blob: a7b4b37d86df7a9232d582a14863c05b5fd34b68 [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>
60#include <linux/net.h>
61#include <linux/interrupt.h>
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -070062#include <linux/thread_info.h>
Stephen Hemminger55737fd2006-09-01 00:23:39 -070063#include <linux/rcupdate.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070064#include <linux/netdevice.h>
65#include <linux/proc_fs.h>
66#include <linux/seq_file.h>
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -080067#include <linux/mutex.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070068#include <linux/if_bridge.h>
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -030069#include <linux/if_vlan.h>
Daniel Borkmann408eccc2014-04-01 16:20:23 +020070#include <linux/ptp_classify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070071#include <linux/init.h>
72#include <linux/poll.h>
73#include <linux/cache.h>
74#include <linux/module.h>
75#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070076#include <linux/mount.h>
David Howellsfba9be42019-03-25 16:38:24 +000077#include <linux/pseudo_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070078#include <linux/security.h>
79#include <linux/syscalls.h>
80#include <linux/compat.h>
81#include <linux/kmod.h>
David Woodhouse3ec3b2f2005-05-17 12:08:48 +010082#include <linux/audit.h>
Adrian Bunkd86b5e02006-01-21 00:46:55 +010083#include <linux/wireless.h>
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -070084#include <linux/nsproxy.h>
Nick Black1fd7317d2009-09-22 16:43:33 -070085#include <linux/magic.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090086#include <linux/slab.h>
Masatake YAMATO600e1772012-08-29 10:44:29 +000087#include <linux/xattr.h>
Jeremy Clinec8e8cd52018-07-27 22:43:01 +000088#include <linux/nospec.h>
Paolo Abeni8c3c4472019-05-03 17:01:39 +020089#include <linux/indirect_call_wrapper.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070090
Linus Torvalds7c0f6ba2016-12-24 11:46:01 -080091#include <linux/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070092#include <asm/unistd.h>
93
94#include <net/compat.h>
David S. Miller87de87d2008-06-03 09:14:03 -070095#include <net/wext.h>
Herbert Xuf8451722010-05-24 00:12:34 -070096#include <net/cls_cgroup.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
98#include <net/sock.h>
99#include <linux/netfilter.h>
100
Arnd Bergmann6b960182009-11-06 23:10:54 -0800101#include <linux/if_tun.h>
102#include <linux/ipv6_route.h>
103#include <linux/route.h>
Arnd Bergmannc7dc5042019-06-03 23:07:12 +0200104#include <linux/termios.h>
Arnd Bergmann6b960182009-11-06 23:10:54 -0800105#include <linux/sockios.h>
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300106#include <net/busy_poll.h>
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400107#include <linux/errqueue.h>
Yangbo Lud7c08822021-06-30 16:12:00 +0800108#include <linux/ptp_clock_kernel.h>
Yunhui Cui6e6eda42023-01-11 14:59:30 +0800109#include <trace/events/sock.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300110
Cong Wange0d10952013-08-01 11:10:25 +0800111#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300112unsigned int sysctl_net_busy_read __read_mostly;
113unsigned int sysctl_net_busy_poll __read_mostly;
Eliezer Tamir06021292013-06-10 11:39:50 +0300114#endif
Arnd Bergmann6b960182009-11-06 23:10:54 -0800115
Al Viro8ae5e0302014-11-28 19:40:50 -0500116static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700118static int sock_mmap(struct file *file, struct vm_area_struct *vma);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119
120static int sock_close(struct inode *inode, struct file *file);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -0700121static __poll_t sock_poll(struct file *file,
122 struct poll_table_struct *wait);
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700123static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800124#ifdef CONFIG_COMPAT
125static long compat_sock_ioctl(struct file *file,
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700126 unsigned int cmd, unsigned long arg);
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800127#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128static int sock_fasync(int fd, struct file *filp, int on);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
Jens Axboe9c55e012007-11-06 23:30:13 -0800131static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700132 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -0800133 unsigned int flags);
Arnd Bergmann542d3062020-01-08 22:44:43 +0100134
135#ifdef CONFIG_PROC_FS
136static void sock_show_fdinfo(struct seq_file *m, struct file *f)
137{
138 struct socket *sock = f->private_data;
139
140 if (sock->ops->show_fdinfo)
141 sock->ops->show_fdinfo(m, sock);
142}
143#else
144#define sock_show_fdinfo NULL
145#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147/*
148 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
149 * in the operation structures but are done directly via the socketcall() multiplexor.
150 */
151
Arjan van de Venda7071d2007-02-12 00:55:36 -0800152static const struct file_operations socket_file_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 .owner = THIS_MODULE,
154 .llseek = no_llseek,
Al Viro8ae5e0302014-11-28 19:40:50 -0500155 .read_iter = sock_read_iter,
156 .write_iter = sock_write_iter,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 .poll = sock_poll,
158 .unlocked_ioctl = sock_ioctl,
Shaun Pereira89bbfc92006-03-21 23:58:08 -0800159#ifdef CONFIG_COMPAT
160 .compat_ioctl = compat_sock_ioctl,
161#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 .mmap = sock_mmap,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 .release = sock_close,
164 .fasync = sock_fasync,
Jens Axboe5274f052006-03-30 15:15:30 +0200165 .sendpage = sock_sendpage,
166 .splice_write = generic_splice_sendpage,
Jens Axboe9c55e012007-11-06 23:30:13 -0800167 .splice_read = sock_splice_read,
Kirill Tkhaib4653342019-12-09 13:03:40 +0300168 .show_fdinfo = sock_show_fdinfo,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169};
170
Yejune Dengfe0bdbd2021-06-21 13:12:25 +0800171static const char * const pf_family_names[] = {
172 [PF_UNSPEC] = "PF_UNSPEC",
173 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
174 [PF_INET] = "PF_INET",
175 [PF_AX25] = "PF_AX25",
176 [PF_IPX] = "PF_IPX",
177 [PF_APPLETALK] = "PF_APPLETALK",
178 [PF_NETROM] = "PF_NETROM",
179 [PF_BRIDGE] = "PF_BRIDGE",
180 [PF_ATMPVC] = "PF_ATMPVC",
181 [PF_X25] = "PF_X25",
182 [PF_INET6] = "PF_INET6",
183 [PF_ROSE] = "PF_ROSE",
184 [PF_DECnet] = "PF_DECnet",
185 [PF_NETBEUI] = "PF_NETBEUI",
186 [PF_SECURITY] = "PF_SECURITY",
187 [PF_KEY] = "PF_KEY",
188 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
189 [PF_PACKET] = "PF_PACKET",
190 [PF_ASH] = "PF_ASH",
191 [PF_ECONET] = "PF_ECONET",
192 [PF_ATMSVC] = "PF_ATMSVC",
193 [PF_RDS] = "PF_RDS",
194 [PF_SNA] = "PF_SNA",
195 [PF_IRDA] = "PF_IRDA",
196 [PF_PPPOX] = "PF_PPPOX",
197 [PF_WANPIPE] = "PF_WANPIPE",
198 [PF_LLC] = "PF_LLC",
199 [PF_IB] = "PF_IB",
200 [PF_MPLS] = "PF_MPLS",
201 [PF_CAN] = "PF_CAN",
202 [PF_TIPC] = "PF_TIPC",
203 [PF_BLUETOOTH] = "PF_BLUETOOTH",
204 [PF_IUCV] = "PF_IUCV",
205 [PF_RXRPC] = "PF_RXRPC",
206 [PF_ISDN] = "PF_ISDN",
207 [PF_PHONET] = "PF_PHONET",
208 [PF_IEEE802154] = "PF_IEEE802154",
209 [PF_CAIF] = "PF_CAIF",
210 [PF_ALG] = "PF_ALG",
211 [PF_NFC] = "PF_NFC",
212 [PF_VSOCK] = "PF_VSOCK",
213 [PF_KCM] = "PF_KCM",
214 [PF_QIPCRTR] = "PF_QIPCRTR",
215 [PF_SMC] = "PF_SMC",
216 [PF_XDP] = "PF_XDP",
Jeremy Kerrbc49d812021-07-29 10:20:39 +0800217 [PF_MCTP] = "PF_MCTP",
Yejune Dengfe0bdbd2021-06-21 13:12:25 +0800218};
219
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220/*
221 * The protocol list. Each protocol is registered in here.
222 */
223
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224static DEFINE_SPINLOCK(net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +0000225static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
Linus Torvalds1da177e2005-04-16 15:20:36 -0700227/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700228 * Support routines.
229 * Move socket addresses back and forth across the kernel/user
230 * divide and look after the messy bits.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231 */
232
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233/**
234 * move_addr_to_kernel - copy a socket address into kernel space
235 * @uaddr: Address in user space
236 * @kaddr: Address in kernel space
237 * @ulen: Length in user space
238 *
239 * The address is copied into kernel space. If the provided address is
240 * too long an error code of -EINVAL is returned. If the copy gives
241 * invalid addresses -EFAULT is returned. On a success 0 is returned.
242 */
243
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000244int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245{
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -0700246 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700248 if (ulen == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 return 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700250 if (copy_from_user(kaddr, uaddr, ulen))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +0100252 return audit_sockaddr(ulen, kaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253}
254
255/**
256 * move_addr_to_user - copy an address to user space
257 * @kaddr: kernel space address
258 * @klen: length of address in kernel
259 * @uaddr: user space address
260 * @ulen: pointer to user length field
261 *
262 * The value pointed to by ulen on entry is the buffer length available.
263 * This is overwritten with the buffer space used. -EINVAL is returned
264 * if an overlong buffer is specified or a negative buffer size. -EFAULT
265 * is returned if either the buffer or the length field are not
266 * accessible.
267 * After copying the data up to the limit the user specifies, the true
268 * length of the data is written over the length limit the user
269 * specified. Zero is returned for a success.
270 */
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700271
Maciej Żenczykowski43db3622012-03-11 12:51:50 +0000272static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
stephen hemminger11165f12010-10-18 14:27:29 +0000273 void __user *uaddr, int __user *ulen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274{
275 int err;
276 int len;
277
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100278 BUG_ON(klen > sizeof(struct sockaddr_storage));
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700279 err = get_user(len, ulen);
280 if (err)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700282 if (len > klen)
283 len = klen;
Hannes Frederic Sowa68c6beb2013-11-21 03:14:34 +0100284 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700286 if (len) {
Steve Grubbd6fe3942006-03-30 12:20:22 -0500287 if (audit_sockaddr(klen, kaddr))
288 return -ENOMEM;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700289 if (copy_to_user(uaddr, kaddr, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 return -EFAULT;
291 }
292 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700293 * "fromlen shall refer to the value before truncation.."
294 * 1003.1g
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295 */
296 return __put_user(klen, ulen);
297}
298
Alexey Dobriyan08009a72018-02-24 21:20:33 +0300299static struct kmem_cache *sock_inode_cachep __ro_after_init;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700300
301static struct inode *sock_alloc_inode(struct super_block *sb)
302{
303 struct socket_alloc *ei;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700304
Muchun Songfd60b282022-03-22 14:41:03 -0700305 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 if (!ei)
307 return NULL;
Al Viro333f7902019-07-05 20:14:16 +0100308 init_waitqueue_head(&ei->socket.wq.wait);
309 ei->socket.wq.fasync_list = NULL;
310 ei->socket.wq.flags = 0;
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700311
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 ei->socket.state = SS_UNCONNECTED;
313 ei->socket.flags = 0;
314 ei->socket.ops = NULL;
315 ei->socket.sk = NULL;
316 ei->socket.file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
318 return &ei->vfs_inode;
319}
320
Al Viro6d7855c2019-07-05 20:13:22 +0100321static void sock_free_inode(struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322{
Eric Dumazet43815482010-04-29 11:01:49 +0000323 struct socket_alloc *ei;
324
325 ei = container_of(inode, struct socket_alloc, vfs_inode);
Eric Dumazet43815482010-04-29 11:01:49 +0000326 kmem_cache_free(sock_inode_cachep, ei);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327}
328
Alexey Dobriyan51cc5062008-07-25 19:45:34 -0700329static void init_once(void *foo)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700331 struct socket_alloc *ei = (struct socket_alloc *)foo;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332
Christoph Lametera35afb82007-05-16 22:10:57 -0700333 inode_init_once(&ei->vfs_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334}
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700335
yuan linyu1e911632017-01-07 17:18:31 +0800336static void init_inodecache(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337{
338 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700339 sizeof(struct socket_alloc),
340 0,
341 (SLAB_HWCACHE_ALIGN |
342 SLAB_RECLAIM_ACCOUNT |
Vladimir Davydov5d097052016-01-14 15:18:21 -0800343 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
Paul Mundt20c2df82007-07-20 10:11:58 +0900344 init_once);
yuan linyu1e911632017-01-07 17:18:31 +0800345 BUG_ON(sock_inode_cachep == NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346}
347
Alexey Dobriyanb87221d2009-09-21 17:01:09 -0700348static const struct super_operations sockfs_ops = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700349 .alloc_inode = sock_alloc_inode,
Al Viro6d7855c2019-07-05 20:13:22 +0100350 .free_inode = sock_free_inode,
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700351 .statfs = simple_statfs,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352};
353
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700354/*
355 * sockfs_dname() is called from d_path().
356 */
357static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
358{
Al Viro0f60d282022-01-30 15:03:49 -0500359 return dynamic_dname(buffer, buflen, "socket:[%lu]",
David Howellsc5ef6032015-03-17 22:26:16 +0000360 d_inode(dentry)->i_ino);
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700361}
362
Al Viro3ba13d12009-02-20 06:02:22 +0000363static const struct dentry_operations sockfs_dentry_operations = {
Eric Dumazetc23fbb62007-05-08 00:26:18 -0700364 .d_dname = sockfs_dname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365};
366
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200367static int sockfs_xattr_get(const struct xattr_handler *handler,
368 struct dentry *dentry, struct inode *inode,
369 const char *suffix, void *value, size_t size)
370{
371 if (value) {
372 if (dentry->d_name.len + 1 > size)
373 return -ERANGE;
374 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
375 }
376 return dentry->d_name.len + 1;
377}
378
379#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
380#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
381#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
382
383static const struct xattr_handler sockfs_xattr_handler = {
384 .name = XATTR_NAME_SOCKPROTONAME,
385 .get = sockfs_xattr_get,
386};
387
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100388static int sockfs_security_xattr_set(const struct xattr_handler *handler,
Christian Brauner39f60c12023-01-13 12:49:23 +0100389 struct mnt_idmap *idmap,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100390 struct dentry *dentry, struct inode *inode,
391 const char *suffix, const void *value,
392 size_t size, int flags)
393{
394 /* Handled by LSM. */
395 return -EAGAIN;
396}
397
398static const struct xattr_handler sockfs_security_xattr_handler = {
399 .prefix = XATTR_SECURITY_PREFIX,
400 .set = sockfs_security_xattr_set,
401};
402
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200403static const struct xattr_handler *sockfs_xattr_handlers[] = {
404 &sockfs_xattr_handler,
Andreas Gruenbacher4a590152016-11-13 21:23:34 +0100405 &sockfs_security_xattr_handler,
Andreas Gruenbacherbba0bd32016-09-29 17:48:35 +0200406 NULL
407};
408
David Howellsfba9be42019-03-25 16:38:24 +0000409static int sockfs_init_fs_context(struct fs_context *fc)
Al Viroc74a1cb2011-01-12 16:59:34 -0500410{
David Howellsfba9be42019-03-25 16:38:24 +0000411 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
412 if (!ctx)
413 return -ENOMEM;
414 ctx->ops = &sockfs_ops;
415 ctx->dops = &sockfs_dentry_operations;
416 ctx->xattr = sockfs_xattr_handlers;
417 return 0;
Al Viroc74a1cb2011-01-12 16:59:34 -0500418}
419
420static struct vfsmount *sock_mnt __read_mostly;
421
422static struct file_system_type sock_fs_type = {
423 .name = "sockfs",
David Howellsfba9be42019-03-25 16:38:24 +0000424 .init_fs_context = sockfs_init_fs_context,
Al Viroc74a1cb2011-01-12 16:59:34 -0500425 .kill_sb = kill_anon_super,
426};
427
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428/*
429 * Obtains the first available file descriptor and sets it up for use.
430 *
David S. Miller39d8c1b2006-03-20 17:13:49 -0800431 * These functions create file structures and maps them to fd space
432 * of the current process. On success it returns file descriptor
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 * and file struct implicitly stored in sock->file.
434 * Note that another thread may close file descriptor before we return
435 * from this function. We use the fact that now we do not refer
436 * to socket after mapping. If one day we will need it, this
437 * function will increment ref. count on file by 1.
438 *
439 * In any case returned fd MAY BE not valid!
440 * This race condition is unavoidable
441 * with shared fd spaces, we cannot solve it inside kernel,
442 * but we take care of internal coherence yet.
443 */
444
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300445/**
446 * sock_alloc_file - Bind a &socket to a &file
447 * @sock: socket
448 * @flags: file status flags
449 * @dname: protocol name
450 *
451 * Returns the &file bound with @sock, implicitly storing it
452 * in sock->file. If dname is %NULL, sets to "".
Thadeu Lima de Souza Cascardo649c15c2023-03-07 14:37:07 -0300453 *
454 * On failure @sock is released, and an ERR pointer is returned.
455 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300456 * This function uses GFP_KERNEL internally.
