_len) {
David S. Millerac4340f2017-01-04 13:24:19 -05002382 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2383 CMSG_ALIGN(sizeof(struct cmsghdr)));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002384 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002386 if (ctl_buf == NULL)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002387 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388 }
2389 err = -EFAULT;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002390 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002392 msg_sys->msg_control = ctl_buf;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002393 msg_sys->msg_control_is_user = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002394 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002395 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396
2397 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002398 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002399 /*
2400 * If this is sendmmsg() and current destination address is same as
2401 * previously succeeded address, omit asking LSM's decision.
2402 * used_address->name_len is initialized to UINT_MAX so that the first
2403 * destination address never matches.
2404 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002405 if (used_address && msg_sys->msg_name &&
2406 used_address->name_len == msg_sys->msg_namelen &&
2407 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002408 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05002409 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002410 goto out_freectl;
2411 }
Al Virod8725c82014-12-11 00:02:50 -05002412 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002413 /*
2414 * If this is sendmmsg() and sending to current destination address was
2415 * successful, remember it.
2416 */
2417 if (used_address && err >= 0) {
2418 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002419 if (msg_sys->msg_name)
2420 memcpy(&used_address->name, msg_sys->msg_name,
2421 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002422 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423
2424out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002425 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
Jens Axboe4257c8c2019-11-25 14:27:34 -07002427out:
2428 return err;
2429}
2430
Jens Axboe03b12302019-12-02 18:50:25 -07002431int sendmsg_copy_msghdr(struct msghdr *msg,
2432 struct user_msghdr __user *umsg, unsigned flags,
2433 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002434{
2435 int err;
2436
2437 if (flags & MSG_CMSG_COMPAT) {
2438 struct compat_msghdr __user *msg_compat;
2439
2440 msg_compat = (struct compat_msghdr __user *) umsg;
2441 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2442 } else {
2443 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2444 }
2445 if (err < 0)
2446 return err;
2447
2448 return 0;
2449}
2450
2451static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2452 struct msghdr *msg_sys, unsigned int flags,
2453 struct used_address *used_address,
2454 unsigned int allowed_msghdr_flags)
2455{
2456 struct sockaddr_storage address;
2457 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2458 ssize_t err;
2459
2460 msg_sys->msg_name = &address;
2461
2462 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2463 if (err < 0)
2464 return err;
2465
2466 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2467 allowed_msghdr_flags);
Al Viroda184282015-03-21 19:29:06 -04002468 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002469 return err;
2470}
2471
2472/*
2473 * BSD sendmsg interface
2474 */
Jens Axboe03b12302019-12-02 18:50:25 -07002475long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
Jens Axboe0fa03c62019-04-19 13:34:07 -06002476 unsigned int flags)
2477{
Jens Axboe03b12302019-12-02 18:50:25 -07002478 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
Jens Axboe0fa03c62019-04-19 13:34:07 -06002479}
Anton Blanchard228e5482011-05-02 20:21:35 +00002480
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002481long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2482 bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002483{
2484 int fput_needed, err;
2485 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002486 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002487
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002488 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2489 return -EINVAL;
2490
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002491 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002492 if (!sock)
2493 goto out;
2494
Tom Herbert28a94d82016-03-07 14:11:02 -08002495 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00002496
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002497 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002498out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002499 return err;
2500}
2501
Al Viro666547f2014-04-06 14:03:05 -04002502SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002503{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002504 return __sys_sendmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002505}
2506
Anton Blanchard228e5482011-05-02 20:21:35 +00002507/*
2508 * Linux sendmmsg interface
2509 */
2510
2511int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002512 unsigned int flags, bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002513{
2514 int fput_needed, err, datagrams;
2515 struct socket *sock;
2516 struct mmsghdr __user *entry;
2517 struct compat_mmsghdr __user *compat_entry;
2518 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002519 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002520 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002521
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002522 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2523 return -EINVAL;
2524
Anton Blanchard98382f42011-08-04 14:07:39 +00002525 if (vlen > UIO_MAXIOV)
2526 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002527
2528 datagrams = 0;
2529
2530 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2531 if (!sock)
2532 return err;
2533
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002534 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002535 entry = mmsg;
2536 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002537 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002538 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002539
2540 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002541 if (datagrams == vlen - 1)
2542 flags = oflags;
2543
Anton Blanchard228e5482011-05-02 20:21:35 +00002544 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002545 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002546 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002547 if (err < 0)
2548 break;
2549 err = __put_user(err, &compat_entry->msg_len);
2550 ++compat_entry;
2551 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002552 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002553 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002554 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002555 if (err < 0)
2556 break;
2557 err = put_user(err, &entry->msg_len);
2558 ++entry;
2559 }
2560
2561 if (err)
2562 break;
2563 ++datagrams;
Soheil Hassas Yeganeh30238982016-11-04 15:36:49 -04002564 if (msg_data_left(&msg_sys))
2565 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002566 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002567 }
2568
Anton Blanchard228e5482011-05-02 20:21:35 +00002569 fput_light(sock->file, fput_needed);
2570
