xref: /linux/net/socket.c (revision 1a9239bb4253f9076b5b4b2a1a4e8d7defd77a95)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
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>
8  *		Ross Biro
9  *		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)
46  *		Tigran Aivazian	:	Made listen(2) backlog sanity checks
47  *					protocol-independent
48  *
49  *	This module is effectively the top level interface to the BSD socket
50  *	paradigm.
51  *
52  *	Based upon Swansea University Computer Society NET3.039
53  */
54 
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
57 #include <linux/mm.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/splice.h>
61 #include <linux/net.h>
62 #include <linux/interrupt.h>
63 #include <linux/thread_info.h>
64 #include <linux/rcupdate.h>
65 #include <linux/netdevice.h>
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68 #include <linux/mutex.h>
69 #include <linux/if_bridge.h>
70 #include <linux/if_vlan.h>
71 #include <linux/ptp_classify.h>
72 #include <linux/init.h>
73 #include <linux/poll.h>
74 #include <linux/cache.h>
75 #include <linux/module.h>
76 #include <linux/highmem.h>
77 #include <linux/mount.h>
78 #include <linux/pseudo_fs.h>
79 #include <linux/security.h>
80 #include <linux/syscalls.h>
81 #include <linux/compat.h>
82 #include <linux/kmod.h>
83 #include <linux/audit.h>
84 #include <linux/wireless.h>
85 #include <linux/nsproxy.h>
86 #include <linux/magic.h>
87 #include <linux/slab.h>
88 #include <linux/xattr.h>
89 #include <linux/nospec.h>
90 #include <linux/indirect_call_wrapper.h>
91 #include <linux/io_uring/net.h>
92 
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
95 
96 #include <net/compat.h>
97 #include <net/wext.h>
98 #include <net/cls_cgroup.h>
99 
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102 
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/termios.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 #include <linux/ptp_clock_kernel.h>
111 #include <trace/events/sock.h>
112 
113 #include "core/dev.h"
114 
115 #ifdef CONFIG_NET_RX_BUSY_POLL
116 unsigned int sysctl_net_busy_read __read_mostly;
117 unsigned int sysctl_net_busy_poll __read_mostly;
118 #endif
119 
120 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
121 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
122 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
123 
124 static int sock_close(struct inode *inode, struct file *file);
125 static __poll_t sock_poll(struct file *file,
126 			      struct poll_table_struct *wait);
127 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
128 #ifdef CONFIG_COMPAT
129 static long compat_sock_ioctl(struct file *file,
130 			      unsigned int cmd, unsigned long arg);
131 #endif
132 static int sock_fasync(int fd, struct file *filp, int on);
133 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
134 				struct pipe_inode_info *pipe, size_t len,
135 				unsigned int flags);
136 static void sock_splice_eof(struct file *file);
137 
138 #ifdef CONFIG_PROC_FS
sock_show_fdinfo(struct seq_file * m,struct file * f)139 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
140 {
141 	struct socket *sock = f->private_data;
142 	const struct proto_ops *ops = READ_ONCE(sock->ops);
143 
144 	if (ops->show_fdinfo)
145 		ops->show_fdinfo(m, sock);
146 }
147 #else
148 #define sock_show_fdinfo NULL
149 #endif
150 
151 /*
152  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
153  *	in the operation structures but are done directly via the socketcall() multiplexor.
154  */
155 
156 static const struct file_operations socket_file_ops = {
157 	.owner =	THIS_MODULE,
158 	.read_iter =	sock_read_iter,
159 	.write_iter =	sock_write_iter,
160 	.poll =		sock_poll,
161 	.unlocked_ioctl = sock_ioctl,
162 #ifdef CONFIG_COMPAT
163 	.compat_ioctl = compat_sock_ioctl,
164 #endif
165 	.uring_cmd =    io_uring_cmd_sock,
166 	.mmap =		sock_mmap,
167 	.release =	sock_close,
168 	.fasync =	sock_fasync,
169 	.splice_write = splice_to_socket,
170 	.splice_read =	sock_splice_read,
171 	.splice_eof =	sock_splice_eof,
172 	.show_fdinfo =	sock_show_fdinfo,
173 };
174 
175 static const char * const pf_family_names[] = {
176 	[PF_UNSPEC]	= "PF_UNSPEC",
177 	[PF_UNIX]	= "PF_UNIX/PF_LOCAL",
178 	[PF_INET]	= "PF_INET",
179 	[PF_AX25]	= "PF_AX25",
180 	[PF_IPX]	= "PF_IPX",
181 	[PF_APPLETALK]	= "PF_APPLETALK",
182 	[PF_NETROM]	= "PF_NETROM",
183 	[PF_BRIDGE]	= "PF_BRIDGE",
184 	[PF_ATMPVC]	= "PF_ATMPVC",
185 	[PF_X25]	= "PF_X25",
186 	[PF_INET6]	= "PF_INET6",
187 	[PF_ROSE]	= "PF_ROSE",
188 	[PF_DECnet]	= "PF_DECnet",
189 	[PF_NETBEUI]	= "PF_NETBEUI",
190 	[PF_SECURITY]	= "PF_SECURITY",
191 	[PF_KEY]	= "PF_KEY",
192 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
193 	[PF_PACKET]	= "PF_PACKET",
194 	[PF_ASH]	= "PF_ASH",
195 	[PF_ECONET]	= "PF_ECONET",
196 	[PF_ATMSVC]	= "PF_ATMSVC",
197 	[PF_RDS]	= "PF_RDS",
198 	[PF_SNA]	= "PF_SNA",
199 	[PF_IRDA]	= "PF_IRDA",
200 	[PF_PPPOX]	= "PF_PPPOX",
201 	[PF_WANPIPE]	= "PF_WANPIPE",
202 	[PF_LLC]	= "PF_LLC",
203 	[PF_IB]		= "PF_IB",
204 	[PF_MPLS]	= "PF_MPLS",
205 	[PF_CAN]	= "PF_CAN",
206 	[PF_TIPC]	= "PF_TIPC",
207 	[PF_BLUETOOTH]	= "PF_BLUETOOTH",
208 	[PF_IUCV]	= "PF_IUCV",
209 	[PF_RXRPC]	= "PF_RXRPC",
210 	[PF_ISDN]	= "PF_ISDN",
211 	[PF_PHONET]	= "PF_PHONET",
212 	[PF_IEEE802154]	= "PF_IEEE802154",
213 	[PF_CAIF]	= "PF_CAIF",
214 	[PF_ALG]	= "PF_ALG",
215 	[PF_NFC]	= "PF_NFC",
216 	[PF_VSOCK]	= "PF_VSOCK",
217 	[PF_KCM]	= "PF_KCM",
218 	[PF_QIPCRTR]	= "PF_QIPCRTR",
219 	[PF_SMC]	= "PF_SMC",
220 	[PF_XDP]	= "PF_XDP",
221 	[PF_MCTP]	= "PF_MCTP",
222 };
223 
224 /*
225  *	The protocol list. Each protocol is registered in here.
226  */
227 
228 static DEFINE_SPINLOCK(net_family_lock);
229 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
230 
231 /*
232  * Support routines.
233  * Move socket addresses back and forth across the kernel/user
234  * divide and look after the messy bits.
235  */
236 
237 /**
238  *	move_addr_to_kernel	-	copy a socket address into kernel space
239  *	@uaddr: Address in user space
240  *	@kaddr: Address in kernel space
241  *	@ulen: Length in user space
242  *
243  *	The address is copied into kernel space. If the provided address is
244  *	too long an error code of -EINVAL is returned. If the copy gives
245  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
246  */
247 
move_addr_to_kernel(void __user * uaddr,int ulen,struct sockaddr_storage * kaddr)248 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
249 {
250 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
251 		return -EINVAL;
252 	if (ulen == 0)
253 		return 0;
254 	if (copy_from_user(kaddr, uaddr, ulen))
255 		return -EFAULT;
256 	return audit_sockaddr(ulen, kaddr);
257 }
258 
259 /**
260  *	move_addr_to_user	-	copy an address to user space
261  *	@kaddr: kernel space address
262  *	@klen: length of address in kernel
263  *	@uaddr: user space address
264  *	@ulen: pointer to user length field
265  *
266  *	The value pointed to by ulen on entry is the buffer length available.
267  *	This is overwritten with the buffer space used. -EINVAL is returned
268  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
269  *	is returned if either the buffer or the length field are not
270  *	accessible.
271  *	After copying the data up to the limit the user specifies, the true
272  *	length of the data is written over the length limit the user
273  *	specified. Zero is returned for a success.
274  */
275 
move_addr_to_user(struct sockaddr_storage * kaddr,int klen,void __user * uaddr,int __user * ulen)276 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
277 			     void __user *uaddr, int __user *ulen)
278 {
279 	int err;
280 	int len;
281 
282 	BUG_ON(klen > sizeof(struct sockaddr_storage));
283 	err = get_user(len, ulen);
284 	if (err)
285 		return err;
286 	if (len > klen)
287 		len = klen;
288 	if (len < 0)
289 		return -EINVAL;
290 	if (len) {
291 		if (audit_sockaddr(klen, kaddr))
292 			return -ENOMEM;
293 		if (copy_to_user(uaddr, kaddr, len))
294 			return -EFAULT;
295 	}
296 	/*
297 	 *      "fromlen shall refer to the value before truncation.."
298 	 *                      1003.1g
299 	 */
300 	return __put_user(klen, ulen);
301 }
302 
303 static struct kmem_cache *sock_inode_cachep __ro_after_init;
304 
sock_alloc_inode(struct super_block * sb)305 static struct inode *sock_alloc_inode(struct super_block *sb)
306 {
307 	struct socket_alloc *ei;
308 
309 	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
310 	if (!ei)
311 		return NULL;
312 	init_waitqueue_head(&ei->socket.wq.wait);
313 	ei->socket.wq.fasync_list = NULL;
314 	ei->socket.wq.flags = 0;
315 
316 	ei->socket.state = SS_UNCONNECTED;
317 	ei->socket.flags = 0;
318 	ei->socket.ops = NULL;
319 	ei->socket.sk = NULL;
320 	ei->socket.file = NULL;
321 
322 	return &ei->vfs_inode;
323 }
324 
sock_free_inode(struct inode * inode)325 static void sock_free_inode(struct inode *inode)
326 {
327 	struct socket_alloc *ei;
328 
329 	ei = container_of(inode, struct socket_alloc, vfs_inode);
330 	kmem_cache_free(sock_inode_cachep, ei);
331 }
332 
init_once(void * foo)333 static void init_once(void *foo)
334 {
335 	struct socket_alloc *ei = (struct socket_alloc *)foo;
336 
337 	inode_init_once(&ei->vfs_inode);
338 }
339 
init_inodecache(void)340 static void init_inodecache(void)
341 {
342 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
343 					      sizeof(struct socket_alloc),
344 					      0,
345 					      (SLAB_HWCACHE_ALIGN |
346 					       SLAB_RECLAIM_ACCOUNT |
347 					       SLAB_ACCOUNT),
348 					      init_once);
349 	BUG_ON(sock_inode_cachep == NULL);
350 }
351 
352 static const struct super_operations sockfs_ops = {
353 	.alloc_inode	= sock_alloc_inode,
354 	.free_inode	= sock_free_inode,
355 	.statfs		= simple_statfs,
356 };
357 
358 /*
359  * sockfs_dname() is called from d_path().
360  */
sockfs_dname(struct dentry * dentry,char * buffer,int buflen)361 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
362 {
363 	return dynamic_dname(buffer, buflen, "socket:[%lu]",
364 				d_inode(dentry)->i_ino);
365 }
366 
367 static const struct dentry_operations sockfs_dentry_operations = {
368 	.d_dname  = sockfs_dname,
369 };
370 
sockfs_xattr_get(const struct xattr_handler * handler,struct dentry * dentry,struct inode * inode,const char * suffix,void * value,size_t size)371 static int sockfs_xattr_get(const struct xattr_handler *handler,
372 			    struct dentry *dentry, struct inode *inode,
373 			    const char *suffix, void *value, size_t size)
374 {
375 	if (value) {
376 		if (dentry->d_name.len + 1 > size)
377 			return -ERANGE;
378 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
379 	}
380 	return dentry->d_name.len + 1;
381 }
382 
383 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
384 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
385 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
386 
387 static const struct xattr_handler sockfs_xattr_handler = {
388 	.name = XATTR_NAME_SOCKPROTONAME,
389 	.get = sockfs_xattr_get,
390 };
391 
sockfs_security_xattr_set(const struct xattr_handler * handler,struct mnt_idmap * idmap,struct dentry * dentry,struct inode * inode,const char * suffix,const void * value,size_t size,int flags)392 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
393 				     struct mnt_idmap *idmap,
394 				     struct dentry *dentry, struct inode *inode,
395 				     const char *suffix, const void *value,
396 				     size_t size, int flags)
397 {
398 	/* Handled by LSM. */
399 	return -EAGAIN;
400 }
401 
402 static const struct xattr_handler sockfs_security_xattr_handler = {
403 	.prefix = XATTR_SECURITY_PREFIX,
404 	.set = sockfs_security_xattr_set,
405 };
406 
407 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
408 	&sockfs_xattr_handler,
409 	&sockfs_security_xattr_handler,
410 	NULL
411 };
412 
sockfs_init_fs_context(struct fs_context * fc)413 static int sockfs_init_fs_context(struct fs_context *fc)
414 {
415 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
416 	if (!ctx)
417 		return -ENOMEM;
418 	ctx->ops = &sockfs_ops;
419 	ctx->dops = &sockfs_dentry_operations;
420 	ctx->xattr = sockfs_xattr_handlers;
421 	return 0;
422 }
423 
424 static struct vfsmount *sock_mnt __read_mostly;
425 
426 static struct file_system_type sock_fs_type = {
427 	.name =		"sockfs",
428 	.init_fs_context = sockfs_init_fs_context,
429 	.kill_sb =	kill_anon_super,
430 };
431 
432 /*
433  *	Obtains the first available file descriptor and sets it up for use.
434  *
435  *	These functions create file structures and maps them to fd space
436  *	of the current process. On success it returns file descriptor
437  *	and file struct implicitly stored in sock->file.
438  *	Note that another thread may close file descriptor before we return
439  *	from this function. We use the fact that now we do not refer
440  *	to socket after mapping. If one day we will need it, this
441  *	function will increment ref. count on file by 1.
442  *
443  *	In any case returned fd MAY BE not valid!
444  *	This race condition is unavoidable
445  *	with shared fd spaces, we cannot solve it inside kernel,
446  *	but we take care of internal coherence yet.
447  */
448 
449 /**
450  *	sock_alloc_file - Bind a &socket to a &file
451  *	@sock: socket
452  *	@flags: file status flags
453  *	@dname: protocol name
454  *
455  *	Returns the &file bound with @sock, implicitly storing it
456  *	in sock->file. If dname is %NULL, sets to "".
457  *
458  *	On failure @sock is released, and an ERR pointer is returned.
459  *
460  *	This function uses GFP_KERNEL internally.
