blob: fcbdd5bc47ac2ff98bb243a2c9ec4f91fd64e778 [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 Kicinskiaef2fed2021-12-15 18:55:37 -080055#include <linux/bpf-cgroup.h>
Jakub Kicinskicc698372020-11-20 14:50:52 -080056#include <linux/ethtool.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058#include <linux/socket.h>
59#include <linux/file.h>
David Howells2dc334f2023-06-07 19:19:09 +010060#include <linux/splice.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070061#include <linux/net.h>
62#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070063#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070064#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#include <linux/netdevice.h>
66#include <linux/proc_fs.h>
67#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080068#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070069#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030070#include <linux/if_vlan.h>
Daniel Borkmann408eccc2014-04-01 16:20:23 +020071#include <linux/ptp_classify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#include <linux/init.h>
73#include <linux/poll.h>
74#include <linux/cache.h>
75#include <linux/module.h>
76#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070077#include <linux/mount.h>
David Howellsfba9be42019-03-25 16:38:24 +000078#include <linux/pseudo_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#include <linux/security.h>
80#include <linux/syscalls.h>
81#include <linux/compat.h>
82#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010083#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010084#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070085#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070086#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090087#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000088#include <linux/xattr.h>
Jeremy Clinec8e8cd52018-07-27 22:43:01 +000089#include <linux/nospec.h>
Paolo Abeni8c3c4472019-05-03 17:01:39 +020090#include <linux/indirect_call_wrapper.h>
Pavel Begunkov8c9a6f52024-04-09 14:05:53 -070091#include <linux/io_uring/net.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080093#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070094#include <asm/unistd.h>
95
96#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070097#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070098#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
Arnd Bergmann6b960182009-11-06 23:10:54 -0800103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
Arnd Bergmannc7dc5042019-06-03 23:07:12 +0200106#include <linux/termios.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800107#include <linux/sockios.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300108#include <net/busy_poll.h>
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400109#include <linux/errqueue.h>
Yangbo Lud7c08822021-06-30 16:12:00 +0800110#include <linux/ptp_clock_kernel.h>
Yunhui Cui6e6eda42023-01-11 14:59:30 +0800111#include <trace/events/sock.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300112
Cong Wange0d10952013-08-01 11:10:25 +0800113#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300114unsigned int sysctl_net_busy_read __read_mostly;
115unsigned int sysctl_net_busy_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300116#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800117
Al Viro8ae5e0302014-11-28 19:40:50 -0500118static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
119static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700120static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121
122static int sock_close(struct inode *inode, struct file *file);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -0700123static __poll_t sock_poll(struct file *file,
124 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700125static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800126#ifdef CONFIG_COMPAT
127static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700128 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800129#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130static int sock_fasync(int fd, struct file *filp, int on);
Jens Axboe9c55e012007-11-06 23:30:13 -0800131static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700132 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800133 unsigned int flags);
David Howells2bfc6682023-06-07 19:19:10 +0100134static void sock_splice_eof(struct file *file);
Arnd Bergmann542d3062020-01-08 22:44:43 +0100135
136#ifdef CONFIG_PROC_FS
137static void sock_show_fdinfo(struct seq_file *m, struct file *f)
138{
139 struct socket *sock = f->private_data;
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000140 const struct proto_ops *ops = READ_ONCE(sock->ops);
Arnd Bergmann542d3062020-01-08 22:44:43 +0100141
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000142 if (ops->show_fdinfo)
143 ops->show_fdinfo(m, sock);
Arnd Bergmann542d3062020-01-08 22:44:43 +0100144}
145#else
146#define sock_show_fdinfo NULL
147#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149/*
150 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
151 * in the operation structures but are done directly via the socketcall() multiplexor.
152 */
153
Arjan van de Venda7071d2007-02-12 00:55:36 -0800154static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 .owner = THIS_MODULE,
156 .llseek = no_llseek,
Al Viro8ae5e0302014-11-28 19:40:50 -0500157 .read_iter = sock_read_iter,
158 .write_iter = sock_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 .poll = sock_poll,
160 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800161#ifdef CONFIG_COMPAT
162 .compat_ioctl = compat_sock_ioctl,
163#endif
Breno Leitao8e9fad02023-06-27 06:44:24 -0700164 .uring_cmd = io_uring_cmd_sock,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 .mmap = sock_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 .release = sock_close,
167 .fasync = sock_fasync,
David Howells2dc334f2023-06-07 19:19:09 +0100168 .splice_write = splice_to_socket,
Jens Axboe9c55e012007-11-06 23:30:13 -0800169 .splice_read = sock_splice_read,
David Howells2bfc6682023-06-07 19:19:10 +0100170 .splice_eof = sock_splice_eof,
Kirill Tkhaib4653342019-12-09 13:03:40 +0300171 .show_fdinfo = sock_show_fdinfo,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172};
173
Yejune Dengfe0bdbd2021-06-21 13:12:25 +0800174static const char * const pf_family_names[] = {
175 [PF_UNSPEC] = "PF_UNSPEC",
176 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
177 [PF_INET] = "PF_INET",
178 [PF_AX25] = "PF_AX25",
179 [PF_IPX] = "PF_IPX",
180 [PF_APPLETALK] = "PF_APPLETALK",
181 [PF_NETROM] = "PF_NETROM",
182 [PF_BRIDGE] = "PF_BRIDGE",
183 [PF_ATMPVC] = "PF_ATMPVC",
184 [PF_X25] = "PF_X25",
185 [PF_INET6] = "PF_INET6",
186 [PF_ROSE] = "PF_ROSE",
187 [PF_DECnet] = "PF_DECnet",
188 [PF_NETBEUI] = "PF_NETBEUI",
189 [PF_SECURITY] = "PF_SECURITY",
190 [PF_KEY] = "PF_KEY",
191 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
192 [PF_PACKET] = "PF_PACKET",
193 [PF_ASH] = "PF_ASH",
194 [PF_ECONET] = "PF_ECONET",
195 [PF_ATMSVC] = "PF_ATMSVC",
196 [PF_RDS] = "PF_RDS",
197 [PF_SNA] = "PF_SNA",
198 [PF_IRDA] = "PF_IRDA",
199 [PF_PPPOX] = "PF_PPPOX",
200 [PF_WANPIPE] = "PF_WANPIPE",
201 [PF_LLC] = "PF_LLC",
202 [PF_IB] = "PF_IB",
203 [PF_MPLS] = "PF_MPLS",
204 [PF_CAN] = "PF_CAN",
205 [PF_TIPC] = "PF_TIPC",
206 [PF_BLUETOOTH] = "PF_BLUETOOTH",
207 [PF_IUCV] = "PF_IUCV",
208 [PF_RXRPC] = "PF_RXRPC",
209 [PF_ISDN] = "PF_ISDN",
210 [PF_PHONET] = "PF_PHONET",
211 [PF_IEEE802154] = "PF_IEEE802154",
212 [PF_CAIF] = "PF_CAIF",
213 [PF_ALG] = "PF_ALG",
214 [PF_NFC] = "PF_NFC",
215 [PF_VSOCK] = "PF_VSOCK",
216 [PF_KCM] = "PF_KCM",
217 [PF_QIPCRTR] = "PF_QIPCRTR",
218 [PF_SMC] = "PF_SMC",
219 [PF_XDP] = "PF_XDP",
Jeremy Kerrbc49d812021-07-29 10:20:39 +0800220 [PF_MCTP] = "PF_MCTP",
Yejune Dengfe0bdbd2021-06-21 13:12:25 +0800221};
222
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223/*
224 * The protocol list. Each protocol is registered in here.
225 */
226
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000228static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700231 * Support routines.
232 * Move socket addresses back and forth across the kernel/user
233 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 */
235
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236/**
237 * move_addr_to_kernel - copy a socket address into kernel space
238 * @uaddr: Address in user space
239 * @kaddr: Address in kernel space
240 * @ulen: Length in user space
241 *
242 * The address is copied into kernel space. If the provided address is
243 * too long an error code of -EINVAL is returned. If the copy gives
244 * invalid addresses -EFAULT is returned. On a success 0 is returned.
245 */
246
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000247int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700249 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700251 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700253 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100255 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256}
257
258/**
259 * move_addr_to_user - copy an address to user space
260 * @kaddr: kernel space address
261 * @klen: length of address in kernel
262 * @uaddr: user space address
263 * @ulen: pointer to user length field
264 *
265 * The value pointed to by ulen on entry is the buffer length available.
266 * This is overwritten with the buffer space used. -EINVAL is returned
267 * if an overlong buffer is specified or a negative buffer size. -EFAULT
268 * is returned if either the buffer or the length field are not
269 * accessible.
270 * After copying the data up to the limit the user specifies, the true
271 * length of the data is written over the length limit the user
272 * specified. Zero is returned for a success.
273 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700274
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000275static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000276 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277{
278 int err;
279 int len;
280
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100281 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700282 err = get_user(len, ulen);
283 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700285 if (len > klen)
286 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100287 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700289 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500290 if (audit_sockaddr(klen, kaddr))
291 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700292 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 return -EFAULT;
294 }
295 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700296 * "fromlen shall refer to the value before truncation.."
297 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298 */
299 return __put_user(klen, ulen);
300}
301
Alexey Dobriyan08009a72018-02-24 21:20:33 +0300302static struct kmem_cache *sock_inode_cachep __ro_after_init;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303
304static struct inode *sock_alloc_inode(struct super_block *sb)
305{
306 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700307
Muchun Songfd60b282022-03-22 14:41:03 -0700308 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 if (!ei)
310 return NULL;
Al Viro333f7902019-07-05 20:14:16 +0100311 init_waitqueue_head(&ei->socket.wq.wait);
312 ei->socket.wq.fasync_list = NULL;
313 ei->socket.wq.flags = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700314
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 ei->socket.state = SS_UNCONNECTED;
316 ei->socket.flags = 0;
317 ei->socket.ops = NULL;
318 ei->socket.sk = NULL;
319 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
321 return &ei->vfs_inode;
322}
323
Al Viro6d7855c2019-07-05 20:13:22 +0100324static void sock_free_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325{
Eric Dumazet43815482010-04-29 11:01:49 +0000326 struct socket_alloc *ei;
327
328 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazet43815482010-04-29 11:01:49 +0000329 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330}
331
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700332static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700334 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335
Christoph Lametera35afb82007-05-16 22:10:57 -0700336 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700338
yuan linyu1e911632017-01-07 17:18:31 +0800339static void init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
341 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700342 sizeof(struct socket_alloc),
343 0,
344 (SLAB_HWCACHE_ALIGN |
345 SLAB_RECLAIM_ACCOUNT |
Chengming Zhoud4f01c52024-02-28 03:06:58 +0000346 SLAB_ACCOUNT),
Paul Mundt20c2df82007-07-20 10:11:58 +0900347 init_once);
yuan linyu1e911632017-01-07 17:18:31 +0800348 BUG_ON(sock_inode_cachep == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349}
350
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700351static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700352 .alloc_inode = sock_alloc_inode,
Al Viro6d7855c2019-07-05 20:13:22 +0100353 .free_inode = sock_free_inode,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700354 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355};
356
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700357/*
358 * sockfs_dname() is called from d_path().
359 */
360static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
361{
Al Viro0f60d282022-01-30 15:03:49 -0500362 return dynamic_dname(buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000363 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700364}
365
Al Viro3ba13d12009-02-20 06:02:22 +0000366static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700367 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368};
369
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200370static int sockfs_xattr_get(const struct xattr_handler *handler,
371 struct dentry *dentry, struct inode *inode,
372 const char *suffix, void *value, size_t size)
373{
374 if (value) {
375 if (dentry->d_name.len + 1 > size)
376 return -ERANGE;
377 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
378 }
379 return dentry->d_name.len + 1;
380}
381
382#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
383#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
384#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
385
386static const struct xattr_handler sockfs_xattr_handler = {
387 .name = XATTR_NAME_SOCKPROTONAME,
388 .get = sockfs_xattr_get,
389};
390
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100391static int sockfs_security_xattr_set(const struct xattr_handler *handler,
Christian Brauner39f60c12023-01-13 12:49:23 +0100392 struct mnt_idmap *idmap,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100393 struct dentry *dentry, struct inode *inode,
394 const char *suffix, const void *value,
395 size_t size, int flags)
396{
397 /* Handled by LSM. */
398 return -EAGAIN;
399}
400
401static const struct xattr_handler sockfs_security_xattr_handler = {
402 .prefix = XATTR_SECURITY_PREFIX,
403 .set = sockfs_security_xattr_set,
404};
405
Wedson Almeida Filho295d3c42023-09-30 02:00:33 -0300406static const struct xattr_handler * const sockfs_xattr_handlers[] = {
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200407 &sockfs_xattr_handler,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100408 &sockfs_security_xattr_handler,
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200409 NULL
410};
411
David Howellsfba9be42019-03-25 16:38:24 +0000412static int sockfs_init_fs_context(struct fs_context *fc)
Al Viroc74a1cb2011-01-12 16:59:34 -0500413{
David Howellsfba9be42019-03-25 16:38:24 +0000414 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
415 if (!ctx)
416 return -ENOMEM;
417 ctx->ops = &sockfs_ops;
418 ctx->dops = &sockfs_dentry_operations;
419 ctx->xattr = sockfs_xattr_handlers;
420 return 0;
Al Viroc74a1cb2011-01-12 16:59:34 -0500421}
422
423static struct vfsmount *sock_mnt __read_mostly;
424
425static struct file_system_type sock_fs_type = {
426 .name = "sockfs",
David Howellsfba9be42019-03-25 16:38:24 +0000427 .init_fs_context = sockfs_init_fs_context,
Al Viroc74a1cb2011-01-12 16:59:34 -0500428 .kill_sb = kill_anon_super,
429};
430
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431/*
432 * Obtains the first available file descriptor and sets it up for use.
433 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800434 * These functions create file structures and maps them to fd space
435 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 * and file struct implicitly stored in sock->file.
437 * Note that another thread may close file descriptor before we return
438 * from this function. We use the fact that now we do not refer
439 * to socket after mapping. If one day we will need it, this
440 * function will increment ref. count on file by 1.
441 *
442 * In any case returned fd MAY BE not valid!
443 * This race condition is unavoidable
444 * with shared fd spaces, we cannot solve it inside kernel,
445 * but we take care of internal coherence yet.
446 */
447
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300448/**
449 * sock_alloc_file - Bind a &socket to a &file
450 * @sock: socket
451 * @flags: file status flags
452 * @dname: protocol name
453 *
454 * Returns the &file bound with @sock, implicitly storing it
455 * in sock->file. If dname is %NULL, sets to "".
Thadeu Lima de Souza Cascardo649c15c2023-03-07 14:37:07 -0300456 *
457 * On failure @sock is released, and an ERR pointer is returned.
458 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300459 * This function uses GFP_KERNEL internally.