Anton Blanchard728ffb82011-08-04 14:07:38 +00002571 /* We only return an error if no datagrams were able to be sent */
2572 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002573 return datagrams;
2574
Anton Blanchard228e5482011-05-02 20:21:35 +00002575 return err;
2576}
2577
2578SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2579 unsigned int, vlen, unsigned int, flags)
2580{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002581 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002582}
2583
Jens Axboe03b12302019-12-02 18:50:25 -07002584int recvmsg_copy_msghdr(struct msghdr *msg,
2585 struct user_msghdr __user *umsg, unsigned flags,
2586 struct sockaddr __user **uaddr,
2587 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002588{
2589 ssize_t err;
2590
2591 if (MSG_CMSG_COMPAT & flags) {
2592 struct compat_msghdr __user *msg_compat;
2593
2594 msg_compat = (struct compat_msghdr __user *) umsg;
2595 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2596 } else {
2597 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2598 }
2599 if (err < 0)
2600 return err;
2601
2602 return 0;
2603}
2604
2605static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2606 struct user_msghdr __user *msg,
2607 struct sockaddr __user *uaddr,
2608 unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002610 struct compat_msghdr __user *msg_compat =
Jens Axboe4257c8c2019-11-25 14:27:34 -07002611 (struct compat_msghdr __user *) msg;
2612 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2613 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002614 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002615 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002616 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002617
Al Viro08adb7d2014-11-10 20:23:13 -05002618 msg_sys->msg_name = &addr;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002619 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2620 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002621
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002622 /* We assume all kernel code knows the size of sockaddr_storage */
2623 msg_sys->msg_namelen = 0;
2624
Linus Torvalds1da177e2005-04-16 15:20:36 -07002625 if (sock->file->f_flags & O_NONBLOCK)
2626 flags |= MSG_DONTWAIT;
Eric Dumazet1af66222019-12-06 09:38:36 -08002627
2628 if (unlikely(nosec))
2629 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2630 else
2631 err = sock_recvmsg(sock, msg_sys, flags);
2632
Linus Torvalds1da177e2005-04-16 15:20:36 -07002633 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002634 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635 len = err;
2636
2637 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002638 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002639 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002640 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002641 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002642 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002644 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002645 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002646 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002647 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002648 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002649 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002650 &msg_compat->msg_controllen);
2651 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002652 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653 &msg->msg_controllen);
2654 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002655 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002656 err = len;
Jens Axboe4257c8c2019-11-25 14:27:34 -07002657out:
2658 return err;
2659}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002660
Jens Axboe4257c8c2019-11-25 14:27:34 -07002661static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2662 struct msghdr *msg_sys, unsigned int flags, int nosec)
2663{
2664 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2665 /* user mode address pointers */
2666 struct sockaddr __user *uaddr;
2667 ssize_t err;
2668
2669 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2670 if (err < 0)
2671 return err;
2672
2673 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
Al Viroda184282015-03-21 19:29:06 -04002674 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002675 return err;
2676}
2677
2678/*
2679 * BSD recvmsg interface
2680 */
2681
Jens Axboe03b12302019-12-02 18:50:25 -07002682long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2683 struct user_msghdr __user *umsg,
2684 struct sockaddr __user *uaddr, unsigned int flags)
Jens Axboeaa1fa282019-04-19 13:38:09 -06002685{
Jens Axboe03b12302019-12-02 18:50:25 -07002686 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
Jens Axboeaa1fa282019-04-19 13:38:09 -06002687}
2688
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002689long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2690 bool forbid_cmsg_compat)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002691{
2692 int fput_needed, err;
2693 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002694 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002695
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002696 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2697 return -EINVAL;
2698
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002699 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002700 if (!sock)
2701 goto out;
2702
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002703 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002704
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002705 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706out:
2707 return err;
2708}
2709
Al Viro666547f2014-04-06 14:03:05 -04002710SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002711 unsigned int, flags)
2712{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002713 return __sys_recvmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002714}
2715
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002716/*
2717 * Linux recvmmsg interface
2718 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002719
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002720static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2721 unsigned int vlen, unsigned int flags,
2722 struct timespec64 *timeout)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002723{
2724 int fput_needed, err, datagrams;
2725 struct socket *sock;
2726 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002727 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002728 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002729 struct timespec64 end_time;
2730 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002731
2732 if (timeout &&
2733 poll_select_set_timeout(&end_time, timeout->tv_sec,
2734 timeout->tv_nsec))
2735 return -EINVAL;
2736
2737 datagrams = 0;
2738
2739 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2740 if (!sock)
2741 return err;
2742
Soheil Hassas Yeganeh7797dc42018-02-27 18:22:40 -05002743 if (likely(!(flags & MSG_ERRQUEUE))) {
2744 err = sock_error(sock->sk);
2745 if (err) {
2746 datagrams = err;
2747 goto out_put;
2748 }
Maxime Jayate623a9e2017-02-21 18:35:51 +01002749 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002750
2751 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002752 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002753
2754 while (datagrams < vlen) {
2755 /*
2756 * No need to ask LSM for more than the first datagram.
