1 /*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61 #include <linux/mm.h>
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
92 #include <linux/nospec.h>
93
94 #include <linux/uaccess.h>
95 #include <asm/unistd.h>
96
97 #include <net/compat.h>
98 #include <net/wext.h>
99 #include <net/cls_cgroup.h>
100
101 #include <net/sock.h>
102 #include <linux/netfilter.h>
103
104 #include <linux/if_tun.h>
105 #include <linux/ipv6_route.h>
106 #include <linux/route.h>
107 #include <linux/sockios.h>
108 #include <net/busy_poll.h>
109 #include <linux/errqueue.h>
110
111 #ifdef CONFIG_NET_RX_BUSY_POLL
112 unsigned int sysctl_net_busy_read __read_mostly;
113 unsigned int sysctl_net_busy_poll __read_mostly;
114 #endif
115
116 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119
120 static int sock_close(struct inode *inode, struct file *file);
121 static __poll_t sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124 #ifdef CONFIG_COMPAT
125 static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
127 #endif
128 static int sock_fasync(int fd, struct file *filp, int on);
129 static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
132 struct pipe_inode_info *pipe, size_t len,
133 unsigned int flags);
134
135 /*
136 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
137 * in the operation structures but are done directly via the socketcall() multiplexor.
138 */
139
140 static const struct file_operations socket_file_ops = {
141 .owner = THIS_MODULE,
142 .llseek = no_llseek,
143 .read_iter = sock_read_iter,
144 .write_iter = sock_write_iter,
145 .poll = sock_poll,
146 .unlocked_ioctl = sock_ioctl,
147 #ifdef CONFIG_COMPAT
148 .compat_ioctl = compat_sock_ioctl,
149 #endif
150 .mmap = sock_mmap,
151 .release = sock_close,
152 .fasync = sock_fasync,
153 .sendpage = sock_sendpage,
154 .splice_write = generic_splice_sendpage,
155 .splice_read = sock_splice_read,
156 };
157
158 /*
159 * The protocol list. Each protocol is registered in here.
160 */
161
162 static DEFINE_SPINLOCK(net_family_lock);
163 static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164
165 /*
166 * Support routines.
167 * Move socket addresses back and forth across the kernel/user
168 * divide and look after the messy bits.
169 */
170
171 /**
172 * move_addr_to_kernel - copy a socket address into kernel space
173 * @uaddr: Address in user space
174 * @kaddr: Address in kernel space
175 * @ulen: Length in user space
176 *
177 * The address is copied into kernel space. If the provided address is
178 * too long an error code of -EINVAL is returned. If the copy gives
179 * invalid addresses -EFAULT is returned. On a success 0 is returned.
180 */
181
182 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
183 {
184 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
185 return -EINVAL;
186 if (ulen == 0)
187 return 0;
188 if (copy_from_user(kaddr, uaddr, ulen))
189 return -EFAULT;
190 return audit_sockaddr(ulen, kaddr);
191 }
192
193 /**
194 * move_addr_to_user - copy an address to user space
195 * @kaddr: kernel space address
196 * @klen: length of address in kernel
197 * @uaddr: user space address
198 * @ulen: pointer to user length field
199 *
200 * The value pointed to by ulen on entry is the buffer length available.
201 * This is overwritten with the buffer space used. -EINVAL is returned
202 * if an overlong buffer is specified or a negative buffer size. -EFAULT
203 * is returned if either the buffer or the length field are not
204 * accessible.
205 * After copying the data up to the limit the user specifies, the true
206 * length of the data is written over the length limit the user
207 * specified. Zero is returned for a success.
208 */
209
210 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
211 void __user *uaddr, int __user *ulen)
212 {
213 int err;
214 int len;
215
216 BUG_ON(klen > sizeof(struct sockaddr_storage));
217 err = get_user(len, ulen);
218 if (err)
219 return err;
220 if (len > klen)
221 len = klen;
222 if (len < 0)
223 return -EINVAL;
224 if (len) {
225 if (audit_sockaddr(klen, kaddr))
226 return -ENOMEM;
227 if (copy_to_user(uaddr, kaddr, len))
228 return -EFAULT;
229 }
230 /*
231 * "fromlen shall refer to the value before truncation.."
