xref: /linux/net/socket.c (revision 89721e3038d181bacbd6be54354b513fdf1b4f10)
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 #ifdef CONFIG_NET_RX_BUSY_POLL
114 unsigned int sysctl_net_busy_read __read_mostly;
115 unsigned int sysctl_net_busy_poll __read_mostly;
116 #endif
117 
118 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
119 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121 
122 static int sock_close(struct inode *inode, struct file *file);
123 static __poll_t sock_poll(struct file *file,
124 			      struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 #ifdef CONFIG_COMPAT
127 static long compat_sock_ioctl(struct file *file,
128 			      unsigned int cmd, unsigned long arg);
129 #endif
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 				struct pipe_inode_info *pipe, size_t len,
133 				unsigned int flags);
134 static void sock_splice_eof(struct file *file);
135 
136 #ifdef CONFIG_PROC_FS
sock_show_fdinfo(struct seq_file * m,struct file * f)137 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
138 {
139 	struct socket *sock = f->private_data;
140 	const struct proto_ops *ops = READ_ONCE(sock->ops);
141 
142 	if (ops->show_fdinfo)
143 		ops->show_fdinfo(m, sock);
144 }
145 #else
146 #define sock_show_fdinfo NULL
147 #endif
148 
149 /*
150  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
151  *	in the operation structures but are done directly via the socketcall() multiplexor.
152  */
153 
154 static const struct file_operations socket_file_ops = {
155 	.owner =	THIS_MODULE,
156 	.llseek =	no_llseek,
157 	.read_iter =	sock_read_iter,
158 	.write_iter =	sock_write_iter,
159 	.poll =		sock_poll,
160 	.unlocked_ioctl = sock_ioctl,
161 #ifdef CONFIG_COMPAT
162 	.compat_ioctl = compat_sock_ioctl,
163 #endif
164 	.uring_cmd =    io_uring_cmd_sock,
165 	.mmap =		sock_mmap,
166 	.release =	sock_close,
167 	.fasync =	sock_fasync,
168 	.splice_write = splice_to_socket,
169 	.splice_read =	sock_splice_read,
170 	.splice_eof =	sock_splice_eof,
171 	.show_fdinfo =	sock_show_fdinfo,
172 };
173 
174 static const char * const pf_family_names[] = {
175 	[PF_UNSPEC]	= "PF_UNSPEC",
176 	[PF_UNIX]	= "PF_UNIX/PF_LOCAL",
177 	[PF_INET]	= "PF_INET",
178 	[PF_AX25]	= "PF_AX25",
179 	[PF_IPX]	= "PF_IPX",
180 	[PF_APPLETALK]	= "PF_APPLETALK",
181 	[PF_NETROM]	= "PF_NETROM",
182 	[PF_BRIDGE]	= "PF_BRIDGE",
183 	[PF_ATMPVC]	= "PF_ATMPVC",
184 	[PF_X25]	= "PF_X25",
185 	[PF_INET6]	= "PF_INET6",
186 	[PF_ROSE]	= "PF_ROSE",
187 	[PF_DECnet]	= "PF_DECnet",
188 	[PF_NETBEUI]	= "PF_NETBEUI",
189 	[PF_SECURITY]	= "PF_SECURITY",
190 	[PF_KEY]	= "PF_KEY",
191 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
192 	[PF_PACKET]	= "PF_PACKET",
193 	[PF_ASH]	= "PF_ASH",
194 	[PF_ECONET]	= "PF_ECONET",
195 	[PF_ATMSVC]	= "PF_ATMSVC",
196 	[PF_RDS]	= "PF_RDS",
197 	[PF_SNA]	= "PF_SNA",
198 	[PF_IRDA]	= "PF_IRDA",
199 	[PF_PPPOX]	= "PF_PPPOX",
200 	[PF_WANPIPE]	= "PF_WANPIPE",
201 	[PF_LLC]	= "PF_LLC",
202 	[PF_IB]		= "PF_IB",
203 	[PF_MPLS]	= "PF_MPLS",
204 	[PF_CAN]	= "PF_CAN",
205 	[PF_TIPC]	= "PF_TIPC",
206 	[PF_BLUETOOTH]	= "PF_BLUETOOTH",
207 	[PF_IUCV]	= "PF_IUCV",
208 	[PF_RXRPC]	= "PF_RXRPC",
209 	[PF_ISDN]	= "PF_ISDN",
210 	[PF_PHONET]	= "PF_PHONET",
211 	[PF_IEEE802154]	= "PF_IEEE802154",
212 	[PF_CAIF]	= "PF_CAIF",
213 	[PF_ALG]	= "PF_ALG",
214 	[PF_NFC]	= "PF_NFC",
215 	[PF_VSOCK]	= "PF_VSOCK",
216 	[PF_KCM]	= "PF_KCM",
217 	[PF_QIPCRTR]	= "PF_QIPCRTR",
218 	[PF_SMC]	= "PF_SMC",
219 	[PF_XDP]	= "PF_XDP",
220 	[PF_MCTP]	= "PF_MCTP",
221 };
222 
223 /*
224  *	The protocol list. Each protocol is registered in here.
225  */
226 
227 static DEFINE_SPINLOCK(net_family_lock);
228 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
229 
230 /*
231  * Support routines.
232  * Move socket addresses back and forth across the kernel/user
233  * divide and look after the messy bits.
234  */
235 
236 /**
237  *	move_addr_to_kernel	-	copy a socket address into kernel space
238  *	@uaddr: Address in user space
239  *	@kaddr: Address in kernel space
240  *	@ulen: Length in user space
241  *
242  *	The address is copied into kernel space. If the provided address is
243  *	too long an error code of -EINVAL is returned. If the copy gives
244  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
245  */
246 
move_addr_to_kernel(void __user * uaddr,int ulen,struct sockaddr_storage * kaddr)247 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
248 {
249 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
250 		return -EINVAL;
251 	if (ulen == 0)
252 		return 0;
253 	if (copy_from_user(kaddr, uaddr, ulen))
254 		return -EFAULT;
255 	return audit_sockaddr(ulen, kaddr);
256 }
257 
258 /**
259  *	move_addr_to_user	-	copy an address to user space
260  *	@kaddr: kernel space address
261  *	@klen: length of address in kernel
262  *	@uaddr: user space address
263  *	@ulen: pointer to user length field
264  *
265  *	The value pointed to by ulen on entry is the buffer length available.
266  *	This is overwritten with the buffer space used. -EINVAL is returned
267  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
268  *	is returned if either the buffer or the length field are not
269  *	accessible.
270  *	After copying the data up to the limit the user specifies, the true
271  *	length of the data is written over the length limit the user
272  *	specified. Zero is returned for a success.
273  */
274 
move_addr_to_user(struct sockaddr_storage * kaddr,int klen,void __user * uaddr,int __user * ulen)275 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
276 			     void __user *uaddr, int __user *ulen)
277 {
278 	int err;
279 	int len;
280 
281 	BUG_ON(klen > sizeof(struct sockaddr_storage));
282 	err = get_user(len, ulen);
283 	if (err)
284 		return err;
285 	if (len > klen)
286 		len = klen;
287 	if (len < 0)
288 		return -EINVAL;
289 	if (len) {
290 		if (audit_sockaddr(klen, kaddr))
291 			return -ENOMEM;
292 		if (copy_to_user(uaddr, kaddr, len))
293 			return -EFAULT;
294 	}
295 	/*
296 	 *      "fromlen shall refer to the value before truncation.."
297 	 *                      1003.1g
298 	 */
299 	return __put_user(klen, ulen);
300 }
301 
302 static struct kmem_cache *sock_inode_cachep __ro_after_init;
303 
sock_alloc_inode(struct super_block * sb)304 static struct inode *sock_alloc_inode(struct super_block *sb)
305 {
306 	struct socket_alloc *ei;
307 
308 	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
309 	if (!ei)
310 		return NULL;
311 	init_waitqueue_head(&ei->socket.wq.wait);
312 	ei->socket.wq.fasync_list = NULL;
313 	ei->socket.wq.flags = 0;
314 
315 	ei->socket.state = SS_UNCONNECTED;
316 	ei->socket.flags = 0;
317 	ei->socket.ops = NULL;
318 	ei->socket.sk = NULL;
319 	ei->socket.file = NULL;
320 
321 	return &ei->vfs_inode;
322 }
323 
sock_free_inode(struct inode * inode)324 static void sock_free_inode(struct inode *inode)
325 {
326 	struct socket_alloc *ei;
327 
328 	ei = container_of(inode, struct socket_alloc, vfs_inode);
329 	kmem_cache_free(sock_inode_cachep, ei);
330 }
331 
init_once(void * foo)332 static void init_once(void *foo)
333 {
334 	struct socket_alloc *ei = (struct socket_alloc *)foo;
335 
336 	inode_init_once(&ei->vfs_inode);
337 }
338 
init_inodecache(void)339 static void init_inodecache(void)
340 {
341 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
342 					      sizeof(struct socket_alloc),
343 					      0,
344 					      (SLAB_HWCACHE_ALIGN |
345 					       SLAB_RECLAIM_ACCOUNT |
346 					       SLAB_ACCOUNT),
347 					      init_once);
348 	BUG_ON(sock_inode_cachep == NULL);
349 }
350 
351 static const struct super_operations sockfs_ops = {
352 	.alloc_inode	= sock_alloc_inode,
353 	.free_inode	= sock_free_inode,
354 	.statfs		= simple_statfs,
355 };
356 
357 /*
358  * sockfs_dname() is called from d_path().
359  */
sockfs_dname(struct dentry * dentry,char * buffer,int buflen)360 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
361 {
362 	return dynamic_dname(buffer, buflen, "socket:[%lu]",
363 				d_inode(dentry)->i_ino);
364 }
365 
366 static const struct dentry_operations sockfs_dentry_operations = {
367 	.d_dname  = sockfs_dname,
368 };
369 
sockfs_xattr_get(const struct xattr_handler * handler,struct dentry * dentry,struct inode * inode,const char * suffix,void * value,size_t size)370 static int sockfs_xattr_get(const struct xattr_handler *handler,
371 			    struct dentry *dentry, struct inode *inode,
372 			    const char *suffix, void *value, size_t size)
373 {
374 	if (value) {
375 		if (dentry->d_name.len + 1 > size)
376 			return -ERANGE;
377 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
378 	}
379 	return dentry->d_name.len + 1;
380 }
381 
382 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
383 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
384 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
385 
386 static const struct xattr_handler sockfs_xattr_handler = {
387 	.name = XATTR_NAME_SOCKPROTONAME,
388 	.get = sockfs_xattr_get,
389 };
390 
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)391 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
392 				     struct mnt_idmap *idmap,
393 				     struct dentry *dentry, struct inode *inode,
394 				     const char *suffix, const void *value,
395 				     size_t size, int flags)
396 {
397 	/* Handled by LSM. */
398 	return -EAGAIN;
399 }
400 
401 static const struct xattr_handler sockfs_security_xattr_handler = {
402 	.prefix = XATTR_SECURITY_PREFIX,
403 	.set = sockfs_security_xattr_set,
404 };
405 
406 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
407 	&sockfs_xattr_handler,
408 	&sockfs_security_xattr_handler,
409 	NULL
410 };
411 
sockfs_init_fs_context(struct fs_context * fc)412 static int sockfs_init_fs_context(struct fs_context *fc)
413 {
414 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
415 	if (!ctx)
416 		return -ENOMEM;
417 	ctx->ops = &sockfs_ops;
418 	ctx->dops = &sockfs_dentry_operations;
419 	ctx->xattr = sockfs_xattr_handlers;
420 	return 0;
421 }
422 
423 static struct vfsmount *sock_mnt __read_mostly;
424 
425 static struct file_system_type sock_fs_type = {
426 	.name =		"sockfs",
427 	.init_fs_context = sockfs_init_fs_context,
428 	.kill_sb =	kill_anon_super,
429 };
430 
431 /*
432  *	Obtains the first available file descriptor and sets it up for use.
433  *
434  *	These functions create file structures and maps them to fd space
435  *	of the current process. On success it returns file descriptor
436  *	and file struct implicitly stored in sock->file.
437  *	Note that another thread may close file descriptor before we return
438  *	from this function. We use the fact that now we do not refer
439  *	to socket after mapping. If one day we will need it, this
440  *	function will increment ref. count on file by 1.
441  *
442  *	In any case returned fd MAY BE not valid!
443  *	This race condition is unavoidable
444  *	with shared fd spaces, we cannot solve it inside kernel,
445  *	but we take care of internal coherence yet.
446  */
447 
448 /**
449  *	sock_alloc_file - Bind a &socket to a &file
450  *	@sock: socket
451  *	@flags: file status flags
452  *	@dname: protocol name
453  *
454  *	Returns the &file bound with @sock, implicitly storing it
455  *	in sock->file. If dname is %NULL, sets to "".
456  *
457  *	On failure @sock is released, and an ERR pointer is returned.
458  *
459  *	This function uses GFP_KERNEL internally.
460  */
461 
sock_alloc_file(struct socket * sock,int flags,const char * dname)462 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
463 {
464 	struct file *file;
465 
466 	if (!dname)
467 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
468 
469 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
470 				O_RDWR | (flags & O_NONBLOCK),
471 				&socket_file_ops);
472 	if (IS_ERR(file)) {
473 		sock_release(sock);
474 		return file;
475 	}
476 
477 	file->f_mode |= FMODE_NOWAIT;
478 	sock->file = file;
479 	file->private_data = sock;
480 	stream_open(SOCK_INODE(sock), file);
481 	return file;
482 }
483 EXPORT_SYMBOL(sock_alloc_file);
484 
sock_map_fd(struct socket * sock,int flags)485 static int sock_map_fd(struct socket *sock, int flags)
486 {
487 	struct file *newfile;
488 	int fd = get_unused_fd_flags(flags);
489 	if (unlikely(fd < 0)) {
490 		sock_release(sock);
491 		return fd;
492 	}
493 
494 	newfile = sock_alloc_file(sock, flags, NULL);
495 	if (!IS_ERR(newfile)) {
496 		fd_install(fd, newfile);
497 		return fd;
498 	}
499 
500 	put_unused_fd(fd);
501 	return PTR_ERR(newfile);
502 }
503 
504 /**
505  *	sock_from_file - Return the &socket bounded to @file.
506  *	@file: file
507  *
508  *	On failure returns %NULL.