2757 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002758 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002759 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002760 &msg_sys, flags & ~MSG_WAITFORONE,
2761 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002762 if (err < 0)
2763 break;
2764 err = __put_user(err, &compat_entry->msg_len);
2765 ++compat_entry;
2766 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002767 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002768 (struct user_msghdr __user *)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, &entry->msg_len);
2774 ++entry;
2775 }
2776
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002777 if (err)
2778 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002779 ++datagrams;
2780
Brandon L Black71c5c1592010-03-26 16:18:03 +00002781 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2782 if (flags & MSG_WAITFORONE)
2783 flags |= MSG_DONTWAIT;
2784
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002785 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002786 ktime_get_ts64(&timeout64);
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002787 *timeout = timespec64_sub(end_time, timeout64);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002788 if (timeout->tv_sec < 0) {
2789 timeout->tv_sec = timeout->tv_nsec = 0;
2790 break;
2791 }
2792
2793 /* Timeout, return less than vlen datagrams */
2794 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2795 break;
2796 }
2797
2798 /* Out of band data, return right away */
2799 if (msg_sys.msg_flags & MSG_OOB)
2800 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002801 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002802 }
2803
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002804 if (err == 0)
2805 goto out_put;
2806
2807 if (datagrams == 0) {
2808 datagrams = err;
2809 goto out_put;
2810 }
2811
2812 /*
2813 * We may return less entries than requested (vlen) if the
2814 * sock is non block and there aren't enough datagrams...
2815 */
2816 if (err != -EAGAIN) {
2817 /*
2818 * ... or if recvmsg returns an error after we
2819 * received some datagrams, where we record the
2820 * error to return on the next call or if the
2821 * app asks about it using getsockopt(SO_ERROR).
2822 */
Eric Dumazet159a81d2023-05-09 16:35:53 +00002823 WRITE_ONCE(sock->sk->sk_err, -err);
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002824 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002825out_put:
2826 fput_light(sock->file, fput_needed);
2827
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002828 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002829}
2830
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002831int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2832 unsigned int vlen, unsigned int flags,
2833 struct __kernel_timespec __user *timeout,
2834 struct old_timespec32 __user *timeout32)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002835{
2836 int datagrams;
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002837 struct timespec64 timeout_sys;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002838
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002839 if (timeout && get_timespec64(&timeout_sys, timeout))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002840 return -EFAULT;
2841
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002842 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2843 return -EFAULT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002844
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002845 if (!timeout && !timeout32)
2846 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2847
2848 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2849
2850 if (datagrams <= 0)
2851 return datagrams;
2852
2853 if (timeout && put_timespec64(&timeout_sys, timeout))
2854 datagrams = -EFAULT;
2855
2856 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002857 datagrams = -EFAULT;
2858
2859 return datagrams;
2860}
2861
Dominik Brodowski1255e262018-03-13 20:44:21 +01002862SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2863 unsigned int, vlen, unsigned int, flags,
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002864 struct __kernel_timespec __user *, timeout)
Dominik Brodowski1255e262018-03-13 20:44:21 +01002865{
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002866 if (flags & MSG_CMSG_COMPAT)
2867 return -EINVAL;
2868
2869 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
Dominik Brodowski1255e262018-03-13 20:44:21 +01002870}
2871
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002872#ifdef CONFIG_COMPAT_32BIT_TIME
2873SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2874 unsigned int, vlen, unsigned int, flags,
2875 struct old_timespec32 __user *, timeout)
2876{
2877 if (flags & MSG_CMSG_COMPAT)
2878 return -EINVAL;
2879
2880 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2881}
2882#endif
2883
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002884#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002885/* Argument list sizes for sys_socketcall */
2886#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002887static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002888 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2889 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2890 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002891 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002892};
2893
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894#undef AL
2895
2896/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002897 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002898 *
2899 * Argument checking cleaned up. Saved 20% in size.
