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/net.h>
61 #include <linux/interrupt.h>
62 #include <linux/thread_info.h>
63 #include <linux/rcupdate.h>
64 #include <linux/netdevice.h>
65 #include <linux/proc_fs.h>
66 #include <linux/seq_file.h>
67 #include <linux/mutex.h>
68 #include <linux/if_bridge.h>
69 #include <linux/if_vlan.h>
70 #include <linux/ptp_classify.h>
71 #include <linux/init.h>
72 #include <linux/poll.h>
73 #include <linux/cache.h>
74 #include <linux/module.h>
75 #include <linux/highmem.h>
76 #include <linux/mount.h>
77 #include <linux/pseudo_fs.h>
78 #include <linux/security.h>
79 #include <linux/syscalls.h>
80 #include <linux/compat.h>
81 #include <linux/kmod.h>
82 #include <linux/audit.h>
83 #include <linux/wireless.h>
84 #include <linux/nsproxy.h>
85 #include <linux/magic.h>
86 #include <linux/slab.h>
87 #include <linux/xattr.h>
88 #include <linux/nospec.h>
89 #include <linux/indirect_call_wrapper.h>
90 
91 #include <linux/uaccess.h>
92 #include <asm/unistd.h>
93 
94 #include <net/compat.h>
95 #include <net/wext.h>
96 #include <net/cls_cgroup.h>
97 
98 #include <net/sock.h>
99 #include <linux/netfilter.h>
100 
101 #include <linux/if_tun.h>
102 #include <linux/ipv6_route.h>
103 #include <linux/route.h>
104 #include <linux/termios.h>
105 #include <linux/sockios.h>
106 #include <net/busy_poll.h>
107 #include <linux/errqueue.h>
108 #include <linux/ptp_clock_kernel.h>
109 
110 #ifdef CONFIG_NET_RX_BUSY_POLL
111 unsigned int sysctl_net_busy_read __read_mostly;
112 unsigned int sysctl_net_busy_poll __read_mostly;
113 #endif
114 
115 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
116 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
117 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
118 
119 static int sock_close(struct inode *inode, struct file *file);
120 static __poll_t sock_poll(struct file *file,
121 			      struct poll_table_struct *wait);
122 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
123 #ifdef CONFIG_COMPAT
124 static long compat_sock_ioctl(struct file *file,
125 			      unsigned int cmd, unsigned long arg);
126 #endif
127 static int sock_fasync(int fd, struct file *filp, int on);
128 static ssize_t sock_sendpage(struct file *file, struct page *page,
129 			     int offset, size_t size, loff_t *ppos, int more);
130 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
131 				struct pipe_inode_info *pipe, size_t len,
132 				unsigned int flags);
133 
134 #ifdef CONFIG_PROC_FS
sock_show_fdinfo(struct seq_file * m,struct file * f)135 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
136 {
137 	struct socket *sock = f->private_data;
138 
139 	if (sock->ops->show_fdinfo)
140 		sock->ops->show_fdinfo(m, sock);
141 }
142 #else
143 #define sock_show_fdinfo NULL
144 #endif
145 
146 /*
147  *	Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
148  *	in the operation structures but are done directly via the socketcall() multiplexor.
149  */
150 
151 static const struct file_operations socket_file_ops = {
152 	.owner =	THIS_MODULE,
153 	.llseek =	no_llseek,
154 	.read_iter =	sock_read_iter,
155 	.write_iter =	sock_write_iter,
156 	.poll =		sock_poll,
157 	.unlocked_ioctl = sock_ioctl,
158 #ifdef CONFIG_COMPAT
159 	.compat_ioctl = compat_sock_ioctl,
160 #endif
161 	.mmap =		sock_mmap,
162 	.release =	sock_close,
163 	.fasync =	sock_fasync,
164 	.sendpage =	sock_sendpage,
165 	.splice_write = generic_splice_sendpage,
166 	.splice_read =	sock_splice_read,
167 	.show_fdinfo =	sock_show_fdinfo,
168 };
169 
170 static const char * const pf_family_names[] = {
171 	[PF_UNSPEC]	= "PF_UNSPEC",
172 	[PF_UNIX]	= "PF_UNIX/PF_LOCAL",
173 	[PF_INET]	= "PF_INET",
174 	[PF_AX25]	= "PF_AX25",
175 	[PF_IPX]	= "PF_IPX",
176 	[PF_APPLETALK]	= "PF_APPLETALK",
177 	[PF_NETROM]	= "PF_NETROM",
178 	[PF_BRIDGE]	= "PF_BRIDGE",
179 	[PF_ATMPVC]	= "PF_ATMPVC",
180 	[PF_X25]	= "PF_X25",
181 	[PF_INET6]	= "PF_INET6",
182 	[PF_ROSE]	= "PF_ROSE",
183 	[PF_DECnet]	= "PF_DECnet",
184 	[PF_NETBEUI]	= "PF_NETBEUI",
185 	[PF_SECURITY]	= "PF_SECURITY",
186 	[PF_KEY]	= "PF_KEY",
187 	[PF_NETLINK]	= "PF_NETLINK/PF_ROUTE",
188 	[PF_PACKET]	= "PF_PACKET",
189 	[PF_ASH]	= "PF_ASH",
190 	[PF_ECONET]	= "PF_ECONET",
191 	[PF_ATMSVC]	= "PF_ATMSVC",
192 	[PF_RDS]	= "PF_RDS",
193 	[PF_SNA]	= "PF_SNA",
194 	[PF_IRDA]	= "PF_IRDA",
195 	[PF_PPPOX]	= "PF_PPPOX",
196 	[PF_WANPIPE]	= "PF_WANPIPE",
197 	[PF_LLC]	= "PF_LLC",
198 	[PF_IB]		= "PF_IB",
199 	[PF_MPLS]	= "PF_MPLS",
200 	[PF_CAN]	= "PF_CAN",
201 	[PF_TIPC]	= "PF_TIPC",
202 	[PF_BLUETOOTH]	= "PF_BLUETOOTH",
203 	[PF_IUCV]	= "PF_IUCV",
204 	[PF_RXRPC]	= "PF_RXRPC",
205 	[PF_ISDN]	= "PF_ISDN",
206 	[PF_PHONET]	= "PF_PHONET",
207 	[PF_IEEE802154]	= "PF_IEEE802154",
208 	[PF_CAIF]	= "PF_CAIF",
209 	[PF_ALG]	= "PF_ALG",
210 	[PF_NFC]	= "PF_NFC",
211 	[PF_VSOCK]	= "PF_VSOCK",
212 	[PF_KCM]	= "PF_KCM",
213 	[PF_QIPCRTR]	= "PF_QIPCRTR",
214 	[PF_SMC]	= "PF_SMC",
215 	[PF_XDP]	= "PF_XDP",
216 	[PF_MCTP]	= "PF_MCTP",
217 };
218 
219 /*
220  *	The protocol list. Each protocol is registered in here.
221  */
222 
223 static DEFINE_SPINLOCK(net_family_lock);
224 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
225 
226 /*
227  * Support routines.
228  * Move socket addresses back and forth across the kernel/user
229  * divide and look after the messy bits.
230  */
231 
232 /**
233  *	move_addr_to_kernel	-	copy a socket address into kernel space
234  *	@uaddr: Address in user space
235  *	@kaddr: Address in kernel space
236  *	@ulen: Length in user space
237  *
238  *	The address is copied into kernel space. If the provided address is
239  *	too long an error code of -EINVAL is returned. If the copy gives
240  *	invalid addresses -EFAULT is returned. On a success 0 is returned.
241  */
242 
move_addr_to_kernel(void __user * uaddr,int ulen,struct sockaddr_storage * kaddr)243 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
244 {
245 	if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
246 		return -EINVAL;
247 	if (ulen == 0)
248 		return 0;
249 	if (copy_from_user(kaddr, uaddr, ulen))
250 		return -EFAULT;
251 	return audit_sockaddr(ulen, kaddr);
252 }
253 
254 /**
255  *	move_addr_to_user	-	copy an address to user space
256  *	@kaddr: kernel space address
257  *	@klen: length of address in kernel
258  *	@uaddr: user space address
259  *	@ulen: pointer to user length field
260  *
261  *	The value pointed to by ulen on entry is the buffer length available.
262  *	This is overwritten with the buffer space used. -EINVAL is returned
263  *	if an overlong buffer is specified or a negative buffer size. -EFAULT
264  *	is returned if either the buffer or the length field are not
265  *	accessible.
266  *	After copying the data up to the limit the user specifies, the true
267  *	length of the data is written over the length limit the user
268  *	specified. Zero is returned for a success.
269  */
270 
move_addr_to_user(struct sockaddr_storage * kaddr,int klen,void __user * uaddr,int __user * ulen)271 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
272 			     void __user *uaddr, int __user *ulen)
273 {
274 	int err;
275 	int len;
276 
277 	BUG_ON(klen > sizeof(struct sockaddr_storage));
278 	err = get_user(len, ulen);
279 	if (err)
280 		return err;
281 	if (len > klen)
282 		len = klen;
283 	if (len < 0)
284 		return -EINVAL;
285 	if (len) {
286 		if (audit_sockaddr(klen, kaddr))
287 			return -ENOMEM;
288 		if (copy_to_user(uaddr, kaddr, len))
289 			return -EFAULT;
290 	}
291 	/*
292 	 *      "fromlen shall refer to the value before truncation.."
293 	 *                      1003.1g
294 	 */
295 	return __put_user(klen, ulen);
296 }
297 
298 static struct kmem_cache *sock_inode_cachep __ro_after_init;
299 
sock_alloc_inode(struct super_block * sb)300 static struct inode *sock_alloc_inode(struct super_block *sb)
301 {
302 	struct socket_alloc *ei;
303 
304 	ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
305 	if (!ei)
306 		return NULL;
307 	init_waitqueue_head(&ei->socket.wq.wait);
308 	ei->socket.wq.fasync_list = NULL;
309 	ei->socket.wq.flags = 0;
310 
311 	ei->socket.state = SS_UNCONNECTED;
312 	ei->socket.flags = 0;
313 	ei->socket.ops = NULL;
314 	ei->socket.sk = NULL;
315 	ei->socket.file = NULL;
316 
317 	return &ei->vfs_inode;
318 }
319 
sock_free_inode(struct inode * inode)320 static void sock_free_inode(struct inode *inode)
321 {
322 	struct socket_alloc *ei;
323 
324 	ei = container_of(inode, struct socket_alloc, vfs_inode);
325 	kmem_cache_free(sock_inode_cachep, ei);
326 }
327 
init_once(void * foo)328 static void init_once(void *foo)
329 {
330 	struct socket_alloc *ei = (struct socket_alloc *)foo;
331 
332 	inode_init_once(&ei->vfs_inode);
333 }
334 
init_inodecache(void)335 static void init_inodecache(void)
336 {
337 	sock_inode_cachep = kmem_cache_create("sock_inode_cache",
338 					      sizeof(struct socket_alloc),
339 					      0,
340 					      (SLAB_HWCACHE_ALIGN |
341 					       SLAB_RECLAIM_ACCOUNT |
342 					       SLAB_MEM_SPREAD | SLAB_ACCOUNT),
343 					      init_once);
344 	BUG_ON(sock_inode_cachep == NULL);
345 }
346 
347 static const struct super_operations sockfs_ops = {
348 	.alloc_inode	= sock_alloc_inode,
349 	.free_inode	= sock_free_inode,
350 	.statfs		= simple_statfs,
351 };
352 
353 /*
354  * sockfs_dname() is called from d_path().
355  */
sockfs_dname(struct dentry * dentry,char * buffer,int buflen)356 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
357 {
358 	return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
359 				d_inode(dentry)->i_ino);
360 }
361 
362 static const struct dentry_operations sockfs_dentry_operations = {
363 	.d_dname  = sockfs_dname,
364 };
365 
sockfs_xattr_get(const struct xattr_handler * handler,struct dentry * dentry,struct inode * inode,const char * suffix,void * value,size_t size)366 static int sockfs_xattr_get(const struct xattr_handler *handler,
367 			    struct dentry *dentry, struct inode *inode,
368 			    const char *suffix, void *value, size_t size)
369 {
370 	if (value) {
371 		if (dentry->d_name.len + 1 > size)
372 			return -ERANGE;
373 		memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
374 	}
375 	return dentry->d_name.len + 1;
376 }
377 
378 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
379 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
380 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
381 
382 static const struct xattr_handler sockfs_xattr_handler = {
383 	.name = XATTR_NAME_SOCKPROTONAME,
384 	.get = sockfs_xattr_get,
385 };
386 
sockfs_security_xattr_set(const struct xattr_handler * handler,struct user_namespace * mnt_userns,struct dentry * dentry,struct inode * inode,const char * suffix,const void * value,size_t size,int flags)387 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
388 				     struct user_namespace *mnt_userns,
389 				     struct dentry *dentry, struct inode *inode,
390 				     const char *suffix, const void *value,
391 				     size_t size, int flags)
392 {
393 	/* Handled by LSM. */
394 	return -EAGAIN;
395 }
396 
397 static const struct xattr_handler sockfs_security_xattr_handler = {
398 	.prefix = XATTR_SECURITY_PREFIX,
399 	.set = sockfs_security_xattr_set,
400 };
401 
402 static const struct xattr_handler *sockfs_xattr_handlers[] = {
403 	&sockfs_xattr_handler,
404 	&sockfs_security_xattr_handler,
405 	NULL
406 };
407 
sockfs_init_fs_context(struct fs_context * fc)408 static int sockfs_init_fs_context(struct fs_context *fc)
409 {
410 	struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
411 	if (!ctx)
412 		return -ENOMEM;
413 	ctx->ops = &sockfs_ops;
414 	ctx->dops = &sockfs_dentry_operations;
415 	ctx->xattr = sockfs_xattr_handlers;
416 	return 0;
417 }
418 
419 static struct vfsmount *sock_mnt __read_mostly;
420 
421 static struct file_system_type sock_fs_type = {
422 	.name =		"sockfs",
423 	.init_fs_context = sockfs_init_fs_context,
424 	.kill_sb =	kill_anon_super,
425 };
426 
427 /*
428  *	Obtains the first available file descriptor and sets it up for use.
429  *
430  *	These functions create file structures and maps them to fd space
431  *	of the current process. On success it returns file descriptor
432  *	and file struct implicitly stored in sock->file.
433  *	Note that another thread may close file descriptor before we return
434  *	from this function. We use the fact that now we do not refer
435  *	to socket after mapping. If one day we will need it, this
436  *	function will increment ref. count on file by 1.
437  *
438  *	In any case returned fd MAY BE not valid!
439  *	This race condition is unavoidable
440  *	with shared fd spaces, we cannot solve it inside kernel,
441  *	but we take care of internal coherence yet.
442  */
443 
444 /**
445  *	sock_alloc_file - Bind a &socket to a &file
446  *	@sock: socket
447  *	@flags: file status flags
448  *	@dname: protocol name
449  *
450  *	Returns the &file bound with @sock, implicitly storing it
451  *	in sock->file. If dname is %NULL, sets to "".
452  *	On failure the return is a ERR pointer (see linux/err.h).
453  *	This function uses GFP_KERNEL internally.