460 */
461
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700462struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463{
Al Viro7cbe66b2009-08-05 19:59:08 +0400464 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800465
Al Virod93aa9d2018-06-09 09:40:05 -0400466 if (!dname)
467 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468
Al Virod93aa9d2018-06-09 09:40:05 -0400469 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
470 O_RDWR | (flags & O_NONBLOCK),
471 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530472 if (IS_ERR(file)) {
Al Viro8e1611e2017-12-05 23:29:09 +0000473 sock_release(sock);
Anatol Pomozov39b652522012-09-12 20:11:55 -0700474 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400475 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800476
Jens Axboefe34db02023-05-09 09:19:08 -0600477 file->f_mode |= FMODE_NOWAIT;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800478 sock->file = file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800479 file->private_data = sock;
Linus Torvaldsd8e464e2019-11-17 11:20:48 -0800480 stream_open(SOCK_INODE(sock), file);
Al Viro28407632012-08-17 23:54:15 -0400481 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800482}
Al Viro56b31d12012-08-18 00:25:51 -0400483EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800484
Al Viro56b31d12012-08-18 00:25:51 -0400485static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800486{
487 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400488 int fd = get_unused_fd_flags(flags);
Al Viroce4bb042018-01-10 18:47:05 -0500489 if (unlikely(fd < 0)) {
490 sock_release(sock);
Al Viro28407632012-08-17 23:54:15 -0400491 return fd;
Al Viroce4bb042018-01-10 18:47:05 -0500492 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800493
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700494 newfile = sock_alloc_file(sock, flags, NULL);
Enrico Weigelt4546e442019-06-05 22:58:50 +0200495 if (!IS_ERR(newfile)) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800496 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400497 return fd;
498 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400499
Al Viro28407632012-08-17 23:54:15 -0400500 put_unused_fd(fd);
501 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502}
503
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300504/**
505 * sock_from_file - Return the &socket bounded to @file.
506 * @file: file
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300507 *
Florent Revestdba4a922020-12-04 12:36:04 +0100508 * On failure returns %NULL.
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300509 */
510
Florent Revestdba4a922020-12-04 12:36:04 +0100511struct socket *sock_from_file(struct file *file)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800512{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800513 if (file->f_op == &socket_file_ops)
Jens Axboeda214a42022-04-12 14:22:39 -0600514 return file->private_data; /* set in sock_alloc_file */
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800515
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800516 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800517}
John Fastabend406a3c62012-07-20 10:39:25 +0000518EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800519
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700521 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 * @fd: file handle
523 * @err: pointer to an error code return
524 *
525 * The file handle passed in is locked and the socket it is bound
Rosen, Rami241c4662017-05-21 22:12:38 +0300526 * to is returned. If an error occurs the err pointer is overwritten
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 * with a negative errno code and NULL is returned. The function checks
528 * for both invalid handles and passing a handle which is not a socket.
529 *
530 * On a success the socket object pointer is returned.
531 */
532
533struct socket *sockfd_lookup(int fd, int *err)
534{
535 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 struct socket *sock;
537
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700538 file = fget(fd);
539 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 *err = -EBADF;
541 return NULL;
542 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700543
Florent Revestdba4a922020-12-04 12:36:04 +0100544 sock = sock_from_file(file);
545 if (!sock) {
546 *err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 fput(file);
Florent Revestdba4a922020-12-04 12:36:04 +0100548 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 return sock;
550}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700551EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800553static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
554{
Al Viro00e188e2014-03-03 23:48:18 -0500555 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800556 struct socket *sock;
557
Hua Zhong36725582006-04-19 15:25:02 -0700558 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500559 if (f.file) {
Florent Revestdba4a922020-12-04 12:36:04 +0100560 sock = sock_from_file(f.file);
Al Viro00e188e2014-03-03 23:48:18 -0500561 if (likely(sock)) {
Miaohe Lince787a52020-08-06 19:53:16 +0800562 *fput_needed = f.flags & FDPUT_FPUT;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800563 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500564 }
Florent Revestdba4a922020-12-04 12:36:04 +0100565 *err = -ENOTSOCK;
Al Viro00e188e2014-03-03 23:48:18 -0500566 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800567 }
568 return NULL;
569}
570
Masatake YAMATO600e1772012-08-29 10:44:29 +0000571static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
572 size_t size)
573{
574 ssize_t len;
575 ssize_t used = 0;
576
David Howellsc5ef6032015-03-17 22:26:16 +0000577 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000578 if (len < 0)
579 return len;
580 used += len;
581 if (buffer) {
582 if (size < used)
583 return -ERANGE;
584 buffer += len;
585 }
586
587 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
588 used += len;
589 if (buffer) {
590 if (size < used)
591 return -ERANGE;
592 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
593 buffer += len;
594 }
595
596 return used;
597}
598
Christian Braunerc1632a02023-01-13 12:49:11 +0100599static int sockfs_setattr(struct mnt_idmap *idmap,
Christian Brauner549c7292021-01-21 14:19:43 +0100600 struct dentry *dentry, struct iattr *iattr)
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900601{
Christian Braunerc1632a02023-01-13 12:49:11 +0100602 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900603
Eric Biggerse1a3a602016-12-30 17:42:32 -0600604 if (!err && (iattr->ia_valid & ATTR_UID)) {
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900605 struct socket *sock = SOCKET_I(d_inode(dentry));
606
Cong Wang6d8c50d2018-06-07 13:39:49 -0700607 if (sock->sk)
608 sock->sk->sk_uid = iattr->ia_uid;
609 else
610 err = -ENOENT;
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900611 }
612
613 return err;
614}
615
Masatake YAMATO600e1772012-08-29 10:44:29 +0000616static const struct inode_operations sockfs_inode_ops = {
Masatake YAMATO600e1772012-08-29 10:44:29 +0000617 .listxattr = sockfs_listxattr,
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900618 .setattr = sockfs_setattr,
Masatake YAMATO600e1772012-08-29 10:44:29 +0000619};
620
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300622 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700623 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 * Allocate a new inode and socket object. The two are bound together
625 * and initialised. The socket is then returned. If we are out of inodes
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300626 * NULL is returned. This functions uses GFP_KERNEL internally.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 */
628
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800629struct socket *sock_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700631 struct inode *inode;
632 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200634 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 if (!inode)
636 return NULL;
637
638 sock = SOCKET_I(inode);
639
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400640 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700641 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100642 inode->i_uid = current_fsuid();
643 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000644 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 return sock;
647}
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800648EXPORT_SYMBOL(sock_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
Cong Wang6d8c50d2018-06-07 13:39:49 -0700650static void __sock_release(struct socket *sock, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651{
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000652 const struct proto_ops *ops = READ_ONCE(sock->ops);
653
654 if (ops) {
655 struct module *owner = ops->owner;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
Cong Wang6d8c50d2018-06-07 13:39:49 -0700657 if (inode)
658 inode_lock(inode);
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000659 ops->release(sock);
Eric Biggersff7b11a2019-02-21 14:13:56 -0800660 sock->sk = NULL;
Cong Wang6d8c50d2018-06-07 13:39:49 -0700661 if (inode)
662 inode_unlock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 sock->ops = NULL;
664 module_put(owner);
665 }
666
Al Viro333f7902019-07-05 20:14:16 +0100667 if (sock->wq.fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800668 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 if (!sock->file) {
671 iput(SOCK_INODE(sock));
672 return;
673 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700674 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675}
Cong Wang6d8c50d2018-06-07 13:39:49 -0700676
Andrew Lunn9a8ad9a2020-07-13 01:15:12 +0200677/**
678 * sock_release - close a socket
679 * @sock: socket to close
680 *
681 * The socket is released from the protocol stack if it has a release
682 * callback, and the inode is then released if the socket is bound to
683 * an inode not a file.
684 */
Cong Wang6d8c50d2018-06-07 13:39:49 -0700685void sock_release(struct socket *sock)
686{
687 __sock_release(sock, NULL);
688}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700689EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400691void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000692{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200693 u8 flags = *tx_flags;
694
Gerhard Engleder51eb7492022-05-06 22:01:38 +0200695 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
Eric Dumazet140c55d2014-08-06 11:49:29 +0200696 flags |= SKBTX_HW_TSTAMP;
697
Gerhard Engleder51eb7492022-05-06 22:01:38 +0200698 /* PTP hardware clocks can provide a free running cycle counter
699 * as a time base for virtual clocks. Tell driver to use the
700 * free running cycle counter for timestamp if socket is bound
701 * to virtual clock.
702 */
703 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
704 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
705 }
706
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400707 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200708 flags |= SKBTX_SW_TSTAMP;
709
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400710 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200711 flags |= SKBTX_SCHED_TSTAMP;
712
Eric Dumazet140c55d2014-08-06 11:49:29 +0200713 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000714}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400715EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000716
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200717INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
718 size_t));
Paolo Abenia648a592019-07-03 16:06:54 +0200719INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
720 size_t));
Yunhui Cui6e6eda42023-01-11 14:59:30 +0800721
722static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
723 int flags)
724{
725 trace_sock_send_length(sk, ret, 0);
726}
727
Al Virod8725c82014-12-11 00:02:50 -0500728static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729{
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000730 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
Paolo Abenia648a592019-07-03 16:06:54 +0200731 inet_sendmsg, sock, msg,
732 msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500733 BUG_ON(ret == -EIOCBQUEUED);
Yunhui Cui6e6eda42023-01-11 14:59:30 +0800734
735 if (trace_sock_send_length_enabled())
736 call_trace_sock_send_length(sock->sk, ret, 0);
Al Virod8725c82014-12-11 00:02:50 -0500737 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800739
Jordan Rife86a7e0b2023-09-21 18:46:41 -0500740static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
741{
742 int err = security_socket_sendmsg(sock, msg,
743 msg_data_left(msg));
744
745 return err ?: sock_sendmsg_nosec(sock, msg);
746}
747
Randy Dunlap85806af2019-05-18 21:23:07 -0700748/**
749 * sock_sendmsg - send a message through @sock
750 * @sock: socket
751 * @msg: message to send
752 *
753 * Sends @msg through @sock, passing through LSM.
754 * Returns the number of bytes sent, or an error code.
755 */
Al Virod8725c82014-12-11 00:02:50 -0500756int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800757{
Jordan Rife86a7e0b2023-09-21 18:46:41 -0500758 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
759 struct sockaddr_storage address;
Marc Dionne01b28852023-12-21 09:12:30 -0400760 int save_len = msg->msg_namelen;
Jordan Rife86a7e0b2023-09-21 18:46:41 -0500761 int ret;
Ying Xue1b784142015-03-02 15:37:48 +0800762
Jordan Rife86a7e0b2023-09-21 18:46:41 -0500763 if (msg->msg_name) {
764 memcpy(&address, msg->msg_name, msg->msg_namelen);
765 msg->msg_name = &address;
766 }
767
768 ret = __sock_sendmsg(sock, msg);
769 msg->msg_name = save_addr;
Marc Dionne01b28852023-12-21 09:12:30 -0400770 msg->msg_namelen = save_len;
Jordan Rife86a7e0b2023-09-21 18:46:41 -0500771
772 return ret;
Gu Zheng0cf00c62014-12-05 15:14:23 +0800773}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700774EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300776/**
777 * kernel_sendmsg - send a message through @sock (kernel-space)
778 * @sock: socket
779 * @msg: message header
780 * @vec: kernel vec
781 * @num: vec array length
782 * @size: total message data size
783 *
784 * Builds the message data with @vec and sends it through @sock.
785 * Returns the number of bytes sent, or an error code.
786 */
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
789 struct kvec *vec, size_t num, size_t size)
790{
Al Virode4eda92022-09-15 20:25:47 -0400791 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500792 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700794EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300796/**
797 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
798 * @sk: sock
799 * @msg: message header
800 * @vec: output s/g array
801 * @num: output s/g array length
802 * @size: total message data size
803 *
804 * Builds the message data with @vec and sends it through @sock.
805 * Returns the number of bytes sent, or an error code.
806 * Caller must hold @sk.
807 */
808
Tom Herbert306b13e2017-07-28 16:22:41 -0700809int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
810 struct kvec *vec, size_t num, size_t size)
811{
812 struct socket *sock = sk->sk_socket;
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000813 const struct proto_ops *ops = READ_ONCE(sock->ops);
Tom Herbert306b13e2017-07-28 16:22:41 -0700814
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000815 if (!ops->sendmsg_locked)
John Fastabenddb5980d2017-08-15 22:31:34 -0700816 return sock_no_sendmsg_locked(sk, msg, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700817
Al Virode4eda92022-09-15 20:25:47 -0400818 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700819
Eric Dumazet1ded5e52023-08-08 13:58:09 +0000820 return ops->sendmsg_locked(sk, msg, msg_data_left(msg));
Tom Herbert306b13e2017-07-28 16:22:41 -0700821}
822EXPORT_SYMBOL(kernel_sendmsg_locked);
823
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400824static bool skb_is_err_queue(const struct sk_buff *skb)
825{
826 /* pkt_type of skbs enqueued on the error queue are set to
827 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
828 * in recvmsg, since skbs received on a local socket will never
829 * have a pkt_type of PACKET_OUTGOING.
830 */
831 return skb->pkt_type == PACKET_OUTGOING;
832}
833
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200834/* On transmit, software and hardware timestamps are returned independently.
835 * As the two skb clones share the hardware timestamp, which may be updated
836 * before the software timestamp is received, a hardware TX timestamp may be
837 * returned only if there is no software TX timestamp. Ignore false software
838 * timestamps, which may be made in the __sock_recv_timestamp() call when the
Deepa Dinamani7f1bc6e2019-02-02 07:34:46 -0800839 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200840 * hardware timestamp.
