xref: /linux/net/socket.c (revision 8be4d31cb8aaeea27bde4b7ddb26e28a89062ebf)
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 			/* Paired with READ_ONCE() in sk_uid() */
597 			WRITE_ONCE(sock->sk->sk_uid, iattr->ia_uid);
598 		} else {
599 			err = -ENOENT;
600 		}
601 	}
602 
603 	return err;
604 }
605 
606 static const struct inode_operations sockfs_inode_ops = {
607 	.listxattr = sockfs_listxattr,
608 	.setattr = sockfs_setattr,
609 };
610 
611 /**
612  *	sock_alloc - allocate a socket
613  *
614  *	Allocate a new inode and socket object. The two are bound together
615  *	and initialised. The socket is then returned. If we are out of inodes
616  *	NULL is returned. This functions uses GFP_KERNEL internally.
617  */
618 
sock_alloc(void)619 struct socket *sock_alloc(void)
620 {
621 	struct inode *inode;
622 	struct socket *sock;
623 
624 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
625 	if (!inode)
626 		return NULL;
627 
628 	sock = SOCKET_I(inode);
629 
630 	inode->i_ino = get_next_ino();
631 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
632 	inode->i_uid = current_fsuid();
633 	inode->i_gid = current_fsgid();
634 	inode->i_op = &sockfs_inode_ops;
635 
636 	return sock;
637 }
638 EXPORT_SYMBOL(sock_alloc);
639 
__sock_release(struct socket * sock,struct inode * inode)640 static void __sock_release(struct socket *sock, struct inode *inode)
641 {
642 	const struct proto_ops *ops = READ_ONCE(sock->ops);
643 
644 	if (ops) {
645 		struct module *owner = ops->owner;
646 
647 		if (inode)
648 			inode_lock(inode);
649 		ops->release(sock);
650 		sock->sk = NULL;
651 		if (inode)
652 			inode_unlock(inode);
653 		sock->ops = NULL;
654 		module_put(owner);
655 	}
656 
657 	if (sock->wq.fasync_list)
658 		pr_err("%s: fasync list not empty!\n", __func__);
659 
660 	if (!sock->file) {
661 		iput(SOCK_INODE(sock));
662 		return;
663 	}
664 	sock->file = NULL;
665 }
666 
667 /**
668  *	sock_release - close a socket
669  *	@sock: socket to close
670  *
671  *	The socket is released from the protocol stack if it has a release
672  *	callback, and the inode is then released if the socket is bound to
673  *	an inode not a file.
674  */
sock_release(struct socket * sock)675 void sock_release(struct socket *sock)
676 {
677 	__sock_release(sock, NULL);
678 }
679 EXPORT_SYMBOL(sock_release);
680 
__sock_tx_timestamp(__u32 tsflags,__u8 * tx_flags)681 void __sock_tx_timestamp(__u32 tsflags, __u8 *tx_flags)
682 {
683 	u8 flags = *tx_flags;
684 
685 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
686 		flags |= SKBTX_HW_TSTAMP_NOBPF;
687 
688 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
689 		flags |= SKBTX_SW_TSTAMP;
690 
691 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
692 		flags |= SKBTX_SCHED_TSTAMP;
693 
694 	if (tsflags & SOF_TIMESTAMPING_TX_COMPLETION)
695 		flags |= SKBTX_COMPLETION_TSTAMP;
696 
697 	*tx_flags = flags;
698 }
699 EXPORT_SYMBOL(__sock_tx_timestamp);
700 
701 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
702 					   size_t));
703 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
704 					    size_t));
705 
call_trace_sock_send_length(struct sock * sk,int ret,int flags)706 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
707 						 int flags)
708 {
709 	trace_sock_send_length(sk, ret, 0);
710 }
711 
sock_sendmsg_nosec(struct socket * sock,struct msghdr * msg)712 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
713 {
714 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
715 				     inet_sendmsg, sock, msg,
716 				     msg_data_left(msg));
717 	BUG_ON(ret == -EIOCBQUEUED);
718 
719 	if (trace_sock_send_length_enabled())
720 		call_trace_sock_send_length(sock->sk, ret, 0);
721 	return ret;
722 }
723 
__sock_sendmsg(struct socket * sock,struct msghdr * msg)724 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
725 {
726 	int err = security_socket_sendmsg(sock, msg,
727 					  msg_data_left(msg));
728 
729 	return err ?: sock_sendmsg_nosec(sock, msg);
730 }
731 
732 /**
733  *	sock_sendmsg - send a message through @sock
734  *	@sock: socket
735  *	@msg: message to send
736  *
737  *	Sends @msg through @sock, passing through LSM.
738  *	Returns the number of bytes sent, or an error code.
739  */
sock_sendmsg(struct socket * sock,struct msghdr * msg)740 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
741 {
742 	struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
743 	struct sockaddr_storage address;
744 	int save_len = msg->msg_namelen;
745 	int ret;
746 
747 	if (msg->msg_name) {
748 		memcpy(&address, msg->msg_name, msg->msg_namelen);
749 		msg->msg_name = &address;
750 	}
751 
752 	ret = __sock_sendmsg(sock, msg);
753 	msg->msg_name = save_addr;
754 	msg->msg_namelen = save_len;
755 
756 	return ret;
757 }
758 EXPORT_SYMBOL(sock_sendmsg);
759 
760 /**
761  *	kernel_sendmsg - send a message through @sock (kernel-space)
762  *	@sock: socket
763  *	@msg: message header
764  *	@vec: kernel vec
765  *	@num: vec array length
766  *	@size: total message data size
767  *
768  *	Builds the message data with @vec and sends it through @sock.
769  *	Returns the number of bytes sent, or an error code.
770  */
771 
kernel_sendmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size)772 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
773 		   struct kvec *vec, size_t num, size_t size)
774 {
775 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
776 	return sock_sendmsg(sock, msg);
777 }
778 EXPORT_SYMBOL(kernel_sendmsg);
779 
skb_is_err_queue(const struct sk_buff * skb)780 static bool skb_is_err_queue(const struct sk_buff *skb)
781 {
782 	/* pkt_type of skbs enqueued on the error queue are set to
783 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
784 	 * in recvmsg, since skbs received on a local socket will never
785 	 * have a pkt_type of PACKET_OUTGOING.
786 	 */
787 	return skb->pkt_type == PACKET_OUTGOING;
788 }
789 
790 /* On transmit, software and hardware timestamps are returned independently.
791  * As the two skb clones share the hardware timestamp, which may be updated
792  * before the software timestamp is received, a hardware TX timestamp may be
793  * returned only if there is no software TX timestamp. Ignore false software
794  * timestamps, which may be made in the __sock_recv_timestamp() call when the
795  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
796  * hardware timestamp.
797  */
skb_is_swtx_tstamp(const struct sk_buff * skb,int false_tstamp)798 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
799 {
800 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
801 }
802 
get_timestamp(struct sock * sk,struct sk_buff * skb,int * if_index)803 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
804 {
805 	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
806 	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
807 	struct net_device *orig_dev;
808 	ktime_t hwtstamp;
809 
810 	rcu_read_lock();
811 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
812 	if (orig_dev) {
813 		*if_index = orig_dev->ifindex;
814 		hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
815 	} else {
816 		hwtstamp = shhwtstamps->hwtstamp;
817 	}
818 	rcu_read_unlock();
819 
820 	return hwtstamp;
821 }
822 
put_ts_pktinfo(struct msghdr * msg,struct sk_buff * skb,int if_index)823 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
824 			   int if_index)
825 {
826 	struct scm_ts_pktinfo ts_pktinfo;
827 	struct net_device *orig_dev;
828 
829 	if (!skb_mac_header_was_set(skb))
830 		return;
831 
832 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
833 
834 	if (!if_index) {
835 		rcu_read_lock();
836 		orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
837 		if (orig_dev)
838 			if_index = orig_dev->ifindex;
839 		rcu_read_unlock();
840 	}
841 	ts_pktinfo.if_index = if_index;
842 
843 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
844 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
845 		 sizeof(ts_pktinfo), &ts_pktinfo);
846 }
847 
skb_has_tx_timestamp(struct sk_buff * skb,const struct sock * sk)848 bool skb_has_tx_timestamp(struct sk_buff *skb, const struct sock *sk)
849 {
850 	const struct sock_exterr_skb *serr = SKB_EXT_ERR(skb);
851 	u32 tsflags = READ_ONCE(sk->sk_tsflags);
852 
853 	if (serr->ee.ee_errno != ENOMSG ||
854 	   serr->ee.ee_origin != SO_EE_ORIGIN_TIMESTAMPING)
855 		return false;
856 
857 	/* software time stamp available and wanted */
858 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) && skb->tstamp)
859 		return true;
860 	/* hardware time stamps available and wanted */
861 	return (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
862 		skb_hwtstamps(skb)->hwtstamp;
863 }
864 
skb_get_tx_timestamp(struct sk_buff * skb,struct sock * sk,struct timespec64 * ts)865 int skb_get_tx_timestamp(struct sk_buff *skb, struct sock *sk,
866 			  struct timespec64 *ts)
867 {
868 	u32 tsflags = READ_ONCE(sk->sk_tsflags);
869 	ktime_t hwtstamp;
870 	int if_index = 0;
871 
872 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
873 	    ktime_to_timespec64_cond(skb->tstamp, ts))
874 		return SOF_TIMESTAMPING_TX_SOFTWARE;
875 
876 	if (!(tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) ||
877 	    skb_is_swtx_tstamp(skb, false))
878 		return -ENOENT;
879 
880 	if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
881 		hwtstamp = get_timestamp(sk, skb, &if_index);
882 	else
883 		hwtstamp = skb_hwtstamps(skb)->hwtstamp;
884 
885 	if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
886 		hwtstamp = ptp_convert_timestamp(&hwtstamp,
887 						READ_ONCE(sk->sk_bind_phc));
888 	if (!ktime_to_timespec64_cond(hwtstamp, ts))
889 		return -ENOENT;
890 
891 	return SOF_TIMESTAMPING_TX_HARDWARE;
892 }
893 
894 /*
895  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
896  */
__sock_recv_timestamp(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)897 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
898 	struct sk_buff *skb)
899 {
900 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
901 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
902 	struct scm_timestamping_internal tss;
903 	int empty = 1, false_tstamp = 0;
904 	struct skb_shared_hwtstamps *shhwtstamps =
905 		skb_hwtstamps(skb);
906 	int if_index;
907 	ktime_t hwtstamp;
908 	u32 tsflags;
909 
910 	/* Race occurred between timestamp enabling and packet
911 	   receiving.  Fill in the current time for now. */
912 	if (need_software_tstamp && skb->tstamp == 0) {
913 		__net_timestamp(skb);
914 		false_tstamp = 1;
915 	}
916 
917 	if (need_software_tstamp) {
918 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
919 			if (new_tstamp) {
920 				struct __kernel_sock_timeval tv;
921 
922 				skb_get_new_timestamp(skb, &tv);
923 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
924 					 sizeof(tv), &tv);
925 			} else {
926 				struct __kernel_old_timeval tv;
927 
928 				skb_get_timestamp(skb, &tv);
929 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
930 					 sizeof(tv), &tv);
931 			}
932 		} else {
933 			if (new_tstamp) {
934 				struct __kernel_timespec ts;
935 
936 				skb_get_new_timestampns(skb, &ts);
937 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
938 					 sizeof(ts), &ts);
939 			} else {
940 				struct __kernel_old_timespec ts;
941 
942 				skb_get_timestampns(skb, &ts);
943 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
944 					 sizeof(ts), &ts);
945 			}
946 		}
947 	}
948 
949 	memset(&tss, 0, sizeof(tss));
950 	tsflags = READ_ONCE(sk->sk_tsflags);
951 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE &&
952 	     (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE ||
953 	      skb_is_err_queue(skb) ||
954 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
955 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
956 		empty = 0;
957 	if (shhwtstamps &&
958 	    (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE &&
959 	     (tsflags & SOF_TIMESTAMPING_RX_HARDWARE ||
960 	      skb_is_err_queue(skb) ||
961 	      !(tsflags & SOF_TIMESTAMPING_OPT_RX_FILTER))) &&
962 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
963 		if_index = 0;
964 		if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
965 			hwtstamp = get_timestamp(sk, skb, &if_index);
966 		else
967 			hwtstamp = shhwtstamps->hwtstamp;
968 
969 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
970 			hwtstamp = ptp_convert_timestamp(&hwtstamp,
971 							 READ_ONCE(sk->sk_bind_phc));
972 
973 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
974 			empty = 0;
975 
976 			if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
977 			    !skb_is_err_queue(skb))
978 				put_ts_pktinfo(msg, skb, if_index);
979 		}
980 	}
981 	if (!empty) {
982 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
983 			put_cmsg_scm_timestamping64(msg, &tss);
984 		else
985 			put_cmsg_scm_timestamping(msg, &tss);
986 
987 		if (skb_is_err_queue(skb) && skb->len &&
988 		    SKB_EXT_ERR(skb)->opt_stats)
989 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
990 				 skb->len, skb->data);
991 	}
992 }
993 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
994 
995 #ifdef CONFIG_WIRELESS
__sock_recv_wifi_status(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)996 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
997 	struct sk_buff *skb)
998 {
999 	int ack;
1000 
1001 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
1002 		return;
1003 	if (!skb->wifi_acked_valid)
1004 		return;
1005 
1006 	ack = skb->wifi_acked;
1007 
1008 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1009 }
1010 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1011 #endif
1012 
sock_recv_drops(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1013 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1014 				   struct sk_buff *skb)
1015 {
1016 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1017 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1018 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1019 }
1020 
sock_recv_mark(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1021 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1022 			   struct sk_buff *skb)
1023 {
1024 	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1025 		/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1026 		__u32 mark = skb->mark;
1027 
1028 		put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1029 	}
1030 }
1031 
sock_recv_priority(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1032 static void sock_recv_priority(struct msghdr *msg, struct sock *sk,
1033 			       struct sk_buff *skb)
1034 {
1035 	if (sock_flag(sk, SOCK_RCVPRIORITY) && skb) {
1036 		__u32 priority = skb->priority;
1037 
1038 		put_cmsg(msg, SOL_SOCKET, SO_PRIORITY, sizeof(__u32), &priority);
1039 	}
1040 }
1041 
__sock_recv_cmsgs(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1042 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1043 		       struct sk_buff *skb)
1044 {
1045 	sock_recv_timestamp(msg, sk, skb);
1046 	sock_recv_drops(msg, sk, skb);
1047 	sock_recv_mark(msg, sk, skb);
1048 	sock_recv_priority(msg, sk, skb);
1049 }
1050 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1051 
1052 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1053 					   size_t, int));
1054 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1055 					    size_t, int));
1056 
call_trace_sock_recv_length(struct sock * sk,int ret,int flags)1057 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1058 {
1059 	trace_sock_recv_length(sk, ret, flags);
1060 }
1061 
sock_recvmsg_nosec(struct socket * sock,struct msghdr * msg,int flags)1062 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1063 				     int flags)
1064 {
1065 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1066 				     inet6_recvmsg,
1067 				     inet_recvmsg, sock, msg,
1068 				     msg_data_left(msg), flags);
1069 	if (trace_sock_recv_length_enabled())
1070 		call_trace_sock_recv_length(sock->sk, ret, flags);
1071 	return ret;
1072 }
1073 
1074 /**
1075  *	sock_recvmsg - receive a message from @sock
1076  *	@sock: socket
1077  *	@msg: message to receive
1078  *	@flags: message flags
1079  *
1080  *	Receives @msg from @sock, passing through LSM. Returns the total number
1081  *	of bytes received, or an error.