509  */
510 
sock_from_file(struct file * file)511 struct socket *sock_from_file(struct file *file)
512 {
513 	if (file->f_op == &socket_file_ops)
514 		return file->private_data;	/* set in sock_alloc_file */
515 
516 	return NULL;
517 }
518 EXPORT_SYMBOL(sock_from_file);
519 
520 /**
521  *	sockfd_lookup - Go from a file number to its socket slot
522  *	@fd: file handle
523  *	@err: pointer to an error code return
524  *
525  *	The file handle passed in is locked and the socket it is bound
526  *	to is returned. If an error occurs the err pointer is overwritten
527  *	with a negative errno code and NULL is returned. The function checks
528  *	for both invalid handles and passing a handle which is not a socket.
529  *
530  *	On a success the socket object pointer is returned.
531  */
532 
sockfd_lookup(int fd,int * err)533 struct socket *sockfd_lookup(int fd, int *err)
534 {
535 	struct file *file;
536 	struct socket *sock;
537 
538 	file = fget(fd);
539 	if (!file) {
540 		*err = -EBADF;
541 		return NULL;
542 	}
543 
544 	sock = sock_from_file(file);
545 	if (!sock) {
546 		*err = -ENOTSOCK;
547 		fput(file);
548 	}
549 	return sock;
550 }
551 EXPORT_SYMBOL(sockfd_lookup);
552 
sockfd_lookup_light(int fd,int * err,int * fput_needed)553 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
554 {
555 	struct fd f = fdget(fd);
556 	struct socket *sock;
557 
558 	*err = -EBADF;
559 	if (f.file) {
560 		sock = sock_from_file(f.file);
561 		if (likely(sock)) {
562 			*fput_needed = f.flags & FDPUT_FPUT;
563 			return sock;
564 		}
565 		*err = -ENOTSOCK;
566 		fdput(f);
567 	}
568 	return NULL;
569 }
570 
sockfs_listxattr(struct dentry * dentry,char * buffer,size_t size)571 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
572 				size_t size)
573 {
574 	ssize_t len;
575 	ssize_t used = 0;
576 
577 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
578 	if (len < 0)
579 		return len;
580 	used += len;
581 	if (buffer) {
582 		if (size < used)
583 			return -ERANGE;
584 		buffer += len;
585 	}
586 
587 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
588 	used += len;
589 	if (buffer) {
590 		if (size < used)
591 			return -ERANGE;
592 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
593 		buffer += len;
594 	}
595 
596 	return used;
597 }
598 
sockfs_setattr(struct mnt_idmap * idmap,struct dentry * dentry,struct iattr * iattr)599 static int sockfs_setattr(struct mnt_idmap *idmap,
600 			  struct dentry *dentry, struct iattr *iattr)
601 {
602 	int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
603 
604 	if (!err && (iattr->ia_valid & ATTR_UID)) {
605 		struct socket *sock = SOCKET_I(d_inode(dentry));
606 
607 		if (sock->sk)
608 			sock->sk->sk_uid = iattr->ia_uid;
609 		else
610 			err = -ENOENT;
611 	}
612 
613 	return err;
614 }
615 
616 static const struct inode_operations sockfs_inode_ops = {
617 	.listxattr = sockfs_listxattr,
618 	.setattr = sockfs_setattr,
619 };
620 
621 /**
622  *	sock_alloc - allocate a socket
623  *
624  *	Allocate a new inode and socket object. The two are bound together
625  *	and initialised. The socket is then returned. If we are out of inodes
626  *	NULL is returned. This functions uses GFP_KERNEL internally.
627  */
628 
sock_alloc(void)629 struct socket *sock_alloc(void)
630 {
631 	struct inode *inode;
632 	struct socket *sock;
633 
634 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
635 	if (!inode)
636 		return NULL;
637 
638 	sock = SOCKET_I(inode);
639 
640 	inode->i_ino = get_next_ino();
641 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
642 	inode->i_uid = current_fsuid();
643 	inode->i_gid = current_fsgid();
644 	inode->i_op = &sockfs_inode_ops;
645 
646 	return sock;
647 }
648 EXPORT_SYMBOL(sock_alloc);
649 
__sock_release(struct socket * sock,struct inode * inode)650 static void __sock_release(struct socket *sock, struct inode *inode)
651 {
652 	const struct proto_ops *ops = READ_ONCE(sock->ops);
653 
654 	if (ops) {
655 		struct module *owner = ops->owner;
656 
657 		if (inode)
658 			inode_lock(inode);
659 		ops->release(sock);
660 		sock->sk = NULL;
661 		if (inode)
662 			inode_unlock(inode);
663 		sock->ops = NULL;
664 		module_put(owner);
665 	}
666 
667 	if (sock->wq.fasync_list)
668 		pr_err("%s: fasync list not empty!\n", __func__);
669 
670 	if (!sock->file) {
671 		iput(SOCK_INODE(sock));
672 		return;
673 	}
674 	sock->file = NULL;
675 }
676 
677 /**
678  *	sock_release - close a socket
679  *	@sock: socket to close
680  *
681  *	The socket is released from the protocol stack if it has a release
682  *	callback, and the inode is then released if the socket is bound to
683  *	an inode not a file.
684  */
sock_release(struct socket * sock)685 void sock_release(struct socket *sock)
686 {
687 	__sock_release(sock, NULL);
688 }
689 EXPORT_SYMBOL(sock_release);
690 
__sock_tx_timestamp(__u16 tsflags,__u8 * tx_flags)691 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
692 {
693 	u8 flags = *tx_flags;
694 
695 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
696 		flags |= SKBTX_HW_TSTAMP;
697 
698 		/* PTP hardware clocks can provide a free running cycle counter
699 		 * as a time base for virtual clocks. Tell driver to use the
700 		 * free running cycle counter for timestamp if socket is bound
701 		 * to virtual clock.
702 		 */
703 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
704 			flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
705 	}
706 
707 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
708 		flags |= SKBTX_SW_TSTAMP;
709 
710 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
711 		flags |= SKBTX_SCHED_TSTAMP;
712 
713 	*tx_flags = flags;
714 }
715 EXPORT_SYMBOL(__sock_tx_timestamp);
716 
717 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
718 					   size_t));
719 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
720 					    size_t));
721 
call_trace_sock_send_length(struct sock * sk,int ret,int flags)722 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
723 						 int flags)
724 {
725 	trace_sock_send_length(sk, ret, 0);
726 }
727 
sock_sendmsg_nosec(struct socket * sock,struct msghdr * msg)728 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
729 {
730 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
731 				     inet_sendmsg, sock, msg,
732 				     msg_data_left(msg));
733 	BUG_ON(ret == -EIOCBQUEUED);
734 
735 	if (trace_sock_send_length_enabled())
736 		call_trace_sock_send_length(sock->sk, ret, 0);
737 	return ret;
738 }
739 
__sock_sendmsg(struct socket * sock,struct msghdr * msg)740 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
741 {
742 	int err = security_socket_sendmsg(sock, msg,
743 					  msg_data_left(msg));
744 
745 	return err ?: sock_sendmsg_nosec(sock, msg);
746 }
747 
748 /**
749  *	sock_sendmsg - send a message through @sock
750  *	@sock: socket
751  *	@msg: message to send
752  *
753  *	Sends @msg through @sock, passing through LSM.
754  *	Returns the number of bytes sent, or an error code.
755  */
sock_sendmsg(struct socket * sock,struct msghdr * msg)756 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
757 {
758 	struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
759 	struct sockaddr_storage address;
760 	int save_len = msg->msg_namelen;
761 	int ret;
762 
763 	if (msg->msg_name) {
764 		memcpy(&address, msg->msg_name, msg->msg_namelen);
765 		msg->msg_name = &address;
766 	}
767 
768 	ret = __sock_sendmsg(sock, msg);
769 	msg->msg_name = save_addr;
770 	msg->msg_namelen = save_len;
771 
772 	return ret;
773 }
774 EXPORT_SYMBOL(sock_sendmsg);
775 
776 /**
777  *	kernel_sendmsg - send a message through @sock (kernel-space)
778  *	@sock: socket
779  *	@msg: message header
780  *	@vec: kernel vec
781  *	@num: vec array length
782  *	@size: total message data size
783  *
784  *	Builds the message data with @vec and sends it through @sock.
785  *	Returns the number of bytes sent, or an error code.
786  */
787 
kernel_sendmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size)788 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
789 		   struct kvec *vec, size_t num, size_t size)
790 {
791 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
792 	return sock_sendmsg(sock, msg);
793 }
794 EXPORT_SYMBOL(kernel_sendmsg);
795 
796 /**
797  *	kernel_sendmsg_locked - send a message through @sock (kernel-space)
798  *	@sk: sock
799  *	@msg: message header
800  *	@vec: output s/g array
801  *	@num: output s/g array length
802  *	@size: total message data size
803  *
804  *	Builds the message data with @vec and sends it through @sock.
805  *	Returns the number of bytes sent, or an error code.
806  *	Caller must hold @sk.
807  */
808 
kernel_sendmsg_locked(struct sock * sk,struct msghdr * msg,struct kvec * vec,size_t num,size_t size)809 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
810 			  struct kvec *vec, size_t num, size_t size)
811 {
812 	struct socket *sock = sk->sk_socket;
813 	const struct proto_ops *ops = READ_ONCE(sock->ops);
814 
815 	if (!ops->sendmsg_locked)
816 		return sock_no_sendmsg_locked(sk, msg, size);
817 
818 	iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
819 
820 	return ops->sendmsg_locked(sk, msg, msg_data_left(msg));
821 }
822 EXPORT_SYMBOL(kernel_sendmsg_locked);
823 
skb_is_err_queue(const struct sk_buff * skb)824 static bool skb_is_err_queue(const struct sk_buff *skb)
825 {
826 	/* pkt_type of skbs enqueued on the error queue are set to
827 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
828 	 * in recvmsg, since skbs received on a local socket will never
829 	 * have a pkt_type of PACKET_OUTGOING.
830 	 */
831 	return skb->pkt_type == PACKET_OUTGOING;
832 }
833 
834 /* On transmit, software and hardware timestamps are returned independently.
835  * As the two skb clones share the hardware timestamp, which may be updated
836  * before the software timestamp is received, a hardware TX timestamp may be
837  * returned only if there is no software TX timestamp. Ignore false software
838  * timestamps, which may be made in the __sock_recv_timestamp() call when the
839  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
840  * hardware timestamp.
841  */
skb_is_swtx_tstamp(const struct sk_buff * skb,int false_tstamp)842 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
843 {
844 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
845 }
846 
get_timestamp(struct sock * sk,struct sk_buff * skb,int * if_index)847 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
848 {
849 	bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
850 	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
851 	struct net_device *orig_dev;
852 	ktime_t hwtstamp;
853 
854 	rcu_read_lock();
855 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
856 	if (orig_dev) {
857 		*if_index = orig_dev->ifindex;
858 		hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
859 	} else {
860 		hwtstamp = shhwtstamps->hwtstamp;
861 	}
862 	rcu_read_unlock();
863 
864 	return hwtstamp;
865 }
866 
put_ts_pktinfo(struct msghdr * msg,struct sk_buff * skb,int if_index)867 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
868 			   int if_index)
869 {
870 	struct scm_ts_pktinfo ts_pktinfo;
871 	struct net_device *orig_dev;
872 
873 	if (!skb_mac_header_was_set(skb))
874 		return;
875 
876 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
877 
878 	if (!if_index) {
879 		rcu_read_lock();
880 		orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
881 		if (orig_dev)
882 			if_index = orig_dev->ifindex;
883 		rcu_read_unlock();
884 	}
885 	ts_pktinfo.if_index = if_index;
886 
887 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
888 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
889 		 sizeof(ts_pktinfo), &ts_pktinfo);
890 }
891 
892 /*
893  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
894  */
__sock_recv_timestamp(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)895 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
896 	struct sk_buff *skb)
897 {
898 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
899 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
900 	struct scm_timestamping_internal tss;
901 	int empty = 1, false_tstamp = 0;
902 	struct skb_shared_hwtstamps *shhwtstamps =
903 		skb_hwtstamps(skb);
904 	int if_index;
905 	ktime_t hwtstamp;
906 	u32 tsflags;
907 
908 	/* Race occurred between timestamp enabling and packet
909 	   receiving.  Fill in the current time for now. */
910 	if (need_software_tstamp && skb->tstamp == 0) {
911 		__net_timestamp(skb);
912 		false_tstamp = 1;
913 	}
914 
915 	if (need_software_tstamp) {
916 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
917 			if (new_tstamp) {
918 				struct __kernel_sock_timeval tv;
919 
920 				skb_get_new_timestamp(skb, &tv);
921 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
922 					 sizeof(tv), &tv);
923 			} else {
924 				struct __kernel_old_timeval tv;
925 
926 				skb_get_timestamp(skb, &tv);
927 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
928 					 sizeof(tv), &tv);
929 			}
930 		} else {
931 			if (new_tstamp) {
932 				struct __kernel_timespec ts;
933 
934 				skb_get_new_timestampns(skb, &ts);
935 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
936 					 sizeof(ts), &ts);
937 			} else {
938 				struct __kernel_old_timespec ts;
939 
940 				skb_get_timestampns(skb, &ts);
941 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
942 					 sizeof(ts), &ts);
943 			}
944 		}
945 	}
946 
947 	memset(&tss, 0, sizeof(tss));
948 	tsflags = READ_ONCE(sk->sk_tsflags);
949 	if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
950 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
951 		empty = 0;
952 	if (shhwtstamps &&
953 	    (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
954 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
955 		if_index = 0;
956 		if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
957 			hwtstamp = get_timestamp(sk, skb, &if_index);
958 		else
959 			hwtstamp = shhwtstamps->hwtstamp;
960 
961 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
962 			hwtstamp = ptp_convert_timestamp(&hwtstamp,
963 							 READ_ONCE(sk->sk_bind_phc));
964 
965 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
966 			empty = 0;
967 
968 			if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
969 			    !skb_is_err_queue(skb))
970 				put_ts_pktinfo(msg, skb, if_index);
971 		}
972 	}
973 	if (!empty) {
974 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
975 			put_cmsg_scm_timestamping64(msg, &tss);
976 		else
977 			put_cmsg_scm_timestamping(msg, &tss);
978 
979 		if (skb_is_err_queue(skb) && skb->len &&
980 		    SKB_EXT_ERR(skb)->opt_stats)
981 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
982 				 skb->len, skb->data);
983 	}
984 }
985 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
986 
987 #ifdef CONFIG_WIRELESS
__sock_recv_wifi_status(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)988 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
989 	struct sk_buff *skb)
990 {
991 	int ack;
992 
993 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
994 		return;
995 	if (!skb->wifi_acked_valid)
996 		return;
997 
998 	ack = skb->wifi_acked;
999 
1000 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1001 }
1002 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1003 #endif
1004 
sock_recv_drops(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1005 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1006 				   struct sk_buff *skb)
1007 {
1008 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1009 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1010 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1011 }
1012 
sock_recv_mark(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1013 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1014 			   struct sk_buff *skb)
1015 {
1016 	if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1017 		/* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1018 		__u32 mark = skb->mark;
1019 
1020 		put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1021 	}
1022 }
1023 
__sock_recv_cmsgs(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)1024 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1025 		       struct sk_buff *skb)
1026 {
1027 	sock_recv_timestamp(msg, sk, skb);
1028 	sock_recv_drops(msg, sk, skb);
1029 	sock_recv_mark(msg, sk, skb);
1030 }
1031 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1032 
1033 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1034 					   size_t, int));
1035 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1036 					    size_t, int));
1037 
call_trace_sock_recv_length(struct sock * sk,int ret,int flags)1038 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1039 {
1040 	trace_sock_recv_length(sk, ret, flags);
1041 }
1042 
sock_recvmsg_nosec(struct socket * sock,struct msghdr * msg,int flags)1043 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1044 				     int flags)
1045 {
1046 	int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1047 				     inet6_recvmsg,
1048 				     inet_recvmsg, sock, msg,
1049 				     msg_data_left(msg), flags);
1050 	if (trace_sock_recv_length_enabled())
1051 		call_trace_sock_recv_length(sock->sk, ret, flags);
1052 	return ret;
1053 }
1054 
1055 /**
1056  *	sock_recvmsg - receive a message from @sock
1057  *	@sock: socket
1058  *	@msg: message to receive
1059  *	@flags: message flags
1060  *
1061  *	Receives @msg from @sock, passing through LSM. Returns the total number
1062  *	of bytes received, or an error.