2900 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002901 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902 */
2903
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002904SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905{
Chen Gang2950fa92013-04-07 16:55:23 +08002906 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002907 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002908 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07002909 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002910
Anton Blanchard228e5482011-05-02 20:21:35 +00002911 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002912 return -EINVAL;
Jeremy Clinec8e8cd52018-07-27 22:43:01 +00002913 call = array_index_nospec(call, SYS_SENDMMSG + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002914
Arjan van de Ven47379052009-09-28 12:57:44 -07002915 len = nargs[call];
2916 if (len > sizeof(a))
2917 return -EINVAL;
2918
Linus Torvalds1da177e2005-04-16 15:20:36 -07002919 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07002920 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002921 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002922
Chen Gang2950fa92013-04-07 16:55:23 +08002923 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2924 if (err)
2925 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01002926
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002927 a0 = a[0];
2928 a1 = a[1];
2929
2930 switch (call) {
2931 case SYS_SOCKET:
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01002932 err = __sys_socket(a0, a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002933 break;
2934 case SYS_BIND:
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01002935 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002936 break;
2937 case SYS_CONNECT:
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01002938 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002939 break;
2940 case SYS_LISTEN:
Dominik Brodowski25e290e2018-03-13 19:36:54 +01002941 err = __sys_listen(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002942 break;
2943 case SYS_ACCEPT:
Dominik Brodowski4541e802018-03-13 19:24:23 +01002944 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2945 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002946 break;
2947 case SYS_GETSOCKNAME:
2948 err =
Dominik Brodowski8882a102018-03-13 19:43:14 +01002949 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2950 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002951 break;
2952 case SYS_GETPEERNAME:
2953 err =
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002954 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2955 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002956 break;
2957 case SYS_SOCKETPAIR:
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01002958 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002959 break;
2960 case SYS_SEND:
Dominik Brodowskif3bf8962018-03-13 19:52:00 +01002961 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2962 NULL, 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002963 break;
2964 case SYS_SENDTO:
Dominik Brodowski211b6342018-03-13 19:18:52 +01002965 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2966 (struct sockaddr __user *)a[4], a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002967 break;
2968 case SYS_RECV:
Dominik Brodowskid27e9af2018-03-13 19:54:17 +01002969 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2970 NULL, NULL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002971 break;
2972 case SYS_RECVFROM:
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002973 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2974 (struct sockaddr __user *)a[4],
2975 (int __user *)a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002976 break;
2977 case SYS_SHUTDOWN:
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002978 err = __sys_shutdown(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002979 break;
2980 case SYS_SETSOCKOPT:
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002981 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2982 a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002983 break;
2984 case SYS_GETSOCKOPT:
2985 err =
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002986 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2987 (int __user *)a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002988 break;
2989 case SYS_SENDMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002990 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2991 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002992 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00002993 case SYS_SENDMMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002994 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2995 a[3], true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002996 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002997 case SYS_RECVMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002998 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2999 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003000 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003001 case SYS_RECVMMSG:
Arnd Bergmann3ca47e92019-04-23 17:43:50 +02003002 if (IS_ENABLED(CONFIG_64BIT))
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003003 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3004 a[2], a[3],
3005 (struct __kernel_timespec __user *)a[4],
3006 NULL);
3007 else
3008 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3009 a[2], a[3], NULL,
3010 (struct old_timespec32 __user *)a[4]);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003011 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08003012 case SYS_ACCEPT4:
Dominik Brodowski4541e802018-03-13 19:24:23 +01003013 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3014 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07003015 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003016 default:
3017 err = -EINVAL;
3018 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003019 }
3020 return err;
3021}
3022
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003023#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003024
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003025/**
3026 * sock_register - add a socket protocol handler
3027 * @ops: description of protocol
3028 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003029 * This function is called by a protocol handler that wants to
3030 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09003031 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003032 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003034int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003035{
3036 int err;
3037
3038 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08003039 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003040 return -ENOBUFS;
3041 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003042
3043 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00003044 if (rcu_dereference_protected(net_families[ops->family],
3045 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003046 err = -EEXIST;
3047 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00003048 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003049 err = 0;
3050 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003051 spin_unlock(&net_family_lock);
3052
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003053 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003054 return err;
3055}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003056EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003057
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003058/**
3059 * sock_unregister - remove a protocol handler
3060 * @family: protocol family to remove
3061 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003062 * This function is called by a protocol handler that wants to
3063 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003064 * new socket creation.
