1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * NET An implementation of the SOCKET network access protocol.
4 *
5 * Version: @(#)socket.c 1.1.93 18/02/95
6 *
7 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Ross Biro
9 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
10 *
11 * Fixes:
12 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * shutdown()
14 * Alan Cox : verify_area() fixes
15 * Alan Cox : Removed DDI
16 * Jonathan Kamens : SOCK_DGRAM reconnect bug
17 * Alan Cox : Moved a load of checks to the very
18 * top level.
19 * Alan Cox : Move address structures to/from user
20 * mode above the protocol layers.
21 * Rob Janssen : Allow 0 length sends.
22 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * tty drivers).
24 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
25 * Jeff Uphoff : Made max number of sockets command-line
26 * configurable.
27 * Matti Aarnio : Made the number of sockets dynamic,
28 * to be allocated when needed, and mr.
29 * Uphoff's max is used as max to be
30 * allowed to allocate.
31 * Linus : Argh. removed all the socket allocation
32 * altogether: it's in the inode now.
33 * Alan Cox : Made sock_alloc()/sock_release() public
34 * for NetROM and future kernel nfsd type
35 * stuff.
36 * Alan Cox : sendmsg/recvmsg basics.
37 * Tom Dyas : Export net symbols.
38 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
39 * Alan Cox : Added thread locking to sys_* calls
40 * for sockets. May have errors at the
41 * moment.
42 * Kevin Buhr : Fixed the dumb errors in the above.
43 * Andi Kleen : Some small cleanups, optimizations,
44 * and fixed a copy_from_user() bug.
45 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
46 * Tigran Aivazian : Made listen(2) backlog sanity checks
47 * protocol-independent
48 *
49 * This module is effectively the top level interface to the BSD socket
50 * paradigm.
51 *
52 * Based upon Swansea University Computer Society NET3.039
53 */
54
55 #include <linux/bpf-cgroup.h>
56 #include <linux/ethtool.h>
57 #include <linux/mm.h>
58 #include <linux/socket.h>
59 #include <linux/file.h>
60 #include <linux/splice.h>
61 #include <linux/net.h>
62 #include <linux/interrupt.h>
63 #include <linux/thread_info.h>
64 #include <linux/rcupdate.h>
65 #include <linux/netdevice.h>
66 #include <linux/proc_fs.h>
67 #include <linux/seq_file.h>
68 #include <linux/mutex.h>
69 #include <linux/if_bridge.h>
70 #include <linux/if_vlan.h>
71 #include <linux/ptp_classify.h>
72 #include <linux/init.h>
73 #include <linux/poll.h>
74 #include <linux/cache.h>
75 #include <linux/module.h>
76 #include <linux/highmem.h>
77 #include <linux/mount.h>
78 #include <linux/pseudo_fs.h>
79 #include <linux/security.h>
80 #include <linux/syscalls.h>
81 #include <linux/compat.h>
82 #include <linux/kmod.h>
83 #include <linux/audit.h>
84 #include <linux/wireless.h>
85 #include <linux/nsproxy.h>
86 #include <linux/magic.h>
87 #include <linux/slab.h>
88 #include <linux/xattr.h>
89 #include <linux/nospec.h>
90 #include <linux/indirect_call_wrapper.h>
91 #include <linux/io_uring/net.h>
92
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
95
96 #include <net/compat.h>
97 #include <net/wext.h>
98 #include <net/cls_cgroup.h>
99
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/termios.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110 #include <linux/ptp_clock_kernel.h>
111 #include <trace/events/sock.h>
112
113 #ifdef CONFIG_NET_RX_BUSY_POLL
114 unsigned int sysctl_net_busy_read __read_mostly;
115 unsigned int sysctl_net_busy_poll __read_mostly;
116 #endif
117
118 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
119 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
120 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
121
122 static int sock_close(struct inode *inode, struct file *file);
123 static __poll_t sock_poll(struct file *file,
124 struct poll_table_struct *wait);
125 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
126 #ifdef CONFIG_COMPAT
127 static long compat_sock_ioctl(struct file *file,
128 unsigned int cmd, unsigned long arg);
129 #endif
130 static int sock_fasync(int fd, struct file *filp, int on);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
133 unsigned int flags);
134 static void sock_splice_eof(struct file *file);
135
136 #ifdef CONFIG_PROC_FS
137 static void sock_show_fdinfo(struct seq_file *m, struct file *f)
138 {
139 struct socket *sock = f->private_data;
140 const struct proto_ops *ops = READ_ONCE(sock->ops);
141
142 if (ops->show_fdinfo)
143 ops->show_fdinfo(m, sock);
144 }
145 #else
146 #define sock_show_fdinfo NULL
147 #endif
148
149 /*
150 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
151 * in the operation structures but are done directly via the socketcall() multiplexor.
152 */
153
154 static const struct file_operations socket_file_ops = {
155 .owner = THIS_MODULE,
156 .llseek = no_llseek,
157 .read_iter = sock_read_iter,
158 .write_iter = sock_write_iter,
159 .poll = sock_poll,
160 .unlocked_ioctl = sock_ioctl,
161 #ifdef CONFIG_COMPAT
162 .compat_ioctl = compat_sock_ioctl,
163 #endif
164 .uring_cmd = io_uring_cmd_sock,
165 .mmap = sock_mmap,
166 .release = sock_close,
167 .fasync = sock_fasync,
168 .splice_write = splice_to_socket,
169 .splice_read = sock_splice_read,
170 .splice_eof = sock_splice_eof,
171 .show_fdinfo = sock_show_fdinfo,
172 };
173
174 static const char * const pf_family_names[] = {
175 [PF_UNSPEC] = "PF_UNSPEC",
176 [PF_UNIX] = "PF_UNIX/PF_LOCAL",
177 [PF_INET] = "PF_INET",
178 [PF_AX25] = "PF_AX25",
179 [PF_IPX] = "PF_IPX",
180 [PF_APPLETALK] = "PF_APPLETALK",
181 [PF_NETROM] = "PF_NETROM",
182 [PF_BRIDGE] = "PF_BRIDGE",
183 [PF_ATMPVC] = "PF_ATMPVC",
184 [PF_X25] = "PF_X25",
185 [PF_INET6] = "PF_INET6",
186 [PF_ROSE] = "PF_ROSE",
187 [PF_DECnet] = "PF_DECnet",
188 [PF_NETBEUI] = "PF_NETBEUI",
189 [PF_SECURITY] = "PF_SECURITY",
190 [PF_KEY] = "PF_KEY",
191 [PF_NETLINK] = "PF_NETLINK/PF_ROUTE",
192 [PF_PACKET] = "PF_PACKET",
193 [PF_ASH] = "PF_ASH",
194 [PF_ECONET] = "PF_ECONET",
195 [PF_ATMSVC] = "PF_ATMSVC",
196 [PF_RDS] = "PF_RDS",
197 [PF_SNA] = "PF_SNA",
198 [PF_IRDA] = "PF_IRDA",
199 [PF_PPPOX] = "PF_PPPOX",
200 [PF_WANPIPE] = "PF_WANPIPE",
201 [PF_LLC] = "PF_LLC",
202 [PF_IB] = "PF_IB",
203 [PF_MPLS] = "PF_MPLS",
204 [PF_CAN] = "PF_CAN",
205 [PF_TIPC] = "PF_TIPC",
206 [PF_BLUETOOTH] = "PF_BLUETOOTH",
207 [PF_IUCV] = "PF_IUCV",
208 [PF_RXRPC] = "PF_RXRPC",
209 [PF_ISDN] = "PF_ISDN",
210 [PF_PHONET] = "PF_PHONET",
211 [PF_IEEE802154] = "PF_IEEE802154",
212 [PF_CAIF] = "PF_CAIF",
213 [PF_ALG] = "PF_ALG",
214 [PF_NFC] = "PF_NFC",
215 [PF_VSOCK] = "PF_VSOCK",
216 [PF_KCM] = "PF_KCM",
217 [PF_QIPCRTR] = "PF_QIPCRTR",
218 [PF_SMC] = "PF_SMC",
219 [PF_XDP] = "PF_XDP",
220 [PF_MCTP] = "PF_MCTP",
221 };
222
223 /*
224 * The protocol list. Each protocol is registered in here.
225 */
226
227 static DEFINE_SPINLOCK(net_family_lock);
228 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
229
230 /*
231 * Support routines.
232 * Move socket addresses back and forth across the kernel/user
233 * divide and look after the messy bits.
234 */
235
236 /**
237 * move_addr_to_kernel - copy a socket address into kernel space
238 * @uaddr: Address in user space
239 * @kaddr: Address in kernel space
240 * @ulen: Length in user space
241 *
242 * The address is copied into kernel space. If the provided address is
243 * too long an error code of -EINVAL is returned. If the copy gives
244 * invalid addresses -EFAULT is returned. On a success 0 is returned.
245 */
246
247 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
248 {
249 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
250 return -EINVAL;
251 if (ulen == 0)
252 return 0;
253 if (copy_from_user(kaddr, uaddr, ulen))
254 return -EFAULT;
255 return audit_sockaddr(ulen, kaddr);
256 }
257
258 /**
259 * move_addr_to_user - copy an address to user space
260 * @kaddr: kernel space address
261 * @klen: length of address in kernel
262 * @uaddr: user space address
263 * @ulen: pointer to user length field
264 *
265 * The value pointed to by ulen on entry is the buffer length available.
266 * This is overwritten with the buffer space used. -EINVAL is returned
267 * if an overlong buffer is specified or a negative buffer size. -EFAULT
268 * is returned if either the buffer or the length field are not
269 * accessible.
270 * After copying the data up to the limit the user specifies, the true
271 * length of the data is written over the length limit the user
272 * specified. Zero is returned for a success.
273 */
274
275 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
276 void __user *uaddr, int __user *ulen)
277 {
278 int err;
279 int len;
280
281 BUG_ON(klen > sizeof(struct sockaddr_storage));
282 err = get_user(len, ulen);
283 if (err)
284 return err;
285 if (len > klen)
286 len = klen;
287 if (len < 0)
288 return -EINVAL;
289 if (len) {
290 if (audit_sockaddr(klen, kaddr))
291 return -ENOMEM;
292 if (copy_to_user(uaddr, kaddr, len))
293 return -EFAULT;
294 }
295 /*
296 * "fromlen shall refer to the value before truncation.."
297 * 1003.1g
298 */
299 return __put_user(klen, ulen);
300 }
301
302 static struct kmem_cache *sock_inode_cachep __ro_after_init;
303
304 static struct inode *sock_alloc_inode(struct super_block *sb)
305 {
306 struct socket_alloc *ei;
307
308 ei = alloc_inode_sb(sb, sock_inode_cachep, GFP_KERNEL);
309 if (!ei)
310 return NULL;
311 init_waitqueue_head(&ei->socket.wq.wait);
312 ei->socket.wq.fasync_list = NULL;
313 ei->socket.wq.flags = 0;
314
315 ei->socket.state = SS_UNCONNECTED;
316 ei->socket.flags = 0;
317 ei->socket.ops = NULL;
318 ei->socket.sk = NULL;
319 ei->socket.file = NULL;
320
321 return &ei->vfs_inode;
322 }
323
324 static void sock_free_inode(struct inode *inode)
325 {
326 struct socket_alloc *ei;
327
328 ei = container_of(inode, struct socket_alloc, vfs_inode);
329 kmem_cache_free(sock_inode_cachep, ei);
330 }
331
332 static void init_once(void *foo)
333 {
334 struct socket_alloc *ei = (struct socket_alloc *)foo;
335
336 inode_init_once(&ei->vfs_inode);
337 }
338
339 static void init_inodecache(void)
340 {
341 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
342 sizeof(struct socket_alloc),
343 0,
344 (SLAB_HWCACHE_ALIGN |
345 SLAB_RECLAIM_ACCOUNT |
346 SLAB_ACCOUNT),
347 init_once);
348 BUG_ON(sock_inode_cachep == NULL);
349 }
350
351 static const struct super_operations sockfs_ops = {
352 .alloc_inode = sock_alloc_inode,
353 .free_inode = sock_free_inode,
354 .statfs = simple_statfs,
355 };
356
357 /*
358 * sockfs_dname() is called from d_path().
359 */
360 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
361 {
362 return dynamic_dname(buffer, buflen, "socket:[%lu]",
363 d_inode(dentry)->i_ino);
364 }
365
366 static const struct dentry_operations sockfs_dentry_operations = {
367 .d_dname = sockfs_dname,
368 };
369
370 static int sockfs_xattr_get(const struct xattr_handler *handler,
371 struct dentry *dentry, struct inode *inode,
372 const char *suffix, void *value, size_t size)
373 {
374 if (value) {
375 if (dentry->d_name.len + 1 > size)
376 return -ERANGE;
377 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
378 }
379 return dentry->d_name.len + 1;
380 }
381
382 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
383 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
384 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
385
386 static const struct xattr_handler sockfs_xattr_handler = {
387 .name = XATTR_NAME_SOCKPROTONAME,
388 .get = sockfs_xattr_get,
389 };
390
391 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
392 struct mnt_idmap *idmap,
393 struct dentry *dentry, struct inode *inode,
394 const char *suffix, const void *value,
395 size_t size, int flags)
396 {
397 /* Handled by LSM. */
398 return -EAGAIN;
399 }
400
401 static const struct xattr_handler sockfs_security_xattr_handler = {
402 .prefix = XATTR_SECURITY_PREFIX,
403 .set = sockfs_security_xattr_set,
404 };
405
406 static const struct xattr_handler * const sockfs_xattr_handlers[] = {
407 &sockfs_xattr_handler,
408 &sockfs_security_xattr_handler,
409 NULL
410 };
411
412 static int sockfs_init_fs_context(struct fs_context *fc)
413 {
414 struct pseudo_fs_context *ctx = init_pseudo(fc, SOCKFS_MAGIC);
415 if (!ctx)
416 return -ENOMEM;
417 ctx->ops = &sockfs_ops;
418 ctx->dops = &sockfs_dentry_operations;
419 ctx->xattr = sockfs_xattr_handlers;
420 return 0;
421 }
422
423 static struct vfsmount *sock_mnt __read_mostly;
424
425 static struct file_system_type sock_fs_type = {
426 .name = "sockfs",
427 .init_fs_context = sockfs_init_fs_context,
428 .kill_sb = kill_anon_super,
429 };
430
431 /*
432 * Obtains the first available file descriptor and sets it up for use.