454  */
455 
sock_alloc_file(struct socket * sock,int flags,const char * dname)456 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
457 {
458 	struct file *file;
459 
460 	if (!dname)
461 		dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
462 
463 	file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
464 				O_RDWR | (flags & O_NONBLOCK),
465 				&socket_file_ops);
466 	if (IS_ERR(file)) {
467 		sock_release(sock);
468 		return file;
469 	}
470 
471 	sock->file = file;
472 	file->private_data = sock;
473 	stream_open(SOCK_INODE(sock), file);
474 	return file;
475 }
476 EXPORT_SYMBOL(sock_alloc_file);
477 
sock_map_fd(struct socket * sock,int flags)478 static int sock_map_fd(struct socket *sock, int flags)
479 {
480 	struct file *newfile;
481 	int fd = get_unused_fd_flags(flags);
482 	if (unlikely(fd < 0)) {
483 		sock_release(sock);
484 		return fd;
485 	}
486 
487 	newfile = sock_alloc_file(sock, flags, NULL);
488 	if (!IS_ERR(newfile)) {
489 		fd_install(fd, newfile);
490 		return fd;
491 	}
492 
493 	put_unused_fd(fd);
494 	return PTR_ERR(newfile);
495 }
496 
497 /**
498  *	sock_from_file - Return the &socket bounded to @file.
499  *	@file: file
500  *
501  *	On failure returns %NULL.
502  */
503 
sock_from_file(struct file * file)504 struct socket *sock_from_file(struct file *file)
505 {
506 	if (file->f_op == &socket_file_ops)
507 		return file->private_data;	/* set in sock_alloc_file */
508 
509 	return NULL;
510 }
511 EXPORT_SYMBOL(sock_from_file);
512 
513 /**
514  *	sockfd_lookup - Go from a file number to its socket slot
515  *	@fd: file handle
516  *	@err: pointer to an error code return
517  *
518  *	The file handle passed in is locked and the socket it is bound
519  *	to is returned. If an error occurs the err pointer is overwritten
520  *	with a negative errno code and NULL is returned. The function checks
521  *	for both invalid handles and passing a handle which is not a socket.
522  *
523  *	On a success the socket object pointer is returned.
524  */
525 
sockfd_lookup(int fd,int * err)526 struct socket *sockfd_lookup(int fd, int *err)
527 {
528 	struct file *file;
529 	struct socket *sock;
530 
531 	file = fget(fd);
532 	if (!file) {
533 		*err = -EBADF;
534 		return NULL;
535 	}
536 
537 	sock = sock_from_file(file);
538 	if (!sock) {
539 		*err = -ENOTSOCK;
540 		fput(file);
541 	}
542 	return sock;
543 }
544 EXPORT_SYMBOL(sockfd_lookup);
545 
sockfd_lookup_light(int fd,int * err,int * fput_needed)546 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
547 {
548 	struct fd f = fdget(fd);
549 	struct socket *sock;
550 
551 	*err = -EBADF;
552 	if (f.file) {
553 		sock = sock_from_file(f.file);
554 		if (likely(sock)) {
555 			*fput_needed = f.flags & FDPUT_FPUT;
556 			return sock;
557 		}
558 		*err = -ENOTSOCK;
559 		fdput(f);
560 	}
561 	return NULL;
562 }
563 
sockfs_listxattr(struct dentry * dentry,char * buffer,size_t size)564 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
565 				size_t size)
566 {
567 	ssize_t len;
568 	ssize_t used = 0;
569 
570 	len = security_inode_listsecurity(d_inode(dentry), buffer, size);
571 	if (len < 0)
572 		return len;
573 	used += len;
574 	if (buffer) {
575 		if (size < used)
576 			return -ERANGE;
577 		buffer += len;
578 	}
579 
580 	len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
581 	used += len;
582 	if (buffer) {
583 		if (size < used)
584 			return -ERANGE;
585 		memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
586 		buffer += len;
587 	}
588 
589 	return used;
590 }
591 
sockfs_setattr(struct user_namespace * mnt_userns,struct dentry * dentry,struct iattr * iattr)592 static int sockfs_setattr(struct user_namespace *mnt_userns,
593 			  struct dentry *dentry, struct iattr *iattr)
594 {
595 	int err = simple_setattr(&init_user_ns, dentry, iattr);
596 
597 	if (!err && (iattr->ia_valid & ATTR_UID)) {
598 		struct socket *sock = SOCKET_I(d_inode(dentry));
599 
600 		if (sock->sk)
601 			sock->sk->sk_uid = iattr->ia_uid;
602 		else
603 			err = -ENOENT;
604 	}
605 
606 	return err;
607 }
608 
609 static const struct inode_operations sockfs_inode_ops = {
610 	.listxattr = sockfs_listxattr,
611 	.setattr = sockfs_setattr,
612 };
613 
614 /**
615  *	sock_alloc - allocate a socket
616  *
617  *	Allocate a new inode and socket object. The two are bound together
618  *	and initialised. The socket is then returned. If we are out of inodes
619  *	NULL is returned. This functions uses GFP_KERNEL internally.
620  */
621 
sock_alloc(void)622 struct socket *sock_alloc(void)
623 {
624 	struct inode *inode;
625 	struct socket *sock;
626 
627 	inode = new_inode_pseudo(sock_mnt->mnt_sb);
628 	if (!inode)
629 		return NULL;
630 
631 	sock = SOCKET_I(inode);
632 
633 	inode->i_ino = get_next_ino();
634 	inode->i_mode = S_IFSOCK | S_IRWXUGO;
635 	inode->i_uid = current_fsuid();
636 	inode->i_gid = current_fsgid();
637 	inode->i_op = &sockfs_inode_ops;
638 
639 	return sock;
640 }
641 EXPORT_SYMBOL(sock_alloc);
642 
__sock_release(struct socket * sock,struct inode * inode)643 static void __sock_release(struct socket *sock, struct inode *inode)
644 {
645 	if (sock->ops) {
646 		struct module *owner = sock->ops->owner;
647 
648 		if (inode)
649 			inode_lock(inode);
650 		sock->ops->release(sock);
651 		sock->sk = NULL;
652 		if (inode)
653 			inode_unlock(inode);
654 		sock->ops = NULL;
655 		module_put(owner);
656 	}
657 
658 	if (sock->wq.fasync_list)
659 		pr_err("%s: fasync list not empty!\n", __func__);
660 
661 	if (!sock->file) {
662 		iput(SOCK_INODE(sock));
663 		return;
664 	}
665 	sock->file = NULL;
666 }
667 
668 /**
669  *	sock_release - close a socket
670  *	@sock: socket to close
671  *
672  *	The socket is released from the protocol stack if it has a release
673  *	callback, and the inode is then released if the socket is bound to
674  *	an inode not a file.
675  */
sock_release(struct socket * sock)676 void sock_release(struct socket *sock)
677 {
678 	__sock_release(sock, NULL);
679 }
680 EXPORT_SYMBOL(sock_release);
681 
__sock_tx_timestamp(__u16 tsflags,__u8 * tx_flags)682 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
683 {
684 	u8 flags = *tx_flags;
685 
686 	if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
687 		flags |= SKBTX_HW_TSTAMP;
688 
689 		/* PTP hardware clocks can provide a free running cycle counter
690 		 * as a time base for virtual clocks. Tell driver to use the
691 		 * free running cycle counter for timestamp if socket is bound
692 		 * to virtual clock.
693 		 */
694 		if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
695 			flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
696 	}
697 
698 	if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
699 		flags |= SKBTX_SW_TSTAMP;
700 
701 	if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
702 		flags |= SKBTX_SCHED_TSTAMP;
703 
704 	*tx_flags = flags;
705 }
706 EXPORT_SYMBOL(__sock_tx_timestamp);
707 
708 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
709 					   size_t));
710 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
711 					    size_t));
sock_sendmsg_nosec(struct socket * sock,struct msghdr * msg)712 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
713 {
714 	int ret = INDIRECT_CALL_INET(sock->ops->sendmsg, inet6_sendmsg,
715 				     inet_sendmsg, sock, msg,
716 				     msg_data_left(msg));
717 	BUG_ON(ret == -EIOCBQUEUED);
718 	return ret;
719 }
720 
721 /**
722  *	sock_sendmsg - send a message through @sock
723  *	@sock: socket
724  *	@msg: message to send
725  *
726  *	Sends @msg through @sock, passing through LSM.
727  *	Returns the number of bytes sent, or an error code.
728  */
sock_sendmsg(struct socket * sock,struct msghdr * msg)729 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
730 {
731 	int err = security_socket_sendmsg(sock, msg,
732 					  msg_data_left(msg));
733 
734 	return err ?: sock_sendmsg_nosec(sock, msg);
735 }
736 EXPORT_SYMBOL(sock_sendmsg);
737 
738 /**
739  *	kernel_sendmsg - send a message through @sock (kernel-space)
740  *	@sock: socket
741  *	@msg: message header
742  *	@vec: kernel vec
743  *	@num: vec array length
744  *	@size: total message data size
745  *
746  *	Builds the message data with @vec and sends it through @sock.
747  *	Returns the number of bytes sent, or an error code.
748  */
749 
kernel_sendmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size)750 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
751 		   struct kvec *vec, size_t num, size_t size)
752 {
753 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
754 	return sock_sendmsg(sock, msg);
755 }
756 EXPORT_SYMBOL(kernel_sendmsg);
757 
758 /**
759  *	kernel_sendmsg_locked - send a message through @sock (kernel-space)
760  *	@sk: sock
761  *	@msg: message header
762  *	@vec: output s/g array
763  *	@num: output s/g array length
764  *	@size: total message data size
765  *
766  *	Builds the message data with @vec and sends it through @sock.
767  *	Returns the number of bytes sent, or an error code.
768  *	Caller must hold @sk.
769  */
770 
kernel_sendmsg_locked(struct sock * sk,struct msghdr * msg,struct kvec * vec,size_t num,size_t size)771 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
772 			  struct kvec *vec, size_t num, size_t size)
773 {
774 	struct socket *sock = sk->sk_socket;
775 
776 	if (!sock->ops->sendmsg_locked)
777 		return sock_no_sendmsg_locked(sk, msg, size);
778 
779 	iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
780 
781 	return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
782 }
783 EXPORT_SYMBOL(kernel_sendmsg_locked);
784 
skb_is_err_queue(const struct sk_buff * skb)785 static bool skb_is_err_queue(const struct sk_buff *skb)
786 {
787 	/* pkt_type of skbs enqueued on the error queue are set to
788 	 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
789 	 * in recvmsg, since skbs received on a local socket will never
790 	 * have a pkt_type of PACKET_OUTGOING.
791 	 */
792 	return skb->pkt_type == PACKET_OUTGOING;
793 }
794 
795 /* On transmit, software and hardware timestamps are returned independently.
796  * As the two skb clones share the hardware timestamp, which may be updated
797  * before the software timestamp is received, a hardware TX timestamp may be
798  * returned only if there is no software TX timestamp. Ignore false software
799  * timestamps, which may be made in the __sock_recv_timestamp() call when the
800  * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
801  * hardware timestamp.
802  */
skb_is_swtx_tstamp(const struct sk_buff * skb,int false_tstamp)803 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
804 {
805 	return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
806 }
807 
get_timestamp(struct sock * sk,struct sk_buff * skb,int * if_index)808 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
809 {
810 	bool cycles = sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC;
811 	struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
812 	struct net_device *orig_dev;
813 	ktime_t hwtstamp;
814 
815 	rcu_read_lock();
816 	orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
817 	if (orig_dev) {
818 		*if_index = orig_dev->ifindex;
819 		hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
820 	} else {
821 		hwtstamp = shhwtstamps->hwtstamp;
822 	}
823 	rcu_read_unlock();
824 
825 	return hwtstamp;
826 }
827 
put_ts_pktinfo(struct msghdr * msg,struct sk_buff * skb,int if_index)828 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
829 			   int if_index)
830 {
831 	struct scm_ts_pktinfo ts_pktinfo;
832 	struct net_device *orig_dev;
833 
834 	if (!skb_mac_header_was_set(skb))
835 		return;
836 
837 	memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
838 
839 	if (!if_index) {
840 		rcu_read_lock();
841 		orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
842 		if (orig_dev)
843 			if_index = orig_dev->ifindex;
844 		rcu_read_unlock();
845 	}
846 	ts_pktinfo.if_index = if_index;
847 
848 	ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
849 	put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
850 		 sizeof(ts_pktinfo), &ts_pktinfo);
851 }
852 
853 /*
854  * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
855  */
__sock_recv_timestamp(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)856 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
857 	struct sk_buff *skb)
858 {
859 	int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
860 	int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
861 	struct scm_timestamping_internal tss;
862 
863 	int empty = 1, false_tstamp = 0;
864 	struct skb_shared_hwtstamps *shhwtstamps =
865 		skb_hwtstamps(skb);
866 	int if_index;
867 	ktime_t hwtstamp;
868 
869 	/* Race occurred between timestamp enabling and packet
870 	   receiving.  Fill in the current time for now. */
871 	if (need_software_tstamp && skb->tstamp == 0) {
872 		__net_timestamp(skb);
873 		false_tstamp = 1;
874 	}
875 
876 	if (need_software_tstamp) {
877 		if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
878 			if (new_tstamp) {
879 				struct __kernel_sock_timeval tv;
880 
881 				skb_get_new_timestamp(skb, &tv);
882 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
883 					 sizeof(tv), &tv);
884 			} else {
885 				struct __kernel_old_timeval tv;
886 
887 				skb_get_timestamp(skb, &tv);
888 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
889 					 sizeof(tv), &tv);
890 			}
891 		} else {
892 			if (new_tstamp) {
893 				struct __kernel_timespec ts;
894 
895 				skb_get_new_timestampns(skb, &ts);
896 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
897 					 sizeof(ts), &ts);
898 			} else {
899 				struct __kernel_old_timespec ts;
900 
901 				skb_get_timestampns(skb, &ts);
902 				put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
903 					 sizeof(ts), &ts);
904 			}
905 		}
906 	}
907 
908 	memset(&tss, 0, sizeof(tss));
909 	if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
910 	    ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
911 		empty = 0;
912 	if (shhwtstamps &&
913 	    (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
914 	    !skb_is_swtx_tstamp(skb, false_tstamp)) {
915 		if_index = 0;
916 		if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
917 			hwtstamp = get_timestamp(sk, skb, &if_index);
918 		else
919 			hwtstamp = shhwtstamps->hwtstamp;
920 
921 		if (sk->sk_tsflags & SOF_TIMESTAMPING_BIND_PHC)
922 			hwtstamp = ptp_convert_timestamp(&hwtstamp,
923 							 sk->sk_bind_phc);
924 
925 		if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
926 			empty = 0;
927 
928 			if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
929 			    !skb_is_err_queue(skb))
930 				put_ts_pktinfo(msg, skb, if_index);
931 		}
932 	}
933 	if (!empty) {
934 		if (sock_flag(sk, SOCK_TSTAMP_NEW))
935 			put_cmsg_scm_timestamping64(msg, &tss);
936 		else
937 			put_cmsg_scm_timestamping(msg, &tss);
938 
939 		if (skb_is_err_queue(skb) && skb->len &&
940 		    SKB_EXT_ERR(skb)->opt_stats)
941 			put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
942 				 skb->len, skb->data);
943 	}
944 }
945 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
946 
__sock_recv_wifi_status(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)947 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
948 	struct sk_buff *skb)
949 {
950 	int ack;
951 
952 	if (!sock_flag(sk, SOCK_WIFI_STATUS))
953 		return;
954 	if (!skb->wifi_acked_valid)
955 		return;
956 
957 	ack = skb->wifi_acked;
958 
959 	put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
960 }
961 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
962 
sock_recv_drops(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)963 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
964 				   struct sk_buff *skb)
965 {
966 	if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
967 		put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
968 			sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
969 }
970 
sock_recv_mark(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)971 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
972 			   struct sk_buff *skb)
973 {
974 	if (sock_flag(sk, SOCK_RCVMARK) && skb)
975 		put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32),
976 			 &skb->mark);
977 }
978 
__sock_recv_cmsgs(struct msghdr * msg,struct sock * sk,struct sk_buff * skb)979 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
980 		       struct sk_buff *skb)
981 {
982 	sock_recv_timestamp(msg, sk, skb);
983 	sock_recv_drops(msg, sk, skb);
984 	sock_recv_mark(msg, sk, skb);
985 }
986 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
987 
988 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
989 					   size_t, int));
990 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
991 					    size_t, int));
sock_recvmsg_nosec(struct socket * sock,struct msghdr * msg,int flags)992 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
993 				     int flags)
994 {
995 	return INDIRECT_CALL_INET(sock->ops->recvmsg, inet6_recvmsg,
996 				  inet_recvmsg, sock, msg, msg_data_left(msg),
997 				  flags);
998 }
999 
1000 /**
1001  *	sock_recvmsg - receive a message from @sock
1002  *	@sock: socket
1003  *	@msg: message to receive
1004  *	@flags: message flags
1005  *
1006  *	Receives @msg from @sock, passing through LSM. Returns the total number
1007  *	of bytes received, or an error.