1063  */
sock_recvmsg(struct socket * sock,struct msghdr * msg,int flags)1064 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1065 {
1066 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1067 
1068 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1069 }
1070 EXPORT_SYMBOL(sock_recvmsg);
1071 
1072 /**
1073  *	kernel_recvmsg - Receive a message from a socket (kernel space)
1074  *	@sock: The socket to receive the message from
1075  *	@msg: Received message
1076  *	@vec: Input s/g array for message data
1077  *	@num: Size of input s/g array
1078  *	@size: Number of bytes to read
1079  *	@flags: Message flags (MSG_DONTWAIT, etc...)
1080  *
1081  *	On return the msg structure contains the scatter/gather array passed in the
1082  *	vec argument. The array is modified so that it consists of the unfilled
1083  *	portion of the original array.
1084  *
1085  *	The returned value is the total number of bytes received, or an error.
1086  */
1087 
kernel_recvmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size,int flags)1088 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1089 		   struct kvec *vec, size_t num, size_t size, int flags)
1090 {
1091 	msg->msg_control_is_user = false;
1092 	iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1093 	return sock_recvmsg(sock, msg, flags);
1094 }
1095 EXPORT_SYMBOL(kernel_recvmsg);
1096 
sock_splice_read(struct file * file,loff_t * ppos,struct pipe_inode_info * pipe,size_t len,unsigned int flags)1097 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1098 				struct pipe_inode_info *pipe, size_t len,
1099 				unsigned int flags)
1100 {
1101 	struct socket *sock = file->private_data;
1102 	const struct proto_ops *ops;
1103 
1104 	ops = READ_ONCE(sock->ops);
1105 	if (unlikely(!ops->splice_read))
1106 		return copy_splice_read(file, ppos, pipe, len, flags);
1107 
1108 	return ops->splice_read(sock, ppos, pipe, len, flags);
1109 }
1110 
sock_splice_eof(struct file * file)1111 static void sock_splice_eof(struct file *file)
1112 {
1113 	struct socket *sock = file->private_data;
1114 	const struct proto_ops *ops;
1115 
1116 	ops = READ_ONCE(sock->ops);
1117 	if (ops->splice_eof)
1118 		ops->splice_eof(sock);
1119 }
1120 
sock_read_iter(struct kiocb * iocb,struct iov_iter * to)1121 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1122 {
1123 	struct file *file = iocb->ki_filp;
1124 	struct socket *sock = file->private_data;
1125 	struct msghdr msg = {.msg_iter = *to,
1126 			     .msg_iocb = iocb};
1127 	ssize_t res;
1128 
1129 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1130 		msg.msg_flags = MSG_DONTWAIT;
1131 
1132 	if (iocb->ki_pos != 0)
1133 		return -ESPIPE;
1134 
1135 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1136 		return 0;
1137 
1138 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1139 	*to = msg.msg_iter;
1140 	return res;
1141 }
1142 
sock_write_iter(struct kiocb * iocb,struct iov_iter * from)1143 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1144 {
1145 	struct file *file = iocb->ki_filp;
1146 	struct socket *sock = file->private_data;
1147 	struct msghdr msg = {.msg_iter = *from,
1148 			     .msg_iocb = iocb};
1149 	ssize_t res;
1150 
1151 	if (iocb->ki_pos != 0)
1152 		return -ESPIPE;
1153 
1154 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1155 		msg.msg_flags = MSG_DONTWAIT;
1156 
1157 	if (sock->type == SOCK_SEQPACKET)
1158 		msg.msg_flags |= MSG_EOR;
1159 
1160 	res = __sock_sendmsg(sock, &msg);
1161 	*from = msg.msg_iter;
1162 	return res;
1163 }
1164 
1165 /*
1166  * Atomic setting of ioctl hooks to avoid race
1167  * with module unload.
1168  */
1169 
1170 static DEFINE_MUTEX(br_ioctl_mutex);
1171 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1172 			    unsigned int cmd, struct ifreq *ifr,
1173 			    void __user *uarg);
1174 
brioctl_set(int (* hook)(struct net * net,struct net_bridge * br,unsigned int cmd,struct ifreq * ifr,void __user * uarg))1175 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1176 			     unsigned int cmd, struct ifreq *ifr,
1177 			     void __user *uarg))
1178 {
1179 	mutex_lock(&br_ioctl_mutex);
1180 	br_ioctl_hook = hook;
1181 	mutex_unlock(&br_ioctl_mutex);
1182 }
1183 EXPORT_SYMBOL(brioctl_set);
1184 
br_ioctl_call(struct net * net,struct net_bridge * br,unsigned int cmd,struct ifreq * ifr,void __user * uarg)1185 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1186 		  struct ifreq *ifr, void __user *uarg)
1187 {
1188 	int err = -ENOPKG;
1189 
1190 	if (!br_ioctl_hook)
1191 		request_module("bridge");
1192 
1193 	mutex_lock(&br_ioctl_mutex);
1194 	if (br_ioctl_hook)
1195 		err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1196 	mutex_unlock(&br_ioctl_mutex);
1197 
1198 	return err;
1199 }
1200 
1201 static DEFINE_MUTEX(vlan_ioctl_mutex);
1202 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1203 
vlan_ioctl_set(int (* hook)(struct net *,void __user *))1204 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1205 {
1206 	mutex_lock(&vlan_ioctl_mutex);
1207 	vlan_ioctl_hook = hook;
1208 	mutex_unlock(&vlan_ioctl_mutex);
1209 }
1210 EXPORT_SYMBOL(vlan_ioctl_set);
1211 
sock_do_ioctl(struct net * net,struct socket * sock,unsigned int cmd,unsigned long arg)1212 static long sock_do_ioctl(struct net *net, struct socket *sock,
1213 			  unsigned int cmd, unsigned long arg)
1214 {
1215 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1216 	struct ifreq ifr;
1217 	bool need_copyout;
1218 	int err;
1219 	void __user *argp = (void __user *)arg;
1220 	void __user *data;
1221 
1222 	err = ops->ioctl(sock, cmd, arg);
1223 
1224 	/*
1225 	 * If this ioctl is unknown try to hand it down
1226 	 * to the NIC driver.
1227 	 */
1228 	if (err != -ENOIOCTLCMD)
1229 		return err;
1230 
1231 	if (!is_socket_ioctl_cmd(cmd))
1232 		return -ENOTTY;
1233 
1234 	if (get_user_ifreq(&ifr, &data, argp))
1235 		return -EFAULT;
1236 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1237 	if (!err && need_copyout)
1238 		if (put_user_ifreq(&ifr, argp))
1239 			return -EFAULT;
1240 
1241 	return err;
1242 }
1243 
1244 /*
1245  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1246  *	what to do with it - that's up to the protocol still.
1247  */
1248 
sock_ioctl(struct file * file,unsigned cmd,unsigned long arg)1249 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1250 {
1251 	const struct proto_ops  *ops;
1252 	struct socket *sock;
1253 	struct sock *sk;
1254 	void __user *argp = (void __user *)arg;
1255 	int pid, err;
1256 	struct net *net;
1257 
1258 	sock = file->private_data;
1259 	ops = READ_ONCE(sock->ops);
1260 	sk = sock->sk;
1261 	net = sock_net(sk);
1262 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1263 		struct ifreq ifr;
1264 		void __user *data;
1265 		bool need_copyout;
1266 		if (get_user_ifreq(&ifr, &data, argp))
1267 			return -EFAULT;
1268 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1269 		if (!err && need_copyout)
1270 			if (put_user_ifreq(&ifr, argp))
1271 				return -EFAULT;
1272 	} else
1273 #ifdef CONFIG_WEXT_CORE
1274 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1275 		err = wext_handle_ioctl(net, cmd, argp);
1276 	} else
1277 #endif
1278 		switch (cmd) {
1279 		case FIOSETOWN:
1280 		case SIOCSPGRP:
1281 			err = -EFAULT;
1282 			if (get_user(pid, (int __user *)argp))
1283 				break;
1284 			err = f_setown(sock->file, pid, 1);
1285 			break;
1286 		case FIOGETOWN:
1287 		case SIOCGPGRP:
1288 			err = put_user(f_getown(sock->file),
1289 				       (int __user *)argp);
1290 			break;
1291 		case SIOCGIFBR:
1292 		case SIOCSIFBR:
1293 		case SIOCBRADDBR:
1294 		case SIOCBRDELBR:
1295 			err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1296 			break;
1297 		case SIOCGIFVLAN:
1298 		case SIOCSIFVLAN:
1299 			err = -ENOPKG;
1300 			if (!vlan_ioctl_hook)
1301 				request_module("8021q");
1302 
1303 			mutex_lock(&vlan_ioctl_mutex);
1304 			if (vlan_ioctl_hook)
1305 				err = vlan_ioctl_hook(net, argp);
1306 			mutex_unlock(&vlan_ioctl_mutex);
1307 			break;
1308 		case SIOCGSKNS:
1309 			err = -EPERM;
1310 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1311 				break;
1312 
1313 			err = open_related_ns(&net->ns, get_net_ns);
1314 			break;
1315 		case SIOCGSTAMP_OLD:
1316 		case SIOCGSTAMPNS_OLD:
1317 			if (!ops->gettstamp) {
1318 				err = -ENOIOCTLCMD;
1319 				break;
1320 			}
1321 			err = ops->gettstamp(sock, argp,
1322 					     cmd == SIOCGSTAMP_OLD,
1323 					     !IS_ENABLED(CONFIG_64BIT));
1324 			break;
1325 		case SIOCGSTAMP_NEW:
1326 		case SIOCGSTAMPNS_NEW:
1327 			if (!ops->gettstamp) {
1328 				err = -ENOIOCTLCMD;
1329 				break;
1330 			}
1331 			err = ops->gettstamp(sock, argp,
1332 					     cmd == SIOCGSTAMP_NEW,
1333 					     false);
1334 			break;
1335 
1336 		case SIOCGIFCONF:
1337 			err = dev_ifconf(net, argp);
1338 			break;
1339 
1340 		default:
1341 			err = sock_do_ioctl(net, sock, cmd, arg);
1342 			break;
1343 		}
1344 	return err;
1345 }
1346 
1347 /**
1348  *	sock_create_lite - creates a socket
1349  *	@family: protocol family (AF_INET, ...)
1350  *	@type: communication type (SOCK_STREAM, ...)
1351  *	@protocol: protocol (0, ...)
1352  *	@res: new socket
1353  *
1354  *	Creates a new socket and assigns it to @res, passing through LSM.
1355  *	The new socket initialization is not complete, see kernel_accept().
1356  *	Returns 0 or an error. On failure @res is set to %NULL.
1357  *	This function internally uses GFP_KERNEL.
1358  */
1359 
sock_create_lite(int family,int type,int protocol,struct socket ** res)1360 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1361 {
1362 	int err;
1363 	struct socket *sock = NULL;
1364 
1365 	err = security_socket_create(family, type, protocol, 1);
1366 	if (err)
1367 		goto out;
1368 
1369 	sock = sock_alloc();
1370 	if (!sock) {
1371 		err = -ENOMEM;
1372 		goto out;
1373 	}
1374 
1375 	sock->type = type;
1376 	err = security_socket_post_create(sock, family, type, protocol, 1);
1377 	if (err)
1378 		goto out_release;
1379 
1380 out:
1381 	*res = sock;
1382 	return err;
1383 out_release:
1384 	sock_release(sock);
1385 	sock = NULL;
1386 	goto out;
1387 }
1388 EXPORT_SYMBOL(sock_create_lite);
1389 
1390 /* No kernel lock held - perfect */
sock_poll(struct file * file,poll_table * wait)1391 static __poll_t sock_poll(struct file *file, poll_table *wait)
1392 {
1393 	struct socket *sock = file->private_data;
1394 	const struct proto_ops *ops = READ_ONCE(sock->ops);
1395 	__poll_t events = poll_requested_events(wait), flag = 0;
1396 
1397 	if (!ops->poll)
1398 		return 0;
1399 
1400 	if (sk_can_busy_loop(sock->sk)) {
1401 		/* poll once if requested by the syscall */
1402 		if (events & POLL_BUSY_LOOP)
1403 			sk_busy_loop(sock->sk, 1);
1404 
1405 		/* if this socket can poll_ll, tell the system call */
1406 		flag = POLL_BUSY_LOOP;
1407 	}
1408 
1409 	return ops->poll(file, sock, wait) | flag;
1410 }
1411 
sock_mmap(struct file * file,struct vm_area_struct * vma)1412 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1413 {
1414 	struct socket *sock = file->private_data;
1415 
1416 	return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1417 }
1418 
sock_close(struct inode * inode,struct file * filp)1419 static int sock_close(struct inode *inode, struct file *filp)
1420 {
1421 	__sock_release(SOCKET_I(inode), inode);
1422 	return 0;
1423 }
1424 
1425 /*
1426  *	Update the socket async list
1427  *
1428  *	Fasync_list locking strategy.
1429  *
1430  *	1. fasync_list is modified only under process context socket lock
1431  *	   i.e. under semaphore.