457 */
458
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700459struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460{
Al Viro7cbe66b2009-08-05 19:59:08 +0400461 struct file *file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800462
Al Virod93aa9d2018-06-09 09:40:05 -0400463 if (!dname)
464 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
Al Virod93aa9d2018-06-09 09:40:05 -0400466 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
467 O_RDWR | (flags & O_NONBLOCK),
468 &socket_file_ops);
Viresh Kumarb5ffe632015-08-12 15:59:47 +0530469 if (IS_ERR(file)) {
Al Viro8e1611e2017-12-05 23:29:09 +0000470 sock_release(sock);
Anatol Pomozov39b652522012-09-12 20:11:55 -0700471 return file;
Al Virocc3808f2009-08-06 09:43:59 +0400472 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800473
474 sock->file = file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800475 file->private_data = sock;
Linus Torvaldsd8e464e2019-11-17 11:20:48 -0800476 stream_open(SOCK_INODE(sock), file);
Al Viro28407632012-08-17 23:54:15 -0400477 return file;
David S. Miller39d8c1b2006-03-20 17:13:49 -0800478}
Al Viro56b31d12012-08-18 00:25:51 -0400479EXPORT_SYMBOL(sock_alloc_file);
David S. Miller39d8c1b2006-03-20 17:13:49 -0800480
Al Viro56b31d12012-08-18 00:25:51 -0400481static int sock_map_fd(struct socket *sock, int flags)
David S. Miller39d8c1b2006-03-20 17:13:49 -0800482{
483 struct file *newfile;
Al Viro28407632012-08-17 23:54:15 -0400484 int fd = get_unused_fd_flags(flags);
Al Viroce4bb042018-01-10 18:47:05 -0500485 if (unlikely(fd < 0)) {
486 sock_release(sock);
Al Viro28407632012-08-17 23:54:15 -0400487 return fd;
Al Viroce4bb042018-01-10 18:47:05 -0500488 }
David S. Miller39d8c1b2006-03-20 17:13:49 -0800489
Linus Torvaldsaab174f2012-10-02 20:25:04 -0700490 newfile = sock_alloc_file(sock, flags, NULL);
Enrico Weigelt4546e442019-06-05 22:58:50 +0200491 if (!IS_ERR(newfile)) {
David S. Miller39d8c1b2006-03-20 17:13:49 -0800492 fd_install(fd, newfile);
Al Viro28407632012-08-17 23:54:15 -0400493 return fd;
494 }
Al Viro7cbe66b2009-08-05 19:59:08 +0400495
Al Viro28407632012-08-17 23:54:15 -0400496 put_unused_fd(fd);
497 return PTR_ERR(newfile);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498}
499
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300500/**
501 * sock_from_file - Return the &socket bounded to @file.
502 * @file: file
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300503 *
Florent Revestdba4a922020-12-04 12:36:04 +0100504 * On failure returns %NULL.
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300505 */
506
Florent Revestdba4a922020-12-04 12:36:04 +0100507struct socket *sock_from_file(struct file *file)
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800508{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800509 if (file->f_op == &socket_file_ops)
Jens Axboeda214a42022-04-12 14:22:39 -0600510 return file->private_data; /* set in sock_alloc_file */
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800511
Eric Dumazet23bb80d2007-02-08 14:59:57 -0800512 return NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800513}
John Fastabend406a3c62012-07-20 10:39:25 +0000514EXPORT_SYMBOL(sock_from_file);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516/**
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700517 * sockfd_lookup - Go from a file number to its socket slot
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 * @fd: file handle
519 * @err: pointer to an error code return
520 *
521 * The file handle passed in is locked and the socket it is bound
Rosen, Rami241c4662017-05-21 22:12:38 +0300522 * to is returned. If an error occurs the err pointer is overwritten
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 * with a negative errno code and NULL is returned. The function checks
524 * for both invalid handles and passing a handle which is not a socket.
525 *
526 * On a success the socket object pointer is returned.
527 */
528
529struct socket *sockfd_lookup(int fd, int *err)
530{
531 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 struct socket *sock;
533
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700534 file = fget(fd);
535 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 *err = -EBADF;
537 return NULL;
538 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700539
Florent Revestdba4a922020-12-04 12:36:04 +0100540 sock = sock_from_file(file);
541 if (!sock) {
542 *err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 fput(file);
Florent Revestdba4a922020-12-04 12:36:04 +0100544 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 return sock;
546}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700547EXPORT_SYMBOL(sockfd_lookup);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800549static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
550{
Al Viro00e188e2014-03-03 23:48:18 -0500551 struct fd f = fdget(fd);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800552 struct socket *sock;
553
Hua Zhong36725582006-04-19 15:25:02 -0700554 *err = -EBADF;
Al Viro00e188e2014-03-03 23:48:18 -0500555 if (f.file) {
Florent Revestdba4a922020-12-04 12:36:04 +0100556 sock = sock_from_file(f.file);
Al Viro00e188e2014-03-03 23:48:18 -0500557 if (likely(sock)) {
Miaohe Lince787a52020-08-06 19:53:16 +0800558 *fput_needed = f.flags & FDPUT_FPUT;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800559 return sock;
Al Viro00e188e2014-03-03 23:48:18 -0500560 }
Florent Revestdba4a922020-12-04 12:36:04 +0100561 *err = -ENOTSOCK;
Al Viro00e188e2014-03-03 23:48:18 -0500562 fdput(f);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -0800563 }
564 return NULL;
565}
566
Masatake YAMATO600e1772012-08-29 10:44:29 +0000567static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
568 size_t size)
569{
570 ssize_t len;
571 ssize_t used = 0;
572
David Howellsc5ef6032015-03-17 22:26:16 +0000573 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
Masatake YAMATO600e1772012-08-29 10:44:29 +0000574 if (len < 0)
575 return len;
576 used += len;
577 if (buffer) {
578 if (size < used)
579 return -ERANGE;
580 buffer += len;
581 }
582
583 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
584 used += len;
585 if (buffer) {
586 if (size < used)
587 return -ERANGE;
588 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
589 buffer += len;
590 }
591
592 return used;
593}
594
Christian Braunerc1632a02023-01-13 12:49:11 +0100595static int sockfs_setattr(struct mnt_idmap *idmap,
Christian Brauner549c7292021-01-21 14:19:43 +0100596 struct dentry *dentry, struct iattr *iattr)
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900597{
Christian Braunerc1632a02023-01-13 12:49:11 +0100598 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900599
Eric Biggerse1a3a602016-12-30 17:42:32 -0600600 if (!err && (iattr->ia_valid & ATTR_UID)) {
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900601 struct socket *sock = SOCKET_I(d_inode(dentry));
602
Cong Wang6d8c50d2018-06-07 13:39:49 -0700603 if (sock->sk)
604 sock->sk->sk_uid = iattr->ia_uid;
605 else
606 err = -ENOENT;
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900607 }
608
609 return err;
610}
611
Masatake YAMATO600e1772012-08-29 10:44:29 +0000612static const struct inode_operations sockfs_inode_ops = {
Masatake YAMATO600e1772012-08-29 10:44:29 +0000613 .listxattr = sockfs_listxattr,
Lorenzo Colitti86741ec2016-11-04 02:23:41 +0900614 .setattr = sockfs_setattr,
Masatake YAMATO600e1772012-08-29 10:44:29 +0000615};
616
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300618 * sock_alloc - allocate a socket
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700619 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 * Allocate a new inode and socket object. The two are bound together
621 * and initialised. The socket is then returned. If we are out of inodes
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300622 * NULL is returned. This functions uses GFP_KERNEL internally.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 */
624
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800625struct socket *sock_alloc(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626{
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700627 struct inode *inode;
628 struct socket *sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
Eric Dumazeta209dfc2011-07-26 11:36:34 +0200630 inode = new_inode_pseudo(sock_mnt->mnt_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 if (!inode)
632 return NULL;
633
634 sock = SOCKET_I(inode);
635
Christoph Hellwig85fe4022010-10-23 11:19:54 -0400636 inode->i_ino = get_next_ino();
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700637 inode->i_mode = S_IFSOCK | S_IRWXUGO;
David Howells8192b0c2008-11-14 10:39:10 +1100638 inode->i_uid = current_fsuid();
639 inode->i_gid = current_fsgid();
Masatake YAMATO600e1772012-08-29 10:44:29 +0000640 inode->i_op = &sockfs_inode_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 return sock;
643}
Tom Herbertf4a00aa2016-03-07 14:11:01 -0800644EXPORT_SYMBOL(sock_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
Cong Wang6d8c50d2018-06-07 13:39:49 -0700646static void __sock_release(struct socket *sock, struct inode *inode)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647{
648 if (sock->ops) {
649 struct module *owner = sock->ops->owner;
650
Cong Wang6d8c50d2018-06-07 13:39:49 -0700651 if (inode)
652 inode_lock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 sock->ops->release(sock);
Eric Biggersff7b11a2019-02-21 14:13:56 -0800654 sock->sk = NULL;
Cong Wang6d8c50d2018-06-07 13:39:49 -0700655 if (inode)
656 inode_unlock(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 sock->ops = NULL;
658 module_put(owner);
659 }
660
Al Viro333f7902019-07-05 20:14:16 +0100661 if (sock->wq.fasync_list)
Yang Yingliang3410f222014-02-12 17:09:55 +0800662 pr_err("%s: fasync list not empty!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 if (!sock->file) {
665 iput(SOCK_INODE(sock));
666 return;
667 }
Stephen Hemminger89bddce2006-09-01 00:19:31 -0700668 sock->file = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669}
Cong Wang6d8c50d2018-06-07 13:39:49 -0700670
Andrew Lunn9a8ad9a2020-07-13 01:15:12 +0200671/**
672 * sock_release - close a socket
673 * @sock: socket to close
674 *
675 * The socket is released from the protocol stack if it has a release
676 * callback, and the inode is then released if the socket is bound to
677 * an inode not a file.
678 */
Cong Wang6d8c50d2018-06-07 13:39:49 -0700679void sock_release(struct socket *sock)
680{
681 __sock_release(sock, NULL);
682}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700683EXPORT_SYMBOL(sock_release);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400685void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
Patrick Ohly20d49472009-02-12 05:03:38 +0000686{
Eric Dumazet140c55d2014-08-06 11:49:29 +0200687 u8 flags = *tx_flags;
688
Gerhard Engleder51eb7492022-05-06 22:01:38 +0200689 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
Eric Dumazet140c55d2014-08-06 11:49:29 +0200690 flags |= SKBTX_HW_TSTAMP;
691
Gerhard Engleder51eb7492022-05-06 22:01:38 +0200692 /* PTP hardware clocks can provide a free running cycle counter
693 * as a time base for virtual clocks. Tell driver to use the
694 * free running cycle counter for timestamp if socket is bound
695 * to virtual clock.
696 */
697 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
698 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
699 }
700
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400701 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200702 flags |= SKBTX_SW_TSTAMP;
703
Soheil Hassas Yeganehc14ac942016-04-02 23:08:12 -0400704 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
Eric Dumazet140c55d2014-08-06 11:49:29 +0200705 flags |= SKBTX_SCHED_TSTAMP;
706
Eric Dumazet140c55d2014-08-06 11:49:29 +0200707 *tx_flags = flags;
Patrick Ohly20d49472009-02-12 05:03:38 +0000708}
Willem de Bruijn67cc0d42014-09-08 19:58:58 -0400709EXPORT_SYMBOL(__sock_tx_timestamp);
Patrick Ohly20d49472009-02-12 05:03:38 +0000710
Paolo Abeni8c3c4472019-05-03 17:01:39 +0200711INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
712 size_t));
Paolo Abenia648a592019-07-03 16:06:54 +0200713INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
714 size_t));
Yunhui Cui6e6eda42023-01-11 14:59:30 +0800715
716static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
717 int flags)
718{
719 trace_sock_send_length(sk, ret, 0);
720}
721
Al Virod8725c82014-12-11 00:02:50 -0500722static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723{
Paolo Abenia648a592019-07-03 16:06:54 +0200724 int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
725 inet_sendmsg, sock, msg,
726 msg_data_left(msg));
Al Virod8725c82014-12-11 00:02:50 -0500727 BUG_ON(ret == -EIOCBQUEUED);
Yunhui Cui6e6eda42023-01-11 14:59:30 +0800728
729 if (trace_sock_send_length_enabled())
730 call_trace_sock_send_length(sock->sk, ret, 0);
Al Virod8725c82014-12-11 00:02:50 -0500731 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732}
Gu Zheng0cf00c62014-12-05 15:14:23 +0800733
Randy Dunlap85806af2019-05-18 21:23:07 -0700734/**
735 * sock_sendmsg - send a message through @sock
736 * @sock: socket
737 * @msg: message to send
738 *
739 * Sends @msg through @sock, passing through LSM.
740 * Returns the number of bytes sent, or an error code.
741 */
Al Virod8725c82014-12-11 00:02:50 -0500742int sock_sendmsg(struct socket *sock, struct msghdr *msg)
Gu Zheng0cf00c62014-12-05 15:14:23 +0800743{
Al Virod8725c82014-12-11 00:02:50 -0500744 int err = security_socket_sendmsg(sock, msg,
Al Viro01e97e62014-12-15 21:39:31 -0500745 msg_data_left(msg));
Ying Xue1b784142015-03-02 15:37:48 +0800746
Al Virod8725c82014-12-11 00:02:50 -0500747 return err ?: sock_sendmsg_nosec(sock, msg);
Gu Zheng0cf00c62014-12-05 15:14:23 +0800748}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700749EXPORT_SYMBOL(sock_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300751/**
752 * kernel_sendmsg - send a message through @sock (kernel-space)
753 * @sock: socket
754 * @msg: message header
755 * @vec: kernel vec
756 * @num: vec array length
757 * @size: total message data size
758 *
759 * Builds the message data with @vec and sends it through @sock.
760 * Returns the number of bytes sent, or an error code.
761 */
762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
764 struct kvec *vec, size_t num, size_t size)
765{
Al Virode4eda92022-09-15 20:25:47 -0400766 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
Al Virod8725c82014-12-11 00:02:50 -0500767 return sock_sendmsg(sock, msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768}
Eric Dumazetc6d409c2010-06-03 20:03:40 -0700769EXPORT_SYMBOL(kernel_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
Pedro Tammela8a3c2452019-03-14 10:45:23 -0300771/**
772 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
773 * @sk: sock
774 * @msg: message header
775 * @vec: output s/g array
776 * @num: output s/g array length
777 * @size: total message data size
778 *
779 * Builds the message data with @vec and sends it through @sock.
780 * Returns the number of bytes sent, or an error code.
781 * Caller must hold @sk.
782 */
783
Tom Herbert306b13e2017-07-28 16:22:41 -0700784int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
785 struct kvec *vec, size_t num, size_t size)
786{
787 struct socket *sock = sk->sk_socket;
788
789 if (!sock->ops->sendmsg_locked)
John Fastabenddb5980d2017-08-15 22:31:34 -0700790 return sock_no_sendmsg_locked(sk, msg, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700791
Al Virode4eda92022-09-15 20:25:47 -0400792 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
Tom Herbert306b13e2017-07-28 16:22:41 -0700793
794 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
795}
796EXPORT_SYMBOL(kernel_sendmsg_locked);
797
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400798static bool skb_is_err_queue(const struct sk_buff *skb)
799{
800 /* pkt_type of skbs enqueued on the error queue are set to
801 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
802 * in recvmsg, since skbs received on a local socket will never
803 * have a pkt_type of PACKET_OUTGOING.
804 */
805 return skb->pkt_type == PACKET_OUTGOING;
806}
807
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200808/* On transmit, software and hardware timestamps are returned independently.
809 * As the two skb clones share the hardware timestamp, which may be updated
810 * before the software timestamp is received, a hardware TX timestamp may be
811 * returned only if there is no software TX timestamp. Ignore false software
812 * timestamps, which may be made in the __sock_recv_timestamp() call when the
Deepa Dinamani7f1bc6e2019-02-02 07:34:46 -0800813 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200814 * hardware timestamp.