1082  */
sock_recvmsg(struct socket * sock,struct msghdr * msg,int flags)1083 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1084 {
1085 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1086 
1087 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1088 }
1089 EXPORT_SYMBOL(sock_recvmsg);
1090 
1091 /**
1092  *	kernel_recvmsg - Receive a message from a socket (kernel space)
1093  *	@sock: The socket to receive the message from
1094  *	@msg: Received message
1095  *	@vec: Input s/g array for message data
1096  *	@num: Size of input s/g array
1097  *	@size: Number of bytes to read
1098  *	@flags: Message flags (MSG_DONTWAIT, etc...)
1099  *
1100  *	On return the msg structure contains the scatter/gather array passed in the
1101  *	vec argument. The array is modified so that it consists of the unfilled
1102  *	portion of the original array.
1103  *
1104  *	The returned value is the total number of bytes received, or an error.
1105  */
1106 
kernel_recvmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size,int flags)1107 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1108 		   struct kvec *vec, size_t num, size_t size, int flags)
1109 {
1110 	msg->msg_control_is_user = false;
1111 	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1112 	return sock_recvmsg(sock, msg, flags);
1113 }
1114 EXPORT_SYMBOL(kernel_recvmsg);
1115 
sock_splice_read(struct file * file,loff_t * ppos,struct pipe_inode_info * pipe,size_t len,unsigned int flags)1116 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1117 				struct pipe_inode_info *pipe, size_t len,
1118 				unsigned int flags)
1119 {
1120 	struct socket *sock = file->private_data;
1121 	const struct proto_ops *ops;
1122 
1123 	ops = READ_ONCE(sock->ops);
1124 	if (unlikely(!ops->splice_read))
1125 		return copy_splice_read(file, ppos, pipe, len, flags);
1126 
1127 	return ops->splice_read(sock, ppos, pipe, len, flags);
1128 }
1129 
sock_splice_eof(struct file * file)1130 static void sock_splice_eof(struct file *file)
1131 {
1132 	struct socket *sock = file->private_data;
1133 	const struct proto_ops *ops;
1134 
1135 	ops = READ_ONCE(sock->ops);
1136 	if (ops->splice_eof)
1137 		ops->splice_eof(sock);
1138 }
1139 
sock_read_iter(struct kiocb * iocb,struct iov_iter * to)1140 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1141 {
1142 	struct file *file = iocb->ki_filp;
1143 	struct socket *sock = file->private_data;
1144 	struct msghdr msg = {.msg_iter = *to,
1145 			     .msg_iocb = iocb};
1146 	ssize_t res;
1147 
1148 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1149 		msg.msg_flags = MSG_DONTWAIT;
1150 
1151 	if (iocb->ki_pos != 0)
1152 		return -ESPIPE;
1153 
1154 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1155 		return 0;
1156 
1157 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1158 	*to = msg.msg_iter;
1159 	return res;
1160 }
1161 
sock_write_iter(struct kiocb * iocb,struct iov_iter * from)1162 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1163 {
1164 	struct file *file = iocb->ki_filp;
1165 	struct socket *sock = file->private_data;
1166 	struct msghdr msg = {.msg_iter = *from,
1167 			     .msg_iocb = iocb};
1168 	ssize_t res;
1169 
1170 	if (iocb->ki_pos != 0)
1171 		return -ESPIPE;
1172 
1173 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1174 		msg.msg_flags = MSG_DONTWAIT;
1175 
1176 	if (sock->type == SOCK_SEQPACKET)
1177 		msg.msg_flags |= MSG_EOR;
1178 
1179 	res = __sock_sendmsg(sock, &msg);
1180 	*from = msg.msg_iter;
1181 	return res;
1182 }
1183 
1184 /*
1185  * Atomic setting of ioctl hooks to avoid race
1186  * with module unload.
1187  */
1188 
1189 static DEFINE_MUTEX(br_ioctl_mutex);
1190 static int (*br_ioctl_hook)(struct net *net, unsigned int cmd,
1191 			    void __user *uarg);
1192 
brioctl_set(int (* hook)(struct net * net,unsigned int cmd,void __user * uarg))1193 void brioctl_set(int (*hook)(struct net *net, unsigned int cmd,
1194 			     void __user *uarg))
1195 {
1196 	mutex_lock(&br_ioctl_mutex);
1197 	br_ioctl_hook = hook;
1198 	mutex_unlock(&br_ioctl_mutex);
1199 }
1200 EXPORT_SYMBOL(brioctl_set);
1201 
br_ioctl_call(struct net * net,unsigned int cmd,void __user * uarg)1202 int br_ioctl_call(struct net *net, unsigned int cmd, void __user *uarg)
1203 {
1204 	int err = -ENOPKG;
1205 
1206 	if (!br_ioctl_hook)
1207 		request_module("bridge");
1208 
1209 	mutex_lock(&br_ioctl_mutex);
1210 	if (br_ioctl_hook)
1211 		err = br_ioctl_hook(net, cmd, uarg);
1212 	mutex_unlock(&br_ioctl_mutex);
1213 
1214 	return err;
1215 }
1216 
1217 static DEFINE_MUTEX(vlan_ioctl_mutex);
1218 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1219 
vlan_ioctl_set(int (* hook)(struct net *,void __user *))1220 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1221 {
1222 	mutex_lock(&vlan_ioctl_mutex);
1223 	vlan_ioctl_hook = hook;
1224 	mutex_unlock(&vlan_ioctl_mutex);
1225 }
1226 EXPORT_SYMBOL(vlan_ioctl_set);
1227 
sock_do_ioctl(struct net * net,struct socket * sock,unsigned int cmd,unsigned long arg)1228 static long sock_do_ioctl(struct net *net, struct socket *sock,
1229 			  unsigned int cmd, unsigned long arg)
1230 {
1231 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1232 	struct ifreq ifr;
1233 	bool need_copyout;
1234 	int err;
1235 	void __user *argp = (void __user *)arg;
1236 	void __user *data;
1237 
1238 	err = ops->ioctl(sock, cmd, arg);
1239 
1240 	/*
1241 	 * If this ioctl is unknown try to hand it down
1242 	 * to the NIC driver.
1243 	 */
1244 	if (err != -ENOIOCTLCMD)
1245 		return err;
1246 
1247 	if (!is_socket_ioctl_cmd(cmd))
1248 		return -ENOTTY;
1249 
1250 	if (get_user_ifreq(&ifr, &data, argp))
1251 		return -EFAULT;
1252 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1253 	if (!err && need_copyout)
1254 		if (put_user_ifreq(&ifr, argp))
1255 			return -EFAULT;
1256 
1257 	return err;
1258 }
1259 
1260 /*
1261  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1262  *	what to do with it - that's up to the protocol still.
1263  */
1264 
sock_ioctl(struct file * file,unsigned cmd,unsigned long arg)1265 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1266 {
1267 	const struct proto_ops  *ops;
1268 	struct socket *sock;
1269 	struct sock *sk;
1270 	void __user *argp = (void __user *)arg;
1271 	int pid, err;
1272 	struct net *net;
1273 
1274 	sock = file->private_data;
1275 	ops = READ_ONCE(sock->ops);
1276 	sk = sock->sk;
1277 	net = sock_net(sk);
1278 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1279 		struct ifreq ifr;
1280 		void __user *data;
1281 		bool need_copyout;
1282 		if (get_user_ifreq(&ifr, &data, argp))
1283 			return -EFAULT;
1284 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1285 		if (!err && need_copyout)
1286 			if (put_user_ifreq(&ifr, argp))
1287 				return -EFAULT;
1288 	} else
1289 #ifdef CONFIG_WEXT_CORE
1290 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1291 		err = wext_handle_ioctl(net, cmd, argp);
1292 	} else
1293 #endif
1294 		switch (cmd) {
1295 		case FIOSETOWN:
1296 		case SIOCSPGRP:
1297 			err = -EFAULT;
1298 			if (get_user(pid, (int __user *)argp))
1299 				break;
1300 			err = f_setown(sock->file, pid, 1);
1301 			break;
1302 		case FIOGETOWN:
1303 		case SIOCGPGRP:
1304 			err = put_user(f_getown(sock->file),
1305 				       (int __user *)argp);
1306 			break;
1307 		case SIOCGIFBR:
1308 		case SIOCSIFBR:
1309 		case SIOCBRADDBR:
1310 		case SIOCBRDELBR:
1311 		case SIOCBRADDIF:
1312 		case SIOCBRDELIF:
1313 			err = br_ioctl_call(net, cmd, argp);
1314 			break;
1315 		case SIOCGIFVLAN:
1316 		case SIOCSIFVLAN:
1317 			err = -ENOPKG;
1318 			if (!vlan_ioctl_hook)
1319 				request_module("8021q");
1320 
1321 			mutex_lock(&vlan_ioctl_mutex);
1322 			if (vlan_ioctl_hook)
1323 				err = vlan_ioctl_hook(net, argp);
1324 			mutex_unlock(&vlan_ioctl_mutex);
1325 			break;
1326 		case SIOCGSKNS:
1327 			err = -EPERM;
1328 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1329 				break;
1330 
1331 			err = open_related_ns(&net->ns, get_net_ns);
1332 			break;
1333 		case SIOCGSTAMP_OLD:
1334 		case SIOCGSTAMPNS_OLD:
1335 			if (!ops->gettstamp) {
1336 				err = -ENOIOCTLCMD;
1337 				break;
1338 			}
1339 			err = ops->gettstamp(sock, argp,
1340 					     cmd == SIOCGSTAMP_OLD,
1341 					     !IS_ENABLED(CONFIG_64BIT));
1342 			break;
1343 		case SIOCGSTAMP_NEW:
1344 		case SIOCGSTAMPNS_NEW:
1345 			if (!ops->gettstamp) {
1346 				err = -ENOIOCTLCMD;
1347 				break;
1348 			}
1349 			err = ops->gettstamp(sock, argp,
1350 					     cmd == SIOCGSTAMP_NEW,
1351 					     false);
1352 			break;
1353 
1354 		case SIOCGIFCONF:
1355 			err = dev_ifconf(net, argp);
1356 			break;
1357 
1358 		default:
1359 			err = sock_do_ioctl(net, sock, cmd, arg);
1360 			break;
1361 		}
1362 	return err;
1363 }
1364 
1365 /**
1366  *	sock_create_lite - creates a socket
1367  *	@family: protocol family (AF_INET, ...)