232 * 1003.1g
233 */
234 return __put_user(klen, ulen);
235 }
236
237 static struct kmem_cache *sock_inode_cachep __ro_after_init;
238
239 static struct inode *sock_alloc_inode(struct super_block *sb)
240 {
241 struct socket_alloc *ei;
242 struct socket_wq *wq;
243
244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
245 if (!ei)
246 return NULL;
247 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
248 if (!wq) {
249 kmem_cache_free(sock_inode_cachep, ei);
250 return NULL;
251 }
252 init_waitqueue_head(&wq->wait);
253 wq->fasync_list = NULL;
254 wq->flags = 0;
255 ei->socket.wq = wq;
256
257 ei->socket.state = SS_UNCONNECTED;
258 ei->socket.flags = 0;
259 ei->socket.ops = NULL;
260 ei->socket.sk = NULL;
261 ei->socket.file = NULL;
262
263 return &ei->vfs_inode;
264 }
265
266 static void sock_destroy_inode(struct inode *inode)
267 {
268 struct socket_alloc *ei;
269
270 ei = container_of(inode, struct socket_alloc, vfs_inode);
271 kfree_rcu(ei->socket.wq, rcu);
272 kmem_cache_free(sock_inode_cachep, ei);
273 }
274
275 static void init_once(void *foo)
276 {
277 struct socket_alloc *ei = (struct socket_alloc *)foo;
278
279 inode_init_once(&ei->vfs_inode);
280 }
281
282 static void init_inodecache(void)
283 {
284 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
285 sizeof(struct socket_alloc),
286 0,
287 (SLAB_HWCACHE_ALIGN |
288 SLAB_RECLAIM_ACCOUNT |
289 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
290 init_once);
291 BUG_ON(sock_inode_cachep == NULL);
292 }
293
294 static const struct super_operations sockfs_ops = {
295 .alloc_inode = sock_alloc_inode,
296 .destroy_inode = sock_destroy_inode,
297 .statfs = simple_statfs,
298 };
299
300 /*
301 * sockfs_dname() is called from d_path().
302 */
303 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
304 {
305 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
306 d_inode(dentry)->i_ino);
307 }
308
309 static const struct dentry_operations sockfs_dentry_operations = {
310 .d_dname = sockfs_dname,
311 };
312
313 static int sockfs_xattr_get(const struct xattr_handler *handler,
314 struct dentry *dentry, struct inode *inode,
315 const char *suffix, void *value, size_t size)
316 {
317 if (value) {
318 if (dentry->d_name.len + 1 > size)
319 return -ERANGE;
320 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
321 }
322 return dentry->d_name.len + 1;
323 }
324
325 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
326 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
327 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
328
329 static const struct xattr_handler sockfs_xattr_handler = {
330 .name = XATTR_NAME_SOCKPROTONAME,
331 .get = sockfs_xattr_get,
332 };
333
334 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
335 struct dentry *dentry, struct inode *inode,
336 const char *suffix, const void *value,
337 size_t size, int flags)
338 {
339 /* Handled by LSM. */
340 return -EAGAIN;
341 }
342
343 static const struct xattr_handler sockfs_security_xattr_handler = {
344 .prefix = XATTR_SECURITY_PREFIX,
345 .set = sockfs_security_xattr_set,
346 };
347
348 static const struct xattr_handler *sockfs_xattr_handlers[] = {
349 &sockfs_xattr_handler,
350 &sockfs_security_xattr_handler,
351 NULL
352 };
353
354 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
355 int flags, const char *dev_name, void *data)
356 {
357 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
358 sockfs_xattr_handlers,
359 &sockfs_dentry_operations, SOCKFS_MAGIC);
360 }
361
362 static struct vfsmount *sock_mnt __read_mostly;
363
364 static struct file_system_type sock_fs_type = {
365 .name = "sockfs",
366 .mount = sockfs_mount,
367 .kill_sb = kill_anon_super,
368 };
369
370 /*
371 * Obtains the first available file descriptor and sets it up for use.
372 *
373 * These functions create file structures and maps them to fd space
374 * of the current process. On success it returns file descriptor
375 * and file struct implicitly stored in sock->file.
376 * Note that another thread may close file descriptor before we return
377 * from this function. We use the fact that now we do not refer
378 * to socket after mapping. If one day we will need it, this
379 * function will increment ref. count on file by 1.
380 *
381 * In any case returned fd MAY BE not valid!
382 * This race condition is unavoidable
383 * with shared fd spaces, we cannot solve it inside kernel,
384 * but we take care of internal coherence yet.