461  */
462 
sock_alloc_file(struct socket * sock,int flags,const char * dname)463 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
464 {
465 	struct file *file;
466 
467 	if (!dname)
468 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
469 
470 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
471 				O_RDWR | (flags & O_NONBLOCK),
472 				&socket_file_ops);
473 	if (IS_ERR(file)) {
474 		sock_release(sock);
475 		return file;
476 	}
477 
478 	file->f_mode |= FMODE_NOWAIT;
479 	sock->file = file;
480 	file->private_data = sock;
481 	stream_open(SOCK_INODE(sock), file);
482 	/*
483 	 * Disable permission and pre-content events, but enable legacy
484 	 * inotify events for legacy users.
485 	 */
486 	file_set_fsnotify_mode(file, FMODE_NONOTIFY_PERM);
487 	return file;
488 }
489 EXPORT_SYMBOL(sock_alloc_file);
490 
sock_map_fd(struct socket * sock,int flags)491 static int sock_map_fd(struct socket *sock, int flags)
492 {
493 	struct file *newfile;
494 	int fd = get_unused_fd_flags(flags);
495 	if (unlikely(fd < 0)) {
496 		sock_release(sock);
497 		return fd;
498 	}
499 
500 	newfile = sock_alloc_file(sock, flags, NULL);
501 	if (!IS_ERR(newfile)) {
502 		fd_install(fd, newfile);
503 		return fd;
504 	}
505 
506 	put_unused_fd(fd);
507 	return PTR_ERR(newfile);
508 }
509 
510 /**
511  *	sock_from_file - Return the &socket bounded to @file.
512  *	@file: file
513  *
514  *	On failure returns %NULL.
515  */
516 
sock_from_file(struct file * file)517 struct socket *sock_from_file(struct file *file)
518 {
519 	if (likely(file->f_op == &socket_file_ops))
520 		return file->private_data;	/* set in sock_alloc_file */
521 
522 	return NULL;
523 }
524 EXPORT_SYMBOL(sock_from_file);
525 
526 /**
527  *	sockfd_lookup - Go from a file number to its socket slot
528  *	@fd: file handle
529  *	@err: pointer to an error code return
530  *
531  *	The file handle passed in is locked and the socket it is bound
532  *	to is returned. If an error occurs the err pointer is overwritten
533  *	with a negative errno code and NULL is returned. The function checks
534  *	for both invalid handles and passing a handle which is not a socket.
535  *
536  *	On a success the socket object pointer is returned.
537  */
538 
sockfd_lookup(int fd,int * err)539 struct socket *sockfd_lookup(int fd, int *err)
540 {
541 	struct file *file;
542 	struct socket *sock;
543 
544 	file = fget(fd);
545 	if (!file) {
546 		*err = -EBADF;
547 		return NULL;
548 	}
549 
550 	sock = sock_from_file(file);
551 	if (!sock) {
552 		*err = -ENOTSOCK;
553 		fput(file);
554 	}
555 	return sock;
556 }
557 EXPORT_SYMBOL(sockfd_lookup);
558 
sockfs_listxattr(struct dentry * dentry,char * buffer,size_t size)559 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
560 				size_t size)
561 {
562 	ssize_t len;
563 	ssize_t used = 0;
564 
565 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
566 	if (len < 0)
567 		return len;
568 	used += len;
569 	if (buffer) {
570 		if (size < used)
571 			return -ERANGE;
572 		buffer += len;
573 	}
574 
575 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
576 	used += len;
577 	if (buffer) {
578 		if (size < used)
579 			return -ERANGE;
580 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
581 		buffer += len;
582 	}
583 
584 	return used;
585 }
586 
sockfs_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * iattr)587 static int sockfs_setattr(struct mnt_idmap *idmap,
588 			  struct dentry *dentry, struct iattr *iattr)
589 {
590 	int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
591 
592 	if (!err && (iattr->ia_valid & ATTR_UID)) {
593 		struct socket *sock = SOCKET_I(d_inode(dentry));
594 
595 		if (sock->sk)
596 			sock->sk->sk_uid = iattr->ia_uid;
597 		else
598 			err = -ENOENT;
599 	}
600 
601 	return err;
602 }
603 
604 static const struct inode_operations sockfs_inode_ops = {
605 	.listxattr = sockfs_listxattr,
606 	.setattr = sockfs_setattr,
607 };
608 
609 /**
610  *	sock_alloc - allocate a socket
611  *
612  *	Allocate a new inode and socket object. The two are bound together
613  *	and initialised. The socket is then returned. If we are out of inodes
614  *	NULL is returned. This functions uses GFP_KERNEL internally.
615  */
616 
sock_alloc(void)617 struct socket *sock_alloc(void)
618 {
619 	struct inode *inode;
620 	struct socket *sock;
621 
622 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
623 	if (!inode)
624 		return NULL;
625 
626 	sock = SOCKET_I(inode);
627 
628 	inode->i_ino = get_next_ino();
629 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
630 	inode->i_uid = current_fsuid();
631 	inode->i_gid = current_fsgid();
632 	inode->i_op = &sockfs_inode_ops;
633 
634 	return sock;
635 }
636 EXPORT_SYMBOL(sock_alloc);
637 
__sock_release(struct socket * sock,struct inode * inode)638 static void __sock_release(struct socket *sock, struct inode *inode)
639 {
640 	const struct proto_ops *ops = READ_ONCE(sock->ops);
641 
642 	if (ops) {
643 		struct module *owner = ops->owner;
644 
645 		if (inode)
646 			inode_lock(inode);
647 		ops->release(sock);
648 		sock->sk = NULL;
649 		if (inode)
650 			inode_unlock(inode);
651 		sock->ops = NULL;
652 		module_put(owner);
653 	}
654 
655 	if (sock->wq.fasync_list)
656 		pr_err("%s: fasync list not empty!\n", __func__);
657 
658 	if (!sock->file) {
659 		iput(SOCK_INODE(sock));
660 		return;
661 	}
662 	sock->file = NULL;
663 }
664 
665 /**
666  *	sock_release - close a socket
667  *	@sock: socket to close
668  *
669  *	The socket is released from the protocol stack if it has a release
670  *	callback, and the inode is then released if the socket is bound to
671  *	an inode not a file.
672  */
sock_release(struct socket * sock)673 void sock_release(struct socket *sock)
674 {
675 	__sock_release(sock, NULL);
676 }
677 EXPORT_SYMBOL(sock_release);
678 
__sock_tx_timestamp(__u32 tsflags,__u8 * tx_flags)679 void __sock_tx_timestamp(__u32 tsflags, __u8 *tx_flags)
680 {
681 	u8 flags = *tx_flags;
682 
683 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
684 		flags |= SKBTX_HW_TSTAMP_NOBPF;
685 
686 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
687 		flags |= SKBTX_SW_TSTAMP;
688 
689 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
690 		flags |= SKBTX_SCHED_TSTAMP;
691 
692 	if (tsflags & SOF_TIMESTAMPING_TX_COMPLETION)
693 		flags |= SKBTX_COMPLETION_TSTAMP;
694 
695 	*tx_flags = flags;
696 }
697 EXPORT_SYMBOL(__sock_tx_timestamp);
698 
699 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
700 					   size_t));
701 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
702 					    size_t));
703 
call_trace_sock_send_length(struct sock * sk,int ret,int flags)704 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
705 						 int flags)
706 {
707 	trace_sock_send_length(sk, ret, 0);
708 }
709 
sock_sendmsg_nosec(struct socket * sock,struct msghdr * msg)710 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
711 {
712 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
713 				     inet_sendmsg, sock, msg,
714 				     msg_data_left(msg));
715 	BUG_ON(ret == -EIOCBQUEUED);
716 
717 	if (trace_sock_send_length_enabled())
718 		call_trace_sock_send_length(sock->sk, ret, 0);
719 	return ret;
720 }
721 
__sock_sendmsg(struct socket * sock,struct msghdr * msg)722 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
723 {
724 	int err = security_socket_sendmsg(sock, msg,
725 					  msg_data_left(msg));
726 
727 	return err ?: sock_sendmsg_nosec(sock, msg);
728 }
729 
730 /**
731  *	sock_sendmsg - send a message through @sock
732  *	@sock: socket
733  *	@msg: message to send
734  *
735  *	Sends @msg through @sock, passing through LSM.
736  *	Returns the number of bytes sent, or an error code.
737  */
sock_sendmsg(struct socket * sock,struct msghdr * msg)738 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
739 {
740 	struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
741 	struct sockaddr_storage address;
742 	int save_len = msg->msg_namelen;
743 	int ret;
744 
745 	if (msg->msg_name) {
746 		memcpy(&address, msg->msg_name, msg->msg_namelen);
747 		msg->msg_name = &address;
748 	}
749 
750 	ret = __sock_sendmsg(sock, msg);
751 	msg->msg_name = save_addr;
752 	msg->msg_namelen = save_len;
753 
754 	return ret;
755 }
756 EXPORT_SYMBOL(sock_sendmsg);
757 
758 /**
759  *	kernel_sendmsg - send a message through @sock (kernel-space)
760  *	@sock: socket
761  *	@msg: message header
762  *	@vec: kernel vec
763  *	@num: vec array length
764  *	@size: total message data size
765  *
766  *	Builds the message data with @vec and sends it through @sock.
767  *	Returns the number of bytes sent, or an error code.
768  */
769 
kernel_sendmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size)770 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
771 		   struct kvec *vec, size_t num, size_t size)
772 {
773 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
774 	return sock_sendmsg(sock, msg);
775 }
776 EXPORT_SYMBOL(kernel_sendmsg);
777 
skb_is_err_queue(const struct sk_buff * skb)778 static bool skb_is_err_queue(const struct sk_buff *skb)
779 {
780 	/* pkt_type of skbs enqueued on the error queue are set to
781 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
782 	 * in recvmsg, since skbs received on a local socket will never
783 	 * have a pkt_type of PACKET_OUTGOING.
784 	 */
785 	return skb->pkt_type == PACKET_OUTGOING;
786 }
787 
788 /* On transmit, software and hardware timestamps are returned independently.
789  * As the two skb clones share the hardware timestamp, which may be updated
790  * before the software timestamp is received, a hardware TX timestamp may be
791  * returned only if there is no software TX timestamp. Ignore false software
792  * timestamps, which may be made in the __sock_recv_timestamp() call when the
793  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
794  * hardware timestamp.
795  */
skb_is_swtx_tstamp(const struct sk_buff * skb,int false_tstamp)796 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
797 {
798 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
799 }
800 
get_timestamp(struct sock * sk,struct sk_buff * skb,int * if_index)801 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
802 {
803 	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
804 	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
805 	struct net_device *orig_dev;
806 	ktime_t hwtstamp;
807 
808 	rcu_read_lock();
809 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
810 	if (orig_dev) {
811 		*if_index = orig_dev->ifindex;
812 		hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
813 	} else {
814 		hwtstamp = shhwtstamps->hwtstamp;
815 	}
816 	rcu_read_unlock();
817 
818 	return hwtstamp;
819 }
820 
put_ts_pktinfo(struct msghdr * msg,struct sk_buff * skb,int if_index)821 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
822 			   int if_index)
823 {
824 	struct scm_ts_pktinfo ts_pktinfo;
825 	struct net_device *orig_dev;
826 
827 	if (!skb_mac_header_was_set(skb))
828 		return;
829 
830 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
831 
832 	if (!if_index) {
833 		rcu_read_lock();
834 		orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
835 		if (orig_dev)
836 			if_index = orig_dev->ifindex;
837 		rcu_read_unlock();
838 	}
839 	ts_pktinfo.if_index = if_index;
840 
841 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
842 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
843 		 sizeof(ts_pktinfo), &ts_pktinfo);
844 }
845 
846 /*
847  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
848  */
__sock_recv_timestamp(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)849 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
850 	struct sk_buff *skb)
851 {
852 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
853 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
854 	struct scm_timestamping_internal tss;
855 	int empty = 1, false_tstamp = 0;
856 	struct skb_shared_hwtstamps *shhwtstamps =
857 		skb_hwtstamps(skb);
858 	int if_index;
859 	ktime_t hwtstamp;
860 	u32 tsflags;
861 
862 	/* Race occurred between timestamp enabling and packet
863 	   receiving.  Fill in the current time for now. */
864 	if (need_software_tstamp && skb->tstamp == 0) {
865 		__net_timestamp(skb);
866 		false_tstamp = 1;
867 	}
868 
869 	if (need_software_tstamp) {
870 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
871 			if (new_tstamp) {
872 				struct __kernel_sock_timeval tv;
873 
874 				skb_get_new_timestamp(skb, &tv);
875 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
876 					 sizeof(tv), &tv);
877 			} else {
878 				struct __kernel_old_timeval tv;
879 
880 				skb_get_timestamp(skb, &tv);
881 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
882 					 sizeof(tv), &tv);
883 			}
884 		} else {
885 			if (new_tstamp) {
886 				struct __kernel_timespec ts;
887 
888 				skb_get_new_timestampns(skb, &ts);
889 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
890 					 sizeof(ts), &ts);
891 			} else {
892 				struct __kernel_old_timespec ts;
893 
894 				skb_get_timestampns(skb, &ts);
895 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
896 					 sizeof(ts), &ts);
897 			}
898 		}
899 	}
900 
901 	memset(&tss, 0, sizeof(tss));
902 	tsflags = READ_ONCE(sk->sk_tsflags);
903 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
904 	     (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE ||
905 	      skb_is_err_queue(skb) ||
906 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
907 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
908 		empty = 0;
909 	if (shhwtstamps &&
910 	    (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
911 	     (tsflags & SOF_TIMESTAMPING_RX_HARDWARE ||
912 	      skb_is_err_queue(skb) ||
913 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
914 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
915 		if_index = 0;
916 		if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
917 			hwtstamp = get_timestamp(sk, skb, &if_index);
918 		else
919 			hwtstamp = shhwtstamps->hwtstamp;
920 
921 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
922 			hwtstamp = ptp_convert_timestamp(&hwtstamp,
923 							 READ_ONCE(sk->sk_bind_phc));
924 
925 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
926 			empty = 0;
927 
928 			if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
929 			    !skb_is_err_queue(skb))
930 				put_ts_pktinfo(msg, skb, if_index);
931 		}
932 	}
933 	if (!empty) {
934 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
935 			put_cmsg_scm_timestamping64(msg, &tss);
936 		else
937 			put_cmsg_scm_timestamping(msg, &tss);
938 
939 		if (skb_is_err_queue(skb) && skb->len &&
940 		    SKB_EXT_ERR(skb)->opt_stats)
941 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
942 				 skb->len, skb->data);
943 	}
944 }
945 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
946 
947 #ifdef CONFIG_WIRELESS
__sock_recv_wifi_status(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)948 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
949 	struct sk_buff *skb)
950 {
951 	int ack;
952 
953 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
954 		return;
955 	if (!skb->wifi_acked_valid)
956 		return;
957 
958 	ack = skb->wifi_acked;
959 
960 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
961 }
962 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
963 #endif
964 
sock_recv_drops(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)965 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
966 				   struct sk_buff *skb)
967 {
968 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
969 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
970 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
971 }
972 
sock_recv_mark(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)973 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
974 			   struct sk_buff *skb)
975 {
976 	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
977 		/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
978 		__u32 mark = skb->mark;
979 
980 		put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
981 	}
982 }
983 
sock_recv_priority(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)984 static void sock_recv_priority(struct msghdr *msg, struct sock *sk,
985 			       struct sk_buff *skb)
986 {
987 	if (sock_flag(sk, SOCK_RCVPRIORITY) && skb) {
988 		__u32 priority = skb->priority;
989 
990 		put_cmsg(msg, SOL_SOCKET, SO_PRIORITY, sizeof(__u32), &priority);
991 	}
992 }
993 
__sock_recv_cmsgs(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)994 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
995 		       struct sk_buff *skb)
996 {
997 	sock_recv_timestamp(msg, sk, skb);
998 	sock_recv_drops(msg, sk, skb);
999 	sock_recv_mark(msg, sk, skb);
1000 	sock_recv_priority(msg, sk, skb);
1001 }
1002 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1003 
1004 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1005 					   size_t, int));
1006 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1007 					    size_t, int));
1008 
call_trace_sock_recv_length(struct sock * sk,int ret,int flags)1009 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1010 {
1011 	trace_sock_recv_length(sk, ret, flags);
1012 }
1013 
sock_recvmsg_nosec(struct socket * sock,struct msghdr * msg,int flags)1014 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1015 				     int flags)
1016 {
1017 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1018 				     inet6_recvmsg,
1019 				     inet_recvmsg, sock, msg,
1020 				     msg_data_left(msg), flags);
1021 	if (trace_sock_recv_length_enabled())
1022 		call_trace_sock_recv_length(sock->sk, ret, flags);
1023 	return ret;
1024 }
1025 
1026 /**
1027  *	sock_recvmsg - receive a message from @sock
1028  *	@sock: socket
1029  *	@msg: message to receive
1030  *	@flags: message flags
1031  *
1032  *	Receives @msg from @sock, passing through LSM. Returns the total number
1033  *	of bytes received, or an error.