1008  */
sock_recvmsg(struct socket * sock,struct msghdr * msg,int flags)1009 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1010 {
1011 	int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1012 
1013 	return err ?: sock_recvmsg_nosec(sock, msg, flags);
1014 }
1015 EXPORT_SYMBOL(sock_recvmsg);
1016 
1017 /**
1018  *	kernel_recvmsg - Receive a message from a socket (kernel space)
1019  *	@sock: The socket to receive the message from
1020  *	@msg: Received message
1021  *	@vec: Input s/g array for message data
1022  *	@num: Size of input s/g array
1023  *	@size: Number of bytes to read
1024  *	@flags: Message flags (MSG_DONTWAIT, etc...)
1025  *
1026  *	On return the msg structure contains the scatter/gather array passed in the
1027  *	vec argument. The array is modified so that it consists of the unfilled
1028  *	portion of the original array.
1029  *
1030  *	The returned value is the total number of bytes received, or an error.
1031  */
1032 
kernel_recvmsg(struct socket * sock,struct msghdr * msg,struct kvec * vec,size_t num,size_t size,int flags)1033 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1034 		   struct kvec *vec, size_t num, size_t size, int flags)
1035 {
1036 	msg->msg_control_is_user = false;
1037 	iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
1038 	return sock_recvmsg(sock, msg, flags);
1039 }
1040 EXPORT_SYMBOL(kernel_recvmsg);
1041 
sock_sendpage(struct file * file,struct page * page,int offset,size_t size,loff_t * ppos,int more)1042 static ssize_t sock_sendpage(struct file *file, struct page *page,
1043 			     int offset, size_t size, loff_t *ppos, int more)
1044 {
1045 	struct socket *sock;
1046 	int flags;
1047 
1048 	sock = file->private_data;
1049 
1050 	flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
1051 	/* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
1052 	flags |= more;
1053 
1054 	return kernel_sendpage(sock, page, offset, size, flags);
1055 }
1056 
sock_splice_read(struct file * file,loff_t * ppos,struct pipe_inode_info * pipe,size_t len,unsigned int flags)1057 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1058 				struct pipe_inode_info *pipe, size_t len,
1059 				unsigned int flags)
1060 {
1061 	struct socket *sock = file->private_data;
1062 
1063 	if (unlikely(!sock->ops->splice_read))
1064 		return generic_file_splice_read(file, ppos, pipe, len, flags);
1065 
1066 	return sock->ops->splice_read(sock, ppos, pipe, len, flags);
1067 }
1068 
sock_read_iter(struct kiocb * iocb,struct iov_iter * to)1069 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1070 {
1071 	struct file *file = iocb->ki_filp;
1072 	struct socket *sock = file->private_data;
1073 	struct msghdr msg = {.msg_iter = *to,
1074 			     .msg_iocb = iocb};
1075 	ssize_t res;
1076 
1077 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1078 		msg.msg_flags = MSG_DONTWAIT;
1079 
1080 	if (iocb->ki_pos != 0)
1081 		return -ESPIPE;
1082 
1083 	if (!iov_iter_count(to))	/* Match SYS5 behaviour */
1084 		return 0;
1085 
1086 	res = sock_recvmsg(sock, &msg, msg.msg_flags);
1087 	*to = msg.msg_iter;
1088 	return res;
1089 }
1090 
sock_write_iter(struct kiocb * iocb,struct iov_iter * from)1091 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1092 {
1093 	struct file *file = iocb->ki_filp;
1094 	struct socket *sock = file->private_data;
1095 	struct msghdr msg = {.msg_iter = *from,
1096 			     .msg_iocb = iocb};
1097 	ssize_t res;
1098 
1099 	if (iocb->ki_pos != 0)
1100 		return -ESPIPE;
1101 
1102 	if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1103 		msg.msg_flags = MSG_DONTWAIT;
1104 
1105 	if (sock->type == SOCK_SEQPACKET)
1106 		msg.msg_flags |= MSG_EOR;
1107 
1108 	res = sock_sendmsg(sock, &msg);
1109 	*from = msg.msg_iter;
1110 	return res;
1111 }
1112 
1113 /*
1114  * Atomic setting of ioctl hooks to avoid race
1115  * with module unload.
1116  */
1117 
1118 static DEFINE_MUTEX(br_ioctl_mutex);
1119 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1120 			    unsigned int cmd, struct ifreq *ifr,
1121 			    void __user *uarg);
1122 
brioctl_set(int (* hook)(struct net * net,struct net_bridge * br,unsigned int cmd,struct ifreq * ifr,void __user * uarg))1123 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1124 			     unsigned int cmd, struct ifreq *ifr,
1125 			     void __user *uarg))
1126 {
1127 	mutex_lock(&br_ioctl_mutex);
1128 	br_ioctl_hook = hook;
1129 	mutex_unlock(&br_ioctl_mutex);
1130 }
1131 EXPORT_SYMBOL(brioctl_set);
1132 
br_ioctl_call(struct net * net,struct net_bridge * br,unsigned int cmd,struct ifreq * ifr,void __user * uarg)1133 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1134 		  struct ifreq *ifr, void __user *uarg)
1135 {
1136 	int err = -ENOPKG;
1137 
1138 	if (!br_ioctl_hook)
1139 		request_module("bridge");
1140 
1141 	mutex_lock(&br_ioctl_mutex);
1142 	if (br_ioctl_hook)
1143 		err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1144 	mutex_unlock(&br_ioctl_mutex);
1145 
1146 	return err;
1147 }
1148 
1149 static DEFINE_MUTEX(vlan_ioctl_mutex);
1150 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1151 
vlan_ioctl_set(int (* hook)(struct net *,void __user *))1152 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1153 {
1154 	mutex_lock(&vlan_ioctl_mutex);
1155 	vlan_ioctl_hook = hook;
1156 	mutex_unlock(&vlan_ioctl_mutex);
1157 }
1158 EXPORT_SYMBOL(vlan_ioctl_set);
1159 
sock_do_ioctl(struct net * net,struct socket * sock,unsigned int cmd,unsigned long arg)1160 static long sock_do_ioctl(struct net *net, struct socket *sock,
1161 			  unsigned int cmd, unsigned long arg)
1162 {
1163 	struct ifreq ifr;
1164 	bool need_copyout;
1165 	int err;
1166 	void __user *argp = (void __user *)arg;
1167 	void __user *data;
1168 
1169 	err = sock->ops->ioctl(sock, cmd, arg);
1170 
1171 	/*
1172 	 * If this ioctl is unknown try to hand it down
1173 	 * to the NIC driver.
1174 	 */
1175 	if (err != -ENOIOCTLCMD)
1176 		return err;
1177 
1178 	if (!is_socket_ioctl_cmd(cmd))
1179 		return -ENOTTY;
1180 
1181 	if (get_user_ifreq(&ifr, &data, argp))
1182 		return -EFAULT;
1183 	err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1184 	if (!err && need_copyout)
1185 		if (put_user_ifreq(&ifr, argp))
1186 			return -EFAULT;
1187 
1188 	return err;
1189 }
1190 
1191 /*
1192  *	With an ioctl, arg may well be a user mode pointer, but we don't know
1193  *	what to do with it - that's up to the protocol still.
1194  */
1195 
sock_ioctl(struct file * file,unsigned cmd,unsigned long arg)1196 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1197 {
1198 	struct socket *sock;
1199 	struct sock *sk;
1200 	void __user *argp = (void __user *)arg;
1201 	int pid, err;
1202 	struct net *net;
1203 
1204 	sock = file->private_data;
1205 	sk = sock->sk;
1206 	net = sock_net(sk);
1207 	if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1208 		struct ifreq ifr;
1209 		void __user *data;
1210 		bool need_copyout;
1211 		if (get_user_ifreq(&ifr, &data, argp))
1212 			return -EFAULT;
1213 		err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1214 		if (!err && need_copyout)
1215 			if (put_user_ifreq(&ifr, argp))
1216 				return -EFAULT;
1217 	} else
1218 #ifdef CONFIG_WEXT_CORE
1219 	if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1220 		err = wext_handle_ioctl(net, cmd, argp);
1221 	} else
1222 #endif
1223 		switch (cmd) {
1224 		case FIOSETOWN:
1225 		case SIOCSPGRP:
1226 			err = -EFAULT;
1227 			if (get_user(pid, (int __user *)argp))
1228 				break;
1229 			err = f_setown(sock->file, pid, 1);
1230 			break;
1231 		case FIOGETOWN:
1232 		case SIOCGPGRP:
1233 			err = put_user(f_getown(sock->file),
1234 				       (int __user *)argp);
1235 			break;
1236 		case SIOCGIFBR:
1237 		case SIOCSIFBR:
1238 		case SIOCBRADDBR:
1239 		case SIOCBRDELBR:
1240 			err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1241 			break;
1242 		case SIOCGIFVLAN:
1243 		case SIOCSIFVLAN:
1244 			err = -ENOPKG;
1245 			if (!vlan_ioctl_hook)
1246 				request_module("8021q");
1247 
1248 			mutex_lock(&vlan_ioctl_mutex);
1249 			if (vlan_ioctl_hook)
1250 				err = vlan_ioctl_hook(net, argp);
1251 			mutex_unlock(&vlan_ioctl_mutex);
1252 			break;
1253 		case SIOCGSKNS:
1254 			err = -EPERM;
1255 			if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1256 				break;
1257 
1258 			err = open_related_ns(&net->ns, get_net_ns);
1259 			break;
1260 		case SIOCGSTAMP_OLD:
1261 		case SIOCGSTAMPNS_OLD:
1262 			if (!sock->ops->gettstamp) {
1263 				err = -ENOIOCTLCMD;
1264 				break;
1265 			}
1266 			err = sock->ops->gettstamp(sock, argp,
1267 						   cmd == SIOCGSTAMP_OLD,
1268 						   !IS_ENABLED(CONFIG_64BIT));
1269 			break;
1270 		case SIOCGSTAMP_NEW:
1271 		case SIOCGSTAMPNS_NEW:
1272 			if (!sock->ops->gettstamp) {
1273 				err = -ENOIOCTLCMD;
1274 				break;
1275 			}
1276 			err = sock->ops->gettstamp(sock, argp,
1277 						   cmd == SIOCGSTAMP_NEW,
1278 						   false);
1279 			break;
1280 
1281 		case SIOCGIFCONF:
1282 			err = dev_ifconf(net, argp);
1283 			break;
1284 
1285 		default:
1286 			err = sock_do_ioctl(net, sock, cmd, arg);
1287 			break;
1288 		}
1289 	return err;
1290 }
1291 
1292 /**
1293  *	sock_create_lite - creates a socket
1294  *	@family: protocol family (AF_INET, ...)
1295  *	@type: communication type (SOCK_STREAM, ...)
1296  *	@protocol: protocol (0, ...)
1297  *	@res: new socket
1298  *
1299  *	Creates a new socket and assigns it to @res, passing through LSM.
1300  *	The new socket initialization is not complete, see kernel_accept().
1301  *	Returns 0 or an error. On failure @res is set to %NULL.
1302  *	This function internally uses GFP_KERNEL.
1303  */
1304 
sock_create_lite(int family,int type,int protocol,struct socket ** res)1305 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1306 {
1307 	int err;
1308 	struct socket *sock = NULL;
1309 
1310 	err = security_socket_create(family, type, protocol, 1);
1311 	if (err)
1312 		goto out;
1313 
1314 	sock = sock_alloc();
1315 	if (!sock) {
1316 		err = -ENOMEM;
1317 		goto out;
1318 	}
1319 
1320 	sock->type = type;
1321 	err = security_socket_post_create(sock, family, type, protocol, 1);
1322 	if (err)
1323 		goto out_release;
1324 
1325 out:
1326 	*res = sock;
1327 	return err;
1328 out_release:
1329 	sock_release(sock);
1330 	sock = NULL;
1331 	goto out;
1332 }
1333 EXPORT_SYMBOL(sock_create_lite);
1334 
1335 /* No kernel lock held - perfect */
sock_poll(struct file * file,poll_table * wait)1336 static __poll_t sock_poll(struct file *file, poll_table *wait)
1337 {
1338 	struct socket *sock = file->private_data;
1339 	__poll_t events = poll_requested_events(wait), flag = 0;
1340 
1341 	if (!sock->ops->poll)
1342 		return 0;
1343 
1344 	if (sk_can_busy_loop(sock->sk)) {
1345 		/* poll once if requested by the syscall */
1346 		if (events & POLL_BUSY_LOOP)
1347 			sk_busy_loop(sock->sk, 1);
1348 
1349 		/* if this socket can poll_ll, tell the system call */
1350 		flag = POLL_BUSY_LOOP;
1351 	}
1352 
1353 	return sock->ops->poll(file, sock, wait) | flag;
1354 }
1355 
sock_mmap(struct file * file,struct vm_area_struct * vma)1356 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1357 {
1358 	struct socket *sock = file->private_data;
1359 
1360 	return sock->ops->mmap(file, sock, vma);
1361 }
1362 
sock_close(struct inode * inode,struct file * filp)1363 static int sock_close(struct inode *inode, struct file *filp)
1364 {
1365 	__sock_release(SOCKET_I(inode), inode);
1366 	return 0;
1367 }
1368 
1369 /*
1370  *	Update the socket async list
1371  *
1372  *	Fasync_list locking strategy.
1373  *
1374  *	1. fasync_list is modified only under process context socket lock
1375  *	   i.e. under semaphore.