385 */
386
387 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
388 {
389 struct file *file;
390
391 if (!dname)
392 dname = sock->sk ? sock->sk->sk_prot_creator->name : "";
393
394 file = alloc_file_pseudo(SOCK_INODE(sock), sock_mnt, dname,
395 O_RDWR | (flags & O_NONBLOCK),
396 &socket_file_ops);
397 if (IS_ERR(file)) {
398 sock_release(sock);
399 return file;
400 }
401
402 sock->file = file;
403 file->private_data = sock;
404 return file;
405 }
406 EXPORT_SYMBOL(sock_alloc_file);
407
408 static int sock_map_fd(struct socket *sock, int flags)
409 {
410 struct file *newfile;
411 int fd = get_unused_fd_flags(flags);
412 if (unlikely(fd < 0)) {
413 sock_release(sock);
414 return fd;
415 }
416
417 newfile = sock_alloc_file(sock, flags, NULL);
418 if (likely(!IS_ERR(newfile))) {
419 fd_install(fd, newfile);
420 return fd;
421 }
422
423 put_unused_fd(fd);
424 return PTR_ERR(newfile);
425 }
426
427 struct socket *sock_from_file(struct file *file, int *err)
428 {
429 if (file->f_op == &socket_file_ops)
430 return file->private_data; /* set in sock_map_fd */
431
432 *err = -ENOTSOCK;
433 return NULL;
434 }
435 EXPORT_SYMBOL(sock_from_file);
436
437 /**
438 * sockfd_lookup - Go from a file number to its socket slot
439 * @fd: file handle
440 * @err: pointer to an error code return
441 *
442 * The file handle passed in is locked and the socket it is bound
443 * to is returned. If an error occurs the err pointer is overwritten
444 * with a negative errno code and NULL is returned. The function checks
445 * for both invalid handles and passing a handle which is not a socket.
446 *
447 * On a success the socket object pointer is returned.
448 */
449
450 struct socket *sockfd_lookup(int fd, int *err)
451 {
452 struct file *file;
453 struct socket *sock;
454
455 file = fget(fd);
456 if (!file) {
457 *err = -EBADF;
458 return NULL;
459 }
460
461 sock = sock_from_file(file, err);
462 if (!sock)
463 fput(file);
464 return sock;
465 }
466 EXPORT_SYMBOL(sockfd_lookup);
467
468 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
469 {
470 struct fd f = fdget(fd);
471 struct socket *sock;
472
473 *err = -EBADF;
474 if (f.file) {
475 sock = sock_from_file(f.file, err);
476 if (likely(sock)) {
477 *fput_needed = f.flags;
478 return sock;
479 }
480 fdput(f);
481 }
482 return NULL;
483 }
484
485 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
486 size_t size)
487 {
488 ssize_t len;
489 ssize_t used = 0;
490
491 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
492 if (len < 0)
493 return len;
494 used += len;
495 if (buffer) {
496 if (size < used)
497 return -ERANGE;
498 buffer += len;
499 }
500
501 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
502 used += len;
503 if (buffer) {
504 if (size < used)
505 return -ERANGE;
506 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
507 buffer += len;
508 }
509
510 return used;
511 }
512
513 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
514 {
515 int err = simple_setattr(dentry, iattr);
516
517 if (!err && (iattr->ia_valid & ATTR_UID)) {
518 struct socket *sock = SOCKET_I(d_inode(dentry));
519
520 if (sock->sk)
521 sock->sk->sk_uid = iattr->ia_uid;
522 else
523 err = -ENOENT;
524 }
525
526 return err;
527 }
528
529 static const struct inode_operations sockfs_inode_ops = {
530 .listxattr = sockfs_listxattr,
531 .setattr = sockfs_setattr,
532 };
533
534 /**
535 * sock_alloc - allocate a socket
536 *
537 * Allocate a new inode and socket object. The two are bound together
538 * and initialised. The socket is then returned. If we are out of inodes
539 * NULL is returned.