1034  */
sock_recvmsg(struct socket * sock,struct msghdr * msg,int flags)1035 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1036 {
1037 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1038 
1039 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1040 }
1041 EXPORT_SYMBOL(sock_recvmsg);
1042 
1043 /**
1044  *	kernel_recvmsg - Receive a message from a socket (kernel space)
1045  *	@sock: The socket to receive the message from
1046  *	@msg: Received message
1047  *	@vec: Input s/g array for message data
1048  *	@num: Size of input s/g array
1049  *	@size: Number of bytes to read
1050  *	@flags: Message flags (MSG_DONTWAIT, etc...)
1051  *
1052  *	On return the msg structure contains the scatter/gather array passed in the
1053  *	vec argument. The array is modified so that it consists of the unfilled
1054  *	portion of the original array.
1055  *
1056  *	The returned value is the total number of bytes received, or an error.
1057  */
1058 
kernel_recvmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size,int flags)1059 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1060 		   struct kvec *vec, size_t num, size_t size, int flags)
1061 {
1062 	msg->msg_control_is_user = false;
1063 	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1064 	return sock_recvmsg(sock, msg, flags);
1065 }
1066 EXPORT_SYMBOL(kernel_recvmsg);
1067 
sock_splice_read(struct file * file,loff_t * ppos,struct pipe_inode_info * pipe,size_t len,unsigned int flags)1068 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1069 				struct pipe_inode_info *pipe, size_t len,
1070 				unsigned int flags)
1071 {
1072 	struct socket *sock = file->private_data;
1073 	const struct proto_ops *ops;
1074 
1075 	ops = READ_ONCE(sock->ops);
1076 	if (unlikely(!ops->splice_read))
1077 		return copy_splice_read(file, ppos, pipe, len, flags);
1078 
1079 	return ops->splice_read(sock, ppos, pipe, len, flags);
1080 }
1081 
sock_splice_eof(struct file * file)1082 static void sock_splice_eof(struct file *file)
1083 {
1084 	struct socket *sock = file->private_data;
1085 	const struct proto_ops *ops;
1086 
1087 	ops = READ_ONCE(sock->ops);
1088 	if (ops->splice_eof)
1089 		ops->splice_eof(sock);
1090 }
1091 
sock_read_iter(struct kiocb * iocb,struct iov_iter * to)1092 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1093 {
1094 	struct file *file = iocb->ki_filp;
1095 	struct socket *sock = file->private_data;
1096 	struct msghdr msg = {.msg_iter = *to,
1097 			     .msg_iocb = iocb};
1098 	ssize_t res;
1099 
1100 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1101 		msg.msg_flags = MSG_DONTWAIT;
1102 
1103 	if (iocb->ki_pos != 0)
1104 		return -ESPIPE;
1105 
1106 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1107 		return 0;
1108 
1109 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1110 	*to = msg.msg_iter;
1111 	return res;
1112 }
1113 
sock_write_iter(struct kiocb * iocb,struct iov_iter * from)1114 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1115 {
1116 	struct file *file = iocb->ki_filp;
1117 	struct socket *sock = file->private_data;
1118 	struct msghdr msg = {.msg_iter = *from,
1119 			     .msg_iocb = iocb};
1120 	ssize_t res;
1121 
1122 	if (iocb->ki_pos != 0)
1123 		return -ESPIPE;
1124 
1125 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1126 		msg.msg_flags = MSG_DONTWAIT;
1127 
1128 	if (sock->type == SOCK_SEQPACKET)
1129 		msg.msg_flags |= MSG_EOR;
1130 
1131 	res = __sock_sendmsg(sock, &msg);
1132 	*from = msg.msg_iter;
1133 	return res;
1134 }
1135 
1136 /*
1137  * Atomic setting of ioctl hooks to avoid race
1138  * with module unload.
1139  */
1140 
1141 static DEFINE_MUTEX(br_ioctl_mutex);
1142 static int (*br_ioctl_hook)(struct net *net, unsigned int cmd,
1143 			    void __user *uarg);
1144 
brioctl_set(int (* hook)(struct net * net,unsigned int cmd,void __user * uarg))1145 void brioctl_set(int (*hook)(struct net *net, unsigned int cmd,
1146 			     void __user *uarg))
1147 {
1148 	mutex_lock(&br_ioctl_mutex);
1149 	br_ioctl_hook = hook;
1150 	mutex_unlock(&br_ioctl_mutex);
1151 }
1152 EXPORT_SYMBOL(brioctl_set);
1153 
br_ioctl_call(struct net * net,unsigned int cmd,void __user * uarg)1154 int br_ioctl_call(struct net *net, unsigned int cmd, void __user *uarg)
1155 {
1156 	int err = -ENOPKG;
1157 
1158 	if (!br_ioctl_hook)
1159 		request_module("bridge");
1160 
1161 	mutex_lock(&br_ioctl_mutex);
1162 	if (br_ioctl_hook)
1163 		err = br_ioctl_hook(net, cmd, uarg);
1164 	mutex_unlock(&br_ioctl_mutex);
1165 
1166 	return err;
1167 }
1168 
1169 static DEFINE_MUTEX(vlan_ioctl_mutex);
1170 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1171 
vlan_ioctl_set(int (* hook)(struct net *,void __user *))1172 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1173 {
1174 	mutex_lock(&vlan_ioctl_mutex);
1175 	vlan_ioctl_hook = hook;
1176 	mutex_unlock(&vlan_ioctl_mutex);
1177 }
1178 EXPORT_SYMBOL(vlan_ioctl_set);
1179 
sock_do_ioctl(struct net * net,struct socket * sock,unsigned int cmd,unsigned long arg)1180 static long sock_do_ioctl(struct net *net, struct socket *sock,
1181 			  unsigned int cmd, unsigned long arg)
1182 {
1183 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1184 	struct ifreq ifr;
1185 	bool need_copyout;
1186 	int err;
1187 	void __user *argp = (void __user *)arg;
1188 	void __user *data;
1189 
1190 	err = ops->ioctl(sock, cmd, arg);
1191 
1192 	/*
1193 	 * If this ioctl is unknown try to hand it down
1194 	 * to the NIC driver.
1195 	 */
1196 	if (err != -ENOIOCTLCMD)
1197 		return err;
1198 
1199 	if (!is_socket_ioctl_cmd(cmd))
1200 		return -ENOTTY;
1201 
1202 	if (get_user_ifreq(&ifr, &data, argp))
1203 		return -EFAULT;
1204 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1205 	if (!err && need_copyout)
1206 		if (put_user_ifreq(&ifr, argp))
1207 			return -EFAULT;
1208 
1209 	return err;
1210 }
1211 
1212 /*
1213  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1214  *	what to do with it - that's up to the protocol still.
1215  */
1216 
sock_ioctl(struct file * file,unsigned cmd,unsigned long arg)1217 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1218 {
1219 	const struct proto_ops  *ops;
1220 	struct socket *sock;
1221 	struct sock *sk;
1222 	void __user *argp = (void __user *)arg;
1223 	int pid, err;
1224 	struct net *net;
1225 
1226 	sock = file->private_data;
1227 	ops = READ_ONCE(sock->ops);
1228 	sk = sock->sk;
1229 	net = sock_net(sk);
1230 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1231 		struct ifreq ifr;
1232 		void __user *data;
1233 		bool need_copyout;
1234 		if (get_user_ifreq(&ifr, &data, argp))
1235 			return -EFAULT;
1236 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1237 		if (!err && need_copyout)
1238 			if (put_user_ifreq(&ifr, argp))
1239 				return -EFAULT;
1240 	} else
1241 #ifdef CONFIG_WEXT_CORE
1242 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1243 		err = wext_handle_ioctl(net, cmd, argp);
1244 	} else
1245 #endif
1246 		switch (cmd) {
1247 		case FIOSETOWN:
1248 		case SIOCSPGRP:
1249 			err = -EFAULT;
1250 			if (get_user(pid, (int __user *)argp))
1251 				break;
1252 			err = f_setown(sock->file, pid, 1);
1253 			break;
1254 		case FIOGETOWN:
1255 		case SIOCGPGRP:
1256 			err = put_user(f_getown(sock->file),
1257 				       (int __user *)argp);
1258 			break;
1259 		case SIOCGIFBR:
1260 		case SIOCSIFBR:
1261 		case SIOCBRADDBR:
1262 		case SIOCBRDELBR:
1263 		case SIOCBRADDIF:
1264 		case SIOCBRDELIF:
1265 			err = br_ioctl_call(net, cmd, argp);
1266 			break;
1267 		case SIOCGIFVLAN:
1268 		case SIOCSIFVLAN:
1269 			err = -ENOPKG;
1270 			if (!vlan_ioctl_hook)
1271 				request_module("8021q");
1272 
1273 			mutex_lock(&vlan_ioctl_mutex);
1274 			if (vlan_ioctl_hook)
1275 				err = vlan_ioctl_hook(net, argp);
1276 			mutex_unlock(&vlan_ioctl_mutex);
1277 			break;
1278 		case SIOCGSKNS:
1279 			err = -EPERM;
1280 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1281 				break;
1282 
1283 			err = open_related_ns(&net->ns, get_net_ns);
1284 			break;
1285 		case SIOCGSTAMP_OLD:
1286 		case SIOCGSTAMPNS_OLD:
1287 			if (!ops->gettstamp) {
1288 				err = -ENOIOCTLCMD;
1289 				break;
1290 			}
1291 			err = ops->gettstamp(sock, argp,
1292 					     cmd == SIOCGSTAMP_OLD,
1293 					     !IS_ENABLED(CONFIG_64BIT));
1294 			break;
1295 		case SIOCGSTAMP_NEW:
1296 		case SIOCGSTAMPNS_NEW:
1297 			if (!ops->gettstamp) {
1298 				err = -ENOIOCTLCMD;
1299 				break;
1300 			}
1301 			err = ops->gettstamp(sock, argp,
1302 					     cmd == SIOCGSTAMP_NEW,
1303 					     false);
1304 			break;
1305 
1306 		case SIOCGIFCONF:
1307 			err = dev_ifconf(net, argp);
1308 			break;
1309 
1310 		default:
1311 			err = sock_do_ioctl(net, sock, cmd, arg);
1312 			break;
1313 		}
1314 	return err;
1315 }
1316 
1317 /**
1318  *	sock_create_lite - creates a socket
1319  *	@family: protocol family (AF_INET, ...)
1320  *	@type: communication type (SOCK_STREAM, ...)
1321  *	@protocol: protocol (0, ...)
1322  *	@res: new socket
1323  *
1324  *	Creates a new socket and assigns it to @res, passing through LSM.
1325  *	The new socket initialization is not complete, see kernel_accept().
1326  *	Returns 0 or an error. On failure @res is set to %NULL.
1327  *	This function internally uses GFP_KERNEL.
1328  */
1329 
sock_create_lite(int family,int type,int protocol,struct socket ** res)1330 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1331 {
1332 	int err;
1333 	struct socket *sock = NULL;
1334 
1335 	err = security_socket_create(family, type, protocol, 1);
1336 	if (err)
1337 		goto out;
1338 
1339 	sock = sock_alloc();
1340 	if (!sock) {
1341 		err = -ENOMEM;
1342 		goto out;
1343 	}
1344 
1345 	sock->type = type;
1346 	err = security_socket_post_create(sock, family, type, protocol, 1);
1347 	if (err)
1348 		goto out_release;
1349 
1350 out:
1351 	*res = sock;
1352 	return err;
1353 out_release:
1354 	sock_release(sock);
1355 	sock = NULL;
1356 	goto out;
1357 }
1358 EXPORT_SYMBOL(sock_create_lite);
1359 
1360 /* No kernel lock held - perfect */
sock_poll(struct file * file,poll_table * wait)1361 static __poll_t sock_poll(struct file *file, poll_table *wait)
1362 {
1363 	struct socket *sock = file->private_data;
1364 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1365 	__poll_t events = poll_requested_events(wait), flag = 0;
1366 
1367 	if (!ops->poll)
1368 		return 0;
1369 
1370 	if (sk_can_busy_loop(sock->sk)) {
1371 		/* poll once if requested by the syscall */
1372 		if (events & POLL_BUSY_LOOP)
1373 			sk_busy_loop(sock->sk, 1);
1374 
1375 		/* if this socket can poll_ll, tell the system call */
1376 		flag = POLL_BUSY_LOOP;
1377 	}
1378 
1379 	return ops->poll(file, sock, wait) | flag;
1380 }
1381 
sock_mmap(struct file * file,struct vm_area_struct * vma)1382 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1383 {
1384 	struct socket *sock = file->private_data;
1385 
1386 	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1387 }
1388 
sock_close(struct inode * inode,struct file * filp)1389 static int sock_close(struct inode *inode, struct file *filp)
1390 {
1391 	__sock_release(SOCKET_I(inode), inode);
1392 	return 0;
1393 }
1394 
1395 /*
1396  *	Update the socket async list
1397  *
1398  *	Fasync_list locking strategy.