1376  *	2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1377  *	   or under socket lock
1378  */
1379 
sock_fasync(int fd,struct file * filp,int on)1380 static int sock_fasync(int fd, struct file *filp, int on)
1381 {
1382 	struct socket *sock = filp->private_data;
1383 	struct sock *sk = sock->sk;
1384 	struct socket_wq *wq = &sock->wq;
1385 
1386 	if (sk == NULL)
1387 		return -EINVAL;
1388 
1389 	lock_sock(sk);
1390 	fasync_helper(fd, filp, on, &wq->fasync_list);
1391 
1392 	if (!wq->fasync_list)
1393 		sock_reset_flag(sk, SOCK_FASYNC);
1394 	else
1395 		sock_set_flag(sk, SOCK_FASYNC);
1396 
1397 	release_sock(sk);
1398 	return 0;
1399 }
1400 
1401 /* This function may be called only under rcu_lock */
1402 
sock_wake_async(struct socket_wq * wq,int how,int band)1403 int sock_wake_async(struct socket_wq *wq, int how, int band)
1404 {
1405 	if (!wq || !wq->fasync_list)
1406 		return -1;
1407 
1408 	switch (how) {
1409 	case SOCK_WAKE_WAITD:
1410 		if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1411 			break;
1412 		goto call_kill;
1413 	case SOCK_WAKE_SPACE:
1414 		if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1415 			break;
1416 		fallthrough;
1417 	case SOCK_WAKE_IO:
1418 call_kill:
1419 		kill_fasync(&wq->fasync_list, SIGIO, band);
1420 		break;
1421 	case SOCK_WAKE_URG:
1422 		kill_fasync(&wq->fasync_list, SIGURG, band);
1423 	}
1424 
1425 	return 0;
1426 }
1427 EXPORT_SYMBOL(sock_wake_async);
1428 
1429 /**
1430  *	__sock_create - creates a socket
1431  *	@net: net namespace
1432  *	@family: protocol family (AF_INET, ...)
1433  *	@type: communication type (SOCK_STREAM, ...)
1434  *	@protocol: protocol (0, ...)
1435  *	@res: new socket
1436  *	@kern: boolean for kernel space sockets
1437  *
1438  *	Creates a new socket and assigns it to @res, passing through LSM.
1439  *	Returns 0 or an error. On failure @res is set to %NULL. @kern must
1440  *	be set to true if the socket resides in kernel space.
1441  *	This function internally uses GFP_KERNEL.
1442  */
1443 
__sock_create(struct net * net,int family,int type,int protocol,struct socket ** res,int kern)1444 int __sock_create(struct net *net, int family, int type, int protocol,
1445 			 struct socket **res, int kern)
1446 {
1447 	int err;
1448 	struct socket *sock;
1449 	const struct net_proto_family *pf;
1450 
1451 	/*
1452 	 *      Check protocol is in range
1453 	 */
1454 	if (family < 0 || family >= NPROTO)
1455 		return -EAFNOSUPPORT;
1456 	if (type < 0 || type >= SOCK_MAX)
1457 		return -EINVAL;
1458 
1459 	/* Compatibility.
1460 
1461 	   This uglymoron is moved from INET layer to here to avoid
1462 	   deadlock in module load.
1463 	 */
1464 	if (family == PF_INET && type == SOCK_PACKET) {
1465 		pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1466 			     current->comm);
1467 		family = PF_PACKET;
1468 	}
1469 
1470 	err = security_socket_create(family, type, protocol, kern);
1471 	if (err)
1472 		return err;
1473 
1474 	/*
1475 	 *	Allocate the socket and allow the family to set things up. if
1476 	 *	the protocol is 0, the family is instructed to select an appropriate
1477 	 *	default.
1478 	 */
1479 	sock = sock_alloc();
1480 	if (!sock) {
1481 		net_warn_ratelimited("socket: no more sockets\n");
1482 		return -ENFILE;	/* Not exactly a match, but its the
1483 				   closest posix thing */
1484 	}
1485 
1486 	sock->type = type;
1487 
1488 #ifdef CONFIG_MODULES
1489 	/* Attempt to load a protocol module if the find failed.
1490 	 *
1491 	 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1492 	 * requested real, full-featured networking support upon configuration.
1493 	 * Otherwise module support will break!
1494 	 */
1495 	if (rcu_access_pointer(net_families[family]) == NULL)
1496 		request_module("net-pf-%d", family);
1497 #endif
1498 
1499 	rcu_read_lock();
1500 	pf = rcu_dereference(net_families[family]);
1501 	err = -EAFNOSUPPORT;
1502 	if (!pf)
1503 		goto out_release;
1504 
1505 	/*
1506 	 * We will call the ->create function, that possibly is in a loadable
1507 	 * module, so we have to bump that loadable module refcnt first.
1508 	 */
1509 	if (!try_module_get(pf->owner))
1510 		goto out_release;
1511 
1512 	/* Now protected by module ref count */
1513 	rcu_read_unlock();
1514 
1515 	err = pf->create(net, sock, protocol, kern);
1516 	if (err < 0)
1517 		goto out_module_put;
1518 
1519 	/*
1520 	 * Now to bump the refcnt of the [loadable] module that owns this
1521 	 * socket at sock_release time we decrement its refcnt.
1522 	 */
1523 	if (!try_module_get(sock->ops->owner))
1524 		goto out_module_busy;
1525 
1526 	/*
1527 	 * Now that we're done with the ->create function, the [loadable]
1528 	 * module can have its refcnt decremented
1529 	 */
1530 	module_put(pf->owner);
1531 	err = security_socket_post_create(sock, family, type, protocol, kern);
1532 	if (err)
1533 		goto out_sock_release;
1534 	*res = sock;
1535 
1536 	return 0;
1537 
1538 out_module_busy:
1539 	err = -EAFNOSUPPORT;
1540 out_module_put:
1541 	sock->ops = NULL;
1542 	module_put(pf->owner);
1543 out_sock_release:
1544 	sock_release(sock);
1545 	return err;
1546 
1547 out_release:
1548 	rcu_read_unlock();
1549 	goto out_sock_release;
1550 }
1551 EXPORT_SYMBOL(__sock_create);
1552 
1553 /**
1554  *	sock_create - creates a socket
1555  *	@family: protocol family (AF_INET, ...)
1556  *	@type: communication type (SOCK_STREAM, ...)
1557  *	@protocol: protocol (0, ...)
1558  *	@res: new socket
1559  *
1560  *	A wrapper around __sock_create().
1561  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1562  */
1563 
sock_create(int family,int type,int protocol,struct socket ** res)1564 int sock_create(int family, int type, int protocol, struct socket **res)
1565 {
1566 	return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1567 }
1568 EXPORT_SYMBOL(sock_create);
1569 
1570 /**
1571  *	sock_create_kern - creates a socket (kernel space)
1572  *	@net: net namespace
1573  *	@family: protocol family (AF_INET, ...)
1574  *	@type: communication type (SOCK_STREAM, ...)
1575  *	@protocol: protocol (0, ...)
1576  *	@res: new socket
1577  *
1578  *	A wrapper around __sock_create().
1579  *	Returns 0 or an error. This function internally uses GFP_KERNEL.
1580  */
1581 
sock_create_kern(struct net * net,int family,int type,int protocol,struct socket ** res)1582 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1583 {
1584 	return __sock_create(net, family, type, protocol, res, 1);
1585 }
1586 EXPORT_SYMBOL(sock_create_kern);
1587 
__sys_socket_create(int family,int type,int protocol)1588 static struct socket *__sys_socket_create(int family, int type, int protocol)
1589 {
1590 	struct socket *sock;
1591 	int retval;
1592 
1593 	/* Check the SOCK_* constants for consistency.  */
1594 	BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1595 	BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1596 	BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1597 	BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1598 
1599 	if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1600 		return ERR_PTR(-EINVAL);
1601 	type &= SOCK_TYPE_MASK;
1602 
1603 	retval = sock_create(family, type, protocol, &sock);
1604 	if (retval < 0)
1605 		return ERR_PTR(retval);
1606 
1607 	return sock;
1608 }
1609 
__sys_socket_file(int family,int type,int protocol)1610 struct file *__sys_socket_file(int family, int type, int protocol)
1611 {
1612 	struct socket *sock;
1613 	struct file *file;
1614 	int flags;
1615 
1616 	sock = __sys_socket_create(family, type, protocol);
1617 	if (IS_ERR(sock))
1618 		return ERR_CAST(sock);
1619 
1620 	flags = type & ~SOCK_TYPE_MASK;
1621 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1622 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1623 
1624 	file = sock_alloc_file(sock, flags, NULL);
1625 	if (IS_ERR(file))
1626 		sock_release(sock);
1627 
1628 	return file;
1629 }
1630 
__sys_socket(int family,int type,int protocol)1631 int __sys_socket(int family, int type, int protocol)
1632 {
1633 	struct socket *sock;
1634 	int flags;
1635 
1636 	sock = __sys_socket_create(family, type, protocol);
1637 	if (IS_ERR(sock))
1638 		return PTR_ERR(sock);
1639 
1640 	flags = type & ~SOCK_TYPE_MASK;
1641 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1642 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1643 
1644 	return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1645 }
1646 
SYSCALL_DEFINE3(socket,int,family,int,type,int,protocol)1647 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1648 {
1649 	return __sys_socket(family, type, protocol);
1650 }
1651 
1652 /*
1653  *	Create a pair of connected sockets.
1654  */
1655 
__sys_socketpair(int family,int type,int protocol,int __user * usockvec)1656 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1657 {
1658 	struct socket *sock1, *sock2;
1659 	int fd1, fd2, err;
1660 	struct file *newfile1, *newfile2;
1661 	int flags;
1662 
1663 	flags = type & ~SOCK_TYPE_MASK;
1664 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1665 		return -EINVAL;
1666 	type &= SOCK_TYPE_MASK;
1667 
1668 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1669 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1670 
1671 	/*
1672 	 * reserve descriptors and make sure we won't fail
1673 	 * to return them to userland.
1674 	 */
1675 	fd1 = get_unused_fd_flags(flags);
1676 	if (unlikely(fd1 < 0))
1677 		return fd1;
1678 
1679 	fd2 = get_unused_fd_flags(flags);
1680 	if (unlikely(fd2 < 0)) {
1681 		put_unused_fd(fd1);
1682 		return fd2;
1683 	}
1684 
1685 	err = put_user(fd1, &usockvec[0]);
1686 	if (err)
1687 		goto out;
1688 
1689 	err = put_user(fd2, &usockvec[1]);
1690 	if (err)
1691 		goto out;
1692 
1693 	/*
1694 	 * Obtain the first socket and check if the underlying protocol
1695 	 * supports the socketpair call.
1696 	 */
1697 
1698 	err = sock_create(family, type, protocol, &sock1);
1699 	if (unlikely(err < 0))
1700 		goto out;
1701 
1702 	err = sock_create(family, type, protocol, &sock2);
1703 	if (unlikely(err < 0)) {
1704 		sock_release(sock1);
1705 		goto out;
1706 	}
1707 
1708 	err = security_socket_socketpair(sock1, sock2);
1709 	if (unlikely(err)) {
1710 		sock_release(sock2);
1711 		sock_release(sock1);
1712 		goto out;
1713 	}
1714 
1715 	err = sock1->ops->socketpair(sock1, sock2);
1716 	if (unlikely(err < 0)) {
1717 		sock_release(sock2);
1718 		sock_release(sock1);
1719 		goto out;
1720 	}
1721 
1722 	newfile1 = sock_alloc_file(sock1, flags, NULL);
1723 	if (IS_ERR(newfile1)) {
1724 		err = PTR_ERR(newfile1);
1725 		sock_release(sock2);
1726 		goto out;
1727 	}
1728 
1729 	newfile2 = sock_alloc_file(sock2, flags, NULL);
1730 	if (IS_ERR(newfile2)) {
1731 		err = PTR_ERR(newfile2);
1732 		fput(newfile1);
1733 		goto out;
1734 	}
1735 
1736 	audit_fd_pair(fd1, fd2);
1737 
1738 	fd_install(fd1, newfile1);
1739 	fd_install(fd2, newfile2);
1740 	return 0;
1741 
1742 out:
1743 	put_unused_fd(fd2);
1744 	put_unused_fd(fd1);
1745 	return err;
1746 }
1747 
SYSCALL_DEFINE4(socketpair,int,family,int,type,int,protocol,int __user *,usockvec)1748 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1749 		int __user *, usockvec)
1750 {
1751 	return __sys_socketpair(family, type, protocol, usockvec);
1752 }
1753 
1754 /*
1755  *	Bind a name to a socket. Nothing much to do here since it's
1756  *	the protocol's responsibility to handle the local address.
1757  *
1758  *	We move the socket address to kernel space before we call
1759  *	the protocol layer (having also checked the address is ok).
1760  */
1761 
__sys_bind(int fd,struct sockaddr __user * umyaddr,int addrlen)1762 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1763 {
1764 	struct socket *sock;
1765 	struct sockaddr_storage address;
1766 	int err, fput_needed;
1767 
1768 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769 	if (sock) {
1770 		err = move_addr_to_kernel(umyaddr, addrlen, &address);
1771 		if (!err) {
1772 			err = security_socket_bind(sock,
1773 						   (struct sockaddr *)&address,
1774 						   addrlen);
1775 			if (!err)
1776 				err = sock->ops->bind(sock,
1777 						      (struct sockaddr *)
1778 						      &address, addrlen);
1779 		}
1780 		fput_light(sock->file, fput_needed);
1781 	}
1782 	return err;
1783 }
1784 
SYSCALL_DEFINE3(bind,int,fd,struct sockaddr __user *,umyaddr,int,addrlen)1785 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1786 {
1787 	return __sys_bind(fd, umyaddr, addrlen);
1788 }
1789 
1790 /*
1791  *	Perform a listen. Basically, we allow the protocol to do anything
1792  *	necessary for a listen, and if that works, we mark the socket as
1793  *	ready for listening.
1794  */
1795 
__sys_listen(int fd,int backlog)1796 int __sys_listen(int fd, int backlog)
1797 {
1798 	struct socket *sock;
1799 	int err, fput_needed;
1800 	int somaxconn;
1801 
1802 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
1803 	if (sock) {
1804 		somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1805 		if ((unsigned int)backlog > somaxconn)
1806 			backlog = somaxconn;
1807 
1808 		err = security_socket_listen(sock, backlog);
1809 		if (!err)
1810 			err = sock->ops->listen(sock, backlog);
1811 
1812 		fput_light(sock->file, fput_needed);
1813 	}
1814 	return err;
1815 }
1816 
SYSCALL_DEFINE2(listen,int,fd,int,backlog)1817 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1818 {
1819 	return __sys_listen(fd, backlog);
1820 }
1821 
do_accept(struct file * file,unsigned file_flags,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1822 struct file *do_accept(struct file *file, unsigned file_flags,
1823 		       struct sockaddr __user *upeer_sockaddr,
1824 		       int __user *upeer_addrlen, int flags)
1825 {
1826 	struct socket *sock, *newsock;
1827 	struct file *newfile;
1828 	int err, len;
1829 	struct sockaddr_storage address;
1830 
1831 	sock = sock_from_file(file);
1832 	if (!sock)
1833 		return ERR_PTR(-ENOTSOCK);
1834 
1835 	newsock = sock_alloc();
1836 	if (!newsock)
1837 		return ERR_PTR(-ENFILE);
1838 
1839 	newsock->type = sock->type;
1840 	newsock->ops = sock->ops;
1841 
1842 	/*
1843 	 * We don't need try_module_get here, as the listening socket (sock)
1844 	 * has the protocol module (sock->ops->owner) held.