841 */
842static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
843{
844 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
845}
846
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200847static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
848{
Eric Dumazete3390b32023-08-31 13:52:11 +0000849 bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200850 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
851 struct net_device *orig_dev;
852 ktime_t hwtstamp;
853
854 rcu_read_lock();
855 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
856 if (orig_dev) {
857 *if_index = orig_dev->ifindex;
858 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
859 } else {
860 hwtstamp = shhwtstamps->hwtstamp;
861 }
862 rcu_read_unlock();
863
864 return hwtstamp;
865}
866
867static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
868 int if_index)
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200869{
870 struct scm_ts_pktinfo ts_pktinfo;
871 struct net_device *orig_dev;
872
873 if (!skb_mac_header_was_set(skb))
874 return;
875
876 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
877
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200878 if (!if_index) {
879 rcu_read_lock();
880 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
881 if (orig_dev)
882 if_index = orig_dev->ifindex;
883 rcu_read_unlock();
884 }
885 ts_pktinfo.if_index = if_index;
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200886
887 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
888 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
889 sizeof(ts_pktinfo), &ts_pktinfo);
890}
891
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700892/*
893 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
894 */
895void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
896 struct sk_buff *skb)
897{
Patrick Ohly20d49472009-02-12 05:03:38 +0000898 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Deepa Dinamani887feae2019-02-02 07:34:50 -0800899 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
Deepa Dinamani97184752019-02-02 07:34:51 -0800900 struct scm_timestamping_internal tss;
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200901 int empty = 1, false_tstamp = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000902 struct skb_shared_hwtstamps *shhwtstamps =
903 skb_hwtstamps(skb);
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200904 int if_index;
Miroslav Lichvar007747a92022-01-05 11:33:26 +0100905 ktime_t hwtstamp;
Eric Dumazete3390b32023-08-31 13:52:11 +0000906 u32 tsflags;
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700907
Patrick Ohly20d49472009-02-12 05:03:38 +0000908 /* Race occurred between timestamp enabling and packet
909 receiving. Fill in the current time for now. */
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200910 if (need_software_tstamp && skb->tstamp == 0) {
Patrick Ohly20d49472009-02-12 05:03:38 +0000911 __net_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200912 false_tstamp = 1;
913 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000914
915 if (need_software_tstamp) {
916 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800917 if (new_tstamp) {
918 struct __kernel_sock_timeval tv;
919
920 skb_get_new_timestamp(skb, &tv);
921 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
922 sizeof(tv), &tv);
923 } else {
924 struct __kernel_old_timeval tv;
925
926 skb_get_timestamp(skb, &tv);
927 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
928 sizeof(tv), &tv);
929 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000930 } else {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800931 if (new_tstamp) {
932 struct __kernel_timespec ts;
933
934 skb_get_new_timestampns(skb, &ts);
935 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
936 sizeof(ts), &ts);
937 } else {
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +0200938 struct __kernel_old_timespec ts;
Deepa Dinamani887feae2019-02-02 07:34:50 -0800939
940 skb_get_timestampns(skb, &ts);
941 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
942 sizeof(ts), &ts);
943 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000944 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700945 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000946
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400947 memset(&tss, 0, sizeof(tss));
Eric Dumazete3390b32023-08-31 13:52:11 +0000948 tsflags = READ_ONCE(sk->sk_tsflags);
949 if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Deepa Dinamani97184752019-02-02 07:34:51 -0800950 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000951 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400952 if (shhwtstamps &&
Eric Dumazete3390b32023-08-31 13:52:11 +0000953 (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Yangbo Lud7c08822021-06-30 16:12:00 +0800954 !skb_is_swtx_tstamp(skb, false_tstamp)) {
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200955 if_index = 0;
956 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
957 hwtstamp = get_timestamp(sk, skb, &if_index);
Miroslav Lichvar007747a92022-01-05 11:33:26 +0100958 else
959 hwtstamp = shhwtstamps->hwtstamp;
Yangbo Lud7c08822021-06-30 16:12:00 +0800960
Eric Dumazete3390b32023-08-31 13:52:11 +0000961 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200962 hwtstamp = ptp_convert_timestamp(&hwtstamp,
Eric Dumazet251cd402023-08-31 13:52:12 +0000963 READ_ONCE(sk->sk_bind_phc));
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200964
Miroslav Lichvar007747a92022-01-05 11:33:26 +0100965 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
Yangbo Lud7c08822021-06-30 16:12:00 +0800966 empty = 0;
967
Eric Dumazete3390b32023-08-31 13:52:11 +0000968 if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
Yangbo Lud7c08822021-06-30 16:12:00 +0800969 !skb_is_err_queue(skb))
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200970 put_ts_pktinfo(msg, skb, if_index);
Yangbo Lud7c08822021-06-30 16:12:00 +0800971 }
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200972 }
Francis Yan1c885802016-11-27 23:07:18 -0800973 if (!empty) {
Deepa Dinamani97184752019-02-02 07:34:51 -0800974 if (sock_flag(sk, SOCK_TSTAMP_NEW))
975 put_cmsg_scm_timestamping64(msg, &tss);
976 else
977 put_cmsg_scm_timestamping(msg, &tss);
Francis Yan1c885802016-11-27 23:07:18 -0800978
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400979 if (skb_is_err_queue(skb) && skb->len &&
Soheil Hassas Yeganeh4ef1b282017-03-18 17:03:00 -0400980 SKB_EXT_ERR(skb)->opt_stats)
Francis Yan1c885802016-11-27 23:07:18 -0800981 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
982 skb->len, skb->data);
983 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700984}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300985EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
986
Jakub Kicinskieb6fba72023-04-17 08:53:46 -0700987#ifdef CONFIG_WIRELESS
Johannes Berg6e3e9392011-11-09 10:15:42 +0100988void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
989 struct sk_buff *skb)
990{
991 int ack;
992
993 if (!sock_flag(sk, SOCK_WIFI_STATUS))
994 return;
995 if (!skb->wifi_acked_valid)
996 return;
997
998 ack = skb->wifi_acked;
999
1000 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1001}
1002EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
Jakub Kicinskieb6fba72023-04-17 08:53:46 -07001003#endif
Johannes Berg6e3e9392011-11-09 10:15:42 +01001004
stephen hemminger11165f12010-10-18 14:27:29 +00001005static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1006 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -07001007{
Eyal Birger744d5a32015-03-01 14:58:31 +02001008 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -07001009 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +02001010 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -07001011}
1012
Erin MacNeil6fd1d512022-04-27 16:02:37 -04001013static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1014 struct sk_buff *skb)
1015{
Eric Dumazet2558b802023-02-13 16:00:59 +00001016 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1017 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1018 __u32 mark = skb->mark;
1019
1020 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1021 }
Erin MacNeil6fd1d512022-04-27 16:02:37 -04001022}
1023
1024void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1025 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -07001026{
1027 sock_recv_timestamp(msg, sk, skb);
1028 sock_recv_drops(msg, sk, skb);
Erin MacNeil6fd1d512022-04-27 16:02:37 -04001029 sock_recv_mark(msg, sk, skb);
Neil Horman3b885782009-10-12 13:26:31 -07001030}
Erin MacNeil6fd1d512022-04-27 16:02:37 -04001031EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
Neil Horman3b885782009-10-12 13:26:31 -07001032
Paolo Abeni8c3c4472019-05-03 17:01:39 +02001033INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
Paolo Abenia648a592019-07-03 16:06:54 +02001034 size_t, int));
1035INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1036 size_t, int));
Yunhui Cui6e6eda42023-01-11 14:59:30 +08001037
1038static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1039{
1040 trace_sock_recv_length(sk, ret, flags);
1041}
1042
Ying Xue1b784142015-03-02 15:37:48 +08001043static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -04001044 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001046 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1047 inet6_recvmsg,
Yunhui Cui6e6eda42023-01-11 14:59:30 +08001048 inet_recvmsg, sock, msg,
1049 msg_data_left(msg), flags);
1050 if (trace_sock_recv_length_enabled())
1051 call_trace_sock_recv_length(sock->sk, ret, flags);
1052 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053}
1054
Randy Dunlap85806af2019-05-18 21:23:07 -07001055/**
1056 * sock_recvmsg - receive a message from @sock
1057 * @sock: socket
1058 * @msg: message to receive
1059 * @flags: message flags
1060 *
1061 * Receives @msg from @sock, passing through LSM. Returns the total number
1062 * of bytes received, or an error.
1063 */
Al Viro2da62902015-03-14 21:13:46 -04001064int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001065{
Al Viro2da62902015-03-14 21:13:46 -04001066 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001067
Al Viro2da62902015-03-14 21:13:46 -04001068 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001070EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Martin Lucinac1249c02010-12-10 00:04:05 +00001072/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001073 * kernel_recvmsg - Receive a message from a socket (kernel space)
1074 * @sock: The socket to receive the message from
1075 * @msg: Received message
1076 * @vec: Input s/g array for message data
1077 * @num: Size of input s/g array
1078 * @size: Number of bytes to read
1079 * @flags: Message flags (MSG_DONTWAIT, etc...)
Martin Lucinac1249c02010-12-10 00:04:05 +00001080 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001081 * On return the msg structure contains the scatter/gather array passed in the
1082 * vec argument. The array is modified so that it consists of the unfilled
1083 * portion of the original array.
Martin Lucinac1249c02010-12-10 00:04:05 +00001084 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001085 * The returned value is the total number of bytes received, or an error.
Martin Lucinac1249c02010-12-10 00:04:05 +00001086 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001087
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001088int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1089 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090{
Christoph Hellwig1f466e12020-05-11 13:59:13 +02001091 msg->msg_control_is_user = false;
Al Virode4eda92022-09-15 20:25:47 -04001092 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
Christoph Hellwig1f466e12020-05-11 13:59:13 +02001093 return sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001095EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096
Jens Axboe9c55e012007-11-06 23:30:13 -08001097static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001098 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -08001099 unsigned int flags)
1100{
1101 struct socket *sock = file->private_data;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001102 const struct proto_ops *ops;
Jens Axboe9c55e012007-11-06 23:30:13 -08001103
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001104 ops = READ_ONCE(sock->ops);
1105 if (unlikely(!ops->splice_read))
David Howells67178fd2023-05-22 14:50:00 +01001106 return copy_splice_read(file, ppos, pipe, len, flags);
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001107
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001108 return ops->splice_read(sock, ppos, pipe, len, flags);
Jens Axboe9c55e012007-11-06 23:30:13 -08001109}
1110
David Howells2bfc6682023-06-07 19:19:10 +01001111static void sock_splice_eof(struct file *file)
1112{
1113 struct socket *sock = file->private_data;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001114 const struct proto_ops *ops;
David Howells2bfc6682023-06-07 19:19:10 +01001115
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001116 ops = READ_ONCE(sock->ops);
1117 if (ops->splice_eof)
1118 ops->splice_eof(sock);
David Howells2bfc6682023-06-07 19:19:10 +01001119}
1120
Al Viro8ae5e0302014-11-28 19:40:50 -05001121static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001122{
Al Viro6d652332015-01-30 16:12:56 -05001123 struct file *file = iocb->ki_filp;
1124 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001125 struct msghdr msg = {.msg_iter = *to,
1126 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001127 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001128
Jens Axboeebfcd892019-12-09 20:58:56 -07001129 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001130 msg.msg_flags = MSG_DONTWAIT;
1131
1132 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001133 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001134
Christoph Hellwig66ee59a2015-02-11 19:56:46 +01001135 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001136 return 0;
1137
Al Viro2da62902015-03-14 21:13:46 -04001138 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e0302014-11-28 19:40:50 -05001139 *to = msg.msg_iter;
1140 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001141}
1142
Al Viro8ae5e0302014-11-28 19:40:50 -05001143static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144{
Al Viro6d652332015-01-30 16:12:56 -05001145 struct file *file = iocb->ki_filp;
1146 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001147 struct msghdr msg = {.msg_iter = *from,
1148 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001149 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001150
Al Viro8ae5e0302014-11-28 19:40:50 -05001151 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001152 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001153
Jens Axboeebfcd892019-12-09 20:58:56 -07001154 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001155 msg.msg_flags = MSG_DONTWAIT;
1156
Al Viro6d652332015-01-30 16:12:56 -05001157 if (sock->type == SOCK_SEQPACKET)
1158 msg.msg_flags |= MSG_EOR;
1159
Jordan Rife86a7e0b2023-09-21 18:46:41 -05001160 res = __sock_sendmsg(sock, &msg);
Al Viro8ae5e0302014-11-28 19:40:50 -05001161 *from = msg.msg_iter;
1162 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163}
1164
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165/*
1166 * Atomic setting of ioctl hooks to avoid race
1167 * with module unload.
1168 */
1169
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001170static DEFINE_MUTEX(br_ioctl_mutex);
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001171static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1172 unsigned int cmd, struct ifreq *ifr,
1173 void __user *uarg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001175void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1176 unsigned int cmd, struct ifreq *ifr,
1177 void __user *uarg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001179 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001181 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182}
1183EXPORT_SYMBOL(brioctl_set);
1184
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001185int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1186 struct ifreq *ifr, void __user *uarg)
1187{
1188 int err = -ENOPKG;
1189
1190 if (!br_ioctl_hook)
1191 request_module("bridge");
1192
1193 mutex_lock(&br_ioctl_mutex);
1194 if (br_ioctl_hook)
1195 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1196 mutex_unlock(&br_ioctl_mutex);
1197
1198 return err;
1199}
1200
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001201static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001202static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001203
Eric W. Biederman881d9662007-09-17 11:56:21 -07001204void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001206 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001208 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209}
1210EXPORT_SYMBOL(vlan_ioctl_set);
1211
Arnd Bergmann6b960182009-11-06 23:10:54 -08001212static long sock_do_ioctl(struct net *net, struct socket *sock,
Johannes Berg63ff03a2019-01-25 22:43:17 +01001213 unsigned int cmd, unsigned long arg)
Arnd Bergmann6b960182009-11-06 23:10:54 -08001214{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001215 const struct proto_ops *ops = READ_ONCE(sock->ops);
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001216 struct ifreq ifr;
1217 bool need_copyout;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001218 int err;
1219 void __user *argp = (void __user *)arg;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001220 void __user *data;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001221
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001222 err = ops->ioctl(sock, cmd, arg);
Arnd Bergmann6b960182009-11-06 23:10:54 -08001223
1224 /*
1225 * If this ioctl is unknown try to hand it down
1226 * to the NIC driver.
1227 */
Al Viro36fd6332017-06-26 13:19:16 -04001228 if (err != -ENOIOCTLCMD)
1229 return err;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001230
Jakub Kicinski29ce8f92021-08-31 09:06:04 -07001231 if (!is_socket_ioctl_cmd(cmd))
1232 return -ENOTTY;
1233
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001234 if (get_user_ifreq(&ifr, &data, argp))
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001235 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001236 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001237 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001238 if (put_user_ifreq(&ifr, argp))
Al Viro36fd6332017-06-26 13:19:16 -04001239 return -EFAULT;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001240
Arnd Bergmann6b960182009-11-06 23:10:54 -08001241 return err;
1242}
1243
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244/*
1245 * With an ioctl, arg may well be a user mode pointer, but we don't know
1246 * what to do with it - that's up to the protocol still.