1432  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1433  *	   or under socket lock
1434  */
1435 
sock_fasync(int fd,struct file * filp,int on)1436 static int sock_fasync(int fd, struct file *filp, int on)
1437 {
1438 	struct socket *sock = filp->private_data;
1439 	struct sock *sk = sock->sk;
1440 	struct socket_wq *wq = &sock->wq;
1441 
1442 	if (sk == NULL)
1443 		return -EINVAL;
1444 
1445 	lock_sock(sk);
1446 	fasync_helper(fd, filp, on, &wq->fasync_list);
1447 
1448 	if (!wq->fasync_list)
1449 		sock_reset_flag(sk, SOCK_FASYNC);
1450 	else
1451 		sock_set_flag(sk, SOCK_FASYNC);
1452 
1453 	release_sock(sk);
1454 	return 0;
1455 }
1456 
1457 /* This function may be called only under rcu_lock */
1458 
sock_wake_async(struct socket_wq * wq,int how,int band)1459 int sock_wake_async(struct socket_wq *wq, int how, int band)
1460 {
1461 	if (!wq || !wq->fasync_list)
1462 		return -1;
1463 
1464 	switch (how) {
1465 	case SOCK_WAKE_WAITD:
1466 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1467 			break;
1468 		goto call_kill;
1469 	case SOCK_WAKE_SPACE:
1470 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1471 			break;
1472 		fallthrough;
1473 	case SOCK_WAKE_IO:
1474 call_kill:
1475 		kill_fasync(&wq->fasync_list, SIGIO, band);
1476 		break;
1477 	case SOCK_WAKE_URG:
1478 		kill_fasync(&wq->fasync_list, SIGURG, band);
1479 	}
1480 
1481 	return 0;
1482 }
1483 EXPORT_SYMBOL(sock_wake_async);
1484 
1485 /**
1486  *	__sock_create - creates a socket
1487  *	@net: net namespace
1488  *	@family: protocol family (AF_INET, ...)
1489  *	@type: communication type (SOCK_STREAM, ...)
1490  *	@protocol: protocol (0, ...)
1491  *	@res: new socket
1492  *	@kern: boolean for kernel space sockets
1493  *
1494  *	Creates a new socket and assigns it to @res, passing through LSM.
1495  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1496  *	be set to true if the socket resides in kernel space.
1497  *	This function internally uses GFP_KERNEL.
1498  */
1499 
__sock_create(struct net * net,int family,int type,int protocol,struct socket ** res,int kern)1500 int __sock_create(struct net *net, int family, int type, int protocol,
1501 			 struct socket **res, int kern)
1502 {
1503 	int err;
1504 	struct socket *sock;
1505 	const struct net_proto_family *pf;
1506 
1507 	/*
1508 	 *      Check protocol is in range
1509 	 */
1510 	if (family < 0 || family >= NPROTO)
1511 		return -EAFNOSUPPORT;
1512 	if (type < 0 || type >= SOCK_MAX)
1513 		return -EINVAL;
1514 
1515 	/* Compatibility.
1516 
1517 	   This uglymoron is moved from INET layer to here to avoid
1518 	   deadlock in module load.
1519 	 */
1520 	if (family == PF_INET && type == SOCK_PACKET) {
1521 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1522 			     current->comm);
1523 		family = PF_PACKET;
1524 	}
1525 
1526 	err = security_socket_create(family, type, protocol, kern);
1527 	if (err)
1528 		return err;
1529 
1530 	/*
1531 	 *	Allocate the socket and allow the family to set things up. if
1532 	 *	the protocol is 0, the family is instructed to select an appropriate
1533 	 *	default.
1534 	 */
1535 	sock = sock_alloc();
1536 	if (!sock) {
1537 		net_warn_ratelimited("socket: no more sockets\n");
1538 		return -ENFILE;	/* Not exactly a match, but its the
1539 				   closest posix thing */
1540 	}
1541 
1542 	sock->type = type;
1543 
1544 #ifdef CONFIG_MODULES
1545 	/* Attempt to load a protocol module if the find failed.
1546 	 *
1547 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1548 	 * requested real, full-featured networking support upon configuration.
1549 	 * Otherwise module support will break!
1550 	 */
1551 	if (rcu_access_pointer(net_families[family]) == NULL)
1552 		request_module("net-pf-%d", family);
1553 #endif
1554 
1555 	rcu_read_lock();
1556 	pf = rcu_dereference(net_families[family]);
1557 	err = -EAFNOSUPPORT;
1558 	if (!pf)
1559 		goto out_release;
1560 
1561 	/*
1562 	 * We will call the ->create function, that possibly is in a loadable
1563 	 * module, so we have to bump that loadable module refcnt first.
1564 	 */
1565 	if (!try_module_get(pf->owner))
1566 		goto out_release;
1567 
1568 	/* Now protected by module ref count */
1569 	rcu_read_unlock();
1570 
1571 	err = pf->create(net, sock, protocol, kern);
1572 	if (err < 0)
1573 		goto out_module_put;
1574 
1575 	/*
1576 	 * Now to bump the refcnt of the [loadable] module that owns this
1577 	 * socket at sock_release time we decrement its refcnt.
1578 	 */
1579 	if (!try_module_get(sock->ops->owner))
1580 		goto out_module_busy;
1581 
1582 	/*
1583 	 * Now that we're done with the ->create function, the [loadable]
1584 	 * module can have its refcnt decremented
1585 	 */
1586 	module_put(pf->owner);
1587 	err = security_socket_post_create(sock, family, type, protocol, kern);
1588 	if (err)
1589 		goto out_sock_release;
1590 	*res = sock;
1591 
1592 	return 0;
1593 
1594 out_module_busy:
1595 	err = -EAFNOSUPPORT;
1596 out_module_put:
1597 	sock->ops = NULL;
1598 	module_put(pf->owner);
1599 out_sock_release:
1600 	sock_release(sock);
1601 	return err;
1602 
1603 out_release:
1604 	rcu_read_unlock();
1605 	goto out_sock_release;
1606 }
1607 EXPORT_SYMBOL(__sock_create);
1608 
1609 /**
1610  *	sock_create - creates a socket
1611  *	@family: protocol family (AF_INET, ...)
1612  *	@type: communication type (SOCK_STREAM, ...)
1613  *	@protocol: protocol (0, ...)
1614  *	@res: new socket
1615  *
1616  *	A wrapper around __sock_create().
1617  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1618  */
1619 
sock_create(int family,int type,int protocol,struct socket ** res)1620 int sock_create(int family, int type, int protocol, struct socket **res)
1621 {
1622 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1623 }
1624 EXPORT_SYMBOL(sock_create);
1625 
1626 /**
1627  *	sock_create_kern - creates a socket (kernel space)
1628  *	@net: net namespace
1629  *	@family: protocol family (AF_INET, ...)
1630  *	@type: communication type (SOCK_STREAM, ...)
1631  *	@protocol: protocol (0, ...)
1632  *	@res: new socket
1633  *
1634  *	A wrapper around __sock_create().
1635  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1636  */
1637 
sock_create_kern(struct net * net,int family,int type,int protocol,struct socket ** res)1638 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1639 {
1640 	return __sock_create(net, family, type, protocol, res, 1);
1641 }
1642 EXPORT_SYMBOL(sock_create_kern);
1643 
__sys_socket_create(int family,int type,int protocol)1644 static struct socket *__sys_socket_create(int family, int type, int protocol)
1645 {
1646 	struct socket *sock;
1647 	int retval;
1648 
1649 	/* Check the SOCK_* constants for consistency.  */
1650 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1651 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1652 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1653 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1654 
1655 	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1656 		return ERR_PTR(-EINVAL);
1657 	type &= SOCK_TYPE_MASK;
1658 
1659 	retval = sock_create(family, type, protocol, &sock);
1660 	if (retval < 0)
1661 		return ERR_PTR(retval);
1662 
1663 	return sock;
1664 }
1665 
__sys_socket_file(int family,int type,int protocol)1666 struct file *__sys_socket_file(int family, int type, int protocol)
1667 {
1668 	struct socket *sock;
1669 	int flags;
1670 
1671 	sock = __sys_socket_create(family, type, protocol);
1672 	if (IS_ERR(sock))
1673 		return ERR_CAST(sock);
1674 
1675 	flags = type & ~SOCK_TYPE_MASK;
1676 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1677 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1678 
1679 	return sock_alloc_file(sock, flags, NULL);
1680 }
1681 
1682 /*	A hook for bpf progs to attach to and update socket protocol.
1683  *
1684  *	A static noinline declaration here could cause the compiler to
1685  *	optimize away the function. A global noinline declaration will
1686  *	keep the definition, but may optimize away the callsite.
1687  *	Therefore, __weak is needed to ensure that the call is still
1688  *	emitted, by telling the compiler that we don't know what the
1689  *	function might eventually be.
1690  */
1691 
1692 __bpf_hook_start();
1693 
update_socket_protocol(int family,int type,int protocol)1694 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1695 {
1696 	return protocol;
1697 }
1698 
1699 __bpf_hook_end();
1700 
__sys_socket(int family,int type,int protocol)1701 int __sys_socket(int family, int type, int protocol)
1702 {
1703 	struct socket *sock;
1704 	int flags;
1705 
1706 	sock = __sys_socket_create(family, type,
1707 				   update_socket_protocol(family, type, protocol));
1708 	if (IS_ERR(sock))
1709 		return PTR_ERR(sock);
1710 
1711 	flags = type & ~SOCK_TYPE_MASK;
1712 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1713 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1714 
1715 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1716 }
1717 
SYSCALL_DEFINE3(socket,int,family,int,type,int,protocol)1718 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1719 {
1720 	return __sys_socket(family, type, protocol);
1721 }
1722 
1723 /*
1724  *	Create a pair of connected sockets.
1725  */
1726 
__sys_socketpair(int family,int type,int protocol,int __user * usockvec)1727 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1728 {
1729 	struct socket *sock1, *sock2;
1730 	int fd1, fd2, err;
1731 	struct file *newfile1, *newfile2;
1732 	int flags;
1733 
1734 	flags = type & ~SOCK_TYPE_MASK;
1735 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1736 		return -EINVAL;
1737 	type &= SOCK_TYPE_MASK;
1738 
1739 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1740 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1741 
1742 	/*
1743 	 * reserve descriptors and make sure we won't fail
1744 	 * to return them to userland.
1745 	 */
1746 	fd1 = get_unused_fd_flags(flags);
1747 	if (unlikely(fd1 < 0))
1748 		return fd1;
1749 
1750 	fd2 = get_unused_fd_flags(flags);
1751 	if (unlikely(fd2 < 0)) {
1752 		put_unused_fd(fd1);
1753 		return fd2;
1754 	}
1755 
1756 	err = put_user(fd1, &usockvec[0]);
1757 	if (err)
1758 		goto out;
1759 
1760 	err = put_user(fd2, &usockvec[1]);
1761 	if (err)
1762 		goto out;
1763 
1764 	/*
1765 	 * Obtain the first socket and check if the underlying protocol
1766 	 * supports the socketpair call.
1767 	 */
1768 
1769 	err = sock_create(family, type, protocol, &sock1);
1770 	if (unlikely(err < 0))
1771 		goto out;
1772 
1773 	err = sock_create(family, type, protocol, &sock2);
1774 	if (unlikely(err < 0)) {
1775 		sock_release(sock1);
1776 		goto out;
1777 	}
1778 
1779 	err = security_socket_socketpair(sock1, sock2);
1780 	if (unlikely(err)) {
1781 		sock_release(sock2);
1782 		sock_release(sock1);
1783 		goto out;
1784 	}
1785 
1786 	err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1787 	if (unlikely(err < 0)) {
1788 		sock_release(sock2);
1789 		sock_release(sock1);
1790 		goto out;
1791 	}
1792 
1793 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1794 	if (IS_ERR(newfile1)) {
1795 		err = PTR_ERR(newfile1);
1796 		sock_release(sock2);
1797 		goto out;
1798 	}
1799 
1800 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1801 	if (IS_ERR(newfile2)) {
1802 		err = PTR_ERR(newfile2);
1803 		fput(newfile1);
1804 		goto out;
1805 	}
1806 
1807 	audit_fd_pair(fd1, fd2);
1808 
1809 	fd_install(fd1, newfile1);
1810 	fd_install(fd2, newfile2);
1811 	return 0;
1812 
1813 out:
1814 	put_unused_fd(fd2);
1815 	put_unused_fd(fd1);
1816 	return err;
1817 }
1818 
SYSCALL_DEFINE4(socketpair,int,family,int,type,int,protocol,int __user *,usockvec)1819 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1820 		int __user *, usockvec)
1821 {
1822 	return __sys_socketpair(family, type, protocol, usockvec);
1823 }
1824 
1825 /*
1826  *	Bind a name to a socket. Nothing much to do here since it's
1827  *	the protocol's responsibility to handle the local address.
1828  *
1829  *	We move the socket address to kernel space before we call
1830  *	the protocol layer (having also checked the address is ok).
1831  */
1832 
__sys_bind(int fd,struct sockaddr __user * umyaddr,int addrlen)1833 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1834 {
1835 	struct socket *sock;
1836 	struct sockaddr_storage address;
1837 	int err, fput_needed;
1838 
1839 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1840 	if (sock) {
1841 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1842 		if (!err) {
1843 			err = security_socket_bind(sock,
1844 						   (struct sockaddr *)&address,
1845 						   addrlen);
1846 			if (!err)
1847 				err = READ_ONCE(sock->ops)->bind(sock,
1848 						      (struct sockaddr *)
1849 						      &address, addrlen);
1850 		}
1851 		fput_light(sock->file, fput_needed);
1852 	}
1853 	return err;
1854 }
1855 
SYSCALL_DEFINE3(bind,int,fd,struct sockaddr __user *,umyaddr,int,addrlen)1856 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1857 {
1858 	return __sys_bind(fd, umyaddr, addrlen);
1859 }
1860 
1861 /*
1862  *	Perform a listen. Basically, we allow the protocol to do anything
1863  *	necessary for a listen, and if that works, we mark the socket as
1864  *	ready for listening.