433 *
434 * These functions create file structures and maps them to fd space
435 * of the current process. On success it returns file descriptor
436 * and file struct implicitly stored in sock->file.
437 * Note that another thread may close file descriptor before we return
438 * from this function. We use the fact that now we do not refer
439 * to socket after mapping. If one day we will need it, this
440 * function will increment ref. count on file by 1.
441 *
442 * In any case returned fd MAY BE not valid!
443 * This race condition is unavoidable
444 * with shared fd spaces, we cannot solve it inside kernel,
445 * but we take care of internal coherence yet.
446 */
447
448 /**
449 * sock_alloc_file - Bind a &socket to a &file
450 * @sock: socket
451 * @flags: file status flags
452 * @dname: protocol name
453 *
454 * Returns the &file bound with @sock, implicitly storing it
455 * in sock->file. If dname is %NULL, sets to "".
456 *
457 * On failure @sock is released, and an ERR pointer is returned.
458 *
459 * This function uses GFP_KERNEL internally.
460 */
461
462 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
463 {
464 struct file *file;
465
466 if (!dname)
467 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
468
469 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
470 O_RDWR | (flags & O_NONBLOCK),
471 &socket_file_ops);
472 if (IS_ERR(file)) {
473 sock_release(sock);
474 return file;
475 }
476
477 file->f_mode |= FMODE_NOWAIT;
478 sock->file = file;
479 file->private_data = sock;
480 stream_open(SOCK_INODE(sock), file);
481 return file;
482 }
483 EXPORT_SYMBOL(sock_alloc_file);
484
485 static int sock_map_fd(struct socket *sock, int flags)
486 {
487 struct file *newfile;
488 int fd = get_unused_fd_flags(flags);
489 if (unlikely(fd < 0)) {
490 sock_release(sock);
491 return fd;
492 }
493
494 newfile = sock_alloc_file(sock, flags, NULL);
495 if (!IS_ERR(newfile)) {
496 fd_install(fd, newfile);
497 return fd;
498 }
499
500 put_unused_fd(fd);
501 return PTR_ERR(newfile);
502 }
503
504 /**
505 * sock_from_file - Return the &socket bounded to @file.
506 * @file: file
507 *
508 * On failure returns %NULL.
509 */
510
511 struct socket *sock_from_file(struct file *file)
512 {
513 if (file->f_op == &socket_file_ops)
514 return file->private_data; /* set in sock_alloc_file */
515
516 return NULL;
517 }
518 EXPORT_SYMBOL(sock_from_file);
519
520 /**
521 * sockfd_lookup - Go from a file number to its socket slot
522 * @fd: file handle
523 * @err: pointer to an error code return
524 *
525 * The file handle passed in is locked and the socket it is bound
526 * to is returned. If an error occurs the err pointer is overwritten
527 * with a negative errno code and NULL is returned. The function checks
528 * for both invalid handles and passing a handle which is not a socket.
529 *
530 * On a success the socket object pointer is returned.
531 */
532
533 struct socket *sockfd_lookup(int fd, int *err)
534 {
535 struct file *file;
536 struct socket *sock;
537
538 file = fget(fd);
539 if (!file) {
540 *err = -EBADF;
541 return NULL;
542 }
543
544 sock = sock_from_file(file);
545 if (!sock) {
546 *err = -ENOTSOCK;
547 fput(file);
548 }
549 return sock;
550 }
551 EXPORT_SYMBOL(sockfd_lookup);
552
553 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
554 {
555 struct fd f = fdget(fd);
556 struct socket *sock;
557
558 *err = -EBADF;
559 if (f.file) {
560 sock = sock_from_file(f.file);
561 if (likely(sock)) {
562 *fput_needed = f.flags & FDPUT_FPUT;
563 return sock;
564 }
565 *err = -ENOTSOCK;
566 fdput(f);
567 }
568 return NULL;
569 }
570
571 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
572 size_t size)
573 {
574 ssize_t len;
575 ssize_t used = 0;
576
577 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
578 if (len < 0)
579 return len;
580 used += len;
581 if (buffer) {
582 if (size < used)
583 return -ERANGE;
584 buffer += len;
585 }
586
587 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
588 used += len;
589 if (buffer) {
590 if (size < used)
591 return -ERANGE;
592 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
593 buffer += len;
594 }
595
596 return used;
597 }
598
599 static int sockfs_setattr(struct mnt_idmap *idmap,
600 struct dentry *dentry, struct iattr *iattr)
601 {
602 int err = simple_setattr(&nop_mnt_idmap, dentry, iattr);
603
604 if (!err && (iattr->ia_valid & ATTR_UID)) {
605 struct socket *sock = SOCKET_I(d_inode(dentry));
606
607 if (sock->sk)
608 sock->sk->sk_uid = iattr->ia_uid;
609 else
610 err = -ENOENT;
611 }
612
613 return err;
614 }
615
616 static const struct inode_operations sockfs_inode_ops = {
617 .listxattr = sockfs_listxattr,
618 .setattr = sockfs_setattr,
619 };
620
621 /**
622 * sock_alloc - allocate a socket
623 *
624 * Allocate a new inode and socket object. The two are bound together
625 * and initialised. The socket is then returned. If we are out of inodes
626 * NULL is returned. This functions uses GFP_KERNEL internally.
627 */
628
629 struct socket *sock_alloc(void)
630 {
631 struct inode *inode;
632 struct socket *sock;
633
634 inode = new_inode_pseudo(sock_mnt->mnt_sb);
635 if (!inode)
636 return NULL;
637
638 sock = SOCKET_I(inode);
639
640 inode->i_ino = get_next_ino();
641 inode->i_mode = S_IFSOCK | S_IRWXUGO;
642 inode->i_uid = current_fsuid();
643 inode->i_gid = current_fsgid();
644 inode->i_op = &sockfs_inode_ops;
645
646 return sock;
647 }
648 EXPORT_SYMBOL(sock_alloc);
649
650 static void __sock_release(struct socket *sock, struct inode *inode)
651 {
652 const struct proto_ops *ops = READ_ONCE(sock->ops);
653
654 if (ops) {
655 struct module *owner = ops->owner;
656
657 if (inode)
658 inode_lock(inode);
659 ops->release(sock);
660 sock->sk = NULL;
661 if (inode)
662 inode_unlock(inode);
663 sock->ops = NULL;
664 module_put(owner);
665 }
666
667 if (sock->wq.fasync_list)
668 pr_err("%s: fasync list not empty!\n", __func__);
669
670 if (!sock->file) {
671 iput(SOCK_INODE(sock));
672 return;
673 }
674 sock->file = NULL;
675 }
676
677 /**
678 * sock_release - close a socket
679 * @sock: socket to close
680 *
681 * The socket is released from the protocol stack if it has a release
682 * callback, and the inode is then released if the socket is bound to
683 * an inode not a file.
684 */
685 void sock_release(struct socket *sock)
686 {
687 __sock_release(sock, NULL);
688 }
689 EXPORT_SYMBOL(sock_release);
690
691 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
692 {
693 u8 flags = *tx_flags;
694
695 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE) {
696 flags |= SKBTX_HW_TSTAMP;
697
698 /* PTP hardware clocks can provide a free running cycle counter
699 * as a time base for virtual clocks. Tell driver to use the
700 * free running cycle counter for timestamp if socket is bound
701 * to virtual clock.
702 */
703 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
704 flags |= SKBTX_HW_TSTAMP_USE_CYCLES;
705 }
706
707 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
708 flags |= SKBTX_SW_TSTAMP;
709
710 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
711 flags |= SKBTX_SCHED_TSTAMP;
712
713 *tx_flags = flags;
714 }
715 EXPORT_SYMBOL(__sock_tx_timestamp);
716
717 INDIRECT_CALLABLE_DECLARE(int inet_sendmsg(struct socket *, struct msghdr *,
718 size_t));
719 INDIRECT_CALLABLE_DECLARE(int inet6_sendmsg(struct socket *, struct msghdr *,
720 size_t));
721
722 static noinline void call_trace_sock_send_length(struct sock *sk, int ret,
723 int flags)
724 {
725 trace_sock_send_length(sk, ret, 0);
726 }
727
728 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
729 {
730 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->sendmsg, inet6_sendmsg,
731 inet_sendmsg, sock, msg,
732 msg_data_left(msg));
733 BUG_ON(ret == -EIOCBQUEUED);
734
735 if (trace_sock_send_length_enabled())
736 call_trace_sock_send_length(sock->sk, ret, 0);
737 return ret;
738 }
739
740 static int __sock_sendmsg(struct socket *sock, struct msghdr *msg)
741 {
742 int err = security_socket_sendmsg(sock, msg,
743 msg_data_left(msg));
744
745 return err ?: sock_sendmsg_nosec(sock, msg);
746 }
747
748 /**
749 * sock_sendmsg - send a message through @sock
750 * @sock: socket
751 * @msg: message to send
752 *
753 * Sends @msg through @sock, passing through LSM.
754 * Returns the number of bytes sent, or an error code.
755 */
756 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
757 {
758 struct sockaddr_storage *save_addr = (struct sockaddr_storage *)msg->msg_name;
759 struct sockaddr_storage address;
760 int save_len = msg->msg_namelen;
761 int ret;
762
763 if (msg->msg_name) {
764 memcpy(&address, msg->msg_name, msg->msg_namelen);
765 msg->msg_name = &address;
766 }
767
768 ret = __sock_sendmsg(sock, msg);
769 msg->msg_name = save_addr;
770 msg->msg_namelen = save_len;
771
772 return ret;
773 }
774 EXPORT_SYMBOL(sock_sendmsg);
775
776 /**
777 * kernel_sendmsg - send a message through @sock (kernel-space)
778 * @sock: socket
779 * @msg: message header
780 * @vec: kernel vec
781 * @num: vec array length
782 * @size: total message data size
783 *
784 * Builds the message data with @vec and sends it through @sock.
785 * Returns the number of bytes sent, or an error code.
786 */
787
788 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
789 struct kvec *vec, size_t num, size_t size)
790 {
791 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
792 return sock_sendmsg(sock, msg);
793 }
794 EXPORT_SYMBOL(kernel_sendmsg);
795
796 /**
797 * kernel_sendmsg_locked - send a message through @sock (kernel-space)
798 * @sk: sock
799 * @msg: message header
800 * @vec: output s/g array
801 * @num: output s/g array length
802 * @size: total message data size
803 *
804 * Builds the message data with @vec and sends it through @sock.
805 * Returns the number of bytes sent, or an error code.
806 * Caller must hold @sk.
807 */
808
809 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
810 struct kvec *vec, size_t num, size_t size)
811 {
812 struct socket *sock = sk->sk_socket;
813 const struct proto_ops *ops = READ_ONCE(sock->ops);
814
815 if (!ops->sendmsg_locked)
816 return sock_no_sendmsg_locked(sk, msg, size);
817
818 iov_iter_kvec(&msg->msg_iter, ITER_SOURCE, vec, num, size);
819
820 return ops->sendmsg_locked(sk, msg, msg_data_left(msg));
821 }
822 EXPORT_SYMBOL(kernel_sendmsg_locked);
823
824 static bool skb_is_err_queue(const struct sk_buff *skb)
825 {
826 /* pkt_type of skbs enqueued on the error queue are set to
827 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
828 * in recvmsg, since skbs received on a local socket will never
829 * have a pkt_type of PACKET_OUTGOING.
830 */
831 return skb->pkt_type == PACKET_OUTGOING;
832 }
833
834 /* On transmit, software and hardware timestamps are returned independently.
835 * As the two skb clones share the hardware timestamp, which may be updated
836 * before the software timestamp is received, a hardware TX timestamp may be
837 * returned only if there is no software TX timestamp. Ignore false software
838 * timestamps, which may be made in the __sock_recv_timestamp() call when the
839 * option SO_TIMESTAMP_OLD(NS) is enabled on the socket, even when the skb has a
840 * hardware timestamp.