1399  *
1400  *	1. fasync_list is modified only under process context socket lock
1401  *	   i.e. under semaphore.
1402  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1403  *	   or under socket lock
1404  */
1405 
sock_fasync(int fd,struct file * filp,int on)1406 static int sock_fasync(int fd, struct file *filp, int on)
1407 {
1408 	struct socket *sock = filp->private_data;
1409 	struct sock *sk = sock->sk;
1410 	struct socket_wq *wq = &sock->wq;
1411 
1412 	if (sk == NULL)
1413 		return -EINVAL;
1414 
1415 	lock_sock(sk);
1416 	fasync_helper(fd, filp, on, &wq->fasync_list);
1417 
1418 	if (!wq->fasync_list)
1419 		sock_reset_flag(sk, SOCK_FASYNC);
1420 	else
1421 		sock_set_flag(sk, SOCK_FASYNC);
1422 
1423 	release_sock(sk);
1424 	return 0;
1425 }
1426 
1427 /* This function may be called only under rcu_lock */
1428 
sock_wake_async(struct socket_wq * wq,int how,int band)1429 int sock_wake_async(struct socket_wq *wq, int how, int band)
1430 {
1431 	if (!wq || !wq->fasync_list)
1432 		return -1;
1433 
1434 	switch (how) {
1435 	case SOCK_WAKE_WAITD:
1436 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1437 			break;
1438 		goto call_kill;
1439 	case SOCK_WAKE_SPACE:
1440 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1441 			break;
1442 		fallthrough;
1443 	case SOCK_WAKE_IO:
1444 call_kill:
1445 		kill_fasync(&wq->fasync_list, SIGIO, band);
1446 		break;
1447 	case SOCK_WAKE_URG:
1448 		kill_fasync(&wq->fasync_list, SIGURG, band);
1449 	}
1450 
1451 	return 0;
1452 }
1453 EXPORT_SYMBOL(sock_wake_async);
1454 
1455 /**
1456  *	__sock_create - creates a socket
1457  *	@net: net namespace
1458  *	@family: protocol family (AF_INET, ...)
1459  *	@type: communication type (SOCK_STREAM, ...)
1460  *	@protocol: protocol (0, ...)
1461  *	@res: new socket
1462  *	@kern: boolean for kernel space sockets
1463  *
1464  *	Creates a new socket and assigns it to @res, passing through LSM.
1465  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1466  *	be set to true if the socket resides in kernel space.
1467  *	This function internally uses GFP_KERNEL.
1468  */
1469 
__sock_create(struct net * net,int family,int type,int protocol,struct socket ** res,int kern)1470 int __sock_create(struct net *net, int family, int type, int protocol,
1471 			 struct socket **res, int kern)
1472 {
1473 	int err;
1474 	struct socket *sock;
1475 	const struct net_proto_family *pf;
1476 
1477 	/*
1478 	 *      Check protocol is in range
1479 	 */
1480 	if (family < 0 || family >= NPROTO)
1481 		return -EAFNOSUPPORT;
1482 	if (type < 0 || type >= SOCK_MAX)
1483 		return -EINVAL;
1484 
1485 	/* Compatibility.
1486 
1487 	   This uglymoron is moved from INET layer to here to avoid
1488 	   deadlock in module load.
1489 	 */
1490 	if (family == PF_INET && type == SOCK_PACKET) {
1491 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1492 			     current->comm);
1493 		family = PF_PACKET;
1494 	}
1495 
1496 	err = security_socket_create(family, type, protocol, kern);
1497 	if (err)
1498 		return err;
1499 
1500 	/*
1501 	 *	Allocate the socket and allow the family to set things up. if
1502 	 *	the protocol is 0, the family is instructed to select an appropriate
1503 	 *	default.
1504 	 */
1505 	sock = sock_alloc();
1506 	if (!sock) {
1507 		net_warn_ratelimited("socket: no more sockets\n");
1508 		return -ENFILE;	/* Not exactly a match, but its the
1509 				   closest posix thing */
1510 	}
1511 
1512 	sock->type = type;
1513 
1514 #ifdef CONFIG_MODULES
1515 	/* Attempt to load a protocol module if the find failed.
1516 	 *
1517 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1518 	 * requested real, full-featured networking support upon configuration.
1519 	 * Otherwise module support will break!
1520 	 */
1521 	if (rcu_access_pointer(net_families[family]) == NULL)
1522 		request_module("net-pf-%d", family);
1523 #endif
1524 
1525 	rcu_read_lock();
1526 	pf = rcu_dereference(net_families[family]);
1527 	err = -EAFNOSUPPORT;
1528 	if (!pf)
1529 		goto out_release;
1530 
1531 	/*
1532 	 * We will call the ->create function, that possibly is in a loadable
1533 	 * module, so we have to bump that loadable module refcnt first.
1534 	 */
1535 	if (!try_module_get(pf->owner))
1536 		goto out_release;
1537 
1538 	/* Now protected by module ref count */
1539 	rcu_read_unlock();
1540 
1541 	err = pf->create(net, sock, protocol, kern);
1542 	if (err < 0) {
1543 		/* ->create should release the allocated sock->sk object on error
1544 		 * and make sure sock->sk is set to NULL to avoid use-after-free
1545 		 */
1546 		DEBUG_NET_WARN_ONCE(sock->sk,
1547 				    "%ps must clear sock->sk on failure, family: %d, type: %d, protocol: %d\n",
1548 				    pf->create, family, type, protocol);
1549 		goto out_module_put;
1550 	}
1551 
1552 	/*
1553 	 * Now to bump the refcnt of the [loadable] module that owns this
1554 	 * socket at sock_release time we decrement its refcnt.
1555 	 */
1556 	if (!try_module_get(sock->ops->owner))
1557 		goto out_module_busy;
1558 
1559 	/*
1560 	 * Now that we're done with the ->create function, the [loadable]
1561 	 * module can have its refcnt decremented
1562 	 */
1563 	module_put(pf->owner);
1564 	err = security_socket_post_create(sock, family, type, protocol, kern);
1565 	if (err)
1566 		goto out_sock_release;
1567 	*res = sock;
1568 
1569 	return 0;
1570 
1571 out_module_busy:
1572 	err = -EAFNOSUPPORT;
1573 out_module_put:
1574 	sock->ops = NULL;
1575 	module_put(pf->owner);
1576 out_sock_release:
1577 	sock_release(sock);
1578 	return err;
1579 
1580 out_release:
1581 	rcu_read_unlock();
1582 	goto out_sock_release;
1583 }
1584 EXPORT_SYMBOL(__sock_create);
1585 
1586 /**
1587  *	sock_create - creates a socket
1588  *	@family: protocol family (AF_INET, ...)
1589  *	@type: communication type (SOCK_STREAM, ...)
1590  *	@protocol: protocol (0, ...)
1591  *	@res: new socket
1592  *
1593  *	A wrapper around __sock_create().
1594  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1595  */
1596 
sock_create(int family,int type,int protocol,struct socket ** res)1597 int sock_create(int family, int type, int protocol, struct socket **res)
1598 {
1599 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1600 }
1601 EXPORT_SYMBOL(sock_create);
1602 
1603 /**
1604  *	sock_create_kern - creates a socket (kernel space)
1605  *	@net: net namespace
1606  *	@family: protocol family (AF_INET, ...)
1607  *	@type: communication type (SOCK_STREAM, ...)
1608  *	@protocol: protocol (0, ...)
1609  *	@res: new socket
1610  *
1611  *	A wrapper around __sock_create().
1612  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1613  */
1614 
sock_create_kern(struct net * net,int family,int type,int protocol,struct socket ** res)1615 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1616 {
1617 	return __sock_create(net, family, type, protocol, res, 1);
1618 }
1619 EXPORT_SYMBOL(sock_create_kern);
1620 
__sys_socket_create(int family,int type,int protocol)1621 static struct socket *__sys_socket_create(int family, int type, int protocol)
1622 {
1623 	struct socket *sock;
1624 	int retval;
1625 
1626 	/* Check the SOCK_* constants for consistency.  */
1627 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1628 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1629 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1630 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1631 
1632 	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1633 		return ERR_PTR(-EINVAL);
1634 	type &= SOCK_TYPE_MASK;
1635 
1636 	retval = sock_create(family, type, protocol, &sock);
1637 	if (retval < 0)
1638 		return ERR_PTR(retval);
1639 
1640 	return sock;
1641 }
1642 
__sys_socket_file(int family,int type,int protocol)1643 struct file *__sys_socket_file(int family, int type, int protocol)
1644 {
1645 	struct socket *sock;
1646 	int flags;
1647 
1648 	sock = __sys_socket_create(family, type, protocol);
1649 	if (IS_ERR(sock))
1650 		return ERR_CAST(sock);
1651 
1652 	flags = type & ~SOCK_TYPE_MASK;
1653 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1654 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1655 
1656 	return sock_alloc_file(sock, flags, NULL);
1657 }
1658 
1659 /*	A hook for bpf progs to attach to and update socket protocol.
1660  *
1661  *	A static noinline declaration here could cause the compiler to
1662  *	optimize away the function. A global noinline declaration will
1663  *	keep the definition, but may optimize away the callsite.
1664  *	Therefore, __weak is needed to ensure that the call is still
1665  *	emitted, by telling the compiler that we don't know what the
1666  *	function might eventually be.
1667  */
1668 
1669 __bpf_hook_start();
1670 
update_socket_protocol(int family,int type,int protocol)1671 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1672 {
1673 	return protocol;
1674 }
1675 
1676 __bpf_hook_end();
1677 
__sys_socket(int family,int type,int protocol)1678 int __sys_socket(int family, int type, int protocol)
1679 {
1680 	struct socket *sock;
1681 	int flags;
1682 
1683 	sock = __sys_socket_create(family, type,
1684 				   update_socket_protocol(family, type, protocol));
1685 	if (IS_ERR(sock))
1686 		return PTR_ERR(sock);
1687 
1688 	flags = type & ~SOCK_TYPE_MASK;
1689 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1690 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1691 
1692 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1693 }
1694 
SYSCALL_DEFINE3(socket,int,family,int,type,int,protocol)1695 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1696 {
1697 	return __sys_socket(family, type, protocol);
1698 }
1699 
1700 /*
1701  *	Create a pair of connected sockets.
1702  */
1703 
__sys_socketpair(int family,int type,int protocol,int __user * usockvec)1704 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1705 {
1706 	struct socket *sock1, *sock2;
1707 	int fd1, fd2, err;
1708 	struct file *newfile1, *newfile2;
1709 	int flags;
1710 
1711 	flags = type & ~SOCK_TYPE_MASK;
1712 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1713 		return -EINVAL;
1714 	type &= SOCK_TYPE_MASK;
1715 
1716 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1717 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1718 
1719 	/*
1720 	 * reserve descriptors and make sure we won't fail
1721 	 * to return them to userland.
1722 	 */
1723 	fd1 = get_unused_fd_flags(flags);
1724 	if (unlikely(fd1 < 0))
1725 		return fd1;
1726 
1727 	fd2 = get_unused_fd_flags(flags);
1728 	if (unlikely(fd2 < 0)) {
1729 		put_unused_fd(fd1);
1730 		return fd2;
1731 	}
1732 
1733 	err = put_user(fd1, &usockvec[0]);
1734 	if (err)
1735 		goto out;
1736 
1737 	err = put_user(fd2, &usockvec[1]);
1738 	if (err)
1739 		goto out;
1740 
1741 	/*
1742 	 * Obtain the first socket and check if the underlying protocol
1743 	 * supports the socketpair call.
1744 	 */
1745 
1746 	err = sock_create(family, type, protocol, &sock1);
1747 	if (unlikely(err < 0))
1748 		goto out;
1749 
1750 	err = sock_create(family, type, protocol, &sock2);
1751 	if (unlikely(err < 0)) {
1752 		sock_release(sock1);
1753 		goto out;
1754 	}
1755 
1756 	err = security_socket_socketpair(sock1, sock2);
1757 	if (unlikely(err)) {
1758 		sock_release(sock2);
1759 		sock_release(sock1);
1760 		goto out;
1761 	}
1762 
1763 	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1764 	if (unlikely(err < 0)) {
1765 		sock_release(sock2);
1766 		sock_release(sock1);
1767 		goto out;
1768 	}
1769 
1770 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1771 	if (IS_ERR(newfile1)) {
1772 		err = PTR_ERR(newfile1);
1773 		sock_release(sock2);
1774 		goto out;
1775 	}
1776 
1777 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1778 	if (IS_ERR(newfile2)) {
1779 		err = PTR_ERR(newfile2);
1780 		fput(newfile1);
1781 		goto out;
1782 	}
1783 
1784 	audit_fd_pair(fd1, fd2);
1785 
1786 	fd_install(fd1, newfile1);
1787 	fd_install(fd2, newfile2);
1788 	return 0;
1789 
1790 out:
1791 	put_unused_fd(fd2);
1792 	put_unused_fd(fd1);
1793 	return err;
1794 }
1795 
SYSCALL_DEFINE4(socketpair,int,family,int,type,int,protocol,int __user *,usockvec)1796 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1797 		int __user *, usockvec)
1798 {
1799 	return __sys_socketpair(family, type, protocol, usockvec);
1800 }
1801 
__sys_bind_socket(struct socket * sock,struct sockaddr_storage * address,int addrlen)1802 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1803 		      int addrlen)
1804 {
1805 	int err;
1806 
1807 	err = security_socket_bind(sock, (struct sockaddr *)address,
1808 				   addrlen);
1809 	if (!err)
1810 		err = READ_ONCE(sock->ops)->bind(sock,
1811 						 (struct sockaddr *)address,
1812 						 addrlen);
1813 	return err;
1814 }
1815 
1816 /*
1817  *	Bind a name to a socket. Nothing much to do here since it's
1818  *	the protocol's responsibility to handle the local address.
1819  *
1820  *	We move the socket address to kernel space before we call
1821  *	the protocol layer (having also checked the address is ok).
1822  */
1823 
__sys_bind(int fd,struct sockaddr __user * umyaddr,int addrlen)1824 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1825 {
1826 	struct socket *sock;
1827 	struct sockaddr_storage address;
1828 	CLASS(fd, f)(fd);
1829 	int err;
1830 
1831 	if (fd_empty(f))
1832 		return -EBADF;
1833 	sock = sock_from_file(fd_file(f));
1834 	if (unlikely(!sock))
1835 		return -ENOTSOCK;
1836 
1837 	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1838 	if (unlikely(err))
1839 		return err;
1840 
1841 	return __sys_bind_socket(sock, &address, addrlen);
1842 }
1843 
SYSCALL_DEFINE3(bind,int,fd,struct sockaddr __user *,umyaddr,int,addrlen)1844 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1845 {
1846 	return __sys_bind(fd, umyaddr, addrlen);
1847 }
1848 
1849 /*
1850  *	Perform a listen. Basically, we allow the protocol to do anything
1851  *	necessary for a listen, and if that works, we mark the socket as
1852  *	ready for listening.