1368  *	@type: communication type (SOCK_STREAM, ...)
1369  *	@protocol: protocol (0, ...)
1370  *	@res: new socket
1371  *
1372  *	Creates a new socket and assigns it to @res, passing through LSM.
1373  *	The new socket initialization is not complete, see kernel_accept().
1374  *	Returns 0 or an error. On failure @res is set to %NULL.
1375  *	This function internally uses GFP_KERNEL.
1376  */
1377 
sock_create_lite(int family,int type,int protocol,struct socket ** res)1378 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1379 {
1380 	int err;
1381 	struct socket *sock = NULL;
1382 
1383 	err = security_socket_create(family, type, protocol, 1);
1384 	if (err)
1385 		goto out;
1386 
1387 	sock = sock_alloc();
1388 	if (!sock) {
1389 		err = -ENOMEM;
1390 		goto out;
1391 	}
1392 
1393 	sock->type = type;
1394 	err = security_socket_post_create(sock, family, type, protocol, 1);
1395 	if (err)
1396 		goto out_release;
1397 
1398 out:
1399 	*res = sock;
1400 	return err;
1401 out_release:
1402 	sock_release(sock);
1403 	sock = NULL;
1404 	goto out;
1405 }
1406 EXPORT_SYMBOL(sock_create_lite);
1407 
1408 /* No kernel lock held - perfect */
sock_poll(struct file * file,poll_table * wait)1409 static __poll_t sock_poll(struct file *file, poll_table *wait)
1410 {
1411 	struct socket *sock = file->private_data;
1412 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1413 	__poll_t events = poll_requested_events(wait), flag = 0;
1414 
1415 	if (!ops->poll)
1416 		return 0;
1417 
1418 	if (sk_can_busy_loop(sock->sk)) {
1419 		/* poll once if requested by the syscall */
1420 		if (events & POLL_BUSY_LOOP)
1421 			sk_busy_loop(sock->sk, 1);
1422 
1423 		/* if this socket can poll_ll, tell the system call */
1424 		flag = POLL_BUSY_LOOP;
1425 	}
1426 
1427 	return ops->poll(file, sock, wait) | flag;
1428 }
1429 
sock_mmap(struct file * file,struct vm_area_struct * vma)1430 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1431 {
1432 	struct socket *sock = file->private_data;
1433 
1434 	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1435 }
1436 
sock_close(struct inode * inode,struct file * filp)1437 static int sock_close(struct inode *inode, struct file *filp)
1438 {
1439 	__sock_release(SOCKET_I(inode), inode);
1440 	return 0;
1441 }
1442 
1443 /*
1444  *	Update the socket async list
1445  *
1446  *	Fasync_list locking strategy.
1447  *
1448  *	1. fasync_list is modified only under process context socket lock
1449  *	   i.e. under semaphore.
1450  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1451  *	   or under socket lock
1452  */
1453 
sock_fasync(int fd,struct file * filp,int on)1454 static int sock_fasync(int fd, struct file *filp, int on)
1455 {
1456 	struct socket *sock = filp->private_data;
1457 	struct sock *sk = sock->sk;
1458 	struct socket_wq *wq = &sock->wq;
1459 
1460 	if (sk == NULL)
1461 		return -EINVAL;
1462 
1463 	lock_sock(sk);
1464 	fasync_helper(fd, filp, on, &wq->fasync_list);
1465 
1466 	if (!wq->fasync_list)
1467 		sock_reset_flag(sk, SOCK_FASYNC);
1468 	else
1469 		sock_set_flag(sk, SOCK_FASYNC);
1470 
1471 	release_sock(sk);
1472 	return 0;
1473 }
1474 
1475 /* This function may be called only under rcu_lock */
1476 
sock_wake_async(struct socket_wq * wq,int how,int band)1477 int sock_wake_async(struct socket_wq *wq, int how, int band)
1478 {
1479 	if (!wq || !wq->fasync_list)
1480 		return -1;
1481 
1482 	switch (how) {
1483 	case SOCK_WAKE_WAITD:
1484 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1485 			break;
1486 		goto call_kill;
1487 	case SOCK_WAKE_SPACE:
1488 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1489 			break;
1490 		fallthrough;
1491 	case SOCK_WAKE_IO:
1492 call_kill:
1493 		kill_fasync(&wq->fasync_list, SIGIO, band);
1494 		break;
1495 	case SOCK_WAKE_URG:
1496 		kill_fasync(&wq->fasync_list, SIGURG, band);
1497 	}
1498 
1499 	return 0;
1500 }
1501 EXPORT_SYMBOL(sock_wake_async);
1502 
1503 /**
1504  *	__sock_create - creates a socket
1505  *	@net: net namespace
1506  *	@family: protocol family (AF_INET, ...)
1507  *	@type: communication type (SOCK_STREAM, ...)
1508  *	@protocol: protocol (0, ...)
1509  *	@res: new socket
1510  *	@kern: boolean for kernel space sockets
1511  *
1512  *	Creates a new socket and assigns it to @res, passing through LSM.
1513  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1514  *	be set to true if the socket resides in kernel space.
1515  *	This function internally uses GFP_KERNEL.
1516  */
1517 
__sock_create(struct net * net,int family,int type,int protocol,struct socket ** res,int kern)1518 int __sock_create(struct net *net, int family, int type, int protocol,
1519 			 struct socket **res, int kern)
1520 {
1521 	int err;
1522 	struct socket *sock;
1523 	const struct net_proto_family *pf;
1524 
1525 	/*
1526 	 *      Check protocol is in range
1527 	 */
1528 	if (family < 0 || family >= NPROTO)
1529 		return -EAFNOSUPPORT;
1530 	if (type < 0 || type >= SOCK_MAX)
1531 		return -EINVAL;
1532 
1533 	/* Compatibility.
1534 
1535 	   This uglymoron is moved from INET layer to here to avoid
1536 	   deadlock in module load.
1537 	 */
1538 	if (family == PF_INET && type == SOCK_PACKET) {
1539 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1540 			     current->comm);
1541 		family = PF_PACKET;
1542 	}
1543 
1544 	err = security_socket_create(family, type, protocol, kern);
1545 	if (err)
1546 		return err;
1547 
1548 	/*
1549 	 *	Allocate the socket and allow the family to set things up. if
1550 	 *	the protocol is 0, the family is instructed to select an appropriate
1551 	 *	default.
1552 	 */
1553 	sock = sock_alloc();
1554 	if (!sock) {
1555 		net_warn_ratelimited("socket: no more sockets\n");
1556 		return -ENFILE;	/* Not exactly a match, but its the
1557 				   closest posix thing */
1558 	}
1559 
1560 	sock->type = type;
1561 
1562 #ifdef CONFIG_MODULES
1563 	/* Attempt to load a protocol module if the find failed.
1564 	 *
1565 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1566 	 * requested real, full-featured networking support upon configuration.
1567 	 * Otherwise module support will break!
1568 	 */
1569 	if (rcu_access_pointer(net_families[family]) == NULL)
1570 		request_module("net-pf-%d", family);
1571 #endif
1572 
1573 	rcu_read_lock();
1574 	pf = rcu_dereference(net_families[family]);
1575 	err = -EAFNOSUPPORT;
1576 	if (!pf)
1577 		goto out_release;
1578 
1579 	/*
1580 	 * We will call the ->create function, that possibly is in a loadable
1581 	 * module, so we have to bump that loadable module refcnt first.
1582 	 */
1583 	if (!try_module_get(pf->owner))
1584 		goto out_release;
1585 
1586 	/* Now protected by module ref count */
1587 	rcu_read_unlock();
1588 
1589 	err = pf->create(net, sock, protocol, kern);
1590 	if (err < 0) {
1591 		/* ->create should release the allocated sock->sk object on error
1592 		 * and make sure sock->sk is set to NULL to avoid use-after-free
1593 		 */
1594 		DEBUG_NET_WARN_ONCE(sock->sk,
1595 				    "%ps must clear sock->sk on failure, family: %d, type: %d, protocol: %d\n",
1596 				    pf->create, family, type, protocol);
1597 		goto out_module_put;
1598 	}
1599 
1600 	/*
1601 	 * Now to bump the refcnt of the [loadable] module that owns this
1602 	 * socket at sock_release time we decrement its refcnt.
1603 	 */
1604 	if (!try_module_get(sock->ops->owner))
1605 		goto out_module_busy;
1606 
1607 	/*
1608 	 * Now that we're done with the ->create function, the [loadable]
1609 	 * module can have its refcnt decremented
1610 	 */
1611 	module_put(pf->owner);
1612 	err = security_socket_post_create(sock, family, type, protocol, kern);
1613 	if (err)
1614 		goto out_sock_release;
1615 	*res = sock;
1616 
1617 	return 0;
1618 
1619 out_module_busy:
1620 	err = -EAFNOSUPPORT;
1621 out_module_put:
1622 	sock->ops = NULL;
1623 	module_put(pf->owner);
1624 out_sock_release:
1625 	sock_release(sock);
1626 	return err;
1627 
1628 out_release:
1629 	rcu_read_unlock();
1630 	goto out_sock_release;
1631 }
1632 EXPORT_SYMBOL(__sock_create);
1633 
1634 /**
1635  *	sock_create - creates a socket
1636  *	@family: protocol family (AF_INET, ...)
1637  *	@type: communication type (SOCK_STREAM, ...)
1638  *	@protocol: protocol (0, ...)
1639  *	@res: new socket
1640  *
1641  *	A wrapper around __sock_create().
1642  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1643  */
1644 
sock_create(int family,int type,int protocol,struct socket ** res)1645 int sock_create(int family, int type, int protocol, struct socket **res)
1646 {
1647 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1648 }
1649 EXPORT_SYMBOL(sock_create);
1650 
1651 /**
1652  *	sock_create_kern - creates a socket (kernel space)
1653  *	@net: net namespace
1654  *	@family: protocol family (AF_INET, ...)
1655  *	@type: communication type (SOCK_STREAM, ...)
1656  *	@protocol: protocol (0, ...)
1657  *	@res: new socket
1658  *
1659  *	A wrapper around __sock_create().
1660  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1661  */
1662 
sock_create_kern(struct net * net,int family,int type,int protocol,struct socket ** res)1663 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1664 {
1665 	return __sock_create(net, family, type, protocol, res, 1);
1666 }
1667 EXPORT_SYMBOL(sock_create_kern);
1668 
__sys_socket_create(int family,int type,int protocol)1669 static struct socket *__sys_socket_create(int family, int type, int protocol)
1670 {
1671 	struct socket *sock;
1672 	int retval;
1673 
1674 	/* Check the SOCK_* constants for consistency.  */
1675 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1676 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1677 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1678 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1679 
1680 	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1681 		return ERR_PTR(-EINVAL);
1682 	type &= SOCK_TYPE_MASK;
1683 
1684 	retval = sock_create(family, type, protocol, &sock);
1685 	if (retval < 0)
1686 		return ERR_PTR(retval);
1687 
1688 	return sock;
1689 }
1690 
__sys_socket_file(int family,int type,int protocol)1691 struct file *__sys_socket_file(int family, int type, int protocol)
1692 {
1693 	struct socket *sock;
1694 	int flags;
1695 
1696 	sock = __sys_socket_create(family, type, protocol);
1697 	if (IS_ERR(sock))
1698 		return ERR_CAST(sock);
1699 
1700 	flags = type & ~SOCK_TYPE_MASK;
1701 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1702 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1703 
1704 	return sock_alloc_file(sock, flags, NULL);
1705 }
1706 
1707 /*	A hook for bpf progs to attach to and update socket protocol.
1708  *
1709  *	A static noinline declaration here could cause the compiler to
1710  *	optimize away the function. A global noinline declaration will
1711  *	keep the definition, but may optimize away the callsite.
1712  *	Therefore, __weak is needed to ensure that the call is still
1713  *	emitted, by telling the compiler that we don't know what the
1714  *	function might eventually be.
1715  */
1716 
1717 __bpf_hook_start();
1718 
update_socket_protocol(int family,int type,int protocol)1719 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1720 {
1721 	return protocol;
1722 }
1723 
1724 __bpf_hook_end();
1725 
__sys_socket(int family,int type,int protocol)1726 int __sys_socket(int family, int type, int protocol)
1727 {
1728 	struct socket *sock;
1729 	int flags;
1730 
1731 	sock = __sys_socket_create(family, type,
1732 				   update_socket_protocol(family, type, protocol));
1733 	if (IS_ERR(sock))
1734 		return PTR_ERR(sock);
1735 
1736 	flags = type & ~SOCK_TYPE_MASK;
1737 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1738 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1739 
1740 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1741 }
1742 
SYSCALL_DEFINE3(socket,int,family,int,type,int,protocol)1743 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1744 {
1745 	return __sys_socket(family, type, protocol);
1746 }
1747 
1748 /*
1749  *	Create a pair of connected sockets.