815 */
816static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
817{
818 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
819}
820
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200821static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
822{
823 bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC;
824 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
825 struct net_device *orig_dev;
826 ktime_t hwtstamp;
827
828 rcu_read_lock();
829 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
830 if (orig_dev) {
831 *if_index = orig_dev->ifindex;
832 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
833 } else {
834 hwtstamp = shhwtstamps->hwtstamp;
835 }
836 rcu_read_unlock();
837
838 return hwtstamp;
839}
840
841static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
842 int if_index)
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200843{
844 struct scm_ts_pktinfo ts_pktinfo;
845 struct net_device *orig_dev;
846
847 if (!skb_mac_header_was_set(skb))
848 return;
849
850 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
851
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200852 if (!if_index) {
853 rcu_read_lock();
854 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
855 if (orig_dev)
856 if_index = orig_dev->ifindex;
857 rcu_read_unlock();
858 }
859 ts_pktinfo.if_index = if_index;
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200860
861 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
862 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
863 sizeof(ts_pktinfo), &ts_pktinfo);
864}
865
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700866/*
867 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
868 */
869void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
870 struct sk_buff *skb)
871{
Patrick Ohly20d49472009-02-12 05:03:38 +0000872 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
Deepa Dinamani887feae2019-02-02 07:34:50 -0800873 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
Deepa Dinamani97184752019-02-02 07:34:51 -0800874 struct scm_timestamping_internal tss;
875
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200876 int empty = 1, false_tstamp = 0;
Patrick Ohly20d49472009-02-12 05:03:38 +0000877 struct skb_shared_hwtstamps *shhwtstamps =
878 skb_hwtstamps(skb);
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200879 int if_index;
Miroslav Lichvar007747a92022-01-05 11:33:26 +0100880 ktime_t hwtstamp;
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700881
Patrick Ohly20d49472009-02-12 05:03:38 +0000882 /* Race occurred between timestamp enabling and packet
883 receiving. Fill in the current time for now. */
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200884 if (need_software_tstamp && skb->tstamp == 0) {
Patrick Ohly20d49472009-02-12 05:03:38 +0000885 __net_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +0200886 false_tstamp = 1;
887 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000888
889 if (need_software_tstamp) {
890 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800891 if (new_tstamp) {
892 struct __kernel_sock_timeval tv;
893
894 skb_get_new_timestamp(skb, &tv);
895 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
896 sizeof(tv), &tv);
897 } else {
898 struct __kernel_old_timeval tv;
899
900 skb_get_timestamp(skb, &tv);
901 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
902 sizeof(tv), &tv);
903 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000904 } else {
Deepa Dinamani887feae2019-02-02 07:34:50 -0800905 if (new_tstamp) {
906 struct __kernel_timespec ts;
907
908 skb_get_new_timestampns(skb, &ts);
909 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
910 sizeof(ts), &ts);
911 } else {
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +0200912 struct __kernel_old_timespec ts;
Deepa Dinamani887feae2019-02-02 07:34:50 -0800913
914 skb_get_timestampns(skb, &ts);
915 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
916 sizeof(ts), &ts);
917 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000918 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700919 }
Patrick Ohly20d49472009-02-12 05:03:38 +0000920
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400921 memset(&tss, 0, sizeof(tss));
Willem de Bruijnc1991052014-09-03 12:01:18 -0400922 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
Deepa Dinamani97184752019-02-02 07:34:51 -0800923 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
Patrick Ohly20d49472009-02-12 05:03:38 +0000924 empty = 0;
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400925 if (shhwtstamps &&
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400926 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
Yangbo Lud7c08822021-06-30 16:12:00 +0800927 !skb_is_swtx_tstamp(skb, false_tstamp)) {
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200928 if_index = 0;
929 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
930 hwtstamp = get_timestamp(sk, skb, &if_index);
Miroslav Lichvar007747a92022-01-05 11:33:26 +0100931 else
932 hwtstamp = shhwtstamps->hwtstamp;
Yangbo Lud7c08822021-06-30 16:12:00 +0800933
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200934 if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
935 hwtstamp = ptp_convert_timestamp(&hwtstamp,
936 sk->sk_bind_phc);
937
Miroslav Lichvar007747a92022-01-05 11:33:26 +0100938 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
Yangbo Lud7c08822021-06-30 16:12:00 +0800939 empty = 0;
940
941 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
942 !skb_is_err_queue(skb))
Gerhard Engleder97dc7cd2022-05-06 22:01:40 +0200943 put_ts_pktinfo(msg, skb, if_index);
Yangbo Lud7c08822021-06-30 16:12:00 +0800944 }
Miroslav Lichvaraad9c8c2017-05-19 17:52:38 +0200945 }
Francis Yan1c885802016-11-27 23:07:18 -0800946 if (!empty) {
Deepa Dinamani97184752019-02-02 07:34:51 -0800947 if (sock_flag(sk, SOCK_TSTAMP_NEW))
948 put_cmsg_scm_timestamping64(msg, &tss);
949 else
950 put_cmsg_scm_timestamping(msg, &tss);
Francis Yan1c885802016-11-27 23:07:18 -0800951
Soheil Hassas Yeganeh8605330a2017-03-18 17:02:59 -0400952 if (skb_is_err_queue(skb) && skb->len &&
Soheil Hassas Yeganeh4ef1b282017-03-18 17:03:00 -0400953 SKB_EXT_ERR(skb)->opt_stats)
Francis Yan1c885802016-11-27 23:07:18 -0800954 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
955 skb->len, skb->data);
956 }
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700957}
Arnaldo Carvalho de Melo7c81fd82007-03-10 00:39:35 -0300958EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
959
Jakub Kicinskieb6fba72023-04-17 08:53:46 -0700960#ifdef CONFIG_WIRELESS
Johannes Berg6e3e9392011-11-09 10:15:42 +0100961void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
962 struct sk_buff *skb)
963{
964 int ack;
965
966 if (!sock_flag(sk, SOCK_WIFI_STATUS))
967 return;
968 if (!skb->wifi_acked_valid)
969 return;
970
971 ack = skb->wifi_acked;
972
973 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
974}
975EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
Jakub Kicinskieb6fba72023-04-17 08:53:46 -0700976#endif
Johannes Berg6e3e9392011-11-09 10:15:42 +0100977
stephen hemminger11165f12010-10-18 14:27:29 +0000978static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
979 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700980{
Eyal Birger744d5a32015-03-01 14:58:31 +0200981 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
Neil Horman3b885782009-10-12 13:26:31 -0700982 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
Eyal Birger744d5a32015-03-01 14:58:31 +0200983 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
Neil Horman3b885782009-10-12 13:26:31 -0700984}
985
Erin MacNeil6fd1d512022-04-27 16:02:37 -0400986static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
987 struct sk_buff *skb)
988{
Eric Dumazet2558b802023-02-13 16:00:59 +0000989 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
990 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
991 __u32 mark = skb->mark;
992
993 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
994 }
Erin MacNeil6fd1d512022-04-27 16:02:37 -0400995}
996
997void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
998 struct sk_buff *skb)
Neil Horman3b885782009-10-12 13:26:31 -0700999{
1000 sock_recv_timestamp(msg, sk, skb);
1001 sock_recv_drops(msg, sk, skb);
Erin MacNeil6fd1d512022-04-27 16:02:37 -04001002 sock_recv_mark(msg, sk, skb);
Neil Horman3b885782009-10-12 13:26:31 -07001003}
Erin MacNeil6fd1d512022-04-27 16:02:37 -04001004EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
Neil Horman3b885782009-10-12 13:26:31 -07001005
Paolo Abeni8c3c4472019-05-03 17:01:39 +02001006INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
Paolo Abenia648a592019-07-03 16:06:54 +02001007 size_t, int));
1008INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1009 size_t, int));
Yunhui Cui6e6eda42023-01-11 14:59:30 +08001010
1011static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1012{
1013 trace_sock_recv_length(sk, ret, flags);
1014}
1015
Ying Xue1b784142015-03-02 15:37:48 +08001016static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
Al Viro2da62902015-03-14 21:13:46 -04001017 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018{
Yunhui Cui6e6eda42023-01-11 14:59:30 +08001019 int ret = INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
1020 inet_recvmsg, sock, msg,
1021 msg_data_left(msg), flags);
1022 if (trace_sock_recv_length_enabled())
1023 call_trace_sock_recv_length(sock->sk, ret, flags);
1024 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025}
1026
Randy Dunlap85806af2019-05-18 21:23:07 -07001027/**
1028 * sock_recvmsg - receive a message from @sock
1029 * @sock: socket
1030 * @msg: message to receive
1031 * @flags: message flags
1032 *
1033 * Receives @msg from @sock, passing through LSM. Returns the total number
1034 * of bytes received, or an error.
1035 */
Al Viro2da62902015-03-14 21:13:46 -04001036int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001037{
Al Viro2da62902015-03-14 21:13:46 -04001038 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07001039
Al Viro2da62902015-03-14 21:13:46 -04001040 return err ?: sock_recvmsg_nosec(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001042EXPORT_SYMBOL(sock_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
Martin Lucinac1249c02010-12-10 00:04:05 +00001044/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001045 * kernel_recvmsg - Receive a message from a socket (kernel space)
1046 * @sock: The socket to receive the message from
1047 * @msg: Received message
1048 * @vec: Input s/g array for message data
1049 * @num: Size of input s/g array
1050 * @size: Number of bytes to read
1051 * @flags: Message flags (MSG_DONTWAIT, etc...)
Martin Lucinac1249c02010-12-10 00:04:05 +00001052 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001053 * On return the msg structure contains the scatter/gather array passed in the
1054 * vec argument. The array is modified so that it consists of the unfilled
1055 * portion of the original array.
Martin Lucinac1249c02010-12-10 00:04:05 +00001056 *
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001057 * The returned value is the total number of bytes received, or an error.
Martin Lucinac1249c02010-12-10 00:04:05 +00001058 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001059
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001060int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1061 struct kvec *vec, size_t num, size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062{
Christoph Hellwig1f466e12020-05-11 13:59:13 +02001063 msg->msg_control_is_user = false;
Al Virode4eda92022-09-15 20:25:47 -04001064 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
Christoph Hellwig1f466e12020-05-11 13:59:13 +02001065 return sock_recvmsg(sock, msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001067EXPORT_SYMBOL(kernel_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001069static ssize_t sock_sendpage(struct file *file, struct page *page,
1070 int offset, size_t size, loff_t *ppos, int more)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071{
1072 struct socket *sock;
1073 int flags;
Yunhui Cui6e6eda42023-01-11 14:59:30 +08001074 int ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
Eric Dumazetb69aee02005-09-06 14:42:45 -07001076 sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077
Eric Dumazet35f9c092012-04-05 03:05:35 +00001078 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1079 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1080 flags |= more;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081
Yunhui Cui6e6eda42023-01-11 14:59:30 +08001082 ret = kernel_sendpage(sock, page, offset, size, flags);
1083
1084 if (trace_sock_send_length_enabled())
1085 call_trace_sock_send_length(sock->sk, ret, 0);
1086 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087}
1088
Jens Axboe9c55e012007-11-06 23:30:13 -08001089static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001090 struct pipe_inode_info *pipe, size_t len,
Jens Axboe9c55e012007-11-06 23:30:13 -08001091 unsigned int flags)
1092{
1093 struct socket *sock = file->private_data;
1094
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001095 if (unlikely(!sock->ops->splice_read))
Slavomir Kaslev95506582018-11-16 11:27:53 +02001096 return generic_file_splice_read(file, ppos, pipe, len, flags);
Rémi Denis-Courmont997b37d2008-02-15 02:35:45 -08001097
Jens Axboe9c55e012007-11-06 23:30:13 -08001098 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1099}
1100
Al Viro8ae5e0302014-11-28 19:40:50 -05001101static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001102{
Al Viro6d652332015-01-30 16:12:56 -05001103 struct file *file = iocb->ki_filp;
1104 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001105 struct msghdr msg = {.msg_iter = *to,
1106 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001107 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001108
Jens Axboeebfcd892019-12-09 20:58:56 -07001109 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001110 msg.msg_flags = MSG_DONTWAIT;
1111
1112 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001113 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001114
Christoph Hellwig66ee59a2015-02-11 19:56:46 +01001115 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001116 return 0;
1117
Al Viro2da62902015-03-14 21:13:46 -04001118 res = sock_recvmsg(sock, &msg, msg.msg_flags);
Al Viro8ae5e0302014-11-28 19:40:50 -05001119 *to = msg.msg_iter;
1120 return res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001121}
1122
Al Viro8ae5e0302014-11-28 19:40:50 -05001123static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124{
Al Viro6d652332015-01-30 16:12:56 -05001125 struct file *file = iocb->ki_filp;
1126 struct socket *sock = file->private_data;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07001127 struct msghdr msg = {.msg_iter = *from,
1128 .msg_iocb = iocb};
Al Viro8ae5e0302014-11-28 19:40:50 -05001129 ssize_t res;
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001130
Al Viro8ae5e0302014-11-28 19:40:50 -05001131 if (iocb->ki_pos != 0)
Christoph Hellwigce1d4d32005-12-22 21:08:46 -08001132 return -ESPIPE;
Badari Pulavarty027445c2006-09-30 23:28:46 -07001133
Jens Axboeebfcd892019-12-09 20:58:56 -07001134 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
Al Viro8ae5e0302014-11-28 19:40:50 -05001135 msg.msg_flags = MSG_DONTWAIT;
1136
Al Viro6d652332015-01-30 16:12:56 -05001137 if (sock->type == SOCK_SEQPACKET)
1138 msg.msg_flags |= MSG_EOR;
1139
Al Virod8725c82014-12-11 00:02:50 -05001140 res = sock_sendmsg(sock, &msg);
Al Viro8ae5e0302014-11-28 19:40:50 -05001141 *from = msg.msg_iter;
1142 return res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143}
1144
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145/*
1146 * Atomic setting of ioctl hooks to avoid race
1147 * with module unload.
1148 */
1149
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001150static DEFINE_MUTEX(br_ioctl_mutex);
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001151static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1152 unsigned int cmd, struct ifreq *ifr,
1153 void __user *uarg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001155void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1156 unsigned int cmd, struct ifreq *ifr,
1157 void __user *uarg))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001159 mutex_lock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 br_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001161 mutex_unlock(&br_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162}
1163EXPORT_SYMBOL(brioctl_set);
1164
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001165int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1166 struct ifreq *ifr, void __user *uarg)
1167{
1168 int err = -ENOPKG;
1169
1170 if (!br_ioctl_hook)
1171 request_module("bridge");
1172
1173 mutex_lock(&br_ioctl_mutex);
1174 if (br_ioctl_hook)
1175 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1176 mutex_unlock(&br_ioctl_mutex);
1177
1178 return err;
1179}
1180
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001181static DEFINE_MUTEX(vlan_ioctl_mutex);
Eric W. Biederman881d9662007-09-17 11:56:21 -07001182static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Eric W. Biederman881d9662007-09-17 11:56:21 -07001184void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185{
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001186 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 vlan_ioctl_hook = hook;
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001188 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189}
1190EXPORT_SYMBOL(vlan_ioctl_set);
1191
Arnd Bergmann6b960182009-11-06 23:10:54 -08001192static long sock_do_ioctl(struct net *net, struct socket *sock,
Johannes Berg63ff03a2019-01-25 22:43:17 +01001193 unsigned int cmd, unsigned long arg)
Arnd Bergmann6b960182009-11-06 23:10:54 -08001194{
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001195 struct ifreq ifr;
1196 bool need_copyout;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001197 int err;
1198 void __user *argp = (void __user *)arg;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001199 void __user *data;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001200
1201 err = sock->ops->ioctl(sock, cmd, arg);
1202
1203 /*
1204 * If this ioctl is unknown try to hand it down
1205 * to the NIC driver.
1206 */
Al Viro36fd6332017-06-26 13:19:16 -04001207 if (err != -ENOIOCTLCMD)
1208 return err;
Arnd Bergmann6b960182009-11-06 23:10:54 -08001209
Jakub Kicinski29ce8f92021-08-31 09:06:04 -07001210 if (!is_socket_ioctl_cmd(cmd))
1211 return -ENOTTY;
1212
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001213 if (get_user_ifreq(&ifr, &data, argp))
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001214 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001215 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001216 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001217 if (put_user_ifreq(&ifr, argp))
Al Viro36fd6332017-06-26 13:19:16 -04001218 return -EFAULT;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001219
Arnd Bergmann6b960182009-11-06 23:10:54 -08001220 return err;
1221}
1222
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223/*
1224 * With an ioctl, arg may well be a user mode pointer, but we don't know
1225 * what to do with it - that's up to the protocol still.
1226 */
1227
1228static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1229{
1230 struct socket *sock;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001231 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 void __user *argp = (void __user *)arg;
1233 int pid, err;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001234 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
Eric Dumazetb69aee02005-09-06 14:42:45 -07001236 sock = file->private_data;
Eric W. Biederman881d9662007-09-17 11:56:21 -07001237 sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001238 net = sock_net(sk);
Al Viro44c02a22017-10-05 12:59:44 -04001239 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1240 struct ifreq ifr;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001241 void __user *data;
Al Viro44c02a22017-10-05 12:59:44 -04001242 bool need_copyout;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001243 if (get_user_ifreq(&ifr, &data, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001244 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001245 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
Al Viro44c02a22017-10-05 12:59:44 -04001246 if (!err && need_copyout)
Arnd Bergmanna554bf92021-07-27 15:45:12 +02001247 if (put_user_ifreq(&ifr, argp))
Al Viro44c02a22017-10-05 12:59:44 -04001248 return -EFAULT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001250#ifdef CONFIG_WEXT_CORE
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
Al Virob1b0c242017-10-01 20:13:08 -04001252 err = wext_handle_ioctl(net, cmd, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 } else
Johannes Berg3d23e342009-09-29 23:27:28 +02001254#endif
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001255 switch (cmd) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 case FIOSETOWN:
1257 case SIOCSPGRP:
1258 err = -EFAULT;
1259 if (get_user(pid, (int __user *)argp))
1260 break;
Jiri Slaby393cc3f2017-06-13 13:35:50 +02001261 err = f_setown(sock->file, pid, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 break;
1263 case FIOGETOWN:
1264 case SIOCGPGRP:
Eric W. Biederman609d7fa2006-10-02 02:17:15 -07001265 err = put_user(f_getown(sock->file),
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001266 (int __user *)argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 break;
1268 case SIOCGIFBR:
1269 case SIOCSIFBR:
1270 case SIOCBRADDBR:
1271 case SIOCBRDELBR:
Arnd Bergmannad2f99a2021-07-27 15:45:16 +02001272 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 break;
1274 case SIOCGIFVLAN:
1275 case SIOCSIFVLAN:
1276 err = -ENOPKG;
1277 if (!vlan_ioctl_hook)
1278 request_module("8021q");
1279
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001280 mutex_lock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 if (vlan_ioctl_hook)
Eric W. Biederman881d9662007-09-17 11:56:21 -07001282 err = vlan_ioctl_hook(net, argp);
Arjan van de Ven4a3e2f72006-03-20 22:33:17 -08001283 mutex_unlock(&vlan_ioctl_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 break;
Andrey Vaginc62cce22016-10-24 18:29:13 -07001285 case SIOCGSKNS:
1286 err = -EPERM;
1287 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1288 break;
1289
1290 err = open_related_ns(&net->ns, get_net_ns);
1291 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001292 case SIOCGSTAMP_OLD:
1293 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001294 if (!sock->ops->gettstamp) {
1295 err = -ENOIOCTLCMD;
1296 break;
1297 }
1298 err = sock->ops->gettstamp(sock, argp,
Arnd Bergmann0768e172019-04-17 22:56:11 +02001299 cmd == SIOCGSTAMP_OLD,
1300 !IS_ENABLED(CONFIG_64BIT));
Gustavo A. R. Silva60747822019-04-24 10:31:24 -05001301 break;
Arnd Bergmann0768e172019-04-17 22:56:11 +02001302 case SIOCGSTAMP_NEW:
1303 case SIOCGSTAMPNS_NEW:
1304 if (!sock->ops->gettstamp) {
1305 err = -ENOIOCTLCMD;
1306 break;
1307 }
1308 err = sock->ops->gettstamp(sock, argp,
1309 cmd == SIOCGSTAMP_NEW,
1310 false);
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02001311 break;
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02001312
1313 case SIOCGIFCONF:
1314 err = dev_ifconf(net, argp);
1315 break;
1316
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 default:
Johannes Berg63ff03a2019-01-25 22:43:17 +01001318 err = sock_do_ioctl(net, sock, cmd, arg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001320 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 return err;
1322}
1323
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001324/**
1325 * sock_create_lite - creates a socket
1326 * @family: protocol family (AF_INET, ...)