3065 *
3066 * If protocol handler is a module, then it can use module reference
3067 * counts to protect against new references. If protocol handler is not
3068 * a module then it needs to provide its own protection in
3069 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003071void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003072{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003073 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003074
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003075 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00003076 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003077 spin_unlock(&net_family_lock);
3078
3079 synchronize_rcu();
3080
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003081 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003082}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003083EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003084
Xin Longbf2ae2e2018-03-10 18:57:50 +08003085bool sock_is_registered(int family)
3086{
Jeremy Cline66b51b02018-08-13 22:23:13 +00003087 return family < NPROTO && rcu_access_pointer(net_families[family]);
Xin Longbf2ae2e2018-03-10 18:57:50 +08003088}
3089
Andi Kleen77d76ea2005-12-22 12:43:42 -08003090static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003091{
Nick Pigginb3e19d92011-01-07 17:50:11 +11003092 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00003093 /*
3094 * Initialize the network sysctl infrastructure.
3095 */
3096 err = net_sysctl_init();
3097 if (err)
3098 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11003099
Linus Torvalds1da177e2005-04-16 15:20:36 -07003100 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003101 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07003102 */
3103 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003104
3105 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003106 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003107 */
3108
3109 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11003110
3111 err = register_filesystem(&sock_fs_type);
3112 if (err)
Miaohe Lin47260ba2020-08-06 19:54:19 +08003113 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003114 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003115 if (IS_ERR(sock_mnt)) {
3116 err = PTR_ERR(sock_mnt);
3117 goto out_mount;
3118 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08003119
3120 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003121 */
3122
3123#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00003124 err = netfilter_init();
3125 if (err)
3126 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003127#endif
David S. Millercbeb3212005-12-22 12:58:55 -08003128
Daniel Borkmann408eccc2014-04-01 16:20:23 +02003129 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00003130
Nick Pigginb3e19d92011-01-07 17:50:11 +11003131out:
3132 return err;
3133
3134out_mount:
3135 unregister_filesystem(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003136 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003137}
3138
Andi Kleen77d76ea2005-12-22 12:43:42 -08003139core_initcall(sock_init); /* early initcall */
3140
Linus Torvalds1da177e2005-04-16 15:20:36 -07003141#ifdef CONFIG_PROC_FS
3142void socket_seq_show(struct seq_file *seq)
3143{
Tonghao Zhang648845a2017-12-14 05:51:58 -08003144 seq_printf(seq, "sockets: used %d\n",
3145 sock_inuse_get(seq->private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003146}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003147#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003148
Arnd Bergmann29c49642021-07-22 16:29:03 +02003149/* Handle the fact that while struct ifreq has the same *layout* on
3150 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3151 * which are handled elsewhere, it still has different *size* due to
3152 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3153 * resulting in struct ifreq being 32 and 40 bytes respectively).
3154 * As a result, if the struct happens to be at the end of a page and
3155 * the next page isn't readable/writable, we get a fault. To prevent
3156 * that, copy back and forth to the full size.
3157 */
3158int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3159{
3160 if (in_compat_syscall()) {
3161 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3162
3163 memset(ifr, 0, sizeof(*ifr));
3164 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3165 return -EFAULT;
3166
3167 if (ifrdata)
3168 *ifrdata = compat_ptr(ifr32->ifr_data);
3169
3170 return 0;
3171 }
3172
3173 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3174 return -EFAULT;
3175
3176 if (ifrdata)
3177 *ifrdata = ifr->ifr_data;
3178
3179 return 0;
3180}
3181EXPORT_SYMBOL(get_user_ifreq);
3182
3183int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3184{
3185 size_t size = sizeof(*ifr);
3186
3187 if (in_compat_syscall())
3188 size = sizeof(struct compat_ifreq);
3189
3190 if (copy_to_user(arg, ifr, size))
3191 return -EFAULT;
3192
3193 return 0;
3194}
3195EXPORT_SYMBOL(put_user_ifreq);
3196
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003197#ifdef CONFIG_COMPAT
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003198static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3199{
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003200 compat_uptr_t uptr32;
Al Viro44c02a22017-10-05 12:59:44 -04003201 struct ifreq ifr;
3202 void __user *saved;
3203 int err;
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003204
Arnd Bergmann29c49642021-07-22 16:29:03 +02003205 if (get_user_ifreq(&ifr, NULL, uifr32))
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003206 return -EFAULT;
3207