1853  */
__sys_listen_socket(struct socket * sock,int backlog)1854 int __sys_listen_socket(struct socket *sock, int backlog)
1855 {
1856 	int somaxconn, err;
1857 
1858 	somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1859 	if ((unsigned int)backlog > somaxconn)
1860 		backlog = somaxconn;
1861 
1862 	err = security_socket_listen(sock, backlog);
1863 	if (!err)
1864 		err = READ_ONCE(sock->ops)->listen(sock, backlog);
1865 	return err;
1866 }
1867 
__sys_listen(int fd,int backlog)1868 int __sys_listen(int fd, int backlog)
1869 {
1870 	CLASS(fd, f)(fd);
1871 	struct socket *sock;
1872 
1873 	if (fd_empty(f))
1874 		return -EBADF;
1875 	sock = sock_from_file(fd_file(f));
1876 	if (unlikely(!sock))
1877 		return -ENOTSOCK;
1878 
1879 	return __sys_listen_socket(sock, backlog);
1880 }
1881 
SYSCALL_DEFINE2(listen,int,fd,int,backlog)1882 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1883 {
1884 	return __sys_listen(fd, backlog);
1885 }
1886 
do_accept(struct file * file,struct proto_accept_arg * arg,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1887 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1888 		       struct sockaddr __user *upeer_sockaddr,
1889 		       int __user *upeer_addrlen, int flags)
1890 {
1891 	struct socket *sock, *newsock;
1892 	struct file *newfile;
1893 	int err, len;
1894 	struct sockaddr_storage address;
1895 	const struct proto_ops *ops;
1896 
1897 	sock = sock_from_file(file);
1898 	if (!sock)
1899 		return ERR_PTR(-ENOTSOCK);
1900 
1901 	newsock = sock_alloc();
1902 	if (!newsock)
1903 		return ERR_PTR(-ENFILE);
1904 	ops = READ_ONCE(sock->ops);
1905 
1906 	newsock->type = sock->type;
1907 	newsock->ops = ops;
1908 
1909 	/*
1910 	 * We don't need try_module_get here, as the listening socket (sock)
1911 	 * has the protocol module (sock->ops->owner) held.
1912 	 */
1913 	__module_get(ops->owner);
1914 
1915 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1916 	if (IS_ERR(newfile))
1917 		return newfile;
1918 
1919 	err = security_socket_accept(sock, newsock);
1920 	if (err)
1921 		goto out_fd;
1922 
1923 	arg->flags |= sock->file->f_flags;
1924 	err = ops->accept(sock, newsock, arg);
1925 	if (err < 0)
1926 		goto out_fd;
1927 
1928 	if (upeer_sockaddr) {
1929 		len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1930 		if (len < 0) {
1931 			err = -ECONNABORTED;
1932 			goto out_fd;
1933 		}
1934 		err = move_addr_to_user(&address,
1935 					len, upeer_sockaddr, upeer_addrlen);
1936 		if (err < 0)
1937 			goto out_fd;
1938 	}
1939 
1940 	/* File flags are not inherited via accept() unlike another OSes. */
1941 	return newfile;
1942 out_fd:
1943 	fput(newfile);
1944 	return ERR_PTR(err);
1945 }
1946 
__sys_accept4_file(struct file * file,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1947 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1948 			      int __user *upeer_addrlen, int flags)
1949 {
1950 	struct proto_accept_arg arg = { };
1951 	struct file *newfile;
1952 	int newfd;
1953 
1954 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1955 		return -EINVAL;
1956 
1957 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1958 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1959 
1960 	newfd = get_unused_fd_flags(flags);
1961 	if (unlikely(newfd < 0))
1962 		return newfd;
1963 
1964 	newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1965 			    flags);
1966 	if (IS_ERR(newfile)) {
1967 		put_unused_fd(newfd);
1968 		return PTR_ERR(newfile);
1969 	}
1970 	fd_install(newfd, newfile);
1971 	return newfd;
1972 }
1973 
1974 /*
1975  *	For accept, we attempt to create a new socket, set up the link
1976  *	with the client, wake up the client, then return the new
1977  *	connected fd. We collect the address of the connector in kernel
1978  *	space and move it to user at the very end. This is unclean because
1979  *	we open the socket then return an error.
1980  *
1981  *	1003.1g adds the ability to recvmsg() to query connection pending
1982  *	status to recvmsg. We need to add that support in a way thats
1983  *	clean when we restructure accept also.
1984  */
1985 
__sys_accept4(int fd,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1986 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1987 		  int __user *upeer_addrlen, int flags)
1988 {
1989 	CLASS(fd, f)(fd);
1990 
1991 	if (fd_empty(f))
1992 		return -EBADF;
1993 	return __sys_accept4_file(fd_file(f), upeer_sockaddr,
1994 					 upeer_addrlen, flags);
1995 }
1996 
SYSCALL_DEFINE4(accept4,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen,int,flags)1997 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1998 		int __user *, upeer_addrlen, int, flags)
1999 {
2000 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2001 }
2002 
SYSCALL_DEFINE3(accept,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen)2003 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2004 		int __user *, upeer_addrlen)
2005 {
2006 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2007 }
2008 
2009 /*
2010  *	Attempt to connect to a socket with the server address.  The address
2011  *	is in user space so we verify it is OK and move it to kernel space.
2012  *
2013  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2014  *	break bindings
2015  *
2016  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2017  *	other SEQPACKET protocols that take time to connect() as it doesn't
2018  *	include the -EINPROGRESS status for such sockets.
2019  */
2020 
__sys_connect_file(struct file * file,struct sockaddr_storage * address,int addrlen,int file_flags)2021 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2022 		       int addrlen, int file_flags)
2023 {
2024 	struct socket *sock;
2025 	int err;
2026 
2027 	sock = sock_from_file(file);
2028 	if (!sock) {
2029 		err = -ENOTSOCK;
2030 		goto out;
2031 	}
2032 
2033 	err =
2034 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2035 	if (err)
2036 		goto out;
2037 
2038 	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2039 				addrlen, sock->file->f_flags | file_flags);
2040 out:
2041 	return err;
2042 }
2043 
__sys_connect(int fd,struct sockaddr __user * uservaddr,int addrlen)2044 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2045 {
2046 	struct sockaddr_storage address;
2047 	CLASS(fd, f)(fd);
2048 	int ret;
2049 
2050 	if (fd_empty(f))
2051 		return -EBADF;
2052 
2053 	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2054 	if (ret)
2055 		return ret;
2056 
2057 	return __sys_connect_file(fd_file(f), &address, addrlen, 0);
2058 }
2059 
SYSCALL_DEFINE3(connect,int,fd,struct sockaddr __user *,uservaddr,int,addrlen)2060 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2061 		int, addrlen)
2062 {
2063 	return __sys_connect(fd, uservaddr, addrlen);
2064 }
2065 
2066 /*
2067  *	Get the local address ('name') of a socket object. Move the obtained
2068  *	name to user space.
2069  */
2070 
__sys_getsockname(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2071 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2072 		      int __user *usockaddr_len)
2073 {
2074 	struct socket *sock;
2075 	struct sockaddr_storage address;
2076 	CLASS(fd, f)(fd);
2077 	int err;
2078 
2079 	if (fd_empty(f))
2080 		return -EBADF;
2081 	sock = sock_from_file(fd_file(f));
2082 	if (unlikely(!sock))
2083 		return -ENOTSOCK;
2084 
2085 	err = security_socket_getsockname(sock);
2086 	if (err)
2087 		return err;
2088 
2089 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2090 	if (err < 0)
2091 		return err;
2092 
2093 	/* "err" is actually length in this case */
2094 	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2095 }
2096 
SYSCALL_DEFINE3(getsockname,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2097 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2098 		int __user *, usockaddr_len)
2099 {
2100 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2101 }
2102 
2103 /*
2104  *	Get the remote address ('name') of a socket object. Move the obtained
2105  *	name to user space.
2106  */
2107 
__sys_getpeername(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2108 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2109 		      int __user *usockaddr_len)
2110 {
2111 	struct socket *sock;
2112 	struct sockaddr_storage address;
2113 	CLASS(fd, f)(fd);
2114 	int err;
2115 
2116 	if (fd_empty(f))
2117 		return -EBADF;
2118 	sock = sock_from_file(fd_file(f));
2119 	if (unlikely(!sock))
2120 		return -ENOTSOCK;
2121 
2122 	err = security_socket_getpeername(sock);
2123 	if (err)
2124 		return err;
2125 
2126 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 1);
2127 	if (err < 0)
2128 		return err;
2129 
2130 	/* "err" is actually length in this case */
2131 	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2132 }
2133 
SYSCALL_DEFINE3(getpeername,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2134 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2135 		int __user *, usockaddr_len)
2136 {
2137 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2138 }
2139 
2140 /*
2141  *	Send a datagram to a given address. We move the address into kernel
2142  *	space and check the user space data area is readable before invoking
2143  *	the protocol.
2144  */
__sys_sendto(int fd,void __user * buff,size_t len,unsigned int flags,struct sockaddr __user * addr,int addr_len)2145 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2146 		 struct sockaddr __user *addr,  int addr_len)
2147 {
2148 	struct socket *sock;
2149 	struct sockaddr_storage address;
2150 	int err;
2151 	struct msghdr msg;
2152 
2153 	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2154 	if (unlikely(err))
2155 		return err;
2156 
2157 	CLASS(fd, f)(fd);
2158 	if (fd_empty(f))
2159 		return -EBADF;
2160 	sock = sock_from_file(fd_file(f));
2161 	if (unlikely(!sock))
2162 		return -ENOTSOCK;
2163 
2164 	msg.msg_name = NULL;
2165 	msg.msg_control = NULL;
2166 	msg.msg_controllen = 0;
2167 	msg.msg_namelen = 0;
2168 	msg.msg_ubuf = NULL;
2169 	if (addr) {
2170 		err = move_addr_to_kernel(addr, addr_len, &address);
2171 		if (err < 0)
2172 			return err;
2173 		msg.msg_name = (struct sockaddr *)&address;
2174 		msg.msg_namelen = addr_len;
2175 	}
2176 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2177 	if (sock->file->f_flags & O_NONBLOCK)
2178 		flags |= MSG_DONTWAIT;
2179 	msg.msg_flags = flags;
2180 	return __sock_sendmsg(sock, &msg);
2181 }
2182 
SYSCALL_DEFINE6(sendto,int,fd,void __user *,buff,size_t,len,unsigned int,flags,struct sockaddr __user *,addr,int,addr_len)2183 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2184 		unsigned int, flags, struct sockaddr __user *, addr,
2185 		int, addr_len)
2186 {
2187 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2188 }
2189 
2190 /*
2191  *	Send a datagram down a socket.
2192  */
2193 
SYSCALL_DEFINE4(send,int,fd,void __user *,buff,size_t,len,unsigned int,flags)2194 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2195 		unsigned int, flags)
2196 {
2197 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2198 }
2199 
2200 /*
2201  *	Receive a frame from the socket and optionally record the address of the
2202  *	sender. We verify the buffers are writable and if needed move the
2203  *	sender address from kernel to user space.
2204  */
__sys_recvfrom(int fd,void __user * ubuf,size_t size,unsigned int flags,struct sockaddr __user * addr,int __user * addr_len)2205 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2206 		   struct sockaddr __user *addr, int __user *addr_len)
2207 {
2208 	struct sockaddr_storage address;
2209 	struct msghdr msg = {
2210 		/* Save some cycles and don't copy the address if not needed */
2211 		.msg_name = addr ? (struct sockaddr *)&address : NULL,
2212 	};
2213 	struct socket *sock;
2214 	int err, err2;
2215 
2216 	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2217 	if (unlikely(err))
2218 		return err;
2219 
2220 	CLASS(fd, f)(fd);
2221 
2222 	if (fd_empty(f))
2223 		return -EBADF;
2224 	sock = sock_from_file(fd_file(f));
2225 	if (unlikely(!sock))
2226 		return -ENOTSOCK;
2227 
2228 	if (sock->file->f_flags & O_NONBLOCK)
2229 		flags |= MSG_DONTWAIT;
2230 	err = sock_recvmsg(sock, &msg, flags);
2231 
2232 	if (err >= 0 && addr != NULL) {
2233 		err2 = move_addr_to_user(&address,
2234 					 msg.msg_namelen, addr, addr_len);
2235 		if (err2 < 0)
2236 			err = err2;
2237 	}
2238 	return err;
2239 }
2240 
SYSCALL_DEFINE6(recvfrom,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags,struct sockaddr __user *,addr,int __user *,addr_len)2241 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2242 		unsigned int, flags, struct sockaddr __user *, addr,
2243 		int __user *, addr_len)
2244 {
2245 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2246 }
2247 
2248 /*
2249  *	Receive a datagram from a socket.
2250  */
2251 
SYSCALL_DEFINE4(recv,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags)2252 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2253 		unsigned int, flags)
2254 {
2255 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2256 }
2257 
sock_use_custom_sol_socket(const struct socket * sock)2258 static bool sock_use_custom_sol_socket(const struct socket *sock)
2259 {
2260 	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2261 }
2262 
do_sock_setsockopt(struct socket * sock,bool compat,int level,int optname,sockptr_t optval,int optlen)2263 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2264 		       int optname, sockptr_t optval, int optlen)
2265 {
2266 	const struct proto_ops *ops;
2267 	char *kernel_optval = NULL;
2268 	int err;
2269 
2270 	if (optlen < 0)
2271 		return -EINVAL;
2272 
2273 	err = security_socket_setsockopt(sock, level, optname);
2274 	if (err)
2275 		goto out_put;
2276 
2277 	if (!compat)
2278 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2279 						     optval, &optlen,
2280 						     &kernel_optval);
2281 	if (err < 0)
2282 		goto out_put;
2283 	if (err > 0) {
2284 		err = 0;
2285 		goto out_put;
2286 	}
2287 
2288 	if (kernel_optval)
2289 		optval = KERNEL_SOCKPTR(kernel_optval);
2290 	ops = READ_ONCE(sock->ops);
2291 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2292 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2293 	else if (unlikely(!ops->setsockopt))
2294 		err = -EOPNOTSUPP;
2295 	else
2296 		err = ops->setsockopt(sock, level, optname, optval,
2297 					    optlen);
2298 	kfree(kernel_optval);
2299 out_put:
2300 	return err;
2301 }
2302 EXPORT_SYMBOL(do_sock_setsockopt);
2303 
2304 /* Set a socket option. Because we don't know the option lengths we have
2305  * to pass the user mode parameter for the protocols to sort out.