540 */
541
542 struct socket *sock_alloc(void)
543 {
544 struct inode *inode;
545 struct socket *sock;
546
547 inode = new_inode_pseudo(sock_mnt->mnt_sb);
548 if (!inode)
549 return NULL;
550
551 sock = SOCKET_I(inode);
552
553 inode->i_ino = get_next_ino();
554 inode->i_mode = S_IFSOCK | S_IRWXUGO;
555 inode->i_uid = current_fsuid();
556 inode->i_gid = current_fsgid();
557 inode->i_op = &sockfs_inode_ops;
558
559 return sock;
560 }
561 EXPORT_SYMBOL(sock_alloc);
562
563 /**
564 * sock_release - close a socket
565 * @sock: socket to close
566 *
567 * The socket is released from the protocol stack if it has a release
568 * callback, and the inode is then released if the socket is bound to
569 * an inode not a file.
570 */
571
572 static void __sock_release(struct socket *sock, struct inode *inode)
573 {
574 if (sock->ops) {
575 struct module *owner = sock->ops->owner;
576
577 if (inode)
578 inode_lock(inode);
579 sock->ops->release(sock);
580 sock->sk = NULL;
581 if (inode)
582 inode_unlock(inode);
583 sock->ops = NULL;
584 module_put(owner);
585 }
586
587 if (sock->wq->fasync_list)
588 pr_err("%s: fasync list not empty!\n", __func__);
589
590 if (!sock->file) {
591 iput(SOCK_INODE(sock));
592 return;
593 }
594 sock->file = NULL;
595 }
596
597 void sock_release(struct socket *sock)
598 {
599 __sock_release(sock, NULL);
600 }
601 EXPORT_SYMBOL(sock_release);
602
603 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
604 {
605 u8 flags = *tx_flags;
606
607 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
608 flags |= SKBTX_HW_TSTAMP;
609
610 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
611 flags |= SKBTX_SW_TSTAMP;
612
613 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
614 flags |= SKBTX_SCHED_TSTAMP;
615
616 *tx_flags = flags;
617 }
618 EXPORT_SYMBOL(__sock_tx_timestamp);
619
620 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
621 {
622 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
623 BUG_ON(ret == -EIOCBQUEUED);
624 return ret;
625 }
626
627 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
628 {
629 int err = security_socket_sendmsg(sock, msg,
630 msg_data_left(msg));
631
632 return err ?: sock_sendmsg_nosec(sock, msg);
633 }
634 EXPORT_SYMBOL(sock_sendmsg);
635
636 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
637 struct kvec *vec, size_t num, size_t size)
638 {
639 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
640 return sock_sendmsg(sock, msg);
641 }
642 EXPORT_SYMBOL(kernel_sendmsg);
643
644 int kernel_sendmsg_locked(struct sock *sk, struct msghdr *msg,
645 struct kvec *vec, size_t num, size_t size)
646 {
647 struct socket *sock = sk->sk_socket;
648
649 if (!sock->ops->sendmsg_locked)
650 return sock_no_sendmsg_locked(sk, msg, size);
651
652 iov_iter_kvec(&msg->msg_iter, WRITE, vec, num, size);
653
654 return sock->ops->sendmsg_locked(sk, msg, msg_data_left(msg));
655 }
656 EXPORT_SYMBOL(kernel_sendmsg_locked);
657
658 static bool skb_is_err_queue(const struct sk_buff *skb)
659 {
660 /* pkt_type of skbs enqueued on the error queue are set to
661 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
662 * in recvmsg, since skbs received on a local socket will never
663 * have a pkt_type of PACKET_OUTGOING.
664 */
665 return skb->pkt_type == PACKET_OUTGOING;
666 }
667
668 /* On transmit, software and hardware timestamps are returned independently.
669 * As the two skb clones share the hardware timestamp, which may be updated
670 * before the software timestamp is received, a hardware TX timestamp may be
671 * returned only if there is no software TX timestamp. Ignore false software
672 * timestamps, which may be made in the __sock_recv_timestamp() call when the
673 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
674 * hardware timestamp.