1247 */
1248
1249static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1250{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001251 const struct proto_ops *ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001253 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 void __user *argp = (void __user *)arg;
1255 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001256 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257
Eric Dumazetb69aee02005-09-06 14:42:45 -07001258 sock = file->private_data;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001259 ops = READ_ONCE(sock->ops);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001260 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001261 net = sock_net(sk);
Al Viro44c02a22017-10-05 12:59:44 -04001262 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1263 struct ifreq ifr;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001264 void __user *data;
Al Viro44c02a22017-10-05 12:59:44 -04001265 bool need_copyout;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001266 if (get_user_ifreq(&ifr, &data, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001267 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001268 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Al Viro44c02a22017-10-05 12:59:44 -04001269 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001270 if (put_user_ifreq(&ifr, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001271 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001273#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Al Virob1b0c242017-10-01 20:13:08 -04001275 err = wext_handle_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001277#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001278 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 case FIOSETOWN:
1280 case SIOCSPGRP:
1281 err = -EFAULT;
1282 if (get_user(pid, (int __user *)argp))
1283 break;
Jiri Slaby393cc3f2017-06-13 13:35:50 +02001284 err = f_setown(sock->file, pid, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 break;
1286 case FIOGETOWN:
1287 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001288 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001289 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 break;
1291 case SIOCGIFBR:
1292 case SIOCSIFBR:
1293 case SIOCBRADDBR:
1294 case SIOCBRDELBR:
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001295 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 break;
1297 case SIOCGIFVLAN:
1298 case SIOCSIFVLAN:
1299 err = -ENOPKG;
1300 if (!vlan_ioctl_hook)
1301 request_module("8021q");
1302
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001303 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001305 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001306 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 break;
Andrey Vaginc62cce22016-10-24 18:29:13 -07001308 case SIOCGSKNS:
1309 err = -EPERM;
1310 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1311 break;
1312
1313 err = open_related_ns(&net->ns, get_net_ns);
1314 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001315 case SIOCGSTAMP_OLD:
1316 case SIOCGSTAMPNS_OLD:
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001317 if (!ops->gettstamp) {
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001318 err = -ENOIOCTLCMD;
1319 break;
1320 }
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001321 err = ops->gettstamp(sock, argp,
1322 cmd == SIOCGSTAMP_OLD,
1323 !IS_ENABLED(CONFIG_64BIT));
Gustavo A. R. Silva60747822019-04-24 10:31:24 -05001324 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001325 case SIOCGSTAMP_NEW:
1326 case SIOCGSTAMPNS_NEW:
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001327 if (!ops->gettstamp) {
Arnd Bergmann0768e172019-04-17 22:56:11 +02001328 err = -ENOIOCTLCMD;
1329 break;
1330 }
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001331 err = ops->gettstamp(sock, argp,
1332 cmd == SIOCGSTAMP_NEW,
1333 false);
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001334 break;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001335
1336 case SIOCGIFCONF:
1337 err = dev_ifconf(net, argp);
1338 break;
1339
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 default:
Johannes Berg63ff03a2019-01-25 22:43:17 +01001341 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001343 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 return err;
1345}
1346
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001347/**
1348 * sock_create_lite - creates a socket
1349 * @family: protocol family (AF_INET, ...)
1350 * @type: communication type (SOCK_STREAM, ...)
1351 * @protocol: protocol (0, ...)
1352 * @res: new socket
1353 *
1354 * Creates a new socket and assigns it to @res, passing through LSM.
1355 * The new socket initialization is not complete, see kernel_accept().
1356 * Returns 0 or an error. On failure @res is set to %NULL.
1357 * This function internally uses GFP_KERNEL.
1358 */
1359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360int sock_create_lite(int family, int type, int protocol, struct socket **res)
1361{
1362 int err;
1363 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 err = security_socket_create(family, type, protocol, 1);
1366 if (err)
1367 goto out;
1368
1369 sock = sock_alloc();
1370 if (!sock) {
1371 err = -ENOMEM;
1372 goto out;
1373 }
1374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001376 err = security_socket_post_create(sock, family, type, protocol, 1);
1377 if (err)
1378 goto out_release;
1379
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380out:
1381 *res = sock;
1382 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001383out_release:
1384 sock_release(sock);
1385 sock = NULL;
1386 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001388EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389
1390/* No kernel lock held - perfect */
Al Viroade994f2017-07-03 00:01:49 -04001391static __poll_t sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392{
Christoph Hellwig3cafb372018-01-09 16:07:34 +01001393 struct socket *sock = file->private_data;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001394 const struct proto_ops *ops = READ_ONCE(sock->ops);
Christoph Hellwiga331de32018-07-30 09:42:13 +02001395 __poll_t events = poll_requested_events(wait), flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001397 if (!ops->poll)
Christoph Hellwige88958e2018-06-29 15:37:24 +02001398 return 0;
Christoph Hellwigf641f13b2018-07-30 09:42:12 +02001399
Christoph Hellwiga331de32018-07-30 09:42:13 +02001400 if (sk_can_busy_loop(sock->sk)) {
1401 /* poll once if requested by the syscall */
1402 if (events & POLL_BUSY_LOOP)
1403 sk_busy_loop(sock->sk, 1);
1404
1405 /* if this socket can poll_ll, tell the system call */
1406 flag = POLL_BUSY_LOOP;
1407 }
1408
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001409 return ops->poll(file, sock, wait) | flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410}
1411
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001412static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001414 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001416 return READ_ONCE(sock->ops)->mmap(file, sock, vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417}
1418
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001419static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420{
Cong Wang6d8c50d2018-06-07 13:39:49 -07001421 __sock_release(SOCKET_I(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 return 0;
1423}
1424
1425/*
1426 * Update the socket async list
1427 *
1428 * Fasync_list locking strategy.
1429 *
1430 * 1. fasync_list is modified only under process context socket lock
1431 * i.e. under semaphore.
1432 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001433 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 */
1435
1436static int sock_fasync(int fd, struct file *filp, int on)
1437{
Eric Dumazet989a2972010-04-14 09:55:35 +00001438 struct socket *sock = filp->private_data;
1439 struct sock *sk = sock->sk;
Al Viro333f7902019-07-05 20:14:16 +01001440 struct socket_wq *wq = &sock->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
Eric Dumazet989a2972010-04-14 09:55:35 +00001442 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444
1445 lock_sock(sk);
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001446 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001448 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001449 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001450 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001451 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452
Eric Dumazet989a2972010-04-14 09:55:35 +00001453 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 return 0;
1455}
1456
Eric Dumazetceb5d582015-11-29 20:03:11 -08001457/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458
Eric Dumazetceb5d582015-11-29 20:03:11 -08001459int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001461 if (!wq || !wq->fasync_list)
1462 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001463
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001464 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001465 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001466 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467 break;
1468 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001469 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001470 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 break;
Miaohe Lin7c7ab582020-08-08 16:23:10 +08001472 fallthrough;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001473 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001474call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001475 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001477 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001478 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001480
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 return 0;
1482}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001483EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001485/**
1486 * __sock_create - creates a socket
1487 * @net: net namespace
1488 * @family: protocol family (AF_INET, ...)
1489 * @type: communication type (SOCK_STREAM, ...)
1490 * @protocol: protocol (0, ...)
1491 * @res: new socket
1492 * @kern: boolean for kernel space sockets
1493 *
1494 * Creates a new socket and assigns it to @res, passing through LSM.
1495 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1496 * be set to true if the socket resides in kernel space.
1497 * This function internally uses GFP_KERNEL.
1498 */
1499
Pavel Emelyanov721db932010-09-29 16:06:32 +04001500int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001501 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502{
1503 int err;
1504 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001505 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506
1507 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001508 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 */
1510 if (family < 0 || family >= NPROTO)
1511 return -EAFNOSUPPORT;
1512 if (type < 0 || type >= SOCK_MAX)
1513 return -EINVAL;
1514
1515 /* Compatibility.
1516
1517 This uglymoron is moved from INET layer to here to avoid
1518 deadlock in module load.
1519 */
1520 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001521 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1522 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 family = PF_PACKET;
1524 }
1525
1526 err = security_socket_create(family, type, protocol, kern);
1527 if (err)
1528 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001529
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001530 /*
1531 * Allocate the socket and allow the family to set things up. if
1532 * the protocol is 0, the family is instructed to select an appropriate
1533 * default.
1534 */
1535 sock = sock_alloc();
1536 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001537 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001538 return -ENFILE; /* Not exactly a match, but its the
1539 closest posix thing */
1540 }
1541
1542 sock->type = type;
1543
Johannes Berg95a5afc2008-10-16 15:24:51 -07001544#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001545 /* Attempt to load a protocol module if the find failed.
1546 *
1547 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 * requested real, full-featured networking support upon configuration.
1549 * Otherwise module support will break!
1550 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001551 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001552 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553#endif
1554
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001555 rcu_read_lock();
1556 pf = rcu_dereference(net_families[family]);
1557 err = -EAFNOSUPPORT;
1558 if (!pf)
1559 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
1561 /*
1562 * We will call the ->create function, that possibly is in a loadable
1563 * module, so we have to bump that loadable module refcnt first.
1564 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001565 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 goto out_release;
1567
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001568 /* Now protected by module ref count */
1569 rcu_read_unlock();
1570
Eric Paris3f378b682009-11-05 22:18:14 -08001571 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001572 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001574
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 /*
1576 * Now to bump the refcnt of the [loadable] module that owns this
1577 * socket at sock_release time we decrement its refcnt.
1578 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001579 if (!try_module_get(sock->ops->owner))
1580 goto out_module_busy;
1581
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 /*
1583 * Now that we're done with the ->create function, the [loadable]
1584 * module can have its refcnt decremented
1585 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001586 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001587 err = security_socket_post_create(sock, family, type, protocol, kern);
1588 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001589 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001590 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001592 return 0;
1593
1594out_module_busy:
1595 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001597 sock->ops = NULL;
1598 module_put(pf->owner);
1599out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001601 return err;
1602
1603out_release:
1604 rcu_read_unlock();
1605 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001607EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001609/**
1610 * sock_create - creates a socket
1611 * @family: protocol family (AF_INET, ...)
1612 * @type: communication type (SOCK_STREAM, ...)
1613 * @protocol: protocol (0, ...)
1614 * @res: new socket
1615 *
1616 * A wrapper around __sock_create().
1617 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1618 */
1619
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620int sock_create(int family, int type, int protocol, struct socket **res)
1621{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001622 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001624EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001626/**
1627 * sock_create_kern - creates a socket (kernel space)
1628 * @net: net namespace
1629 * @family: protocol family (AF_INET, ...)
1630 * @type: communication type (SOCK_STREAM, ...)
1631 * @protocol: protocol (0, ...)
1632 * @res: new socket
1633 *
1634 * A wrapper around __sock_create().
1635 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1636 */
1637
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001638int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001640 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001642EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643
Jens Axboeda214a42022-04-12 14:22:39 -06001644static struct socket *__sys_socket_create(int family, int type, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 struct socket *sock;
Jens Axboeda214a42022-04-12 14:22:39 -06001647 int retval;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001648
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001649 /* Check the SOCK_* constants for consistency. */
1650 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1651 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1652 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1653 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1654
Jens Axboeda214a42022-04-12 14:22:39 -06001655 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1656 return ERR_PTR(-EINVAL);
Ulrich Dreppera677a032008-07-23 21:29:17 -07001657 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658
1659 retval = sock_create(family, type, protocol, &sock);
1660 if (retval < 0)
Jens Axboeda214a42022-04-12 14:22:39 -06001661 return ERR_PTR(retval);
1662
1663 return sock;
1664}
1665
1666struct file *__sys_socket_file(int family, int type, int protocol)
1667{
1668 struct socket *sock;
Jens Axboeda214a42022-04-12 14:22:39 -06001669 int flags;
1670
1671 sock = __sys_socket_create(family, type, protocol);
1672 if (IS_ERR(sock))
1673 return ERR_CAST(sock);
1674
1675 flags = type & ~SOCK_TYPE_MASK;
1676 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1677 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1678
Thadeu Lima de Souza Cascardo649c15c2023-03-07 14:37:07 -03001679 return sock_alloc_file(sock, flags, NULL);
Jens Axboeda214a42022-04-12 14:22:39 -06001680}
1681
Geliang Tang0dd061a2023-08-16 09:11:56 +08001682/* A hook for bpf progs to attach to and update socket protocol.
1683 *
1684 * A static noinline declaration here could cause the compiler to
1685 * optimize away the function. A global noinline declaration will
1686 * keep the definition, but may optimize away the callsite.
1687 * Therefore, __weak is needed to ensure that the call is still
1688 * emitted, by telling the compiler that we don't know what the
1689 * function might eventually be.
Geliang Tang0dd061a2023-08-16 09:11:56 +08001690 */
1691
Dave Marchevsky15fb6f2b2023-10-31 14:56:25 -07001692__bpf_hook_start();
Geliang Tang0dd061a2023-08-16 09:11:56 +08001693
1694__weak noinline int update_socket_protocol(int family, int type, int protocol)
1695{
1696 return protocol;
1697}
1698
Dave Marchevsky15fb6f2b2023-10-31 14:56:25 -07001699__bpf_hook_end();
Geliang Tang0dd061a2023-08-16 09:11:56 +08001700
Jens Axboeda214a42022-04-12 14:22:39 -06001701int __sys_socket(int family, int type, int protocol)
1702{
1703 struct socket *sock;
1704 int flags;
1705
Geliang Tang0dd061a2023-08-16 09:11:56 +08001706 sock = __sys_socket_create(family, type,
1707 update_socket_protocol(family, type, protocol));
Jens Axboeda214a42022-04-12 14:22:39 -06001708 if (IS_ERR(sock))
1709 return PTR_ERR(sock);
1710
1711 flags = type & ~SOCK_TYPE_MASK;
1712 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1713 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714
Al Viro8e1611e2017-12-05 23:29:09 +00001715 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716}
1717
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001718SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1719{
1720 return __sys_socket(family, type, protocol);
1721}
1722
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723/*
1724 * Create a pair of connected sockets.
1725 */
1726
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001727int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728{
1729 struct socket *sock1, *sock2;
1730 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001731 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001732 int flags;
1733
1734 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001735 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001736 return -EINVAL;
1737 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001739 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1740 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1741
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 /*
Al Viro016a2662017-12-05 23:28:38 +00001743 * reserve descriptors and make sure we won't fail
1744 * to return them to userland.
1745 */
1746 fd1 = get_unused_fd_flags(flags);
1747 if (unlikely(fd1 < 0))
1748 return fd1;
1749
1750 fd2 = get_unused_fd_flags(flags);
1751 if (unlikely(fd2 < 0)) {
1752 put_unused_fd(fd1);
1753 return fd2;
1754 }
1755
1756 err = put_user(fd1, &usockvec[0]);
1757 if (err)
1758 goto out;
1759
1760 err = put_user(fd2, &usockvec[1]);
1761 if (err)
1762 goto out;
1763
1764 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 * Obtain the first socket and check if the underlying protocol
1766 * supports the socketpair call.