2306  */
__sys_setsockopt(int fd,int level,int optname,char __user * user_optval,int optlen)2307 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2308 		     int optlen)
2309 {
2310 	sockptr_t optval = USER_SOCKPTR(user_optval);
2311 	bool compat = in_compat_syscall();
2312 	struct socket *sock;
2313 	CLASS(fd, f)(fd);
2314 
2315 	if (fd_empty(f))
2316 		return -EBADF;
2317 	sock = sock_from_file(fd_file(f));
2318 	if (unlikely(!sock))
2319 		return -ENOTSOCK;
2320 
2321 	return do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2322 }
2323 
SYSCALL_DEFINE5(setsockopt,int,fd,int,level,int,optname,char __user *,optval,int,optlen)2324 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2325 		char __user *, optval, int, optlen)
2326 {
2327 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2328 }
2329 
2330 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2331 							 int optname));
2332 
do_sock_getsockopt(struct socket * sock,bool compat,int level,int optname,sockptr_t optval,sockptr_t optlen)2333 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2334 		       int optname, sockptr_t optval, sockptr_t optlen)
2335 {
2336 	int max_optlen __maybe_unused = 0;
2337 	const struct proto_ops *ops;
2338 	int err;
2339 
2340 	err = security_socket_getsockopt(sock, level, optname);
2341 	if (err)
2342 		return err;
2343 
2344 	if (!compat)
2345 		copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2346 
2347 	ops = READ_ONCE(sock->ops);
2348 	if (level == SOL_SOCKET) {
2349 		err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2350 	} else if (unlikely(!ops->getsockopt)) {
2351 		err = -EOPNOTSUPP;
2352 	} else {
2353 		if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2354 			      "Invalid argument type"))
2355 			return -EOPNOTSUPP;
2356 
2357 		err = ops->getsockopt(sock, level, optname, optval.user,
2358 				      optlen.user);
2359 	}
2360 
2361 	if (!compat)
2362 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2363 						     optval, optlen, max_optlen,
2364 						     err);
2365 
2366 	return err;
2367 }
2368 EXPORT_SYMBOL(do_sock_getsockopt);
2369 
2370 /*
2371  *	Get a socket option. Because we don't know the option lengths we have
2372  *	to pass a user mode parameter for the protocols to sort out.
2373  */
__sys_getsockopt(int fd,int level,int optname,char __user * optval,int __user * optlen)2374 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2375 		int __user *optlen)
2376 {
2377 	struct socket *sock;
2378 	CLASS(fd, f)(fd);
2379 
2380 	if (fd_empty(f))
2381 		return -EBADF;
2382 	sock = sock_from_file(fd_file(f));
2383 	if (unlikely(!sock))
2384 		return -ENOTSOCK;
2385 
2386 	return do_sock_getsockopt(sock, in_compat_syscall(), level, optname,
2387 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2388 }
2389 
SYSCALL_DEFINE5(getsockopt,int,fd,int,level,int,optname,char __user *,optval,int __user *,optlen)2390 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2391 		char __user *, optval, int __user *, optlen)
2392 {
2393 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2394 }
2395 
2396 /*
2397  *	Shutdown a socket.
2398  */
2399 
__sys_shutdown_sock(struct socket * sock,int how)2400 int __sys_shutdown_sock(struct socket *sock, int how)
2401 {
2402 	int err;
2403 
2404 	err = security_socket_shutdown(sock, how);
2405 	if (!err)
2406 		err = READ_ONCE(sock->ops)->shutdown(sock, how);
2407 
2408 	return err;
2409 }
2410 
__sys_shutdown(int fd,int how)2411 int __sys_shutdown(int fd, int how)
2412 {
2413 	struct socket *sock;
2414 	CLASS(fd, f)(fd);
2415 
2416 	if (fd_empty(f))
2417 		return -EBADF;
2418 	sock = sock_from_file(fd_file(f));
2419 	if (unlikely(!sock))
2420 		return -ENOTSOCK;
2421 
2422 	return __sys_shutdown_sock(sock, how);
2423 }
2424 
SYSCALL_DEFINE2(shutdown,int,fd,int,how)2425 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2426 {
2427 	return __sys_shutdown(fd, how);
2428 }
2429 
2430 /* A couple of helpful macros for getting the address of the 32/64 bit
2431  * fields which are the same type (int / unsigned) on our platforms.
2432  */
2433 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2434 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2435 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2436 
2437 struct used_address {
2438 	struct sockaddr_storage name;
2439 	unsigned int name_len;
2440 };
2441 
__copy_msghdr(struct msghdr * kmsg,struct user_msghdr * msg,struct sockaddr __user ** save_addr)2442 int __copy_msghdr(struct msghdr *kmsg,
2443 		  struct user_msghdr *msg,
2444 		  struct sockaddr __user **save_addr)
2445 {
2446 	ssize_t err;
2447 
2448 	kmsg->msg_control_is_user = true;
2449 	kmsg->msg_get_inq = 0;
2450 	kmsg->msg_control_user = msg->msg_control;
2451 	kmsg->msg_controllen = msg->msg_controllen;
2452 	kmsg->msg_flags = msg->msg_flags;
2453 
2454 	kmsg->msg_namelen = msg->msg_namelen;
2455 	if (!msg->msg_name)
2456 		kmsg->msg_namelen = 0;
2457 
2458 	if (kmsg->msg_namelen < 0)
2459 		return -EINVAL;
2460 
2461 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2462 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2463 
2464 	if (save_addr)
2465 		*save_addr = msg->msg_name;
2466 
2467 	if (msg->msg_name && kmsg->msg_namelen) {
2468 		if (!save_addr) {
2469 			err = move_addr_to_kernel(msg->msg_name,
2470 						  kmsg->msg_namelen,
2471 						  kmsg->msg_name);
2472 			if (err < 0)
2473 				return err;
2474 		}
2475 	} else {
2476 		kmsg->msg_name = NULL;
2477 		kmsg->msg_namelen = 0;
2478 	}
2479 
2480 	if (msg->msg_iovlen > UIO_MAXIOV)
2481 		return -EMSGSIZE;
2482 
2483 	kmsg->msg_iocb = NULL;
2484 	kmsg->msg_ubuf = NULL;
2485 	return 0;
2486 }
2487 
copy_msghdr_from_user(struct msghdr * kmsg,struct user_msghdr __user * umsg,struct sockaddr __user ** save_addr,struct iovec ** iov)2488 static int copy_msghdr_from_user(struct msghdr *kmsg,
2489 				 struct user_msghdr __user *umsg,
2490 				 struct sockaddr __user **save_addr,
2491 				 struct iovec **iov)
2492 {
2493 	struct user_msghdr msg;
2494 	ssize_t err;
2495 
2496 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2497 		return -EFAULT;
2498 
2499 	err = __copy_msghdr(kmsg, &msg, save_addr);
2500 	if (err)
2501 		return err;
2502 
2503 	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2504 			    msg.msg_iov, msg.msg_iovlen,
2505 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2506 	return err < 0 ? err : 0;
2507 }
2508 
____sys_sendmsg(struct socket * sock,struct msghdr * msg_sys,unsigned int flags,struct used_address * used_address,unsigned int allowed_msghdr_flags)2509 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2510 			   unsigned int flags, struct used_address *used_address,
2511 			   unsigned int allowed_msghdr_flags)
2512 {
2513 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2514 				__aligned(sizeof(__kernel_size_t));
2515 	/* 20 is size of ipv6_pktinfo */
2516 	unsigned char *ctl_buf = ctl;
2517 	int ctl_len;
2518 	ssize_t err;
2519 
2520 	err = -ENOBUFS;
2521 
2522 	if (msg_sys->msg_controllen > INT_MAX)
2523 		goto out;
2524 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2525 	ctl_len = msg_sys->msg_controllen;
2526 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2527 		err =
2528 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2529 						     sizeof(ctl));
2530 		if (err)
2531 			goto out;
2532 		ctl_buf = msg_sys->msg_control;
2533 		ctl_len = msg_sys->msg_controllen;
2534 	} else if (ctl_len) {
2535 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2536 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2537 		if (ctl_len > sizeof(ctl)) {
2538 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2539 			if (ctl_buf == NULL)
2540 				goto out;
2541 		}
2542 		err = -EFAULT;
2543 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2544 			goto out_freectl;
2545 		msg_sys->msg_control = ctl_buf;
2546 		msg_sys->msg_control_is_user = false;
2547 	}
2548 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2549 	msg_sys->msg_flags = flags;
2550 
2551 	if (sock->file->f_flags & O_NONBLOCK)
2552 		msg_sys->msg_flags |= MSG_DONTWAIT;
2553 	/*
2554 	 * If this is sendmmsg() and current destination address is same as
2555 	 * previously succeeded address, omit asking LSM's decision.
2556 	 * used_address->name_len is initialized to UINT_MAX so that the first
2557 	 * destination address never matches.
2558 	 */
2559 	if (used_address && msg_sys->msg_name &&
2560 	    used_address->name_len == msg_sys->msg_namelen &&
2561 	    !memcmp(&used_address->name, msg_sys->msg_name,
2562 		    used_address->name_len)) {
2563 		err = sock_sendmsg_nosec(sock, msg_sys);
2564 		goto out_freectl;
2565 	}
2566 	err = __sock_sendmsg(sock, msg_sys);
2567 	/*
2568 	 * If this is sendmmsg() and sending to current destination address was
2569 	 * successful, remember it.
2570 	 */
2571 	if (used_address && err >= 0) {
2572 		used_address->name_len = msg_sys->msg_namelen;
2573 		if (msg_sys->msg_name)
2574 			memcpy(&used_address->name, msg_sys->msg_name,
2575 			       used_address->name_len);
2576 	}
2577 
2578 out_freectl:
2579 	if (ctl_buf != ctl)
2580 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2581 out:
2582 	return err;
2583 }
2584 
sendmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct iovec ** iov)2585 static int sendmsg_copy_msghdr(struct msghdr *msg,
2586 			       struct user_msghdr __user *umsg, unsigned flags,
2587 			       struct iovec **iov)
2588 {
2589 	int err;
2590 
2591 	if (flags & MSG_CMSG_COMPAT) {
2592 		struct compat_msghdr __user *msg_compat;
2593 
2594 		msg_compat = (struct compat_msghdr __user *) umsg;
2595 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2596 	} else {
2597 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2598 	}
2599 	if (err < 0)
2600 		return err;
2601 
2602 	return 0;
2603 }
2604 
___sys_sendmsg(struct socket * sock,struct user_msghdr __user * msg,struct msghdr * msg_sys,unsigned int flags,struct used_address * used_address,unsigned int allowed_msghdr_flags)2605 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2606 			 struct msghdr *msg_sys, unsigned int flags,
2607 			 struct used_address *used_address,
2608 			 unsigned int allowed_msghdr_flags)
2609 {
2610 	struct sockaddr_storage address;
2611 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2612 	ssize_t err;
2613 
2614 	msg_sys->msg_name = &address;
2615 
2616 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2617 	if (err < 0)
2618 		return err;
2619 
2620 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2621 				allowed_msghdr_flags);
2622 	kfree(iov);
2623 	return err;
2624 }
2625 
2626 /*
2627  *	BSD sendmsg interface
2628  */
__sys_sendmsg_sock(struct socket * sock,struct msghdr * msg,unsigned int flags)2629 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2630 			unsigned int flags)
2631 {
2632 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2633 }
2634 
__sys_sendmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2635 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2636 		   bool forbid_cmsg_compat)
2637 {
2638 	struct msghdr msg_sys;
2639 	struct socket *sock;
2640 
2641 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2642 		return -EINVAL;
2643 
2644 	CLASS(fd, f)(fd);
2645 
2646 	if (fd_empty(f))
2647 		return -EBADF;
2648 	sock = sock_from_file(fd_file(f));
2649 	if (unlikely(!sock))
2650 		return -ENOTSOCK;
2651 
2652 	return ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2653 }
2654 
SYSCALL_DEFINE3(sendmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2655 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2656 {
2657 	return __sys_sendmsg(fd, msg, flags, true);
2658 }
2659 
2660 /*
2661  *	Linux sendmmsg interface
2662  */
2663 
__sys_sendmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,bool forbid_cmsg_compat)2664 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2665 		   unsigned int flags, bool forbid_cmsg_compat)
2666 {
2667 	int err, datagrams;
2668 	struct socket *sock;
2669 	struct mmsghdr __user *entry;
2670 	struct compat_mmsghdr __user *compat_entry;
2671 	struct msghdr msg_sys;
2672 	struct used_address used_address;
2673 	unsigned int oflags = flags;
2674 
2675 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2676 		return -EINVAL;
2677 
2678 	if (vlen > UIO_MAXIOV)
2679 		vlen = UIO_MAXIOV;
2680 
2681 	datagrams = 0;
2682 
2683 	CLASS(fd, f)(fd);
2684 
2685 	if (fd_empty(f))
2686 		return -EBADF;
2687 	sock = sock_from_file(fd_file(f));
2688 	if (unlikely(!sock))
2689 		return -ENOTSOCK;
2690 
2691 	used_address.name_len = UINT_MAX;
2692 	entry = mmsg;
2693 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2694 	err = 0;
2695 	flags |= MSG_BATCH;
2696 
2697 	while (datagrams < vlen) {
2698 		if (datagrams == vlen - 1)
2699 			flags = oflags;
2700 
2701 		if (MSG_CMSG_COMPAT & flags) {
2702 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2703 					     &msg_sys, flags, &used_address, MSG_EOR);
2704 			if (err < 0)
2705 				break;
2706 			err = __put_user(err, &compat_entry->msg_len);
2707 			++compat_entry;
2708 		} else {
2709 			err = ___sys_sendmsg(sock,
2710 					     (struct user_msghdr __user *)entry,
2711 					     &msg_sys, flags, &used_address, MSG_EOR);
2712 			if (err < 0)
2713 				break;
2714 			err = put_user(err, &entry->msg_len);
2715 			++entry;
2716 		}
2717 
2718 		if (err)
2719 			break;
2720 		++datagrams;
2721 		if (msg_data_left(&msg_sys))
2722 			break;
2723 		cond_resched();
2724 	}
2725 
2726 	/* We only return an error if no datagrams were able to be sent */
2727 	if (datagrams != 0)
2728 		return datagrams;
2729 
2730 	return err;
2731 }
2732 
SYSCALL_DEFINE4(sendmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags)2733 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2734 		unsigned int, vlen, unsigned int, flags)
2735 {
2736 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2737 }
2738 
recvmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct sockaddr __user ** uaddr,struct iovec ** iov)2739 static int recvmsg_copy_msghdr(struct msghdr *msg,
2740 			       struct user_msghdr __user *umsg, unsigned flags,
2741 			       struct sockaddr __user **uaddr,
2742 			       struct iovec **iov)
2743 {
2744 	ssize_t err;
2745 
2746 	if (MSG_CMSG_COMPAT & flags) {
2747 		struct compat_msghdr __user *msg_compat;
2748 
2749 		msg_compat = (struct compat_msghdr __user *) umsg;
2750 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2751 	} else {
2752 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2753 	}
2754 	if (err < 0)
2755 		return err;
2756 
2757 	return 0;
2758 }
2759 
____sys_recvmsg(struct socket * sock,struct msghdr * msg_sys,struct user_msghdr __user * msg,struct sockaddr __user * uaddr,unsigned int flags,int nosec)2760 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2761 			   struct user_msghdr __user *msg,
2762 			   struct sockaddr __user *uaddr,
2763 			   unsigned int flags, int nosec)
2764 {
2765 	struct compat_msghdr __user *msg_compat =
2766 					(struct compat_msghdr __user *) msg;
2767 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2768 	struct sockaddr_storage addr;
2769 	unsigned long cmsg_ptr;
2770 	int len;
2771 	ssize_t err;
2772 
2773 	msg_sys->msg_name = &addr;
2774 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2775 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2776 
2777 	/* We assume all kernel code knows the size of sockaddr_storage */
2778 	msg_sys->msg_namelen = 0;
2779 
2780 	if (sock->file->f_flags & O_NONBLOCK)
2781 		flags |= MSG_DONTWAIT;
2782 
2783 	if (unlikely(nosec))
2784 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2785 	else
2786 		err = sock_recvmsg(sock, msg_sys, flags);
2787 
2788 	if (err < 0)
2789 		goto out;
2790 	len = err;
2791 
2792 	if (uaddr != NULL) {
2793 		err = move_addr_to_user(&addr,
2794 					msg_sys->msg_namelen, uaddr,
2795 					uaddr_len);
2796 		if (err < 0)
2797 			goto out;
2798 	}
2799 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2800 			 COMPAT_FLAGS(msg));
2801 	if (err)
2802 		goto out;
2803 	if (MSG_CMSG_COMPAT & flags)
2804 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2805 				 &msg_compat->msg_controllen);
2806 	else
2807 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2808 				 &msg->msg_controllen);
2809 	if (err)
2810 		goto out;
2811 	err = len;
2812 out:
2813 	return err;
2814 }
2815 
___sys_recvmsg(struct socket * sock,struct user_msghdr __user * msg,struct msghdr * msg_sys,unsigned int flags,int nosec)2816 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2817 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2818 {
2819 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2820 	/* user mode address pointers */
2821 	struct sockaddr __user *uaddr;
2822 	ssize_t err;
2823 
2824 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2825 	if (err < 0)
2826 		return err;
2827 
2828 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2829 	kfree(iov);
2830 	return err;
2831 }
2832 
2833 /*
2834  *	BSD recvmsg interface
2835  */
2836 
__sys_recvmsg_sock(struct socket * sock,struct msghdr * msg,struct user_msghdr __user * umsg,struct sockaddr __user * uaddr,unsigned int flags)2837 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2838 			struct user_msghdr __user *umsg,
2839 			struct sockaddr __user *uaddr, unsigned int flags)
2840 {
2841 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2842 }
2843 
__sys_recvmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2844 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2845 		   bool forbid_cmsg_compat)
2846 {
2847 	struct msghdr msg_sys;
2848 	struct socket *sock;
2849 
2850 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2851 		return -EINVAL;
2852 
2853 	CLASS(fd, f)(fd);
2854 
2855 	if (fd_empty(f))
2856 		return -EBADF;
2857 	sock = sock_from_file(fd_file(f));
2858 	if (unlikely(!sock))
2859 		return -ENOTSOCK;
2860 
2861 	return ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2862 }
2863 
SYSCALL_DEFINE3(recvmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2864 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2865 		unsigned int, flags)
2866 {
2867 	return __sys_recvmsg(fd, msg, flags, true);
2868 }
2869 
2870 /*
2871  *     Linux recvmmsg interface
2872  */
2873 
do_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct timespec64 * timeout)2874 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2875 			  unsigned int vlen, unsigned int flags,
2876 			  struct timespec64 *timeout)
2877 {
2878 	int err = 0, datagrams;
2879 	struct socket *sock;
2880 	struct mmsghdr __user *entry;
2881 	struct compat_mmsghdr __user *compat_entry;
2882 	struct msghdr msg_sys;
2883 	struct timespec64 end_time;
2884 	struct timespec64 timeout64;
2885 
2886 	if (timeout &&
2887 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2888 				    timeout->tv_nsec))
2889 		return -EINVAL;
2890 
2891 	datagrams = 0;
2892 
2893 	CLASS(fd, f)(fd);
2894 
2895 	if (fd_empty(f))
2896 		return -EBADF;
2897 	sock = sock_from_file(fd_file(f));
2898 	if (unlikely(!sock))
2899 		return -ENOTSOCK;
2900 
2901 	if (likely(!(flags & MSG_ERRQUEUE))) {
2902 		err = sock_error(sock->sk);
2903 		if (err)
2904 			return err;
2905 	}
2906 
2907 	entry = mmsg;
2908 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2909 
2910 	while (datagrams < vlen) {
2911 		/*
2912 		 * No need to ask LSM for more than the first datagram.