1327 * @type: communication type (SOCK_STREAM, ...)
1328 * @protocol: protocol (0, ...)
1329 * @res: new socket
1330 *
1331 * Creates a new socket and assigns it to @res, passing through LSM.
1332 * The new socket initialization is not complete, see kernel_accept().
1333 * Returns 0 or an error. On failure @res is set to %NULL.
1334 * This function internally uses GFP_KERNEL.
1335 */
1336
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337int sock_create_lite(int family, int type, int protocol, struct socket **res)
1338{
1339 int err;
1340 struct socket *sock = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001341
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 err = security_socket_create(family, type, protocol, 1);
1343 if (err)
1344 goto out;
1345
1346 sock = sock_alloc();
1347 if (!sock) {
1348 err = -ENOMEM;
1349 goto out;
1350 }
1351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 sock->type = type;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001353 err = security_socket_post_create(sock, family, type, protocol, 1);
1354 if (err)
1355 goto out_release;
1356
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357out:
1358 *res = sock;
1359 return err;
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001360out_release:
1361 sock_release(sock);
1362 sock = NULL;
1363 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001365EXPORT_SYMBOL(sock_create_lite);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366
1367/* No kernel lock held - perfect */
Al Viroade994f2017-07-03 00:01:49 -04001368static __poll_t sock_poll(struct file *file, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369{
Christoph Hellwig3cafb372018-01-09 16:07:34 +01001370 struct socket *sock = file->private_data;
Christoph Hellwiga331de32018-07-30 09:42:13 +02001371 __poll_t events = poll_requested_events(wait), flag = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
Christoph Hellwige88958e2018-06-29 15:37:24 +02001373 if (!sock->ops->poll)
1374 return 0;
Christoph Hellwigf641f13b2018-07-30 09:42:12 +02001375
Christoph Hellwiga331de32018-07-30 09:42:13 +02001376 if (sk_can_busy_loop(sock->sk)) {
1377 /* poll once if requested by the syscall */
1378 if (events & POLL_BUSY_LOOP)
1379 sk_busy_loop(sock->sk, 1);
1380
1381 /* if this socket can poll_ll, tell the system call */
1382 flag = POLL_BUSY_LOOP;
1383 }
1384
1385 return sock->ops->poll(file, sock, wait) | flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386}
1387
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001388static int sock_mmap(struct file *file, struct vm_area_struct *vma)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389{
Eric Dumazetb69aee02005-09-06 14:42:45 -07001390 struct socket *sock = file->private_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391
1392 return sock->ops->mmap(file, sock, vma);
1393}
1394
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001395static int sock_close(struct inode *inode, struct file *filp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396{
Cong Wang6d8c50d2018-06-07 13:39:49 -07001397 __sock_release(SOCKET_I(inode), inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 return 0;
1399}
1400
1401/*
1402 * Update the socket async list
1403 *
1404 * Fasync_list locking strategy.
1405 *
1406 * 1. fasync_list is modified only under process context socket lock
1407 * i.e. under semaphore.
1408 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
Eric Dumazet989a2972010-04-14 09:55:35 +00001409 * or under socket lock
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 */
1411
1412static int sock_fasync(int fd, struct file *filp, int on)
1413{
Eric Dumazet989a2972010-04-14 09:55:35 +00001414 struct socket *sock = filp->private_data;
1415 struct sock *sk = sock->sk;
Al Viro333f7902019-07-05 20:14:16 +01001416 struct socket_wq *wq = &sock->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417
Eric Dumazet989a2972010-04-14 09:55:35 +00001418 if (sk == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
1421 lock_sock(sk);
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001422 fasync_helper(fd, filp, on, &wq->fasync_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423
Eric Dumazeteaefd1102011-02-18 03:26:36 +00001424 if (!wq->fasync_list)
Eric Dumazet989a2972010-04-14 09:55:35 +00001425 sock_reset_flag(sk, SOCK_FASYNC);
Jonathan Corbet76398422009-02-01 14:26:59 -07001426 else
Eric Dumazetbcdce712009-10-06 17:28:29 -07001427 sock_set_flag(sk, SOCK_FASYNC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
Eric Dumazet989a2972010-04-14 09:55:35 +00001429 release_sock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 return 0;
1431}
1432
Eric Dumazetceb5d582015-11-29 20:03:11 -08001433/* This function may be called only under rcu_lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
Eric Dumazetceb5d582015-11-29 20:03:11 -08001435int sock_wake_async(struct socket_wq *wq, int how, int band)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436{
Eric Dumazetceb5d582015-11-29 20:03:11 -08001437 if (!wq || !wq->fasync_list)
1438 return -1;
Eric Dumazet43815482010-04-29 11:01:49 +00001439
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001440 switch (how) {
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001441 case SOCK_WAKE_WAITD:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001442 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 break;
1444 goto call_kill;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001445 case SOCK_WAKE_SPACE:
Eric Dumazetceb5d582015-11-29 20:03:11 -08001446 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 break;
Miaohe Lin7c7ab582020-08-08 16:23:10 +08001448 fallthrough;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001449 case SOCK_WAKE_IO:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001450call_kill:
Eric Dumazet43815482010-04-29 11:01:49 +00001451 kill_fasync(&wq->fasync_list, SIGIO, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 break;
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001453 case SOCK_WAKE_URG:
Eric Dumazet43815482010-04-29 11:01:49 +00001454 kill_fasync(&wq->fasync_list, SIGURG, band);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 }
Eric Dumazetceb5d582015-11-29 20:03:11 -08001456
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 return 0;
1458}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001459EXPORT_SYMBOL(sock_wake_async);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001461/**
1462 * __sock_create - creates a socket
1463 * @net: net namespace
1464 * @family: protocol family (AF_INET, ...)
1465 * @type: communication type (SOCK_STREAM, ...)
1466 * @protocol: protocol (0, ...)
1467 * @res: new socket
1468 * @kern: boolean for kernel space sockets
1469 *
1470 * Creates a new socket and assigns it to @res, passing through LSM.
1471 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1472 * be set to true if the socket resides in kernel space.
1473 * This function internally uses GFP_KERNEL.
1474 */
1475
Pavel Emelyanov721db932010-09-29 16:06:32 +04001476int __sock_create(struct net *net, int family, int type, int protocol,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001477 struct socket **res, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478{
1479 int err;
1480 struct socket *sock;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001481 const struct net_proto_family *pf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482
1483 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001484 * Check protocol is in range
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 */
1486 if (family < 0 || family >= NPROTO)
1487 return -EAFNOSUPPORT;
1488 if (type < 0 || type >= SOCK_MAX)
1489 return -EINVAL;
1490
1491 /* Compatibility.
1492
1493 This uglymoron is moved from INET layer to here to avoid
1494 deadlock in module load.
1495 */
1496 if (family == PF_INET && type == SOCK_PACKET) {
liping.zhangf3c98692016-03-11 23:08:36 +08001497 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1498 current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 family = PF_PACKET;
1500 }
1501
1502 err = security_socket_create(family, type, protocol, kern);
1503 if (err)
1504 return err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001505
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001506 /*
1507 * Allocate the socket and allow the family to set things up. if
1508 * the protocol is 0, the family is instructed to select an appropriate
1509 * default.
1510 */
1511 sock = sock_alloc();
1512 if (!sock) {
Joe Perchese87cc472012-05-13 21:56:26 +00001513 net_warn_ratelimited("socket: no more sockets\n");
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001514 return -ENFILE; /* Not exactly a match, but its the
1515 closest posix thing */
1516 }
1517
1518 sock->type = type;
1519
Johannes Berg95a5afc2008-10-16 15:24:51 -07001520#ifdef CONFIG_MODULES
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001521 /* Attempt to load a protocol module if the find failed.
1522 *
1523 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 * requested real, full-featured networking support upon configuration.
1525 * Otherwise module support will break!
1526 */
Eric Dumazet190683a2010-11-10 10:50:44 +00001527 if (rcu_access_pointer(net_families[family]) == NULL)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001528 request_module("net-pf-%d", family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529#endif
1530
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001531 rcu_read_lock();
1532 pf = rcu_dereference(net_families[family]);
1533 err = -EAFNOSUPPORT;
1534 if (!pf)
1535 goto out_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
1537 /*
1538 * We will call the ->create function, that possibly is in a loadable
1539 * module, so we have to bump that loadable module refcnt first.
1540 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001541 if (!try_module_get(pf->owner))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 goto out_release;
1543
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001544 /* Now protected by module ref count */
1545 rcu_read_unlock();
1546
Eric Paris3f378b682009-11-05 22:18:14 -08001547 err = pf->create(net, sock, protocol, kern);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001548 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 goto out_module_put;
Frank Filza79af592005-09-27 15:23:38 -07001550
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 /*
1552 * Now to bump the refcnt of the [loadable] module that owns this
1553 * socket at sock_release time we decrement its refcnt.
1554 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001555 if (!try_module_get(sock->ops->owner))
1556 goto out_module_busy;
1557
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 /*
1559 * Now that we're done with the ->create function, the [loadable]
1560 * module can have its refcnt decremented
1561 */
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001562 module_put(pf->owner);
Venkat Yekkirala7420ed22006-08-04 23:17:57 -07001563 err = security_socket_post_create(sock, family, type, protocol, kern);
1564 if (err)
Herbert Xu3b185522007-08-15 14:46:02 -07001565 goto out_sock_release;
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001566 *res = sock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001568 return 0;
1569
1570out_module_busy:
1571 err = -EAFNOSUPPORT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572out_module_put:
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001573 sock->ops = NULL;
1574 module_put(pf->owner);
1575out_sock_release:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576 sock_release(sock);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07001577 return err;
1578
1579out_release:
1580 rcu_read_unlock();
1581 goto out_sock_release;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582}
Pavel Emelyanov721db932010-09-29 16:06:32 +04001583EXPORT_SYMBOL(__sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001585/**
1586 * sock_create - creates a socket
1587 * @family: protocol family (AF_INET, ...)
1588 * @type: communication type (SOCK_STREAM, ...)
1589 * @protocol: protocol (0, ...)
1590 * @res: new socket
1591 *
1592 * A wrapper around __sock_create().
1593 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1594 */
1595
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596int sock_create(int family, int type, int protocol, struct socket **res)
1597{
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001598 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001600EXPORT_SYMBOL(sock_create);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601
Pedro Tammela8a3c2452019-03-14 10:45:23 -03001602/**
1603 * sock_create_kern - creates a socket (kernel space)
1604 * @net: net namespace
1605 * @family: protocol family (AF_INET, ...)
1606 * @type: communication type (SOCK_STREAM, ...)
1607 * @protocol: protocol (0, ...)
1608 * @res: new socket
1609 *
1610 * A wrapper around __sock_create().
1611 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1612 */
1613
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001614int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615{
Eric W. Biedermaneeb1bd52015-05-08 21:08:05 -05001616 return __sock_create(net, family, type, protocol, res, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001618EXPORT_SYMBOL(sock_create_kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619
Jens Axboeda214a42022-04-12 14:22:39 -06001620static struct socket *__sys_socket_create(int family, int type, int protocol)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 struct socket *sock;
Jens Axboeda214a42022-04-12 14:22:39 -06001623 int retval;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001624
Ulrich Dreppere38b36f2008-07-23 21:29:42 -07001625 /* Check the SOCK_* constants for consistency. */
1626 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1627 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1628 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1629 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1630
Jens Axboeda214a42022-04-12 14:22:39 -06001631 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1632 return ERR_PTR(-EINVAL);
Ulrich Dreppera677a032008-07-23 21:29:17 -07001633 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634
1635 retval = sock_create(family, type, protocol, &sock);
1636 if (retval < 0)
Jens Axboeda214a42022-04-12 14:22:39 -06001637 return ERR_PTR(retval);
1638
1639 return sock;
1640}
1641
1642struct file *__sys_socket_file(int family, int type, int protocol)
1643{
1644 struct socket *sock;
Jens Axboeda214a42022-04-12 14:22:39 -06001645 int flags;
1646
1647 sock = __sys_socket_create(family, type, protocol);
1648 if (IS_ERR(sock))
1649 return ERR_CAST(sock);
1650
1651 flags = type & ~SOCK_TYPE_MASK;
1652 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1653 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1654
Thadeu Lima de Souza Cascardo649c15c2023-03-07 14:37:07 -03001655 return sock_alloc_file(sock, flags, NULL);
Jens Axboeda214a42022-04-12 14:22:39 -06001656}
1657
1658int __sys_socket(int family, int type, int protocol)
1659{
1660 struct socket *sock;
1661 int flags;
1662
1663 sock = __sys_socket_create(family, type, protocol);
1664 if (IS_ERR(sock))
1665 return PTR_ERR(sock);
1666
1667 flags = type & ~SOCK_TYPE_MASK;
1668 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1669 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670
Al Viro8e1611e2017-12-05 23:29:09 +00001671 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672}
1673
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01001674SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1675{
1676 return __sys_socket(family, type, protocol);
1677}
1678
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679/*
1680 * Create a pair of connected sockets.
1681 */
1682
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001683int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684{
1685 struct socket *sock1, *sock2;
1686 int fd1, fd2, err;
Al Virodb349502007-02-07 01:48:00 -05001687 struct file *newfile1, *newfile2;
Ulrich Dreppera677a032008-07-23 21:29:17 -07001688 int flags;
1689
1690 flags = type & ~SOCK_TYPE_MASK;
Ulrich Drepper77d27202008-07-23 21:29:35 -07001691 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
Ulrich Dreppera677a032008-07-23 21:29:17 -07001692 return -EINVAL;
1693 type &= SOCK_TYPE_MASK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001695 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1696 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1697
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 /*
Al Viro016a2662017-12-05 23:28:38 +00001699 * reserve descriptors and make sure we won't fail
1700 * to return them to userland.
1701 */
1702 fd1 = get_unused_fd_flags(flags);
1703 if (unlikely(fd1 < 0))
1704 return fd1;
1705
1706 fd2 = get_unused_fd_flags(flags);
1707 if (unlikely(fd2 < 0)) {
1708 put_unused_fd(fd1);
1709 return fd2;
1710 }
1711
1712 err = put_user(fd1, &usockvec[0]);
1713 if (err)
1714 goto out;
1715
1716 err = put_user(fd2, &usockvec[1]);
1717 if (err)
1718 goto out;
1719
1720 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 * Obtain the first socket and check if the underlying protocol
1722 * supports the socketpair call.
1723 */
1724
1725 err = sock_create(family, type, protocol, &sock1);
Al Viro016a2662017-12-05 23:28:38 +00001726 if (unlikely(err < 0))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 goto out;
1728
1729 err = sock_create(family, type, protocol, &sock2);
Al Viro016a2662017-12-05 23:28:38 +00001730 if (unlikely(err < 0)) {
1731 sock_release(sock1);
1732 goto out;
David S. Millerbf3c23d2007-10-29 21:54:02 -07001733 }
Yann Droneaudd73aa282013-12-09 22:42:20 +01001734
David Herrmannd47cd942018-05-04 16:28:20 +02001735 err = security_socket_socketpair(sock1, sock2);
1736 if (unlikely(err)) {
1737 sock_release(sock2);
1738 sock_release(sock1);
1739 goto out;
1740 }
1741
Al Viro016a2662017-12-05 23:28:38 +00001742 err = sock1->ops->socketpair(sock1, sock2);
1743 if (unlikely(err < 0)) {
1744 sock_release(sock2);
1745 sock_release(sock1);
1746 goto out;
Al Viro28407632012-08-17 23:54:15 -04001747 }
1748
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001749 newfile1 = sock_alloc_file(sock1, flags, NULL);
Viresh Kumarb5ffe632015-08-12 15:59:47 +05301750 if (IS_ERR(newfile1)) {
Al Viro28407632012-08-17 23:54:15 -04001751 err = PTR_ERR(newfile1);
Al Viro016a2662017-12-05 23:28:38 +00001752 sock_release(sock2);
1753 goto out;
Al Viro28407632012-08-17 23:54:15 -04001754 }
1755
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001756 newfile2 = sock_alloc_file(sock2, flags, NULL);
Al Viro28407632012-08-17 23:54:15 -04001757 if (IS_ERR(newfile2)) {
1758 err = PTR_ERR(newfile2);
Al Viro016a2662017-12-05 23:28:38 +00001759 fput(newfile1);
1760 goto out;
Al Virodb349502007-02-07 01:48:00 -05001761 }
1762
Al Viro157cf642008-12-14 04:57:47 -05001763 audit_fd_pair(fd1, fd2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001764
Al Virodb349502007-02-07 01:48:00 -05001765 fd_install(fd1, newfile1);
1766 fd_install(fd2, newfile2);
Yann Droneaudd73aa282013-12-09 22:42:20 +01001767 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769out:
Al Viro016a2662017-12-05 23:28:38 +00001770 put_unused_fd(fd2);
1771 put_unused_fd(fd1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 return err;
1773}
1774
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01001775SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1776 int __user *, usockvec)
1777{
1778 return __sys_socketpair(family, type, protocol, usockvec);
1779}
1780
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781/*
1782 * Bind a name to a socket. Nothing much to do here since it's
1783 * the protocol's responsibility to handle the local address.
1784 *
1785 * We move the socket address to kernel space before we call
1786 * the protocol layer (having also checked the address is ok).
1787 */
1788
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001789int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790{
1791 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001792 struct sockaddr_storage address;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001793 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001795 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001796 if (sock) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001797 err = move_addr_to_kernel(umyaddr, addrlen, &address);
Jakub Sitnicki068b88c2018-10-04 11:09:40 +02001798 if (!err) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001799 err = security_socket_bind(sock,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001800 (struct sockaddr *)&address,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001801 addrlen);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001802 if (!err)
1803 err = sock->ops->bind(sock,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001804 (struct sockaddr *)
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001805 &address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 }
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001807 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001808 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 return err;
1810}
1811
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01001812SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1813{
1814 return __sys_bind(fd, umyaddr, addrlen);
1815}
1816
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817/*
1818 * Perform a listen. Basically, we allow the protocol to do anything
1819 * necessary for a listen, and if that works, we mark the socket as
1820 * ready for listening.