1767 */
1768
1769 err = sock_create(family, type, protocol, &sock1);
Al Viro016a2662017-12-05 23:28:38 +00001770 if (unlikely(err < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771 goto out;
1772
1773 err = sock_create(family, type, protocol, &sock2);
Al Viro016a2662017-12-05 23:28:38 +00001774 if (unlikely(err < 0)) {
1775 sock_release(sock1);
1776 goto out;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001777 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001778
David Herrmannd47cd942018-05-04 16:28:20 +02001779 err = security_socket_socketpair(sock1, sock2);
1780 if (unlikely(err)) {
1781 sock_release(sock2);
1782 sock_release(sock1);
1783 goto out;
1784 }
1785
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001786 err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
Al Viro016a2662017-12-05 23:28:38 +00001787 if (unlikely(err < 0)) {
1788 sock_release(sock2);
1789 sock_release(sock1);
1790 goto out;
Al Viro28407632012-08-17 23:54:15 -04001791 }
1792
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001793 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301794 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001795 err = PTR_ERR(newfile1);
Al Viro016a2662017-12-05 23:28:38 +00001796 sock_release(sock2);
1797 goto out;
Al Viro28407632012-08-17 23:54:15 -04001798 }
1799
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001800 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001801 if (IS_ERR(newfile2)) {
1802 err = PTR_ERR(newfile2);
Al Viro016a2662017-12-05 23:28:38 +00001803 fput(newfile1);
1804 goto out;
Al Virodb349502007-02-07 01:48:00 -05001805 }
1806
Al Viro157cf642008-12-14 04:57:47 -05001807 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001808
Al Virodb349502007-02-07 01:48:00 -05001809 fd_install(fd1, newfile1);
1810 fd_install(fd2, newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001811 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813out:
Al Viro016a2662017-12-05 23:28:38 +00001814 put_unused_fd(fd2);
1815 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816 return err;
1817}
1818
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001819SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1820 int __user *, usockvec)
1821{
1822 return __sys_socketpair(family, type, protocol, usockvec);
1823}
1824
Gabriel Krisman Bertazidc2e77972024-06-14 12:30:44 -04001825int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1826 int addrlen)
1827{
1828 int err;
1829
1830 err = security_socket_bind(sock, (struct sockaddr *)address,
1831 addrlen);
1832 if (!err)
1833 err = READ_ONCE(sock->ops)->bind(sock,
1834 (struct sockaddr *)address,
1835 addrlen);
1836 return err;
1837}
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839/*
1840 * Bind a name to a socket. Nothing much to do here since it's
1841 * the protocol's responsibility to handle the local address.
1842 *
1843 * We move the socket address to kernel space before we call
1844 * the protocol layer (having also checked the address is ok).
1845 */
1846
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001847int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848{
1849 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001850 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001851 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001853 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001854 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001855 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Gabriel Krisman Bertazidc2e77972024-06-14 12:30:44 -04001856 if (!err)
1857 err = __sys_bind_socket(sock, &address, addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001858 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001859 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 return err;
1861}
1862
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001863SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1864{
1865 return __sys_bind(fd, umyaddr, addrlen);
1866}
1867
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868/*
1869 * Perform a listen. Basically, we allow the protocol to do anything
1870 * necessary for a listen, and if that works, we mark the socket as
1871 * ready for listening.
1872 */
Gabriel Krisman Bertazibb6aaf72024-06-14 12:30:45 -04001873int __sys_listen_socket(struct socket *sock, int backlog)
1874{
1875 int somaxconn, err;
1876
1877 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1878 if ((unsigned int)backlog > somaxconn)
1879 backlog = somaxconn;
1880
1881 err = security_socket_listen(sock, backlog);
1882 if (!err)
1883 err = READ_ONCE(sock->ops)->listen(sock, backlog);
1884 return err;
1885}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001887int __sys_listen(int fd, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888{
1889 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001890 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001891
1892 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1893 if (sock) {
Gabriel Krisman Bertazibb6aaf72024-06-14 12:30:45 -04001894 err = __sys_listen_socket(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001895 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 }
1897 return err;
1898}
1899
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001900SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1901{
1902 return __sys_listen(fd, backlog);
1903}
1904
Jens Axboe0645fbe2024-05-09 09:31:05 -06001905struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
Jens Axboede2ea4b2019-10-17 14:41:29 -06001906 struct sockaddr __user *upeer_sockaddr,
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001907 int __user *upeer_addrlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908{
1909 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001910 struct file *newfile;
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001911 int err, len;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001912 struct sockaddr_storage address;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001913 const struct proto_ops *ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914
Florent Revestdba4a922020-12-04 12:36:04 +01001915 sock = sock_from_file(file);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001916 if (!sock)
1917 return ERR_PTR(-ENOTSOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001919 newsock = sock_alloc();
1920 if (!newsock)
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001921 return ERR_PTR(-ENFILE);
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001922 ops = READ_ONCE(sock->ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923
1924 newsock->type = sock->type;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001925 newsock->ops = ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 /*
1928 * We don't need try_module_get here, as the listening socket (sock)
1929 * has the protocol module (sock->ops->owner) held.
1930 */
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001931 __module_get(ops->owner);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001933 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001934 if (IS_ERR(newfile))
1935 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001936
Frank Filza79af592005-09-27 15:23:38 -07001937 err = security_socket_accept(sock, newsock);
1938 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001939 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001940
Jens Axboe0645fbe2024-05-09 09:31:05 -06001941 arg->flags |= sock->file->f_flags;
1942 err = ops->accept(sock, newsock, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001944 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945
1946 if (upeer_sockaddr) {
Eric Dumazet1ded5e52023-08-08 13:58:09 +00001947 len = ops->getname(newsock, (struct sockaddr *)&address, 2);
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001948 if (len < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001950 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001952 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001953 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001955 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956 }
1957
1958 /* File flags are not inherited via accept() unlike another OSes. */
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001959 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001960out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001961 fput(newfile);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001962 return ERR_PTR(err);
1963}
Jens Axboede2ea4b2019-10-17 14:41:29 -06001964
Yajun Dengc0424532022-06-10 17:10:17 +08001965static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1966 int __user *upeer_addrlen, int flags)
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001967{
Jens Axboe0645fbe2024-05-09 09:31:05 -06001968 struct proto_accept_arg arg = { };
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001969 struct file *newfile;
1970 int newfd;
1971
1972 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1973 return -EINVAL;
1974
1975 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1976 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1977
Yajun Dengc0424532022-06-10 17:10:17 +08001978 newfd = get_unused_fd_flags(flags);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001979 if (unlikely(newfd < 0))
1980 return newfd;
1981
Jens Axboe0645fbe2024-05-09 09:31:05 -06001982 newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001983 flags);
1984 if (IS_ERR(newfile)) {
1985 put_unused_fd(newfd);
1986 return PTR_ERR(newfile);
1987 }
1988 fd_install(newfd, newfile);
1989 return newfd;
Jens Axboede2ea4b2019-10-17 14:41:29 -06001990}
1991
1992/*
1993 * For accept, we attempt to create a new socket, set up the link
1994 * with the client, wake up the client, then return the new
1995 * connected fd. We collect the address of the connector in kernel
1996 * space and move it to user at the very end. This is unclean because
1997 * we open the socket then return an error.
1998 *
1999 * 1003.1g adds the ability to recvmsg() to query connection pending
2000 * status to recvmsg. We need to add that support in a way thats
2001 * clean when we restructure accept also.
2002 */
2003
2004int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2005 int __user *upeer_addrlen, int flags)
2006{
2007 int ret = -EBADF;
2008 struct fd f;
2009
2010 f = fdget(fd);
2011 if (f.file) {
Yajun Dengc0424532022-06-10 17:10:17 +08002012 ret = __sys_accept4_file(f.file, upeer_sockaddr,
2013 upeer_addrlen, flags);
Miaohe Lin6b07ede2020-08-06 19:52:24 +08002014 fdput(f);
Jens Axboede2ea4b2019-10-17 14:41:29 -06002015 }
2016
2017 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018}
2019
Dominik Brodowski4541e802018-03-13 19:24:23 +01002020SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2021 int __user *, upeer_addrlen, int, flags)
2022{
2023 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2024}
2025
Heiko Carstens20f37032009-01-14 14:14:23 +01002026SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2027 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002028{
Dominik Brodowski4541e802018-03-13 19:24:23 +01002029 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07002030}
2031
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032/*
2033 * Attempt to connect to a socket with the server address. The address
2034 * is in user space so we verify it is OK and move it to kernel space.
2035 *
2036 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2037 * break bindings
2038 *
2039 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2040 * other SEQPACKET protocols that take time to connect() as it doesn't
2041 * include the -EINPROGRESS status for such sockets.
2042 */
2043
Jens Axboef499a022019-12-02 16:28:46 -07002044int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
Jens Axboebd3ded32019-11-23 14:17:16 -07002045 int addrlen, int file_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046{
2047 struct socket *sock;
Jens Axboebd3ded32019-11-23 14:17:16 -07002048 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049
Florent Revestdba4a922020-12-04 12:36:04 +01002050 sock = sock_from_file(file);
2051 if (!sock) {
2052 err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053 goto out;
Florent Revestdba4a922020-12-04 12:36:04 +01002054 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002056 err =
Jens Axboef499a022019-12-02 16:28:46 -07002057 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 if (err)
Jens Axboebd3ded32019-11-23 14:17:16 -07002059 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002061 err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2062 addrlen, sock->file->f_flags | file_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063out:
2064 return err;
2065}
2066
Jens Axboebd3ded32019-11-23 14:17:16 -07002067int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2068{
2069 int ret = -EBADF;
2070 struct fd f;
2071
2072 f = fdget(fd);
2073 if (f.file) {
Jens Axboef499a022019-12-02 16:28:46 -07002074 struct sockaddr_storage address;
2075
2076 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2077 if (!ret)
2078 ret = __sys_connect_file(f.file, &address, addrlen, 0);
Miaohe Lin6b07ede2020-08-06 19:52:24 +08002079 fdput(f);
Jens Axboebd3ded32019-11-23 14:17:16 -07002080 }
2081
2082 return ret;
2083}
2084
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01002085SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2086 int, addrlen)
2087{
2088 return __sys_connect(fd, uservaddr, addrlen);
2089}
2090
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091/*
2092 * Get the local address ('name') of a socket object. Move the obtained
2093 * name to user space.
2094 */
2095
Dominik Brodowski8882a102018-03-13 19:43:14 +01002096int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2097 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098{
2099 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002100 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002101 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002102
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002103 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 if (!sock)
2105 goto out;
2106
2107 err = security_socket_getsockname(sock);
2108 if (err)
2109 goto out_put;
2110
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002111 err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002112 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113 goto out_put;
Hamish MacDonalde44ef1d2022-01-01 22:38:25 +10002114 /* "err" is actually length in this case */
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002115 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002116
2117out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002118 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119out:
2120 return err;
2121}
2122
Dominik Brodowski8882a102018-03-13 19:43:14 +01002123SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2124 int __user *, usockaddr_len)
2125{
2126 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2127}
2128
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129/*
2130 * Get the remote address ('name') of a socket object. Move the obtained
2131 * name to user space.
2132 */
2133
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002134int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2135 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136{
2137 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002138 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002139 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002141 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2142 if (sock != NULL) {
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002143 const struct proto_ops *ops = READ_ONCE(sock->ops);
2144
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 err = security_socket_getpeername(sock);
2146 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002147 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148 return err;
2149 }
2150
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002151 err = ops->getname(sock, (struct sockaddr *)&address, 1);
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002152 if (err >= 0)
2153 /* "err" is actually length in this case */
2154 err = move_addr_to_user(&address, err, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002155 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002156 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157 }
2158 return err;
2159}
2160
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002161SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2162 int __user *, usockaddr_len)
2163{
2164 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2165}
2166
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167/*
2168 * Send a datagram to a given address. We move the address into kernel
2169 * space and check the user space data area is readable before invoking
2170 * the protocol.
2171 */
Dominik Brodowski211b6342018-03-13 19:18:52 +01002172int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2173 struct sockaddr __user *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174{
2175 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002176 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002177 int err;
2178 struct msghdr msg;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002179 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002180
Jens Axboe9fd78742023-12-04 10:47:50 -07002181 err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
Al Viro602bd0e2015-03-21 19:12:32 -04002182 if (unlikely(err))
2183 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002184 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2185 if (!sock)
David S. Miller4387ff752007-02-08 15:06:08 -08002186 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002187
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002188 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002189 msg.msg_control = NULL;
2190 msg.msg_controllen = 0;
2191 msg.msg_namelen = 0;
Pavel Begunkov7c701d92022-07-12 21:52:29 +01002192 msg.msg_ubuf = NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002193 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002194 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 if (err < 0)
2196 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002197 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002198 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199 }
David Howellsb841b902023-05-22 13:11:10 +01002200 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002201 if (sock->file->f_flags & O_NONBLOCK)
2202 flags |= MSG_DONTWAIT;
2203 msg.msg_flags = flags;
Jordan Rife86a7e0b2023-09-21 18:46:41 -05002204 err = __sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002206out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002207 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08002208out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 return err;
2210}
2211
Dominik Brodowski211b6342018-03-13 19:18:52 +01002212SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2213 unsigned int, flags, struct sockaddr __user *, addr,
2214 int, addr_len)
2215{
2216 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2217}
2218
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002220 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002221 */
2222
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002223SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00002224 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225{
Dominik Brodowski211b6342018-03-13 19:18:52 +01002226 return __sys_sendto(fd, buff, len, flags, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227}
2228
2229/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002230 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 * sender. We verify the buffers are writable and if needed move the
2232 * sender address from kernel to user space.
2233 */
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002234int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2235 struct sockaddr __user *addr, int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236{
Eric Dumazet1228b342022-06-22 15:02:20 +00002237 struct sockaddr_storage address;
2238 struct msghdr msg = {
2239 /* Save some cycles and don't copy the address if not needed */
2240 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2241 };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 struct socket *sock;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002243 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002244 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245
Jens Axboe9fd78742023-12-04 10:47:50 -07002246 err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
Al Viro602bd0e2015-03-21 19:12:32 -04002247 if (unlikely(err))
2248 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002249 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002251 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253 if (sock->file->f_flags & O_NONBLOCK)
2254 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002255 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002257 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002258 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002259 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002260 if (err2 < 0)
2261 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002262 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002263
2264 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08002265out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266 return err;
2267}
2268
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002269SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2270 unsigned int, flags, struct sockaddr __user *, addr,
2271 int __user *, addr_len)
2272{
2273 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2274}
2275
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002277 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278 */
2279
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02002280SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2281 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282{
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002283 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284}
2285
Florian Westphal83f0c102020-07-05 01:30:15 +02002286static bool sock_use_custom_sol_socket(const struct socket *sock)
2287{
Paolo Abenia5ef0582022-10-20 19:48:51 +02002288 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
Florian Westphal83f0c102020-07-05 01:30:15 +02002289}
2290
Breno Leitao14062452023-10-16 06:47:41 -07002291int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2292 int optname, sockptr_t optval, int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002294 const struct proto_ops *ops;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002295 char *kernel_optval = NULL;
Breno Leitao14062452023-10-16 06:47:41 -07002296 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297
2298 if (optlen < 0)
2299 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002300
Christoph Hellwig4a367292020-07-17 08:23:11 +02002301 err = security_socket_setsockopt(sock, level, optname);
2302 if (err)
2303 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002304
Breno Leitao14062452023-10-16 06:47:41 -07002305 if (!compat)
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002306 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
Breno Leitao3f31e0d2023-10-16 06:47:40 -07002307 optval, &optlen,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002308 &kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002309 if (err < 0)
2310 goto out_put;
2311 if (err > 0) {
2312 err = 0;
2313 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314 }
Christoph Hellwig4a367292020-07-17 08:23:11 +02002315
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002316 if (kernel_optval)
2317 optval = KERNEL_SOCKPTR(kernel_optval);
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002318 ops = READ_ONCE(sock->ops);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002319 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002320 err = sock_setsockopt(sock, level, optname, optval, optlen);
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002321 else if (unlikely(!ops->setsockopt))
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002322 err = -EOPNOTSUPP;
Christoph Hellwig4a367292020-07-17 08:23:11 +02002323 else
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002324 err = ops->setsockopt(sock, level, optname, optval,
Christoph Hellwig4a367292020-07-17 08:23:11 +02002325 optlen);
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002326 kfree(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002327out_put:
Breno Leitao14062452023-10-16 06:47:41 -07002328 return err;
2329}
2330EXPORT_SYMBOL(do_sock_setsockopt);
2331
2332/* Set a socket option. Because we don't know the option lengths we have
2333 * to pass the user mode parameter for the protocols to sort out.