675 */
676 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
677 {
678 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
679 }
680
681 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
682 {
683 struct scm_ts_pktinfo ts_pktinfo;
684 struct net_device *orig_dev;
685
686 if (!skb_mac_header_was_set(skb))
687 return;
688
689 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
690
691 rcu_read_lock();
692 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
693 if (orig_dev)
694 ts_pktinfo.if_index = orig_dev->ifindex;
695 rcu_read_unlock();
696
697 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
698 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
699 sizeof(ts_pktinfo), &ts_pktinfo);
700 }
701
702 /*
703 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
704 */
705 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
706 struct sk_buff *skb)
707 {
708 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
709 struct scm_timestamping tss;
710 int empty = 1, false_tstamp = 0;
711 struct skb_shared_hwtstamps *shhwtstamps =
712 skb_hwtstamps(skb);
713
714 /* Race occurred between timestamp enabling and packet
715 receiving. Fill in the current time for now. */
716 if (need_software_tstamp && skb->tstamp == 0) {
717 __net_timestamp(skb);
718 false_tstamp = 1;
719 }
720
721 if (need_software_tstamp) {
722 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
723 struct timeval tv;
724 skb_get_timestamp(skb, &tv);
725 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
726 sizeof(tv), &tv);
727 } else {
728 struct timespec ts;
729 skb_get_timestampns(skb, &ts);
730 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
731 sizeof(ts), &ts);
732 }
733 }
734
735 memset(&tss, 0, sizeof(tss));
736 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
737 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
738 empty = 0;
739 if (shhwtstamps &&
740 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
741 !skb_is_swtx_tstamp(skb, false_tstamp) &&
742 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
743 empty = 0;
744 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
745 !skb_is_err_queue(skb))
746 put_ts_pktinfo(msg, skb);
747 }
748 if (!empty) {
749 put_cmsg(msg, SOL_SOCKET,
750 SCM_TIMESTAMPING, sizeof(tss), &tss);
751
752 if (skb_is_err_queue(skb) && skb->len &&
753 SKB_EXT_ERR(skb)->opt_stats)
754 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
755 skb->len, skb->data);
756 }
757 }
758 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
759
760 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
761 struct sk_buff *skb)
762 {
763 int ack;
764
765 if (!sock_flag(sk, SOCK_WIFI_STATUS))
766 return;
767 if (!skb->wifi_acked_valid)
768 return;
769
770 ack = skb->wifi_acked;
771
772 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
773 }
774 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
775
776 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
777 struct sk_buff *skb)
778 {
779 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
780 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
781 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
782 }
783
784 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
785 struct sk_buff *skb)
786 {
787 sock_recv_timestamp(msg, sk, skb);
788 sock_recv_drops(msg, sk, skb);
789 }
790 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
791
792 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
793 int flags)
794 {
795 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
796 }
797
798 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
799 {
800 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
801
802 return err ?: sock_recvmsg_nosec(sock, msg, flags);
803 }
804 EXPORT_SYMBOL(sock_recvmsg);
805
806 /**
807 * kernel_recvmsg - Receive a message from a socket (kernel space)
808 * @sock: The socket to receive the message from
809 * @msg: Received message
810 * @vec: Input s/g array for message data
811 * @num: Size of input s/g array
812 * @size: Number of bytes to read
813 * @flags: Message flags (MSG_DONTWAIT, etc...)
814 *
815 * On return the msg structure contains the scatter/gather array passed in the
816 * vec argument. The array is modified so that it consists of the unfilled
817 * portion of the original array.
818 *
819 * The returned value is the total number of bytes received, or an error.
820 */
821 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
822 struct kvec *vec, size_t num, size_t size, int flags)
823 {
824 mm_segment_t oldfs = get_fs();
825 int result;
826
827 iov_iter_kvec(&msg->msg_iter, READ, vec, num, size);
828 set_fs(KERNEL_DS);
829 result = sock_recvmsg(sock, msg, flags);
830 set_fs(oldfs);
831 return result;
832 }
833 EXPORT_SYMBOL(kernel_recvmsg);
834
835 static ssize_t sock_sendpage(struct file *file, struct page *page,
836 int offset, size_t size, loff_t *ppos, int more)
837 {
838 struct socket *sock;
839 int flags;
840
841 sock = file->private_data;
842
843 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
844 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
845 flags |= more;
846
847 return kernel_sendpage(sock, page, offset, size, flags);
848 }
849
850 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
851 struct pipe_inode_info *pipe, size_t len,
852 unsigned int flags)
853 {
854 struct socket *sock = file->private_data;
855
856 if (unlikely(!sock->ops->splice_read))
857 return generic_file_splice_read(file, ppos, pipe, len, flags);
858
859 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
860 }
861
862 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
863 {
864 struct file *file = iocb->ki_filp;
865 struct socket *sock = file->private_data;
866 struct msghdr msg = {.msg_iter = *to,
867 .msg_iocb = iocb};
868 ssize_t res;
869
870 if (file->f_flags & O_NONBLOCK)
871 msg.msg_flags = MSG_DONTWAIT;
872
873 if (iocb->ki_pos != 0)
874 return -ESPIPE;
875
876 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
877 return 0;
878
879 res = sock_recvmsg(sock, &msg, msg.msg_flags);
880 *to = msg.msg_iter;
881 return res;
882 }
883
884 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
885 {
886 struct file *file = iocb->ki_filp;
887 struct socket *sock = file->private_data;
888 struct msghdr msg = {.msg_iter = *from,
889 .msg_iocb = iocb};
890 ssize_t res;
891
892 if (iocb->ki_pos != 0)
893 return -ESPIPE;
894
895 if (file->f_flags & O_NONBLOCK)
896 msg.msg_flags = MSG_DONTWAIT;
897
898 if (sock->type == SOCK_SEQPACKET)
899 msg.msg_flags |= MSG_EOR;
900
901 res = sock_sendmsg(sock, &msg);
902 *from = msg.msg_iter;
903 return res;
904 }
905
906 /*
907 * Atomic setting of ioctl hooks to avoid race
908 * with module unload.