1845 	 */
1846 	__module_get(newsock->ops->owner);
1847 
1848 	newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1849 	if (IS_ERR(newfile))
1850 		return newfile;
1851 
1852 	err = security_socket_accept(sock, newsock);
1853 	if (err)
1854 		goto out_fd;
1855 
1856 	err = sock->ops->accept(sock, newsock, sock->file->f_flags | file_flags,
1857 					false);
1858 	if (err < 0)
1859 		goto out_fd;
1860 
1861 	if (upeer_sockaddr) {
1862 		len = newsock->ops->getname(newsock,
1863 					(struct sockaddr *)&address, 2);
1864 		if (len < 0) {
1865 			err = -ECONNABORTED;
1866 			goto out_fd;
1867 		}
1868 		err = move_addr_to_user(&address,
1869 					len, upeer_sockaddr, upeer_addrlen);
1870 		if (err < 0)
1871 			goto out_fd;
1872 	}
1873 
1874 	/* File flags are not inherited via accept() unlike another OSes. */
1875 	return newfile;
1876 out_fd:
1877 	fput(newfile);
1878 	return ERR_PTR(err);
1879 }
1880 
__sys_accept4_file(struct file * file,unsigned file_flags,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags,unsigned long nofile)1881 int __sys_accept4_file(struct file *file, unsigned file_flags,
1882 		       struct sockaddr __user *upeer_sockaddr,
1883 		       int __user *upeer_addrlen, int flags,
1884 		       unsigned long nofile)
1885 {
1886 	struct file *newfile;
1887 	int newfd;
1888 
1889 	if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1890 		return -EINVAL;
1891 
1892 	if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1893 		flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1894 
1895 	newfd = __get_unused_fd_flags(flags, nofile);
1896 	if (unlikely(newfd < 0))
1897 		return newfd;
1898 
1899 	newfile = do_accept(file, file_flags, upeer_sockaddr, upeer_addrlen,
1900 			    flags);
1901 	if (IS_ERR(newfile)) {
1902 		put_unused_fd(newfd);
1903 		return PTR_ERR(newfile);
1904 	}
1905 	fd_install(newfd, newfile);
1906 	return newfd;
1907 }
1908 
1909 /*
1910  *	For accept, we attempt to create a new socket, set up the link
1911  *	with the client, wake up the client, then return the new
1912  *	connected fd. We collect the address of the connector in kernel
1913  *	space and move it to user at the very end. This is unclean because
1914  *	we open the socket then return an error.
1915  *
1916  *	1003.1g adds the ability to recvmsg() to query connection pending
1917  *	status to recvmsg. We need to add that support in a way thats
1918  *	clean when we restructure accept also.
1919  */
1920 
__sys_accept4(int fd,struct sockaddr __user * upeer_sockaddr,int __user * upeer_addrlen,int flags)1921 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1922 		  int __user *upeer_addrlen, int flags)
1923 {
1924 	int ret = -EBADF;
1925 	struct fd f;
1926 
1927 	f = fdget(fd);
1928 	if (f.file) {
1929 		ret = __sys_accept4_file(f.file, 0, upeer_sockaddr,
1930 						upeer_addrlen, flags,
1931 						rlimit(RLIMIT_NOFILE));
1932 		fdput(f);
1933 	}
1934 
1935 	return ret;
1936 }
1937 
SYSCALL_DEFINE4(accept4,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen,int,flags)1938 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1939 		int __user *, upeer_addrlen, int, flags)
1940 {
1941 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1942 }
1943 
SYSCALL_DEFINE3(accept,int,fd,struct sockaddr __user *,upeer_sockaddr,int __user *,upeer_addrlen)1944 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1945 		int __user *, upeer_addrlen)
1946 {
1947 	return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1948 }
1949 
1950 /*
1951  *	Attempt to connect to a socket with the server address.  The address
1952  *	is in user space so we verify it is OK and move it to kernel space.
1953  *
1954  *	For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1955  *	break bindings
1956  *
1957  *	NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1958  *	other SEQPACKET protocols that take time to connect() as it doesn't
1959  *	include the -EINPROGRESS status for such sockets.
1960  */
1961 
__sys_connect_file(struct file * file,struct sockaddr_storage * address,int addrlen,int file_flags)1962 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
1963 		       int addrlen, int file_flags)
1964 {
1965 	struct socket *sock;
1966 	int err;
1967 
1968 	sock = sock_from_file(file);
1969 	if (!sock) {
1970 		err = -ENOTSOCK;
1971 		goto out;
1972 	}
1973 
1974 	err =
1975 	    security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1976 	if (err)
1977 		goto out;
1978 
1979 	err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1980 				 sock->file->f_flags | file_flags);
1981 out:
1982 	return err;
1983 }
1984 
__sys_connect(int fd,struct sockaddr __user * uservaddr,int addrlen)1985 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1986 {
1987 	int ret = -EBADF;
1988 	struct fd f;
1989 
1990 	f = fdget(fd);
1991 	if (f.file) {
1992 		struct sockaddr_storage address;
1993 
1994 		ret = move_addr_to_kernel(uservaddr, addrlen, &address);
1995 		if (!ret)
1996 			ret = __sys_connect_file(f.file, &address, addrlen, 0);
1997 		fdput(f);
1998 	}
1999 
2000 	return ret;
2001 }
2002 
SYSCALL_DEFINE3(connect,int,fd,struct sockaddr __user *,uservaddr,int,addrlen)2003 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2004 		int, addrlen)
2005 {
2006 	return __sys_connect(fd, uservaddr, addrlen);
2007 }
2008 
2009 /*
2010  *	Get the local address ('name') of a socket object. Move the obtained
2011  *	name to user space.
2012  */
2013 
__sys_getsockname(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2014 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2015 		      int __user *usockaddr_len)
2016 {
2017 	struct socket *sock;
2018 	struct sockaddr_storage address;
2019 	int err, fput_needed;
2020 
2021 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2022 	if (!sock)
2023 		goto out;
2024 
2025 	err = security_socket_getsockname(sock);
2026 	if (err)
2027 		goto out_put;
2028 
2029 	err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
2030 	if (err < 0)
2031 		goto out_put;
2032 	/* "err" is actually length in this case */
2033 	err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2034 
2035 out_put:
2036 	fput_light(sock->file, fput_needed);
2037 out:
2038 	return err;
2039 }
2040 
SYSCALL_DEFINE3(getsockname,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2041 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2042 		int __user *, usockaddr_len)
2043 {
2044 	return __sys_getsockname(fd, usockaddr, usockaddr_len);
2045 }
2046 
2047 /*
2048  *	Get the remote address ('name') of a socket object. Move the obtained
2049  *	name to user space.
2050  */
2051 
__sys_getpeername(int fd,struct sockaddr __user * usockaddr,int __user * usockaddr_len)2052 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2053 		      int __user *usockaddr_len)
2054 {
2055 	struct socket *sock;
2056 	struct sockaddr_storage address;
2057 	int err, fput_needed;
2058 
2059 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2060 	if (sock != NULL) {
2061 		err = security_socket_getpeername(sock);
2062 		if (err) {
2063 			fput_light(sock->file, fput_needed);
2064 			return err;
2065 		}
2066 
2067 		err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
2068 		if (err >= 0)
2069 			/* "err" is actually length in this case */
2070 			err = move_addr_to_user(&address, err, usockaddr,
2071 						usockaddr_len);
2072 		fput_light(sock->file, fput_needed);
2073 	}
2074 	return err;
2075 }
2076 
SYSCALL_DEFINE3(getpeername,int,fd,struct sockaddr __user *,usockaddr,int __user *,usockaddr_len)2077 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2078 		int __user *, usockaddr_len)
2079 {
2080 	return __sys_getpeername(fd, usockaddr, usockaddr_len);
2081 }
2082 
2083 /*
2084  *	Send a datagram to a given address. We move the address into kernel
2085  *	space and check the user space data area is readable before invoking
2086  *	the protocol.
2087  */
__sys_sendto(int fd,void __user * buff,size_t len,unsigned int flags,struct sockaddr __user * addr,int addr_len)2088 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2089 		 struct sockaddr __user *addr,  int addr_len)
2090 {
2091 	struct socket *sock;
2092 	struct sockaddr_storage address;
2093 	int err;
2094 	struct msghdr msg;
2095 	struct iovec iov;
2096 	int fput_needed;
2097 
2098 	err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
2099 	if (unlikely(err))
2100 		return err;
2101 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2102 	if (!sock)
2103 		goto out;
2104 
2105 	msg.msg_name = NULL;
2106 	msg.msg_control = NULL;
2107 	msg.msg_controllen = 0;
2108 	msg.msg_namelen = 0;
2109 	if (addr) {
2110 		err = move_addr_to_kernel(addr, addr_len, &address);
2111 		if (err < 0)
2112 			goto out_put;
2113 		msg.msg_name = (struct sockaddr *)&address;
2114 		msg.msg_namelen = addr_len;
2115 	}
2116 	if (sock->file->f_flags & O_NONBLOCK)
2117 		flags |= MSG_DONTWAIT;
2118 	msg.msg_flags = flags;
2119 	err = sock_sendmsg(sock, &msg);
2120 
2121 out_put:
2122 	fput_light(sock->file, fput_needed);
2123 out:
2124 	return err;
2125 }
2126 
SYSCALL_DEFINE6(sendto,int,fd,void __user *,buff,size_t,len,unsigned int,flags,struct sockaddr __user *,addr,int,addr_len)2127 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2128 		unsigned int, flags, struct sockaddr __user *, addr,
2129 		int, addr_len)
2130 {
2131 	return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2132 }
2133 
2134 /*
2135  *	Send a datagram down a socket.
2136  */
2137 
SYSCALL_DEFINE4(send,int,fd,void __user *,buff,size_t,len,unsigned int,flags)2138 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2139 		unsigned int, flags)
2140 {
2141 	return __sys_sendto(fd, buff, len, flags, NULL, 0);
2142 }
2143 
2144 /*
2145  *	Receive a frame from the socket and optionally record the address of the
2146  *	sender. We verify the buffers are writable and if needed move the
2147  *	sender address from kernel to user space.
2148  */
__sys_recvfrom(int fd,void __user * ubuf,size_t size,unsigned int flags,struct sockaddr __user * addr,int __user * addr_len)2149 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2150 		   struct sockaddr __user *addr, int __user *addr_len)
2151 {
2152 	struct sockaddr_storage address;
2153 	struct msghdr msg = {
2154 		/* Save some cycles and don't copy the address if not needed */
2155 		.msg_name = addr ? (struct sockaddr *)&address : NULL,
2156 	};
2157 	struct socket *sock;
2158 	struct iovec iov;
2159 	int err, err2;
2160 	int fput_needed;
2161 
2162 	err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
2163 	if (unlikely(err))
2164 		return err;
2165 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2166 	if (!sock)
2167 		goto out;
2168 
2169 	if (sock->file->f_flags & O_NONBLOCK)
2170 		flags |= MSG_DONTWAIT;
2171 	err = sock_recvmsg(sock, &msg, flags);
2172 
2173 	if (err >= 0 && addr != NULL) {
2174 		err2 = move_addr_to_user(&address,
2175 					 msg.msg_namelen, addr, addr_len);
2176 		if (err2 < 0)
2177 			err = err2;
2178 	}
2179 
2180 	fput_light(sock->file, fput_needed);
2181 out:
2182 	return err;
2183 }
2184 
SYSCALL_DEFINE6(recvfrom,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags,struct sockaddr __user *,addr,int __user *,addr_len)2185 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2186 		unsigned int, flags, struct sockaddr __user *, addr,
2187 		int __user *, addr_len)
2188 {
2189 	return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2190 }
2191 
2192 /*
2193  *	Receive a datagram from a socket.
2194  */
2195 
SYSCALL_DEFINE4(recv,int,fd,void __user *,ubuf,size_t,size,unsigned int,flags)2196 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2197 		unsigned int, flags)
2198 {
2199 	return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2200 }
2201 
sock_use_custom_sol_socket(const struct socket * sock)2202 static bool sock_use_custom_sol_socket(const struct socket *sock)
2203 {
2204 	const struct sock *sk = sock->sk;
2205 
2206 	/* Use sock->ops->setsockopt() for MPTCP */
2207 	return IS_ENABLED(CONFIG_MPTCP) &&
2208 	       sk->sk_protocol == IPPROTO_MPTCP &&
2209 	       sk->sk_type == SOCK_STREAM &&
2210 	       (sk->sk_family == AF_INET || sk->sk_family == AF_INET6);
2211 }
2212 
2213 /*
2214  *	Set a socket option. Because we don't know the option lengths we have
2215  *	to pass the user mode parameter for the protocols to sort out.
2216  */
__sys_setsockopt(int fd,int level,int optname,char __user * user_optval,int optlen)2217 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2218 		int optlen)
2219 {
2220 	sockptr_t optval = USER_SOCKPTR(user_optval);
2221 	char *kernel_optval = NULL;
2222 	int err, fput_needed;
2223 	struct socket *sock;
2224 
2225 	if (optlen < 0)
2226 		return -EINVAL;
2227 
2228 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2229 	if (!sock)
2230 		return err;
2231 
2232 	err = security_socket_setsockopt(sock, level, optname);
2233 	if (err)
2234 		goto out_put;
2235 
2236 	if (!in_compat_syscall())
2237 		err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2238 						     user_optval, &optlen,
2239 						     &kernel_optval);
2240 	if (err < 0)
2241 		goto out_put;
2242 	if (err > 0) {
2243 		err = 0;
2244 		goto out_put;
2245 	}
2246 
2247 	if (kernel_optval)
2248 		optval = KERNEL_SOCKPTR(kernel_optval);
2249 	if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2250 		err = sock_setsockopt(sock, level, optname, optval, optlen);
2251 	else if (unlikely(!sock->ops->setsockopt))
2252 		err = -EOPNOTSUPP;
2253 	else
2254 		err = sock->ops->setsockopt(sock, level, optname, optval,
2255 					    optlen);
2256 	kfree(kernel_optval);
2257 out_put:
2258 	fput_light(sock->file, fput_needed);
2259 	return err;
2260 }
2261 
SYSCALL_DEFINE5(setsockopt,int,fd,int,level,int,optname,char __user *,optval,int,optlen)2262 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2263 		char __user *, optval, int, optlen)
2264 {
2265 	return __sys_setsockopt(fd, level, optname, optval, optlen);
2266 }
2267 
2268 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2269 							 int optname));
2270 
2271 /*
2272  *	Get a socket option. Because we don't know the option lengths we have
2273  *	to pass a user mode parameter for the protocols to sort out.
2274  */
__sys_getsockopt(int fd,int level,int optname,char __user * optval,int __user * optlen)2275 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2276 		int __user *optlen)
2277 {
2278 	int err, fput_needed;
2279 	struct socket *sock;
2280 	int max_optlen;
2281 
2282 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2283 	if (!sock)
2284 		return err;
2285 
2286 	err = security_socket_getsockopt(sock, level, optname);
2287 	if (err)
2288 		goto out_put;
2289 
2290 	if (!in_compat_syscall())
2291 		max_optlen = BPF_CGROUP_GETSOCKOPT_MAX_OPTLEN(optlen);
2292 
2293 	if (level == SOL_SOCKET)
2294 		err = sock_getsockopt(sock, level, optname, optval, optlen);
2295 	else if (unlikely(!sock->ops->getsockopt))
2296 		err = -EOPNOTSUPP;
2297 	else
2298 		err = sock->ops->getsockopt(sock, level, optname, optval,
2299 					    optlen);
2300 
2301 	if (!in_compat_syscall())
2302 		err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2303 						     optval, optlen, max_optlen,
2304 						     err);
2305 out_put:
2306 	fput_light(sock->file, fput_needed);
2307 	return err;
2308 }
2309 
SYSCALL_DEFINE5(getsockopt,int,fd,int,level,int,optname,char __user *,optval,int __user *,optlen)2310 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2311 		char __user *, optval, int __user *, optlen)
2312 {
2313 	return __sys_getsockopt(fd, level, optname, optval, optlen);
2314 }
2315 
2316 /*
2317  *	Shutdown a socket.