1750  */
1751 
__sys_socketpair(int family,int type,int protocol,int __user * usockvec)1752 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1753 {
1754 	struct socket *sock1, *sock2;
1755 	int fd1, fd2, err;
1756 	struct file *newfile1, *newfile2;
1757 	int flags;
1758 
1759 	flags = type & ~SOCK_TYPE_MASK;
1760 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1761 		return -EINVAL;
1762 	type &= SOCK_TYPE_MASK;
1763 
1764 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1765 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1766 
1767 	/*
1768 	 * reserve descriptors and make sure we won't fail
1769 	 * to return them to userland.
1770 	 */
1771 	fd1 = get_unused_fd_flags(flags);
1772 	if (unlikely(fd1 < 0))
1773 		return fd1;
1774 
1775 	fd2 = get_unused_fd_flags(flags);
1776 	if (unlikely(fd2 < 0)) {
1777 		put_unused_fd(fd1);
1778 		return fd2;
1779 	}
1780 
1781 	err = put_user(fd1, &usockvec[0]);
1782 	if (err)
1783 		goto out;
1784 
1785 	err = put_user(fd2, &usockvec[1]);
1786 	if (err)
1787 		goto out;
1788 
1789 	/*
1790 	 * Obtain the first socket and check if the underlying protocol
1791 	 * supports the socketpair call.
1792 	 */
1793 
1794 	err = sock_create(family, type, protocol, &sock1);
1795 	if (unlikely(err < 0))
1796 		goto out;
1797 
1798 	err = sock_create(family, type, protocol, &sock2);
1799 	if (unlikely(err < 0)) {
1800 		sock_release(sock1);
1801 		goto out;
1802 	}
1803 
1804 	err = security_socket_socketpair(sock1, sock2);
1805 	if (unlikely(err)) {
1806 		sock_release(sock2);
1807 		sock_release(sock1);
1808 		goto out;
1809 	}
1810 
1811 	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1812 	if (unlikely(err < 0)) {
1813 		sock_release(sock2);
1814 		sock_release(sock1);
1815 		goto out;
1816 	}
1817 
1818 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1819 	if (IS_ERR(newfile1)) {
1820 		err = PTR_ERR(newfile1);
1821 		sock_release(sock2);
1822 		goto out;
1823 	}
1824 
1825 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1826 	if (IS_ERR(newfile2)) {
1827 		err = PTR_ERR(newfile2);
1828 		fput(newfile1);
1829 		goto out;
1830 	}
1831 
1832 	audit_fd_pair(fd1, fd2);
1833 
1834 	fd_install(fd1, newfile1);
1835 	fd_install(fd2, newfile2);
1836 	return 0;
1837 
1838 out:
1839 	put_unused_fd(fd2);
1840 	put_unused_fd(fd1);
1841 	return err;
1842 }
1843 
SYSCALL_DEFINE4(socketpair,int,family,int,type,int,protocol,int __user *,usockvec)1844 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1845 		int __user *, usockvec)
1846 {
1847 	return __sys_socketpair(family, type, protocol, usockvec);
1848 }
1849 
__sys_bind_socket(struct socket * sock,struct sockaddr_storage * address,int addrlen)1850 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1851 		      int addrlen)
1852 {
1853 	int err;
1854 
1855 	err = security_socket_bind(sock, (struct sockaddr *)address,
1856 				   addrlen);
1857 	if (!err)
1858 		err = READ_ONCE(sock->ops)->bind(sock,
1859 						 (struct sockaddr *)address,
1860 						 addrlen);
1861 	return err;
1862 }
1863 
1864 /*
1865  *	Bind a name to a socket. Nothing much to do here since it's
1866  *	the protocol's responsibility to handle the local address.
1867  *
1868  *	We move the socket address to kernel space before we call
1869  *	the protocol layer (having also checked the address is ok).
1870  */
1871 
__sys_bind(int fd,struct sockaddr __user * umyaddr,int addrlen)1872 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1873 {
1874 	struct socket *sock;
1875 	struct sockaddr_storage address;
1876 	CLASS(fd, f)(fd);
1877 	int err;
1878 
1879 	if (fd_empty(f))
1880 		return -EBADF;
1881 	sock = sock_from_file(fd_file(f));
1882 	if (unlikely(!sock))
1883 		return -ENOTSOCK;
1884 
1885 	err = move_addr_to_kernel(umyaddr, addrlen, &address);
1886 	if (unlikely(err))
1887 		return err;
1888 
1889 	return __sys_bind_socket(sock, &address, addrlen);
1890 }
1891 
SYSCALL_DEFINE3(bind,int,fd,struct sockaddr __user *,umyaddr,int,addrlen)1892 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1893 {
1894 	return __sys_bind(fd, umyaddr, addrlen);
1895 }
1896 
1897 /*
1898  *	Perform a listen. Basically, we allow the protocol to do anything
1899  *	necessary for a listen, and if that works, we mark the socket as
1900  *	ready for listening.
1901  */
__sys_listen_socket(struct socket * sock,int backlog)1902 int __sys_listen_socket(struct socket *sock, int backlog)
1903 {
1904 	int somaxconn, err;
1905 
1906 	somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1907 	if ((unsigned int)backlog > somaxconn)
1908 		backlog = somaxconn;
1909 
1910 	err = security_socket_listen(sock, backlog);
1911 	if (!err)
1912 		err = READ_ONCE(sock->ops)->listen(sock, backlog);
1913 	return err;
1914 }
1915 
__sys_listen(int fd,int backlog)1916 int __sys_listen(int fd, int backlog)
1917 {
1918 	CLASS(fd, f)(fd);
1919 	struct socket *sock;
1920 
1921 	if (fd_empty(f))
1922 		return -EBADF;
1923 	sock = sock_from_file(fd_file(f));
1924 	if (unlikely(!sock))
1925 		return -ENOTSOCK;
1926 
1927 	return __sys_listen_socket(sock, backlog);
1928 }
1929 
SYSCALL_DEFINE2(listen,int,fd,int,backlog)1930 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1931 {
1932 	return __sys_listen(fd, backlog);
1933 }
1934 
do_accept(struct file * file,struct proto_accept_arg * arg,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1935 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1936 		       struct sockaddr __user *upeer_sockaddr,
1937 		       int __user *upeer_addrlen, int flags)
1938 {
1939 	struct socket *sock, *newsock;
1940 	struct file *newfile;
1941 	int err, len;
1942 	struct sockaddr_storage address;
1943 	const struct proto_ops *ops;
1944 
1945 	sock = sock_from_file(file);
1946 	if (!sock)
1947 		return ERR_PTR(-ENOTSOCK);
1948 
1949 	newsock = sock_alloc();
1950 	if (!newsock)
1951 		return ERR_PTR(-ENFILE);
1952 	ops = READ_ONCE(sock->ops);
1953 
1954 	newsock->type = sock->type;
1955 	newsock->ops = ops;
1956 
1957 	/*
1958 	 * We don't need try_module_get here, as the listening socket (sock)
1959 	 * has the protocol module (sock->ops->owner) held.
1960 	 */
1961 	__module_get(ops->owner);
1962 
1963 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1964 	if (IS_ERR(newfile))
1965 		return newfile;
1966 
1967 	err = security_socket_accept(sock, newsock);
1968 	if (err)
1969 		goto out_fd;
1970 
1971 	arg->flags |= sock->file->f_flags;
1972 	err = ops->accept(sock, newsock, arg);
1973 	if (err < 0)
1974 		goto out_fd;
1975 
1976 	if (upeer_sockaddr) {
1977 		len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1978 		if (len < 0) {
1979 			err = -ECONNABORTED;
1980 			goto out_fd;
1981 		}
1982 		err = move_addr_to_user(&address,
1983 					len, upeer_sockaddr, upeer_addrlen);
1984 		if (err < 0)
1985 			goto out_fd;
1986 	}
1987 
1988 	/* File flags are not inherited via accept() unlike another OSes. */
1989 	return newfile;
1990 out_fd:
1991 	fput(newfile);
1992 	return ERR_PTR(err);
1993 }
1994 
__sys_accept4_file(struct file * file,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1995 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1996 			      int __user *upeer_addrlen, int flags)
1997 {
1998 	struct proto_accept_arg arg = { };
1999 	struct file *newfile;
2000 	int newfd;
2001 
2002 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
2003 		return -EINVAL;
2004 
2005 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
2006 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
2007 
2008 	newfd = get_unused_fd_flags(flags);
2009 	if (unlikely(newfd < 0))
2010 		return newfd;
2011 
2012 	newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
2013 			    flags);
2014 	if (IS_ERR(newfile)) {
2015 		put_unused_fd(newfd);
2016 		return PTR_ERR(newfile);
2017 	}
2018 	fd_install(newfd, newfile);
2019 	return newfd;
2020 }
2021 
2022 /*
2023  *	For accept, we attempt to create a new socket, set up the link
2024  *	with the client, wake up the client, then return the new
2025  *	connected fd. We collect the address of the connector in kernel
2026  *	space and move it to user at the very end. This is unclean because
2027  *	we open the socket then return an error.
2028  *
2029  *	1003.1g adds the ability to recvmsg() to query connection pending
2030  *	status to recvmsg. We need to add that support in a way thats
2031  *	clean when we restructure accept also.
2032  */
2033 
__sys_accept4(int fd,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)2034 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2035 		  int __user *upeer_addrlen, int flags)
2036 {
2037 	CLASS(fd, f)(fd);
2038 
2039 	if (fd_empty(f))
2040 		return -EBADF;
2041 	return __sys_accept4_file(fd_file(f), upeer_sockaddr,
2042 					 upeer_addrlen, flags);
2043 }
2044 
SYSCALL_DEFINE4(accept4,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen,int,flags)2045 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2046 		int __user *, upeer_addrlen, int, flags)
2047 {
2048 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2049 }
2050 
SYSCALL_DEFINE3(accept,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen)2051 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2052 		int __user *, upeer_addrlen)
2053 {
2054 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2055 }
2056 
2057 /*
2058  *	Attempt to connect to a socket with the server address.  The address
2059  *	is in user space so we verify it is OK and move it to kernel space.
2060  *
2061  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2062  *	break bindings
2063  *
2064  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2065  *	other SEQPACKET protocols that take time to connect() as it doesn't
2066  *	include the -EINPROGRESS status for such sockets.
2067  */
2068 
__sys_connect_file(struct file * file,struct sockaddr_storage * address,int addrlen,int file_flags)2069 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2070 		       int addrlen, int file_flags)
2071 {
2072 	struct socket *sock;
2073 	int err;
2074 
2075 	sock = sock_from_file(file);
2076 	if (!sock) {
2077 		err = -ENOTSOCK;
2078 		goto out;
2079 	}
2080 
2081 	err =
2082 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2083 	if (err)
2084 		goto out;
2085 
2086 	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2087 				addrlen, sock->file->f_flags | file_flags);
2088 out:
2089 	return err;
2090 }
2091 
__sys_connect(int fd,struct sockaddr __user * uservaddr,int addrlen)2092 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2093 {
2094 	struct sockaddr_storage address;
2095 	CLASS(fd, f)(fd);
2096 	int ret;
2097 
2098 	if (fd_empty(f))
2099 		return -EBADF;
2100 
2101 	ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2102 	if (ret)
2103 		return ret;
2104 
2105 	return __sys_connect_file(fd_file(f), &address, addrlen, 0);
2106 }
2107 
SYSCALL_DEFINE3(connect,int,fd,struct sockaddr __user *,uservaddr,int,addrlen)2108 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2109 		int, addrlen)
2110 {
2111 	return __sys_connect(fd, uservaddr, addrlen);
2112 }
2113 
2114 /*
2115  *	Get the local address ('name') of a socket object. Move the obtained
2116  *	name to user space.
2117  */
2118 
__sys_getsockname(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2119 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2120 		      int __user *usockaddr_len)
2121 {
2122 	struct socket *sock;
2123 	struct sockaddr_storage address;
2124 	CLASS(fd, f)(fd);
2125 	int err;
2126 
2127 	if (fd_empty(f))
2128 		return -EBADF;
2129 	sock = sock_from_file(fd_file(f));
2130 	if (unlikely(!sock))
2131 		return -ENOTSOCK;
2132 
2133 	err = security_socket_getsockname(sock);
2134 	if (err)
2135 		return err;
2136 
2137 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2138 	if (err < 0)
2139 		return err;
2140 
2141 	/* "err" is actually length in this case */
2142 	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2143 }
2144 
SYSCALL_DEFINE3(getsockname,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2145 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2146 		int __user *, usockaddr_len)
2147 {
2148 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2149 }
2150 
2151 /*
2152  *	Get the remote address ('name') of a socket object. Move the obtained
2153  *	name to user space.