909 */
910
911 static DEFINE_MUTEX(br_ioctl_mutex);
912 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
913
914 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
915 {
916 mutex_lock(&br_ioctl_mutex);
917 br_ioctl_hook = hook;
918 mutex_unlock(&br_ioctl_mutex);
919 }
920 EXPORT_SYMBOL(brioctl_set);
921
922 static DEFINE_MUTEX(vlan_ioctl_mutex);
923 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
924
925 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
926 {
927 mutex_lock(&vlan_ioctl_mutex);
928 vlan_ioctl_hook = hook;
929 mutex_unlock(&vlan_ioctl_mutex);
930 }
931 EXPORT_SYMBOL(vlan_ioctl_set);
932
933 static DEFINE_MUTEX(dlci_ioctl_mutex);
934 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
935
936 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
937 {
938 mutex_lock(&dlci_ioctl_mutex);
939 dlci_ioctl_hook = hook;
940 mutex_unlock(&dlci_ioctl_mutex);
941 }
942 EXPORT_SYMBOL(dlci_ioctl_set);
943
944 static long sock_do_ioctl(struct net *net, struct socket *sock,
945 unsigned int cmd, unsigned long arg)
946 {
947 int err;
948 void __user *argp = (void __user *)arg;
949
950 err = sock->ops->ioctl(sock, cmd, arg);
951
952 /*
953 * If this ioctl is unknown try to hand it down
954 * to the NIC driver.
955 */
956 if (err != -ENOIOCTLCMD)
957 return err;
958
959 if (cmd == SIOCGIFCONF) {
960 struct ifconf ifc;
961 if (copy_from_user(&ifc, argp, sizeof(struct ifconf)))
962 return -EFAULT;
963 rtnl_lock();
964 err = dev_ifconf(net, &ifc, sizeof(struct ifreq));
965 rtnl_unlock();
966 if (!err && copy_to_user(argp, &ifc, sizeof(struct ifconf)))
967 err = -EFAULT;
968 } else {
969 struct ifreq ifr;
970 bool need_copyout;
971 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
972 return -EFAULT;
973 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
974 if (!err && need_copyout)
975 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
976 return -EFAULT;
977 }
978 return err;
979 }
980
981 /*
982 * With an ioctl, arg may well be a user mode pointer, but we don't know
983 * what to do with it - that's up to the protocol still.