2334 */
2335int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2336 int optlen)
2337{
2338 sockptr_t optval = USER_SOCKPTR(user_optval);
2339 bool compat = in_compat_syscall();
2340 int err, fput_needed;
2341 struct socket *sock;
2342
2343 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2344 if (!sock)
2345 return err;
2346
2347 err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2348
Christoph Hellwig4a367292020-07-17 08:23:11 +02002349 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350 return err;
2351}
2352
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002353SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2354 char __user *, optval, int, optlen)
2355{
2356 return __sys_setsockopt(fd, level, optname, optval, optlen);
2357}
2358
Stanislav Fomichev9cacf812021-01-15 08:34:59 -08002359INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2360 int optname));
2361
Breno Leitao0b05b0c2023-10-16 06:47:42 -07002362int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2363 int optname, sockptr_t optval, sockptr_t optlen)
2364{
2365 int max_optlen __maybe_unused;
2366 const struct proto_ops *ops;
2367 int err;
2368
2369 err = security_socket_getsockopt(sock, level, optname);
2370 if (err)
2371 return err;
2372
2373 if (!compat)
2374 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2375
2376 ops = READ_ONCE(sock->ops);
2377 if (level == SOL_SOCKET) {
2378 err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2379 } else if (unlikely(!ops->getsockopt)) {
2380 err = -EOPNOTSUPP;
2381 } else {
2382 if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2383 "Invalid argument type"))
2384 return -EOPNOTSUPP;
2385
2386 err = ops->getsockopt(sock, level, optname, optval.user,
2387 optlen.user);
2388 }
2389
2390 if (!compat)
2391 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2392 optval, optlen, max_optlen,
2393 err);
2394
2395 return err;
2396}
2397EXPORT_SYMBOL(do_sock_getsockopt);
2398
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399/*
2400 * Get a socket option. Because we don't know the option lengths we have
2401 * to pass a user mode parameter for the protocols to sort out.
2402 */
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002403int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2404 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002405{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002406 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407 struct socket *sock;
Breno Leitao0b05b0c2023-10-16 06:47:42 -07002408 bool compat;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002410 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002411 if (!sock)
2412 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002413
Breno Leitao0b05b0c2023-10-16 06:47:42 -07002414 compat = in_compat_syscall();
2415 err = do_sock_getsockopt(sock, compat, level, optname,
2416 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002417
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002418 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419 return err;
2420}
2421
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002422SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2423 char __user *, optval, int __user *, optlen)
2424{
2425 return __sys_getsockopt(fd, level, optname, optval, optlen);
2426}
2427
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428/*
2429 * Shutdown a socket.
2430 */
2431
Jens Axboeb713c192020-09-05 11:13:35 -06002432int __sys_shutdown_sock(struct socket *sock, int how)
2433{
2434 int err;
2435
2436 err = security_socket_shutdown(sock, how);
2437 if (!err)
Eric Dumazet1ded5e52023-08-08 13:58:09 +00002438 err = READ_ONCE(sock->ops)->shutdown(sock, how);
Jens Axboeb713c192020-09-05 11:13:35 -06002439
2440 return err;
2441}
2442
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002443int __sys_shutdown(int fd, int how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002445 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446 struct socket *sock;
2447
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002448 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2449 if (sock != NULL) {
Jens Axboeb713c192020-09-05 11:13:35 -06002450 err = __sys_shutdown_sock(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002451 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 }
2453 return err;
2454}
2455
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002456SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2457{
2458 return __sys_shutdown(fd, how);
2459}
2460
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002461/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 * fields which are the same type (int / unsigned) on our platforms.
2463 */
2464#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2465#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2466#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2467
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002468struct used_address {
2469 struct sockaddr_storage name;
2470 unsigned int name_len;
2471};
2472
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002473int __copy_msghdr(struct msghdr *kmsg,
2474 struct user_msghdr *msg,
2475 struct sockaddr __user **save_addr)
Dan Carpenter1661bf32013-10-03 00:27:20 +03002476{
Al Viro08adb7d2014-11-10 20:23:13 -05002477 ssize_t err;
2478
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002479 kmsg->msg_control_is_user = true;
Eric Dumazet1228b342022-06-22 15:02:20 +00002480 kmsg->msg_get_inq = 0;
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002481 kmsg->msg_control_user = msg->msg_control;
2482 kmsg->msg_controllen = msg->msg_controllen;
2483 kmsg->msg_flags = msg->msg_flags;
Al Viroffb07552017-06-27 19:32:04 -04002484
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002485 kmsg->msg_namelen = msg->msg_namelen;
2486 if (!msg->msg_name)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07002487 kmsg->msg_namelen = 0;
2488
Matthew Leachdbb490b2014-03-11 11:58:27 +00002489 if (kmsg->msg_namelen < 0)
2490 return -EINVAL;
2491
Dan Carpenter1661bf32013-10-03 00:27:20 +03002492 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03002493 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05002494
2495 if (save_addr)
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002496 *save_addr = msg->msg_name;
Al Viro08adb7d2014-11-10 20:23:13 -05002497
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002498 if (msg->msg_name && kmsg->msg_namelen) {
Al Viro08adb7d2014-11-10 20:23:13 -05002499 if (!save_addr) {
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002500 err = move_addr_to_kernel(msg->msg_name,
Paolo Abeni864d9662017-07-21 18:49:45 +02002501 kmsg->msg_namelen,
Al Viro08adb7d2014-11-10 20:23:13 -05002502 kmsg->msg_name);
2503 if (err < 0)
2504 return err;
2505 }
2506 } else {
2507 kmsg->msg_name = NULL;
2508 kmsg->msg_namelen = 0;
2509 }
2510
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002511 if (msg->msg_iovlen > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05002512 return -EMSGSIZE;
2513
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002514 kmsg->msg_iocb = NULL;
Pavel Begunkov7c701d92022-07-12 21:52:29 +01002515 kmsg->msg_ubuf = NULL;
Jens Axboe0a384ab2020-02-27 08:11:20 -07002516 return 0;
2517}
2518
2519static int copy_msghdr_from_user(struct msghdr *kmsg,
2520 struct user_msghdr __user *umsg,
2521 struct sockaddr __user **save_addr,
2522 struct iovec **iov)
2523{
2524 struct user_msghdr msg;
2525 ssize_t err;
2526
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002527 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2528 return -EFAULT;
2529
2530 err = __copy_msghdr(kmsg, &msg, save_addr);
Jens Axboe0a384ab2020-02-27 08:11:20 -07002531 if (err)
2532 return err;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002533
Al Virode4eda92022-09-15 20:25:47 -04002534 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
Al Viroffb07552017-06-27 19:32:04 -04002535 msg.msg_iov, msg.msg_iovlen,
Al Viroda184282015-03-21 19:29:06 -04002536 UIO_FASTIOV, iov, &kmsg->msg_iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002537 return err < 0 ? err : 0;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002538}
2539
Jens Axboe4257c8c2019-11-25 14:27:34 -07002540static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2541 unsigned int flags, struct used_address *used_address,
2542 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002543{
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002544 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Amit Kushwaha846cc122016-12-08 18:21:53 +05302545 __aligned(sizeof(__kernel_size_t));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002546 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002547 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05002548 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05002549 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002550
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551 err = -ENOBUFS;
2552
Anton Blanchard228e5482011-05-02 20:21:35 +00002553 if (msg_sys->msg_controllen > INT_MAX)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002554 goto out;
Tom Herbert28a94d82016-03-07 14:11:02 -08002555 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00002556 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002558 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002559 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002560 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002562 goto out;
Anton Blanchard228e5482011-05-02 20:21:35 +00002563 ctl_buf = msg_sys->msg_control;
2564 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565 } else if (ctl_len) {
David S. Millerac4340f2017-01-04 13:24:19 -05002566 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2567 CMSG_ALIGN(sizeof(struct cmsghdr)));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002568 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002569 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002570 if (ctl_buf == NULL)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002571 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572 }
2573 err = -EFAULT;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002574 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002576 msg_sys->msg_control = ctl_buf;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002577 msg_sys->msg_control_is_user = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578 }
David Howellsb841b902023-05-22 13:11:10 +01002579 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
Anton Blanchard228e5482011-05-02 20:21:35 +00002580 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002581
2582 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002583 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002584 /*
2585 * If this is sendmmsg() and current destination address is same as
2586 * previously succeeded address, omit asking LSM's decision.
2587 * used_address->name_len is initialized to UINT_MAX so that the first
2588 * destination address never matches.
2589 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002590 if (used_address && msg_sys->msg_name &&
2591 used_address->name_len == msg_sys->msg_namelen &&
2592 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002593 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05002594 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002595 goto out_freectl;
2596 }
Jordan Rife86a7e0b2023-09-21 18:46:41 -05002597 err = __sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002598 /*
2599 * If this is sendmmsg() and sending to current destination address was
2600 * successful, remember it.
2601 */
2602 if (used_address && err >= 0) {
2603 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002604 if (msg_sys->msg_name)
2605 memcpy(&used_address->name, msg_sys->msg_name,
2606 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002607 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608
2609out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002610 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
Jens Axboe4257c8c2019-11-25 14:27:34 -07002612out:
2613 return err;
2614}
2615
Jens Axboee54e09c2024-03-12 09:55:45 -06002616static int sendmsg_copy_msghdr(struct msghdr *msg,
2617 struct user_msghdr __user *umsg, unsigned flags,
2618 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002619{
2620 int err;
2621
2622 if (flags & MSG_CMSG_COMPAT) {
2623 struct compat_msghdr __user *msg_compat;
2624
2625 msg_compat = (struct compat_msghdr __user *) umsg;
2626 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2627 } else {
2628 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2629 }
2630 if (err < 0)
2631 return err;
2632
2633 return 0;
2634}
2635
2636static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2637 struct msghdr *msg_sys, unsigned int flags,
2638 struct used_address *used_address,
2639 unsigned int allowed_msghdr_flags)
2640{
2641 struct sockaddr_storage address;
2642 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2643 ssize_t err;
2644
2645 msg_sys->msg_name = &address;
2646
2647 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2648 if (err < 0)
2649 return err;
2650
2651 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2652 allowed_msghdr_flags);
Al Viroda184282015-03-21 19:29:06 -04002653 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002654 return err;
2655}
2656
2657/*
2658 * BSD sendmsg interface
2659 */
Jens Axboe03b12302019-12-02 18:50:25 -07002660long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
Jens Axboe0fa03c62019-04-19 13:34:07 -06002661 unsigned int flags)
2662{
Jens Axboe03b12302019-12-02 18:50:25 -07002663 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
Jens Axboe0fa03c62019-04-19 13:34:07 -06002664}
Anton Blanchard228e5482011-05-02 20:21:35 +00002665
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002666long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2667 bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002668{
2669 int fput_needed, err;
2670 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002671 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002672
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002673 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2674 return -EINVAL;
2675
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002676 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002677 if (!sock)
2678 goto out;
2679
Tom Herbert28a94d82016-03-07 14:11:02 -08002680 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00002681
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002682 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002683out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002684 return err;
2685}
2686
Al Viro666547f2014-04-06 14:03:05 -04002687SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002688{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002689 return __sys_sendmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002690}
2691
Anton Blanchard228e5482011-05-02 20:21:35 +00002692/*
2693 * Linux sendmmsg interface
2694 */
2695
2696int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002697 unsigned int flags, bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002698{
2699 int fput_needed, err, datagrams;
2700 struct socket *sock;
2701 struct mmsghdr __user *entry;
2702 struct compat_mmsghdr __user *compat_entry;
2703 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002704 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002705 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002706
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002707 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2708 return -EINVAL;
2709
Anton Blanchard98382f42011-08-04 14:07:39 +00002710 if (vlen > UIO_MAXIOV)
2711 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002712
2713 datagrams = 0;
2714
2715 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2716 if (!sock)
2717 return err;
2718
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002719 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002720 entry = mmsg;
2721 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002722 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002723 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002724
2725 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002726 if (datagrams == vlen - 1)
2727 flags = oflags;
2728
Anton Blanchard228e5482011-05-02 20:21:35 +00002729 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002730 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002731 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002732 if (err < 0)
2733 break;
2734 err = __put_user(err, &compat_entry->msg_len);
2735 ++compat_entry;
2736 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002737 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002738 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002739 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002740 if (err < 0)
2741 break;
2742 err = put_user(err, &entry->msg_len);
2743 ++entry;
2744 }
2745
2746 if (err)
2747 break;
2748 ++datagrams;
Soheil Hassas Yeganeh30238982016-11-04 15:36:49 -04002749 if (msg_data_left(&msg_sys))
2750 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002751 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002752 }
2753
Anton Blanchard228e5482011-05-02 20:21:35 +00002754 fput_light(sock->file, fput_needed);
2755
Anton Blanchard728ffb82011-08-04 14:07:38 +00002756 /* We only return an error if no datagrams were able to be sent */
2757 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002758 return datagrams;
2759
Anton Blanchard228e5482011-05-02 20:21:35 +00002760 return err;
2761}
2762
2763SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2764 unsigned int, vlen, unsigned int, flags)
2765{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002766 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002767}
2768
Jens Axboee54e09c2024-03-12 09:55:45 -06002769static int recvmsg_copy_msghdr(struct msghdr *msg,
2770 struct user_msghdr __user *umsg, unsigned flags,
2771 struct sockaddr __user **uaddr,
2772 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002773{
2774 ssize_t err;
2775
2776 if (MSG_CMSG_COMPAT & flags) {
2777 struct compat_msghdr __user *msg_compat;
2778
2779 msg_compat = (struct compat_msghdr __user *) umsg;
2780 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2781 } else {
2782 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2783 }
2784 if (err < 0)
2785 return err;
2786
2787 return 0;
2788}
2789
2790static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2791 struct user_msghdr __user *msg,
2792 struct sockaddr __user *uaddr,
2793 unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002795 struct compat_msghdr __user *msg_compat =
Jens Axboe4257c8c2019-11-25 14:27:34 -07002796 (struct compat_msghdr __user *) msg;
2797 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2798 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002799 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002800 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002801 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802
Al Viro08adb7d2014-11-10 20:23:13 -05002803 msg_sys->msg_name = &addr;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002804 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2805 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002806
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002807 /* We assume all kernel code knows the size of sockaddr_storage */
2808 msg_sys->msg_namelen = 0;
2809
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810 if (sock->file->f_flags & O_NONBLOCK)
2811 flags |= MSG_DONTWAIT;
Eric Dumazet1af66222019-12-06 09:38:36 -08002812
2813 if (unlikely(nosec))
2814 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2815 else
2816 err = sock_recvmsg(sock, msg_sys, flags);
2817
Linus Torvalds1da177e2005-04-16 15:20:36 -07002818 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002819 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002820 len = err;
2821
2822 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002823 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002824 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002825 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002826 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002827 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002828 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002829 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002830 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002832 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002834 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835 &msg_compat->msg_controllen);
2836 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002837 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002838 &msg->msg_controllen);
2839 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002840 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002841 err = len;
Jens Axboe4257c8c2019-11-25 14:27:34 -07002842out:
2843 return err;
2844}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845
Jens Axboe4257c8c2019-11-25 14:27:34 -07002846static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2847 struct msghdr *msg_sys, unsigned int flags, int nosec)
2848{
2849 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2850 /* user mode address pointers */
2851 struct sockaddr __user *uaddr;
2852 ssize_t err;
2853
2854 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2855 if (err < 0)
2856 return err;
2857
2858 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
Al Viroda184282015-03-21 19:29:06 -04002859 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002860 return err;
2861}
2862
2863/*
2864 * BSD recvmsg interface
2865 */
2866
Jens Axboe03b12302019-12-02 18:50:25 -07002867long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2868 struct user_msghdr __user *umsg,
2869 struct sockaddr __user *uaddr, unsigned int flags)
Jens Axboeaa1fa282019-04-19 13:38:09 -06002870{
Jens Axboe03b12302019-12-02 18:50:25 -07002871 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
Jens Axboeaa1fa282019-04-19 13:38:09 -06002872}
2873
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002874long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2875 bool forbid_cmsg_compat)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002876{
2877 int fput_needed, err;
2878 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002879 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002880
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002881 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2882 return -EINVAL;
2883
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002884 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002885 if (!sock)
2886 goto out;
2887
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002888 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002889
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002890 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002891out:
2892 return err;
2893}
2894
Al Viro666547f2014-04-06 14:03:05 -04002895SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002896 unsigned int, flags)
2897{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002898 return __sys_recvmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002899}
2900
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002901/*
2902 * Linux recvmmsg interface
2903 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002905static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2906 unsigned int vlen, unsigned int flags,
2907 struct timespec64 *timeout)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002908{
2909 int fput_needed, err, datagrams;
2910 struct socket *sock;
2911 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002912 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002913 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002914 struct timespec64 end_time;
2915 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002916
2917 if (timeout &&
2918 poll_select_set_timeout(&end_time, timeout->tv_sec,
2919 timeout->tv_nsec))
2920 return -EINVAL;
2921
2922 datagrams = 0;
2923
2924 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2925 if (!sock)
2926 return err;
2927
Soheil Hassas Yeganeh7797dc42018-02-27 18:22:40 -05002928 if (likely(!(flags & MSG_ERRQUEUE))) {
2929 err = sock_error(sock->sk);
2930 if (err) {
2931 datagrams = err;
2932 goto out_put;
2933 }
Maxime Jayate623a9e2017-02-21 18:35:51 +01002934 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002935
2936 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002937 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002938
2939 while (datagrams < vlen) {
2940 /*
2941 * No need to ask LSM for more than the first datagram.