2154  */
2155 
__sys_getpeername(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2156 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2157 		      int __user *usockaddr_len)
2158 {
2159 	struct socket *sock;
2160 	struct sockaddr_storage address;
2161 	CLASS(fd, f)(fd);
2162 	int err;
2163 
2164 	if (fd_empty(f))
2165 		return -EBADF;
2166 	sock = sock_from_file(fd_file(f));
2167 	if (unlikely(!sock))
2168 		return -ENOTSOCK;
2169 
2170 	err = security_socket_getpeername(sock);
2171 	if (err)
2172 		return err;
2173 
2174 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 1);
2175 	if (err < 0)
2176 		return err;
2177 
2178 	/* "err" is actually length in this case */
2179 	return move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2180 }
2181 
SYSCALL_DEFINE3(getpeername,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2182 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2183 		int __user *, usockaddr_len)
2184 {
2185 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2186 }
2187 
2188 /*
2189  *	Send a datagram to a given address. We move the address into kernel
2190  *	space and check the user space data area is readable before invoking
2191  *	the protocol.
2192  */
__sys_sendto(int fd,void __user * buff,size_t len,unsigned int flags,struct sockaddr __user * addr,int addr_len)2193 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2194 		 struct sockaddr __user *addr,  int addr_len)
2195 {
2196 	struct socket *sock;
2197 	struct sockaddr_storage address;
2198 	int err;
2199 	struct msghdr msg;
2200 
2201 	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2202 	if (unlikely(err))
2203 		return err;
2204 
2205 	CLASS(fd, f)(fd);
2206 	if (fd_empty(f))
2207 		return -EBADF;
2208 	sock = sock_from_file(fd_file(f));
2209 	if (unlikely(!sock))
2210 		return -ENOTSOCK;
2211 
2212 	msg.msg_name = NULL;
2213 	msg.msg_control = NULL;
2214 	msg.msg_controllen = 0;
2215 	msg.msg_namelen = 0;
2216 	msg.msg_ubuf = NULL;
2217 	if (addr) {
2218 		err = move_addr_to_kernel(addr, addr_len, &address);
2219 		if (err < 0)
2220 			return err;
2221 		msg.msg_name = (struct sockaddr *)&address;
2222 		msg.msg_namelen = addr_len;
2223 	}
2224 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2225 	if (sock->file->f_flags & O_NONBLOCK)
2226 		flags |= MSG_DONTWAIT;
2227 	msg.msg_flags = flags;
2228 	return __sock_sendmsg(sock, &msg);
2229 }
2230 
SYSCALL_DEFINE6(sendto,int,fd,void __user *,buff,size_t,len,unsigned int,flags,struct sockaddr __user *,addr,int,addr_len)2231 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2232 		unsigned int, flags, struct sockaddr __user *, addr,
2233 		int, addr_len)
2234 {
2235 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2236 }
2237 
2238 /*
2239  *	Send a datagram down a socket.
2240  */
2241 
SYSCALL_DEFINE4(send,int,fd,void __user *,buff,size_t,len,unsigned int,flags)2242 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2243 		unsigned int, flags)
2244 {
2245 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2246 }
2247 
2248 /*
2249  *	Receive a frame from the socket and optionally record the address of the
2250  *	sender. We verify the buffers are writable and if needed move the
2251  *	sender address from kernel to user space.
2252  */
__sys_recvfrom(int fd,void __user * ubuf,size_t size,unsigned int flags,struct sockaddr __user * addr,int __user * addr_len)2253 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2254 		   struct sockaddr __user *addr, int __user *addr_len)
2255 {
2256 	struct sockaddr_storage address;
2257 	struct msghdr msg = {
2258 		/* Save some cycles and don't copy the address if not needed */
2259 		.msg_name = addr ? (struct sockaddr *)&address : NULL,
2260 	};
2261 	struct socket *sock;
2262 	int err, err2;
2263 
2264 	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2265 	if (unlikely(err))
2266 		return err;
2267 
2268 	CLASS(fd, f)(fd);
2269 
2270 	if (fd_empty(f))
2271 		return -EBADF;
2272 	sock = sock_from_file(fd_file(f));
2273 	if (unlikely(!sock))
2274 		return -ENOTSOCK;
2275 
2276 	if (sock->file->f_flags & O_NONBLOCK)
2277 		flags |= MSG_DONTWAIT;
2278 	err = sock_recvmsg(sock, &msg, flags);
2279 
2280 	if (err >= 0 && addr != NULL) {
2281 		err2 = move_addr_to_user(&address,
2282 					 msg.msg_namelen, addr, addr_len);
2283 		if (err2 < 0)
2284 			err = err2;
2285 	}
2286 	return err;
2287 }
2288 
SYSCALL_DEFINE6(recvfrom,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags,struct sockaddr __user *,addr,int __user *,addr_len)2289 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2290 		unsigned int, flags, struct sockaddr __user *, addr,
2291 		int __user *, addr_len)
2292 {
2293 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2294 }
2295 
2296 /*
2297  *	Receive a datagram from a socket.
2298  */
2299 
SYSCALL_DEFINE4(recv,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags)2300 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2301 		unsigned int, flags)
2302 {
2303 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2304 }
2305 
sock_use_custom_sol_socket(const struct socket * sock)2306 static bool sock_use_custom_sol_socket(const struct socket *sock)
2307 {
2308 	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2309 }
2310 
do_sock_setsockopt(struct socket * sock,bool compat,int level,int optname,sockptr_t optval,int optlen)2311 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2312 		       int optname, sockptr_t optval, int optlen)
2313 {
2314 	const struct proto_ops *ops;
2315 	char *kernel_optval = NULL;
2316 	int err;
2317 
2318 	if (optlen < 0)
2319 		return -EINVAL;
2320 
2321 	err = security_socket_setsockopt(sock, level, optname);
2322 	if (err)
2323 		goto out_put;
2324 
2325 	if (!compat)
2326 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2327 						     optval, &optlen,
2328 						     &kernel_optval);
2329 	if (err < 0)
2330 		goto out_put;
2331 	if (err > 0) {
2332 		err = 0;
2333 		goto out_put;
2334 	}
2335 
2336 	if (kernel_optval)
2337 		optval = KERNEL_SOCKPTR(kernel_optval);
2338 	ops = READ_ONCE(sock->ops);
2339 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2340 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2341 	else if (unlikely(!ops->setsockopt))
2342 		err = -EOPNOTSUPP;
2343 	else
2344 		err = ops->setsockopt(sock, level, optname, optval,
2345 					    optlen);
2346 	kfree(kernel_optval);
2347 out_put:
2348 	return err;
2349 }
2350 EXPORT_SYMBOL(do_sock_setsockopt);
2351 
2352 /* Set a socket option. Because we don't know the option lengths we have
2353  * to pass the user mode parameter for the protocols to sort out.
2354  */
__sys_setsockopt(int fd,int level,int optname,char __user * user_optval,int optlen)2355 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2356 		     int optlen)
2357 {
2358 	sockptr_t optval = USER_SOCKPTR(user_optval);
2359 	bool compat = in_compat_syscall();
2360 	struct socket *sock;
2361 	CLASS(fd, f)(fd);
2362 
2363 	if (fd_empty(f))
2364 		return -EBADF;
2365 	sock = sock_from_file(fd_file(f));
2366 	if (unlikely(!sock))
2367 		return -ENOTSOCK;
2368 
2369 	return do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2370 }
2371 
SYSCALL_DEFINE5(setsockopt,int,fd,int,level,int,optname,char __user *,optval,int,optlen)2372 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2373 		char __user *, optval, int, optlen)
2374 {
2375 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2376 }
2377 
2378 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2379 							 int optname));
2380 
do_sock_getsockopt(struct socket * sock,bool compat,int level,int optname,sockptr_t optval,sockptr_t optlen)2381 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2382 		       int optname, sockptr_t optval, sockptr_t optlen)
2383 {
2384 	int max_optlen __maybe_unused = 0;
2385 	const struct proto_ops *ops;
2386 	int err;
2387 
2388 	err = security_socket_getsockopt(sock, level, optname);
2389 	if (err)
2390 		return err;
2391 
2392 	if (!compat)
2393 		copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2394 
2395 	ops = READ_ONCE(sock->ops);
2396 	if (level == SOL_SOCKET) {
2397 		err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2398 	} else if (unlikely(!ops->getsockopt)) {
2399 		err = -EOPNOTSUPP;
2400 	} else {
2401 		if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2402 			      "Invalid argument type"))
2403 			return -EOPNOTSUPP;
2404 
2405 		err = ops->getsockopt(sock, level, optname, optval.user,
2406 				      optlen.user);
2407 	}
2408 
2409 	if (!compat)
2410 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2411 						     optval, optlen, max_optlen,
2412 						     err);
2413 
2414 	return err;
2415 }
2416 EXPORT_SYMBOL(do_sock_getsockopt);
2417 
2418 /*
2419  *	Get a socket option. Because we don't know the option lengths we have
2420  *	to pass a user mode parameter for the protocols to sort out.
2421  */
__sys_getsockopt(int fd,int level,int optname,char __user * optval,int __user * optlen)2422 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2423 		int __user *optlen)
2424 {
2425 	struct socket *sock;
2426 	CLASS(fd, f)(fd);
2427 
2428 	if (fd_empty(f))
2429 		return -EBADF;
2430 	sock = sock_from_file(fd_file(f));
2431 	if (unlikely(!sock))
2432 		return -ENOTSOCK;
2433 
2434 	return do_sock_getsockopt(sock, in_compat_syscall(), level, optname,
2435 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2436 }
2437 
SYSCALL_DEFINE5(getsockopt,int,fd,int,level,int,optname,char __user *,optval,int __user *,optlen)2438 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2439 		char __user *, optval, int __user *, optlen)
2440 {
2441 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2442 }
2443 
2444 /*
2445  *	Shutdown a socket.
2446  */
2447 
__sys_shutdown_sock(struct socket * sock,int how)2448 int __sys_shutdown_sock(struct socket *sock, int how)
2449 {
2450 	int err;
2451 
2452 	err = security_socket_shutdown(sock, how);
2453 	if (!err)
2454 		err = READ_ONCE(sock->ops)->shutdown(sock, how);
2455 
2456 	return err;
2457 }
2458 
__sys_shutdown(int fd,int how)2459 int __sys_shutdown(int fd, int how)
2460 {
2461 	struct socket *sock;
2462 	CLASS(fd, f)(fd);
2463 
2464 	if (fd_empty(f))
2465 		return -EBADF;
2466 	sock = sock_from_file(fd_file(f));
2467 	if (unlikely(!sock))
2468 		return -ENOTSOCK;
2469 
2470 	return __sys_shutdown_sock(sock, how);
2471 }
2472 
SYSCALL_DEFINE2(shutdown,int,fd,int,how)2473 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2474 {
2475 	return __sys_shutdown(fd, how);
2476 }
2477 
2478 /* A couple of helpful macros for getting the address of the 32/64 bit
2479  * fields which are the same type (int / unsigned) on our platforms.
