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