1821 */
1822
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001823int __sys_listen(int fd, int backlog)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824{
1825 struct socket *sock;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001826 int err, fput_needed;
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001827 int somaxconn;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001828
1829 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1830 if (sock) {
Kuniyuki Iwashima3c9ba812022-08-23 10:47:00 -07001831 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
Eric Dumazet95c96172012-04-15 05:58:06 +00001832 if ((unsigned int)backlog > somaxconn)
Pavel Emelyanovb8e1f9b2007-12-08 00:12:33 -08001833 backlog = somaxconn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834
1835 err = security_socket_listen(sock, backlog);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001836 if (!err)
1837 err = sock->ops->listen(sock, backlog);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08001839 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 }
1841 return err;
1842}
1843
Dominik Brodowski25e290e2018-03-13 19:36:54 +01001844SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1845{
1846 return __sys_listen(fd, backlog);
1847}
1848
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001849struct file *do_accept(struct file *file, unsigned file_flags,
Jens Axboede2ea4b2019-10-17 14:41:29 -06001850 struct sockaddr __user *upeer_sockaddr,
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001851 int __user *upeer_addrlen, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852{
1853 struct socket *sock, *newsock;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001854 struct file *newfile;
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001855 int err, len;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001856 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857
Florent Revestdba4a922020-12-04 12:36:04 +01001858 sock = sock_from_file(file);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001859 if (!sock)
1860 return ERR_PTR(-ENOTSOCK);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Eric Dumazetc6d409c2010-06-03 20:03:40 -07001862 newsock = sock_alloc();
1863 if (!newsock)
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001864 return ERR_PTR(-ENFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
1866 newsock->type = sock->type;
1867 newsock->ops = sock->ops;
1868
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 /*
1870 * We don't need try_module_get here, as the listening socket (sock)
1871 * has the protocol module (sock->ops->owner) held.
1872 */
1873 __module_get(newsock->ops->owner);
1874
Linus Torvaldsaab174f2012-10-02 20:25:04 -07001875 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001876 if (IS_ERR(newfile))
1877 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001878
Frank Filza79af592005-09-27 15:23:38 -07001879 err = security_socket_accept(sock, newsock);
1880 if (err)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001881 goto out_fd;
Frank Filza79af592005-09-27 15:23:38 -07001882
Jens Axboede2ea4b2019-10-17 14:41:29 -06001883 err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1884 false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001886 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887
1888 if (upeer_sockaddr) {
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01001889 len = newsock->ops->getname(newsock,
1890 (struct sockaddr *)&address, 2);
1891 if (len < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 err = -ECONNABORTED;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001893 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 }
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00001895 err = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07001896 len, upeer_sockaddr, upeer_addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897 if (err < 0)
David S. Miller39d8c1b2006-03-20 17:13:49 -08001898 goto out_fd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 }
1900
1901 /* File flags are not inherited via accept() unlike another OSes. */
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001902 return newfile;
David S. Miller39d8c1b2006-03-20 17:13:49 -08001903out_fd:
David S. Miller9606a212006-04-01 01:00:14 -08001904 fput(newfile);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001905 return ERR_PTR(err);
1906}
Jens Axboede2ea4b2019-10-17 14:41:29 -06001907
Yajun Dengc0424532022-06-10 17:10:17 +08001908static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1909 int __user *upeer_addrlen, int flags)
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001910{
1911 struct file *newfile;
1912 int newfd;
1913
1914 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1915 return -EINVAL;
1916
1917 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1918 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1919
Yajun Dengc0424532022-06-10 17:10:17 +08001920 newfd = get_unused_fd_flags(flags);
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001921 if (unlikely(newfd < 0))
1922 return newfd;
1923
Yajun Dengc0424532022-06-10 17:10:17 +08001924 newfile = do_accept(file, 0, upeer_sockaddr, upeer_addrlen,
Pavel Begunkovd32f89d2021-08-25 12:25:44 +01001925 flags);
1926 if (IS_ERR(newfile)) {
1927 put_unused_fd(newfd);
1928 return PTR_ERR(newfile);
1929 }
1930 fd_install(newfd, newfile);
1931 return newfd;
Jens Axboede2ea4b2019-10-17 14:41:29 -06001932}
1933
1934/*
1935 * For accept, we attempt to create a new socket, set up the link
1936 * with the client, wake up the client, then return the new
1937 * connected fd. We collect the address of the connector in kernel
1938 * space and move it to user at the very end. This is unclean because
1939 * we open the socket then return an error.
1940 *
1941 * 1003.1g adds the ability to recvmsg() to query connection pending
1942 * status to recvmsg. We need to add that support in a way thats
1943 * clean when we restructure accept also.
1944 */
1945
1946int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1947 int __user *upeer_addrlen, int flags)
1948{
1949 int ret = -EBADF;
1950 struct fd f;
1951
1952 f = fdget(fd);
1953 if (f.file) {
Yajun Dengc0424532022-06-10 17:10:17 +08001954 ret = __sys_accept4_file(f.file, upeer_sockaddr,
1955 upeer_addrlen, flags);
Miaohe Lin6b07ede2020-08-06 19:52:24 +08001956 fdput(f);
Jens Axboede2ea4b2019-10-17 14:41:29 -06001957 }
1958
1959 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960}
1961
Dominik Brodowski4541e802018-03-13 19:24:23 +01001962SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1963 int __user *, upeer_addrlen, int, flags)
1964{
1965 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1966}
1967
Heiko Carstens20f37032009-01-14 14:14:23 +01001968SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1969 int __user *, upeer_addrlen)
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001970{
Dominik Brodowski4541e802018-03-13 19:24:23 +01001971 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07001972}
1973
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974/*
1975 * Attempt to connect to a socket with the server address. The address
1976 * is in user space so we verify it is OK and move it to kernel space.
1977 *
1978 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1979 * break bindings
1980 *
1981 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1982 * other SEQPACKET protocols that take time to connect() as it doesn't
1983 * include the -EINPROGRESS status for such sockets.
1984 */
1985
Jens Axboef499a022019-12-02 16:28:46 -07001986int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
Jens Axboebd3ded32019-11-23 14:17:16 -07001987 int addrlen, int file_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988{
1989 struct socket *sock;
Jens Axboebd3ded32019-11-23 14:17:16 -07001990 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991
Florent Revestdba4a922020-12-04 12:36:04 +01001992 sock = sock_from_file(file);
1993 if (!sock) {
1994 err = -ENOTSOCK;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995 goto out;
Florent Revestdba4a922020-12-04 12:36:04 +01001996 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997
Stephen Hemminger89bddce2006-09-01 00:19:31 -07001998 err =
Jens Axboef499a022019-12-02 16:28:46 -07001999 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 if (err)
Jens Axboebd3ded32019-11-23 14:17:16 -07002001 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002
Jens Axboef499a022019-12-02 16:28:46 -07002003 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
Jens Axboebd3ded32019-11-23 14:17:16 -07002004 sock->file->f_flags | file_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002005out:
2006 return err;
2007}
2008
Jens Axboebd3ded32019-11-23 14:17:16 -07002009int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2010{
2011 int ret = -EBADF;
2012 struct fd f;
2013
2014 f = fdget(fd);
2015 if (f.file) {
Jens Axboef499a022019-12-02 16:28:46 -07002016 struct sockaddr_storage address;
2017
2018 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2019 if (!ret)
2020 ret = __sys_connect_file(f.file, &address, addrlen, 0);
Miaohe Lin6b07ede2020-08-06 19:52:24 +08002021 fdput(f);
Jens Axboebd3ded32019-11-23 14:17:16 -07002022 }
2023
2024 return ret;
2025}
2026
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01002027SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2028 int, addrlen)
2029{
2030 return __sys_connect(fd, uservaddr, addrlen);
2031}
2032
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033/*
2034 * Get the local address ('name') of a socket object. Move the obtained
2035 * name to user space.
2036 */
2037
Dominik Brodowski8882a102018-03-13 19:43:14 +01002038int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2039 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040{
2041 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002042 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002043 int err, fput_needed;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002044
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002045 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 if (!sock)
2047 goto out;
2048
2049 err = security_socket_getsockname(sock);
2050 if (err)
2051 goto out_put;
2052
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002053 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
2054 if (err < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055 goto out_put;
Hamish MacDonalde44ef1d2022-01-01 22:38:25 +10002056 /* "err" is actually length in this case */
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002057 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058
2059out_put:
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002060 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061out:
2062 return err;
2063}
2064
Dominik Brodowski8882a102018-03-13 19:43:14 +01002065SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2066 int __user *, usockaddr_len)
2067{
2068 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2069}
2070
Linus Torvalds1da177e2005-04-16 15:20:36 -07002071/*
2072 * Get the remote address ('name') of a socket object. Move the obtained
2073 * name to user space.
2074 */
2075
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002076int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2077 int __user *usockaddr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078{
2079 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002080 struct sockaddr_storage address;
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002081 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002083 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2084 if (sock != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 err = security_socket_getpeername(sock);
2086 if (err) {
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002087 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 return err;
2089 }
2090
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01002091 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
2092 if (err >= 0)
2093 /* "err" is actually length in this case */
2094 err = move_addr_to_user(&address, err, usockaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002095 usockaddr_len);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002096 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 }
2098 return err;
2099}
2100
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01002101SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2102 int __user *, usockaddr_len)
2103{
2104 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2105}
2106
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107/*
2108 * Send a datagram to a given address. We move the address into kernel
2109 * space and check the user space data area is readable before invoking
2110 * the protocol.
2111 */
Dominik Brodowski211b6342018-03-13 19:18:52 +01002112int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2113 struct sockaddr __user *addr, int addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114{
2115 struct socket *sock;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002116 struct sockaddr_storage address;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 int err;
2118 struct msghdr msg;
2119 struct iovec iov;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002120 int fput_needed;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002121
Al Virode4eda92022-09-15 20:25:47 -04002122 err = import_single_range(ITER_SOURCE, buff, len, &iov, &msg.msg_iter);
Al Viro602bd0e2015-03-21 19:12:32 -04002123 if (unlikely(err))
2124 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002125 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2126 if (!sock)
David S. Miller4387ff752007-02-08 15:06:08 -08002127 goto out;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002128
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002129 msg.msg_name = NULL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002130 msg.msg_control = NULL;
2131 msg.msg_controllen = 0;
2132 msg.msg_namelen = 0;
Pavel Begunkov7c701d92022-07-12 21:52:29 +01002133 msg.msg_ubuf = NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002134 if (addr) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002135 err = move_addr_to_kernel(addr, addr_len, &address);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136 if (err < 0)
2137 goto out_put;
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002138 msg.msg_name = (struct sockaddr *)&address;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002139 msg.msg_namelen = addr_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140 }
2141 if (sock->file->f_flags & O_NONBLOCK)
2142 flags |= MSG_DONTWAIT;
2143 msg.msg_flags = flags;
Al Virod8725c82014-12-11 00:02:50 -05002144 err = sock_sendmsg(sock, &msg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002146out_put:
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002147 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08002148out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149 return err;
2150}
2151
Dominik Brodowski211b6342018-03-13 19:18:52 +01002152SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2153 unsigned int, flags, struct sockaddr __user *, addr,
2154 int, addr_len)
2155{
2156 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2157}
2158
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002160 * Send a datagram down a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161 */
2162
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002163SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
Eric Dumazet95c96172012-04-15 05:58:06 +00002164 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165{
Dominik Brodowski211b6342018-03-13 19:18:52 +01002166 return __sys_sendto(fd, buff, len, flags, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167}
2168
2169/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002170 * Receive a frame from the socket and optionally record the address of the
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171 * sender. We verify the buffers are writable and if needed move the
2172 * sender address from kernel to user space.
2173 */
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002174int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2175 struct sockaddr __user *addr, int __user *addr_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176{
Eric Dumazet1228b342022-06-22 15:02:20 +00002177 struct sockaddr_storage address;
2178 struct msghdr msg = {
2179 /* Save some cycles and don't copy the address if not needed */
2180 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2181 };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182 struct socket *sock;
2183 struct iovec iov;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002184 int err, err2;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002185 int fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186
Al Virode4eda92022-09-15 20:25:47 -04002187 err = import_single_range(ITER_DEST, ubuf, size, &iov, &msg.msg_iter);
Al Viro602bd0e2015-03-21 19:12:32 -04002188 if (unlikely(err))
2189 return err;
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002190 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191 if (!sock)
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002192 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 if (sock->file->f_flags & O_NONBLOCK)
2195 flags |= MSG_DONTWAIT;
Al Viro2da62902015-03-14 21:13:46 -04002196 err = sock_recvmsg(sock, &msg, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002198 if (err >= 0 && addr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002199 err2 = move_addr_to_user(&address,
YOSHIFUJI Hideaki230b1832008-07-19 22:35:47 -07002200 msg.msg_namelen, addr, addr_len);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002201 if (err2 < 0)
2202 err = err2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203 }
Pavel Emelyanovde0fa952007-11-14 16:01:43 -08002204
2205 fput_light(sock->file, fput_needed);
David S. Miller4387ff752007-02-08 15:06:08 -08002206out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207 return err;
2208}
2209
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002210SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2211 unsigned int, flags, struct sockaddr __user *, addr,
2212 int __user *, addr_len)
2213{
2214 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2215}
2216
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002218 * Receive a datagram from a socket.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219 */
2220
Jan Glauberb7c0ddf2014-04-16 09:32:48 +02002221SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2222 unsigned int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223{
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01002224 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225}
2226
Florian Westphal83f0c102020-07-05 01:30:15 +02002227static bool sock_use_custom_sol_socket(const struct socket *sock)
2228{
Paolo Abenia5ef0582022-10-20 19:48:51 +02002229 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
Florian Westphal83f0c102020-07-05 01:30:15 +02002230}
2231
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232/*
2233 * Set a socket option. Because we don't know the option lengths we have
2234 * to pass the user mode parameter for the protocols to sort out.
2235 */
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002236int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002237 int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238{
Christoph Hellwig519a8a62020-08-10 18:42:14 +02002239 sockptr_t optval = USER_SOCKPTR(user_optval);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002240 char *kernel_optval = NULL;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002241 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 struct socket *sock;
2243
2244 if (optlen < 0)
2245 return -EINVAL;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002246
2247 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002248 if (!sock)
2249 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250
Christoph Hellwig4a367292020-07-17 08:23:11 +02002251 err = security_socket_setsockopt(sock, level, optname);
2252 if (err)
2253 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002254
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002255 if (!in_compat_syscall())
2256 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002257 user_optval, &optlen,
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002258 &kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002259 if (err < 0)
2260 goto out_put;
2261 if (err > 0) {
2262 err = 0;
2263 goto out_put;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002264 }
Christoph Hellwig4a367292020-07-17 08:23:11 +02002265
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002266 if (kernel_optval)
2267 optval = KERNEL_SOCKPTR(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002268 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002269 err = sock_setsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002270 else if (unlikely(!sock->ops->setsockopt))
2271 err = -EOPNOTSUPP;
Christoph Hellwig4a367292020-07-17 08:23:11 +02002272 else
2273 err = sock->ops->setsockopt(sock, level, optname, optval,
2274 optlen);
Christoph Hellwiga7b75c52020-07-23 08:09:07 +02002275 kfree(kernel_optval);
Christoph Hellwig4a367292020-07-17 08:23:11 +02002276out_put:
2277 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278 return err;
2279}
2280
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01002281SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2282 char __user *, optval, int, optlen)
2283{
2284 return __sys_setsockopt(fd, level, optname, optval, optlen);
2285}
2286
Stanislav Fomichev9cacf812021-01-15 08:34:59 -08002287INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2288 int optname));
2289
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290/*
2291 * Get a socket option. Because we don't know the option lengths we have
2292 * to pass a user mode parameter for the protocols to sort out.
2293 */
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002294int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2295 int __user *optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296{
Vincenzo Palazzoad4bf5f2023-03-10 23:18:51 +01002297 int max_optlen __maybe_unused;
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002298 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299 struct socket *sock;
2300
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002301 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002302 if (!sock)
2303 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002305 err = security_socket_getsockopt(sock, level, optname);
2306 if (err)
2307 goto out_put;
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002308
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002309 if (!in_compat_syscall())
2310 max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002311
2312 if (level == SOL_SOCKET)
2313 err = sock_getsockopt(sock, level, optname, optval, optlen);
Christoph Hellwiga44d9e72020-07-17 08:23:31 +02002314 else if (unlikely(!sock->ops->getsockopt))
2315 err = -EOPNOTSUPP;
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002316 else
2317 err = sock->ops->getsockopt(sock, level, optname, optval,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002318 optlen);
Stanislav Fomichev0d01da62019-06-27 13:38:47 -07002319
Christoph Hellwig55db9c02020-07-17 08:23:15 +02002320 if (!in_compat_syscall())
2321 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2322 optval, optlen, max_optlen,
2323 err);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002324out_put:
Christoph Hellwigd8a9b382020-07-17 08:23:12 +02002325 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 return err;
2327}
2328
Dominik Brodowski13a2d702018-03-13 20:15:04 +01002329SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2330 char __user *, optval, int __user *, optlen)
2331{
2332 return __sys_getsockopt(fd, level, optname, optval, optlen);
2333}
2334
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335/*
2336 * Shutdown a socket.