2318  */
2319 
__sys_shutdown_sock(struct socket * sock,int how)2320 int __sys_shutdown_sock(struct socket *sock, int how)
2321 {
2322 	int err;
2323 
2324 	err = security_socket_shutdown(sock, how);
2325 	if (!err)
2326 		err = sock->ops->shutdown(sock, how);
2327 
2328 	return err;
2329 }
2330 
__sys_shutdown(int fd,int how)2331 int __sys_shutdown(int fd, int how)
2332 {
2333 	int err, fput_needed;
2334 	struct socket *sock;
2335 
2336 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2337 	if (sock != NULL) {
2338 		err = __sys_shutdown_sock(sock, how);
2339 		fput_light(sock->file, fput_needed);
2340 	}
2341 	return err;
2342 }
2343 
SYSCALL_DEFINE2(shutdown,int,fd,int,how)2344 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2345 {
2346 	return __sys_shutdown(fd, how);
2347 }
2348 
2349 /* A couple of helpful macros for getting the address of the 32/64 bit
2350  * fields which are the same type (int / unsigned) on our platforms.
2351  */
2352 #define COMPAT_MSG(msg, member)	((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2353 #define COMPAT_NAMELEN(msg)	COMPAT_MSG(msg, msg_namelen)
2354 #define COMPAT_FLAGS(msg)	COMPAT_MSG(msg, msg_flags)
2355 
2356 struct used_address {
2357 	struct sockaddr_storage name;
2358 	unsigned int name_len;
2359 };
2360 
__copy_msghdr_from_user(struct msghdr * kmsg,struct user_msghdr __user * umsg,struct sockaddr __user ** save_addr,struct iovec __user ** uiov,size_t * nsegs)2361 int __copy_msghdr_from_user(struct msghdr *kmsg,
2362 			    struct user_msghdr __user *umsg,
2363 			    struct sockaddr __user **save_addr,
2364 			    struct iovec __user **uiov, size_t *nsegs)
2365 {
2366 	struct user_msghdr msg;
2367 	ssize_t err;
2368 
2369 	if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2370 		return -EFAULT;
2371 
2372 	kmsg->msg_control_is_user = true;
2373 	kmsg->msg_get_inq = 0;
2374 	kmsg->msg_control_user = msg.msg_control;
2375 	kmsg->msg_controllen = msg.msg_controllen;
2376 	kmsg->msg_flags = msg.msg_flags;
2377 
2378 	kmsg->msg_namelen = msg.msg_namelen;
2379 	if (!msg.msg_name)
2380 		kmsg->msg_namelen = 0;
2381 
2382 	if (kmsg->msg_namelen < 0)
2383 		return -EINVAL;
2384 
2385 	if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2386 		kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2387 
2388 	if (save_addr)
2389 		*save_addr = msg.msg_name;
2390 
2391 	if (msg.msg_name && kmsg->msg_namelen) {
2392 		if (!save_addr) {
2393 			err = move_addr_to_kernel(msg.msg_name,
2394 						  kmsg->msg_namelen,
2395 						  kmsg->msg_name);
2396 			if (err < 0)
2397 				return err;
2398 		}
2399 	} else {
2400 		kmsg->msg_name = NULL;
2401 		kmsg->msg_namelen = 0;
2402 	}
2403 
2404 	if (msg.msg_iovlen > UIO_MAXIOV)
2405 		return -EMSGSIZE;
2406 
2407 	kmsg->msg_iocb = NULL;
2408 	*uiov = msg.msg_iov;
2409 	*nsegs = msg.msg_iovlen;
2410 	return 0;
2411 }
2412 
copy_msghdr_from_user(struct msghdr * kmsg,struct user_msghdr __user * umsg,struct sockaddr __user ** save_addr,struct iovec ** iov)2413 static int copy_msghdr_from_user(struct msghdr *kmsg,
2414 				 struct user_msghdr __user *umsg,
2415 				 struct sockaddr __user **save_addr,
2416 				 struct iovec **iov)
2417 {
2418 	struct user_msghdr msg;
2419 	ssize_t err;
2420 
2421 	err = __copy_msghdr_from_user(kmsg, umsg, save_addr, &msg.msg_iov,
2422 					&msg.msg_iovlen);
2423 	if (err)
2424 		return err;
2425 
2426 	err = import_iovec(save_addr ? READ : WRITE,
2427 			    msg.msg_iov, msg.msg_iovlen,
2428 			    UIO_FASTIOV, iov, &kmsg->msg_iter);
2429 	return err < 0 ? err : 0;
2430 }
2431 
____sys_sendmsg(struct socket * sock,struct msghdr * msg_sys,unsigned int flags,struct used_address * used_address,unsigned int allowed_msghdr_flags)2432 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2433 			   unsigned int flags, struct used_address *used_address,
2434 			   unsigned int allowed_msghdr_flags)
2435 {
2436 	unsigned char ctl[sizeof(struct cmsghdr) + 20]
2437 				__aligned(sizeof(__kernel_size_t));
2438 	/* 20 is size of ipv6_pktinfo */
2439 	unsigned char *ctl_buf = ctl;
2440 	int ctl_len;
2441 	ssize_t err;
2442 
2443 	err = -ENOBUFS;
2444 
2445 	if (msg_sys->msg_controllen > INT_MAX)
2446 		goto out;
2447 	flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2448 	ctl_len = msg_sys->msg_controllen;
2449 	if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2450 		err =
2451 		    cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2452 						     sizeof(ctl));
2453 		if (err)
2454 			goto out;
2455 		ctl_buf = msg_sys->msg_control;
2456 		ctl_len = msg_sys->msg_controllen;
2457 	} else if (ctl_len) {
2458 		BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2459 			     CMSG_ALIGN(sizeof(struct cmsghdr)));
2460 		if (ctl_len > sizeof(ctl)) {
2461 			ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2462 			if (ctl_buf == NULL)
2463 				goto out;
2464 		}
2465 		err = -EFAULT;
2466 		if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2467 			goto out_freectl;
2468 		msg_sys->msg_control = ctl_buf;
2469 		msg_sys->msg_control_is_user = false;
2470 	}
2471 	msg_sys->msg_flags = flags;
2472 
2473 	if (sock->file->f_flags & O_NONBLOCK)
2474 		msg_sys->msg_flags |= MSG_DONTWAIT;
2475 	/*
2476 	 * If this is sendmmsg() and current destination address is same as
2477 	 * previously succeeded address, omit asking LSM's decision.
2478 	 * used_address->name_len is initialized to UINT_MAX so that the first
2479 	 * destination address never matches.
2480 	 */
2481 	if (used_address && msg_sys->msg_name &&
2482 	    used_address->name_len == msg_sys->msg_namelen &&
2483 	    !memcmp(&used_address->name, msg_sys->msg_name,
2484 		    used_address->name_len)) {
2485 		err = sock_sendmsg_nosec(sock, msg_sys);
2486 		goto out_freectl;
2487 	}
2488 	err = sock_sendmsg(sock, msg_sys);
2489 	/*
2490 	 * If this is sendmmsg() and sending to current destination address was
2491 	 * successful, remember it.
2492 	 */
2493 	if (used_address && err >= 0) {
2494 		used_address->name_len = msg_sys->msg_namelen;
2495 		if (msg_sys->msg_name)
2496 			memcpy(&used_address->name, msg_sys->msg_name,
2497 			       used_address->name_len);
2498 	}
2499 
2500 out_freectl:
2501 	if (ctl_buf != ctl)
2502 		sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2503 out:
2504 	return err;
2505 }
2506 
sendmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct iovec ** iov)2507 int sendmsg_copy_msghdr(struct msghdr *msg,
2508 			struct user_msghdr __user *umsg, unsigned flags,
2509 			struct iovec **iov)
2510 {
2511 	int err;
2512 
2513 	if (flags & MSG_CMSG_COMPAT) {
2514 		struct compat_msghdr __user *msg_compat;
2515 
2516 		msg_compat = (struct compat_msghdr __user *) umsg;
2517 		err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2518 	} else {
2519 		err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2520 	}
2521 	if (err < 0)
2522 		return err;
2523 
2524 	return 0;
2525 }
2526 
___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)2527 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2528 			 struct msghdr *msg_sys, unsigned int flags,
2529 			 struct used_address *used_address,
2530 			 unsigned int allowed_msghdr_flags)
2531 {
2532 	struct sockaddr_storage address;
2533 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2534 	ssize_t err;
2535 
2536 	msg_sys->msg_name = &address;
2537 
2538 	err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2539 	if (err < 0)
2540 		return err;
2541 
2542 	err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2543 				allowed_msghdr_flags);
2544 	kfree(iov);
2545 	return err;
2546 }
2547 
2548 /*
2549  *	BSD sendmsg interface
2550  */
__sys_sendmsg_sock(struct socket * sock,struct msghdr * msg,unsigned int flags)2551 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2552 			unsigned int flags)
2553 {
2554 	return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2555 }
2556 
__sys_sendmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2557 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2558 		   bool forbid_cmsg_compat)
2559 {
2560 	int fput_needed, err;
2561 	struct msghdr msg_sys;
2562 	struct socket *sock;
2563 
2564 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2565 		return -EINVAL;
2566 
2567 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2568 	if (!sock)
2569 		goto out;
2570 
2571 	err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2572 
2573 	fput_light(sock->file, fput_needed);
2574 out:
2575 	return err;
2576 }
2577 
SYSCALL_DEFINE3(sendmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2578 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2579 {
2580 	return __sys_sendmsg(fd, msg, flags, true);
2581 }
2582 
2583 /*
2584  *	Linux sendmmsg interface
2585  */
2586 
__sys_sendmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,bool forbid_cmsg_compat)2587 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2588 		   unsigned int flags, bool forbid_cmsg_compat)
2589 {
2590 	int fput_needed, err, datagrams;
2591 	struct socket *sock;
2592 	struct mmsghdr __user *entry;
2593 	struct compat_mmsghdr __user *compat_entry;
2594 	struct msghdr msg_sys;
2595 	struct used_address used_address;
2596 	unsigned int oflags = flags;
2597 
2598 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2599 		return -EINVAL;
2600 
2601 	if (vlen > UIO_MAXIOV)
2602 		vlen = UIO_MAXIOV;
2603 
2604 	datagrams = 0;
2605 
2606 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2607 	if (!sock)
2608 		return err;
2609 
2610 	used_address.name_len = UINT_MAX;
2611 	entry = mmsg;
2612 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2613 	err = 0;
2614 	flags |= MSG_BATCH;
2615 
2616 	while (datagrams < vlen) {
2617 		if (datagrams == vlen - 1)
2618 			flags = oflags;
2619 
2620 		if (MSG_CMSG_COMPAT & flags) {
2621 			err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2622 					     &msg_sys, flags, &used_address, MSG_EOR);
2623 			if (err < 0)
2624 				break;
2625 			err = __put_user(err, &compat_entry->msg_len);
2626 			++compat_entry;
2627 		} else {
2628 			err = ___sys_sendmsg(sock,
2629 					     (struct user_msghdr __user *)entry,
2630 					     &msg_sys, flags, &used_address, MSG_EOR);
2631 			if (err < 0)
2632 				break;
2633 			err = put_user(err, &entry->msg_len);
2634 			++entry;
2635 		}
2636 
2637 		if (err)
2638 			break;
2639 		++datagrams;
2640 		if (msg_data_left(&msg_sys))
2641 			break;
2642 		cond_resched();
2643 	}
2644 
2645 	fput_light(sock->file, fput_needed);
2646 
2647 	/* We only return an error if no datagrams were able to be sent */
2648 	if (datagrams != 0)
2649 		return datagrams;
2650 
2651 	return err;
2652 }
2653 
SYSCALL_DEFINE4(sendmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags)2654 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2655 		unsigned int, vlen, unsigned int, flags)
2656 {
2657 	return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2658 }
2659 
recvmsg_copy_msghdr(struct msghdr * msg,struct user_msghdr __user * umsg,unsigned flags,struct sockaddr __user ** uaddr,struct iovec ** iov)2660 int recvmsg_copy_msghdr(struct msghdr *msg,
2661 			struct user_msghdr __user *umsg, unsigned flags,
2662 			struct sockaddr __user **uaddr,
2663 			struct iovec **iov)
2664 {
2665 	ssize_t err;
2666 
2667 	if (MSG_CMSG_COMPAT & flags) {
2668 		struct compat_msghdr __user *msg_compat;
2669 
2670 		msg_compat = (struct compat_msghdr __user *) umsg;
2671 		err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2672 	} else {
2673 		err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2674 	}
2675 	if (err < 0)
2676 		return err;
2677 
2678 	return 0;
2679 }
2680 
____sys_recvmsg(struct socket * sock,struct msghdr * msg_sys,struct user_msghdr __user * msg,struct sockaddr __user * uaddr,unsigned int flags,int nosec)2681 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2682 			   struct user_msghdr __user *msg,
2683 			   struct sockaddr __user *uaddr,
2684 			   unsigned int flags, int nosec)
2685 {
2686 	struct compat_msghdr __user *msg_compat =
2687 					(struct compat_msghdr __user *) msg;
2688 	int __user *uaddr_len = COMPAT_NAMELEN(msg);
2689 	struct sockaddr_storage addr;
2690 	unsigned long cmsg_ptr;
2691 	int len;
2692 	ssize_t err;
2693 
2694 	msg_sys->msg_name = &addr;
2695 	cmsg_ptr = (unsigned long)msg_sys->msg_control;
2696 	msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2697 
2698 	/* We assume all kernel code knows the size of sockaddr_storage */
2699 	msg_sys->msg_namelen = 0;
2700 
2701 	if (sock->file->f_flags & O_NONBLOCK)
2702 		flags |= MSG_DONTWAIT;
2703 
2704 	if (unlikely(nosec))
2705 		err = sock_recvmsg_nosec(sock, msg_sys, flags);
2706 	else
2707 		err = sock_recvmsg(sock, msg_sys, flags);
2708 
2709 	if (err < 0)
2710 		goto out;
2711 	len = err;
2712 
2713 	if (uaddr != NULL) {
2714 		err = move_addr_to_user(&addr,
2715 					msg_sys->msg_namelen, uaddr,
2716 					uaddr_len);
2717 		if (err < 0)
2718 			goto out;
2719 	}
2720 	err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2721 			 COMPAT_FLAGS(msg));
2722 	if (err)
2723 		goto out;
2724 	if (MSG_CMSG_COMPAT & flags)
2725 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2726 				 &msg_compat->msg_controllen);
2727 	else
2728 		err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2729 				 &msg->msg_controllen);
2730 	if (err)
2731 		goto out;
2732 	err = len;
2733 out:
2734 	return err;
2735 }
2736 
___sys_recvmsg(struct socket * sock,struct user_msghdr __user * msg,struct msghdr * msg_sys,unsigned int flags,int nosec)2737 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2738 			 struct msghdr *msg_sys, unsigned int flags, int nosec)
2739 {
2740 	struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2741 	/* user mode address pointers */
2742 	struct sockaddr __user *uaddr;
2743 	ssize_t err;
2744 
2745 	err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2746 	if (err < 0)
2747 		return err;
2748 
2749 	err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2750 	kfree(iov);
2751 	return err;
2752 }
2753 
2754 /*
2755  *	BSD recvmsg interface
2756  */
2757 
__sys_recvmsg_sock(struct socket * sock,struct msghdr * msg,struct user_msghdr __user * umsg,struct sockaddr __user * uaddr,unsigned int flags)2758 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2759 			struct user_msghdr __user *umsg,
2760 			struct sockaddr __user *uaddr, unsigned int flags)
2761 {
2762 	return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2763 }
2764 
__sys_recvmsg(int fd,struct user_msghdr __user * msg,unsigned int flags,bool forbid_cmsg_compat)2765 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2766 		   bool forbid_cmsg_compat)
2767 {
2768 	int fput_needed, err;
2769 	struct msghdr msg_sys;
2770 	struct socket *sock;
2771 
2772 	if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2773 		return -EINVAL;
2774 
2775 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2776 	if (!sock)
2777 		goto out;
2778 
2779 	err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2780 
2781 	fput_light(sock->file, fput_needed);
2782 out:
2783 	return err;
2784 }
2785 
SYSCALL_DEFINE3(recvmsg,int,fd,struct user_msghdr __user *,msg,unsigned int,flags)2786 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2787 		unsigned int, flags)
2788 {
2789 	return __sys_recvmsg(fd, msg, flags, true);
2790 }
2791 
2792 /*
2793  *     Linux recvmmsg interface
2794  */
2795 
do_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct timespec64 * timeout)2796 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2797 			  unsigned int vlen, unsigned int flags,
2798 			  struct timespec64 *timeout)
2799 {
2800 	int fput_needed, err, datagrams;
2801 	struct socket *sock;
2802 	struct mmsghdr __user *entry;
2803 	struct compat_mmsghdr __user *compat_entry;
2804 	struct msghdr msg_sys;
2805 	struct timespec64 end_time;
2806 	struct timespec64 timeout64;
2807 
2808 	if (timeout &&
2809 	    poll_select_set_timeout(&end_time, timeout->tv_sec,
2810 				    timeout->tv_nsec))
2811 		return -EINVAL;
2812 
2813 	datagrams = 0;
2814 
2815 	sock = sockfd_lookup_light(fd, &err, &fput_needed);
2816 	if (!sock)
2817 		return err;
2818 
2819 	if (likely(!(flags & MSG_ERRQUEUE))) {
2820 		err = sock_error(sock->sk);
2821 		if (err) {
2822 			datagrams = err;
2823 			goto out_put;
2824 		}
2825 	}
2826 
2827 	entry = mmsg;
2828 	compat_entry = (struct compat_mmsghdr __user *)mmsg;
2829 
2830 	while (datagrams < vlen) {
2831 		/*
2832 		 * No need to ask LSM for more than the first datagram.