1865  */
1866 
__sys_listen(int fd,int backlog)1867 int __sys_listen(int fd, int backlog)
1868 {
1869 	struct socket *sock;
1870 	int err, fput_needed;
1871 	int somaxconn;
1872 
1873 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1874 	if (sock) {
1875 		somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1876 		if ((unsigned int)backlog > somaxconn)
1877 			backlog = somaxconn;
1878 
1879 		err = security_socket_listen(sock, backlog);
1880 		if (!err)
1881 			err = READ_ONCE(sock->ops)->listen(sock, backlog);
1882 
1883 		fput_light(sock->file, fput_needed);
1884 	}
1885 	return err;
1886 }
1887 
SYSCALL_DEFINE2(listen,int,fd,int,backlog)1888 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1889 {
1890 	return __sys_listen(fd, backlog);
1891 }
1892 
do_accept(struct file * file,struct proto_accept_arg * arg,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1893 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1894 		       struct sockaddr __user *upeer_sockaddr,
1895 		       int __user *upeer_addrlen, int flags)
1896 {
1897 	struct socket *sock, *newsock;
1898 	struct file *newfile;
1899 	int err, len;
1900 	struct sockaddr_storage address;
1901 	const struct proto_ops *ops;
1902 
1903 	sock = sock_from_file(file);
1904 	if (!sock)
1905 		return ERR_PTR(-ENOTSOCK);
1906 
1907 	newsock = sock_alloc();
1908 	if (!newsock)
1909 		return ERR_PTR(-ENFILE);
1910 	ops = READ_ONCE(sock->ops);
1911 
1912 	newsock->type = sock->type;
1913 	newsock->ops = ops;
1914 
1915 	/*
1916 	 * We don't need try_module_get here, as the listening socket (sock)
1917 	 * has the protocol module (sock->ops->owner) held.
1918 	 */
1919 	__module_get(ops->owner);
1920 
1921 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1922 	if (IS_ERR(newfile))
1923 		return newfile;
1924 
1925 	err = security_socket_accept(sock, newsock);
1926 	if (err)
1927 		goto out_fd;
1928 
1929 	arg->flags |= sock->file->f_flags;
1930 	err = ops->accept(sock, newsock, arg);
1931 	if (err < 0)
1932 		goto out_fd;
1933 
1934 	if (upeer_sockaddr) {
1935 		len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1936 		if (len < 0) {
1937 			err = -ECONNABORTED;
1938 			goto out_fd;
1939 		}
1940 		err = move_addr_to_user(&address,
1941 					len, upeer_sockaddr, upeer_addrlen);
1942 		if (err < 0)
1943 			goto out_fd;
1944 	}
1945 
1946 	/* File flags are not inherited via accept() unlike another OSes. */
1947 	return newfile;
1948 out_fd:
1949 	fput(newfile);
1950 	return ERR_PTR(err);
1951 }
1952 
__sys_accept4_file(struct file * file,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1953 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1954 			      int __user *upeer_addrlen, int flags)
1955 {
1956 	struct proto_accept_arg arg = { };
1957 	struct file *newfile;
1958 	int newfd;
1959 
1960 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1961 		return -EINVAL;
1962 
1963 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1964 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1965 
1966 	newfd = get_unused_fd_flags(flags);
1967 	if (unlikely(newfd < 0))
1968 		return newfd;
1969 
1970 	newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1971 			    flags);
1972 	if (IS_ERR(newfile)) {
1973 		put_unused_fd(newfd);
1974 		return PTR_ERR(newfile);
1975 	}
1976 	fd_install(newfd, newfile);
1977 	return newfd;
1978 }
1979 
1980 /*
1981  *	For accept, we attempt to create a new socket, set up the link
1982  *	with the client, wake up the client, then return the new
1983  *	connected fd. We collect the address of the connector in kernel
1984  *	space and move it to user at the very end. This is unclean because
1985  *	we open the socket then return an error.
1986  *
1987  *	1003.1g adds the ability to recvmsg() to query connection pending
1988  *	status to recvmsg. We need to add that support in a way thats
1989  *	clean when we restructure accept also.
1990  */
1991 
__sys_accept4(int fd,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1992 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1993 		  int __user *upeer_addrlen, int flags)
1994 {
1995 	int ret = -EBADF;
1996 	struct fd f;
1997 
1998 	f = fdget(fd);
1999 	if (f.file) {
2000 		ret = __sys_accept4_file(f.file, upeer_sockaddr,
2001 					 upeer_addrlen, flags);
2002 		fdput(f);
2003 	}
2004 
2005 	return ret;
2006 }
2007 
SYSCALL_DEFINE4(accept4,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen,int,flags)2008 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2009 		int __user *, upeer_addrlen, int, flags)
2010 {
2011 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2012 }
2013 
SYSCALL_DEFINE3(accept,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen)2014 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2015 		int __user *, upeer_addrlen)
2016 {
2017 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2018 }
2019 
2020 /*
2021  *	Attempt to connect to a socket with the server address.  The address
2022  *	is in user space so we verify it is OK and move it to kernel space.
2023  *
2024  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2025  *	break bindings
2026  *
2027  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2028  *	other SEQPACKET protocols that take time to connect() as it doesn't
2029  *	include the -EINPROGRESS status for such sockets.
2030  */
2031 
__sys_connect_file(struct file * file,struct sockaddr_storage * address,int addrlen,int file_flags)2032 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2033 		       int addrlen, int file_flags)
2034 {
2035 	struct socket *sock;
2036 	int err;
2037 
2038 	sock = sock_from_file(file);
2039 	if (!sock) {
2040 		err = -ENOTSOCK;
2041 		goto out;
2042 	}
2043 
2044 	err =
2045 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2046 	if (err)
2047 		goto out;
2048 
2049 	err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2050 				addrlen, sock->file->f_flags | file_flags);
2051 out:
2052 	return err;
2053 }
2054 
__sys_connect(int fd,struct sockaddr __user * uservaddr,int addrlen)2055 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2056 {
2057 	int ret = -EBADF;
2058 	struct fd f;
2059 
2060 	f = fdget(fd);
2061 	if (f.file) {
2062 		struct sockaddr_storage address;
2063 
2064 		ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2065 		if (!ret)
2066 			ret = __sys_connect_file(f.file, &address, addrlen, 0);
2067 		fdput(f);
2068 	}
2069 
2070 	return ret;
2071 }
2072 
SYSCALL_DEFINE3(connect,int,fd,struct sockaddr __user *,uservaddr,int,addrlen)2073 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2074 		int, addrlen)
2075 {
2076 	return __sys_connect(fd, uservaddr, addrlen);
2077 }
2078 
2079 /*
2080  *	Get the local address ('name') of a socket object. Move the obtained
2081  *	name to user space.
2082  */
2083 
__sys_getsockname(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2084 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2085 		      int __user *usockaddr_len)
2086 {
2087 	struct socket *sock;
2088 	struct sockaddr_storage address;
2089 	int err, fput_needed;
2090 
2091 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2092 	if (!sock)
2093 		goto out;
2094 
2095 	err = security_socket_getsockname(sock);
2096 	if (err)
2097 		goto out_put;
2098 
2099 	err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2100 	if (err < 0)
2101 		goto out_put;
2102 	/* "err" is actually length in this case */
2103 	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2104 
2105 out_put:
2106 	fput_light(sock->file, fput_needed);
2107 out:
2108 	return err;
2109 }
2110 
SYSCALL_DEFINE3(getsockname,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2111 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2112 		int __user *, usockaddr_len)
2113 {
2114 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2115 }
2116 
2117 /*
2118  *	Get the remote address ('name') of a socket object. Move the obtained
2119  *	name to user space.
2120  */
2121 
__sys_getpeername(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2122 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2123 		      int __user *usockaddr_len)
2124 {
2125 	struct socket *sock;
2126 	struct sockaddr_storage address;
2127 	int err, fput_needed;
2128 
2129 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2130 	if (sock != NULL) {
2131 		const struct proto_ops *ops = READ_ONCE(sock->ops);
2132 
2133 		err = security_socket_getpeername(sock);
2134 		if (err) {
2135 			fput_light(sock->file, fput_needed);
2136 			return err;
2137 		}
2138 
2139 		err = ops->getname(sock, (struct sockaddr *)&address, 1);
2140 		if (err >= 0)
2141 			/* "err" is actually length in this case */
2142 			err = move_addr_to_user(&address, err, usockaddr,
2143 						usockaddr_len);
2144 		fput_light(sock->file, fput_needed);
2145 	}
2146 	return err;
2147 }
2148 
SYSCALL_DEFINE3(getpeername,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2149 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2150 		int __user *, usockaddr_len)
2151 {
2152 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2153 }
2154 
2155 /*
2156  *	Send a datagram to a given address. We move the address into kernel
2157  *	space and check the user space data area is readable before invoking
2158  *	the protocol.
2159  */
__sys_sendto(int fd,void __user * buff,size_t len,unsigned int flags,struct sockaddr __user * addr,int addr_len)2160 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2161 		 struct sockaddr __user *addr,  int addr_len)
2162 {
2163 	struct socket *sock;
2164 	struct sockaddr_storage address;
2165 	int err;
2166 	struct msghdr msg;
2167 	int fput_needed;
2168 
2169 	err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2170 	if (unlikely(err))
2171 		return err;
2172 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2173 	if (!sock)
2174 		goto out;
2175 
2176 	msg.msg_name = NULL;
2177 	msg.msg_control = NULL;
2178 	msg.msg_controllen = 0;
2179 	msg.msg_namelen = 0;
2180 	msg.msg_ubuf = NULL;
2181 	if (addr) {
2182 		err = move_addr_to_kernel(addr, addr_len, &address);
2183 		if (err < 0)
2184 			goto out_put;
2185 		msg.msg_name = (struct sockaddr *)&address;
2186 		msg.msg_namelen = addr_len;
2187 	}
2188 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2189 	if (sock->file->f_flags & O_NONBLOCK)
2190 		flags |= MSG_DONTWAIT;
2191 	msg.msg_flags = flags;
2192 	err = __sock_sendmsg(sock, &msg);
2193 
2194 out_put:
2195 	fput_light(sock->file, fput_needed);
2196 out:
2197 	return err;
2198 }
2199 
SYSCALL_DEFINE6(sendto,int,fd,void __user *,buff,size_t,len,unsigned int,flags,struct sockaddr __user *,addr,int,addr_len)2200 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2201 		unsigned int, flags, struct sockaddr __user *, addr,
2202 		int, addr_len)
2203 {
2204 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2205 }
2206 
2207 /*
2208  *	Send a datagram down a socket.
2209  */
2210 
SYSCALL_DEFINE4(send,int,fd,void __user *,buff,size_t,len,unsigned int,flags)2211 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2212 		unsigned int, flags)
2213 {
2214 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2215 }
2216 
2217 /*
2218  *	Receive a frame from the socket and optionally record the address of the
2219  *	sender. We verify the buffers are writable and if needed move the
2220  *	sender address from kernel to user space.
2221  */
__sys_recvfrom(int fd,void __user * ubuf,size_t size,unsigned int flags,struct sockaddr __user * addr,int __user * addr_len)2222 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2223 		   struct sockaddr __user *addr, int __user *addr_len)
2224 {
2225 	struct sockaddr_storage address;
2226 	struct msghdr msg = {
2227 		/* Save some cycles and don't copy the address if not needed */
2228 		.msg_name = addr ? (struct sockaddr *)&address : NULL,
2229 	};
2230 	struct socket *sock;
2231 	int err, err2;
2232 	int fput_needed;
2233 
2234 	err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2235 	if (unlikely(err))
2236 		return err;
2237 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2238 	if (!sock)
2239 		goto out;
2240 
2241 	if (sock->file->f_flags & O_NONBLOCK)
2242 		flags |= MSG_DONTWAIT;
2243 	err = sock_recvmsg(sock, &msg, flags);
2244 
2245 	if (err >= 0 && addr != NULL) {
2246 		err2 = move_addr_to_user(&address,
2247 					 msg.msg_namelen, addr, addr_len);
2248 		if (err2 < 0)
2249 			err = err2;
2250 	}
2251 
2252 	fput_light(sock->file, fput_needed);
2253 out:
2254 	return err;
2255 }
2256 
SYSCALL_DEFINE6(recvfrom,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags,struct sockaddr __user *,addr,int __user *,addr_len)2257 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2258 		unsigned int, flags, struct sockaddr __user *, addr,
2259 		int __user *, addr_len)
2260 {
2261 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2262 }
2263 
2264 /*
2265  *	Receive a datagram from a socket.
2266  */
2267 
SYSCALL_DEFINE4(recv,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags)2268 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2269 		unsigned int, flags)
2270 {
2271 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2272 }
2273 
sock_use_custom_sol_socket(const struct socket * sock)2274 static bool sock_use_custom_sol_socket(const struct socket *sock)
2275 {
2276 	return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2277 }
2278 
do_sock_setsockopt(struct socket * sock,bool compat,int level,int optname,sockptr_t optval,int optlen)2279 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2280 		       int optname, sockptr_t optval, int optlen)
2281 {
2282 	const struct proto_ops *ops;
2283 	char *kernel_optval = NULL;
2284 	int err;
2285 
2286 	if (optlen < 0)
2287 		return -EINVAL;
2288 
2289 	err = security_socket_setsockopt(sock, level, optname);
2290 	if (err)
2291 		goto out_put;
2292 
2293 	if (!compat)
2294 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2295 						     optval, &optlen,
2296 						     &kernel_optval);
2297 	if (err < 0)
2298 		goto out_put;
2299 	if (err > 0) {
2300 		err = 0;
2301 		goto out_put;
2302 	}
2303 
2304 	if (kernel_optval)
2305 		optval = KERNEL_SOCKPTR(kernel_optval);
2306 	ops = READ_ONCE(sock->ops);
2307 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2308 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2309 	else if (unlikely(!ops->setsockopt))
2310 		err = -EOPNOTSUPP;
2311 	else
2312 		err = ops->setsockopt(sock, level, optname, optval,
2313 					    optlen);
2314 	kfree(kernel_optval);
2315 out_put:
2316 	return err;
2317 }
2318 EXPORT_SYMBOL(do_sock_setsockopt);
2319 
2320 /* Set a socket option. Because we don't know the option lengths we have
2321  * to pass the user mode parameter for the protocols to sort out.