841 */
842 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
843 {
844 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
845 }
846
847 static ktime_t get_timestamp(struct sock *sk, struct sk_buff *skb, int *if_index)
848 {
849 bool cycles = READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_BIND_PHC;
850 struct skb_shared_hwtstamps *shhwtstamps = skb_hwtstamps(skb);
851 struct net_device *orig_dev;
852 ktime_t hwtstamp;
853
854 rcu_read_lock();
855 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
856 if (orig_dev) {
857 *if_index = orig_dev->ifindex;
858 hwtstamp = netdev_get_tstamp(orig_dev, shhwtstamps, cycles);
859 } else {
860 hwtstamp = shhwtstamps->hwtstamp;
861 }
862 rcu_read_unlock();
863
864 return hwtstamp;
865 }
866
867 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb,
868 int if_index)
869 {
870 struct scm_ts_pktinfo ts_pktinfo;
871 struct net_device *orig_dev;
872
873 if (!skb_mac_header_was_set(skb))
874 return;
875
876 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
877
878 if (!if_index) {
879 rcu_read_lock();
880 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
881 if (orig_dev)
882 if_index = orig_dev->ifindex;
883 rcu_read_unlock();
884 }
885 ts_pktinfo.if_index = if_index;
886
887 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
888 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
889 sizeof(ts_pktinfo), &ts_pktinfo);
890 }
891
892 /*
893 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
894 */
895 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
896 struct sk_buff *skb)
897 {
898 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
899 int new_tstamp = sock_flag(sk, SOCK_TSTAMP_NEW);
900 struct scm_timestamping_internal tss;
901 int empty = 1, false_tstamp = 0;
902 struct skb_shared_hwtstamps *shhwtstamps =
903 skb_hwtstamps(skb);
904 int if_index;
905 ktime_t hwtstamp;
906 u32 tsflags;
907
908 /* Race occurred between timestamp enabling and packet
909 receiving. Fill in the current time for now. */
910 if (need_software_tstamp && skb->tstamp == 0) {
911 __net_timestamp(skb);
912 false_tstamp = 1;
913 }
914
915 if (need_software_tstamp) {
916 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
917 if (new_tstamp) {
918 struct __kernel_sock_timeval tv;
919
920 skb_get_new_timestamp(skb, &tv);
921 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_NEW,
922 sizeof(tv), &tv);
923 } else {
924 struct __kernel_old_timeval tv;
925
926 skb_get_timestamp(skb, &tv);
927 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMP_OLD,
928 sizeof(tv), &tv);
929 }
930 } else {
931 if (new_tstamp) {
932 struct __kernel_timespec ts;
933
934 skb_get_new_timestampns(skb, &ts);
935 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_NEW,
936 sizeof(ts), &ts);
937 } else {
938 struct __kernel_old_timespec ts;
939
940 skb_get_timestampns(skb, &ts);
941 put_cmsg(msg, SOL_SOCKET, SO_TIMESTAMPNS_OLD,
942 sizeof(ts), &ts);
943 }
944 }
945 }
946
947 memset(&tss, 0, sizeof(tss));
948 tsflags = READ_ONCE(sk->sk_tsflags);
949 if ((tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
950 ktime_to_timespec64_cond(skb->tstamp, tss.ts + 0))
951 empty = 0;
952 if (shhwtstamps &&
953 (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
954 !skb_is_swtx_tstamp(skb, false_tstamp)) {
955 if_index = 0;
956 if (skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP_NETDEV)
957 hwtstamp = get_timestamp(sk, skb, &if_index);
958 else
959 hwtstamp = shhwtstamps->hwtstamp;
960
961 if (tsflags & SOF_TIMESTAMPING_BIND_PHC)
962 hwtstamp = ptp_convert_timestamp(&hwtstamp,
963 READ_ONCE(sk->sk_bind_phc));
964
965 if (ktime_to_timespec64_cond(hwtstamp, tss.ts + 2)) {
966 empty = 0;
967
968 if ((tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
969 !skb_is_err_queue(skb))
970 put_ts_pktinfo(msg, skb, if_index);
971 }
972 }
973 if (!empty) {
974 if (sock_flag(sk, SOCK_TSTAMP_NEW))
975 put_cmsg_scm_timestamping64(msg, &tss);
976 else
977 put_cmsg_scm_timestamping(msg, &tss);
978
979 if (skb_is_err_queue(skb) && skb->len &&
980 SKB_EXT_ERR(skb)->opt_stats)
981 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
982 skb->len, skb->data);
983 }
984 }
985 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
986
987 #ifdef CONFIG_WIRELESS
988 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
989 struct sk_buff *skb)
990 {
991 int ack;
992
993 if (!sock_flag(sk, SOCK_WIFI_STATUS))
994 return;
995 if (!skb->wifi_acked_valid)
996 return;
997
998 ack = skb->wifi_acked;
999
1000 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
1001 }
1002 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
1003 #endif
1004
1005 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
1006 struct sk_buff *skb)
1007 {
1008 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
1009 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
1010 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
1011 }
1012
1013 static void sock_recv_mark(struct msghdr *msg, struct sock *sk,
1014 struct sk_buff *skb)
1015 {
1016 if (sock_flag(sk, SOCK_RCVMARK) && skb) {
1017 /* We must use a bounce buffer for CONFIG_HARDENED_USERCOPY=y */
1018 __u32 mark = skb->mark;
1019
1020 put_cmsg(msg, SOL_SOCKET, SO_MARK, sizeof(__u32), &mark);
1021 }
1022 }
1023
1024 void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk,
1025 struct sk_buff *skb)
1026 {
1027 sock_recv_timestamp(msg, sk, skb);
1028 sock_recv_drops(msg, sk, skb);
1029 sock_recv_mark(msg, sk, skb);
1030 }
1031 EXPORT_SYMBOL_GPL(__sock_recv_cmsgs);
1032
1033 INDIRECT_CALLABLE_DECLARE(int inet_recvmsg(struct socket *, struct msghdr *,
1034 size_t, int));
1035 INDIRECT_CALLABLE_DECLARE(int inet6_recvmsg(struct socket *, struct msghdr *,
1036 size_t, int));
1037
1038 static noinline void call_trace_sock_recv_length(struct sock *sk, int ret, int flags)
1039 {
1040 trace_sock_recv_length(sk, ret, flags);
1041 }
1042
1043 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
1044 int flags)
1045 {
1046 int ret = INDIRECT_CALL_INET(READ_ONCE(sock->ops)->recvmsg,
1047 inet6_recvmsg,
1048 inet_recvmsg, sock, msg,
1049 msg_data_left(msg), flags);
1050 if (trace_sock_recv_length_enabled())
1051 call_trace_sock_recv_length(sock->sk, ret, flags);
1052 return ret;
1053 }
1054
1055 /**
1056 * sock_recvmsg - receive a message from @sock
1057 * @sock: socket
1058 * @msg: message to receive
1059 * @flags: message flags
1060 *
1061 * Receives @msg from @sock, passing through LSM. Returns the total number
1062 * of bytes received, or an error.
1063 */
1064 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
1065 {
1066 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
1067
1068 return err ?: sock_recvmsg_nosec(sock, msg, flags);
1069 }
1070 EXPORT_SYMBOL(sock_recvmsg);
1071
1072 /**
1073 * kernel_recvmsg - Receive a message from a socket (kernel space)
1074 * @sock: The socket to receive the message from
1075 * @msg: Received message
1076 * @vec: Input s/g array for message data
1077 * @num: Size of input s/g array
1078 * @size: Number of bytes to read
1079 * @flags: Message flags (MSG_DONTWAIT, etc...)
1080 *
1081 * On return the msg structure contains the scatter/gather array passed in the
1082 * vec argument. The array is modified so that it consists of the unfilled
1083 * portion of the original array.
1084 *
1085 * The returned value is the total number of bytes received, or an error.
1086 */
1087
1088 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
1089 struct kvec *vec, size_t num, size_t size, int flags)
1090 {
1091 msg->msg_control_is_user = false;
1092 iov_iter_kvec(&msg->msg_iter, ITER_DEST, vec, num, size);
1093 return sock_recvmsg(sock, msg, flags);
1094 }
1095 EXPORT_SYMBOL(kernel_recvmsg);
1096
1097 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
1098 struct pipe_inode_info *pipe, size_t len,
1099 unsigned int flags)
1100 {
1101 struct socket *sock = file->private_data;
1102 const struct proto_ops *ops;
1103
1104 ops = READ_ONCE(sock->ops);
1105 if (unlikely(!ops->splice_read))
1106 return copy_splice_read(file, ppos, pipe, len, flags);
1107
1108 return ops->splice_read(sock, ppos, pipe, len, flags);
1109 }
1110
1111 static void sock_splice_eof(struct file *file)
1112 {
1113 struct socket *sock = file->private_data;
1114 const struct proto_ops *ops;
1115
1116 ops = READ_ONCE(sock->ops);
1117 if (ops->splice_eof)
1118 ops->splice_eof(sock);
1119 }
1120
1121 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
1122 {
1123 struct file *file = iocb->ki_filp;
1124 struct socket *sock = file->private_data;
1125 struct msghdr msg = {.msg_iter = *to,
1126 .msg_iocb = iocb};
1127 ssize_t res;
1128
1129 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1130 msg.msg_flags = MSG_DONTWAIT;
1131
1132 if (iocb->ki_pos != 0)
1133 return -ESPIPE;
1134
1135 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
1136 return 0;
1137
1138 res = sock_recvmsg(sock, &msg, msg.msg_flags);
1139 *to = msg.msg_iter;
1140 return res;
1141 }
1142
1143 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
1144 {
1145 struct file *file = iocb->ki_filp;
1146 struct socket *sock = file->private_data;
1147 struct msghdr msg = {.msg_iter = *from,
1148 .msg_iocb = iocb};
1149 ssize_t res;
1150
1151 if (iocb->ki_pos != 0)
1152 return -ESPIPE;
1153
1154 if (file->f_flags & O_NONBLOCK || (iocb->ki_flags & IOCB_NOWAIT))
1155 msg.msg_flags = MSG_DONTWAIT;
1156
1157 if (sock->type == SOCK_SEQPACKET)
1158 msg.msg_flags |= MSG_EOR;
1159
1160 res = __sock_sendmsg(sock, &msg);
1161 *from = msg.msg_iter;
1162 return res;
1163 }
1164
1165 /*
1166 * Atomic setting of ioctl hooks to avoid race
1167 * with module unload.
1168 */
1169
1170 static DEFINE_MUTEX(br_ioctl_mutex);
1171 static int (*br_ioctl_hook)(struct net *net, struct net_bridge *br,
1172 unsigned int cmd, struct ifreq *ifr,
1173 void __user *uarg);
1174
1175 void brioctl_set(int (*hook)(struct net *net, struct net_bridge *br,
1176 unsigned int cmd, struct ifreq *ifr,
1177 void __user *uarg))
1178 {
1179 mutex_lock(&br_ioctl_mutex);
1180 br_ioctl_hook = hook;
1181 mutex_unlock(&br_ioctl_mutex);
1182 }
1183 EXPORT_SYMBOL(brioctl_set);
1184
1185 int br_ioctl_call(struct net *net, struct net_bridge *br, unsigned int cmd,
1186 struct ifreq *ifr, void __user *uarg)
1187 {
1188 int err = -ENOPKG;
1189
1190 if (!br_ioctl_hook)
1191 request_module("bridge");
1192
1193 mutex_lock(&br_ioctl_mutex);
1194 if (br_ioctl_hook)
1195 err = br_ioctl_hook(net, br, cmd, ifr, uarg);
1196 mutex_unlock(&br_ioctl_mutex);
1197
1198 return err;
1199 }
1200
1201 static DEFINE_MUTEX(vlan_ioctl_mutex);
1202 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
1203
1204 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
1205 {
1206 mutex_lock(&vlan_ioctl_mutex);
1207 vlan_ioctl_hook = hook;
1208 mutex_unlock(&vlan_ioctl_mutex);
1209 }
1210 EXPORT_SYMBOL(vlan_ioctl_set);
1211
1212 static long sock_do_ioctl(struct net *net, struct socket *sock,
1213 unsigned int cmd, unsigned long arg)
1214 {
1215 const struct proto_ops *ops = READ_ONCE(sock->ops);
1216 struct ifreq ifr;
1217 bool need_copyout;
1218 int err;
1219 void __user *argp = (void __user *)arg;
1220 void __user *data;
1221
1222 err = ops->ioctl(sock, cmd, arg);
1223
1224 /*
1225 * If this ioctl is unknown try to hand it down
1226 * to the NIC driver.
1227 */
1228 if (err != -ENOIOCTLCMD)
1229 return err;
1230
1231 if (!is_socket_ioctl_cmd(cmd))
1232 return -ENOTTY;
1233
1234 if (get_user_ifreq(&ifr, &data, argp))
1235 return -EFAULT;
1236 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1237 if (!err && need_copyout)
1238 if (put_user_ifreq(&ifr, argp))
1239 return -EFAULT;
1240
1241 return err;
1242 }
1243
1244 /*
1245 * With an ioctl, arg may well be a user mode pointer, but we don't know
1246 * what to do with it - that's up to the protocol still.
1247 */
1248
1249 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
1250 {
1251 const struct proto_ops *ops;
1252 struct socket *sock;
1253 struct sock *sk;
1254 void __user *argp = (void __user *)arg;
1255 int pid, err;
1256 struct net *net;
1257
1258 sock = file->private_data;
1259 ops = READ_ONCE(sock->ops);
1260 sk = sock->sk;
1261 net = sock_net(sk);
1262 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1263 struct ifreq ifr;
1264 void __user *data;
1265 bool need_copyout;
1266 if (get_user_ifreq(&ifr, &data, argp))
1267 return -EFAULT;
1268 err = dev_ioctl(net, cmd, &ifr, data, &need_copyout);
1269 if (!err && need_copyout)
1270 if (put_user_ifreq(&ifr, argp))
1271 return -EFAULT;
1272 } else
1273 #ifdef CONFIG_WEXT_CORE
1274 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1275 err = wext_handle_ioctl(net, cmd, argp);
1276 } else
1277 #endif
1278 switch (cmd) {
1279 case FIOSETOWN:
1280 case SIOCSPGRP:
1281 err = -EFAULT;
1282 if (get_user(pid, (int __user *)argp))
1283 break;
1284 err = f_setown(sock->file, pid, 1);
1285 break;
1286 case FIOGETOWN:
1287 case SIOCGPGRP:
1288 err = put_user(f_getown(sock->file),
1289 (int __user *)argp);
1290 break;
1291 case SIOCGIFBR:
1292 case SIOCSIFBR:
1293 case SIOCBRADDBR:
1294 case SIOCBRDELBR:
1295 err = br_ioctl_call(net, NULL, cmd, NULL, argp);
1296 break;
1297 case SIOCGIFVLAN:
1298 case SIOCSIFVLAN:
1299 err = -ENOPKG;
1300 if (!vlan_ioctl_hook)
1301 request_module("8021q");
1302
1303 mutex_lock(&vlan_ioctl_mutex);
1304 if (vlan_ioctl_hook)
1305 err = vlan_ioctl_hook(net, argp);
1306 mutex_unlock(&vlan_ioctl_mutex);
1307 break;
1308 case SIOCGSKNS:
1309 err = -EPERM;
1310 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1311 break;
1312
1313 err = open_related_ns(&net->ns, get_net_ns);
1314 break;
1315 case SIOCGSTAMP_OLD:
1316 case SIOCGSTAMPNS_OLD:
1317 if (!ops->gettstamp) {
1318 err = -ENOIOCTLCMD;
1319 break;
1320 }
1321 err = ops->gettstamp(sock, argp,
1322 cmd == SIOCGSTAMP_OLD,
1323 !IS_ENABLED(CONFIG_64BIT));
1324 break;
1325 case SIOCGSTAMP_NEW:
1326 case SIOCGSTAMPNS_NEW:
1327 if (!ops->gettstamp) {
1328 err = -ENOIOCTLCMD;
1329 break;
1330 }
1331 err = ops->gettstamp(sock, argp,
1332 cmd == SIOCGSTAMP_NEW,
1333 false);
1334 break;
1335
1336 case SIOCGIFCONF:
1337 err = dev_ifconf(net, argp);
1338 break;
1339
1340 default:
1341 err = sock_do_ioctl(net, sock, cmd, arg);
1342 break;
1343 }
1344 return err;
1345 }
1346
1347 /**
1348 * sock_create_lite - creates a socket
1349 * @family: protocol family (AF_INET, ...)
1350 * @type: communication type (SOCK_STREAM, ...)
1351 * @protocol: protocol (0, ...)
1352 * @res: new socket
1353 *
1354 * Creates a new socket and assigns it to @res, passing through LSM.
1355 * The new socket initialization is not complete, see kernel_accept().
1356 * Returns 0 or an error. On failure @res is set to %NULL.
1357 * This function internally uses GFP_KERNEL.