2480  */
2481 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2482 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2483 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2484 
2485 struct used_address {
2486 	struct sockaddr_storage name;
2487 	unsigned int name_len;
2488 };
2489 
__copy_msghdr(struct msghdr * kmsg,struct user_msghdr * msg,struct sockaddr __user ** save_addr)2490 int __copy_msghdr(struct msghdr *kmsg,
2491 		  struct user_msghdr *msg,
2492 		  struct sockaddr __user **save_addr)
2493 {
2494 	ssize_t err;
2495 
2496 	kmsg->msg_control_is_user = true;
2497 	kmsg->msg_get_inq = 0;
2498 	kmsg->msg_control_user = msg->msg_control;
2499 	kmsg->msg_controllen = msg->msg_controllen;
2500 	kmsg->msg_flags = msg->msg_flags;
2501 
2502 	kmsg->msg_namelen = msg->msg_namelen;
2503 	if (!msg->msg_name)
2504 		kmsg->msg_namelen = 0;
2505 
2506 	if (kmsg->msg_namelen < 0)
2507 		return -EINVAL;
2508 
2509 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2510 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2511 
2512 	if (save_addr)
2513 		*save_addr = msg->msg_name;
2514 
2515 	if (msg->msg_name && kmsg->msg_namelen) {
2516 		if (!save_addr) {
2517 			err = move_addr_to_kernel(msg->msg_name,
2518 						  kmsg->msg_namelen,
2519 						  kmsg->msg_name);
2520 			if (err < 0)
2521 				return err;
2522 		}
2523 	} else {
2524 		kmsg->msg_name = NULL;
2525 		kmsg->msg_namelen = 0;
2526 	}
2527 
2528 	if (msg->msg_iovlen > UIO_MAXIOV)
2529 		return -EMSGSIZE;
2530 
2531 	kmsg->msg_iocb = NULL;
2532 	kmsg->msg_ubuf = NULL;
2533 	return 0;
2534 }
2535 
copy_msghdr_from_user(struct msghdr * kmsg,struct user_msghdr __user * umsg,struct sockaddr __user ** save_addr,struct iovec ** iov)2536 static int copy_msghdr_from_user(struct msghdr *kmsg,
2537 				 struct user_msghdr __user *umsg,
2538 				 struct sockaddr __user **save_addr,
2539 				 struct iovec **iov)
2540 {
2541 	struct user_msghdr msg;
2542 	ssize_t err;
2543 
2544 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2545 		return -EFAULT;
2546 
2547 	err = __copy_msghdr(kmsg, &msg, save_addr);
2548 	if (err)
2549 		return err;
2550 
2551 	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2552 			    msg.msg_iov, msg.msg_iovlen,
2553 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2554 	return err < 0 ? err : 0;
2555 }
2556 
____sys_sendmsg(struct socket * sock,struct msghdr * msg_sys,unsigned int flags,struct used_address * used_address,unsigned int allowed_msghdr_flags)2557 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2558 			   unsigned int flags, struct used_address *used_address,
2559 			   unsigned int allowed_msghdr_flags)
2560 {
2561 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2562 				__aligned(sizeof(__kernel_size_t));
2563 	/* 20 is size of ipv6_pktinfo */
2564 	unsigned char *ctl_buf = ctl;
2565 	int ctl_len;
2566 	ssize_t err;
2567 
2568 	err = -ENOBUFS;
2569 
2570 	if (msg_sys->msg_controllen > INT_MAX)
2571 		goto out;
2572 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2573 	ctl_len = msg_sys->msg_controllen;
2574 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2575 		err =
2576 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2577 						     sizeof(ctl));
2578 		if (err)
2579 			goto out;
2580 		ctl_buf = msg_sys->msg_control;
2581 		ctl_len = msg_sys->msg_controllen;
2582 	} else if (ctl_len) {
2583 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2584 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2585 		if (ctl_len > sizeof(ctl)) {
2586 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2587 			if (ctl_buf == NULL)
2588 				goto out;
2589 		}
2590 		err = -EFAULT;
2591 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2592 			goto out_freectl;
2593 		msg_sys->msg_control = ctl_buf;
2594 		msg_sys->msg_control_is_user = false;
2595 	}
2596 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2597 	msg_sys->msg_flags = flags;
2598 
2599 	if (sock->file->f_flags & O_NONBLOCK)
2600 		msg_sys->msg_flags |= MSG_DONTWAIT;
2601 	/*
2602 	 * If this is sendmmsg() and current destination address is same as
2603 	 * previously succeeded address, omit asking LSM's decision.
2604 	 * used_address->name_len is initialized to UINT_MAX so that the first
2605 	 * destination address never matches.
2606 	 */
2607 	if (used_address && msg_sys->msg_name &&
2608 	    used_address->name_len == msg_sys->msg_namelen &&
2609 	    !memcmp(&used_address->name, msg_sys->msg_name,
2610 		    used_address->name_len)) {
2611 		err = sock_sendmsg_nosec(sock, msg_sys);
2612 		goto out_freectl;
2613 	}
2614 	err = __sock_sendmsg(sock, msg_sys);
2615 	/*
2616 	 * If this is sendmmsg() and sending to current destination address was
2617 	 * successful, remember it.
2618 	 */
2619 	if (used_address && err >= 0) {
2620 		used_address->name_len = msg_sys->msg_namelen;
2621 		if (msg_sys->msg_name)
2622 			memcpy(&used_address->name, msg_sys->msg_name,
2623 			       used_address->name_len);
2624 	}
2625 
2626 out_freectl:
2627 	if (ctl_buf != ctl)
2628 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2629 out:
2630 	return err;
2631 }
2632 
sendmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct iovec ** iov)2633 static int sendmsg_copy_msghdr(struct msghdr *msg,
2634 			       struct user_msghdr __user *umsg, unsigned flags,
2635 			       struct iovec **iov)
2636 {
2637 	int err;
2638 
2639 	if (flags & MSG_CMSG_COMPAT) {
2640 		struct compat_msghdr __user *msg_compat;
2641 
2642 		msg_compat = (struct compat_msghdr __user *) umsg;
2643 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2644 	} else {
2645 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2646 	}
2647 	if (err < 0)
2648 		return err;
2649 
2650 	return 0;
2651 }
2652 
___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)2653 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2654 			 struct msghdr *msg_sys, unsigned int flags,
2655 			 struct used_address *used_address,
2656 			 unsigned int allowed_msghdr_flags)
2657 {
2658 	struct sockaddr_storage address;
2659 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2660 	ssize_t err;
2661 
2662 	msg_sys->msg_name = &address;
2663 
2664 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2665 	if (err < 0)
2666 		return err;
2667 
2668 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2669 				allowed_msghdr_flags);
2670 	kfree(iov);
2671 	return err;
2672 }
2673 
2674 /*
2675  *	BSD sendmsg interface
2676  */
__sys_sendmsg_sock(struct socket * sock,struct msghdr * msg,unsigned int flags)2677 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2678 			unsigned int flags)
2679 {
2680 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2681 }
2682 
__sys_sendmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2683 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2684 		   bool forbid_cmsg_compat)
2685 {
2686 	struct msghdr msg_sys;
2687 	struct socket *sock;
2688 
2689 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2690 		return -EINVAL;
2691 
2692 	CLASS(fd, f)(fd);
2693 
2694 	if (fd_empty(f))
2695 		return -EBADF;
2696 	sock = sock_from_file(fd_file(f));
2697 	if (unlikely(!sock))
2698 		return -ENOTSOCK;
2699 
2700 	return ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2701 }
2702 
SYSCALL_DEFINE3(sendmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2703 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2704 {
2705 	return __sys_sendmsg(fd, msg, flags, true);
2706 }
2707 
2708 /*
2709  *	Linux sendmmsg interface
2710  */
2711 
__sys_sendmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,bool forbid_cmsg_compat)2712 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2713 		   unsigned int flags, bool forbid_cmsg_compat)
2714 {
2715 	int err, datagrams;
2716 	struct socket *sock;
2717 	struct mmsghdr __user *entry;
2718 	struct compat_mmsghdr __user *compat_entry;
2719 	struct msghdr msg_sys;
2720 	struct used_address used_address;
2721 	unsigned int oflags = flags;
2722 
2723 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2724 		return -EINVAL;
2725 
2726 	if (vlen > UIO_MAXIOV)
2727 		vlen = UIO_MAXIOV;
2728 
2729 	datagrams = 0;
2730 
2731 	CLASS(fd, f)(fd);
2732 
2733 	if (fd_empty(f))
2734 		return -EBADF;
2735 	sock = sock_from_file(fd_file(f));
2736 	if (unlikely(!sock))
2737 		return -ENOTSOCK;
2738 
2739 	used_address.name_len = UINT_MAX;
2740 	entry = mmsg;
2741 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2742 	err = 0;
2743 	flags |= MSG_BATCH;
2744 
2745 	while (datagrams < vlen) {
2746 		if (datagrams == vlen - 1)
2747 			flags = oflags;
2748 
2749 		if (MSG_CMSG_COMPAT & flags) {
2750 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2751 					     &msg_sys, flags, &used_address, MSG_EOR);
2752 			if (err < 0)
2753 				break;
2754 			err = __put_user(err, &compat_entry->msg_len);
2755 			++compat_entry;
2756 		} else {
2757 			err = ___sys_sendmsg(sock,
2758 					     (struct user_msghdr __user *)entry,
2759 					     &msg_sys, flags, &used_address, MSG_EOR);
2760 			if (err < 0)
2761 				break;
2762 			err = put_user(err, &entry->msg_len);
2763 			++entry;
2764 		}
2765 
2766 		if (err)
2767 			break;
2768 		++datagrams;
2769 		if (msg_data_left(&msg_sys))
2770 			break;
2771 		cond_resched();
2772 	}
2773 
2774 	/* We only return an error if no datagrams were able to be sent */
2775 	if (datagrams != 0)
2776 		return datagrams;
2777 
2778 	return err;
2779 }
2780 
SYSCALL_DEFINE4(sendmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags)2781 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2782 		unsigned int, vlen, unsigned int, flags)
2783 {
2784 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2785 }
2786 
recvmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct sockaddr __user ** uaddr,struct iovec ** iov)2787 static int recvmsg_copy_msghdr(struct msghdr *msg,
2788 			       struct user_msghdr __user *umsg, unsigned flags,
2789 			       struct sockaddr __user **uaddr,
2790 			       struct iovec **iov)
2791 {
2792 	ssize_t err;
2793 
2794 	if (MSG_CMSG_COMPAT & flags) {
2795 		struct compat_msghdr __user *msg_compat;
2796 
2797 		msg_compat = (struct compat_msghdr __user *) umsg;
2798 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2799 	} else {
2800 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2801 	}
2802 	if (err < 0)
2803 		return err;
2804 
2805 	return 0;
2806 }
2807 
____sys_recvmsg(struct socket * sock,struct msghdr * msg_sys,struct user_msghdr __user * msg,struct sockaddr __user * uaddr,unsigned int flags,int nosec)2808 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2809 			   struct user_msghdr __user *msg,
2810 			   struct sockaddr __user *uaddr,
2811 			   unsigned int flags, int nosec)
2812 {
2813 	struct compat_msghdr __user *msg_compat =
2814 					(struct compat_msghdr __user *) msg;
2815 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2816 	struct sockaddr_storage addr;
2817 	unsigned long cmsg_ptr;
2818 	int len;
2819 	ssize_t err;
2820 
2821 	msg_sys->msg_name = &addr;
2822 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2823 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2824 
2825 	/* We assume all kernel code knows the size of sockaddr_storage */
2826 	msg_sys->msg_namelen = 0;
2827 
2828 	if (sock->file->f_flags & O_NONBLOCK)
2829 		flags |= MSG_DONTWAIT;
2830 
2831 	if (unlikely(nosec))
2832 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2833 	else
2834 		err = sock_recvmsg(sock, msg_sys, flags);
2835 
2836 	if (err < 0)
2837 		goto out;
2838 	len = err;
2839 
2840 	if (uaddr != NULL) {
2841 		err = move_addr_to_user(&addr,
2842 					msg_sys->msg_namelen, uaddr,
2843 					uaddr_len);
2844 		if (err < 0)
2845 			goto out;
2846 	}
2847 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2848 			 COMPAT_FLAGS(msg));
2849 	if (err)
2850 		goto out;
2851 	if (MSG_CMSG_COMPAT & flags)
2852 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2853 				 &msg_compat->msg_controllen);
2854 	else
2855 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2856 				 &msg->msg_controllen);
2857 	if (err)
2858 		goto out;
2859 	err = len;
2860 out:
2861 	return err;
2862 }
2863 
___sys_recvmsg(struct socket * sock,struct user_msghdr __user * msg,struct msghdr * msg_sys,unsigned int flags,int nosec)2864 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2865 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2866 {
2867 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2868 	/* user mode address pointers */
2869 	struct sockaddr __user *uaddr;
2870 	ssize_t err;
2871 
2872 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2873 	if (err < 0)
2874 		return err;
2875 
2876 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2877 	kfree(iov);
2878 	return err;
2879 }
2880 
2881 /*
2882  *	BSD recvmsg interface
2883  */
2884 
__sys_recvmsg_sock(struct socket * sock,struct msghdr * msg,struct user_msghdr __user * umsg,struct sockaddr __user * uaddr,unsigned int flags)2885 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2886 			struct user_msghdr __user *umsg,
2887 			struct sockaddr __user *uaddr, unsigned int flags)
2888 {
2889 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2890 }
2891 
__sys_recvmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2892 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2893 		   bool forbid_cmsg_compat)
2894 {
2895 	struct msghdr msg_sys;
2896 	struct socket *sock;
2897 
2898 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2899 		return -EINVAL;
2900 
2901 	CLASS(fd, f)(fd);
2902 
2903 	if (fd_empty(f))
2904 		return -EBADF;
2905 	sock = sock_from_file(fd_file(f));
2906 	if (unlikely(!sock))
2907 		return -ENOTSOCK;
2908 
2909 	return ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2910 }
2911 
SYSCALL_DEFINE3(recvmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2912 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2913 		unsigned int, flags)
2914 {
2915 	return __sys_recvmsg(fd, msg, flags, true);
2916 }
2917 
2918 /*
2919  *     Linux recvmmsg interface
2920  */
2921 
do_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct timespec64 * timeout)2922 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2923 			  unsigned int vlen, unsigned int flags,
2924 			  struct timespec64 *timeout)
2925 {
2926 	int err = 0, datagrams;
2927 	struct socket *sock;
2928 	struct mmsghdr __user *entry;
2929 	struct compat_mmsghdr __user *compat_entry;
2930 	struct msghdr msg_sys;
2931 	struct timespec64 end_time;
2932 	struct timespec64 timeout64;
2933 
2934 	if (timeout &&
2935 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2936 				    timeout->tv_nsec))
2937 		return -EINVAL;
2938 
2939 	datagrams = 0;
2940 
2941 	CLASS(fd, f)(fd);
2942 
2943 	if (fd_empty(f))
2944 		return -EBADF;
2945 	sock = sock_from_file(fd_file(f));
2946 	if (unlikely(!sock))
2947 		return -ENOTSOCK;
2948 
2949 	if (likely(!(flags & MSG_ERRQUEUE))) {
2950 		err = sock_error(sock->sk);
2951 		if (err)
2952 			return err;
2953 	}
2954 
2955 	entry = mmsg;
2956 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2957 
2958 	while (datagrams < vlen) {
2959 		/*
2960 		 * No need to ask LSM for more than the first datagram.