2337 */
2338
Jens Axboeb713c192020-09-05 11:13:35 -06002339int __sys_shutdown_sock(struct socket *sock, int how)
2340{
2341 int err;
2342
2343 err = security_socket_shutdown(sock, how);
2344 if (!err)
2345 err = sock->ops->shutdown(sock, how);
2346
2347 return err;
2348}
2349
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002350int __sys_shutdown(int fd, int how)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351{
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002352 int err, fput_needed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353 struct socket *sock;
2354
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002355 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2356 if (sock != NULL) {
Jens Axboeb713c192020-09-05 11:13:35 -06002357 err = __sys_shutdown_sock(sock, how);
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002358 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359 }
2360 return err;
2361}
2362
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01002363SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2364{
2365 return __sys_shutdown(fd, how);
2366}
2367
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002368/* A couple of helpful macros for getting the address of the 32/64 bit
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 * fields which are the same type (int / unsigned) on our platforms.
2370 */
2371#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2372#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2373#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2374
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002375struct used_address {
2376 struct sockaddr_storage name;
2377 unsigned int name_len;
2378};
2379
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002380int __copy_msghdr(struct msghdr *kmsg,
2381 struct user_msghdr *msg,
2382 struct sockaddr __user **save_addr)
Dan Carpenter1661bf32013-10-03 00:27:20 +03002383{
Al Viro08adb7d2014-11-10 20:23:13 -05002384 ssize_t err;
2385
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002386 kmsg->msg_control_is_user = true;
Eric Dumazet1228b342022-06-22 15:02:20 +00002387 kmsg->msg_get_inq = 0;
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002388 kmsg->msg_control_user = msg->msg_control;
2389 kmsg->msg_controllen = msg->msg_controllen;
2390 kmsg->msg_flags = msg->msg_flags;
Al Viroffb07552017-06-27 19:32:04 -04002391
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002392 kmsg->msg_namelen = msg->msg_namelen;
2393 if (!msg->msg_name)
Ani Sinha6a2a2b32014-09-08 14:49:59 -07002394 kmsg->msg_namelen = 0;
2395
Matthew Leachdbb490b2014-03-11 11:58:27 +00002396 if (kmsg->msg_namelen < 0)
2397 return -EINVAL;
2398
Dan Carpenter1661bf32013-10-03 00:27:20 +03002399 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
Dan Carpenterdb31c552013-11-27 15:40:21 +03002400 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
Al Viro08adb7d2014-11-10 20:23:13 -05002401
2402 if (save_addr)
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002403 *save_addr = msg->msg_name;
Al Viro08adb7d2014-11-10 20:23:13 -05002404
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002405 if (msg->msg_name && kmsg->msg_namelen) {
Al Viro08adb7d2014-11-10 20:23:13 -05002406 if (!save_addr) {
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002407 err = move_addr_to_kernel(msg->msg_name,
Paolo Abeni864d9662017-07-21 18:49:45 +02002408 kmsg->msg_namelen,
Al Viro08adb7d2014-11-10 20:23:13 -05002409 kmsg->msg_name);
2410 if (err < 0)
2411 return err;
2412 }
2413 } else {
2414 kmsg->msg_name = NULL;
2415 kmsg->msg_namelen = 0;
2416 }
2417
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002418 if (msg->msg_iovlen > UIO_MAXIOV)
Al Viro08adb7d2014-11-10 20:23:13 -05002419 return -EMSGSIZE;
2420
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002421 kmsg->msg_iocb = NULL;
Pavel Begunkov7c701d92022-07-12 21:52:29 +01002422 kmsg->msg_ubuf = NULL;
Jens Axboe0a384ab2020-02-27 08:11:20 -07002423 return 0;
2424}
2425
2426static int copy_msghdr_from_user(struct msghdr *kmsg,
2427 struct user_msghdr __user *umsg,
2428 struct sockaddr __user **save_addr,
2429 struct iovec **iov)
2430{
2431 struct user_msghdr msg;
2432 ssize_t err;
2433
Dylan Yudaken7fa875b2022-07-14 04:02:56 -07002434 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2435 return -EFAULT;
2436
2437 err = __copy_msghdr(kmsg, &msg, save_addr);
Jens Axboe0a384ab2020-02-27 08:11:20 -07002438 if (err)
2439 return err;
tadeusz.struk@intel.com0345f932015-03-19 12:31:25 -07002440
Al Virode4eda92022-09-15 20:25:47 -04002441 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
Al Viroffb07552017-06-27 19:32:04 -04002442 msg.msg_iov, msg.msg_iovlen,
Al Viroda184282015-03-21 19:29:06 -04002443 UIO_FASTIOV, iov, &kmsg->msg_iter);
Jens Axboe87e5e6d2019-05-14 16:02:22 -06002444 return err < 0 ? err : 0;
Dan Carpenter1661bf32013-10-03 00:27:20 +03002445}
2446
Jens Axboe4257c8c2019-11-25 14:27:34 -07002447static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2448 unsigned int flags, struct used_address *used_address,
2449 unsigned int allowed_msghdr_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450{
Alex Williamsonb9d717a2005-09-26 14:28:02 -07002451 unsigned char ctl[sizeof(struct cmsghdr) + 20]
Amit Kushwaha846cc122016-12-08 18:21:53 +05302452 __aligned(sizeof(__kernel_size_t));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002453 /* 20 is size of ipv6_pktinfo */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454 unsigned char *ctl_buf = ctl;
Al Virod8725c82014-12-11 00:02:50 -05002455 int ctl_len;
Al Viro08adb7d2014-11-10 20:23:13 -05002456 ssize_t err;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002457
Linus Torvalds1da177e2005-04-16 15:20:36 -07002458 err = -ENOBUFS;
2459
Anton Blanchard228e5482011-05-02 20:21:35 +00002460 if (msg_sys->msg_controllen > INT_MAX)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002461 goto out;
Tom Herbert28a94d82016-03-07 14:11:02 -08002462 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
Anton Blanchard228e5482011-05-02 20:21:35 +00002463 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002464 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002465 err =
Anton Blanchard228e5482011-05-02 20:21:35 +00002466 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002467 sizeof(ctl));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002469 goto out;
Anton Blanchard228e5482011-05-02 20:21:35 +00002470 ctl_buf = msg_sys->msg_control;
2471 ctl_len = msg_sys->msg_controllen;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472 } else if (ctl_len) {
David S. Millerac4340f2017-01-04 13:24:19 -05002473 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2474 CMSG_ALIGN(sizeof(struct cmsghdr)));
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002475 if (ctl_len > sizeof(ctl)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002476 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002477 if (ctl_buf == NULL)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002478 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479 }
2480 err = -EFAULT;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002481 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 goto out_freectl;
Anton Blanchard228e5482011-05-02 20:21:35 +00002483 msg_sys->msg_control = ctl_buf;
Christoph Hellwig1f466e12020-05-11 13:59:13 +02002484 msg_sys->msg_control_is_user = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002485 }
Anton Blanchard228e5482011-05-02 20:21:35 +00002486 msg_sys->msg_flags = flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487
2488 if (sock->file->f_flags & O_NONBLOCK)
Anton Blanchard228e5482011-05-02 20:21:35 +00002489 msg_sys->msg_flags |= MSG_DONTWAIT;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002490 /*
2491 * If this is sendmmsg() and current destination address is same as
2492 * previously succeeded address, omit asking LSM's decision.
2493 * used_address->name_len is initialized to UINT_MAX so that the first
2494 * destination address never matches.
2495 */
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002496 if (used_address && msg_sys->msg_name &&
2497 used_address->name_len == msg_sys->msg_namelen &&
2498 !memcmp(&used_address->name, msg_sys->msg_name,
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002499 used_address->name_len)) {
Al Virod8725c82014-12-11 00:02:50 -05002500 err = sock_sendmsg_nosec(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002501 goto out_freectl;
2502 }
Al Virod8725c82014-12-11 00:02:50 -05002503 err = sock_sendmsg(sock, msg_sys);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002504 /*
2505 * If this is sendmmsg() and sending to current destination address was
2506 * successful, remember it.
2507 */
2508 if (used_address && err >= 0) {
2509 used_address->name_len = msg_sys->msg_namelen;
Mathieu Desnoyersbc909d92011-08-24 19:45:03 -07002510 if (msg_sys->msg_name)
2511 memcpy(&used_address->name, msg_sys->msg_name,
2512 used_address->name_len);
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002513 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514
2515out_freectl:
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002516 if (ctl_buf != ctl)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002517 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
Jens Axboe4257c8c2019-11-25 14:27:34 -07002518out:
2519 return err;
2520}
2521
Jens Axboe03b12302019-12-02 18:50:25 -07002522int sendmsg_copy_msghdr(struct msghdr *msg,
2523 struct user_msghdr __user *umsg, unsigned flags,
2524 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002525{
2526 int err;
2527
2528 if (flags & MSG_CMSG_COMPAT) {
2529 struct compat_msghdr __user *msg_compat;
2530
2531 msg_compat = (struct compat_msghdr __user *) umsg;
2532 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2533 } else {
2534 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2535 }
2536 if (err < 0)
2537 return err;
2538
2539 return 0;
2540}
2541
2542static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2543 struct msghdr *msg_sys, unsigned int flags,
2544 struct used_address *used_address,
2545 unsigned int allowed_msghdr_flags)
2546{
2547 struct sockaddr_storage address;
2548 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2549 ssize_t err;
2550
2551 msg_sys->msg_name = &address;
2552
2553 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2554 if (err < 0)
2555 return err;
2556
2557 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2558 allowed_msghdr_flags);
Al Viroda184282015-03-21 19:29:06 -04002559 kfree(iov);
Anton Blanchard228e5482011-05-02 20:21:35 +00002560 return err;
2561}
2562
2563/*
2564 * BSD sendmsg interface
2565 */
Jens Axboe03b12302019-12-02 18:50:25 -07002566long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
Jens Axboe0fa03c62019-04-19 13:34:07 -06002567 unsigned int flags)
2568{
Jens Axboe03b12302019-12-02 18:50:25 -07002569 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
Jens Axboe0fa03c62019-04-19 13:34:07 -06002570}
Anton Blanchard228e5482011-05-02 20:21:35 +00002571
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002572long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2573 bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002574{
2575 int fput_needed, err;
2576 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002577 struct socket *sock;
Anton Blanchard228e5482011-05-02 20:21:35 +00002578
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002579 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2580 return -EINVAL;
2581
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002582 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Anton Blanchard228e5482011-05-02 20:21:35 +00002583 if (!sock)
2584 goto out;
2585
Tom Herbert28a94d82016-03-07 14:11:02 -08002586 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
Anton Blanchard228e5482011-05-02 20:21:35 +00002587
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002588 fput_light(sock->file, fput_needed);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002589out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590 return err;
2591}
2592
Al Viro666547f2014-04-06 14:03:05 -04002593SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002594{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002595 return __sys_sendmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002596}
2597
Anton Blanchard228e5482011-05-02 20:21:35 +00002598/*
2599 * Linux sendmmsg interface
2600 */
2601
2602int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002603 unsigned int flags, bool forbid_cmsg_compat)
Anton Blanchard228e5482011-05-02 20:21:35 +00002604{
2605 int fput_needed, err, datagrams;
2606 struct socket *sock;
2607 struct mmsghdr __user *entry;
2608 struct compat_mmsghdr __user *compat_entry;
2609 struct msghdr msg_sys;
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002610 struct used_address used_address;
Tom Herbertf0922762016-03-07 14:11:03 -08002611 unsigned int oflags = flags;
Anton Blanchard228e5482011-05-02 20:21:35 +00002612
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002613 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2614 return -EINVAL;
2615
Anton Blanchard98382f42011-08-04 14:07:39 +00002616 if (vlen > UIO_MAXIOV)
2617 vlen = UIO_MAXIOV;
Anton Blanchard228e5482011-05-02 20:21:35 +00002618
2619 datagrams = 0;
2620
2621 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2622 if (!sock)
2623 return err;
2624
Tetsuo Handac71d8eb2011-08-04 14:07:40 +00002625 used_address.name_len = UINT_MAX;
Anton Blanchard228e5482011-05-02 20:21:35 +00002626 entry = mmsg;
2627 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Anton Blanchard728ffb82011-08-04 14:07:38 +00002628 err = 0;
Tom Herbertf0922762016-03-07 14:11:03 -08002629 flags |= MSG_BATCH;
Anton Blanchard228e5482011-05-02 20:21:35 +00002630
2631 while (datagrams < vlen) {
Tom Herbertf0922762016-03-07 14:11:03 -08002632 if (datagrams == vlen - 1)
2633 flags = oflags;
2634
Anton Blanchard228e5482011-05-02 20:21:35 +00002635 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002636 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002637 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002638 if (err < 0)
2639 break;
2640 err = __put_user(err, &compat_entry->msg_len);
2641 ++compat_entry;
2642 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002643 err = ___sys_sendmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002644 (struct user_msghdr __user *)entry,
Tom Herbert28a94d82016-03-07 14:11:02 -08002645 &msg_sys, flags, &used_address, MSG_EOR);
Anton Blanchard228e5482011-05-02 20:21:35 +00002646 if (err < 0)
2647 break;
2648 err = put_user(err, &entry->msg_len);
2649 ++entry;
2650 }
2651
2652 if (err)
2653 break;
2654 ++datagrams;
Soheil Hassas Yeganeh30238982016-11-04 15:36:49 -04002655 if (msg_data_left(&msg_sys))
2656 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002657 cond_resched();
Anton Blanchard228e5482011-05-02 20:21:35 +00002658 }
2659
Anton Blanchard228e5482011-05-02 20:21:35 +00002660 fput_light(sock->file, fput_needed);
2661
Anton Blanchard728ffb82011-08-04 14:07:38 +00002662 /* We only return an error if no datagrams were able to be sent */
2663 if (datagrams != 0)
Anton Blanchard228e5482011-05-02 20:21:35 +00002664 return datagrams;
2665
Anton Blanchard228e5482011-05-02 20:21:35 +00002666 return err;
2667}
2668
2669SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2670 unsigned int, vlen, unsigned int, flags)
2671{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002672 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
Anton Blanchard228e5482011-05-02 20:21:35 +00002673}
2674
Jens Axboe03b12302019-12-02 18:50:25 -07002675int recvmsg_copy_msghdr(struct msghdr *msg,
2676 struct user_msghdr __user *umsg, unsigned flags,
2677 struct sockaddr __user **uaddr,
2678 struct iovec **iov)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002679{
2680 ssize_t err;
2681
2682 if (MSG_CMSG_COMPAT & flags) {
2683 struct compat_msghdr __user *msg_compat;
2684
2685 msg_compat = (struct compat_msghdr __user *) umsg;
2686 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2687 } else {
2688 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2689 }
2690 if (err < 0)
2691 return err;
2692
2693 return 0;
2694}
2695
2696static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2697 struct user_msghdr __user *msg,
2698 struct sockaddr __user *uaddr,
2699 unsigned int flags, int nosec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002700{
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002701 struct compat_msghdr __user *msg_compat =
Jens Axboe4257c8c2019-11-25 14:27:34 -07002702 (struct compat_msghdr __user *) msg;
2703 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2704 struct sockaddr_storage addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002705 unsigned long cmsg_ptr;
Al Viro2da62902015-03-14 21:13:46 -04002706 int len;
Al Viro08adb7d2014-11-10 20:23:13 -05002707 ssize_t err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002708
Al Viro08adb7d2014-11-10 20:23:13 -05002709 msg_sys->msg_name = &addr;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002710 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2711 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002712
Hannes Frederic Sowaf3d33422013-11-21 03:14:22 +01002713 /* We assume all kernel code knows the size of sockaddr_storage */
2714 msg_sys->msg_namelen = 0;
2715
Linus Torvalds1da177e2005-04-16 15:20:36 -07002716 if (sock->file->f_flags & O_NONBLOCK)
2717 flags |= MSG_DONTWAIT;
Eric Dumazet1af66222019-12-06 09:38:36 -08002718
2719 if (unlikely(nosec))
2720 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2721 else
2722 err = sock_recvmsg(sock, msg_sys, flags);
2723
Linus Torvalds1da177e2005-04-16 15:20:36 -07002724 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002725 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726 len = err;
2727
2728 if (uaddr != NULL) {
Maciej Żenczykowski43db3622012-03-11 12:51:50 +00002729 err = move_addr_to_user(&addr,
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002730 msg_sys->msg_namelen, uaddr,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002731 uaddr_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732 if (err < 0)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002733 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002735 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
David S. Miller37f7f422005-09-16 16:51:01 -07002736 COMPAT_FLAGS(msg));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002737 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002738 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002739 if (MSG_CMSG_COMPAT & flags)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002740 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741 &msg_compat->msg_controllen);
2742 else
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002743 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002744 &msg->msg_controllen);
2745 if (err)
Jens Axboe4257c8c2019-11-25 14:27:34 -07002746 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747 err = len;
Jens Axboe4257c8c2019-11-25 14:27:34 -07002748out:
2749 return err;
2750}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002751
Jens Axboe4257c8c2019-11-25 14:27:34 -07002752static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2753 struct msghdr *msg_sys, unsigned int flags, int nosec)
2754{
2755 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2756 /* user mode address pointers */
2757 struct sockaddr __user *uaddr;
2758 ssize_t err;
2759
2760 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2761 if (err < 0)
2762 return err;
2763
2764 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
Al Viroda184282015-03-21 19:29:06 -04002765 kfree(iov);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002766 return err;
2767}
2768
2769/*
2770 * BSD recvmsg interface
2771 */
2772
Jens Axboe03b12302019-12-02 18:50:25 -07002773long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2774 struct user_msghdr __user *umsg,
2775 struct sockaddr __user *uaddr, unsigned int flags)
Jens Axboeaa1fa282019-04-19 13:38:09 -06002776{
Jens Axboe03b12302019-12-02 18:50:25 -07002777 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
Jens Axboeaa1fa282019-04-19 13:38:09 -06002778}
2779
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002780long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2781 bool forbid_cmsg_compat)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002782{
2783 int fput_needed, err;
2784 struct msghdr msg_sys;
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002785 struct socket *sock;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002786
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002787 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2788 return -EINVAL;
2789
Andy Lutomirski1be374a2013-05-22 14:07:44 -07002790 sock = sockfd_lookup_light(fd, &err, &fput_needed);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002791 if (!sock)
2792 goto out;
2793
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002794 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002795
Benjamin LaHaise6cb153c2006-03-20 22:27:12 -08002796 fput_light(sock->file, fput_needed);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002797out:
2798 return err;
2799}
2800
Al Viro666547f2014-04-06 14:03:05 -04002801SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002802 unsigned int, flags)
2803{
Dominik Brodowskie1834a32018-03-13 20:35:57 +01002804 return __sys_recvmsg(fd, msg, flags, true);
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002805}
2806
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002807/*
2808 * Linux recvmmsg interface
2809 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002811static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2812 unsigned int vlen, unsigned int flags,
2813 struct timespec64 *timeout)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002814{
2815 int fput_needed, err, datagrams;
2816 struct socket *sock;
2817 struct mmsghdr __user *entry;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002818 struct compat_mmsghdr __user *compat_entry;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002819 struct msghdr msg_sys;
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002820 struct timespec64 end_time;
2821 struct timespec64 timeout64;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002822
2823 if (timeout &&
2824 poll_select_set_timeout(&end_time, timeout->tv_sec,
2825 timeout->tv_nsec))
2826 return -EINVAL;
2827
2828 datagrams = 0;
2829
2830 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2831 if (!sock)
2832 return err;
2833
Soheil Hassas Yeganeh7797dc42018-02-27 18:22:40 -05002834 if (likely(!(flags & MSG_ERRQUEUE))) {
2835 err = sock_error(sock->sk);
2836 if (err) {
2837 datagrams = err;
2838 goto out_put;
2839 }
Maxime Jayate623a9e2017-02-21 18:35:51 +01002840 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002841
2842 entry = mmsg;
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002843 compat_entry = (struct compat_mmsghdr __user *)mmsg;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002844
2845 while (datagrams < vlen) {
2846 /*
2847 * No need to ask LSM for more than the first datagram.