1358 */
1359
1360 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1361 {
1362 int err;
1363 struct socket *sock = NULL;
1364
1365 err = security_socket_create(family, type, protocol, 1);
1366 if (err)
1367 goto out;
1368
1369 sock = sock_alloc();
1370 if (!sock) {
1371 err = -ENOMEM;
1372 goto out;
1373 }
1374
1375 sock->type = type;
1376 err = security_socket_post_create(sock, family, type, protocol, 1);
1377 if (err)
1378 goto out_release;
1379
1380 out:
1381 *res = sock;
1382 return err;
1383 out_release:
1384 sock_release(sock);
1385 sock = NULL;
1386 goto out;
1387 }
1388 EXPORT_SYMBOL(sock_create_lite);
1389
1390 /* No kernel lock held - perfect */
1391 static __poll_t sock_poll(struct file *file, poll_table *wait)
1392 {
1393 struct socket *sock = file->private_data;
1394 const struct proto_ops *ops = READ_ONCE(sock->ops);
1395 __poll_t events = poll_requested_events(wait), flag = 0;
1396
1397 if (!ops->poll)
1398 return 0;
1399
1400 if (sk_can_busy_loop(sock->sk)) {
1401 /* poll once if requested by the syscall */
1402 if (events & POLL_BUSY_LOOP)
1403 sk_busy_loop(sock->sk, 1);
1404
1405 /* if this socket can poll_ll, tell the system call */
1406 flag = POLL_BUSY_LOOP;
1407 }
1408
1409 return ops->poll(file, sock, wait) | flag;
1410 }
1411
1412 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1413 {
1414 struct socket *sock = file->private_data;
1415
1416 return READ_ONCE(sock->ops)->mmap(file, sock, vma);
1417 }
1418
1419 static int sock_close(struct inode *inode, struct file *filp)
1420 {
1421 __sock_release(SOCKET_I(inode), inode);
1422 return 0;
1423 }
1424
1425 /*
1426 * Update the socket async list
1427 *
1428 * Fasync_list locking strategy.
1429 *
1430 * 1. fasync_list is modified only under process context socket lock
1431 * i.e. under semaphore.
1432 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1433 * or under socket lock
1434 */
1435
1436 static int sock_fasync(int fd, struct file *filp, int on)
1437 {
1438 struct socket *sock = filp->private_data;
1439 struct sock *sk = sock->sk;
1440 struct socket_wq *wq = &sock->wq;
1441
1442 if (sk == NULL)
1443 return -EINVAL;
1444
1445 lock_sock(sk);
1446 fasync_helper(fd, filp, on, &wq->fasync_list);
1447
1448 if (!wq->fasync_list)
1449 sock_reset_flag(sk, SOCK_FASYNC);
1450 else
1451 sock_set_flag(sk, SOCK_FASYNC);
1452
1453 release_sock(sk);
1454 return 0;
1455 }
1456
1457 /* This function may be called only under rcu_lock */
1458
1459 int sock_wake_async(struct socket_wq *wq, int how, int band)
1460 {
1461 if (!wq || !wq->fasync_list)
1462 return -1;
1463
1464 switch (how) {
1465 case SOCK_WAKE_WAITD:
1466 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1467 break;
1468 goto call_kill;
1469 case SOCK_WAKE_SPACE:
1470 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1471 break;
1472 fallthrough;
1473 case SOCK_WAKE_IO:
1474 call_kill:
1475 kill_fasync(&wq->fasync_list, SIGIO, band);
1476 break;
1477 case SOCK_WAKE_URG:
1478 kill_fasync(&wq->fasync_list, SIGURG, band);
1479 }
1480
1481 return 0;
1482 }
1483 EXPORT_SYMBOL(sock_wake_async);
1484
1485 /**
1486 * __sock_create - creates a socket
1487 * @net: net namespace
1488 * @family: protocol family (AF_INET, ...)
1489 * @type: communication type (SOCK_STREAM, ...)
1490 * @protocol: protocol (0, ...)
1491 * @res: new socket
1492 * @kern: boolean for kernel space sockets
1493 *
1494 * Creates a new socket and assigns it to @res, passing through LSM.
1495 * Returns 0 or an error. On failure @res is set to %NULL. @kern must
1496 * be set to true if the socket resides in kernel space.
1497 * This function internally uses GFP_KERNEL.
1498 */
1499
1500 int __sock_create(struct net *net, int family, int type, int protocol,
1501 struct socket **res, int kern)
1502 {
1503 int err;
1504 struct socket *sock;
1505 const struct net_proto_family *pf;
1506
1507 /*
1508 * Check protocol is in range
1509 */
1510 if (family < 0 || family >= NPROTO)
1511 return -EAFNOSUPPORT;
1512 if (type < 0 || type >= SOCK_MAX)
1513 return -EINVAL;
1514
1515 /* Compatibility.
1516
1517 This uglymoron is moved from INET layer to here to avoid
1518 deadlock in module load.
1519 */
1520 if (family == PF_INET && type == SOCK_PACKET) {
1521 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1522 current->comm);
1523 family = PF_PACKET;
1524 }
1525
1526 err = security_socket_create(family, type, protocol, kern);
1527 if (err)
1528 return err;
1529
1530 /*
1531 * Allocate the socket and allow the family to set things up. if
1532 * the protocol is 0, the family is instructed to select an appropriate
1533 * default.
1534 */
1535 sock = sock_alloc();
1536 if (!sock) {
1537 net_warn_ratelimited("socket: no more sockets\n");
1538 return -ENFILE; /* Not exactly a match, but its the
1539 closest posix thing */
1540 }
1541
1542 sock->type = type;
1543
1544 #ifdef CONFIG_MODULES
1545 /* Attempt to load a protocol module if the find failed.
1546 *
1547 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1548 * requested real, full-featured networking support upon configuration.
1549 * Otherwise module support will break!
1550 */
1551 if (rcu_access_pointer(net_families[family]) == NULL)
1552 request_module("net-pf-%d", family);
1553 #endif
1554
1555 rcu_read_lock();
1556 pf = rcu_dereference(net_families[family]);
1557 err = -EAFNOSUPPORT;
1558 if (!pf)
1559 goto out_release;
1560
1561 /*
1562 * We will call the ->create function, that possibly is in a loadable
1563 * module, so we have to bump that loadable module refcnt first.
1564 */
1565 if (!try_module_get(pf->owner))
1566 goto out_release;
1567
1568 /* Now protected by module ref count */
1569 rcu_read_unlock();
1570
1571 err = pf->create(net, sock, protocol, kern);
1572 if (err < 0) {
1573 /* ->create should release the allocated sock->sk object on error
1574 * but it may leave the dangling pointer
1575 */
1576 sock->sk = NULL;
1577 goto out_module_put;
1578 }
1579
1580 /*
1581 * Now to bump the refcnt of the [loadable] module that owns this
1582 * socket at sock_release time we decrement its refcnt.
1583 */
1584 if (!try_module_get(sock->ops->owner))
1585 goto out_module_busy;
1586
1587 /*
1588 * Now that we're done with the ->create function, the [loadable]
1589 * module can have its refcnt decremented
1590 */
1591 module_put(pf->owner);
1592 err = security_socket_post_create(sock, family, type, protocol, kern);
1593 if (err)
1594 goto out_sock_release;
1595 *res = sock;
1596
1597 return 0;
1598
1599 out_module_busy:
1600 err = -EAFNOSUPPORT;
1601 out_module_put:
1602 sock->ops = NULL;
1603 module_put(pf->owner);
1604 out_sock_release:
1605 sock_release(sock);
1606 return err;
1607
1608 out_release:
1609 rcu_read_unlock();
1610 goto out_sock_release;
1611 }
1612 EXPORT_SYMBOL(__sock_create);
1613
1614 /**
1615 * sock_create - creates a socket
1616 * @family: protocol family (AF_INET, ...)
1617 * @type: communication type (SOCK_STREAM, ...)
1618 * @protocol: protocol (0, ...)
1619 * @res: new socket
1620 *
1621 * A wrapper around __sock_create().
1622 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1623 */
1624
1625 int sock_create(int family, int type, int protocol, struct socket **res)
1626 {
1627 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1628 }
1629 EXPORT_SYMBOL(sock_create);
1630
1631 /**
1632 * sock_create_kern - creates a socket (kernel space)
1633 * @net: net namespace
1634 * @family: protocol family (AF_INET, ...)
1635 * @type: communication type (SOCK_STREAM, ...)
1636 * @protocol: protocol (0, ...)
1637 * @res: new socket
1638 *
1639 * A wrapper around __sock_create().
1640 * Returns 0 or an error. This function internally uses GFP_KERNEL.
1641 */
1642
1643 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1644 {
1645 return __sock_create(net, family, type, protocol, res, 1);
1646 }
1647 EXPORT_SYMBOL(sock_create_kern);
1648
1649 static struct socket *__sys_socket_create(int family, int type, int protocol)
1650 {
1651 struct socket *sock;
1652 int retval;
1653
1654 /* Check the SOCK_* constants for consistency. */
1655 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1656 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1657 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1658 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1659
1660 if ((type & ~SOCK_TYPE_MASK) & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1661 return ERR_PTR(-EINVAL);
1662 type &= SOCK_TYPE_MASK;
1663
1664 retval = sock_create(family, type, protocol, &sock);
1665 if (retval < 0)
1666 return ERR_PTR(retval);
1667
1668 return sock;
1669 }
1670
1671 struct file *__sys_socket_file(int family, int type, int protocol)
1672 {
1673 struct socket *sock;
1674 int flags;
1675
1676 sock = __sys_socket_create(family, type, protocol);
1677 if (IS_ERR(sock))
1678 return ERR_CAST(sock);
1679
1680 flags = type & ~SOCK_TYPE_MASK;
1681 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1682 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1683
1684 return sock_alloc_file(sock, flags, NULL);
1685 }
1686
1687 /* A hook for bpf progs to attach to and update socket protocol.
1688 *
1689 * A static noinline declaration here could cause the compiler to
1690 * optimize away the function. A global noinline declaration will
1691 * keep the definition, but may optimize away the callsite.
1692 * Therefore, __weak is needed to ensure that the call is still
1693 * emitted, by telling the compiler that we don't know what the
1694 * function might eventually be.
1695 */
1696
1697 __bpf_hook_start();
1698
1699 __weak noinline int update_socket_protocol(int family, int type, int protocol)
1700 {
1701 return protocol;
1702 }
1703
1704 __bpf_hook_end();
1705
1706 int __sys_socket(int family, int type, int protocol)
1707 {
1708 struct socket *sock;
1709 int flags;
1710
1711 sock = __sys_socket_create(family, type,
1712 update_socket_protocol(family, type, protocol));
1713 if (IS_ERR(sock))
1714 return PTR_ERR(sock);
1715
1716 flags = type & ~SOCK_TYPE_MASK;
1717 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1718 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1719
1720 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1721 }
1722
1723 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1724 {
1725 return __sys_socket(family, type, protocol);
1726 }
1727
1728 /*
1729 * Create a pair of connected sockets.
1730 */
1731
1732 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1733 {
1734 struct socket *sock1, *sock2;
1735 int fd1, fd2, err;
1736 struct file *newfile1, *newfile2;
1737 int flags;
1738
1739 flags = type & ~SOCK_TYPE_MASK;
1740 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1741 return -EINVAL;
1742 type &= SOCK_TYPE_MASK;
1743
1744 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1745 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1746
1747 /*
1748 * reserve descriptors and make sure we won't fail
1749 * to return them to userland.
1750 */
1751 fd1 = get_unused_fd_flags(flags);
1752 if (unlikely(fd1 < 0))
1753 return fd1;
1754
1755 fd2 = get_unused_fd_flags(flags);
1756 if (unlikely(fd2 < 0)) {
1757 put_unused_fd(fd1);
1758 return fd2;
1759 }
1760
1761 err = put_user(fd1, &usockvec[0]);
1762 if (err)
1763 goto out;
1764
1765 err = put_user(fd2, &usockvec[1]);
1766 if (err)
1767 goto out;
1768
1769 /*
1770 * Obtain the first socket and check if the underlying protocol
1771 * supports the socketpair call.
1772 */
1773
1774 err = sock_create(family, type, protocol, &sock1);
1775 if (unlikely(err < 0))
1776 goto out;
1777
1778 err = sock_create(family, type, protocol, &sock2);
1779 if (unlikely(err < 0)) {
1780 sock_release(sock1);
1781 goto out;
1782 }
1783
1784 err = security_socket_socketpair(sock1, sock2);
1785 if (unlikely(err)) {
1786 sock_release(sock2);
1787 sock_release(sock1);
1788 goto out;
1789 }
1790
1791 err = READ_ONCE(sock1->ops)->socketpair(sock1, sock2);
1792 if (unlikely(err < 0)) {
1793 sock_release(sock2);
1794 sock_release(sock1);
1795 goto out;
1796 }
1797
1798 newfile1 = sock_alloc_file(sock1, flags, NULL);
1799 if (IS_ERR(newfile1)) {
1800 err = PTR_ERR(newfile1);
1801 sock_release(sock2);
1802 goto out;
1803 }
1804
1805 newfile2 = sock_alloc_file(sock2, flags, NULL);
1806 if (IS_ERR(newfile2)) {
1807 err = PTR_ERR(newfile2);
1808 fput(newfile1);
1809 goto out;
1810 }
1811
1812 audit_fd_pair(fd1, fd2);
1813
1814 fd_install(fd1, newfile1);
1815 fd_install(fd2, newfile2);
1816 return 0;
1817
1818 out:
1819 put_unused_fd(fd2);
1820 put_unused_fd(fd1);
1821 return err;
1822 }
1823
1824 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1825 int __user *, usockvec)
1826 {
1827 return __sys_socketpair(family, type, protocol, usockvec);
1828 }
1829
1830 int __sys_bind_socket(struct socket *sock, struct sockaddr_storage *address,
1831 int addrlen)
1832 {
1833 int err;
1834
1835 err = security_socket_bind(sock, (struct sockaddr *)address,
1836 addrlen);
1837 if (!err)
1838 err = READ_ONCE(sock->ops)->bind(sock,
1839 (struct sockaddr *)address,
1840 addrlen);
1841 return err;
1842 }
1843
1844 /*
1845 * Bind a name to a socket. Nothing much to do here since it's
1846 * the protocol's responsibility to handle the local address.
1847 *
1848 * We move the socket address to kernel space before we call
1849 * the protocol layer (having also checked the address is ok).