3208 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3209 return -EFAULT;
3210
Al Viro44c02a22017-10-05 12:59:44 -04003211 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3212 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003213
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003214 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
Al Viro44c02a22017-10-05 12:59:44 -04003215 if (!err) {
3216 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
Arnd Bergmann29c49642021-07-22 16:29:03 +02003217 if (put_user_ifreq(&ifr, uifr32))
Al Viro44c02a22017-10-05 12:59:44 -04003218 err = -EFAULT;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003219 }
Al Viro44c02a22017-10-05 12:59:44 -04003220 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003221}
3222
Ben Hutchings590d4692013-11-18 17:04:13 +00003223/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3224static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08003225 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003226{
Al Viro44c02a22017-10-05 12:59:44 -04003227 struct ifreq ifreq;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003228 void __user *data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003229
Peter Collingbourned0efb162021-08-26 12:46:01 -07003230 if (!is_socket_ioctl_cmd(cmd))
3231 return -ENOTTY;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003232 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003233 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003234 ifreq.ifr_data = data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003235
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003236 return dev_ioctl(net, cmd, &ifreq, data, NULL);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003237}
3238
Arnd Bergmann7a229382009-11-06 23:00:29 -08003239/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3240 * for some operations; this forces use of the newer bridge-utils that
Lucas De Marchi25985ed2011-03-30 22:57:33 -03003241 * use compatible ioctls
Arnd Bergmann7a229382009-11-06 23:00:29 -08003242 */
Arnd Bergmann6b960182009-11-06 23:10:54 -08003243static int old_bridge_ioctl(compat_ulong_t __user *argp)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003244{
Arnd Bergmann6b960182009-11-06 23:10:54 -08003245 compat_ulong_t tmp;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003246
Arnd Bergmann6b960182009-11-06 23:10:54 -08003247 if (get_user(tmp, argp))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003248 return -EFAULT;
3249 if (tmp == BRCTL_GET_VERSION)
3250 return BRCTL_VERSION + 1;
3251 return -EINVAL;
3252}
3253
Arnd Bergmann6b960182009-11-06 23:10:54 -08003254static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3255 unsigned int cmd, unsigned long arg)
3256{
3257 void __user *argp = compat_ptr(arg);
3258 struct sock *sk = sock->sk;
3259 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003260
Arnd Bergmann6b960182009-11-06 23:10:54 -08003261 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Arnd Bergmann88fc0232021-07-27 15:45:15 +02003262 return sock_ioctl(file, cmd, (unsigned long)argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003263
Arnd Bergmann6b960182009-11-06 23:10:54 -08003264 switch (cmd) {
3265 case SIOCSIFBR:
3266 case SIOCGIFBR:
3267 return old_bridge_ioctl(argp);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003268 case SIOCWANDEV:
3269 return compat_siocwandev(net, argp);
Arnd Bergmann0768e172019-04-17 22:56:11 +02003270 case SIOCGSTAMP_OLD:
3271 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003272 if (!sock->ops->gettstamp)
3273 return -ENOIOCTLCMD;
Arnd Bergmann0768e172019-04-17 22:56:11 +02003274 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003275 !COMPAT_USE_64BIT_TIME);
3276
Arnd Bergmanndd98d282021-07-22 16:28:59 +02003277 case SIOCETHTOOL:
Ben Hutchings590d4692013-11-18 17:04:13 +00003278 case SIOCBONDSLAVEINFOQUERY:
3279 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003280 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003281 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003282 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003283
Arnd Bergmann6b960182009-11-06 23:10:54 -08003284 case FIOSETOWN:
3285 case SIOCSPGRP:
3286 case FIOGETOWN:
3287 case SIOCGPGRP:
3288 case SIOCBRADDBR:
3289 case SIOCBRDELBR:
3290 case SIOCGIFVLAN:
3291 case SIOCSIFVLAN:
Andrey Vaginc62cce22016-10-24 18:29:13 -07003292 case SIOCGSKNS:
Arnd Bergmann0768e172019-04-17 22:56:11 +02003293 case SIOCGSTAMP_NEW:
3294 case SIOCGSTAMPNS_NEW:
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02003295 case SIOCGIFCONF:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003296 return sock_ioctl(file, cmd, arg);
3297
3298 case SIOCGIFFLAGS:
3299 case SIOCSIFFLAGS:
Arnd Bergmann709566d2021-07-22 16:29:00 +02003300 case SIOCGIFMAP:
3301 case SIOCSIFMAP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003302 case SIOCGIFMETRIC:
3303 case SIOCSIFMETRIC:
3304 case SIOCGIFMTU:
3305 case SIOCSIFMTU:
3306 case SIOCGIFMEM:
3307 case SIOCSIFMEM:
3308 case SIOCGIFHWADDR:
3309 case SIOCSIFHWADDR:
3310 case SIOCADDMULTI:
3311 case SIOCDELMULTI:
3312 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003313 case SIOCGIFADDR:
3314 case SIOCSIFADDR:
3315 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003316 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003317 case SIOCGIFBRDADDR:
3318 case SIOCSIFBRDADDR:
3319 case SIOCGIFDSTADDR:
3320 case SIOCSIFDSTADDR:
3321 case SIOCGIFNETMASK:
3322 case SIOCSIFNETMASK:
3323 case SIOCSIFPFLAGS:
3324 case SIOCGIFPFLAGS:
3325 case SIOCGIFTXQLEN:
3326 case SIOCSIFTXQLEN:
3327 case SIOCBRADDIF:
3328 case SIOCBRDELIF:
Johannes Bergc6c9fee2019-01-25 22:43:19 +01003329 case SIOCGIFNAME:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003330 case SIOCSIFNAME:
3331 case SIOCGMIIPHY:
3332 case SIOCGMIIREG:
3333 case SIOCSMIIREG:
Al Virof92d4fc2017-09-30 19:32:17 -04003334 case SIOCBONDENSLAVE:
3335 case SIOCBONDRELEASE:
3336 case SIOCBONDSETHWADDR:
3337 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003338 case SIOCSARP:
3339 case SIOCGARP:
3340 case SIOCDARP:
Arnd Bergmannc7dc5042019-06-03 23:07:12 +02003341 case SIOCOUTQ:
Arnd Bergmann9d7bf412019-06-03 23:06:00 +02003342 case SIOCOUTQNSD:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003343 case SIOCATMARK:
Johannes Berg63ff03a2019-01-25 22:43:17 +01003344 return sock_do_ioctl(net, sock, cmd, arg);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003345 }
3346
Arnd Bergmann6b960182009-11-06 23:10:54 -08003347 return -ENOIOCTLCMD;
3348}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003349
Eric Dumazet95c96172012-04-15 05:58:06 +00003350static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003351 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003352{
3353 struct socket *sock = file->private_data;
3354 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003355 struct sock *sk;
3356 struct net *net;
3357
3358 sk = sock->sk;
3359 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003360
3361 if (sock->ops->compat_ioctl)
3362 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3363
David S. Miller87de87d2008-06-03 09:14:03 -07003364 if (ret == -ENOIOCTLCMD &&
3365 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3366 ret = compat_wext_handle_ioctl(net, cmd, arg);
3367
Arnd Bergmann6b960182009-11-06 23:10:54 -08003368 if (ret == -ENOIOCTLCMD)
3369 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3370
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003371 return ret;
3372}
3373#endif
3374
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003375/**
3376 * kernel_bind - bind an address to a socket (kernel space)
3377 * @sock: socket
3378 * @addr: address
3379 * @addrlen: length of address
3380 *
3381 * Returns 0 or an error.
3382 */
3383
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003384int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3385{
3386 return sock->ops->bind(sock, addr, addrlen);
3387}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003388EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003389
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003390/**
3391 * kernel_listen - move socket to listening state (kernel space)
3392 * @sock: socket
3393 * @backlog: pending connections queue size
3394 *
3395 * Returns 0 or an error.
3396 */
3397
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003398int kernel_listen(struct socket *sock, int backlog)
3399{
3400 return sock->ops->listen(sock, backlog);
3401}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003402EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003403
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003404/**
3405 * kernel_accept - accept a connection (kernel space)
3406 * @sock: listening socket
3407 * @newsock: new connected socket
3408 * @flags: flags
3409 *
3410 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3411 * If it fails, @newsock is guaranteed to be %NULL.
3412 * Returns 0 or an error.
3413 */
3414
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003415int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3416{
3417 struct sock *sk = sock->sk;
3418 int err;
3419
3420 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3421 newsock);
3422 if (err < 0)
3423 goto done;
3424
David Howellscdfbabf2017-03-09 08:09:05 +00003425 err = sock->ops->accept(sock, *newsock, flags, true);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003426 if (err < 0) {
3427 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003428 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003429 goto done;
3430 }
3431
3432 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003433 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003434
3435done:
3436 return err;
3437}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003438EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003439
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003440/**
3441 * kernel_connect - connect a socket (kernel space)
3442 * @sock: socket
3443 * @addr: address
3444 * @addrlen: address length
3445 * @flags: flags (O_NONBLOCK, ...)
3446 *
Lu Weif1dcffc2021-03-25 11:01:55 +08003447 * For datagram sockets, @addr is the address to which datagrams are sent
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003448 * by default, and the only address from which datagrams are received.
3449 * For stream sockets, attempts to connect to @addr.
3450 * Returns 0 or an error code.
3451 */
3452
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003453int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003454 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003455{
Jordan Rife759d51e2023-08-21 16:45:23 -05003456 struct sockaddr_storage address;
3457
3458 memcpy(&address, addr, addrlen);
3459
3460 return sock->ops->connect(sock, (struct sockaddr *)&address, addrlen, flags);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003461}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003462EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003463
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003464/**
3465 * kernel_getsockname - get the address which the socket is bound (kernel space)
3466 * @sock: socket
3467 * @addr: address holder
3468 *
3469 * Fills the @addr pointer with the address which the socket is bound.