2322  */
__sys_setsockopt(int fd,int level,int optname,char __user * user_optval,int optlen)2323 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2324 		     int optlen)
2325 {
2326 	sockptr_t optval = USER_SOCKPTR(user_optval);
2327 	bool compat = in_compat_syscall();
2328 	int err, fput_needed;
2329 	struct socket *sock;
2330 
2331 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2332 	if (!sock)
2333 		return err;
2334 
2335 	err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2336 
2337 	fput_light(sock->file, fput_needed);
2338 	return err;
2339 }
2340 
SYSCALL_DEFINE5(setsockopt,int,fd,int,level,int,optname,char __user *,optval,int,optlen)2341 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2342 		char __user *, optval, int, optlen)
2343 {
2344 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2345 }
2346 
2347 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2348 							 int optname));
2349 
do_sock_getsockopt(struct socket * sock,bool compat,int level,int optname,sockptr_t optval,sockptr_t optlen)2350 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2351 		       int optname, sockptr_t optval, sockptr_t optlen)
2352 {
2353 	int max_optlen __maybe_unused;
2354 	const struct proto_ops *ops;
2355 	int err;
2356 
2357 	err = security_socket_getsockopt(sock, level, optname);
2358 	if (err)
2359 		return err;
2360 
2361 	if (!compat)
2362 		max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2363 
2364 	ops = READ_ONCE(sock->ops);
2365 	if (level == SOL_SOCKET) {
2366 		err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2367 	} else if (unlikely(!ops->getsockopt)) {
2368 		err = -EOPNOTSUPP;
2369 	} else {
2370 		if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2371 			      "Invalid argument type"))
2372 			return -EOPNOTSUPP;
2373 
2374 		err = ops->getsockopt(sock, level, optname, optval.user,
2375 				      optlen.user);
2376 	}
2377 
2378 	if (!compat)
2379 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2380 						     optval, optlen, max_optlen,
2381 						     err);
2382 
2383 	return err;
2384 }
2385 EXPORT_SYMBOL(do_sock_getsockopt);
2386 
2387 /*
2388  *	Get a socket option. Because we don't know the option lengths we have
2389  *	to pass a user mode parameter for the protocols to sort out.
2390  */
__sys_getsockopt(int fd,int level,int optname,char __user * optval,int __user * optlen)2391 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2392 		int __user *optlen)
2393 {
2394 	int err, fput_needed;
2395 	struct socket *sock;
2396 	bool compat;
2397 
2398 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2399 	if (!sock)
2400 		return err;
2401 
2402 	compat = in_compat_syscall();
2403 	err = do_sock_getsockopt(sock, compat, level, optname,
2404 				 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2405 
2406 	fput_light(sock->file, fput_needed);
2407 	return err;
2408 }
2409 
SYSCALL_DEFINE5(getsockopt,int,fd,int,level,int,optname,char __user *,optval,int __user *,optlen)2410 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2411 		char __user *, optval, int __user *, optlen)
2412 {
2413 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2414 }
2415 
2416 /*
2417  *	Shutdown a socket.
2418  */
2419 
__sys_shutdown_sock(struct socket * sock,int how)2420 int __sys_shutdown_sock(struct socket *sock, int how)
2421 {
2422 	int err;
2423 
2424 	err = security_socket_shutdown(sock, how);
2425 	if (!err)
2426 		err = READ_ONCE(sock->ops)->shutdown(sock, how);
2427 
2428 	return err;
2429 }
2430 
__sys_shutdown(int fd,int how)2431 int __sys_shutdown(int fd, int how)
2432 {
2433 	int err, fput_needed;
2434 	struct socket *sock;
2435 
2436 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2437 	if (sock != NULL) {
2438 		err = __sys_shutdown_sock(sock, how);
2439 		fput_light(sock->file, fput_needed);
2440 	}
2441 	return err;
2442 }
2443 
SYSCALL_DEFINE2(shutdown,int,fd,int,how)2444 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2445 {
2446 	return __sys_shutdown(fd, how);
2447 }
2448 
2449 /* A couple of helpful macros for getting the address of the 32/64 bit
2450  * fields which are the same type (int / unsigned) on our platforms.
2451  */
2452 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2453 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2454 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2455 
2456 struct used_address {
2457 	struct sockaddr_storage name;
2458 	unsigned int name_len;
2459 };
2460 
__copy_msghdr(struct msghdr * kmsg,struct user_msghdr * msg,struct sockaddr __user ** save_addr)2461 int __copy_msghdr(struct msghdr *kmsg,
2462 		  struct user_msghdr *msg,
2463 		  struct sockaddr __user **save_addr)
2464 {
2465 	ssize_t err;
2466 
2467 	kmsg->msg_control_is_user = true;
2468 	kmsg->msg_get_inq = 0;
2469 	kmsg->msg_control_user = msg->msg_control;
2470 	kmsg->msg_controllen = msg->msg_controllen;
2471 	kmsg->msg_flags = msg->msg_flags;
2472 
2473 	kmsg->msg_namelen = msg->msg_namelen;
2474 	if (!msg->msg_name)
2475 		kmsg->msg_namelen = 0;
2476 
2477 	if (kmsg->msg_namelen < 0)
2478 		return -EINVAL;
2479 
2480 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2481 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2482 
2483 	if (save_addr)
2484 		*save_addr = msg->msg_name;
2485 
2486 	if (msg->msg_name && kmsg->msg_namelen) {
2487 		if (!save_addr) {
2488 			err = move_addr_to_kernel(msg->msg_name,
2489 						  kmsg->msg_namelen,
2490 						  kmsg->msg_name);
2491 			if (err < 0)
2492 				return err;
2493 		}
2494 	} else {
2495 		kmsg->msg_name = NULL;
2496 		kmsg->msg_namelen = 0;
2497 	}
2498 
2499 	if (msg->msg_iovlen > UIO_MAXIOV)
2500 		return -EMSGSIZE;
2501 
2502 	kmsg->msg_iocb = NULL;
2503 	kmsg->msg_ubuf = NULL;
2504 	return 0;
2505 }
2506 
copy_msghdr_from_user(struct msghdr * kmsg,struct user_msghdr __user * umsg,struct sockaddr __user ** save_addr,struct iovec ** iov)2507 static int copy_msghdr_from_user(struct msghdr *kmsg,
2508 				 struct user_msghdr __user *umsg,
2509 				 struct sockaddr __user **save_addr,
2510 				 struct iovec **iov)
2511 {
2512 	struct user_msghdr msg;
2513 	ssize_t err;
2514 
2515 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2516 		return -EFAULT;
2517 
2518 	err = __copy_msghdr(kmsg, &msg, save_addr);
2519 	if (err)
2520 		return err;
2521 
2522 	err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2523 			    msg.msg_iov, msg.msg_iovlen,
2524 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2525 	return err < 0 ? err : 0;
2526 }
2527 
____sys_sendmsg(struct socket * sock,struct msghdr * msg_sys,unsigned int flags,struct used_address * used_address,unsigned int allowed_msghdr_flags)2528 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2529 			   unsigned int flags, struct used_address *used_address,
2530 			   unsigned int allowed_msghdr_flags)
2531 {
2532 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2533 				__aligned(sizeof(__kernel_size_t));
2534 	/* 20 is size of ipv6_pktinfo */
2535 	unsigned char *ctl_buf = ctl;
2536 	int ctl_len;
2537 	ssize_t err;
2538 
2539 	err = -ENOBUFS;
2540 
2541 	if (msg_sys->msg_controllen > INT_MAX)
2542 		goto out;
2543 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2544 	ctl_len = msg_sys->msg_controllen;
2545 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2546 		err =
2547 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2548 						     sizeof(ctl));
2549 		if (err)
2550 			goto out;
2551 		ctl_buf = msg_sys->msg_control;
2552 		ctl_len = msg_sys->msg_controllen;
2553 	} else if (ctl_len) {
2554 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2555 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2556 		if (ctl_len > sizeof(ctl)) {
2557 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2558 			if (ctl_buf == NULL)
2559 				goto out;
2560 		}
2561 		err = -EFAULT;
2562 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2563 			goto out_freectl;
2564 		msg_sys->msg_control = ctl_buf;
2565 		msg_sys->msg_control_is_user = false;
2566 	}
2567 	flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2568 	msg_sys->msg_flags = flags;
2569 
2570 	if (sock->file->f_flags & O_NONBLOCK)
2571 		msg_sys->msg_flags |= MSG_DONTWAIT;
2572 	/*
2573 	 * If this is sendmmsg() and current destination address is same as
2574 	 * previously succeeded address, omit asking LSM's decision.
2575 	 * used_address->name_len is initialized to UINT_MAX so that the first
2576 	 * destination address never matches.
2577 	 */
2578 	if (used_address && msg_sys->msg_name &&
2579 	    used_address->name_len == msg_sys->msg_namelen &&
2580 	    !memcmp(&used_address->name, msg_sys->msg_name,
2581 		    used_address->name_len)) {
2582 		err = sock_sendmsg_nosec(sock, msg_sys);
2583 		goto out_freectl;
2584 	}
2585 	err = __sock_sendmsg(sock, msg_sys);
2586 	/*
2587 	 * If this is sendmmsg() and sending to current destination address was
2588 	 * successful, remember it.
2589 	 */
2590 	if (used_address && err >= 0) {
2591 		used_address->name_len = msg_sys->msg_namelen;
2592 		if (msg_sys->msg_name)
2593 			memcpy(&used_address->name, msg_sys->msg_name,
2594 			       used_address->name_len);
2595 	}
2596 
2597 out_freectl:
2598 	if (ctl_buf != ctl)
2599 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2600 out:
2601 	return err;
2602 }
2603 
sendmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct iovec ** iov)2604 static int sendmsg_copy_msghdr(struct msghdr *msg,
2605 			       struct user_msghdr __user *umsg, unsigned flags,
2606 			       struct iovec **iov)
2607 {
2608 	int err;
2609 
2610 	if (flags & MSG_CMSG_COMPAT) {
2611 		struct compat_msghdr __user *msg_compat;
2612 
2613 		msg_compat = (struct compat_msghdr __user *) umsg;
2614 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2615 	} else {
2616 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2617 	}
2618 	if (err < 0)
2619 		return err;
2620 
2621 	return 0;
2622 }
2623 
___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)2624 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2625 			 struct msghdr *msg_sys, unsigned int flags,
2626 			 struct used_address *used_address,
2627 			 unsigned int allowed_msghdr_flags)
2628 {
2629 	struct sockaddr_storage address;
2630 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2631 	ssize_t err;
2632 
2633 	msg_sys->msg_name = &address;
2634 
2635 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2636 	if (err < 0)
2637 		return err;
2638 
2639 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2640 				allowed_msghdr_flags);
2641 	kfree(iov);
2642 	return err;
2643 }
2644 
2645 /*
2646  *	BSD sendmsg interface
2647  */
__sys_sendmsg_sock(struct socket * sock,struct msghdr * msg,unsigned int flags)2648 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2649 			unsigned int flags)
2650 {
2651 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2652 }
2653 
__sys_sendmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2654 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2655 		   bool forbid_cmsg_compat)
2656 {
2657 	int fput_needed, err;
2658 	struct msghdr msg_sys;
2659 	struct socket *sock;
2660 
2661 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2662 		return -EINVAL;
2663 
2664 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2665 	if (!sock)
2666 		goto out;
2667 
2668 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2669 
2670 	fput_light(sock->file, fput_needed);
2671 out:
2672 	return err;
2673 }
2674 
SYSCALL_DEFINE3(sendmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2675 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2676 {
2677 	return __sys_sendmsg(fd, msg, flags, true);
2678 }
2679 
2680 /*
2681  *	Linux sendmmsg interface
2682  */
2683 
__sys_sendmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,bool forbid_cmsg_compat)2684 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2685 		   unsigned int flags, bool forbid_cmsg_compat)
2686 {
2687 	int fput_needed, err, datagrams;
2688 	struct socket *sock;
2689 	struct mmsghdr __user *entry;
2690 	struct compat_mmsghdr __user *compat_entry;
2691 	struct msghdr msg_sys;
2692 	struct used_address used_address;
2693 	unsigned int oflags = flags;
2694 
2695 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2696 		return -EINVAL;
2697 
2698 	if (vlen > UIO_MAXIOV)
2699 		vlen = UIO_MAXIOV;
2700 
2701 	datagrams = 0;
2702 
2703 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2704 	if (!sock)
2705 		return err;
2706 
2707 	used_address.name_len = UINT_MAX;
2708 	entry = mmsg;
2709 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2710 	err = 0;
2711 	flags |= MSG_BATCH;
2712 
2713 	while (datagrams < vlen) {
2714 		if (datagrams == vlen - 1)
2715 			flags = oflags;
2716 
2717 		if (MSG_CMSG_COMPAT & flags) {
2718 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2719 					     &msg_sys, flags, &used_address, MSG_EOR);
2720 			if (err < 0)
2721 				break;
2722 			err = __put_user(err, &compat_entry->msg_len);
2723 			++compat_entry;
2724 		} else {
2725 			err = ___sys_sendmsg(sock,
2726 					     (struct user_msghdr __user *)entry,
2727 					     &msg_sys, flags, &used_address, MSG_EOR);
2728 			if (err < 0)
2729 				break;
2730 			err = put_user(err, &entry->msg_len);
2731 			++entry;
2732 		}
2733 
2734 		if (err)
2735 			break;
2736 		++datagrams;
2737 		if (msg_data_left(&msg_sys))
2738 			break;
2739 		cond_resched();
2740 	}
2741 
2742 	fput_light(sock->file, fput_needed);
2743 
2744 	/* We only return an error if no datagrams were able to be sent */
2745 	if (datagrams != 0)
2746 		return datagrams;
2747 
2748 	return err;
2749 }
2750 
SYSCALL_DEFINE4(sendmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags)2751 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2752 		unsigned int, vlen, unsigned int, flags)
2753 {
2754 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2755 }
2756 
recvmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct sockaddr __user ** uaddr,struct iovec ** iov)2757 static int recvmsg_copy_msghdr(struct msghdr *msg,
2758 			       struct user_msghdr __user *umsg, unsigned flags,
2759 			       struct sockaddr __user **uaddr,
2760 			       struct iovec **iov)
2761 {
2762 	ssize_t err;
2763 
2764 	if (MSG_CMSG_COMPAT & flags) {
2765 		struct compat_msghdr __user *msg_compat;
2766 
2767 		msg_compat = (struct compat_msghdr __user *) umsg;
2768 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2769 	} else {
2770 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2771 	}
2772 	if (err < 0)
2773 		return err;
2774 
2775 	return 0;
2776 }
2777 
____sys_recvmsg(struct socket * sock,struct msghdr * msg_sys,struct user_msghdr __user * msg,struct sockaddr __user * uaddr,unsigned int flags,int nosec)2778 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2779 			   struct user_msghdr __user *msg,
2780 			   struct sockaddr __user *uaddr,
2781 			   unsigned int flags, int nosec)
2782 {
2783 	struct compat_msghdr __user *msg_compat =
2784 					(struct compat_msghdr __user *) msg;
2785 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2786 	struct sockaddr_storage addr;
2787 	unsigned long cmsg_ptr;
2788 	int len;
2789 	ssize_t err;
2790 
2791 	msg_sys->msg_name = &addr;
2792 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2793 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2794 
2795 	/* We assume all kernel code knows the size of sockaddr_storage */
2796 	msg_sys->msg_namelen = 0;
2797 
2798 	if (sock->file->f_flags & O_NONBLOCK)
2799 		flags |= MSG_DONTWAIT;
2800 
2801 	if (unlikely(nosec))
2802 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2803 	else
2804 		err = sock_recvmsg(sock, msg_sys, flags);
2805 
2806 	if (err < 0)
2807 		goto out;
2808 	len = err;
2809 
2810 	if (uaddr != NULL) {
2811 		err = move_addr_to_user(&addr,
2812 					msg_sys->msg_namelen, uaddr,
2813 					uaddr_len);
2814 		if (err < 0)
2815 			goto out;
2816 	}
2817 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2818 			 COMPAT_FLAGS(msg));
2819 	if (err)
2820 		goto out;
2821 	if (MSG_CMSG_COMPAT & flags)
2822 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2823 				 &msg_compat->msg_controllen);
2824 	else
2825 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2826 				 &msg->msg_controllen);
2827 	if (err)
2828 		goto out;
2829 	err = len;
2830 out:
2831 	return err;
2832 }
2833 
___sys_recvmsg(struct socket * sock,struct user_msghdr __user * msg,struct msghdr * msg_sys,unsigned int flags,int nosec)2834 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2835 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2836 {
2837 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2838 	/* user mode address pointers */
2839 	struct sockaddr __user *uaddr;
2840 	ssize_t err;
2841 
2842 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2843 	if (err < 0)
2844 		return err;
2845 
2846 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2847 	kfree(iov);
2848 	return err;
2849 }
2850 
2851 /*
2852  *	BSD recvmsg interface
2853  */
2854 
__sys_recvmsg_sock(struct socket * sock,struct msghdr * msg,struct user_msghdr __user * umsg,struct sockaddr __user * uaddr,unsigned int flags)2855 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2856 			struct user_msghdr __user *umsg,
2857 			struct sockaddr __user *uaddr, unsigned int flags)
2858 {
2859 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2860 }
2861 
__sys_recvmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2862 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2863 		   bool forbid_cmsg_compat)
2864 {
2865 	int fput_needed, err;
2866 	struct msghdr msg_sys;
2867 	struct socket *sock;
2868 
2869 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2870 		return -EINVAL;
2871 
2872 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2873 	if (!sock)
2874 		goto out;
2875 
2876 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2877 
2878 	fput_light(sock->file, fput_needed);
2879 out:
2880 	return err;
2881 }
2882 
SYSCALL_DEFINE3(recvmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2883 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2884 		unsigned int, flags)
2885 {
2886 	return __sys_recvmsg(fd, msg, flags, true);
2887 }
2888 
2889 /*
2890  *     Linux recvmmsg interface
2891  */
2892 
do_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct timespec64 * timeout)2893 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2894 			  unsigned int vlen, unsigned int flags,
2895 			  struct timespec64 *timeout)
2896 {
2897 	int fput_needed, err, datagrams;
2898 	struct socket *sock;
2899 	struct mmsghdr __user *entry;
2900 	struct compat_mmsghdr __user *compat_entry;
2901 	struct msghdr msg_sys;
2902 	struct timespec64 end_time;
2903 	struct timespec64 timeout64;
2904 
2905 	if (timeout &&
2906 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2907 				    timeout->tv_nsec))
2908 		return -EINVAL;
2909 
2910 	datagrams = 0;
2911 
2912 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2913 	if (!sock)
2914 		return err;
2915 
2916 	if (likely(!(flags & MSG_ERRQUEUE))) {
2917 		err = sock_error(sock->sk);
2918 		if (err) {
2919 			datagrams = err;
2920 			goto out_put;
2921 		}
2922 	}
2923 
2924 	entry = mmsg;
2925 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2926 
2927 	while (datagrams < vlen) {
2928 		/*
2929 		 * No need to ask LSM for more than the first datagram.