1850 */
1851
1852 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1853 {
1854 struct socket *sock;
1855 struct sockaddr_storage address;
1856 int err, fput_needed;
1857
1858 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1859 if (sock) {
1860 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1861 if (!err)
1862 err = __sys_bind_socket(sock, &address, addrlen);
1863 fput_light(sock->file, fput_needed);
1864 }
1865 return err;
1866 }
1867
1868 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1869 {
1870 return __sys_bind(fd, umyaddr, addrlen);
1871 }
1872
1873 /*
1874 * Perform a listen. Basically, we allow the protocol to do anything
1875 * necessary for a listen, and if that works, we mark the socket as
1876 * ready for listening.
1877 */
1878 int __sys_listen_socket(struct socket *sock, int backlog)
1879 {
1880 int somaxconn, err;
1881
1882 somaxconn = READ_ONCE(sock_net(sock->sk)->core.sysctl_somaxconn);
1883 if ((unsigned int)backlog > somaxconn)
1884 backlog = somaxconn;
1885
1886 err = security_socket_listen(sock, backlog);
1887 if (!err)
1888 err = READ_ONCE(sock->ops)->listen(sock, backlog);
1889 return err;
1890 }
1891
1892 int __sys_listen(int fd, int backlog)
1893 {
1894 struct socket *sock;
1895 int err, fput_needed;
1896
1897 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1898 if (sock) {
1899 err = __sys_listen_socket(sock, backlog);
1900 fput_light(sock->file, fput_needed);
1901 }
1902 return err;
1903 }
1904
1905 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1906 {
1907 return __sys_listen(fd, backlog);
1908 }
1909
1910 struct file *do_accept(struct file *file, struct proto_accept_arg *arg,
1911 struct sockaddr __user *upeer_sockaddr,
1912 int __user *upeer_addrlen, int flags)
1913 {
1914 struct socket *sock, *newsock;
1915 struct file *newfile;
1916 int err, len;
1917 struct sockaddr_storage address;
1918 const struct proto_ops *ops;
1919
1920 sock = sock_from_file(file);
1921 if (!sock)
1922 return ERR_PTR(-ENOTSOCK);
1923
1924 newsock = sock_alloc();
1925 if (!newsock)
1926 return ERR_PTR(-ENFILE);
1927 ops = READ_ONCE(sock->ops);
1928
1929 newsock->type = sock->type;
1930 newsock->ops = ops;
1931
1932 /*
1933 * We don't need try_module_get here, as the listening socket (sock)
1934 * has the protocol module (sock->ops->owner) held.
1935 */
1936 __module_get(ops->owner);
1937
1938 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1939 if (IS_ERR(newfile))
1940 return newfile;
1941
1942 err = security_socket_accept(sock, newsock);
1943 if (err)
1944 goto out_fd;
1945
1946 arg->flags |= sock->file->f_flags;
1947 err = ops->accept(sock, newsock, arg);
1948 if (err < 0)
1949 goto out_fd;
1950
1951 if (ccs_socket_post_accept_permission(sock, newsock)) {
1952 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1953 goto out_fd;
1954 }
1955 if (upeer_sockaddr) {
1956 len = ops->getname(newsock, (struct sockaddr *)&address, 2);
1957 if (len < 0) {
1958 err = -ECONNABORTED;
1959 goto out_fd;
1960 }
1961 err = move_addr_to_user(&address,
1962 len, upeer_sockaddr, upeer_addrlen);
1963 if (err < 0)
1964 goto out_fd;
1965 }
1966
1967 /* File flags are not inherited via accept() unlike another OSes. */
1968 return newfile;
1969 out_fd:
1970 fput(newfile);
1971 return ERR_PTR(err);
1972 }
1973
1974 static int __sys_accept4_file(struct file *file, struct sockaddr __user *upeer_sockaddr,
1975 int __user *upeer_addrlen, int flags)
1976 {
1977 struct proto_accept_arg arg = { };
1978 struct file *newfile;
1979 int newfd;
1980
1981 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1982 return -EINVAL;
1983
1984 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1985 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1986
1987 newfd = get_unused_fd_flags(flags);
1988 if (unlikely(newfd < 0))
1989 return newfd;
1990
1991 newfile = do_accept(file, &arg, upeer_sockaddr, upeer_addrlen,
1992 flags);
1993 if (IS_ERR(newfile)) {
1994 put_unused_fd(newfd);
1995 return PTR_ERR(newfile);
1996 }
1997 fd_install(newfd, newfile);
1998 return newfd;
1999 }
2000
2001 /*
2002 * For accept, we attempt to create a new socket, set up the link
2003 * with the client, wake up the client, then return the new
2004 * connected fd. We collect the address of the connector in kernel
2005 * space and move it to user at the very end. This is unclean because
2006 * we open the socket then return an error.
2007 *
2008 * 1003.1g adds the ability to recvmsg() to query connection pending
2009 * status to recvmsg. We need to add that support in a way thats
2010 * clean when we restructure accept also.
2011 */
2012
2013 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
2014 int __user *upeer_addrlen, int flags)
2015 {
2016 int ret = -EBADF;
2017 struct fd f;
2018
2019 f = fdget(fd);
2020 if (f.file) {
2021 ret = __sys_accept4_file(f.file, upeer_sockaddr,
2022 upeer_addrlen, flags);
2023 fdput(f);
2024 }
2025
2026 return ret;
2027 }
2028
2029 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
2030 int __user *, upeer_addrlen, int, flags)
2031 {
2032 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
2033 }
2034
2035 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
2036 int __user *, upeer_addrlen)
2037 {
2038 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
2039 }
2040
2041 /*
2042 * Attempt to connect to a socket with the server address. The address
2043 * is in user space so we verify it is OK and move it to kernel space.
2044 *
2045 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
2046 * break bindings
2047 *
2048 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
2049 * other SEQPACKET protocols that take time to connect() as it doesn't
2050 * include the -EINPROGRESS status for such sockets.
2051 */
2052
2053 int __sys_connect_file(struct file *file, struct sockaddr_storage *address,
2054 int addrlen, int file_flags)
2055 {
2056 struct socket *sock;
2057 int err;
2058
2059 sock = sock_from_file(file);
2060 if (!sock) {
2061 err = -ENOTSOCK;
2062 goto out;
2063 }
2064
2065 err =
2066 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
2067 if (err)
2068 goto out;
2069
2070 err = READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)address,
2071 addrlen, sock->file->f_flags | file_flags);
2072 out:
2073 return err;
2074 }
2075
2076 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
2077 {
2078 int ret = -EBADF;
2079 struct fd f;
2080
2081 f = fdget(fd);
2082 if (f.file) {
2083 struct sockaddr_storage address;
2084
2085 ret = move_addr_to_kernel(uservaddr, addrlen, &address);
2086 if (!ret)
2087 ret = __sys_connect_file(f.file, &address, addrlen, 0);
2088 fdput(f);
2089 }
2090
2091 return ret;
2092 }
2093
2094 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
2095 int, addrlen)
2096 {
2097 return __sys_connect(fd, uservaddr, addrlen);
2098 }
2099
2100 /*
2101 * Get the local address ('name') of a socket object. Move the obtained
2102 * name to user space.
2103 */
2104
2105 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
2106 int __user *usockaddr_len)
2107 {
2108 struct socket *sock;
2109 struct sockaddr_storage address;
2110 int err, fput_needed;
2111
2112 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2113 if (!sock)
2114 goto out;
2115
2116 err = security_socket_getsockname(sock);
2117 if (err)
2118 goto out_put;
2119
2120 err = READ_ONCE(sock->ops)->getname(sock, (struct sockaddr *)&address, 0);
2121 if (err < 0)
2122 goto out_put;
2123 /* "err" is actually length in this case */
2124 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
2125
2126 out_put:
2127 fput_light(sock->file, fput_needed);
2128 out:
2129 return err;
2130 }
2131
2132 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
2133 int __user *, usockaddr_len)
2134 {
2135 return __sys_getsockname(fd, usockaddr, usockaddr_len);
2136 }
2137
2138 /*
2139 * Get the remote address ('name') of a socket object. Move the obtained
2140 * name to user space.
2141 */
2142
2143 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
2144 int __user *usockaddr_len)
2145 {
2146 struct socket *sock;
2147 struct sockaddr_storage address;
2148 int err, fput_needed;
2149
2150 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2151 if (sock != NULL) {
2152 const struct proto_ops *ops = READ_ONCE(sock->ops);
2153
2154 err = security_socket_getpeername(sock);
2155 if (err) {
2156 fput_light(sock->file, fput_needed);
2157 return err;
2158 }
2159
2160 err = ops->getname(sock, (struct sockaddr *)&address, 1);
2161 if (err >= 0)
2162 /* "err" is actually length in this case */
2163 err = move_addr_to_user(&address, err, usockaddr,
2164 usockaddr_len);
2165 fput_light(sock->file, fput_needed);
2166 }
2167 return err;
2168 }
2169
2170 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
2171 int __user *, usockaddr_len)
2172 {
2173 return __sys_getpeername(fd, usockaddr, usockaddr_len);
2174 }
2175
2176 /*
2177 * Send a datagram to a given address. We move the address into kernel
2178 * space and check the user space data area is readable before invoking
2179 * the protocol.
2180 */
2181 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
2182 struct sockaddr __user *addr, int addr_len)
2183 {
2184 struct socket *sock;
2185 struct sockaddr_storage address;
2186 int err;
2187 struct msghdr msg;
2188 int fput_needed;
2189
2190 err = import_ubuf(ITER_SOURCE, buff, len, &msg.msg_iter);
2191 if (unlikely(err))
2192 return err;
2193 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2194 if (!sock)
2195 goto out;
2196
2197 msg.msg_name = NULL;
2198 msg.msg_control = NULL;
2199 msg.msg_controllen = 0;
2200 msg.msg_namelen = 0;
2201 msg.msg_ubuf = NULL;
2202 if (addr) {
2203 err = move_addr_to_kernel(addr, addr_len, &address);
2204 if (err < 0)
2205 goto out_put;
2206 msg.msg_name = (struct sockaddr *)&address;
2207 msg.msg_namelen = addr_len;
2208 }
2209 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2210 if (sock->file->f_flags & O_NONBLOCK)
2211 flags |= MSG_DONTWAIT;
2212 msg.msg_flags = flags;
2213 err = __sock_sendmsg(sock, &msg);
2214
2215 out_put:
2216 fput_light(sock->file, fput_needed);
2217 out:
2218 return err;
2219 }
2220
2221 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
2222 unsigned int, flags, struct sockaddr __user *, addr,
2223 int, addr_len)
2224 {
2225 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
2226 }
2227
2228 /*
2229 * Send a datagram down a socket.
2230 */
2231
2232 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
2233 unsigned int, flags)
2234 {
2235 return __sys_sendto(fd, buff, len, flags, NULL, 0);
2236 }
2237
2238 /*
2239 * Receive a frame from the socket and optionally record the address of the
2240 * sender. We verify the buffers are writable and if needed move the
2241 * sender address from kernel to user space.
2242 */
2243 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
2244 struct sockaddr __user *addr, int __user *addr_len)
2245 {
2246 struct sockaddr_storage address;
2247 struct msghdr msg = {
2248 /* Save some cycles and don't copy the address if not needed */
2249 .msg_name = addr ? (struct sockaddr *)&address : NULL,
2250 };
2251 struct socket *sock;
2252 int err, err2;
2253 int fput_needed;
2254
2255 err = import_ubuf(ITER_DEST, ubuf, size, &msg.msg_iter);
2256 if (unlikely(err))
2257 return err;
2258 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2259 if (!sock)
2260 goto out;
2261
2262 if (sock->file->f_flags & O_NONBLOCK)
2263 flags |= MSG_DONTWAIT;
2264 err = sock_recvmsg(sock, &msg, flags);
2265
2266 if (err >= 0 && addr != NULL) {
2267 err2 = move_addr_to_user(&address,
2268 msg.msg_namelen, addr, addr_len);
2269 if (err2 < 0)
2270 err = err2;
2271 }
2272
2273 fput_light(sock->file, fput_needed);
2274 out:
2275 return err;
2276 }
2277
2278 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
2279 unsigned int, flags, struct sockaddr __user *, addr,
2280 int __user *, addr_len)
2281 {
2282 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
2283 }
2284
2285 /*
2286 * Receive a datagram from a socket.
2287 */
2288
2289 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
2290 unsigned int, flags)
2291 {
2292 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
2293 }
2294
2295 static bool sock_use_custom_sol_socket(const struct socket *sock)
2296 {
2297 return test_bit(SOCK_CUSTOM_SOCKOPT, &sock->flags);
2298 }
2299
2300 int do_sock_setsockopt(struct socket *sock, bool compat, int level,
2301 int optname, sockptr_t optval, int optlen)
2302 {
2303 const struct proto_ops *ops;
2304 char *kernel_optval = NULL;
2305 int err;
2306
2307 if (optlen < 0)
2308 return -EINVAL;
2309
2310 err = security_socket_setsockopt(sock, level, optname);
2311 if (err)
2312 goto out_put;
2313
2314 if (!compat)
2315 err = BPF_CGROUP_RUN_PROG_SETSOCKOPT(sock->sk, &level, &optname,
2316 optval, &optlen,
2317 &kernel_optval);
2318 if (err < 0)
2319 goto out_put;
2320 if (err > 0) {
2321 err = 0;
2322 goto out_put;
2323 }
2324
2325 if (kernel_optval)
2326 optval = KERNEL_SOCKPTR(kernel_optval);
2327 ops = READ_ONCE(sock->ops);
2328 if (level == SOL_SOCKET && !sock_use_custom_sol_socket(sock))
2329 err = sock_setsockopt(sock, level, optname, optval, optlen);
2330 else if (unlikely(!ops->setsockopt))
2331 err = -EOPNOTSUPP;
2332 else
2333 err = ops->setsockopt(sock, level, optname, optval,
2334 optlen);
2335 kfree(kernel_optval);
2336 out_put:
2337 return err;
2338 }
2339 EXPORT_SYMBOL(do_sock_setsockopt);
2340
2341 /* Set a socket option. Because we don't know the option lengths we have
2342 * to pass the user mode parameter for the protocols to sort out.