2913 		 */
2914 		if (MSG_CMSG_COMPAT & flags) {
2915 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2916 					     &msg_sys, flags & ~MSG_WAITFORONE,
2917 					     datagrams);
2918 			if (err < 0)
2919 				break;
2920 			err = __put_user(err, &compat_entry->msg_len);
2921 			++compat_entry;
2922 		} else {
2923 			err = ___sys_recvmsg(sock,
2924 					     (struct user_msghdr __user *)entry,
2925 					     &msg_sys, flags & ~MSG_WAITFORONE,
2926 					     datagrams);
2927 			if (err < 0)
2928 				break;
2929 			err = put_user(err, &entry->msg_len);
2930 			++entry;
2931 		}
2932 
2933 		if (err)
2934 			break;
2935 		++datagrams;
2936 
2937 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2938 		if (flags & MSG_WAITFORONE)
2939 			flags |= MSG_DONTWAIT;
2940 
2941 		if (timeout) {
2942 			ktime_get_ts64(&timeout64);
2943 			*timeout = timespec64_sub(end_time, timeout64);
2944 			if (timeout->tv_sec < 0) {
2945 				timeout->tv_sec = timeout->tv_nsec = 0;
2946 				break;
2947 			}
2948 
2949 			/* Timeout, return less than vlen datagrams */
2950 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2951 				break;
2952 		}
2953 
2954 		/* Out of band data, return right away */
2955 		if (msg_sys.msg_flags & MSG_OOB)
2956 			break;
2957 		cond_resched();
2958 	}
2959 
2960 	if (err == 0)
2961 		return datagrams;
2962 
2963 	if (datagrams == 0)
2964 		return err;
2965 
2966 	/*
2967 	 * We may return less entries than requested (vlen) if the
2968 	 * sock is non block and there aren't enough datagrams...
2969 	 */
2970 	if (err != -EAGAIN) {
2971 		/*
2972 		 * ... or  if recvmsg returns an error after we
2973 		 * received some datagrams, where we record the
2974 		 * error to return on the next call or if the
2975 		 * app asks about it using getsockopt(SO_ERROR).
2976 		 */
2977 		WRITE_ONCE(sock->sk->sk_err, -err);
2978 	}
2979 	return datagrams;
2980 }
2981 
__sys_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct __kernel_timespec __user * timeout,struct old_timespec32 __user * timeout32)2982 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2983 		   unsigned int vlen, unsigned int flags,
2984 		   struct __kernel_timespec __user *timeout,
2985 		   struct old_timespec32 __user *timeout32)
2986 {
2987 	int datagrams;
2988 	struct timespec64 timeout_sys;
2989 
2990 	if (timeout && get_timespec64(&timeout_sys, timeout))
2991 		return -EFAULT;
2992 
2993 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2994 		return -EFAULT;
2995 
2996 	if (!timeout && !timeout32)
2997 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2998 
2999 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3000 
3001 	if (datagrams <= 0)
3002 		return datagrams;
3003 
3004 	if (timeout && put_timespec64(&timeout_sys, timeout))
3005 		datagrams = -EFAULT;
3006 
3007 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3008 		datagrams = -EFAULT;
3009 
3010 	return datagrams;
3011 }
3012 
SYSCALL_DEFINE5(recvmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags,struct __kernel_timespec __user *,timeout)3013 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3014 		unsigned int, vlen, unsigned int, flags,
3015 		struct __kernel_timespec __user *, timeout)
3016 {
3017 	if (flags & MSG_CMSG_COMPAT)
3018 		return -EINVAL;
3019 
3020 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3021 }
3022 
3023 #ifdef CONFIG_COMPAT_32BIT_TIME
SYSCALL_DEFINE5(recvmmsg_time32,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags,struct old_timespec32 __user *,timeout)3024 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3025 		unsigned int, vlen, unsigned int, flags,
3026 		struct old_timespec32 __user *, timeout)
3027 {
3028 	if (flags & MSG_CMSG_COMPAT)
3029 		return -EINVAL;
3030 
3031 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3032 }
3033 #endif
3034 
3035 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3036 /* Argument list sizes for sys_socketcall */
3037 #define AL(x) ((x) * sizeof(unsigned long))
3038 static const unsigned char nargs[21] = {
3039 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3040 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3041 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3042 	AL(4), AL(5), AL(4)
3043 };
3044 
3045 #undef AL
3046 
3047 /*
3048  *	System call vectors.
3049  *
3050  *	Argument checking cleaned up. Saved 20% in size.
3051  *  This function doesn't need to set the kernel lock because
3052  *  it is set by the callees.
3053  */
3054 
SYSCALL_DEFINE2(socketcall,int,call,unsigned long __user *,args)3055 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3056 {
3057 	unsigned long a[AUDITSC_ARGS];
3058 	unsigned long a0, a1;
3059 	int err;
3060 	unsigned int len;
3061 
3062 	if (call < 1 || call > SYS_SENDMMSG)
3063 		return -EINVAL;
3064 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3065 
3066 	len = nargs[call];
3067 	if (len > sizeof(a))
3068 		return -EINVAL;
3069 
3070 	/* copy_from_user should be SMP safe. */
3071 	if (copy_from_user(a, args, len))
3072 		return -EFAULT;
3073 
3074 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3075 	if (err)
3076 		return err;
3077 
3078 	a0 = a[0];
3079 	a1 = a[1];
3080 
3081 	switch (call) {
3082 	case SYS_SOCKET:
3083 		err = __sys_socket(a0, a1, a[2]);
3084 		break;
3085 	case SYS_BIND:
3086 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3087 		break;
3088 	case SYS_CONNECT:
3089 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3090 		break;
3091 	case SYS_LISTEN:
3092 		err = __sys_listen(a0, a1);
3093 		break;
3094 	case SYS_ACCEPT:
3095 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3096 				    (int __user *)a[2], 0);
3097 		break;
3098 	case SYS_GETSOCKNAME:
3099 		err =
3100 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
3101 				      (int __user *)a[2]);
3102 		break;
3103 	case SYS_GETPEERNAME:
3104 		err =
3105 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
3106 				      (int __user *)a[2]);
3107 		break;
3108 	case SYS_SOCKETPAIR:
3109 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3110 		break;
3111 	case SYS_SEND:
3112 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3113 				   NULL, 0);
3114 		break;
3115 	case SYS_SENDTO:
3116 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3117 				   (struct sockaddr __user *)a[4], a[5]);
3118 		break;
3119 	case SYS_RECV:
3120 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3121 				     NULL, NULL);
3122 		break;
3123 	case SYS_RECVFROM:
3124 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3125 				     (struct sockaddr __user *)a[4],
3126 				     (int __user *)a[5]);
3127 		break;
3128 	case SYS_SHUTDOWN:
3129 		err = __sys_shutdown(a0, a1);
3130 		break;
3131 	case SYS_SETSOCKOPT:
3132 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3133 				       a[4]);
3134 		break;
3135 	case SYS_GETSOCKOPT:
3136 		err =
3137 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3138 				     (int __user *)a[4]);
3139 		break;
3140 	case SYS_SENDMSG:
3141 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3142 				    a[2], true);
3143 		break;
3144 	case SYS_SENDMMSG:
3145 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3146 				     a[3], true);
3147 		break;
3148 	case SYS_RECVMSG:
3149 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3150 				    a[2], true);
3151 		break;
3152 	case SYS_RECVMMSG:
3153 		if (IS_ENABLED(CONFIG_64BIT))
3154 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3155 					     a[2], a[3],
3156 					     (struct __kernel_timespec __user *)a[4],
3157 					     NULL);
3158 		else
3159 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3160 					     a[2], a[3], NULL,
3161 					     (struct old_timespec32 __user *)a[4]);
3162 		break;
3163 	case SYS_ACCEPT4:
3164 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3165 				    (int __user *)a[2], a[3]);
3166 		break;
3167 	default:
3168 		err = -EINVAL;
3169 		break;
3170 	}
3171 	return err;
3172 }
3173 
3174 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3175 
3176 /**
3177  *	sock_register - add a socket protocol handler
3178  *	@ops: description of protocol
3179  *
3180  *	This function is called by a protocol handler that wants to
3181  *	advertise its address family, and have it linked into the
3182  *	socket interface. The value ops->family corresponds to the
3183  *	socket system call protocol family.
3184  */
sock_register(const struct net_proto_family * ops)3185 int sock_register(const struct net_proto_family *ops)
3186 {
3187 	int err;
3188 
3189 	if (ops->family >= NPROTO) {
3190 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3191 		return -ENOBUFS;
3192 	}
3193 
3194 	spin_lock(&net_family_lock);
3195 	if (rcu_dereference_protected(net_families[ops->family],
3196 				      lockdep_is_held(&net_family_lock)))
3197 		err = -EEXIST;
3198 	else {
3199 		rcu_assign_pointer(net_families[ops->family], ops);
3200 		err = 0;
3201 	}
3202 	spin_unlock(&net_family_lock);
3203 
3204 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3205 	return err;
3206 }
3207 EXPORT_SYMBOL(sock_register);
3208 
3209 /**
3210  *	sock_unregister - remove a protocol handler
3211  *	@family: protocol family to remove
3212  *
3213  *	This function is called by a protocol handler that wants to
3214  *	remove its address family, and have it unlinked from the
3215  *	new socket creation.
3216  *
3217  *	If protocol handler is a module, then it can use module reference
3218  *	counts to protect against new references. If protocol handler is not
3219  *	a module then it needs to provide its own protection in
3220  *	the ops->create routine.
3221  */
sock_unregister(int family)3222 void sock_unregister(int family)
3223 {
3224 	BUG_ON(family < 0 || family >= NPROTO);
3225 
3226 	spin_lock(&net_family_lock);
3227 	RCU_INIT_POINTER(net_families[family], NULL);
3228 	spin_unlock(&net_family_lock);
3229 
3230 	synchronize_rcu();
3231 
3232 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3233 }
3234 EXPORT_SYMBOL(sock_unregister);
3235 
sock_is_registered(int family)3236 bool sock_is_registered(int family)
3237 {
3238 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3239 }
3240 
sock_init(void)3241 static int __init sock_init(void)
3242 {
3243 	int err;
3244 	/*
3245 	 *      Initialize the network sysctl infrastructure.
3246 	 */
3247 	err = net_sysctl_init();
3248 	if (err)
3249 		goto out;
3250 
3251 	/*
3252 	 *      Initialize skbuff SLAB cache
3253 	 */
3254 	skb_init();
3255 
3256 	/*
3257 	 *      Initialize the protocols module.