984 */
985
986 struct ns_common *get_net_ns(struct ns_common *ns)
987 {
988 return &get_net(container_of(ns, struct net, ns))->ns;
989 }
990 EXPORT_SYMBOL_GPL(get_net_ns);
991
992 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
993 {
994 struct socket *sock;
995 struct sock *sk;
996 void __user *argp = (void __user *)arg;
997 int pid, err;
998 struct net *net;
999
1000 sock = file->private_data;
1001 sk = sock->sk;
1002 net = sock_net(sk);
1003 if (unlikely(cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))) {
1004 struct ifreq ifr;
1005 bool need_copyout;
1006 if (copy_from_user(&ifr, argp, sizeof(struct ifreq)))
1007 return -EFAULT;
1008 err = dev_ioctl(net, cmd, &ifr, &need_copyout);
1009 if (!err && need_copyout)
1010 if (copy_to_user(argp, &ifr, sizeof(struct ifreq)))
1011 return -EFAULT;
1012 } else
1013 #ifdef CONFIG_WEXT_CORE
1014 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
1015 err = wext_handle_ioctl(net, cmd, argp);
1016 } else
1017 #endif
1018 switch (cmd) {
1019 case FIOSETOWN:
1020 case SIOCSPGRP:
1021 err = -EFAULT;
1022 if (get_user(pid, (int __user *)argp))
1023 break;
1024 err = f_setown(sock->file, pid, 1);
1025 break;
1026 case FIOGETOWN:
1027 case SIOCGPGRP:
1028 err = put_user(f_getown(sock->file),
1029 (int __user *)argp);
1030 break;
1031 case SIOCGIFBR:
1032 case SIOCSIFBR:
1033 case SIOCBRADDBR:
1034 case SIOCBRDELBR:
1035 err = -ENOPKG;
1036 if (!br_ioctl_hook)
1037 request_module("bridge");
1038
1039 mutex_lock(&br_ioctl_mutex);
1040 if (br_ioctl_hook)
1041 err = br_ioctl_hook(net, cmd, argp);
1042 mutex_unlock(&br_ioctl_mutex);
1043 break;
1044 case SIOCGIFVLAN:
1045 case SIOCSIFVLAN:
1046 err = -ENOPKG;
1047 if (!vlan_ioctl_hook)
1048 request_module("8021q");
1049
1050 mutex_lock(&vlan_ioctl_mutex);
1051 if (vlan_ioctl_hook)
1052 err = vlan_ioctl_hook(net, argp);
1053 mutex_unlock(&vlan_ioctl_mutex);
1054 break;
1055 case SIOCADDDLCI:
1056 case SIOCDELDLCI:
1057 err = -ENOPKG;
1058 if (!dlci_ioctl_hook)
1059 request_module("dlci");
1060
1061 mutex_lock(&dlci_ioctl_mutex);
1062 if (dlci_ioctl_hook)
1063 err = dlci_ioctl_hook(cmd, argp);
1064 mutex_unlock(&dlci_ioctl_mutex);
1065 break;
1066 case SIOCGSKNS:
1067 err = -EPERM;
1068 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1069 break;
1070
1071 err = open_related_ns(&net->ns, get_net_ns);
1072 break;
1073 default:
1074 err = sock_do_ioctl(net, sock, cmd, arg);
1075 break;
1076 }
1077 return err;
1078 }
1079
1080 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1081 {
1082 int err;
1083 struct socket *sock = NULL;
1084
1085 err = security_socket_create(family, type, protocol, 1);
1086 if (err)
1087 goto out;
1088
1089 sock = sock_alloc();
1090 if (!sock) {
1091 err = -ENOMEM;
1092 goto out;
1093 }
1094
1095 sock->type = type;
1096 err = security_socket_post_create(sock, family, type, protocol, 1);
1097 if (err)
1098 goto out_release;
1099
1100 out:
1101 *res = sock;
1102 return err;
1103 out_release:
1104 sock_release(sock);
1105 sock = NULL;
1106 goto out;
1107 }
1108 EXPORT_SYMBOL(sock_create_lite);
1109
1110 /* No kernel lock held - perfect */
1111 static __poll_t sock_poll(struct file *file, poll_table *wait)
1112 {
1113 struct socket *sock = file->private_data;
1114 __poll_t events = poll_requested_events(wait), flag = 0;
1115
1116 if (!sock->ops->poll)
1117 return 0;
1118
1119 if (sk_can_busy_loop(sock->sk)) {
1120 /* poll once if requested by the syscall */
1121 if (events & POLL_BUSY_LOOP)
1122 sk_busy_loop(sock->sk, 1);
1123
1124 /* if this socket can poll_ll, tell the system call */
1125 flag = POLL_BUSY_LOOP;
1126 }
1127
1128 return sock->ops->poll(file, sock, wait) | flag;
1129 }
1130
1131 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1132 {
1133 struct socket *sock = file->private_data;
1134
1135 return sock->ops->mmap(file, sock, vma);
1136 }
1137
1138 static int sock_close(struct inode *inode, struct file *filp)
1139 {
1140 __sock_release(SOCKET_I(inode), inode);
1141 return 0;
1142 }
1143
1144 /*
1145 * Update the socket async list
1146 *
1147 * Fasync_list locking strategy.