2343 */
2344 int __sys_setsockopt(int fd, int level, int optname, char __user *user_optval,
2345 int optlen)
2346 {
2347 sockptr_t optval = USER_SOCKPTR(user_optval);
2348 bool compat = in_compat_syscall();
2349 int err, fput_needed;
2350 struct socket *sock;
2351
2352 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2353 if (!sock)
2354 return err;
2355
2356 err = do_sock_setsockopt(sock, compat, level, optname, optval, optlen);
2357
2358 fput_light(sock->file, fput_needed);
2359 return err;
2360 }
2361
2362 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
2363 char __user *, optval, int, optlen)
2364 {
2365 return __sys_setsockopt(fd, level, optname, optval, optlen);
2366 }
2367
2368 INDIRECT_CALLABLE_DECLARE(bool tcp_bpf_bypass_getsockopt(int level,
2369 int optname));
2370
2371 int do_sock_getsockopt(struct socket *sock, bool compat, int level,
2372 int optname, sockptr_t optval, sockptr_t optlen)
2373 {
2374 int max_optlen __maybe_unused = 0;
2375 const struct proto_ops *ops;
2376 int err;
2377
2378 err = security_socket_getsockopt(sock, level, optname);
2379 if (err)
2380 return err;
2381
2382 if (!compat)
2383 copy_from_sockptr(&max_optlen, optlen, sizeof(int));
2384
2385 ops = READ_ONCE(sock->ops);
2386 if (level == SOL_SOCKET) {
2387 err = sk_getsockopt(sock->sk, level, optname, optval, optlen);
2388 } else if (unlikely(!ops->getsockopt)) {
2389 err = -EOPNOTSUPP;
2390 } else {
2391 if (WARN_ONCE(optval.is_kernel || optlen.is_kernel,
2392 "Invalid argument type"))
2393 return -EOPNOTSUPP;
2394
2395 err = ops->getsockopt(sock, level, optname, optval.user,
2396 optlen.user);
2397 }
2398
2399 if (!compat)
2400 err = BPF_CGROUP_RUN_PROG_GETSOCKOPT(sock->sk, level, optname,
2401 optval, optlen, max_optlen,
2402 err);
2403
2404 return err;
2405 }
2406 EXPORT_SYMBOL(do_sock_getsockopt);
2407
2408 /*
2409 * Get a socket option. Because we don't know the option lengths we have
2410 * to pass a user mode parameter for the protocols to sort out.
2411 */
2412 int __sys_getsockopt(int fd, int level, int optname, char __user *optval,
2413 int __user *optlen)
2414 {
2415 int err, fput_needed;
2416 struct socket *sock;
2417 bool compat;
2418
2419 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2420 if (!sock)
2421 return err;
2422
2423 compat = in_compat_syscall();
2424 err = do_sock_getsockopt(sock, compat, level, optname,
2425 USER_SOCKPTR(optval), USER_SOCKPTR(optlen));
2426
2427 fput_light(sock->file, fput_needed);
2428 return err;
2429 }
2430
2431 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
2432 char __user *, optval, int __user *, optlen)
2433 {
2434 return __sys_getsockopt(fd, level, optname, optval, optlen);
2435 }
2436
2437 /*
2438 * Shutdown a socket.
2439 */
2440
2441 int __sys_shutdown_sock(struct socket *sock, int how)
2442 {
2443 int err;
2444
2445 err = security_socket_shutdown(sock, how);
2446 if (!err)
2447 err = READ_ONCE(sock->ops)->shutdown(sock, how);
2448
2449 return err;
2450 }
2451
2452 int __sys_shutdown(int fd, int how)
2453 {
2454 int err, fput_needed;
2455 struct socket *sock;
2456
2457 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2458 if (sock != NULL) {
2459 err = __sys_shutdown_sock(sock, how);
2460 fput_light(sock->file, fput_needed);
2461 }
2462 return err;
2463 }
2464
2465 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
2466 {
2467 return __sys_shutdown(fd, how);
2468 }
2469
2470 /* A couple of helpful macros for getting the address of the 32/64 bit
2471 * fields which are the same type (int / unsigned) on our platforms.
2472 */
2473 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
2474 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
2475 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
2476
2477 struct used_address {
2478 struct sockaddr_storage name;
2479 unsigned int name_len;
2480 };
2481
2482 int __copy_msghdr(struct msghdr *kmsg,
2483 struct user_msghdr *msg,
2484 struct sockaddr __user **save_addr)
2485 {
2486 ssize_t err;
2487
2488 kmsg->msg_control_is_user = true;
2489 kmsg->msg_get_inq = 0;
2490 kmsg->msg_control_user = msg->msg_control;
2491 kmsg->msg_controllen = msg->msg_controllen;
2492 kmsg->msg_flags = msg->msg_flags;
2493
2494 kmsg->msg_namelen = msg->msg_namelen;
2495 if (!msg->msg_name)
2496 kmsg->msg_namelen = 0;
2497
2498 if (kmsg->msg_namelen < 0)
2499 return -EINVAL;
2500
2501 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2502 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2503
2504 if (save_addr)
2505 *save_addr = msg->msg_name;
2506
2507 if (msg->msg_name && kmsg->msg_namelen) {
2508 if (!save_addr) {
2509 err = move_addr_to_kernel(msg->msg_name,
2510 kmsg->msg_namelen,
2511 kmsg->msg_name);
2512 if (err < 0)
2513 return err;
2514 }
2515 } else {
2516 kmsg->msg_name = NULL;
2517 kmsg->msg_namelen = 0;
2518 }
2519
2520 if (msg->msg_iovlen > UIO_MAXIOV)
2521 return -EMSGSIZE;
2522
2523 kmsg->msg_iocb = NULL;
2524 kmsg->msg_ubuf = NULL;
2525 return 0;
2526 }
2527
2528 static int copy_msghdr_from_user(struct msghdr *kmsg,
2529 struct user_msghdr __user *umsg,
2530 struct sockaddr __user **save_addr,
2531 struct iovec **iov)
2532 {
2533 struct user_msghdr msg;
2534 ssize_t err;
2535
2536 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2537 return -EFAULT;
2538
2539 err = __copy_msghdr(kmsg, &msg, save_addr);
2540 if (err)
2541 return err;
2542
2543 err = import_iovec(save_addr ? ITER_DEST : ITER_SOURCE,
2544 msg.msg_iov, msg.msg_iovlen,
2545 UIO_FASTIOV, iov, &kmsg->msg_iter);
2546 return err < 0 ? err : 0;
2547 }
2548
2549 static int ____sys_sendmsg(struct socket *sock, struct msghdr *msg_sys,
2550 unsigned int flags, struct used_address *used_address,
2551 unsigned int allowed_msghdr_flags)
2552 {
2553 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2554 __aligned(sizeof(__kernel_size_t));
2555 /* 20 is size of ipv6_pktinfo */
2556 unsigned char *ctl_buf = ctl;
2557 int ctl_len;
2558 ssize_t err;
2559
2560 err = -ENOBUFS;
2561
2562 if (msg_sys->msg_controllen > INT_MAX)
2563 goto out;
2564 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2565 ctl_len = msg_sys->msg_controllen;
2566 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2567 err =
2568 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2569 sizeof(ctl));
2570 if (err)
2571 goto out;
2572 ctl_buf = msg_sys->msg_control;
2573 ctl_len = msg_sys->msg_controllen;
2574 } else if (ctl_len) {
2575 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2576 CMSG_ALIGN(sizeof(struct cmsghdr)));
2577 if (ctl_len > sizeof(ctl)) {
2578 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2579 if (ctl_buf == NULL)
2580 goto out;
2581 }
2582 err = -EFAULT;
2583 if (copy_from_user(ctl_buf, msg_sys->msg_control_user, ctl_len))
2584 goto out_freectl;
2585 msg_sys->msg_control = ctl_buf;
2586 msg_sys->msg_control_is_user = false;
2587 }
2588 flags &= ~MSG_INTERNAL_SENDMSG_FLAGS;
2589 msg_sys->msg_flags = flags;
2590
2591 if (sock->file->f_flags & O_NONBLOCK)
2592 msg_sys->msg_flags |= MSG_DONTWAIT;
2593 /*
2594 * If this is sendmmsg() and current destination address is same as
2595 * previously succeeded address, omit asking LSM's decision.
2596 * used_address->name_len is initialized to UINT_MAX so that the first
2597 * destination address never matches.
2598 */
2599 if (used_address && msg_sys->msg_name &&
2600 used_address->name_len == msg_sys->msg_namelen &&
2601 !memcmp(&used_address->name, msg_sys->msg_name,
2602 used_address->name_len)) {
2603 err = sock_sendmsg_nosec(sock, msg_sys);
2604 goto out_freectl;
2605 }
2606 err = __sock_sendmsg(sock, msg_sys);
2607 /*
2608 * If this is sendmmsg() and sending to current destination address was
2609 * successful, remember it.
2610 */
2611 if (used_address && err >= 0) {
2612 used_address->name_len = msg_sys->msg_namelen;
2613 if (msg_sys->msg_name)
2614 memcpy(&used_address->name, msg_sys->msg_name,
2615 used_address->name_len);
2616 }
2617
2618 out_freectl:
2619 if (ctl_buf != ctl)
2620 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2621 out:
2622 return err;
2623 }
2624
2625 static int sendmsg_copy_msghdr(struct msghdr *msg,
2626 struct user_msghdr __user *umsg, unsigned flags,
2627 struct iovec **iov)
2628 {
2629 int err;
2630
2631 if (flags & MSG_CMSG_COMPAT) {
2632 struct compat_msghdr __user *msg_compat;
2633
2634 msg_compat = (struct compat_msghdr __user *) umsg;
2635 err = get_compat_msghdr(msg, msg_compat, NULL, iov);
2636 } else {
2637 err = copy_msghdr_from_user(msg, umsg, NULL, iov);
2638 }
2639 if (err < 0)
2640 return err;
2641
2642 return 0;
2643 }
2644
2645 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2646 struct msghdr *msg_sys, unsigned int flags,
2647 struct used_address *used_address,
2648 unsigned int allowed_msghdr_flags)
2649 {
2650 struct sockaddr_storage address;
2651 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2652 ssize_t err;
2653
2654 msg_sys->msg_name = &address;
2655
2656 err = sendmsg_copy_msghdr(msg_sys, msg, flags, &iov);
2657 if (err < 0)
2658 return err;
2659
2660 err = ____sys_sendmsg(sock, msg_sys, flags, used_address,
2661 allowed_msghdr_flags);
2662 kfree(iov);
2663 return err;
2664 }
2665
2666 /*
2667 * BSD sendmsg interface
2668 */
2669 long __sys_sendmsg_sock(struct socket *sock, struct msghdr *msg,
2670 unsigned int flags)
2671 {
2672 return ____sys_sendmsg(sock, msg, flags, NULL, 0);
2673 }
2674
2675 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2676 bool forbid_cmsg_compat)
2677 {
2678 int fput_needed, err;
2679 struct msghdr msg_sys;
2680 struct socket *sock;
2681
2682 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2683 return -EINVAL;
2684
2685 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2686 if (!sock)
2687 goto out;
2688
2689 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2690
2691 fput_light(sock->file, fput_needed);
2692 out:
2693 return err;
2694 }
2695
2696 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2697 {
2698 return __sys_sendmsg(fd, msg, flags, true);
2699 }
2700
2701 /*
2702 * Linux sendmmsg interface
2703 */
2704
2705 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2706 unsigned int flags, bool forbid_cmsg_compat)
2707 {
2708 int fput_needed, err, datagrams;
2709 struct socket *sock;
2710 struct mmsghdr __user *entry;
2711 struct compat_mmsghdr __user *compat_entry;
2712 struct msghdr msg_sys;
2713 struct used_address used_address;
2714 unsigned int oflags = flags;
2715
2716 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2717 return -EINVAL;
2718
2719 if (vlen > UIO_MAXIOV)
2720 vlen = UIO_MAXIOV;
2721
2722 datagrams = 0;
2723
2724 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2725 if (!sock)
2726 return err;
2727
2728 used_address.name_len = UINT_MAX;
2729 entry = mmsg;
2730 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2731 err = 0;
2732 flags |= MSG_BATCH;
2733
2734 while (datagrams < vlen) {
2735 if (datagrams == vlen - 1)
2736 flags = oflags;
2737
2738 if (MSG_CMSG_COMPAT & flags) {
2739 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2740 &msg_sys, flags, &used_address, MSG_EOR);
2741 if (err < 0)
2742 break;
2743 err = __put_user(err, &compat_entry->msg_len);
2744 ++compat_entry;
2745 } else {
2746 err = ___sys_sendmsg(sock,
2747 (struct user_msghdr __user *)entry,
2748 &msg_sys, flags, &used_address, MSG_EOR);
2749 if (err < 0)
2750 break;
2751 err = put_user(err, &entry->msg_len);
2752 ++entry;
2753 }
2754
2755 if (err)
2756 break;
2757 ++datagrams;
2758 if (msg_data_left(&msg_sys))
2759 break;
2760 cond_resched();
2761 }
2762
2763 fput_light(sock->file, fput_needed);
2764
2765 /* We only return an error if no datagrams were able to be sent */
2766 if (datagrams != 0)
2767 return datagrams;
2768
2769 return err;
2770 }
2771
2772 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2773 unsigned int, vlen, unsigned int, flags)
2774 {
2775 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2776 }
2777
2778 static int recvmsg_copy_msghdr(struct msghdr *msg,
2779 struct user_msghdr __user *umsg, unsigned flags,
2780 struct sockaddr __user **uaddr,
2781 struct iovec **iov)
2782 {
2783 ssize_t err;
2784
2785 if (MSG_CMSG_COMPAT & flags) {
2786 struct compat_msghdr __user *msg_compat;
2787
2788 msg_compat = (struct compat_msghdr __user *) umsg;
2789 err = get_compat_msghdr(msg, msg_compat, uaddr, iov);
2790 } else {
2791 err = copy_msghdr_from_user(msg, umsg, uaddr, iov);
2792 }
2793 if (err < 0)
2794 return err;
2795
2796 return 0;
2797 }
2798
2799 static int ____sys_recvmsg(struct socket *sock, struct msghdr *msg_sys,
2800 struct user_msghdr __user *msg,
2801 struct sockaddr __user *uaddr,
2802 unsigned int flags, int nosec)
2803 {
2804 struct compat_msghdr __user *msg_compat =
2805 (struct compat_msghdr __user *) msg;
2806 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2807 struct sockaddr_storage addr;
2808 unsigned long cmsg_ptr;
2809 int len;
2810 ssize_t err;
2811
2812 msg_sys->msg_name = &addr;
2813 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2814 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2815
2816 /* We assume all kernel code knows the size of sockaddr_storage */
2817 msg_sys->msg_namelen = 0;
2818
2819 if (sock->file->f_flags & O_NONBLOCK)
2820 flags |= MSG_DONTWAIT;
2821
2822 if (unlikely(nosec))
2823 err = sock_recvmsg_nosec(sock, msg_sys, flags);
2824 else
2825 err = sock_recvmsg(sock, msg_sys, flags);
2826
2827 if (err < 0)
2828 goto out;
2829 len = err;
2830
2831 if (uaddr != NULL) {
2832 err = move_addr_to_user(&addr,
2833 msg_sys->msg_namelen, uaddr,
2834 uaddr_len);
2835 if (err < 0)
2836 goto out;
2837 }
2838 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2839 COMPAT_FLAGS(msg));
2840 if (err)
2841 goto out;
2842 if (MSG_CMSG_COMPAT & flags)
2843 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2844 &msg_compat->msg_controllen);
2845 else
2846 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2847 &msg->msg_controllen);
2848 if (err)
2849 goto out;
2850 err = len;
2851 out:
2852 return err;
2853 }
2854
2855 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2856 struct msghdr *msg_sys, unsigned int flags, int nosec)
2857 {
2858 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2859 /* user mode address pointers */
2860 struct sockaddr __user *uaddr;
2861 ssize_t err;
2862
2863 err = recvmsg_copy_msghdr(msg_sys, msg, flags, &uaddr, &iov);
2864 if (err < 0)
2865 return err;
2866
2867 err = ____sys_recvmsg(sock, msg_sys, msg, uaddr, flags, nosec);
2868 kfree(iov);
2869 return err;
2870 }
2871
2872 /*
2873 * BSD recvmsg interface
2874 */
2875
2876 long __sys_recvmsg_sock(struct socket *sock, struct msghdr *msg,
2877 struct user_msghdr __user *umsg,
2878 struct sockaddr __user *uaddr, unsigned int flags)
2879 {
2880 return ____sys_recvmsg(sock, msg, umsg, uaddr, flags, 0);
2881 }
2882
2883 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2884 bool forbid_cmsg_compat)
2885 {
2886 int fput_needed, err;
2887 struct msghdr msg_sys;
2888 struct socket *sock;
2889
2890 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2891 return -EINVAL;
2892
2893 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2894 if (!sock)
2895 goto out;
2896
2897 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2898
2899 fput_light(sock->file, fput_needed);
2900 out:
2901 return err;
2902 }
2903
2904 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2905 unsigned int, flags)
2906 {
2907 return __sys_recvmsg(fd, msg, flags, true);
2908 }
2909
2910 /*
2911 * Linux recvmmsg interface
2912 */
2913
2914 static int do_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2915 unsigned int vlen, unsigned int flags,
2916 struct timespec64 *timeout)
2917 {
2918 int fput_needed, err, datagrams;
2919 struct socket *sock;
2920 struct mmsghdr __user *entry;
2921 struct compat_mmsghdr __user *compat_entry;
2922 struct msghdr msg_sys;
2923 struct timespec64 end_time;
2924 struct timespec64 timeout64;
2925
2926 if (timeout &&
2927 poll_select_set_timeout(&end_time, timeout->tv_sec,
2928 timeout->tv_nsec))
2929 return -EINVAL;
2930
2931 datagrams = 0;
2932
2933 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2934 if (!sock)
2935 return err;
2936
2937 if (likely(!(flags & MSG_ERRQUEUE))) {
2938 err = sock_error(sock->sk);
2939 if (err) {
2940 datagrams = err;
2941 goto out_put;
2942 }
2943 }
2944
2945 entry = mmsg;
2946 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2947
2948 while (datagrams < vlen) {
2949 /*
2950 * No need to ask LSM for more than the first datagram.