2961 		 */
2962 		if (MSG_CMSG_COMPAT & flags) {
2963 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2964 					     &msg_sys, flags & ~MSG_WAITFORONE,
2965 					     datagrams);
2966 			if (err < 0)
2967 				break;
2968 			err = __put_user(err, &compat_entry->msg_len);
2969 			++compat_entry;
2970 		} else {
2971 			err = ___sys_recvmsg(sock,
2972 					     (struct user_msghdr __user *)entry,
2973 					     &msg_sys, flags & ~MSG_WAITFORONE,
2974 					     datagrams);
2975 			if (err < 0)
2976 				break;
2977 			err = put_user(err, &entry->msg_len);
2978 			++entry;
2979 		}
2980 
2981 		if (err)
2982 			break;
2983 		++datagrams;
2984 
2985 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2986 		if (flags & MSG_WAITFORONE)
2987 			flags |= MSG_DONTWAIT;
2988 
2989 		if (timeout) {
2990 			ktime_get_ts64(&timeout64);
2991 			*timeout = timespec64_sub(end_time, timeout64);
2992 			if (timeout->tv_sec < 0) {
2993 				timeout->tv_sec = timeout->tv_nsec = 0;
2994 				break;
2995 			}
2996 
2997 			/* Timeout, return less than vlen datagrams */
2998 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2999 				break;
3000 		}
3001 
3002 		/* Out of band data, return right away */
3003 		if (msg_sys.msg_flags & MSG_OOB)
3004 			break;
3005 		cond_resched();
3006 	}
3007 
3008 	if (err == 0)
3009 		return datagrams;
3010 
3011 	if (datagrams == 0)
3012 		return err;
3013 
3014 	/*
3015 	 * We may return less entries than requested (vlen) if the
3016 	 * sock is non block and there aren't enough datagrams...
3017 	 */
3018 	if (err != -EAGAIN) {
3019 		/*
3020 		 * ... or  if recvmsg returns an error after we
3021 		 * received some datagrams, where we record the
3022 		 * error to return on the next call or if the
3023 		 * app asks about it using getsockopt(SO_ERROR).
3024 		 */
3025 		WRITE_ONCE(sock->sk->sk_err, -err);
3026 	}
3027 	return datagrams;
3028 }
3029 
__sys_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct __kernel_timespec __user * timeout,struct old_timespec32 __user * timeout32)3030 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3031 		   unsigned int vlen, unsigned int flags,
3032 		   struct __kernel_timespec __user *timeout,
3033 		   struct old_timespec32 __user *timeout32)
3034 {
3035 	int datagrams;
3036 	struct timespec64 timeout_sys;
3037 
3038 	if (timeout && get_timespec64(&timeout_sys, timeout))
3039 		return -EFAULT;
3040 
3041 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3042 		return -EFAULT;
3043 
3044 	if (!timeout && !timeout32)
3045 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3046 
3047 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3048 
3049 	if (datagrams <= 0)
3050 		return datagrams;
3051 
3052 	if (timeout && put_timespec64(&timeout_sys, timeout))
3053 		datagrams = -EFAULT;
3054 
3055 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3056 		datagrams = -EFAULT;
3057 
3058 	return datagrams;
3059 }
3060 
SYSCALL_DEFINE5(recvmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags,struct __kernel_timespec __user *,timeout)3061 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3062 		unsigned int, vlen, unsigned int, flags,
3063 		struct __kernel_timespec __user *, timeout)
3064 {
3065 	if (flags & MSG_CMSG_COMPAT)
3066 		return -EINVAL;
3067 
3068 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3069 }
3070 
3071 #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)3072 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3073 		unsigned int, vlen, unsigned int, flags,
3074 		struct old_timespec32 __user *, timeout)
3075 {
3076 	if (flags & MSG_CMSG_COMPAT)
3077 		return -EINVAL;
3078 
3079 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3080 }
3081 #endif
3082 
3083 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3084 /* Argument list sizes for sys_socketcall */
3085 #define AL(x) ((x) * sizeof(unsigned long))
3086 static const unsigned char nargs[21] = {
3087 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3088 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3089 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3090 	AL(4), AL(5), AL(4)
3091 };
3092 
3093 #undef AL
3094 
3095 /*
3096  *	System call vectors.
3097  *
3098  *	Argument checking cleaned up. Saved 20% in size.
3099  *  This function doesn't need to set the kernel lock because
3100  *  it is set by the callees.
3101  */
3102 
SYSCALL_DEFINE2(socketcall,int,call,unsigned long __user *,args)3103 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3104 {
3105 	unsigned long a[AUDITSC_ARGS];
3106 	unsigned long a0, a1;
3107 	int err;
3108 	unsigned int len;
3109 
3110 	if (call < 1 || call > SYS_SENDMMSG)
3111 		return -EINVAL;
3112 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3113 
3114 	len = nargs[call];
3115 	if (len > sizeof(a))
3116 		return -EINVAL;
3117 
3118 	/* copy_from_user should be SMP safe. */
3119 	if (copy_from_user(a, args, len))
3120 		return -EFAULT;
3121 
3122 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3123 	if (err)
3124 		return err;
3125 
3126 	a0 = a[0];
3127 	a1 = a[1];
3128 
3129 	switch (call) {
3130 	case SYS_SOCKET:
3131 		err = __sys_socket(a0, a1, a[2]);
3132 		break;
3133 	case SYS_BIND:
3134 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3135 		break;
3136 	case SYS_CONNECT:
3137 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3138 		break;
3139 	case SYS_LISTEN:
3140 		err = __sys_listen(a0, a1);
3141 		break;
3142 	case SYS_ACCEPT:
3143 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3144 				    (int __user *)a[2], 0);
3145 		break;
3146 	case SYS_GETSOCKNAME:
3147 		err =
3148 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
3149 				      (int __user *)a[2]);
3150 		break;
3151 	case SYS_GETPEERNAME:
3152 		err =
3153 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
3154 				      (int __user *)a[2]);
3155 		break;
3156 	case SYS_SOCKETPAIR:
3157 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3158 		break;
3159 	case SYS_SEND:
3160 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3161 				   NULL, 0);
3162 		break;
3163 	case SYS_SENDTO:
3164 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3165 				   (struct sockaddr __user *)a[4], a[5]);
3166 		break;
3167 	case SYS_RECV:
3168 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3169 				     NULL, NULL);
3170 		break;
3171 	case SYS_RECVFROM:
3172 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3173 				     (struct sockaddr __user *)a[4],
3174 				     (int __user *)a[5]);
3175 		break;
3176 	case SYS_SHUTDOWN:
3177 		err = __sys_shutdown(a0, a1);
3178 		break;
3179 	case SYS_SETSOCKOPT:
3180 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3181 				       a[4]);
3182 		break;
3183 	case SYS_GETSOCKOPT:
3184 		err =
3185 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3186 				     (int __user *)a[4]);
3187 		break;
3188 	case SYS_SENDMSG:
3189 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3190 				    a[2], true);
3191 		break;
3192 	case SYS_SENDMMSG:
3193 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3194 				     a[3], true);
3195 		break;
3196 	case SYS_RECVMSG:
3197 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3198 				    a[2], true);
3199 		break;
3200 	case SYS_RECVMMSG:
3201 		if (IS_ENABLED(CONFIG_64BIT))
3202 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3203 					     a[2], a[3],
3204 					     (struct __kernel_timespec __user *)a[4],
3205 					     NULL);
3206 		else
3207 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3208 					     a[2], a[3], NULL,
3209 					     (struct old_timespec32 __user *)a[4]);
3210 		break;
3211 	case SYS_ACCEPT4:
3212 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3213 				    (int __user *)a[2], a[3]);
3214 		break;
3215 	default:
3216 		err = -EINVAL;
3217 		break;
3218 	}
3219 	return err;
3220 }
3221 
3222 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3223 
3224 /**
3225  *	sock_register - add a socket protocol handler
3226  *	@ops: description of protocol
3227  *
3228  *	This function is called by a protocol handler that wants to
3229  *	advertise its address family, and have it linked into the
3230  *	socket interface. The value ops->family corresponds to the
3231  *	socket system call protocol family.
3232  */
sock_register(const struct net_proto_family * ops)3233 int sock_register(const struct net_proto_family *ops)
3234 {
3235 	int err;
3236 
3237 	if (ops->family >= NPROTO) {
3238 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3239 		return -ENOBUFS;
3240 	}
3241 
3242 	spin_lock(&net_family_lock);
3243 	if (rcu_dereference_protected(net_families[ops->family],
3244 				      lockdep_is_held(&net_family_lock)))
3245 		err = -EEXIST;
3246 	else {
3247 		rcu_assign_pointer(net_families[ops->family], ops);
3248 		err = 0;
3249 	}
3250 	spin_unlock(&net_family_lock);
3251 
3252 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3253 	return err;
3254 }
3255 EXPORT_SYMBOL(sock_register);
3256 
3257 /**
3258  *	sock_unregister - remove a protocol handler
3259  *	@family: protocol family to remove
3260  *
3261  *	This function is called by a protocol handler that wants to
3262  *	remove its address family, and have it unlinked from the
3263  *	new socket creation.
3264  *
3265  *	If protocol handler is a module, then it can use module reference
3266  *	counts to protect against new references. If protocol handler is not
3267  *	a module then it needs to provide its own protection in
3268  *	the ops->create routine.
3269  */
sock_unregister(int family)3270 void sock_unregister(int family)
3271 {
3272 	BUG_ON(family < 0 || family >= NPROTO);
3273 
3274 	spin_lock(&net_family_lock);
3275 	RCU_INIT_POINTER(net_families[family], NULL);
3276 	spin_unlock(&net_family_lock);
3277 
3278 	synchronize_rcu();
3279 
3280 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3281 }
3282 EXPORT_SYMBOL(sock_unregister);
3283 
sock_is_registered(int family)3284 bool sock_is_registered(int family)
3285 {
3286 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3287 }
3288 
sock_init(void)3289 static int __init sock_init(void)
3290 {
3291 	int err;
3292 	/*
3293 	 *      Initialize the network sysctl infrastructure.
3294 	 */
3295 	err = net_sysctl_init();
3296 	if (err)
3297 		goto out;
3298 
3299 	/*
3300 	 *      Initialize skbuff SLAB cache
3301 	 */
3302 	skb_init();
3303 
3304 	/*
3305 	 *      Initialize the protocols module.
3306 	 */
3307 
3308 	init_inodecache();
3309 
3310 	err = register_filesystem(&sock_fs_type);
3311 	if (err)
3312 		goto out;
3313 	sock_mnt = kern_mount(&sock_fs_type);
3314 	if (IS_ERR(sock_mnt)) {
3315 		err = PTR_ERR(sock_mnt);
3316 		goto out_mount;
3317 	}
3318 
3319 	/* The real protocol initialization is performed in later initcalls.
3320 	 */
3321 
3322 #ifdef CONFIG_NETFILTER
3323 	err = netfilter_init();
3324 	if (err)
3325 		goto out;
3326 #endif
3327 
3328 	ptp_classifier_init();
3329 
3330 out:
3331 	return err;
3332 
3333 out_mount:
3334 	unregister_filesystem(&sock_fs_type);
3335 	goto out;
3336 }
3337 
3338 core_initcall(sock_init);	/* early initcall */
3339 
3340 #ifdef CONFIG_PROC_FS
socket_seq_show(struct seq_file * seq)3341 void socket_seq_show(struct seq_file *seq)
3342 {
3343 	seq_printf(seq, "sockets: used %d\n",
3344 		   sock_inuse_get(seq->private));
3345 }
3346 #endif				/* CONFIG_PROC_FS */
3347 
3348 /* Handle the fact that while struct ifreq has the same *layout* on
3349  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3350  * which are handled elsewhere, it still has different *size* due to
3351  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3352  * resulting in struct ifreq being 32 and 40 bytes respectively).
3353  * As a result, if the struct happens to be at the end of a page and
3354  * the next page isn't readable/writable, we get a fault. To prevent
3355  * that, copy back and forth to the full size.