3258 	 */
3259 
3260 	init_inodecache();
3261 
3262 	err = register_filesystem(&sock_fs_type);
3263 	if (err)
3264 		goto out;
3265 	sock_mnt = kern_mount(&sock_fs_type);
3266 	if (IS_ERR(sock_mnt)) {
3267 		err = PTR_ERR(sock_mnt);
3268 		goto out_mount;
3269 	}
3270 
3271 	/* The real protocol initialization is performed in later initcalls.
3272 	 */
3273 
3274 #ifdef CONFIG_NETFILTER
3275 	err = netfilter_init();
3276 	if (err)
3277 		goto out;
3278 #endif
3279 
3280 	ptp_classifier_init();
3281 
3282 out:
3283 	return err;
3284 
3285 out_mount:
3286 	unregister_filesystem(&sock_fs_type);
3287 	goto out;
3288 }
3289 
3290 core_initcall(sock_init);	/* early initcall */
3291 
3292 #ifdef CONFIG_PROC_FS
socket_seq_show(struct seq_file * seq)3293 void socket_seq_show(struct seq_file *seq)
3294 {
3295 	seq_printf(seq, "sockets: used %d\n",
3296 		   sock_inuse_get(seq->private));
3297 }
3298 #endif				/* CONFIG_PROC_FS */
3299 
3300 /* Handle the fact that while struct ifreq has the same *layout* on
3301  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3302  * which are handled elsewhere, it still has different *size* due to
3303  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3304  * resulting in struct ifreq being 32 and 40 bytes respectively).
3305  * As a result, if the struct happens to be at the end of a page and
3306  * the next page isn't readable/writable, we get a fault. To prevent
3307  * that, copy back and forth to the full size.
3308  */
get_user_ifreq(struct ifreq * ifr,void __user ** ifrdata,void __user * arg)3309 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3310 {
3311 	if (in_compat_syscall()) {
3312 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3313 
3314 		memset(ifr, 0, sizeof(*ifr));
3315 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3316 			return -EFAULT;
3317 
3318 		if (ifrdata)
3319 			*ifrdata = compat_ptr(ifr32->ifr_data);
3320 
3321 		return 0;
3322 	}
3323 
3324 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3325 		return -EFAULT;
3326 
3327 	if (ifrdata)
3328 		*ifrdata = ifr->ifr_data;
3329 
3330 	return 0;
3331 }
3332 EXPORT_SYMBOL(get_user_ifreq);
3333 
put_user_ifreq(struct ifreq * ifr,void __user * arg)3334 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3335 {
3336 	size_t size = sizeof(*ifr);
3337 
3338 	if (in_compat_syscall())
3339 		size = sizeof(struct compat_ifreq);
3340 
3341 	if (copy_to_user(arg, ifr, size))
3342 		return -EFAULT;
3343 
3344 	return 0;
3345 }
3346 EXPORT_SYMBOL(put_user_ifreq);
3347 
3348 #ifdef CONFIG_COMPAT
compat_siocwandev(struct net * net,struct compat_ifreq __user * uifr32)3349 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3350 {
3351 	compat_uptr_t uptr32;
3352 	struct ifreq ifr;
3353 	void __user *saved;
3354 	int err;
3355 
3356 	if (get_user_ifreq(&ifr, NULL, uifr32))
3357 		return -EFAULT;
3358 
3359 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3360 		return -EFAULT;
3361 
3362 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3363 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3364 
3365 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3366 	if (!err) {
3367 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3368 		if (put_user_ifreq(&ifr, uifr32))
3369 			err = -EFAULT;
3370 	}
3371 	return err;
3372 }
3373 
3374 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
compat_ifr_data_ioctl(struct net * net,unsigned int cmd,struct compat_ifreq __user * u_ifreq32)3375 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3376 				 struct compat_ifreq __user *u_ifreq32)
3377 {
3378 	struct ifreq ifreq;
3379 	void __user *data;
3380 
3381 	if (!is_socket_ioctl_cmd(cmd))
3382 		return -ENOTTY;
3383 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3384 		return -EFAULT;
3385 	ifreq.ifr_data = data;
3386 
3387 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3388 }
3389 
compat_sock_ioctl_trans(struct file * file,struct socket * sock,unsigned int cmd,unsigned long arg)3390 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3391 			 unsigned int cmd, unsigned long arg)
3392 {
3393 	void __user *argp = compat_ptr(arg);
3394 	struct sock *sk = sock->sk;
3395 	struct net *net = sock_net(sk);
3396 	const struct proto_ops *ops;
3397 
3398 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3399 		return sock_ioctl(file, cmd, (unsigned long)argp);
3400 
3401 	switch (cmd) {
3402 	case SIOCWANDEV:
3403 		return compat_siocwandev(net, argp);
3404 	case SIOCGSTAMP_OLD:
3405 	case SIOCGSTAMPNS_OLD:
3406 		ops = READ_ONCE(sock->ops);
3407 		if (!ops->gettstamp)
3408 			return -ENOIOCTLCMD;
3409 		return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3410 				      !COMPAT_USE_64BIT_TIME);
3411 
3412 	case SIOCETHTOOL:
3413 	case SIOCBONDSLAVEINFOQUERY:
3414 	case SIOCBONDINFOQUERY:
3415 	case SIOCSHWTSTAMP:
3416 	case SIOCGHWTSTAMP:
3417 		return compat_ifr_data_ioctl(net, cmd, argp);
3418 
3419 	case FIOSETOWN:
3420 	case SIOCSPGRP:
3421 	case FIOGETOWN:
3422 	case SIOCGPGRP:
3423 	case SIOCBRADDBR:
3424 	case SIOCBRDELBR:
3425 	case SIOCBRADDIF:
3426 	case SIOCBRDELIF:
3427 	case SIOCGIFVLAN:
3428 	case SIOCSIFVLAN:
3429 	case SIOCGSKNS:
3430 	case SIOCGSTAMP_NEW:
3431 	case SIOCGSTAMPNS_NEW:
3432 	case SIOCGIFCONF:
3433 	case SIOCSIFBR:
3434 	case SIOCGIFBR:
3435 		return sock_ioctl(file, cmd, arg);
3436 
3437 	case SIOCGIFFLAGS:
3438 	case SIOCSIFFLAGS:
3439 	case SIOCGIFMAP:
3440 	case SIOCSIFMAP:
3441 	case SIOCGIFMETRIC:
3442 	case SIOCSIFMETRIC:
3443 	case SIOCGIFMTU:
3444 	case SIOCSIFMTU:
3445 	case SIOCGIFMEM:
3446 	case SIOCSIFMEM:
3447 	case SIOCGIFHWADDR:
3448 	case SIOCSIFHWADDR:
3449 	case SIOCADDMULTI:
3450 	case SIOCDELMULTI:
3451 	case SIOCGIFINDEX:
3452 	case SIOCGIFADDR:
3453 	case SIOCSIFADDR:
3454 	case SIOCSIFHWBROADCAST:
3455 	case SIOCDIFADDR:
3456 	case SIOCGIFBRDADDR:
3457 	case SIOCSIFBRDADDR:
3458 	case SIOCGIFDSTADDR:
3459 	case SIOCSIFDSTADDR:
3460 	case SIOCGIFNETMASK:
3461 	case SIOCSIFNETMASK:
3462 	case SIOCSIFPFLAGS:
3463 	case SIOCGIFPFLAGS:
3464 	case SIOCGIFTXQLEN:
3465 	case SIOCSIFTXQLEN:
3466 	case SIOCGIFNAME:
3467 	case SIOCSIFNAME:
3468 	case SIOCGMIIPHY:
3469 	case SIOCGMIIREG:
3470 	case SIOCSMIIREG:
3471 	case SIOCBONDENSLAVE:
3472 	case SIOCBONDRELEASE:
3473 	case SIOCBONDSETHWADDR:
3474 	case SIOCBONDCHANGEACTIVE:
3475 	case SIOCSARP:
3476 	case SIOCGARP:
3477 	case SIOCDARP:
3478 	case SIOCOUTQ:
3479 	case SIOCOUTQNSD:
3480 	case SIOCATMARK:
3481 		return sock_do_ioctl(net, sock, cmd, arg);
3482 	}
3483 
3484 	return -ENOIOCTLCMD;
3485 }
3486 
compat_sock_ioctl(struct file * file,unsigned int cmd,unsigned long arg)3487 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3488 			      unsigned long arg)
3489 {
3490 	struct socket *sock = file->private_data;
3491 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3492 	int ret = -ENOIOCTLCMD;
3493 	struct sock *sk;
3494 	struct net *net;
3495 
3496 	sk = sock->sk;
3497 	net = sock_net(sk);
3498 
3499 	if (ops->compat_ioctl)
3500 		ret = ops->compat_ioctl(sock, cmd, arg);
3501 
3502 	if (ret == -ENOIOCTLCMD &&
3503 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3504 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3505 
3506 	if (ret == -ENOIOCTLCMD)
3507 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3508 
3509 	return ret;
3510 }
3511 #endif
3512 
3513 /**
3514  *	kernel_bind - bind an address to a socket (kernel space)
3515  *	@sock: socket
3516  *	@addr: address
3517  *	@addrlen: length of address
3518  *
3519  *	Returns 0 or an error.
3520  */
3521 
kernel_bind(struct socket * sock,struct sockaddr * addr,int addrlen)3522 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3523 {
3524 	struct sockaddr_storage address;
3525 
3526 	memcpy(&address, addr, addrlen);
3527 
3528 	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3529 					  addrlen);
3530 }
3531 EXPORT_SYMBOL(kernel_bind);
3532 
3533 /**
3534  *	kernel_listen - move socket to listening state (kernel space)
3535  *	@sock: socket
3536  *	@backlog: pending connections queue size
3537  *
3538  *	Returns 0 or an error.
3539  */
3540 
kernel_listen(struct socket * sock,int backlog)3541 int kernel_listen(struct socket *sock, int backlog)
3542 {
3543 	return READ_ONCE(sock->ops)->listen(sock, backlog);
3544 }
3545 EXPORT_SYMBOL(kernel_listen);
3546 
3547 /**
3548  *	kernel_accept - accept a connection (kernel space)
3549  *	@sock: listening socket
3550  *	@newsock: new connected socket
3551  *	@flags: flags
3552  *
3553  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3554  *	If it fails, @newsock is guaranteed to be %NULL.
3555  *	Returns 0 or an error.
3556  */
3557 
kernel_accept(struct socket * sock,struct socket ** newsock,int flags)3558 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3559 {
3560 	struct sock *sk = sock->sk;
3561 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3562 	struct proto_accept_arg arg = {
3563 		.flags = flags,
3564 		.kern = true,
3565 	};
3566 	int err;
3567 
3568 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3569 			       newsock);
3570 	if (err < 0)
3571 		goto done;
3572 
3573 	err = ops->accept(sock, *newsock, &arg);
3574 	if (err < 0) {
3575 		sock_release(*newsock);
3576 		*newsock = NULL;
3577 		goto done;
3578 	}
3579 
3580 	(*newsock)->ops = ops;
3581 	__module_get(ops->owner);
3582 
3583 done:
3584 	return err;
3585 }
3586 EXPORT_SYMBOL(kernel_accept);
3587 
3588 /**
3589  *	kernel_connect - connect a socket (kernel space)
3590  *	@sock: socket
3591  *	@addr: address
3592  *	@addrlen: address length
3593  *	@flags: flags (O_NONBLOCK, ...)
3594  *
3595  *	For datagram sockets, @addr is the address to which datagrams are sent
3596  *	by default, and the only address from which datagrams are received.
3597  *	For stream sockets, attempts to connect to @addr.
3598  *	Returns 0 or an error code.
3599  */
3600 
kernel_connect(struct socket * sock,struct sockaddr * addr,int addrlen,int flags)3601 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3602 		   int flags)
3603 {
3604 	struct sockaddr_storage address;
3605 
3606 	memcpy(&address, addr, addrlen);
3607 
3608 	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3609 					     addrlen, flags);
3610 }
3611 EXPORT_SYMBOL(kernel_connect);
3612 
3613 /**
3614  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3615  *	@sock: socket
3616  *	@addr: address holder
3617  *
3618  * 	Fills the @addr pointer with the address which the socket is bound.
3619  *	Returns the length of the address in bytes or an error code.
3620  */
3621 
kernel_getsockname(struct socket * sock,struct sockaddr * addr)3622 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3623 {
3624 	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3625 }
3626 EXPORT_SYMBOL(kernel_getsockname);
3627 
3628 /**
3629  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3630  *	@sock: socket
3631  *	@addr: address holder
3632  *
3633  * 	Fills the @addr pointer with the address which the socket is connected.
3634  *	Returns the length of the address in bytes or an error code.
3635  */
3636 
kernel_getpeername(struct socket * sock,struct sockaddr * addr)3637 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3638 {
3639 	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3640 }
3641 EXPORT_SYMBOL(kernel_getpeername);
3642 
3643 /**
3644  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3645  *	@sock: socket
3646  *	@how: connection part
3647  *
3648  *	Returns 0 or an error.
3649  */
3650 
kernel_sock_shutdown(struct socket * sock,enum sock_shutdown_cmd how)3651 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3652 {
3653 	return READ_ONCE(sock->ops)->shutdown(sock, how);
3654 }
3655 EXPORT_SYMBOL(kernel_sock_shutdown);
3656 
3657 /**
3658  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3659  *	@sk: socket
3660  *
3661  *	This routine returns the IP overhead imposed by a socket i.e.
3662  *	the length of the underlying IP header, depending on whether
3663  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3664  *	on at the socket. Assumes that the caller has a lock on the socket.
3665  */
3666 
kernel_sock_ip_overhead(struct sock * sk)3667 u32 kernel_sock_ip_overhead(struct sock *sk)
3668 {
3669 	struct inet_sock *inet;
3670 	struct ip_options_rcu *opt;
3671 	u32 overhead = 0;
3672 #if IS_ENABLED(CONFIG_IPV6)
3673 	struct ipv6_pinfo *np;
3674 	struct ipv6_txoptions *optv6 = NULL;
3675 #endif /* IS_ENABLED(CONFIG_IPV6) */
3676 
3677 	if (!sk)
3678 		return overhead;
3679 
3680 	switch (sk->sk_family) {
3681 	case AF_INET:
3682 		inet = inet_sk(sk);
3683 		overhead += sizeof(struct iphdr);
3684 		opt = rcu_dereference_protected(inet->inet_opt,
3685 						sock_owned_by_user(sk));
3686 		if (opt)
3687 			overhead += opt->opt.optlen;
3688 		return overhead;
3689 #if IS_ENABLED(CONFIG_IPV6)
3690 	case AF_INET6:
3691 		np = inet6_sk(sk);
3692 		overhead += sizeof(struct ipv6hdr);
3693 		if (np)
3694 			optv6 = rcu_dereference_protected(np->opt,
3695 							  sock_owned_by_user(sk));
3696 		if (optv6)
3697 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3698 		return overhead;
3699 #endif /* IS_ENABLED(CONFIG_IPV6) */
3700 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3701 		return overhead;
3702 	}
3703 }
3704 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3705