2833 		 */
2834 		if (MSG_CMSG_COMPAT & flags) {
2835 			err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2836 					     &msg_sys, flags & ~MSG_WAITFORONE,
2837 					     datagrams);
2838 			if (err < 0)
2839 				break;
2840 			err = __put_user(err, &compat_entry->msg_len);
2841 			++compat_entry;
2842 		} else {
2843 			err = ___sys_recvmsg(sock,
2844 					     (struct user_msghdr __user *)entry,
2845 					     &msg_sys, flags & ~MSG_WAITFORONE,
2846 					     datagrams);
2847 			if (err < 0)
2848 				break;
2849 			err = put_user(err, &entry->msg_len);
2850 			++entry;
2851 		}
2852 
2853 		if (err)
2854 			break;
2855 		++datagrams;
2856 
2857 		/* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2858 		if (flags & MSG_WAITFORONE)
2859 			flags |= MSG_DONTWAIT;
2860 
2861 		if (timeout) {
2862 			ktime_get_ts64(&timeout64);
2863 			*timeout = timespec64_sub(end_time, timeout64);
2864 			if (timeout->tv_sec < 0) {
2865 				timeout->tv_sec = timeout->tv_nsec = 0;
2866 				break;
2867 			}
2868 
2869 			/* Timeout, return less than vlen datagrams */
2870 			if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2871 				break;
2872 		}
2873 
2874 		/* Out of band data, return right away */
2875 		if (msg_sys.msg_flags & MSG_OOB)
2876 			break;
2877 		cond_resched();
2878 	}
2879 
2880 	if (err == 0)
2881 		goto out_put;
2882 
2883 	if (datagrams == 0) {
2884 		datagrams = err;
2885 		goto out_put;
2886 	}
2887 
2888 	/*
2889 	 * We may return less entries than requested (vlen) if the
2890 	 * sock is non block and there aren't enough datagrams...
2891 	 */
2892 	if (err != -EAGAIN) {
2893 		/*
2894 		 * ... or  if recvmsg returns an error after we
2895 		 * received some datagrams, where we record the
2896 		 * error to return on the next call or if the
2897 		 * app asks about it using getsockopt(SO_ERROR).
2898 		 */
2899 		sock->sk->sk_err = -err;
2900 	}
2901 out_put:
2902 	fput_light(sock->file, fput_needed);
2903 
2904 	return datagrams;
2905 }
2906 
__sys_recvmmsg(int fd,struct mmsghdr __user * mmsg,unsigned int vlen,unsigned int flags,struct __kernel_timespec __user * timeout,struct old_timespec32 __user * timeout32)2907 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2908 		   unsigned int vlen, unsigned int flags,
2909 		   struct __kernel_timespec __user *timeout,
2910 		   struct old_timespec32 __user *timeout32)
2911 {
2912 	int datagrams;
2913 	struct timespec64 timeout_sys;
2914 
2915 	if (timeout && get_timespec64(&timeout_sys, timeout))
2916 		return -EFAULT;
2917 
2918 	if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
2919 		return -EFAULT;
2920 
2921 	if (!timeout && !timeout32)
2922 		return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
2923 
2924 	datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2925 
2926 	if (datagrams <= 0)
2927 		return datagrams;
2928 
2929 	if (timeout && put_timespec64(&timeout_sys, timeout))
2930 		datagrams = -EFAULT;
2931 
2932 	if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
2933 		datagrams = -EFAULT;
2934 
2935 	return datagrams;
2936 }
2937 
SYSCALL_DEFINE5(recvmmsg,int,fd,struct mmsghdr __user *,mmsg,unsigned int,vlen,unsigned int,flags,struct __kernel_timespec __user *,timeout)2938 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2939 		unsigned int, vlen, unsigned int, flags,
2940 		struct __kernel_timespec __user *, timeout)
2941 {
2942 	if (flags & MSG_CMSG_COMPAT)
2943 		return -EINVAL;
2944 
2945 	return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
2946 }
2947 
2948 #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)2949 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
2950 		unsigned int, vlen, unsigned int, flags,
2951 		struct old_timespec32 __user *, timeout)
2952 {
2953 	if (flags & MSG_CMSG_COMPAT)
2954 		return -EINVAL;
2955 
2956 	return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
2957 }
2958 #endif
2959 
2960 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2961 /* Argument list sizes for sys_socketcall */
2962 #define AL(x) ((x) * sizeof(unsigned long))
2963 static const unsigned char nargs[21] = {
2964 	AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2965 	AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2966 	AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2967 	AL(4), AL(5), AL(4)
2968 };
2969 
2970 #undef AL
2971 
2972 /*
2973  *	System call vectors.
2974  *
2975  *	Argument checking cleaned up. Saved 20% in size.
2976  *  This function doesn't need to set the kernel lock because
2977  *  it is set by the callees.
2978  */
2979 
SYSCALL_DEFINE2(socketcall,int,call,unsigned long __user *,args)2980 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2981 {
2982 	unsigned long a[AUDITSC_ARGS];
2983 	unsigned long a0, a1;
2984 	int err;
2985 	unsigned int len;
2986 
2987 	if (call < 1 || call > SYS_SENDMMSG)
2988 		return -EINVAL;
2989 	call = array_index_nospec(call, SYS_SENDMMSG + 1);
2990 
2991 	len = nargs[call];
2992 	if (len > sizeof(a))
2993 		return -EINVAL;
2994 
2995 	/* copy_from_user should be SMP safe. */
2996 	if (copy_from_user(a, args, len))
2997 		return -EFAULT;
2998 
2999 	err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3000 	if (err)
3001 		return err;
3002 
3003 	a0 = a[0];
3004 	a1 = a[1];
3005 
3006 	switch (call) {
3007 	case SYS_SOCKET:
3008 		err = __sys_socket(a0, a1, a[2]);
3009 		break;
3010 	case SYS_BIND:
3011 		err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3012 		break;
3013 	case SYS_CONNECT:
3014 		err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3015 		break;
3016 	case SYS_LISTEN:
3017 		err = __sys_listen(a0, a1);
3018 		break;
3019 	case SYS_ACCEPT:
3020 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3021 				    (int __user *)a[2], 0);
3022 		break;
3023 	case SYS_GETSOCKNAME:
3024 		err =
3025 		    __sys_getsockname(a0, (struct sockaddr __user *)a1,
3026 				      (int __user *)a[2]);
3027 		break;
3028 	case SYS_GETPEERNAME:
3029 		err =
3030 		    __sys_getpeername(a0, (struct sockaddr __user *)a1,
3031 				      (int __user *)a[2]);
3032 		break;
3033 	case SYS_SOCKETPAIR:
3034 		err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3035 		break;
3036 	case SYS_SEND:
3037 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3038 				   NULL, 0);
3039 		break;
3040 	case SYS_SENDTO:
3041 		err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3042 				   (struct sockaddr __user *)a[4], a[5]);
3043 		break;
3044 	case SYS_RECV:
3045 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3046 				     NULL, NULL);
3047 		break;
3048 	case SYS_RECVFROM:
3049 		err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3050 				     (struct sockaddr __user *)a[4],
3051 				     (int __user *)a[5]);
3052 		break;
3053 	case SYS_SHUTDOWN:
3054 		err = __sys_shutdown(a0, a1);
3055 		break;
3056 	case SYS_SETSOCKOPT:
3057 		err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3058 				       a[4]);
3059 		break;
3060 	case SYS_GETSOCKOPT:
3061 		err =
3062 		    __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3063 				     (int __user *)a[4]);
3064 		break;
3065 	case SYS_SENDMSG:
3066 		err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3067 				    a[2], true);
3068 		break;
3069 	case SYS_SENDMMSG:
3070 		err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3071 				     a[3], true);
3072 		break;
3073 	case SYS_RECVMSG:
3074 		err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3075 				    a[2], true);
3076 		break;
3077 	case SYS_RECVMMSG:
3078 		if (IS_ENABLED(CONFIG_64BIT))
3079 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3080 					     a[2], a[3],
3081 					     (struct __kernel_timespec __user *)a[4],
3082 					     NULL);
3083 		else
3084 			err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3085 					     a[2], a[3], NULL,
3086 					     (struct old_timespec32 __user *)a[4]);
3087 		break;
3088 	case SYS_ACCEPT4:
3089 		err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3090 				    (int __user *)a[2], a[3]);
3091 		break;
3092 	default:
3093 		err = -EINVAL;
3094 		break;
3095 	}
3096 	return err;
3097 }
3098 
3099 #endif				/* __ARCH_WANT_SYS_SOCKETCALL */
3100 
3101 /**
3102  *	sock_register - add a socket protocol handler
3103  *	@ops: description of protocol
3104  *
3105  *	This function is called by a protocol handler that wants to
3106  *	advertise its address family, and have it linked into the
3107  *	socket interface. The value ops->family corresponds to the
3108  *	socket system call protocol family.
3109  */
sock_register(const struct net_proto_family * ops)3110 int sock_register(const struct net_proto_family *ops)
3111 {
3112 	int err;
3113 
3114 	if (ops->family >= NPROTO) {
3115 		pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3116 		return -ENOBUFS;
3117 	}
3118 
3119 	spin_lock(&net_family_lock);
3120 	if (rcu_dereference_protected(net_families[ops->family],
3121 				      lockdep_is_held(&net_family_lock)))
3122 		err = -EEXIST;
3123 	else {
3124 		rcu_assign_pointer(net_families[ops->family], ops);
3125 		err = 0;
3126 	}
3127 	spin_unlock(&net_family_lock);
3128 
3129 	pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3130 	return err;
3131 }
3132 EXPORT_SYMBOL(sock_register);
3133 
3134 /**
3135  *	sock_unregister - remove a protocol handler
3136  *	@family: protocol family to remove
3137  *
3138  *	This function is called by a protocol handler that wants to
3139  *	remove its address family, and have it unlinked from the
3140  *	new socket creation.
3141  *
3142  *	If protocol handler is a module, then it can use module reference
3143  *	counts to protect against new references. If protocol handler is not
3144  *	a module then it needs to provide its own protection in
3145  *	the ops->create routine.
3146  */
sock_unregister(int family)3147 void sock_unregister(int family)
3148 {
3149 	BUG_ON(family < 0 || family >= NPROTO);
3150 
3151 	spin_lock(&net_family_lock);
3152 	RCU_INIT_POINTER(net_families[family], NULL);
3153 	spin_unlock(&net_family_lock);
3154 
3155 	synchronize_rcu();
3156 
3157 	pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3158 }
3159 EXPORT_SYMBOL(sock_unregister);
3160 
sock_is_registered(int family)3161 bool sock_is_registered(int family)
3162 {
3163 	return family < NPROTO && rcu_access_pointer(net_families[family]);
3164 }
3165 
sock_init(void)3166 static int __init sock_init(void)
3167 {
3168 	int err;
3169 	/*
3170 	 *      Initialize the network sysctl infrastructure.
3171 	 */
3172 	err = net_sysctl_init();
3173 	if (err)
3174 		goto out;
3175 
3176 	/*
3177 	 *      Initialize skbuff SLAB cache
3178 	 */
3179 	skb_init();
3180 
3181 	/*
3182 	 *      Initialize the protocols module.
3183 	 */
3184 
3185 	init_inodecache();
3186 
3187 	err = register_filesystem(&sock_fs_type);
3188 	if (err)
3189 		goto out;
3190 	sock_mnt = kern_mount(&sock_fs_type);
3191 	if (IS_ERR(sock_mnt)) {
3192 		err = PTR_ERR(sock_mnt);
3193 		goto out_mount;
3194 	}
3195 
3196 	/* The real protocol initialization is performed in later initcalls.
3197 	 */
3198 
3199 #ifdef CONFIG_NETFILTER
3200 	err = netfilter_init();
3201 	if (err)
3202 		goto out;
3203 #endif
3204 
3205 	ptp_classifier_init();
3206 
3207 out:
3208 	return err;
3209 
3210 out_mount:
3211 	unregister_filesystem(&sock_fs_type);
3212 	goto out;
3213 }
3214 
3215 core_initcall(sock_init);	/* early initcall */
3216 
3217 #ifdef CONFIG_PROC_FS
socket_seq_show(struct seq_file * seq)3218 void socket_seq_show(struct seq_file *seq)
3219 {
3220 	seq_printf(seq, "sockets: used %d\n",
3221 		   sock_inuse_get(seq->private));
3222 }
3223 #endif				/* CONFIG_PROC_FS */
3224 
3225 /* Handle the fact that while struct ifreq has the same *layout* on
3226  * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3227  * which are handled elsewhere, it still has different *size* due to
3228  * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3229  * resulting in struct ifreq being 32 and 40 bytes respectively).
3230  * As a result, if the struct happens to be at the end of a page and
3231  * the next page isn't readable/writable, we get a fault. To prevent
3232  * that, copy back and forth to the full size.