2951 */
2952 if (MSG_CMSG_COMPAT & flags) {
2953 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2954 &msg_sys, flags & ~MSG_WAITFORONE,
2955 datagrams);
2956 if (err < 0)
2957 break;
2958 err = __put_user(err, &compat_entry->msg_len);
2959 ++compat_entry;
2960 } else {
2961 err = ___sys_recvmsg(sock,
2962 (struct user_msghdr __user *)entry,
2963 &msg_sys, flags & ~MSG_WAITFORONE,
2964 datagrams);
2965 if (err < 0)
2966 break;
2967 err = put_user(err, &entry->msg_len);
2968 ++entry;
2969 }
2970
2971 if (err)
2972 break;
2973 ++datagrams;
2974
2975 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2976 if (flags & MSG_WAITFORONE)
2977 flags |= MSG_DONTWAIT;
2978
2979 if (timeout) {
2980 ktime_get_ts64(&timeout64);
2981 *timeout = timespec64_sub(end_time, timeout64);
2982 if (timeout->tv_sec < 0) {
2983 timeout->tv_sec = timeout->tv_nsec = 0;
2984 break;
2985 }
2986
2987 /* Timeout, return less than vlen datagrams */
2988 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2989 break;
2990 }
2991
2992 /* Out of band data, return right away */
2993 if (msg_sys.msg_flags & MSG_OOB)
2994 break;
2995 cond_resched();
2996 }
2997
2998 if (err == 0)
2999 goto out_put;
3000
3001 if (datagrams == 0) {
3002 datagrams = err;
3003 goto out_put;
3004 }
3005
3006 /*
3007 * We may return less entries than requested (vlen) if the
3008 * sock is non block and there aren't enough datagrams...
3009 */
3010 if (err != -EAGAIN) {
3011 /*
3012 * ... or if recvmsg returns an error after we
3013 * received some datagrams, where we record the
3014 * error to return on the next call or if the
3015 * app asks about it using getsockopt(SO_ERROR).
3016 */
3017 WRITE_ONCE(sock->sk->sk_err, -err);
3018 }
3019 out_put:
3020 fput_light(sock->file, fput_needed);
3021
3022 return datagrams;
3023 }
3024
3025 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
3026 unsigned int vlen, unsigned int flags,
3027 struct __kernel_timespec __user *timeout,
3028 struct old_timespec32 __user *timeout32)
3029 {
3030 int datagrams;
3031 struct timespec64 timeout_sys;
3032
3033 if (timeout && get_timespec64(&timeout_sys, timeout))
3034 return -EFAULT;
3035
3036 if (timeout32 && get_old_timespec32(&timeout_sys, timeout32))
3037 return -EFAULT;
3038
3039 if (!timeout && !timeout32)
3040 return do_recvmmsg(fd, mmsg, vlen, flags, NULL);
3041
3042 datagrams = do_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
3043
3044 if (datagrams <= 0)
3045 return datagrams;
3046
3047 if (timeout && put_timespec64(&timeout_sys, timeout))
3048 datagrams = -EFAULT;
3049
3050 if (timeout32 && put_old_timespec32(&timeout_sys, timeout32))
3051 datagrams = -EFAULT;
3052
3053 return datagrams;
3054 }
3055
3056 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
3057 unsigned int, vlen, unsigned int, flags,
3058 struct __kernel_timespec __user *, timeout)
3059 {
3060 if (flags & MSG_CMSG_COMPAT)
3061 return -EINVAL;
3062
3063 return __sys_recvmmsg(fd, mmsg, vlen, flags, timeout, NULL);
3064 }
3065
3066 #ifdef CONFIG_COMPAT_32BIT_TIME
3067 SYSCALL_DEFINE5(recvmmsg_time32, int, fd, struct mmsghdr __user *, mmsg,
3068 unsigned int, vlen, unsigned int, flags,
3069 struct old_timespec32 __user *, timeout)
3070 {
3071 if (flags & MSG_CMSG_COMPAT)
3072 return -EINVAL;
3073
3074 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL, timeout);
3075 }
3076 #endif
3077
3078 #ifdef __ARCH_WANT_SYS_SOCKETCALL
3079 /* Argument list sizes for sys_socketcall */
3080 #define AL(x) ((x) * sizeof(unsigned long))
3081 static const unsigned char nargs[21] = {
3082 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
3083 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
3084 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
3085 AL(4), AL(5), AL(4)
3086 };
3087
3088 #undef AL
3089
3090 /*
3091 * System call vectors.
3092 *
3093 * Argument checking cleaned up. Saved 20% in size.
3094 * This function doesn't need to set the kernel lock because
3095 * it is set by the callees.
3096 */
3097
3098 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
3099 {
3100 unsigned long a[AUDITSC_ARGS];
3101 unsigned long a0, a1;
3102 int err;
3103 unsigned int len;
3104
3105 if (call < 1 || call > SYS_SENDMMSG)
3106 return -EINVAL;
3107 call = array_index_nospec(call, SYS_SENDMMSG + 1);
3108
3109 len = nargs[call];
3110 if (len > sizeof(a))
3111 return -EINVAL;
3112
3113 /* copy_from_user should be SMP safe. */
3114 if (copy_from_user(a, args, len))
3115 return -EFAULT;
3116
3117 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
3118 if (err)
3119 return err;
3120
3121 a0 = a[0];
3122 a1 = a[1];
3123
3124 switch (call) {
3125 case SYS_SOCKET:
3126 err = __sys_socket(a0, a1, a[2]);
3127 break;
3128 case SYS_BIND:
3129 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
3130 break;
3131 case SYS_CONNECT:
3132 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
3133 break;
3134 case SYS_LISTEN:
3135 err = __sys_listen(a0, a1);
3136 break;
3137 case SYS_ACCEPT:
3138 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3139 (int __user *)a[2], 0);
3140 break;
3141 case SYS_GETSOCKNAME:
3142 err =
3143 __sys_getsockname(a0, (struct sockaddr __user *)a1,
3144 (int __user *)a[2]);
3145 break;
3146 case SYS_GETPEERNAME:
3147 err =
3148 __sys_getpeername(a0, (struct sockaddr __user *)a1,
3149 (int __user *)a[2]);
3150 break;
3151 case SYS_SOCKETPAIR:
3152 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
3153 break;
3154 case SYS_SEND:
3155 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3156 NULL, 0);
3157 break;
3158 case SYS_SENDTO:
3159 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
3160 (struct sockaddr __user *)a[4], a[5]);
3161 break;
3162 case SYS_RECV:
3163 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3164 NULL, NULL);
3165 break;
3166 case SYS_RECVFROM:
3167 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
3168 (struct sockaddr __user *)a[4],
3169 (int __user *)a[5]);
3170 break;
3171 case SYS_SHUTDOWN:
3172 err = __sys_shutdown(a0, a1);
3173 break;
3174 case SYS_SETSOCKOPT:
3175 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
3176 a[4]);
3177 break;
3178 case SYS_GETSOCKOPT:
3179 err =
3180 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
3181 (int __user *)a[4]);
3182 break;
3183 case SYS_SENDMSG:
3184 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
3185 a[2], true);
3186 break;
3187 case SYS_SENDMMSG:
3188 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
3189 a[3], true);
3190 break;
3191 case SYS_RECVMSG:
3192 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
3193 a[2], true);
3194 break;
3195 case SYS_RECVMMSG:
3196 if (IS_ENABLED(CONFIG_64BIT))
3197 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3198 a[2], a[3],
3199 (struct __kernel_timespec __user *)a[4],
3200 NULL);
3201 else
3202 err = __sys_recvmmsg(a0, (struct mmsghdr __user *)a1,
3203 a[2], a[3], NULL,
3204 (struct old_timespec32 __user *)a[4]);
3205 break;
3206 case SYS_ACCEPT4:
3207 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
3208 (int __user *)a[2], a[3]);
3209 break;
3210 default:
3211 err = -EINVAL;
3212 break;
3213 }
3214 return err;
3215 }
3216
3217 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
3218
3219 /**
3220 * sock_register - add a socket protocol handler
3221 * @ops: description of protocol
3222 *
3223 * This function is called by a protocol handler that wants to
3224 * advertise its address family, and have it linked into the
3225 * socket interface. The value ops->family corresponds to the
3226 * socket system call protocol family.
3227 */
3228 int sock_register(const struct net_proto_family *ops)
3229 {
3230 int err;
3231
3232 if (ops->family >= NPROTO) {
3233 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
3234 return -ENOBUFS;
3235 }
3236
3237 spin_lock(&net_family_lock);
3238 if (rcu_dereference_protected(net_families[ops->family],
3239 lockdep_is_held(&net_family_lock)))
3240 err = -EEXIST;
3241 else {
3242 rcu_assign_pointer(net_families[ops->family], ops);
3243 err = 0;
3244 }
3245 spin_unlock(&net_family_lock);
3246
3247 pr_info("NET: Registered %s protocol family\n", pf_family_names[ops->family]);
3248 return err;
3249 }
3250 EXPORT_SYMBOL(sock_register);
3251
3252 /**
3253 * sock_unregister - remove a protocol handler
3254 * @family: protocol family to remove
3255 *
3256 * This function is called by a protocol handler that wants to
3257 * remove its address family, and have it unlinked from the
3258 * new socket creation.
3259 *
3260 * If protocol handler is a module, then it can use module reference
3261 * counts to protect against new references. If protocol handler is not
3262 * a module then it needs to provide its own protection in
3263 * the ops->create routine.
3264 */
3265 void sock_unregister(int family)
3266 {
3267 BUG_ON(family < 0 || family >= NPROTO);
3268
3269 spin_lock(&net_family_lock);
3270 RCU_INIT_POINTER(net_families[family], NULL);
3271 spin_unlock(&net_family_lock);
3272
3273 synchronize_rcu();
3274
3275 pr_info("NET: Unregistered %s protocol family\n", pf_family_names[family]);
3276 }
3277 EXPORT_SYMBOL(sock_unregister);
3278
3279 bool sock_is_registered(int family)
3280 {
3281 return family < NPROTO && rcu_access_pointer(net_families[family]);
3282 }
3283
3284 static int __init sock_init(void)
3285 {
3286 int err;
3287 /*
3288 * Initialize the network sysctl infrastructure.
3289 */
3290 err = net_sysctl_init();
3291 if (err)
3292 goto out;
3293
3294 /*
3295 * Initialize skbuff SLAB cache
3296 */
3297 skb_init();
3298
3299 /*
3300 * Initialize the protocols module.
3301 */
3302
3303 init_inodecache();
3304
3305 err = register_filesystem(&sock_fs_type);
3306 if (err)
3307 goto out;
3308 sock_mnt = kern_mount(&sock_fs_type);
3309 if (IS_ERR(sock_mnt)) {
3310 err = PTR_ERR(sock_mnt);
3311 goto out_mount;
3312 }
3313
3314 /* The real protocol initialization is performed in later initcalls.