2848 */
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002849 if (MSG_CMSG_COMPAT & flags) {
Al Viro666547f2014-04-06 14:03:05 -04002850 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002851 &msg_sys, flags & ~MSG_WAITFORONE,
2852 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002853 if (err < 0)
2854 break;
2855 err = __put_user(err, &compat_entry->msg_len);
2856 ++compat_entry;
2857 } else {
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002858 err = ___sys_recvmsg(sock,
Al Viro666547f2014-04-06 14:03:05 -04002859 (struct user_msghdr __user *)entry,
Andy Lutomirskia7526eb2013-06-05 19:38:26 +00002860 &msg_sys, flags & ~MSG_WAITFORONE,
2861 datagrams);
Jean-Mickael Guerind7256d02009-12-01 08:47:26 +00002862 if (err < 0)
2863 break;
2864 err = put_user(err, &entry->msg_len);
2865 ++entry;
2866 }
2867
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002868 if (err)
2869 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002870 ++datagrams;
2871
Brandon L Black71c5c1592010-03-26 16:18:03 +00002872 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2873 if (flags & MSG_WAITFORONE)
2874 flags |= MSG_DONTWAIT;
2875
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002876 if (timeout) {
Deepa Dinamani766b9f92016-05-19 17:09:05 -07002877 ktime_get_ts64(&timeout64);
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002878 *timeout = timespec64_sub(end_time, timeout64);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002879 if (timeout->tv_sec < 0) {
2880 timeout->tv_sec = timeout->tv_nsec = 0;
2881 break;
2882 }
2883
2884 /* Timeout, return less than vlen datagrams */
2885 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2886 break;
2887 }
2888
2889 /* Out of band data, return right away */
2890 if (msg_sys.msg_flags & MSG_OOB)
2891 break;
Eric Dumazeta78cb842016-01-08 08:37:20 -08002892 cond_resched();
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002893 }
2894
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002895 if (err == 0)
2896 goto out_put;
2897
2898 if (datagrams == 0) {
2899 datagrams = err;
2900 goto out_put;
2901 }
2902
2903 /*
2904 * We may return less entries than requested (vlen) if the
2905 * sock is non block and there aren't enough datagrams...
2906 */
2907 if (err != -EAGAIN) {
2908 /*
2909 * ... or if recvmsg returns an error after we
2910 * received some datagrams, where we record the
2911 * error to return on the next call or if the
2912 * app asks about it using getsockopt(SO_ERROR).
2913 */
2914 sock->sk->sk_err = -err;
2915 }
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002916out_put:
2917 fput_light(sock->file, fput_needed);
2918
Arnaldo Carvalho de Melo34b88a62016-03-14 09:56:35 -03002919 return datagrams;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002920}
2921
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002922int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2923 unsigned int vlen, unsigned int flags,
2924 struct __kernel_timespec __user *timeout,
2925 struct old_timespec32 __user *timeout32)
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002926{
2927 int datagrams;
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002928 struct timespec64 timeout_sys;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002929
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002930 if (timeout && get_timespec64(&timeout_sys, timeout))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002931 return -EFAULT;
2932
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002933 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2934 return -EFAULT;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002935
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002936 if (!timeout && !timeout32)
2937 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2938
2939 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2940
2941 if (datagrams <= 0)
2942 return datagrams;
2943
2944 if (timeout && put_timespec64(&timeout_sys, timeout))
2945 datagrams = -EFAULT;
2946
2947 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002948 datagrams = -EFAULT;
2949
2950 return datagrams;
2951}
2952
Dominik Brodowski1255e262018-03-13 20:44:21 +01002953SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2954 unsigned int, vlen, unsigned int, flags,
Arnd Bergmannc2e6c852018-04-18 13:42:25 +02002955 struct __kernel_timespec __user *, timeout)
Dominik Brodowski1255e262018-03-13 20:44:21 +01002956{
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002957 if (flags & MSG_CMSG_COMPAT)
2958 return -EINVAL;
2959
2960 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
Dominik Brodowski1255e262018-03-13 20:44:21 +01002961}
2962
Arnd Bergmanne11d4282018-04-18 13:43:52 +02002963#ifdef CONFIG_COMPAT_32BIT_TIME
2964SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2965 unsigned int, vlen, unsigned int, flags,
2966 struct old_timespec32 __user *, timeout)
2967{
2968 if (flags & MSG_CMSG_COMPAT)
2969 return -EINVAL;
2970
2971 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2972}
2973#endif
2974
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07002975#ifdef __ARCH_WANT_SYS_SOCKETCALL
Linus Torvalds1da177e2005-04-16 15:20:36 -07002976/* Argument list sizes for sys_socketcall */
2977#define AL(x) ((x) * sizeof(unsigned long))
Anton Blanchard228e5482011-05-02 20:21:35 +00002978static const unsigned char nargs[21] = {
Eric Dumazetc6d409c2010-06-03 20:03:40 -07002979 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2980 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2981 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
Anton Blanchard228e5482011-05-02 20:21:35 +00002982 AL(4), AL(5), AL(4)
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002983};
2984
Linus Torvalds1da177e2005-04-16 15:20:36 -07002985#undef AL
2986
2987/*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002988 * System call vectors.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002989 *
2990 * Argument checking cleaned up. Saved 20% in size.
2991 * This function doesn't need to set the kernel lock because
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002992 * it is set by the callees.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002993 */
2994
Heiko Carstens3e0fa652009-01-14 14:14:24 +01002995SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002996{
Chen Gang2950fa92013-04-07 16:55:23 +08002997 unsigned long a[AUDITSC_ARGS];
Stephen Hemminger89bddce2006-09-01 00:19:31 -07002998 unsigned long a0, a1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999 int err;
Arjan van de Ven47379052009-09-28 12:57:44 -07003000 unsigned int len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003001
Anton Blanchard228e5482011-05-02 20:21:35 +00003002 if (call < 1 || call > SYS_SENDMMSG)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003003 return -EINVAL;
Jeremy Clinec8e8cd52018-07-27 22:43:01 +00003004 call = array_index_nospec(call, SYS_SENDMMSG + 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003005
Arjan van de Ven47379052009-09-28 12:57:44 -07003006 len = nargs[call];
3007 if (len > sizeof(a))
3008 return -EINVAL;
3009
Linus Torvalds1da177e2005-04-16 15:20:36 -07003010 /* copy_from_user should be SMP safe. */
Arjan van de Ven47379052009-09-28 12:57:44 -07003011 if (copy_from_user(a, args, len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003012 return -EFAULT;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01003013
Chen Gang2950fa92013-04-07 16:55:23 +08003014 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3015 if (err)
3016 return err;
David Woodhouse3ec3b2f2005-05-17 12:08:48 +01003017
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003018 a0 = a[0];
3019 a1 = a[1];
3020
3021 switch (call) {
3022 case SYS_SOCKET:
Dominik Brodowski9d6a15c2018-03-13 19:29:43 +01003023 err = __sys_socket(a0, a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003024 break;
3025 case SYS_BIND:
Dominik Brodowskia87d35d2018-03-13 19:33:09 +01003026 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003027 break;
3028 case SYS_CONNECT:
Dominik Brodowski1387c2c2018-03-13 19:35:09 +01003029 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003030 break;
3031 case SYS_LISTEN:
Dominik Brodowski25e290e2018-03-13 19:36:54 +01003032 err = __sys_listen(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003033 break;
3034 case SYS_ACCEPT:
Dominik Brodowski4541e802018-03-13 19:24:23 +01003035 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3036 (int __user *)a[2], 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003037 break;
3038 case SYS_GETSOCKNAME:
3039 err =
Dominik Brodowski8882a102018-03-13 19:43:14 +01003040 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3041 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003042 break;
3043 case SYS_GETPEERNAME:
3044 err =
Dominik Brodowskib21c8f82018-03-13 19:47:00 +01003045 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3046 (int __user *)a[2]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003047 break;
3048 case SYS_SOCKETPAIR:
Dominik Brodowski6debc8d2018-03-13 19:49:23 +01003049 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003050 break;
3051 case SYS_SEND:
Dominik Brodowskif3bf8962018-03-13 19:52:00 +01003052 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3053 NULL, 0);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003054 break;
3055 case SYS_SENDTO:
Dominik Brodowski211b6342018-03-13 19:18:52 +01003056 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3057 (struct sockaddr __user *)a[4], a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003058 break;
3059 case SYS_RECV:
Dominik Brodowskid27e9af2018-03-13 19:54:17 +01003060 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3061 NULL, NULL);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003062 break;
3063 case SYS_RECVFROM:
Dominik Brodowski7a09e1e2018-03-13 19:10:06 +01003064 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3065 (struct sockaddr __user *)a[4],
3066 (int __user *)a[5]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003067 break;
3068 case SYS_SHUTDOWN:
Dominik Brodowski005a1ae2018-03-13 20:07:05 +01003069 err = __sys_shutdown(a0, a1);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003070 break;
3071 case SYS_SETSOCKOPT:
Dominik Brodowskicc36dca2018-03-13 20:10:59 +01003072 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3073 a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003074 break;
3075 case SYS_GETSOCKOPT:
3076 err =
Dominik Brodowski13a2d702018-03-13 20:15:04 +01003077 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3078 (int __user *)a[4]);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003079 break;
3080 case SYS_SENDMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003081 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3082 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003083 break;
Anton Blanchard228e5482011-05-02 20:21:35 +00003084 case SYS_SENDMMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003085 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3086 a[3], true);
Anton Blanchard228e5482011-05-02 20:21:35 +00003087 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003088 case SYS_RECVMSG:
Dominik Brodowskie1834a32018-03-13 20:35:57 +01003089 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3090 a[2], true);
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003091 break;
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003092 case SYS_RECVMMSG:
Arnd Bergmann3ca47e92019-04-23 17:43:50 +02003093 if (IS_ENABLED(CONFIG_64BIT))
Arnd Bergmanne11d4282018-04-18 13:43:52 +02003094 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3095 a[2], a[3],
3096 (struct __kernel_timespec __user *)a[4],
3097 NULL);
3098 else
3099 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3100 a[2], a[3], NULL,
3101 (struct old_timespec32 __user *)a[4]);
Arnaldo Carvalho de Meloa2e27252009-10-12 23:40:10 -07003102 break;
Ulrich Drepperde11def2008-11-19 15:36:14 -08003103 case SYS_ACCEPT4:
Dominik Brodowski4541e802018-03-13 19:24:23 +01003104 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3105 (int __user *)a[2], a[3]);
Ulrich Drepperaaca0bd2008-07-23 21:29:20 -07003106 break;
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003107 default:
3108 err = -EINVAL;
3109 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003110 }
3111 return err;
3112}
3113
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003114#endif /* __ARCH_WANT_SYS_SOCKETCALL */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003115
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003116/**
3117 * sock_register - add a socket protocol handler
3118 * @ops: description of protocol
3119 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003120 * This function is called by a protocol handler that wants to
3121 * advertise its address family, and have it linked into the
Masanari Iidae793c0f2014-09-04 23:44:36 +09003122 * socket interface. The value ops->family corresponds to the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003123 * socket system call protocol family.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003124 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003125int sock_register(const struct net_proto_family *ops)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003126{
3127 int err;
3128
3129 if (ops->family >= NPROTO) {
Yang Yingliang3410f222014-02-12 17:09:55 +08003130 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003131 return -ENOBUFS;
3132 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003133
3134 spin_lock(&net_family_lock);
Eric Dumazet190683a2010-11-10 10:50:44 +00003135 if (rcu_dereference_protected(net_families[ops->family],
3136 lockdep_is_held(&net_family_lock)))
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003137 err = -EEXIST;
3138 else {
Eric Dumazetcf778b02012-01-12 04:41:32 +00003139 rcu_assign_pointer(net_families[ops->family], ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003140 err = 0;
3141 }
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003142 spin_unlock(&net_family_lock);
3143
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003144 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003145 return err;
3146}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003147EXPORT_SYMBOL(sock_register);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003148
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003149/**
3150 * sock_unregister - remove a protocol handler
3151 * @family: protocol family to remove
3152 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07003153 * This function is called by a protocol handler that wants to
3154 * remove its address family, and have it unlinked from the
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003155 * new socket creation.
3156 *
3157 * If protocol handler is a module, then it can use module reference
3158 * counts to protect against new references. If protocol handler is not
3159 * a module then it needs to provide its own protection in
3160 * the ops->create routine.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003161 */
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003162void sock_unregister(int family)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003163{
Stephen Hemmingerf0fd27d2006-08-09 21:03:17 -07003164 BUG_ON(family < 0 || family >= NPROTO);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003165
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003166 spin_lock(&net_family_lock);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00003167 RCU_INIT_POINTER(net_families[family], NULL);
Stephen Hemminger55737fd2006-09-01 00:23:39 -07003168 spin_unlock(&net_family_lock);
3169
3170 synchronize_rcu();
3171
Yejune Dengfe0bdbd2021-06-21 13:12:25 +08003172 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003173}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003174EXPORT_SYMBOL(sock_unregister);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003175
Xin Longbf2ae2e2018-03-10 18:57:50 +08003176bool sock_is_registered(int family)
3177{
Jeremy Cline66b51b02018-08-13 22:23:13 +00003178 return family < NPROTO && rcu_access_pointer(net_families[family]);
Xin Longbf2ae2e2018-03-10 18:57:50 +08003179}
3180
Andi Kleen77d76ea2005-12-22 12:43:42 -08003181static int __init sock_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003182{
Nick Pigginb3e19d92011-01-07 17:50:11 +11003183 int err;
Eric W. Biederman2ca794e2012-04-19 13:20:32 +00003184 /*
3185 * Initialize the network sysctl infrastructure.
3186 */
3187 err = net_sysctl_init();
3188 if (err)
3189 goto out;
Nick Pigginb3e19d92011-01-07 17:50:11 +11003190
Linus Torvalds1da177e2005-04-16 15:20:36 -07003191 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003192 * Initialize skbuff SLAB cache
Linus Torvalds1da177e2005-04-16 15:20:36 -07003193 */
3194 skb_init();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003195
3196 /*
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003197 * Initialize the protocols module.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003198 */
3199
3200 init_inodecache();
Nick Pigginb3e19d92011-01-07 17:50:11 +11003201
3202 err = register_filesystem(&sock_fs_type);
3203 if (err)
Miaohe Lin47260ba2020-08-06 19:54:19 +08003204 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003205 sock_mnt = kern_mount(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003206 if (IS_ERR(sock_mnt)) {
3207 err = PTR_ERR(sock_mnt);
3208 goto out_mount;
3209 }
Andi Kleen77d76ea2005-12-22 12:43:42 -08003210
3211 /* The real protocol initialization is performed in later initcalls.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003212 */
3213
3214#ifdef CONFIG_NETFILTER
Pablo Neira Ayuso6d11cfd2013-05-22 22:42:36 +00003215 err = netfilter_init();
3216 if (err)
3217 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003218#endif
David S. Millercbeb3212005-12-22 12:58:55 -08003219
Daniel Borkmann408eccc2014-04-01 16:20:23 +02003220 ptp_classifier_init();
Richard Cochranc1f19b52010-07-17 08:49:36 +00003221
Nick Pigginb3e19d92011-01-07 17:50:11 +11003222out:
3223 return err;
3224
3225out_mount:
3226 unregister_filesystem(&sock_fs_type);
Nick Pigginb3e19d92011-01-07 17:50:11 +11003227 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003228}
3229
Andi Kleen77d76ea2005-12-22 12:43:42 -08003230core_initcall(sock_init); /* early initcall */
3231
Linus Torvalds1da177e2005-04-16 15:20:36 -07003232#ifdef CONFIG_PROC_FS
3233void socket_seq_show(struct seq_file *seq)
3234{
Tonghao Zhang648845a2017-12-14 05:51:58 -08003235 seq_printf(seq, "sockets: used %d\n",
3236 sock_inuse_get(seq->private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003237}
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003238#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003239
Arnd Bergmann29c49642021-07-22 16:29:03 +02003240/* Handle the fact that while struct ifreq has the same *layout* on
3241 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3242 * which are handled elsewhere, it still has different *size* due to
3243 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3244 * resulting in struct ifreq being 32 and 40 bytes respectively).
3245 * As a result, if the struct happens to be at the end of a page and
3246 * the next page isn't readable/writable, we get a fault. To prevent
3247 * that, copy back and forth to the full size.