2930 		 */
2931 		if (MSG_CMSG_COMPAT & flags) {
2932 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2933 					     &msg_sys, flags & ~MSG_WAITFORONE,
2934 					     datagrams);
2935 			if (err < 0)
2936 				break;
2937 			err = __put_user(err, &compat_entry->msg_len);
2938 			++compat_entry;
2939 		} else {
2940 			err = ___sys_recvmsg(sock,
2941 					     (struct user_msghdr __user *)entry,
2942 					     &msg_sys, flags & ~MSG_WAITFORONE,
2943 					     datagrams);
2944 			if (err < 0)
2945 				break;
2946 			err = put_user(err, &entry->msg_len);
2947 			++entry;
2948 		}
2949 
2950 		if (err)
2951 			break;
2952 		++datagrams;
2953 
2954 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2955 		if (flags & MSG_WAITFORONE)
2956 			flags |= MSG_DONTWAIT;
2957 
2958 		if (timeout) {
2959 			ktime_get_ts64(&timeout64);
2960 			*timeout = timespec64_sub(end_time, timeout64);
2961 			if (timeout->tv_sec < 0) {
2962 				timeout->tv_sec = timeout->tv_nsec = 0;
2963 				break;
2964 			}
2965 
2966 			/* Timeout, return less than vlen datagrams */
2967 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2968 				break;
2969 		}
2970 
2971 		/* Out of band data, return right away */
2972 		if (msg_sys.msg_flags & MSG_OOB)
2973 			break;
2974 		cond_resched();
2975 	}
2976 
2977 	if (err == 0)
2978 		goto out_put;
2979 
2980 	if (datagrams == 0) {
2981 		datagrams = err;
2982 		goto out_put;
2983 	}
2984 
2985 	/*
2986 	 * We may return less entries than requested (vlen) if the
2987 	 * sock is non block and there aren't enough datagrams...
2988 	 */
2989 	if (err != -EAGAIN) {
2990 		/*
2991 		 * ... or  if recvmsg returns an error after we
2992 		 * received some datagrams, where we record the
2993 		 * error to return on the next call or if the
2994 		 * app asks about it using getsockopt(SO_ERROR).
2995 		 */
2996 		WRITE_ONCE(sock->sk->sk_err, -err);
2997 	}
2998 out_put:
2999 	fput_light(sock->file, fput_needed);
3000 
3001 	return datagrams;
3002 }
3003 
__sys_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct __kernel_timespec __user * timeout,struct old_timespec32 __user * timeout32)3004 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3005 		   unsigned int vlen, unsigned int flags,
3006 		   struct __kernel_timespec __user *timeout,
3007 		   struct old_timespec32 __user *timeout32)
3008 {
3009 	int datagrams;
3010 	struct timespec64 timeout_sys;
3011 
3012 	if (timeout && get_timespec64(&timeout_sys, timeout))
3013 		return -EFAULT;
3014 
3015 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3016 		return -EFAULT;
3017 
3018 	if (!timeout && !timeout32)
3019 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3020 
3021 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3022 
3023 	if (datagrams <= 0)
3024 		return datagrams;
3025 
3026 	if (timeout && put_timespec64(&timeout_sys, timeout))
3027 		datagrams = -EFAULT;
3028 
3029 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3030 		datagrams = -EFAULT;
3031 
3032 	return datagrams;
3033 }
3034 
SYSCALL_DEFINE5(recvmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags,struct __kernel_timespec __user *,timeout)3035 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3036 		unsigned int, vlen, unsigned int, flags,
3037 		struct __kernel_timespec __user *, timeout)
3038 {
3039 	if (flags & MSG_CMSG_COMPAT)
3040 		return -EINVAL;
3041 
3042 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3043 }
3044 
3045 #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)3046 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3047 		unsigned int, vlen, unsigned int, flags,
3048 		struct old_timespec32 __user *, timeout)
3049 {
3050 	if (flags & MSG_CMSG_COMPAT)
3051 		return -EINVAL;
3052 
3053 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3054 }
3055 #endif
3056 
3057 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3058 /* Argument list sizes for sys_socketcall */
3059 #define AL(x) ((x) * sizeof(unsigned long))
3060 static const unsigned char nargs[21] = {
3061 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3062 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3063 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3064 	AL(4), AL(5), AL(4)
3065 };
3066 
3067 #undef AL
3068 
3069 /*
3070  *	System call vectors.
3071  *
3072  *	Argument checking cleaned up. Saved 20% in size.
3073  *  This function doesn't need to set the kernel lock because
3074  *  it is set by the callees.
3075  */
3076 
SYSCALL_DEFINE2(socketcall,int,call,unsigned long __user *,args)3077 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3078 {
3079 	unsigned long a[AUDITSC_ARGS];
3080 	unsigned long a0, a1;
3081 	int err;
3082 	unsigned int len;
3083 
3084 	if (call < 1 || call > SYS_SENDMMSG)
3085 		return -EINVAL;
3086 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
3087 
3088 	len = nargs[call];
3089 	if (len > sizeof(a))
3090 		return -EINVAL;
3091 
3092 	/* copy_from_user should be SMP safe. */
3093 	if (copy_from_user(a, args, len))
3094 		return -EFAULT;
3095 
3096 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3097 	if (err)
3098 		return err;
3099 
3100 	a0 = a[0];
3101 	a1 = a[1];
3102 
3103 	switch (call) {
3104 	case SYS_SOCKET:
3105 		err = __sys_socket(a0, a1, a[2]);
3106 		break;
3107 	case SYS_BIND:
3108 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3109 		break;
3110 	case SYS_CONNECT:
3111 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3112 		break;
3113 	case SYS_LISTEN:
3114 		err = __sys_listen(a0, a1);
3115 		break;
3116 	case SYS_ACCEPT:
3117 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3118 				    (int __user *)a[2], 0);
3119 		break;
3120 	case SYS_GETSOCKNAME:
3121 		err =
3122 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
3123 				      (int __user *)a[2]);
3124 		break;
3125 	case SYS_GETPEERNAME:
3126 		err =
3127 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
3128 				      (int __user *)a[2]);
3129 		break;
3130 	case SYS_SOCKETPAIR:
3131 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3132 		break;
3133 	case SYS_SEND:
3134 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3135 				   NULL, 0);
3136 		break;
3137 	case SYS_SENDTO:
3138 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3139 				   (struct sockaddr __user *)a[4], a[5]);
3140 		break;
3141 	case SYS_RECV:
3142 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3143 				     NULL, NULL);
3144 		break;
3145 	case SYS_RECVFROM:
3146 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3147 				     (struct sockaddr __user *)a[4],
3148 				     (int __user *)a[5]);
3149 		break;
3150 	case SYS_SHUTDOWN:
3151 		err = __sys_shutdown(a0, a1);
3152 		break;
3153 	case SYS_SETSOCKOPT:
3154 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3155 				       a[4]);
3156 		break;
3157 	case SYS_GETSOCKOPT:
3158 		err =
3159 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3160 				     (int __user *)a[4]);
3161 		break;
3162 	case SYS_SENDMSG:
3163 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3164 				    a[2], true);
3165 		break;
3166 	case SYS_SENDMMSG:
3167 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3168 				     a[3], true);
3169 		break;
3170 	case SYS_RECVMSG:
3171 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3172 				    a[2], true);
3173 		break;
3174 	case SYS_RECVMMSG:
3175 		if (IS_ENABLED(CONFIG_64BIT))
3176 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3177 					     a[2], a[3],
3178 					     (struct __kernel_timespec __user *)a[4],
3179 					     NULL);
3180 		else
3181 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3182 					     a[2], a[3], NULL,
3183 					     (struct old_timespec32 __user *)a[4]);
3184 		break;
3185 	case SYS_ACCEPT4:
3186 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3187 				    (int __user *)a[2], a[3]);
3188 		break;
3189 	default:
3190 		err = -EINVAL;
3191 		break;
3192 	}
3193 	return err;
3194 }
3195 
3196 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3197 
3198 /**
3199  *	sock_register - add a socket protocol handler
3200  *	@ops: description of protocol
3201  *
3202  *	This function is called by a protocol handler that wants to
3203  *	advertise its address family, and have it linked into the
3204  *	socket interface. The value ops->family corresponds to the
3205  *	socket system call protocol family.
3206  */
sock_register(const struct net_proto_family * ops)3207 int sock_register(const struct net_proto_family *ops)
3208 {
3209 	int err;
3210 
3211 	if (ops->family >= NPROTO) {
3212 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3213 		return -ENOBUFS;
3214 	}
3215 
3216 	spin_lock(&net_family_lock);
3217 	if (rcu_dereference_protected(net_families[ops->family],
3218 				      lockdep_is_held(&net_family_lock)))
3219 		err = -EEXIST;
3220 	else {
3221 		rcu_assign_pointer(net_families[ops->family], ops);
3222 		err = 0;
3223 	}
3224 	spin_unlock(&net_family_lock);
3225 
3226 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3227 	return err;
3228 }
3229 EXPORT_SYMBOL(sock_register);
3230 
3231 /**
3232  *	sock_unregister - remove a protocol handler
3233  *	@family: protocol family to remove
3234  *
3235  *	This function is called by a protocol handler that wants to
3236  *	remove its address family, and have it unlinked from the
3237  *	new socket creation.
3238  *
3239  *	If protocol handler is a module, then it can use module reference
3240  *	counts to protect against new references. If protocol handler is not
3241  *	a module then it needs to provide its own protection in
3242  *	the ops->create routine.
3243  */
sock_unregister(int family)3244 void sock_unregister(int family)
3245 {
3246 	BUG_ON(family < 0 || family >= NPROTO);
3247 
3248 	spin_lock(&net_family_lock);
3249 	RCU_INIT_POINTER(net_families[family], NULL);
3250 	spin_unlock(&net_family_lock);
3251 
3252 	synchronize_rcu();
3253 
3254 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3255 }
3256 EXPORT_SYMBOL(sock_unregister);
3257 
sock_is_registered(int family)3258 bool sock_is_registered(int family)
3259 {
3260 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3261 }
3262 
sock_init(void)3263 static int __init sock_init(void)
3264 {
3265 	int err;
3266 	/*
3267 	 *      Initialize the network sysctl infrastructure.
3268 	 */
3269 	err = net_sysctl_init();
3270 	if (err)
3271 		goto out;
3272 
3273 	/*
3274 	 *      Initialize skbuff SLAB cache
3275 	 */
3276 	skb_init();
3277 
3278 	/*
3279 	 *      Initialize the protocols module.
3280 	 */
3281 
3282 	init_inodecache();
3283 
3284 	err = register_filesystem(&sock_fs_type);
3285 	if (err)
3286 		goto out;
3287 	sock_mnt = kern_mount(&sock_fs_type);
3288 	if (IS_ERR(sock_mnt)) {
3289 		err = PTR_ERR(sock_mnt);
3290 		goto out_mount;
3291 	}
3292 
3293 	/* The real protocol initialization is performed in later initcalls.