3315 */
3316
3317 #ifdef CONFIG_NETFILTER
3318 err = netfilter_init();
3319 if (err)
3320 goto out;
3321 #endif
3322
3323 ptp_classifier_init();
3324
3325 out:
3326 return err;
3327
3328 out_mount:
3329 unregister_filesystem(&sock_fs_type);
3330 goto out;
3331 }
3332
3333 core_initcall(sock_init); /* early initcall */
3334
3335 #ifdef CONFIG_PROC_FS
3336 void socket_seq_show(struct seq_file *seq)
3337 {
3338 seq_printf(seq, "sockets: used %d\n",
3339 sock_inuse_get(seq->private));
3340 }
3341 #endif /* CONFIG_PROC_FS */
3342
3343 /* Handle the fact that while struct ifreq has the same *layout* on
3344 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
3345 * which are handled elsewhere, it still has different *size* due to
3346 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
3347 * resulting in struct ifreq being 32 and 40 bytes respectively).
3348 * As a result, if the struct happens to be at the end of a page and
3349 * the next page isn't readable/writable, we get a fault. To prevent
3350 * that, copy back and forth to the full size.
3351 */
3352 int get_user_ifreq(struct ifreq *ifr, void __user **ifrdata, void __user *arg)
3353 {
3354 if (in_compat_syscall()) {
3355 struct compat_ifreq *ifr32 = (struct compat_ifreq *)ifr;
3356
3357 memset(ifr, 0, sizeof(*ifr));
3358 if (copy_from_user(ifr32, arg, sizeof(*ifr32)))
3359 return -EFAULT;
3360
3361 if (ifrdata)
3362 *ifrdata = compat_ptr(ifr32->ifr_data);
3363
3364 return 0;
3365 }
3366
3367 if (copy_from_user(ifr, arg, sizeof(*ifr)))
3368 return -EFAULT;
3369
3370 if (ifrdata)
3371 *ifrdata = ifr->ifr_data;
3372
3373 return 0;
3374 }
3375 EXPORT_SYMBOL(get_user_ifreq);
3376
3377 int put_user_ifreq(struct ifreq *ifr, void __user *arg)
3378 {
3379 size_t size = sizeof(*ifr);
3380
3381 if (in_compat_syscall())
3382 size = sizeof(struct compat_ifreq);
3383
3384 if (copy_to_user(arg, ifr, size))
3385 return -EFAULT;
3386
3387 return 0;
3388 }
3389 EXPORT_SYMBOL(put_user_ifreq);
3390
3391 #ifdef CONFIG_COMPAT
3392 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
3393 {
3394 compat_uptr_t uptr32;
3395 struct ifreq ifr;
3396 void __user *saved;
3397 int err;
3398
3399 if (get_user_ifreq(&ifr, NULL, uifr32))
3400 return -EFAULT;
3401
3402 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
3403 return -EFAULT;
3404
3405 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
3406 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
3407
3408 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL, NULL);
3409 if (!err) {
3410 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
3411 if (put_user_ifreq(&ifr, uifr32))
3412 err = -EFAULT;
3413 }
3414 return err;
3415 }
3416
3417 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
3418 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
3419 struct compat_ifreq __user *u_ifreq32)
3420 {
3421 struct ifreq ifreq;
3422 void __user *data;
3423
3424 if (!is_socket_ioctl_cmd(cmd))
3425 return -ENOTTY;
3426 if (get_user_ifreq(&ifreq, &data, u_ifreq32))
3427 return -EFAULT;
3428 ifreq.ifr_data = data;
3429
3430 return dev_ioctl(net, cmd, &ifreq, data, NULL);
3431 }
3432
3433 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3434 unsigned int cmd, unsigned long arg)
3435 {
3436 void __user *argp = compat_ptr(arg);
3437 struct sock *sk = sock->sk;
3438 struct net *net = sock_net(sk);
3439 const struct proto_ops *ops;
3440
3441 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3442 return sock_ioctl(file, cmd, (unsigned long)argp);
3443
3444 switch (cmd) {
3445 case SIOCWANDEV:
3446 return compat_siocwandev(net, argp);
3447 case SIOCGSTAMP_OLD:
3448 case SIOCGSTAMPNS_OLD:
3449 ops = READ_ONCE(sock->ops);
3450 if (!ops->gettstamp)
3451 return -ENOIOCTLCMD;
3452 return ops->gettstamp(sock, argp, cmd == SIOCGSTAMP_OLD,
3453 !COMPAT_USE_64BIT_TIME);
3454
3455 case SIOCETHTOOL:
3456 case SIOCBONDSLAVEINFOQUERY:
3457 case SIOCBONDINFOQUERY:
3458 case SIOCSHWTSTAMP:
3459 case SIOCGHWTSTAMP:
3460 return compat_ifr_data_ioctl(net, cmd, argp);
3461
3462 case FIOSETOWN:
3463 case SIOCSPGRP:
3464 case FIOGETOWN:
3465 case SIOCGPGRP:
3466 case SIOCBRADDBR:
3467 case SIOCBRDELBR:
3468 case SIOCGIFVLAN:
3469 case SIOCSIFVLAN:
3470 case SIOCGSKNS:
3471 case SIOCGSTAMP_NEW:
3472 case SIOCGSTAMPNS_NEW:
3473 case SIOCGIFCONF:
3474 case SIOCSIFBR:
3475 case SIOCGIFBR:
3476 return sock_ioctl(file, cmd, arg);
3477
3478 case SIOCGIFFLAGS:
3479 case SIOCSIFFLAGS:
3480 case SIOCGIFMAP:
3481 case SIOCSIFMAP:
3482 case SIOCGIFMETRIC:
3483 case SIOCSIFMETRIC:
3484 case SIOCGIFMTU:
3485 case SIOCSIFMTU:
3486 case SIOCGIFMEM:
3487 case SIOCSIFMEM:
3488 case SIOCGIFHWADDR:
3489 case SIOCSIFHWADDR:
3490 case SIOCADDMULTI:
3491 case SIOCDELMULTI:
3492 case SIOCGIFINDEX:
3493 case SIOCGIFADDR:
3494 case SIOCSIFADDR:
3495 case SIOCSIFHWBROADCAST:
3496 case SIOCDIFADDR:
3497 case SIOCGIFBRDADDR:
3498 case SIOCSIFBRDADDR:
3499 case SIOCGIFDSTADDR:
3500 case SIOCSIFDSTADDR:
3501 case SIOCGIFNETMASK:
3502 case SIOCSIFNETMASK:
3503 case SIOCSIFPFLAGS:
3504 case SIOCGIFPFLAGS:
3505 case SIOCGIFTXQLEN:
3506 case SIOCSIFTXQLEN:
3507 case SIOCBRADDIF:
3508 case SIOCBRDELIF:
3509 case SIOCGIFNAME:
3510 case SIOCSIFNAME:
3511 case SIOCGMIIPHY:
3512 case SIOCGMIIREG:
3513 case SIOCSMIIREG:
3514 case SIOCBONDENSLAVE:
3515 case SIOCBONDRELEASE:
3516 case SIOCBONDSETHWADDR:
3517 case SIOCBONDCHANGEACTIVE:
3518 case SIOCSARP:
3519 case SIOCGARP:
3520 case SIOCDARP:
3521 case SIOCOUTQ:
3522 case SIOCOUTQNSD:
3523 case SIOCATMARK:
3524 return sock_do_ioctl(net, sock, cmd, arg);
3525 }
3526
3527 return -ENOIOCTLCMD;
3528 }
3529
3530 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3531 unsigned long arg)
3532 {
3533 struct socket *sock = file->private_data;
3534 const struct proto_ops *ops = READ_ONCE(sock->ops);
3535 int ret = -ENOIOCTLCMD;
3536 struct sock *sk;
3537 struct net *net;
3538
3539 sk = sock->sk;
3540 net = sock_net(sk);
3541
3542 if (ops->compat_ioctl)
3543 ret = ops->compat_ioctl(sock, cmd, arg);
3544
3545 if (ret == -ENOIOCTLCMD &&
3546 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3547 ret = compat_wext_handle_ioctl(net, cmd, arg);
3548
3549 if (ret == -ENOIOCTLCMD)
3550 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3551
3552 return ret;
3553 }
3554 #endif
3555
3556 /**
3557 * kernel_bind - bind an address to a socket (kernel space)
3558 * @sock: socket
3559 * @addr: address
3560 * @addrlen: length of address
3561 *
3562 * Returns 0 or an error.
3563 */
3564
3565 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3566 {
3567 struct sockaddr_storage address;
3568
3569 memcpy(&address, addr, addrlen);
3570
3571 return READ_ONCE(sock->ops)->bind(sock, (struct sockaddr *)&address,
3572 addrlen);
3573 }
3574 EXPORT_SYMBOL(kernel_bind);
3575
3576 /**
3577 * kernel_listen - move socket to listening state (kernel space)
3578 * @sock: socket
3579 * @backlog: pending connections queue size
3580 *
3581 * Returns 0 or an error.
3582 */
3583
3584 int kernel_listen(struct socket *sock, int backlog)
3585 {
3586 return READ_ONCE(sock->ops)->listen(sock, backlog);
3587 }
3588 EXPORT_SYMBOL(kernel_listen);
3589
3590 /**
3591 * kernel_accept - accept a connection (kernel space)
3592 * @sock: listening socket
3593 * @newsock: new connected socket
3594 * @flags: flags
3595 *
3596 * @flags must be SOCK_CLOEXEC, SOCK_NONBLOCK or 0.
3597 * If it fails, @newsock is guaranteed to be %NULL.
3598 * Returns 0 or an error.
3599 */
3600
3601 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3602 {
3603 struct sock *sk = sock->sk;
3604 const struct proto_ops *ops = READ_ONCE(sock->ops);
3605 struct proto_accept_arg arg = {
3606 .flags = flags,
3607 .kern = true,
3608 };
3609 int err;
3610
3611 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3612 newsock);
3613 if (err < 0)
3614 goto done;
3615
3616 err = ops->accept(sock, *newsock, &arg);
3617 if (err < 0) {
3618 sock_release(*newsock);
3619 *newsock = NULL;
3620 goto done;
3621 }
3622
3623 (*newsock)->ops = ops;
3624 __module_get(ops->owner);
3625
3626 done:
3627 return err;
3628 }
3629 EXPORT_SYMBOL(kernel_accept);
3630
3631 /**
3632 * kernel_connect - connect a socket (kernel space)
3633 * @sock: socket
3634 * @addr: address
3635 * @addrlen: address length
3636 * @flags: flags (O_NONBLOCK, ...)
3637 *
3638 * For datagram sockets, @addr is the address to which datagrams are sent
3639 * by default, and the only address from which datagrams are received.
3640 * For stream sockets, attempts to connect to @addr.
3641 * Returns 0 or an error code.
3642 */
3643
3644 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3645 int flags)
3646 {
3647 struct sockaddr_storage address;
3648
3649 memcpy(&address, addr, addrlen);
3650
3651 return READ_ONCE(sock->ops)->connect(sock, (struct sockaddr *)&address,
3652 addrlen, flags);
3653 }
3654 EXPORT_SYMBOL(kernel_connect);
3655
3656 /**
3657 * kernel_getsockname - get the address which the socket is bound (kernel space)
3658 * @sock: socket
3659 * @addr: address holder
3660 *
3661 * Fills the @addr pointer with the address which the socket is bound.
3662 * Returns the length of the address in bytes or an error code.
3663 */
3664
3665 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3666 {
3667 return READ_ONCE(sock->ops)->getname(sock, addr, 0);
3668 }
3669 EXPORT_SYMBOL(kernel_getsockname);
3670
3671 /**
3672 * kernel_getpeername - get the address which the socket is connected (kernel space)
3673 * @sock: socket
3674 * @addr: address holder
3675 *
3676 * Fills the @addr pointer with the address which the socket is connected.
3677 * Returns the length of the address in bytes or an error code.
3678 */
3679
3680 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3681 {
3682 return READ_ONCE(sock->ops)->getname(sock, addr, 1);
3683 }
3684 EXPORT_SYMBOL(kernel_getpeername);
3685
3686 /**
3687 * kernel_sock_shutdown - shut down part of a full-duplex connection (kernel space)
3688 * @sock: socket
3689 * @how: connection part
3690 *
3691 * Returns 0 or an error.
3692 */
3693
3694 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3695 {
3696 return READ_ONCE(sock->ops)->shutdown(sock, how);
3697 }
3698 EXPORT_SYMBOL(kernel_sock_shutdown);
3699
3700 /**
3701 * kernel_sock_ip_overhead - returns the IP overhead imposed by a socket
3702 * @sk: socket
3703 *
3704 * This routine returns the IP overhead imposed by a socket i.e.
3705 * the length of the underlying IP header, depending on whether
3706 * this is an IPv4 or IPv6 socket and the length from IP options turned
3707 * on at the socket. Assumes that the caller has a lock on the socket.
3708 */
3709
3710 u32 kernel_sock_ip_overhead(struct sock *sk)
3711 {
3712 struct inet_sock *inet;
3713 struct ip_options_rcu *opt;
3714 u32 overhead = 0;
3715 #if IS_ENABLED(CONFIG_IPV6)
3716 struct ipv6_pinfo *np;
3717 struct ipv6_txoptions *optv6 = NULL;
3718 #endif /* IS_ENABLED(CONFIG_IPV6) */
3719
3720 if (!sk)
3721 return overhead;
3722
3723 switch (sk->sk_family) {
3724 case AF_INET:
3725 inet = inet_sk(sk);
3726 overhead += sizeof(struct iphdr);
3727 opt = rcu_dereference_protected(inet->inet_opt,
3728 sock_owned_by_user(sk));
3729 if (opt)
3730 overhead += opt->opt.optlen;
3731 return overhead;
3732 #if IS_ENABLED(CONFIG_IPV6)
3733 case AF_INET6:
3734 np = inet6_sk(sk);
3735 overhead += sizeof(struct ipv6hdr);
3736 if (np)
3737 optv6 = rcu_dereference_protected(np->opt,
3738 sock_owned_by_user(sk));
3739 if (optv6)
3740 overhead += (optv6->opt_flen + optv6->opt_nflen);
3741 return overhead;
3742 #endif /* IS_ENABLED(CONFIG_IPV6) */
3743 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3744 return overhead;
3745 }
3746 }
3747 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3748
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.