3248 */
3249int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3250{
3251 if (in_compat_syscall()) {
3252 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3253
3254 memset(ifr, 0, sizeof(*ifr));
3255 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3256 return -EFAULT;
3257
3258 if (ifrdata)
3259 *ifrdata = compat_ptr(ifr32->ifr_data);
3260
3261 return 0;
3262 }
3263
3264 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3265 return -EFAULT;
3266
3267 if (ifrdata)
3268 *ifrdata = ifr->ifr_data;
3269
3270 return 0;
3271}
3272EXPORT_SYMBOL(get_user_ifreq);
3273
3274int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3275{
3276 size_t size = sizeof(*ifr);
3277
3278 if (in_compat_syscall())
3279 size = sizeof(struct compat_ifreq);
3280
3281 if (copy_to_user(arg, ifr, size))
3282 return -EFAULT;
3283
3284 return 0;
3285}
3286EXPORT_SYMBOL(put_user_ifreq);
3287
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003288#ifdef CONFIG_COMPAT
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003289static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3290{
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003291 compat_uptr_t uptr32;
Al Viro44c02a22017-10-05 12:59:44 -04003292 struct ifreq ifr;
3293 void __user *saved;
3294 int err;
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003295
Arnd Bergmann29c49642021-07-22 16:29:03 +02003296 if (get_user_ifreq(&ifr, NULL, uifr32))
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003297 return -EFAULT;
3298
3299 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3300 return -EFAULT;
3301
Al Viro44c02a22017-10-05 12:59:44 -04003302 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3303 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003304
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003305 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
Al Viro44c02a22017-10-05 12:59:44 -04003306 if (!err) {
3307 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
Arnd Bergmann29c49642021-07-22 16:29:03 +02003308 if (put_user_ifreq(&ifr, uifr32))
Al Viro44c02a22017-10-05 12:59:44 -04003309 err = -EFAULT;
Joe Perchesccbd6a52010-05-14 10:58:26 +00003310 }
Al Viro44c02a22017-10-05 12:59:44 -04003311 return err;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003312}
3313
Ben Hutchings590d4692013-11-18 17:04:13 +00003314/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3315static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
Arnd Bergmann6b960182009-11-06 23:10:54 -08003316 struct compat_ifreq __user *u_ifreq32)
Arnd Bergmann7a229382009-11-06 23:00:29 -08003317{
Al Viro44c02a22017-10-05 12:59:44 -04003318 struct ifreq ifreq;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003319 void __user *data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003320
Peter Collingbourned0efb162021-08-26 12:46:01 -07003321 if (!is_socket_ioctl_cmd(cmd))
3322 return -ENOTTY;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003323 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
Arnd Bergmann7a229382009-11-06 23:00:29 -08003324 return -EFAULT;
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003325 ifreq.ifr_data = data;
Arnd Bergmann7a229382009-11-06 23:00:29 -08003326
Arnd Bergmanna554bf92021-07-27 15:45:12 +02003327 return dev_ioctl(net, cmd, &ifreq, data, NULL);
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003328}
3329
Arnd Bergmann6b960182009-11-06 23:10:54 -08003330static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3331 unsigned int cmd, unsigned long arg)
3332{
3333 void __user *argp = compat_ptr(arg);
3334 struct sock *sk = sock->sk;
3335 struct net *net = sock_net(sk);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003336
Arnd Bergmann6b960182009-11-06 23:10:54 -08003337 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
Arnd Bergmann88fc0232021-07-27 15:45:15 +02003338 return sock_ioctl(file, cmd, (unsigned long)argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003339
Arnd Bergmann6b960182009-11-06 23:10:54 -08003340 switch (cmd) {
Arnd Bergmann7a50a242009-11-08 20:57:03 -08003341 case SIOCWANDEV:
3342 return compat_siocwandev(net, argp);
Arnd Bergmann0768e172019-04-17 22:56:11 +02003343 case SIOCGSTAMP_OLD:
3344 case SIOCGSTAMPNS_OLD:
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003345 if (!sock->ops->gettstamp)
3346 return -ENOIOCTLCMD;
Arnd Bergmann0768e172019-04-17 22:56:11 +02003347 return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
Arnd Bergmannc7cbdbf2019-04-17 22:51:48 +02003348 !COMPAT_USE_64BIT_TIME);
3349
Arnd Bergmanndd98d282021-07-22 16:28:59 +02003350 case SIOCETHTOOL:
Ben Hutchings590d4692013-11-18 17:04:13 +00003351 case SIOCBONDSLAVEINFOQUERY:
3352 case SIOCBONDINFOQUERY:
Arnd Bergmanna2116ed2009-11-11 03:39:40 +00003353 case SIOCSHWTSTAMP:
Ben Hutchingsfd468c72013-11-14 01:19:29 +00003354 case SIOCGHWTSTAMP:
Ben Hutchings590d4692013-11-18 17:04:13 +00003355 return compat_ifr_data_ioctl(net, cmd, argp);
Arnd Bergmann7a229382009-11-06 23:00:29 -08003356
Arnd Bergmann6b960182009-11-06 23:10:54 -08003357 case FIOSETOWN:
3358 case SIOCSPGRP:
3359 case FIOGETOWN:
3360 case SIOCGPGRP:
3361 case SIOCBRADDBR:
3362 case SIOCBRDELBR:
3363 case SIOCGIFVLAN:
3364 case SIOCSIFVLAN:
Andrey Vaginc62cce22016-10-24 18:29:13 -07003365 case SIOCGSKNS:
Arnd Bergmann0768e172019-04-17 22:56:11 +02003366 case SIOCGSTAMP_NEW:
3367 case SIOCGSTAMPNS_NEW:
Arnd Bergmann876f0bf2021-07-22 16:29:02 +02003368 case SIOCGIFCONF:
Remi Pommarelfd3a4592021-12-24 12:46:40 +01003369 case SIOCSIFBR:
3370 case SIOCGIFBR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003371 return sock_ioctl(file, cmd, arg);
3372
3373 case SIOCGIFFLAGS:
3374 case SIOCSIFFLAGS:
Arnd Bergmann709566d2021-07-22 16:29:00 +02003375 case SIOCGIFMAP:
3376 case SIOCSIFMAP:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003377 case SIOCGIFMETRIC:
3378 case SIOCSIFMETRIC:
3379 case SIOCGIFMTU:
3380 case SIOCSIFMTU:
3381 case SIOCGIFMEM:
3382 case SIOCSIFMEM:
3383 case SIOCGIFHWADDR:
3384 case SIOCSIFHWADDR:
3385 case SIOCADDMULTI:
3386 case SIOCDELMULTI:
3387 case SIOCGIFINDEX:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003388 case SIOCGIFADDR:
3389 case SIOCSIFADDR:
3390 case SIOCSIFHWBROADCAST:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003391 case SIOCDIFADDR:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003392 case SIOCGIFBRDADDR:
3393 case SIOCSIFBRDADDR:
3394 case SIOCGIFDSTADDR:
3395 case SIOCSIFDSTADDR:
3396 case SIOCGIFNETMASK:
3397 case SIOCSIFNETMASK:
3398 case SIOCSIFPFLAGS:
3399 case SIOCGIFPFLAGS:
3400 case SIOCGIFTXQLEN:
3401 case SIOCSIFTXQLEN:
3402 case SIOCBRADDIF:
3403 case SIOCBRDELIF:
Johannes Bergc6c9fee2019-01-25 22:43:19 +01003404 case SIOCGIFNAME:
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003405 case SIOCSIFNAME:
3406 case SIOCGMIIPHY:
3407 case SIOCGMIIREG:
3408 case SIOCSMIIREG:
Al Virof92d4fc2017-09-30 19:32:17 -04003409 case SIOCBONDENSLAVE:
3410 case SIOCBONDRELEASE:
3411 case SIOCBONDSETHWADDR:
3412 case SIOCBONDCHANGEACTIVE:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003413 case SIOCSARP:
3414 case SIOCGARP:
3415 case SIOCDARP:
Arnd Bergmannc7dc5042019-06-03 23:07:12 +02003416 case SIOCOUTQ:
Arnd Bergmann9d7bf412019-06-03 23:06:00 +02003417 case SIOCOUTQNSD:
Arnd Bergmann6b960182009-11-06 23:10:54 -08003418 case SIOCATMARK:
Johannes Berg63ff03a2019-01-25 22:43:17 +01003419 return sock_do_ioctl(net, sock, cmd, arg);
Arnd Bergmann9177efd2009-11-06 08:09:09 +00003420 }
3421
Arnd Bergmann6b960182009-11-06 23:10:54 -08003422 return -ENOIOCTLCMD;
3423}
Arnd Bergmann7a229382009-11-06 23:00:29 -08003424
Eric Dumazet95c96172012-04-15 05:58:06 +00003425static long compat_sock_ioctl(struct file *file, unsigned int cmd,
Stephen Hemminger89bddce2006-09-01 00:19:31 -07003426 unsigned long arg)
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003427{
3428 struct socket *sock = file->private_data;
3429 int ret = -ENOIOCTLCMD;
David S. Miller87de87d2008-06-03 09:14:03 -07003430 struct sock *sk;
3431 struct net *net;
3432
3433 sk = sock->sk;
3434 net = sock_net(sk);
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003435
3436 if (sock->ops->compat_ioctl)
3437 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3438
David S. Miller87de87d2008-06-03 09:14:03 -07003439 if (ret == -ENOIOCTLCMD &&
3440 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3441 ret = compat_wext_handle_ioctl(net, cmd, arg);
3442
Arnd Bergmann6b960182009-11-06 23:10:54 -08003443 if (ret == -ENOIOCTLCMD)
3444 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3445
Shaun Pereira89bbfc92006-03-21 23:58:08 -08003446 return ret;
3447}
3448#endif
3449
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003450/**
3451 * kernel_bind - bind an address to a socket (kernel space)
3452 * @sock: socket
3453 * @addr: address
3454 * @addrlen: length of address
3455 *
3456 * Returns 0 or an error.
3457 */
3458
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003459int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3460{
3461 return sock->ops->bind(sock, addr, addrlen);
3462}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003463EXPORT_SYMBOL(kernel_bind);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003464
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003465/**
3466 * kernel_listen - move socket to listening state (kernel space)
3467 * @sock: socket
3468 * @backlog: pending connections queue size
3469 *
3470 * Returns 0 or an error.
3471 */
3472
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003473int kernel_listen(struct socket *sock, int backlog)
3474{
3475 return sock->ops->listen(sock, backlog);
3476}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003477EXPORT_SYMBOL(kernel_listen);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003478
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003479/**
3480 * kernel_accept - accept a connection (kernel space)
3481 * @sock: listening socket
3482 * @newsock: new connected socket
3483 * @flags: flags
3484 *
3485 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3486 * If it fails, @newsock is guaranteed to be %NULL.
3487 * Returns 0 or an error.
3488 */
3489
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003490int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3491{
3492 struct sock *sk = sock->sk;
3493 int err;
3494
3495 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3496 newsock);
3497 if (err < 0)
3498 goto done;
3499
David Howellscdfbabf2017-03-09 08:09:05 +00003500 err = sock->ops->accept(sock, *newsock, flags, true);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003501 if (err < 0) {
3502 sock_release(*newsock);
Tony Battersbyfa8705b2007-10-10 21:09:04 -07003503 *newsock = NULL;
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003504 goto done;
3505 }
3506
3507 (*newsock)->ops = sock->ops;
Wei Yongjun1b085342008-12-18 19:35:10 -08003508 __module_get((*newsock)->ops->owner);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003509
3510done:
3511 return err;
3512}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003513EXPORT_SYMBOL(kernel_accept);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003514
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003515/**
3516 * kernel_connect - connect a socket (kernel space)
3517 * @sock: socket
3518 * @addr: address
3519 * @addrlen: address length
3520 * @flags: flags (O_NONBLOCK, ...)
3521 *
Lu Weif1dcffc2021-03-25 11:01:55 +08003522 * For datagram sockets, @addr is the address to which datagrams are sent
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003523 * by default, and the only address from which datagrams are received.
3524 * For stream sockets, attempts to connect to @addr.
3525 * Returns 0 or an error code.
3526 */
3527
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003528int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
YOSHIFUJI Hideaki4768fbc2007-02-09 23:25:31 +09003529 int flags)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003530{
3531 return sock->ops->connect(sock, addr, addrlen, flags);
3532}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003533EXPORT_SYMBOL(kernel_connect);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003534
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003535/**
3536 * kernel_getsockname - get the address which the socket is bound (kernel space)
3537 * @sock: socket
3538 * @addr: address holder
3539 *
3540 * Fills the @addr pointer with the address which the socket is bound.
Alex Maydanik0fc95de2022-02-12 12:29:27 +02003541 * Returns the length of the address in bytes or an error code.
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003542 */
3543
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003544int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003545{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003546 return sock->ops->getname(sock, addr, 0);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003547}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003548EXPORT_SYMBOL(kernel_getsockname);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003549
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003550/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003551 * kernel_getpeername - get the address which the socket is connected (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003552 * @sock: socket
3553 * @addr: address holder
3554 *
3555 * Fills the @addr pointer with the address which the socket is connected.
Alex Maydanik0fc95de2022-02-12 12:29:27 +02003556 * Returns the length of the address in bytes or an error code.
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003557 */
3558
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003559int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003560{
Denys Vlasenko9b2c45d2018-02-12 20:00:20 +01003561 return sock->ops->getname(sock, addr, 1);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003562}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003563EXPORT_SYMBOL(kernel_getpeername);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003564
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003565/**
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003566 * kernel_sendpage - send a &page through a socket (kernel space)
3567 * @sock: socket
3568 * @page: page
3569 * @offset: page offset
3570 * @size: total size in bytes
3571 * @flags: flags (MSG_DONTWAIT, ...)
3572 *
3573 * Returns the total amount sent in bytes or an error.
3574 */
3575
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003576int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3577 size_t size, int flags)
3578{
Coly Li7b62d312020-10-02 16:27:29 +08003579 if (sock->ops->sendpage) {
3580 /* Warn in case the improper page to zero-copy send */
3581 WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003582 return sock->ops->sendpage(sock, page, offset, size, flags);
Coly Li7b62d312020-10-02 16:27:29 +08003583 }
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003584 return sock_no_sendpage(sock, page, offset, size, flags);
3585}
Eric Dumazetc6d409c2010-06-03 20:03:40 -07003586EXPORT_SYMBOL(kernel_sendpage);
Sridhar Samudralaac5a4882006-08-07 20:57:31 -07003587
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003588/**
3589 * kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3590 * @sk: sock
3591 * @page: page
3592 * @offset: page offset
3593 * @size: total size in bytes
3594 * @flags: flags (MSG_DONTWAIT, ...)
3595 *
3596 * Returns the total amount sent in bytes or an error.
3597 * Caller must hold @sk.
3598 */
3599
Tom Herbert306b13e2017-07-28 16:22:41 -07003600int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3601 size_t size, int flags)
3602{
3603 struct socket *sock = sk->sk_socket;
3604
3605 if (sock->ops->sendpage_locked)
3606 return sock->ops->sendpage_locked(sk, page, offset, size,
3607 flags);
3608
3609 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3610}
3611EXPORT_SYMBOL(kernel_sendpage_locked);
3612
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003613/**
Miaohe Lin645f0892020-08-27 07:27:49 -04003614 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003615 * @sock: socket
3616 * @how: connection part
3617 *
3618 * Returns 0 or an error.
3619 */
3620
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003621int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3622{
3623 return sock->ops->shutdown(sock, how);
3624}
Trond Myklebust91cf45f2007-11-12 18:10:39 -08003625EXPORT_SYMBOL(kernel_sock_shutdown);
R. Parameswaran113c3072017-04-05 16:50:35 -07003626
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003627/**
3628 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3629 * @sk: socket
3630 *
3631 * This routine returns the IP overhead imposed by a socket i.e.
3632 * the length of the underlying IP header, depending on whether
3633 * this is an IPv4 or IPv6 socket and the length from IP options turned
3634 * on at the socket. Assumes that the caller has a lock on the socket.
R. Parameswaran113c3072017-04-05 16:50:35 -07003635 */
Pedro Tammela8a3c2452019-03-14 10:45:23 -03003636
R. Parameswaran113c3072017-04-05 16:50:35 -07003637u32 kernel_sock_ip_overhead(struct sock *sk)
3638{
3639 struct inet_sock *inet;
3640 struct ip_options_rcu *opt;
3641 u32 overhead = 0;
R. Parameswaran113c3072017-04-05 16:50:35 -07003642#if IS_ENABLED(CONFIG_IPV6)
3643 struct ipv6_pinfo *np;
3644 struct ipv6_txoptions *optv6 = NULL;
3645#endif /* IS_ENABLED(CONFIG_IPV6) */
3646
3647 if (!sk)
3648 return overhead;
3649
R. Parameswaran113c3072017-04-05 16:50:35 -07003650 switch (sk->sk_family) {
3651 case AF_INET:
3652 inet = inet_sk(sk);
3653 overhead += sizeof(struct iphdr);
3654 opt = rcu_dereference_protected(inet->inet_opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003655 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003656 if (opt)
3657 overhead += opt->opt.optlen;
3658 return overhead;
3659#if IS_ENABLED(CONFIG_IPV6)
3660 case AF_INET6:
3661 np = inet6_sk(sk);
3662 overhead += sizeof(struct ipv6hdr);
3663 if (np)
3664 optv6 = rcu_dereference_protected(np->opt,
stephen hemminger614d79c2017-07-24 10:25:22 -07003665 sock_owned_by_user(sk));
R. Parameswaran113c3072017-04-05 16:50:35 -07003666 if (optv6)
3667 overhead += (optv6->opt_flen + optv6->opt_nflen);
3668 return overhead;
3669#endif /* IS_ENABLED(CONFIG_IPV6) */
3670 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3671 return overhead;
3672 }
3673}
3674EXPORT_SYMBOL(kernel_sock_ip_overhead);