3294 	 */
3295 
3296 #ifdef CONFIG_NETFILTER
3297 	err = netfilter_init();
3298 	if (err)
3299 		goto out;
3300 #endif
3301 
3302 	ptp_classifier_init();
3303 
3304 out:
3305 	return err;
3306 
3307 out_mount:
3308 	unregister_filesystem(&sock_fs_type);
3309 	goto out;
3310 }
3311 
3312 core_initcall(sock_init);	/* early initcall */
3313 
3314 #ifdef CONFIG_PROC_FS
socket_seq_show(struct seq_file * seq)3315 void socket_seq_show(struct seq_file *seq)
3316 {
3317 	seq_printf(seq, "sockets: used %d\n",
3318 		   sock_inuse_get(seq->private));
3319 }
3320 #endif				/* CONFIG_PROC_FS */
3321 
3322 /* Handle the fact that while struct ifreq has the same *layout* on
3323  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3324  * which are handled elsewhere, it still has different *size* due to
3325  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3326  * resulting in struct ifreq being 32 and 40 bytes respectively).
3327  * As a result, if the struct happens to be at the end of a page and
3328  * the next page isn't readable/writable, we get a fault. To prevent
3329  * that, copy back and forth to the full size.
3330  */
get_user_ifreq(struct ifreq * ifr,void __user ** ifrdata,void __user * arg)3331 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3332 {
3333 	if (in_compat_syscall()) {
3334 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3335 
3336 		memset(ifr, 0, sizeof(*ifr));
3337 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3338 			return -EFAULT;
3339 
3340 		if (ifrdata)
3341 			*ifrdata = compat_ptr(ifr32->ifr_data);
3342 
3343 		return 0;
3344 	}
3345 
3346 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3347 		return -EFAULT;
3348 
3349 	if (ifrdata)
3350 		*ifrdata = ifr->ifr_data;
3351 
3352 	return 0;
3353 }
3354 EXPORT_SYMBOL(get_user_ifreq);
3355 
put_user_ifreq(struct ifreq * ifr,void __user * arg)3356 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3357 {
3358 	size_t size = sizeof(*ifr);
3359 
3360 	if (in_compat_syscall())
3361 		size = sizeof(struct compat_ifreq);
3362 
3363 	if (copy_to_user(arg, ifr, size))
3364 		return -EFAULT;
3365 
3366 	return 0;
3367 }
3368 EXPORT_SYMBOL(put_user_ifreq);
3369 
3370 #ifdef CONFIG_COMPAT
compat_siocwandev(struct net * net,struct compat_ifreq __user * uifr32)3371 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3372 {
3373 	compat_uptr_t uptr32;
3374 	struct ifreq ifr;
3375 	void __user *saved;
3376 	int err;
3377 
3378 	if (get_user_ifreq(&ifr, NULL, uifr32))
3379 		return -EFAULT;
3380 
3381 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3382 		return -EFAULT;
3383 
3384 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3385 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3386 
3387 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3388 	if (!err) {
3389 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3390 		if (put_user_ifreq(&ifr, uifr32))
3391 			err = -EFAULT;
3392 	}
3393 	return err;
3394 }
3395 
3396 /* 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)3397 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3398 				 struct compat_ifreq __user *u_ifreq32)
3399 {
3400 	struct ifreq ifreq;
3401 	void __user *data;
3402 
3403 	if (!is_socket_ioctl_cmd(cmd))
3404 		return -ENOTTY;
3405 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3406 		return -EFAULT;
3407 	ifreq.ifr_data = data;
3408 
3409 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3410 }
3411 
compat_sock_ioctl_trans(struct file * file,struct socket * sock,unsigned int cmd,unsigned long arg)3412 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3413 			 unsigned int cmd, unsigned long arg)
3414 {
3415 	void __user *argp = compat_ptr(arg);
3416 	struct sock *sk = sock->sk;
3417 	struct net *net = sock_net(sk);
3418 	const struct proto_ops *ops;
3419 
3420 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3421 		return sock_ioctl(file, cmd, (unsigned long)argp);
3422 
3423 	switch (cmd) {
3424 	case SIOCWANDEV:
3425 		return compat_siocwandev(net, argp);
3426 	case SIOCGSTAMP_OLD:
3427 	case SIOCGSTAMPNS_OLD:
3428 		ops = READ_ONCE(sock->ops);
3429 		if (!ops->gettstamp)
3430 			return -ENOIOCTLCMD;
3431 		return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3432 				      !COMPAT_USE_64BIT_TIME);
3433 
3434 	case SIOCETHTOOL:
3435 	case SIOCBONDSLAVEINFOQUERY:
3436 	case SIOCBONDINFOQUERY:
3437 	case SIOCSHWTSTAMP:
3438 	case SIOCGHWTSTAMP:
3439 		return compat_ifr_data_ioctl(net, cmd, argp);
3440 
3441 	case FIOSETOWN:
3442 	case SIOCSPGRP:
3443 	case FIOGETOWN:
3444 	case SIOCGPGRP:
3445 	case SIOCBRADDBR:
3446 	case SIOCBRDELBR:
3447 	case SIOCGIFVLAN:
3448 	case SIOCSIFVLAN:
3449 	case SIOCGSKNS:
3450 	case SIOCGSTAMP_NEW:
3451 	case SIOCGSTAMPNS_NEW:
3452 	case SIOCGIFCONF:
3453 	case SIOCSIFBR:
3454 	case SIOCGIFBR:
3455 		return sock_ioctl(file, cmd, arg);
3456 
3457 	case SIOCGIFFLAGS:
3458 	case SIOCSIFFLAGS:
3459 	case SIOCGIFMAP:
3460 	case SIOCSIFMAP:
3461 	case SIOCGIFMETRIC:
3462 	case SIOCSIFMETRIC:
3463 	case SIOCGIFMTU:
3464 	case SIOCSIFMTU:
3465 	case SIOCGIFMEM:
3466 	case SIOCSIFMEM:
3467 	case SIOCGIFHWADDR:
3468 	case SIOCSIFHWADDR:
3469 	case SIOCADDMULTI:
3470 	case SIOCDELMULTI:
3471 	case SIOCGIFINDEX:
3472 	case SIOCGIFADDR:
3473 	case SIOCSIFADDR:
3474 	case SIOCSIFHWBROADCAST:
3475 	case SIOCDIFADDR:
3476 	case SIOCGIFBRDADDR:
3477 	case SIOCSIFBRDADDR:
3478 	case SIOCGIFDSTADDR:
3479 	case SIOCSIFDSTADDR:
3480 	case SIOCGIFNETMASK:
3481 	case SIOCSIFNETMASK:
3482 	case SIOCSIFPFLAGS:
3483 	case SIOCGIFPFLAGS:
3484 	case SIOCGIFTXQLEN:
3485 	case SIOCSIFTXQLEN:
3486 	case SIOCBRADDIF:
3487 	case SIOCBRDELIF:
3488 	case SIOCGIFNAME:
3489 	case SIOCSIFNAME:
3490 	case SIOCGMIIPHY:
3491 	case SIOCGMIIREG:
3492 	case SIOCSMIIREG:
3493 	case SIOCBONDENSLAVE:
3494 	case SIOCBONDRELEASE:
3495 	case SIOCBONDSETHWADDR:
3496 	case SIOCBONDCHANGEACTIVE:
3497 	case SIOCSARP:
3498 	case SIOCGARP:
3499 	case SIOCDARP:
3500 	case SIOCOUTQ:
3501 	case SIOCOUTQNSD:
3502 	case SIOCATMARK:
3503 		return sock_do_ioctl(net, sock, cmd, arg);
3504 	}
3505 
3506 	return -ENOIOCTLCMD;
3507 }
3508 
compat_sock_ioctl(struct file * file,unsigned int cmd,unsigned long arg)3509 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3510 			      unsigned long arg)
3511 {
3512 	struct socket *sock = file->private_data;
3513 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3514 	int ret = -ENOIOCTLCMD;
3515 	struct sock *sk;
3516 	struct net *net;
3517 
3518 	sk = sock->sk;
3519 	net = sock_net(sk);
3520 
3521 	if (ops->compat_ioctl)
3522 		ret = ops->compat_ioctl(sock, cmd, arg);
3523 
3524 	if (ret == -ENOIOCTLCMD &&
3525 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3526 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3527 
3528 	if (ret == -ENOIOCTLCMD)
3529 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3530 
3531 	return ret;
3532 }
3533 #endif
3534 
3535 /**
3536  *	kernel_bind - bind an address to a socket (kernel space)
3537  *	@sock: socket
3538  *	@addr: address
3539  *	@addrlen: length of address
3540  *
3541  *	Returns 0 or an error.
3542  */
3543 
kernel_bind(struct socket * sock,struct sockaddr * addr,int addrlen)3544 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3545 {
3546 	struct sockaddr_storage address;
3547 
3548 	memcpy(&address, addr, addrlen);
3549 
3550 	return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3551 					  addrlen);
3552 }
3553 EXPORT_SYMBOL(kernel_bind);
3554 
3555 /**
3556  *	kernel_listen - move socket to listening state (kernel space)
3557  *	@sock: socket
3558  *	@backlog: pending connections queue size
3559  *
3560  *	Returns 0 or an error.
3561  */
3562 
kernel_listen(struct socket * sock,int backlog)3563 int kernel_listen(struct socket *sock, int backlog)
3564 {
3565 	return READ_ONCE(sock->ops)->listen(sock, backlog);
3566 }
3567 EXPORT_SYMBOL(kernel_listen);
3568 
3569 /**
3570  *	kernel_accept - accept a connection (kernel space)
3571  *	@sock: listening socket
3572  *	@newsock: new connected socket
3573  *	@flags: flags
3574  *
3575  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3576  *	If it fails, @newsock is guaranteed to be %NULL.
3577  *	Returns 0 or an error.
3578  */
3579 
kernel_accept(struct socket * sock,struct socket ** newsock,int flags)3580 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3581 {
3582 	struct sock *sk = sock->sk;
3583 	const struct proto_ops *ops = READ_ONCE(sock->ops);
3584 	struct proto_accept_arg arg = {
3585 		.flags = flags,
3586 		.kern = true,
3587 	};
3588 	int err;
3589 
3590 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3591 			       newsock);
3592 	if (err < 0)
3593 		goto done;
3594 
3595 	err = ops->accept(sock, *newsock, &arg);
3596 	if (err < 0) {
3597 		sock_release(*newsock);
3598 		*newsock = NULL;
3599 		goto done;
3600 	}
3601 
3602 	(*newsock)->ops = ops;
3603 	__module_get(ops->owner);
3604 
3605 done:
3606 	return err;
3607 }
3608 EXPORT_SYMBOL(kernel_accept);
3609 
3610 /**
3611  *	kernel_connect - connect a socket (kernel space)
3612  *	@sock: socket
3613  *	@addr: address
3614  *	@addrlen: address length
3615  *	@flags: flags (O_NONBLOCK, ...)
3616  *
3617  *	For datagram sockets, @addr is the address to which datagrams are sent
3618  *	by default, and the only address from which datagrams are received.
3619  *	For stream sockets, attempts to connect to @addr.
3620  *	Returns 0 or an error code.
3621  */
3622 
kernel_connect(struct socket * sock,struct sockaddr * addr,int addrlen,int flags)3623 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3624 		   int flags)
3625 {
3626 	struct sockaddr_storage address;
3627 
3628 	memcpy(&address, addr, addrlen);
3629 
3630 	return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3631 					     addrlen, flags);
3632 }
3633 EXPORT_SYMBOL(kernel_connect);
3634 
3635 /**
3636  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3637  *	@sock: socket
3638  *	@addr: address holder
3639  *
3640  * 	Fills the @addr pointer with the address which the socket is bound.
3641  *	Returns the length of the address in bytes or an error code.
3642  */
3643 
kernel_getsockname(struct socket * sock,struct sockaddr * addr)3644 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3645 {
3646 	return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3647 }
3648 EXPORT_SYMBOL(kernel_getsockname);
3649 
3650 /**
3651  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3652  *	@sock: socket
3653  *	@addr: address holder
3654  *
3655  * 	Fills the @addr pointer with the address which the socket is connected.
3656  *	Returns the length of the address in bytes or an error code.
3657  */
3658 
kernel_getpeername(struct socket * sock,struct sockaddr * addr)3659 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3660 {
3661 	return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3662 }
3663 EXPORT_SYMBOL(kernel_getpeername);
3664 
3665 /**
3666  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3667  *	@sock: socket
3668  *	@how: connection part
3669  *
3670  *	Returns 0 or an error.
3671  */
3672 
kernel_sock_shutdown(struct socket * sock,enum sock_shutdown_cmd how)3673 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3674 {
3675 	return READ_ONCE(sock->ops)->shutdown(sock, how);
3676 }
3677 EXPORT_SYMBOL(kernel_sock_shutdown);
3678 
3679 /**
3680  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3681  *	@sk: socket
3682  *
3683  *	This routine returns the IP overhead imposed by a socket i.e.
3684  *	the length of the underlying IP header, depending on whether
3685  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3686  *	on at the socket. Assumes that the caller has a lock on the socket.
3687  */
3688 
kernel_sock_ip_overhead(struct sock * sk)3689 u32 kernel_sock_ip_overhead(struct sock *sk)
3690 {
3691 	struct inet_sock *inet;
3692 	struct ip_options_rcu *opt;
3693 	u32 overhead = 0;
3694 #if IS_ENABLED(CONFIG_IPV6)
3695 	struct ipv6_pinfo *np;
3696 	struct ipv6_txoptions *optv6 = NULL;
3697 #endif /* IS_ENABLED(CONFIG_IPV6) */
3698 
3699 	if (!sk)
3700 		return overhead;
3701 
3702 	switch (sk->sk_family) {
3703 	case AF_INET:
3704 		inet = inet_sk(sk);
3705 		overhead += sizeof(struct iphdr);
3706 		opt = rcu_dereference_protected(inet->inet_opt,
3707 						sock_owned_by_user(sk));
3708 		if (opt)
3709 			overhead += opt->opt.optlen;
3710 		return overhead;
3711 #if IS_ENABLED(CONFIG_IPV6)
3712 	case AF_INET6:
3713 		np = inet6_sk(sk);
3714 		overhead += sizeof(struct ipv6hdr);
3715 		if (np)
3716 			optv6 = rcu_dereference_protected(np->opt,
3717 							  sock_owned_by_user(sk));
3718 		if (optv6)
3719 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3720 		return overhead;
3721 #endif /* IS_ENABLED(CONFIG_IPV6) */
3722 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3723 		return overhead;
3724 	}
3725 }
3726 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3727