3470 * Returns 0 or an error code.
3471 */
3472
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003473int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003474{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003475 return sock->ops->getname(sock, addr, 0);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003476}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003477EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003478
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003479/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003480 * kernel_getpeername - get the address which the socket is connected (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003481 * @sock: socket
3482 * @addr: address holder
3483 *
3484 * Fills the @addr pointer with the address which the socket is connected.
3485 * Returns 0 or an error code.
3486 */
3487
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003488int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003489{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003490 return sock->ops->getname(sock, addr, 1);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003491}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003492EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003493
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003494/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003495 * kernel_sendpage - send a &page through a socket (kernel space)
3496 * @sock: socket
3497 * @page: page
3498 * @offset: page offset
3499 * @size: total size in bytes
3500 * @flags: flags (MSG_DONTWAIT, ...)
3501 *
3502 * Returns the total amount sent in bytes or an error.
3503 */
3504
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003505int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3506 size_t size, int flags)
3507{
Coly Li7b62d312020-10-02 16:27:29 +08003508 if (sock->ops->sendpage) {
3509 /* Warn in case the improper page to zero-copy send */
3510 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003511 return sock->ops->sendpage(sock, page, offset, size, flags);
Coly Li7b62d312020-10-02 16:27:29 +08003512 }
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003513 return sock_no_sendpage(sock, page, offset, size, flags);
3514}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003515EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003516
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003517/**
3518 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3519 * @sk: sock
3520 * @page: page
3521 * @offset: page offset
3522 * @size: total size in bytes
3523 * @flags: flags (MSG_DONTWAIT, ...)
3524 *
3525 * Returns the total amount sent in bytes or an error.
3526 * Caller must hold @sk.
3527 */
3528
Tom Herbert306b13e2017-07-28 16:22:41 -07003529int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3530 size_t size, int flags)
3531{
3532 struct socket *sock = sk->sk_socket;
3533
3534 if (sock->ops->sendpage_locked)
3535 return sock->ops->sendpage_locked(sk, page, offset, size,
3536 flags);
3537
3538 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3539}
3540EXPORT_SYMBOL(kernel_sendpage_locked);
3541
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003542/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003543 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003544 * @sock: socket
3545 * @how: connection part
3546 *
3547 * Returns 0 or an error.
3548 */
3549
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003550int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3551{
3552 return sock->ops->shutdown(sock, how);
3553}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003554EXPORT_SYMBOL(kernel_sock_shutdown);
R. Parameswaran113c3072017-04-05 16:50:35 -07003555
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003556/**
3557 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3558 * @sk: socket
3559 *
3560 * This routine returns the IP overhead imposed by a socket i.e.
3561 * the length of the underlying IP header, depending on whether
3562 * this is an IPv4 or IPv6 socket and the length from IP options turned
3563 * on at the socket. Assumes that the caller has a lock on the socket.
R. Parameswaran113c3072017-04-05 16:50:35 -07003564 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003565
R. Parameswaran113c3072017-04-05 16:50:35 -07003566u32 kernel_sock_ip_overhead(struct sock *sk)
3567{
3568 struct inet_sock *inet;
3569 struct ip_options_rcu *opt;
3570 u32 overhead = 0;
R. Parameswaran113c3072017-04-05 16:50:35 -07003571#if IS_ENABLED(CONFIG_IPV6)
3572 struct ipv6_pinfo *np;
3573 struct ipv6_txoptions *optv6 = NULL;
3574#endif /* IS_ENABLED(CONFIG_IPV6) */
3575
3576 if (!sk)
3577 return overhead;
3578
R. Parameswaran113c3072017-04-05 16:50:35 -07003579 switch (sk->sk_family) {
3580 case AF_INET:
3581 inet = inet_sk(sk);
3582 overhead += sizeof(struct iphdr);
3583 opt = rcu_dereference_protected(inet->inet_opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003584 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003585 if (opt)
3586 overhead += opt->opt.optlen;
3587 return overhead;
3588#if IS_ENABLED(CONFIG_IPV6)
3589 case AF_INET6:
3590 np = inet6_sk(sk);
3591 overhead += sizeof(struct ipv6hdr);
3592 if (np)
3593 optv6 = rcu_dereference_protected(np->opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003594 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003595 if (optv6)
3596 overhead += (optv6->opt_flen + optv6->opt_nflen);
3597 return overhead;
3598#endif /* IS_ENABLED(CONFIG_IPV6) */
3599 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3600 return overhead;
3601 }
3602}
3603EXPORT_SYMBOL(kernel_sock_ip_overhead);