2942 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002943 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002944 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002945 &msg_sys, flags & ~MSG_WAITFORONE,
2946 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002947 if (err < 0)
2948 break;
2949 err = __put_user(err, &compat_entry->msg_len);
2950 ++compat_entry;
2951 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002952 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002953 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002954 &msg_sys, flags & ~MSG_WAITFORONE,
2955 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002956 if (err < 0)
2957 break;
2958 err = put_user(err, &entry->msg_len);
2959 ++entry;
2960 }
2961
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002962 if (err)
2963 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002964 ++datagrams;
2965
Brandon L Black71c5c1592010-03-26 16:18:03 +00002966 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2967 if (flags & MSG_WAITFORONE)
2968 flags |= MSG_DONTWAIT;
2969
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002970 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002971 ktime_get_ts64(&timeout64);
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002972 *timeout = timespec64_sub(end_time, timeout64);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002973 if (timeout->tv_sec < 0) {
2974 timeout->tv_sec = timeout->tv_nsec = 0;
2975 break;
2976 }
2977
2978 /* Timeout, return less than vlen datagrams */
2979 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2980 break;
2981 }
2982
2983 /* Out of band data, return right away */
2984 if (msg_sys.msg_flags & MSG_OOB)
2985 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002986 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002987 }
2988
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002989 if (err == 0)
2990 goto out_put;
2991
2992 if (datagrams == 0) {
2993 datagrams = err;
2994 goto out_put;
2995 }
2996
2997 /*
2998 * We may return less entries than requested (vlen) if the
2999 * sock is non block and there aren't enough datagrams...
3000 */
3001 if (err != -EAGAIN) {
3002 /*
3003 * ... or if recvmsg returns an error after we
3004 * received some datagrams, where we record the
3005 * error to return on the next call or if the
3006 * app asks about it using getsockopt(SO_ERROR).
3007 */
Eric Dumazete05a5f52023-05-09 16:35:53 +00003008 WRITE_ONCE(sock->sk->sk_err, -err);
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03003009 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003010out_put:
3011 fput_light(sock->file, fput_needed);
3012
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03003013 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003014}
3015
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003016int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3017 unsigned int vlen, unsigned int flags,
3018 struct __kernel_timespec __user *timeout,
3019 struct old_timespec32 __user *timeout32)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003020{
3021 int datagrams;
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02003022 struct timespec64 timeout_sys;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003023
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003024 if (timeout && get_timespec64(&timeout_sys, timeout))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003025 return -EFAULT;
3026
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003027 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3028 return -EFAULT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003029
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003030 if (!timeout && !timeout32)
3031 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3032
3033 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3034
3035 if (datagrams <= 0)
3036 return datagrams;
3037
3038 if (timeout && put_timespec64(&timeout_sys, timeout))
3039 datagrams = -EFAULT;
3040
3041 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003042 datagrams = -EFAULT;
3043
3044 return datagrams;
3045}
3046
Dominik Brodowski1255e262018-03-13 20:44:21 +01003047SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3048 unsigned int, vlen, unsigned int, flags,
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02003049 struct __kernel_timespec __user *, timeout)
Dominik Brodowski1255e262018-03-13 20:44:21 +01003050{
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003051 if (flags & MSG_CMSG_COMPAT)
3052 return -EINVAL;
3053
3054 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
Dominik Brodowski1255e262018-03-13 20:44:21 +01003055}
3056
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003057#ifdef CONFIG_COMPAT_32BIT_TIME
3058SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3059 unsigned int, vlen, unsigned int, flags,
3060 struct old_timespec32 __user *, timeout)
3061{
3062 if (flags & MSG_CMSG_COMPAT)
3063 return -EINVAL;
3064
3065 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3066}
3067#endif
3068
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003069#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07003070/* Argument list sizes for sys_socketcall */
3071#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00003072static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003073 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3074 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3075 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00003076 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003077};
3078
Linus Torvalds1da177e2005-04-16 15:20:36 -07003079#undef AL
3080
3081/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003082 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003083 *
3084 * Argument checking cleaned up. Saved 20% in size.
3085 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003086 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003087 */
3088
Heiko Carstens3e0fa652009-01-14 14:14:24 +01003089SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003090{
Chen Gang2950fa92013-04-07 16:55:23 +08003091 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003092 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003093 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07003094 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003095
Anton Blanchard228e5482011-05-02 20:21:35 +00003096 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003097 return -EINVAL;
Jeremy Clinec8e8cd52018-07-27 22:43:01 +00003098 call = array_index_nospec(call, SYS_SENDMMSG + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003099
Arjan van de Ven47379052009-09-28 12:57:44 -07003100 len = nargs[call];
3101 if (len > sizeof(a))
3102 return -EINVAL;
3103
Linus Torvalds1da177e2005-04-16 15:20:36 -07003104 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07003105 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003106 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01003107
Chen Gang2950fa92013-04-07 16:55:23 +08003108 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3109 if (err)
3110 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01003111
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003112 a0 = a[0];
3113 a1 = a[1];
3114
3115 switch (call) {
3116 case SYS_SOCKET:
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01003117 err = __sys_socket(a0, a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003118 break;
3119 case SYS_BIND:
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01003120 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003121 break;
3122 case SYS_CONNECT:
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01003123 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003124 break;
3125 case SYS_LISTEN:
Dominik Brodowski25e290e2018-03-13 19:36:54 +01003126 err = __sys_listen(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003127 break;
3128 case SYS_ACCEPT:
Dominik Brodowski4541e802018-03-13 19:24:23 +01003129 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3130 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003131 break;
3132 case SYS_GETSOCKNAME:
3133 err =
Dominik Brodowski8882a102018-03-13 19:43:14 +01003134 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3135 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003136 break;
3137 case SYS_GETPEERNAME:
3138 err =
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01003139 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3140 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003141 break;
3142 case SYS_SOCKETPAIR:
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01003143 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003144 break;
3145 case SYS_SEND:
Dominik Brodowskif3bf8962018-03-13 19:52:00 +01003146 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3147 NULL, 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003148 break;
3149 case SYS_SENDTO:
Dominik Brodowski211b6342018-03-13 19:18:52 +01003150 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3151 (struct sockaddr __user *)a[4], a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003152 break;
3153 case SYS_RECV:
Dominik Brodowskid27e9af2018-03-13 19:54:17 +01003154 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3155 NULL, NULL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003156 break;
3157 case SYS_RECVFROM:
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01003158 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3159 (struct sockaddr __user *)a[4],
3160 (int __user *)a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003161 break;
3162 case SYS_SHUTDOWN:
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01003163 err = __sys_shutdown(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003164 break;
3165 case SYS_SETSOCKOPT:
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01003166 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3167 a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003168 break;
3169 case SYS_GETSOCKOPT:
3170 err =
Dominik Brodowski13a2d702018-03-13 20:15:04 +01003171 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3172 (int __user *)a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003173 break;
3174 case SYS_SENDMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003175 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3176 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003177 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00003178 case SYS_SENDMMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003179 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3180 a[3], true);
Anton Blanchard228e5482011-05-02 20:21:35 +00003181 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003182 case SYS_RECVMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003183 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3184 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003185 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003186 case SYS_RECVMMSG:
Arnd Bergmann3ca47e92019-04-23 17:43:50 +02003187 if (IS_ENABLED(CONFIG_64BIT))
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003188 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3189 a[2], a[3],
3190 (struct __kernel_timespec __user *)a[4],
3191 NULL);
3192 else
3193 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3194 a[2], a[3], NULL,
3195 (struct old_timespec32 __user *)a[4]);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003196 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08003197 case SYS_ACCEPT4:
Dominik Brodowski4541e802018-03-13 19:24:23 +01003198 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3199 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07003200 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003201 default:
3202 err = -EINVAL;
3203 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003204 }
3205 return err;
3206}
3207
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003208#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003209
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003210/**
3211 * sock_register - add a socket protocol handler
3212 * @ops: description of protocol
3213 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003214 * This function is called by a protocol handler that wants to
3215 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09003216 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003217 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003218 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003219int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003220{
3221 int err;
3222
3223 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08003224 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003225 return -ENOBUFS;
3226 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003227
3228 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00003229 if (rcu_dereference_protected(net_families[ops->family],
3230 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003231 err = -EEXIST;
3232 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00003233 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003234 err = 0;
3235 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003236 spin_unlock(&net_family_lock);
3237
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003238 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003239 return err;
3240}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003241EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003242
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003243/**
3244 * sock_unregister - remove a protocol handler
3245 * @family: protocol family to remove
3246 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003247 * This function is called by a protocol handler that wants to
3248 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003249 * new socket creation.
3250 *
3251 * If protocol handler is a module, then it can use module reference
3252 * counts to protect against new references. If protocol handler is not
3253 * a module then it needs to provide its own protection in
3254 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003255 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003256void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003257{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003258 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003259
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003260 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00003261 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003262 spin_unlock(&net_family_lock);
3263
3264 synchronize_rcu();
3265
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003266 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003267}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003268EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003269
Xin Longbf2ae2e2018-03-10 18:57:50 +08003270bool sock_is_registered(int family)
3271{
Jeremy Cline66b51b02018-08-13 22:23:13 +00003272 return family < NPROTO && rcu_access_pointer(net_families[family]);
Xin Longbf2ae2e2018-03-10 18:57:50 +08003273}
3274
Andi Kleen77d76ea2005-12-22 12:43:42 -08003275static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003276{
Nick Pigginb3e19d92011-01-07 17:50:11 +11003277 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00003278 /*
3279 * Initialize the network sysctl infrastructure.
3280 */
3281 err = net_sysctl_init();
3282 if (err)
3283 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11003284
Linus Torvalds1da177e2005-04-16 15:20:36 -07003285 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003286 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07003287 */
3288 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003289
3290 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003291 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003292 */
3293
3294 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11003295
3296 err = register_filesystem(&sock_fs_type);
3297 if (err)
Miaohe Lin47260ba2020-08-06 19:54:19 +08003298 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003299 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003300 if (IS_ERR(sock_mnt)) {
3301 err = PTR_ERR(sock_mnt);
3302 goto out_mount;
3303 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08003304
3305 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003306 */
3307
3308#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00003309 err = netfilter_init();
3310 if (err)
3311 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003312#endif
David S. Millercbeb3212005-12-22 12:58:55 -08003313
Daniel Borkmann408eccc2014-04-01 16:20:23 +02003314 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00003315
Nick Pigginb3e19d92011-01-07 17:50:11 +11003316out:
3317 return err;
3318
3319out_mount:
3320 unregister_filesystem(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003321 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322}
3323
Andi Kleen77d76ea2005-12-22 12:43:42 -08003324core_initcall(sock_init); /* early initcall */
3325
Linus Torvalds1da177e2005-04-16 15:20:36 -07003326#ifdef CONFIG_PROC_FS
3327void socket_seq_show(struct seq_file *seq)
3328{
Tonghao Zhang648845a2017-12-14 05:51:58 -08003329 seq_printf(seq, "sockets: used %d\n",
3330 sock_inuse_get(seq->private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003331}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003332#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003333
Arnd Bergmann29c49642021-07-22 16:29:03 +02003334/* Handle the fact that while struct ifreq has the same *layout* on
3335 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3336 * which are handled elsewhere, it still has different *size* due to
3337 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3338 * resulting in struct ifreq being 32 and 40 bytes respectively).