3233  */
get_user_ifreq(struct ifreq * ifr,void __user ** ifrdata,void __user * arg)3234 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3235 {
3236 	if (in_compat_syscall()) {
3237 		struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3238 
3239 		memset(ifr, 0, sizeof(*ifr));
3240 		if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3241 			return -EFAULT;
3242 
3243 		if (ifrdata)
3244 			*ifrdata = compat_ptr(ifr32->ifr_data);
3245 
3246 		return 0;
3247 	}
3248 
3249 	if (copy_from_user(ifr, arg, sizeof(*ifr)))
3250 		return -EFAULT;
3251 
3252 	if (ifrdata)
3253 		*ifrdata = ifr->ifr_data;
3254 
3255 	return 0;
3256 }
3257 EXPORT_SYMBOL(get_user_ifreq);
3258 
put_user_ifreq(struct ifreq * ifr,void __user * arg)3259 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3260 {
3261 	size_t size = sizeof(*ifr);
3262 
3263 	if (in_compat_syscall())
3264 		size = sizeof(struct compat_ifreq);
3265 
3266 	if (copy_to_user(arg, ifr, size))
3267 		return -EFAULT;
3268 
3269 	return 0;
3270 }
3271 EXPORT_SYMBOL(put_user_ifreq);
3272 
3273 #ifdef CONFIG_COMPAT
compat_siocwandev(struct net * net,struct compat_ifreq __user * uifr32)3274 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3275 {
3276 	compat_uptr_t uptr32;
3277 	struct ifreq ifr;
3278 	void __user *saved;
3279 	int err;
3280 
3281 	if (get_user_ifreq(&ifr, NULL, uifr32))
3282 		return -EFAULT;
3283 
3284 	if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3285 		return -EFAULT;
3286 
3287 	saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3288 	ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3289 
3290 	err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3291 	if (!err) {
3292 		ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3293 		if (put_user_ifreq(&ifr, uifr32))
3294 			err = -EFAULT;
3295 	}
3296 	return err;
3297 }
3298 
3299 /* 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)3300 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3301 				 struct compat_ifreq __user *u_ifreq32)
3302 {
3303 	struct ifreq ifreq;
3304 	void __user *data;
3305 
3306 	if (!is_socket_ioctl_cmd(cmd))
3307 		return -ENOTTY;
3308 	if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3309 		return -EFAULT;
3310 	ifreq.ifr_data = data;
3311 
3312 	return dev_ioctl(net, cmd, &ifreq, data, NULL);
3313 }
3314 
compat_sock_ioctl_trans(struct file * file,struct socket * sock,unsigned int cmd,unsigned long arg)3315 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3316 			 unsigned int cmd, unsigned long arg)
3317 {
3318 	void __user *argp = compat_ptr(arg);
3319 	struct sock *sk = sock->sk;
3320 	struct net *net = sock_net(sk);
3321 
3322 	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3323 		return sock_ioctl(file, cmd, (unsigned long)argp);
3324 
3325 	switch (cmd) {
3326 	case SIOCWANDEV:
3327 		return compat_siocwandev(net, argp);
3328 	case SIOCGSTAMP_OLD:
3329 	case SIOCGSTAMPNS_OLD:
3330 		if (!sock->ops->gettstamp)
3331 			return -ENOIOCTLCMD;
3332 		return sock->ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3333 					    !COMPAT_USE_64BIT_TIME);
3334 
3335 	case SIOCETHTOOL:
3336 	case SIOCBONDSLAVEINFOQUERY:
3337 	case SIOCBONDINFOQUERY:
3338 	case SIOCSHWTSTAMP:
3339 	case SIOCGHWTSTAMP:
3340 		return compat_ifr_data_ioctl(net, cmd, argp);
3341 
3342 	case FIOSETOWN:
3343 	case SIOCSPGRP:
3344 	case FIOGETOWN:
3345 	case SIOCGPGRP:
3346 	case SIOCBRADDBR:
3347 	case SIOCBRDELBR:
3348 	case SIOCGIFVLAN:
3349 	case SIOCSIFVLAN:
3350 	case SIOCGSKNS:
3351 	case SIOCGSTAMP_NEW:
3352 	case SIOCGSTAMPNS_NEW:
3353 	case SIOCGIFCONF:
3354 	case SIOCSIFBR:
3355 	case SIOCGIFBR:
3356 		return sock_ioctl(file, cmd, arg);
3357 
3358 	case SIOCGIFFLAGS:
3359 	case SIOCSIFFLAGS:
3360 	case SIOCGIFMAP:
3361 	case SIOCSIFMAP:
3362 	case SIOCGIFMETRIC:
3363 	case SIOCSIFMETRIC:
3364 	case SIOCGIFMTU:
3365 	case SIOCSIFMTU:
3366 	case SIOCGIFMEM:
3367 	case SIOCSIFMEM:
3368 	case SIOCGIFHWADDR:
3369 	case SIOCSIFHWADDR:
3370 	case SIOCADDMULTI:
3371 	case SIOCDELMULTI:
3372 	case SIOCGIFINDEX:
3373 	case SIOCGIFADDR:
3374 	case SIOCSIFADDR:
3375 	case SIOCSIFHWBROADCAST:
3376 	case SIOCDIFADDR:
3377 	case SIOCGIFBRDADDR:
3378 	case SIOCSIFBRDADDR:
3379 	case SIOCGIFDSTADDR:
3380 	case SIOCSIFDSTADDR:
3381 	case SIOCGIFNETMASK:
3382 	case SIOCSIFNETMASK:
3383 	case SIOCSIFPFLAGS:
3384 	case SIOCGIFPFLAGS:
3385 	case SIOCGIFTXQLEN:
3386 	case SIOCSIFTXQLEN:
3387 	case SIOCBRADDIF:
3388 	case SIOCBRDELIF:
3389 	case SIOCGIFNAME:
3390 	case SIOCSIFNAME:
3391 	case SIOCGMIIPHY:
3392 	case SIOCGMIIREG:
3393 	case SIOCSMIIREG:
3394 	case SIOCBONDENSLAVE:
3395 	case SIOCBONDRELEASE:
3396 	case SIOCBONDSETHWADDR:
3397 	case SIOCBONDCHANGEACTIVE:
3398 	case SIOCSARP:
3399 	case SIOCGARP:
3400 	case SIOCDARP:
3401 	case SIOCOUTQ:
3402 	case SIOCOUTQNSD:
3403 	case SIOCATMARK:
3404 		return sock_do_ioctl(net, sock, cmd, arg);
3405 	}
3406 
3407 	return -ENOIOCTLCMD;
3408 }
3409 
compat_sock_ioctl(struct file * file,unsigned int cmd,unsigned long arg)3410 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3411 			      unsigned long arg)
3412 {
3413 	struct socket *sock = file->private_data;
3414 	int ret = -ENOIOCTLCMD;
3415 	struct sock *sk;
3416 	struct net *net;
3417 
3418 	sk = sock->sk;
3419 	net = sock_net(sk);
3420 
3421 	if (sock->ops->compat_ioctl)
3422 		ret = sock->ops->compat_ioctl(sock, cmd, arg);
3423 
3424 	if (ret == -ENOIOCTLCMD &&
3425 	    (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3426 		ret = compat_wext_handle_ioctl(net, cmd, arg);
3427 
3428 	if (ret == -ENOIOCTLCMD)
3429 		ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3430 
3431 	return ret;
3432 }
3433 #endif
3434 
3435 /**
3436  *	kernel_bind - bind an address to a socket (kernel space)
3437  *	@sock: socket
3438  *	@addr: address
3439  *	@addrlen: length of address
3440  *
3441  *	Returns 0 or an error.
3442  */
3443 
kernel_bind(struct socket * sock,struct sockaddr * addr,int addrlen)3444 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3445 {
3446 	return sock->ops->bind(sock, addr, addrlen);
3447 }
3448 EXPORT_SYMBOL(kernel_bind);
3449 
3450 /**
3451  *	kernel_listen - move socket to listening state (kernel space)
3452  *	@sock: socket
3453  *	@backlog: pending connections queue size
3454  *
3455  *	Returns 0 or an error.
3456  */
3457 
kernel_listen(struct socket * sock,int backlog)3458 int kernel_listen(struct socket *sock, int backlog)
3459 {
3460 	return sock->ops->listen(sock, backlog);
3461 }
3462 EXPORT_SYMBOL(kernel_listen);
3463 
3464 /**
3465  *	kernel_accept - accept a connection (kernel space)
3466  *	@sock: listening socket
3467  *	@newsock: new connected socket
3468  *	@flags: flags
3469  *
3470  *	@flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3471  *	If it fails, @newsock is guaranteed to be %NULL.
3472  *	Returns 0 or an error.
3473  */
3474 
kernel_accept(struct socket * sock,struct socket ** newsock,int flags)3475 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3476 {
3477 	struct sock *sk = sock->sk;
3478 	int err;
3479 
3480 	err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3481 			       newsock);
3482 	if (err < 0)
3483 		goto done;
3484 
3485 	err = sock->ops->accept(sock, *newsock, flags, true);
3486 	if (err < 0) {
3487 		sock_release(*newsock);
3488 		*newsock = NULL;
3489 		goto done;
3490 	}
3491 
3492 	(*newsock)->ops = sock->ops;
3493 	__module_get((*newsock)->ops->owner);
3494 
3495 done:
3496 	return err;
3497 }
3498 EXPORT_SYMBOL(kernel_accept);
3499 
3500 /**
3501  *	kernel_connect - connect a socket (kernel space)
3502  *	@sock: socket
3503  *	@addr: address
3504  *	@addrlen: address length
3505  *	@flags: flags (O_NONBLOCK, ...)
3506  *
3507  *	For datagram sockets, @addr is the address to which datagrams are sent
3508  *	by default, and the only address from which datagrams are received.
3509  *	For stream sockets, attempts to connect to @addr.
3510  *	Returns 0 or an error code.
3511  */
3512 
kernel_connect(struct socket * sock,struct sockaddr * addr,int addrlen,int flags)3513 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3514 		   int flags)
3515 {
3516 	return sock->ops->connect(sock, addr, addrlen, flags);
3517 }
3518 EXPORT_SYMBOL(kernel_connect);
3519 
3520 /**
3521  *	kernel_getsockname - get the address which the socket is bound (kernel space)
3522  *	@sock: socket
3523  *	@addr: address holder
3524  *
3525  * 	Fills the @addr pointer with the address which the socket is bound.
3526  *	Returns the length of the address in bytes or an error code.
3527  */
3528 
kernel_getsockname(struct socket * sock,struct sockaddr * addr)3529 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3530 {
3531 	return sock->ops->getname(sock, addr, 0);
3532 }
3533 EXPORT_SYMBOL(kernel_getsockname);
3534 
3535 /**
3536  *	kernel_getpeername - get the address which the socket is connected (kernel space)
3537  *	@sock: socket
3538  *	@addr: address holder
3539  *
3540  * 	Fills the @addr pointer with the address which the socket is connected.
3541  *	Returns the length of the address in bytes or an error code.
3542  */
3543 
kernel_getpeername(struct socket * sock,struct sockaddr * addr)3544 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3545 {
3546 	return sock->ops->getname(sock, addr, 1);
3547 }
3548 EXPORT_SYMBOL(kernel_getpeername);
3549 
3550 /**
3551  *	kernel_sendpage - send a &page through a socket (kernel space)
3552  *	@sock: socket
3553  *	@page: page
3554  *	@offset: page offset
3555  *	@size: total size in bytes
3556  *	@flags: flags (MSG_DONTWAIT, ...)
3557  *
3558  *	Returns the total amount sent in bytes or an error.
3559  */
3560 
kernel_sendpage(struct socket * sock,struct page * page,int offset,size_t size,int flags)3561 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3562 		    size_t size, int flags)
3563 {
3564 	if (sock->ops->sendpage) {
3565 		/* Warn in case the improper page to zero-copy send */
3566 		WARN_ONCE(!sendpage_ok(page), "improper page for zero-copy send");
3567 		return sock->ops->sendpage(sock, page, offset, size, flags);
3568 	}
3569 	return sock_no_sendpage(sock, page, offset, size, flags);
3570 }
3571 EXPORT_SYMBOL(kernel_sendpage);
3572 
3573 /**
3574  *	kernel_sendpage_locked - send a &page through the locked sock (kernel space)
3575  *	@sk: sock
3576  *	@page: page
3577  *	@offset: page offset
3578  *	@size: total size in bytes
3579  *	@flags: flags (MSG_DONTWAIT, ...)
3580  *
3581  *	Returns the total amount sent in bytes or an error.
3582  *	Caller must hold @sk.
3583  */
3584 
kernel_sendpage_locked(struct sock * sk,struct page * page,int offset,size_t size,int flags)3585 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3586 			   size_t size, int flags)
3587 {
3588 	struct socket *sock = sk->sk_socket;
3589 
3590 	if (sock->ops->sendpage_locked)
3591 		return sock->ops->sendpage_locked(sk, page, offset, size,
3592 						  flags);
3593 
3594 	return sock_no_sendpage_locked(sk, page, offset, size, flags);
3595 }
3596 EXPORT_SYMBOL(kernel_sendpage_locked);
3597 
3598 /**
3599  *	kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3600  *	@sock: socket
3601  *	@how: connection part
3602  *
3603  *	Returns 0 or an error.
3604  */
3605 
kernel_sock_shutdown(struct socket * sock,enum sock_shutdown_cmd how)3606 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3607 {
3608 	return sock->ops->shutdown(sock, how);
3609 }
3610 EXPORT_SYMBOL(kernel_sock_shutdown);
3611 
3612 /**
3613  *	kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3614  *	@sk: socket
3615  *
3616  *	This routine returns the IP overhead imposed by a socket i.e.
3617  *	the length of the underlying IP header, depending on whether
3618  *	this is an IPv4 or IPv6 socket and the length from IP options turned
3619  *	on at the socket. Assumes that the caller has a lock on the socket.
3620  */
3621 
kernel_sock_ip_overhead(struct sock * sk)3622 u32 kernel_sock_ip_overhead(struct sock *sk)
3623 {
3624 	struct inet_sock *inet;
3625 	struct ip_options_rcu *opt;
3626 	u32 overhead = 0;
3627 #if IS_ENABLED(CONFIG_IPV6)
3628 	struct ipv6_pinfo *np;
3629 	struct ipv6_txoptions *optv6 = NULL;
3630 #endif /* IS_ENABLED(CONFIG_IPV6) */
3631 
3632 	if (!sk)
3633 		return overhead;
3634 
3635 	switch (sk->sk_family) {
3636 	case AF_INET:
3637 		inet = inet_sk(sk);
3638 		overhead += sizeof(struct iphdr);
3639 		opt = rcu_dereference_protected(inet->inet_opt,
3640 						sock_owned_by_user(sk));
3641 		if (opt)
3642 			overhead += opt->opt.optlen;
3643 		return overhead;
3644 #if IS_ENABLED(CONFIG_IPV6)
3645 	case AF_INET6:
3646 		np = inet6_sk(sk);
3647 		overhead += sizeof(struct ipv6hdr);
3648 		if (np)
3649 			optv6 = rcu_dereference_protected(np->opt,
3650 							  sock_owned_by_user(sk));
3651 		if (optv6)
3652 			overhead += (optv6->opt_flen + optv6->opt_nflen);
3653 		return overhead;
3654 #endif /* IS_ENABLED(CONFIG_IPV6) */
3655 	default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3656 		return overhead;
3657 	}
3658 }
3659 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3660