3356  */
get_user_ifreq(struct ifreq * ifr,void __user ** ifrdata,void __user * arg)3357 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3358 {
3359 	if (in_compat_syscall()) {
3360 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3361 
3362 		memset(ifr, 0, sizeof(*ifr));
3363 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3364 			return -EFAULT;
3365 
3366 		if (ifrdata)
3367 			*ifrdata = compat_ptr(ifr32->ifr_data);
3368 
3369 		return 0;
3370 	}
3371 
3372 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3373 		return -EFAULT;
3374 
3375 	if (ifrdata)
3376 		*ifrdata = ifr->ifr_data;
3377 
3378 	return 0;
3379 }
3380 EXPORT_SYMBOL(get_user_ifreq);
3381 
put_user_ifreq(struct ifreq * ifr,void __user * arg)3382 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3383 {
3384 	size_t size = sizeof(*ifr);
3385 
3386 	if (in_compat_syscall())
3387 		size = sizeof(struct compat_ifreq);
3388 
3389 	if (copy_to_user(arg, ifr, size))
3390 		return -EFAULT;
3391 
3392 	return 0;
3393 }
3394 EXPORT_SYMBOL(put_user_ifreq);
3395 
3396 #ifdef CONFIG_COMPAT
compat_siocwandev(struct net * net,struct compat_ifreq __user * uifr32)3397 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3398 {
3399 	compat_uptr_t uptr32;
3400 	struct ifreq ifr;
3401 	void __user *saved;
3402 	int err;
3403 
3404 	if (get_user_ifreq(&ifr, NULL, uifr32))
3405 		return -EFAULT;
3406 
3407 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3408 		return -EFAULT;
3409 
3410 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3411 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3412 
3413 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3414 	if (!err) {
3415 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3416 		if (put_user_ifreq(&ifr, uifr32))
3417 			err = -EFAULT;
3418 	}
3419 	return err;
3420 }
3421 
3422 /* 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)3423 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3424 				 struct compat_ifreq __user *u_ifreq32)
3425 {
3426 	struct ifreq ifreq;
3427 	void __user *data;
3428 
3429 	if (!is_socket_ioctl_cmd(cmd))
3430 		return -ENOTTY;
3431 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3432 		return -EFAULT;
3433 	ifreq.ifr_data = data;
3434 
3435 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3436 }
3437 
compat_sock_ioctl_trans(struct file * file,struct socket * sock,unsigned int cmd,unsigned long arg)3438 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3439 			 unsigned int cmd, unsigned long arg)
3440 {
3441 	void __user *argp = compat_ptr(arg);
3442 	struct sock *sk = sock->sk;
3443 	struct net *net = sock_net(sk);
3444 	const struct proto_ops *ops;
3445 
3446 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3447 		return sock_ioctl(file, cmd, (unsigned long)argp);
3448 
3449 	switch (cmd) {
3450 	case SIOCWANDEV:
3451 		return compat_siocwandev(net, argp);
3452 	case SIOCGSTAMP_OLD:
3453 	case SIOCGSTAMPNS_OLD:
3454 		ops = READ_ONCE(sock->ops);
3455 		if (!ops->gettstamp)
3456 			return -ENOIOCTLCMD;
3457 		return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3458 				      !COMPAT_USE_64BIT_TIME);
3459 
3460 	case SIOCETHTOOL:
3461 	case SIOCBONDSLAVEINFOQUERY:
3462 	case SIOCBONDINFOQUERY:
3463 	case SIOCSHWTSTAMP:
3464 	case SIOCGHWTSTAMP:
3465 		return compat_ifr_data_ioctl(net, cmd, argp);
3466 
3467 	case FIOSETOWN:
3468 	case SIOCSPGRP:
3469 	case FIOGETOWN:
3470 	case SIOCGPGRP:
3471 	case SIOCBRADDBR:
3472 	case SIOCBRDELBR:
3473 	case SIOCBRADDIF:
3474 	case SIOCBRDELIF:
3475 	case SIOCGIFVLAN:
3476 	case SIOCSIFVLAN:
3477 	case SIOCGSKNS:
3478 	case SIOCGSTAMP_NEW:
3479 	case SIOCGSTAMPNS_NEW:
3480 	case SIOCGIFCONF:
3481 	case SIOCSIFBR:
3482 	case SIOCGIFBR:
3483 		return sock_ioctl(file, cmd, arg);
3484 
3485 	case SIOCGIFFLAGS:
3486 	case SIOCSIFFLAGS:
3487 	case SIOCGIFMAP:
3488 	case SIOCSIFMAP:
3489 	case SIOCGIFMETRIC:
3490 	case SIOCSIFMETRIC:
3491 	case SIOCGIFMTU:
3492 	case SIOCSIFMTU:
3493 	case SIOCGIFMEM:
3494 	case SIOCSIFMEM:
3495 	case SIOCGIFHWADDR:
3496 	case SIOCSIFHWADDR:
3497 	case SIOCADDMULTI:
3498 	case SIOCDELMULTI:
3499 	case SIOCGIFINDEX:
3500 	case SIOCGIFADDR:
3501 	case SIOCSIFADDR:
3502 	case SIOCSIFHWBROADCAST:
3503 	case SIOCDIFADDR:
3504 	case SIOCGIFBRDADDR:
3505 	case SIOCSIFBRDADDR:
3506 	case SIOCGIFDSTADDR:
3507 	case SIOCSIFDSTADDR:
3508 	case SIOCGIFNETMASK:
3509 	case SIOCSIFNETMASK:
3510 	case SIOCSIFPFLAGS:
3511 	case SIOCGIFPFLAGS:
3512 	case SIOCGIFTXQLEN:
3513 	case SIOCSIFTXQLEN:
3514 	case SIOCGIFNAME:
3515 	case SIOCSIFNAME:
3516 	case SIOCGMIIPHY:
3517 	case SIOCGMIIREG:
3518 	case SIOCSMIIREG:
3519 	case SIOCBONDENSLAVE:
3520 	case SIOCBONDRELEASE:
3521 	case SIOCBONDSETHWADDR:
3522 	case SIOCBONDCHANGEACTIVE:
3523 	case SIOCSARP:
3524 	case SIOCGARP:
3525 	case SIOCDARP:
3526 	case SIOCOUTQ:
3527 	case SIOCOUTQNSD:
3528 	case SIOCATMARK:
3529 		return sock_do_ioctl(net, sock, cmd, arg);
3530 	}
3531 
3532 	return -ENOIOCTLCMD;
3533 }
3534 
compat_sock_ioctl(struct file * file,unsigned int cmd,unsigned long arg)3535 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3536 			      unsigned long arg)
3537 {
3538 	struct socket *sock = file->private_data;
3539 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3540 	int ret = -ENOIOCTLCMD;
3541 	struct sock *sk;
3542 	struct net *net;
3543 
3544 	sk = sock->sk;
3545 	net = sock_net(sk);
3546 
3547 	if (ops->compat_ioctl)
3548 		ret = ops->compat_ioctl(sock, cmd, arg);
3549 
3550 	if (ret == -ENOIOCTLCMD &&
3551 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3552 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3553 
3554 	if (ret == -ENOIOCTLCMD)
3555 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3556 
3557 	return ret;
3558 }
3559 #endif
3560 
3561 /**
3562  *	kernel_bind - bind an address to a socket (kernel space)
3563  *	@sock: socket
3564  *	@addr: address
3565  *	@addrlen: length of address
3566  *
3567  *	Returns 0 or an error.
3568  */
3569 
kernel_bind(struct socket * sock,struct sockaddr * addr,int addrlen)3570 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3571 {
3572 	struct sockaddr_storage address;
3573 
3574 	memcpy(&address, addr, addrlen);
3575 
3576 	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3577 					  addrlen);
3578 }
3579 EXPORT_SYMBOL(kernel_bind);
3580 
3581 /**
3582  *	kernel_listen - move socket to listening state (kernel space)
3583  *	@sock: socket
3584  *	@backlog: pending connections queue size
3585  *
3586  *	Returns 0 or an error.
3587  */
3588 
kernel_listen(struct socket * sock,int backlog)3589 int kernel_listen(struct socket *sock, int backlog)
3590 {
3591 	return READ_ONCE(sock->ops)->listen(sock, backlog);
3592 }
3593 EXPORT_SYMBOL(kernel_listen);
3594 
3595 /**
3596  *	kernel_accept - accept a connection (kernel space)
3597  *	@sock: listening socket
3598  *	@newsock: new connected socket
3599  *	@flags: flags
3600  *
3601  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3602  *	If it fails, @newsock is guaranteed to be %NULL.
3603  *	Returns 0 or an error.
3604  */
3605 
kernel_accept(struct socket * sock,struct socket ** newsock,int flags)3606 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3607 {
3608 	struct sock *sk = sock->sk;
3609 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3610 	struct proto_accept_arg arg = {
3611 		.flags = flags,
3612 		.kern = true,
3613 	};
3614 	int err;
3615 
3616 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3617 			       newsock);
3618 	if (err < 0)
3619 		goto done;
3620 
3621 	err = ops->accept(sock, *newsock, &arg);
3622 	if (err < 0) {
3623 		sock_release(*newsock);
3624 		*newsock = NULL;
3625 		goto done;
3626 	}
3627 
3628 	(*newsock)->ops = ops;
3629 	__module_get(ops->owner);
3630 
3631 done:
3632 	return err;
3633 }
3634 EXPORT_SYMBOL(kernel_accept);
3635 
3636 /**
3637  *	kernel_connect - connect a socket (kernel space)
3638  *	@sock: socket
3639  *	@addr: address
3640  *	@addrlen: address length
3641  *	@flags: flags (O_NONBLOCK, ...)
3642  *
3643  *	For datagram sockets, @addr is the address to which datagrams are sent
3644  *	by default, and the only address from which datagrams are received.
3645  *	For stream sockets, attempts to connect to @addr.
3646  *	Returns 0 or an error code.
3647  */
3648 
kernel_connect(struct socket * sock,struct sockaddr * addr,int addrlen,int flags)3649 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3650 		   int flags)
3651 {
3652 	struct sockaddr_storage address;
3653 
3654 	memcpy(&address, addr, addrlen);
3655 
3656 	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3657 					     addrlen, flags);
3658 }
3659 EXPORT_SYMBOL(kernel_connect);
3660 
3661 /**
3662  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3663  *	@sock: socket
3664  *	@addr: address holder
3665  *
3666  * 	Fills the @addr pointer with the address which the socket is bound.
3667  *	Returns the length of the address in bytes or an error code.
3668  */
3669 
kernel_getsockname(struct socket * sock,struct sockaddr * addr)3670 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3671 {
3672 	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3673 }
3674 EXPORT_SYMBOL(kernel_getsockname);
3675 
3676 /**
3677  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3678  *	@sock: socket
3679  *	@addr: address holder
3680  *
3681  * 	Fills the @addr pointer with the address which the socket is connected.
3682  *	Returns the length of the address in bytes or an error code.
3683  */
3684 
kernel_getpeername(struct socket * sock,struct sockaddr * addr)3685 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3686 {
3687 	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3688 }
3689 EXPORT_SYMBOL(kernel_getpeername);
3690 
3691 /**
3692  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3693  *	@sock: socket
3694  *	@how: connection part
3695  *
3696  *	Returns 0 or an error.
3697  */
3698 
kernel_sock_shutdown(struct socket * sock,enum sock_shutdown_cmd how)3699 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3700 {
3701 	return READ_ONCE(sock->ops)->shutdown(sock, how);
3702 }
3703 EXPORT_SYMBOL(kernel_sock_shutdown);
3704 
3705 /**
3706  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3707  *	@sk: socket
3708  *
3709  *	This routine returns the IP overhead imposed by a socket i.e.
3710  *	the length of the underlying IP header, depending on whether
3711  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3712  *	on at the socket. Assumes that the caller has a lock on the socket.
3713  */
3714 
kernel_sock_ip_overhead(struct sock * sk)3715 u32 kernel_sock_ip_overhead(struct sock *sk)
3716 {
3717 	struct inet_sock *inet;
3718 	struct ip_options_rcu *opt;
3719 	u32 overhead = 0;
3720 #if IS_ENABLED(CONFIG_IPV6)
3721 	struct ipv6_pinfo *np;
3722 	struct ipv6_txoptions *optv6 = NULL;
3723 #endif /* IS_ENABLED(CONFIG_IPV6) */
3724 
3725 	if (!sk)
3726 		return overhead;
3727 
3728 	switch (sk->sk_family) {
3729 	case AF_INET:
3730 		inet = inet_sk(sk);
3731 		overhead += sizeof(struct iphdr);
3732 		opt = rcu_dereference_protected(inet->inet_opt,
3733 						sock_owned_by_user(sk));
3734 		if (opt)
3735 			overhead += opt->opt.optlen;
3736 		return overhead;
3737 #if IS_ENABLED(CONFIG_IPV6)
3738 	case AF_INET6:
3739 		np = inet6_sk(sk);
3740 		overhead += sizeof(struct ipv6hdr);
3741 		if (np)
3742 			optv6 = rcu_dereference_protected(np->opt,
3743 							  sock_owned_by_user(sk));
3744 		if (optv6)
3745 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3746 		return overhead;
3747 #endif /* IS_ENABLED(CONFIG_IPV6) */
3748 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3749 		return overhead;
3750 	}
3751 }
3752 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3753