3339 * As a result, if the struct happens to be at the end of a page and
3340 * the next page isn't readable/writable, we get a fault. To prevent
3341 * that, copy back and forth to the full size.
3342 */
3343int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3344{
3345 if (in_compat_syscall()) {
3346 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3347
3348 memset(ifr, 0, sizeof(*ifr));
3349 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3350 return -EFAULT;
3351
3352 if (ifrdata)
3353 *ifrdata = compat_ptr(ifr32->ifr_data);
3354
3355 return 0;
3356 }
3357
3358 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3359 return -EFAULT;
3360
3361 if (ifrdata)
3362 *ifrdata = ifr->ifr_data;
3363
3364 return 0;
3365}
3366EXPORT_SYMBOL(get_user_ifreq);
3367
3368int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3369{
3370 size_t size = sizeof(*ifr);
3371
3372 if (in_compat_syscall())
3373 size = sizeof(struct compat_ifreq);
3374
3375 if (copy_to_user(arg, ifr, size))
3376 return -EFAULT;
3377
3378 return 0;
3379}
3380EXPORT_SYMBOL(put_user_ifreq);
3381
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003382#ifdef CONFIG_COMPAT
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003383static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3384{
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003385 compat_uptr_t uptr32;
Al Viro44c02a22017-10-05 12:59:44 -04003386 struct ifreq ifr;
3387 void __user *saved;
3388 int err;
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003389
Arnd Bergmann29c49642021-07-22 16:29:03 +02003390 if (get_user_ifreq(&ifr, NULL, uifr32))
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003391 return -EFAULT;
3392
3393 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3394 return -EFAULT;
3395
Al Viro44c02a22017-10-05 12:59:44 -04003396 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3397 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003398
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003399 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
Al Viro44c02a22017-10-05 12:59:44 -04003400 if (!err) {
3401 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
Arnd Bergmann29c49642021-07-22 16:29:03 +02003402 if (put_user_ifreq(&ifr, uifr32))
Al Viro44c02a22017-10-05 12:59:44 -04003403 err = -EFAULT;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003404 }
Al Viro44c02a22017-10-05 12:59:44 -04003405 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003406}
3407
Ben Hutchings590d4692013-11-18 17:04:13 +00003408/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3409static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08003410 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003411{
Al Viro44c02a22017-10-05 12:59:44 -04003412 struct ifreq ifreq;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003413 void __user *data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003414
Peter Collingbourned0efb162021-08-26 12:46:01 -07003415 if (!is_socket_ioctl_cmd(cmd))
3416 return -ENOTTY;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003417 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003418 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003419 ifreq.ifr_data = data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003420
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003421 return dev_ioctl(net, cmd, &ifreq, data, NULL);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003422}
3423
Arnd Bergmann6b960182009-11-06 23:10:54 -08003424static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3425 unsigned int cmd, unsigned long arg)
3426{
3427 void __user *argp = compat_ptr(arg);
3428 struct sock *sk = sock->sk;
3429 struct net *net = sock_net(sk);
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003430 const struct proto_ops *ops;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003431
Arnd Bergmann6b960182009-11-06 23:10:54 -08003432 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Arnd Bergmann88fc0232021-07-27 15:45:15 +02003433 return sock_ioctl(file, cmd, (unsigned long)argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003434
Arnd Bergmann6b960182009-11-06 23:10:54 -08003435 switch (cmd) {
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003436 case SIOCWANDEV:
3437 return compat_siocwandev(net, argp);
Arnd Bergmann0768e172019-04-17 22:56:11 +02003438 case SIOCGSTAMP_OLD:
3439 case SIOCGSTAMPNS_OLD:
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003440 ops = READ_ONCE(sock->ops);
3441 if (!ops->gettstamp)
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003442 return -ENOIOCTLCMD;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003443 return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3444 !COMPAT_USE_64BIT_TIME);
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003445
Arnd Bergmanndd98d282021-07-22 16:28:59 +02003446 case SIOCETHTOOL:
Ben Hutchings590d4692013-11-18 17:04:13 +00003447 case SIOCBONDSLAVEINFOQUERY:
3448 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003449 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003450 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003451 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003452
Arnd Bergmann6b960182009-11-06 23:10:54 -08003453 case FIOSETOWN:
3454 case SIOCSPGRP:
3455 case FIOGETOWN:
3456 case SIOCGPGRP:
3457 case SIOCBRADDBR:
3458 case SIOCBRDELBR:
3459 case SIOCGIFVLAN:
3460 case SIOCSIFVLAN:
Andrey Vaginc62cce22016-10-24 18:29:13 -07003461 case SIOCGSKNS:
Arnd Bergmann0768e172019-04-17 22:56:11 +02003462 case SIOCGSTAMP_NEW:
3463 case SIOCGSTAMPNS_NEW:
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02003464 case SIOCGIFCONF:
Remi Pommarelfd3a4592021-12-24 12:46:40 +01003465 case SIOCSIFBR:
3466 case SIOCGIFBR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003467 return sock_ioctl(file, cmd, arg);
3468
3469 case SIOCGIFFLAGS:
3470 case SIOCSIFFLAGS:
Arnd Bergmann709566d2021-07-22 16:29:00 +02003471 case SIOCGIFMAP:
3472 case SIOCSIFMAP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003473 case SIOCGIFMETRIC:
3474 case SIOCSIFMETRIC:
3475 case SIOCGIFMTU:
3476 case SIOCSIFMTU:
3477 case SIOCGIFMEM:
3478 case SIOCSIFMEM:
3479 case SIOCGIFHWADDR:
3480 case SIOCSIFHWADDR:
3481 case SIOCADDMULTI:
3482 case SIOCDELMULTI:
3483 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003484 case SIOCGIFADDR:
3485 case SIOCSIFADDR:
3486 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003487 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003488 case SIOCGIFBRDADDR:
3489 case SIOCSIFBRDADDR:
3490 case SIOCGIFDSTADDR:
3491 case SIOCSIFDSTADDR:
3492 case SIOCGIFNETMASK:
3493 case SIOCSIFNETMASK:
3494 case SIOCSIFPFLAGS:
3495 case SIOCGIFPFLAGS:
3496 case SIOCGIFTXQLEN:
3497 case SIOCSIFTXQLEN:
3498 case SIOCBRADDIF:
3499 case SIOCBRDELIF:
Johannes Bergc6c9fee2019-01-25 22:43:19 +01003500 case SIOCGIFNAME:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003501 case SIOCSIFNAME:
3502 case SIOCGMIIPHY:
3503 case SIOCGMIIREG:
3504 case SIOCSMIIREG:
Al Virof92d4fc2017-09-30 19:32:17 -04003505 case SIOCBONDENSLAVE:
3506 case SIOCBONDRELEASE:
3507 case SIOCBONDSETHWADDR:
3508 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003509 case SIOCSARP:
3510 case SIOCGARP:
3511 case SIOCDARP:
Arnd Bergmannc7dc5042019-06-03 23:07:12 +02003512 case SIOCOUTQ:
Arnd Bergmann9d7bf412019-06-03 23:06:00 +02003513 case SIOCOUTQNSD:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003514 case SIOCATMARK:
Johannes Berg63ff03a2019-01-25 22:43:17 +01003515 return sock_do_ioctl(net, sock, cmd, arg);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003516 }
3517
Arnd Bergmann6b960182009-11-06 23:10:54 -08003518 return -ENOIOCTLCMD;
3519}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003520
Eric Dumazet95c96172012-04-15 05:58:06 +00003521static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003522 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003523{
3524 struct socket *sock = file->private_data;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003525 const struct proto_ops *ops = READ_ONCE(sock->ops);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003526 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003527 struct sock *sk;
3528 struct net *net;
3529
3530 sk = sock->sk;
3531 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003532
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003533 if (ops->compat_ioctl)
3534 ret = ops->compat_ioctl(sock, cmd, arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003535
David S. Miller87de87d2008-06-03 09:14:03 -07003536 if (ret == -ENOIOCTLCMD &&
3537 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3538 ret = compat_wext_handle_ioctl(net, cmd, arg);
3539
Arnd Bergmann6b960182009-11-06 23:10:54 -08003540 if (ret == -ENOIOCTLCMD)
3541 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3542
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003543 return ret;
3544}
3545#endif
3546
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003547/**
3548 * kernel_bind - bind an address to a socket (kernel space)
3549 * @sock: socket
3550 * @addr: address
3551 * @addrlen: length of address
3552 *
3553 * Returns 0 or an error.
3554 */
3555
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003556int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3557{
Jordan Rifec889a99a2023-09-21 18:46:42 -05003558 struct sockaddr_storage address;
3559
3560 memcpy(&address, addr, addrlen);
3561
3562 return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3563 addrlen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003564}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003565EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003566
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003567/**
3568 * kernel_listen - move socket to listening state (kernel space)
3569 * @sock: socket
3570 * @backlog: pending connections queue size
3571 *
3572 * Returns 0 or an error.
3573 */
3574
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003575int kernel_listen(struct socket *sock, int backlog)
3576{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003577 return READ_ONCE(sock->ops)->listen(sock, backlog);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003578}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003579EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003580
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003581/**
3582 * kernel_accept - accept a connection (kernel space)
3583 * @sock: listening socket
3584 * @newsock: new connected socket
3585 * @flags: flags
3586 *
3587 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3588 * If it fails, @newsock is guaranteed to be %NULL.
3589 * Returns 0 or an error.
3590 */
3591
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003592int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3593{
3594 struct sock *sk = sock->sk;
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003595 const struct proto_ops *ops = READ_ONCE(sock->ops);
Jens Axboe92ef0fd2024-05-09 09:20:08 -06003596 struct proto_accept_arg arg = {
3597 .flags = flags,
3598 .kern = true,
3599 };
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003600 int err;
3601
3602 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3603 newsock);
3604 if (err < 0)
3605 goto done;
3606
Jens Axboe92ef0fd2024-05-09 09:20:08 -06003607 err = ops->accept(sock, *newsock, &arg);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003608 if (err < 0) {
3609 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003610 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003611 goto done;
3612 }
3613
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003614 (*newsock)->ops = ops;
3615 __module_get(ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003616
3617done:
3618 return err;
3619}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003620EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003621
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003622/**
3623 * kernel_connect - connect a socket (kernel space)
3624 * @sock: socket
3625 * @addr: address
3626 * @addrlen: address length
3627 * @flags: flags (O_NONBLOCK, ...)
3628 *
Lu Weif1dcffc2021-03-25 11:01:55 +08003629 * For datagram sockets, @addr is the address to which datagrams are sent
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003630 * by default, and the only address from which datagrams are received.
3631 * For stream sockets, attempts to connect to @addr.
3632 * Returns 0 or an error code.
3633 */
3634
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003635int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003636 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003637{
Jordan Rife0bdf3992023-08-21 16:45:23 -05003638 struct sockaddr_storage address;
3639
3640 memcpy(&address, addr, addrlen);
3641
3642 return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3643 addrlen, flags);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003644}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003645EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003646
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003647/**
3648 * kernel_getsockname - get the address which the socket is bound (kernel space)
3649 * @sock: socket
3650 * @addr: address holder
3651 *
3652 * Fills the @addr pointer with the address which the socket is bound.
Alex Maydanik0fc95de2022-02-12 12:29:27 +02003653 * Returns the length of the address in bytes or an error code.
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003654 */
3655
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003656int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003657{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003658 return READ_ONCE(sock->ops)->getname(sock, addr, 0);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003659}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003660EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003661
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003662/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003663 * kernel_getpeername - get the address which the socket is connected (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003664 * @sock: socket
3665 * @addr: address holder
3666 *
3667 * Fills the @addr pointer with the address which the socket is connected.
Alex Maydanik0fc95de2022-02-12 12:29:27 +02003668 * Returns the length of the address in bytes or an error code.
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003669 */
3670
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003671int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003672{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003673 return READ_ONCE(sock->ops)->getname(sock, addr, 1);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003674}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003675EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003676
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003677/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003678 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003679 * @sock: socket
3680 * @how: connection part
3681 *
3682 * Returns 0 or an error.
3683 */
3684
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003685int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3686{
Eric Dumazet1ded5e52023-08-08 13:58:09 +00003687 return READ_ONCE(sock->ops)->shutdown(sock, how);
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003688}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003689EXPORT_SYMBOL(kernel_sock_shutdown);
R. Parameswaran113c3072017-04-05 16:50:35 -07003690
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003691/**
3692 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3693 * @sk: socket
3694 *
3695 * This routine returns the IP overhead imposed by a socket i.e.
3696 * the length of the underlying IP header, depending on whether
3697 * this is an IPv4 or IPv6 socket and the length from IP options turned
3698 * on at the socket. Assumes that the caller has a lock on the socket.
R. Parameswaran113c3072017-04-05 16:50:35 -07003699 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003700
R. Parameswaran113c3072017-04-05 16:50:35 -07003701u32 kernel_sock_ip_overhead(struct sock *sk)
3702{
3703 struct inet_sock *inet;
3704 struct ip_options_rcu *opt;
3705 u32 overhead = 0;
R. Parameswaran113c3072017-04-05 16:50:35 -07003706#if IS_ENABLED(CONFIG_IPV6)
3707 struct ipv6_pinfo *np;
3708 struct ipv6_txoptions *optv6 = NULL;
3709#endif /* IS_ENABLED(CONFIG_IPV6) */
3710
3711 if (!sk)
3712 return overhead;
3713
R. Parameswaran113c3072017-04-05 16:50:35 -07003714 switch (sk->sk_family) {
3715 case AF_INET:
3716 inet = inet_sk(sk);
3717 overhead += sizeof(struct iphdr);
3718 opt = rcu_dereference_protected(inet->inet_opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003719 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003720 if (opt)
3721 overhead += opt->opt.optlen;
3722 return overhead;
3723#if IS_ENABLED(CONFIG_IPV6)
3724 case AF_INET6:
3725 np = inet6_sk(sk);
3726 overhead += sizeof(struct ipv6hdr);
3727 if (np)
3728 optv6 = rcu_dereference_protected(np->opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003729 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003730 if (optv6)
3731 overhead += (optv6->opt_flen + optv6->opt_nflen);
3732 return overhead;
3733#endif /* IS_ENABLED(CONFIG_IPV6) */
3734 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3735 return overhead;
3736 }
3737}
3738EXPORT_SYMBOL(kernel_sock_ip_overhead);