1148 *
1149 * 1. fasync_list is modified only under process context socket lock
1150 * i.e. under semaphore.
1151 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1152 * or under socket lock
1153 */
1154
1155 static int sock_fasync(int fd, struct file *filp, int on)
1156 {
1157 struct socket *sock = filp->private_data;
1158 struct sock *sk = sock->sk;
1159 struct socket_wq *wq;
1160
1161 if (sk == NULL)
1162 return -EINVAL;
1163
1164 lock_sock(sk);
1165 wq = sock->wq;
1166 fasync_helper(fd, filp, on, &wq->fasync_list);
1167
1168 if (!wq->fasync_list)
1169 sock_reset_flag(sk, SOCK_FASYNC);
1170 else
1171 sock_set_flag(sk, SOCK_FASYNC);
1172
1173 release_sock(sk);
1174 return 0;
1175 }
1176
1177 /* This function may be called only under rcu_lock */
1178
1179 int sock_wake_async(struct socket_wq *wq, int how, int band)
1180 {
1181 if (!wq || !wq->fasync_list)
1182 return -1;
1183
1184 switch (how) {
1185 case SOCK_WAKE_WAITD:
1186 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1187 break;
1188 goto call_kill;
1189 case SOCK_WAKE_SPACE:
1190 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1191 break;
1192 /* fall through */
1193 case SOCK_WAKE_IO:
1194 call_kill:
1195 kill_fasync(&wq->fasync_list, SIGIO, band);
1196 break;
1197 case SOCK_WAKE_URG:
1198 kill_fasync(&wq->fasync_list, SIGURG, band);
1199 }
1200
1201 return 0;
1202 }
1203 EXPORT_SYMBOL(sock_wake_async);
1204
1205 int __sock_create(struct net *net, int family, int type, int protocol,
1206 struct socket **res, int kern)
1207 {
1208 int err;
1209 struct socket *sock;
1210 const struct net_proto_family *pf;
1211
1212 /*
1213 * Check protocol is in range
1214 */
1215 if (family < 0 || family >= NPROTO)
1216 return -EAFNOSUPPORT;
1217 if (type < 0 || type >= SOCK_MAX)
1218 return -EINVAL;
1219
1220 /* Compatibility.
1221
1222 This uglymoron is moved from INET layer to here to avoid
1223 deadlock in module load.
1224 */
1225 if (family == PF_INET && type == SOCK_PACKET) {
1226 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1227 current->comm);
1228 family = PF_PACKET;
1229 }
1230
1231 err = security_socket_create(family, type, protocol, kern);
1232 if (err)
1233 return err;
1234
1235 /*
1236 * Allocate the socket and allow the family to set things up. if
1237 * the protocol is 0, the family is instructed to select an appropriate
1238 * default.
1239 */
1240 sock = sock_alloc();
1241 if (!sock) {
1242 net_warn_ratelimited("socket: no more sockets\n");
1243 return -ENFILE; /* Not exactly a match, but its the
1244 closest posix thing */
1245 }
1246
1247 sock->type = type;
1248
1249 #ifdef CONFIG_MODULES
1250 /* Attempt to load a protocol module if the find failed.
1251 *
1252 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1253 * requested real, full-featured networking support upon configuration.
1254 * Otherwise module support will break!
1255 */
1256 if (rcu_access_pointer(net_families[family]) == NULL)
1257 request_module("net-pf-%d", family);
1258 #endif
1259
1260 rcu_read_lock();
1261 pf = rcu_dereference(net_families[family]);
1262 err = -EAFNOSUPPORT;
1263 if (!pf)
1264 goto out_release;
1265
1266 /*
1267 * We will call the ->create function, that possibly is in a loadable
1268 * module, so we have to bump that loadable module refcnt first.
1269 */
1270 if (!try_module_get(pf->owner))
1271 goto out_release;
1272
1273 /* Now protected by module ref count */
1274 rcu_read_unlock();
1275
1276 err = pf->create(net, sock, protocol, kern);
1277 if (err < 0)
1278 goto out_module_put;
1279
1280 /*
1281 * Now to bump the refcnt of the [loadable] module that owns this
1282 * socket at sock_release time we decrement its refcnt.
1283 */
1284 if (!try_module_get(sock->ops->owner))
1285 goto out_module_busy;
1286
1287 /*
1288 * Now that we're done with the ->create function, the [loadable]
1289 * module can have its refcnt decremented
1290 */
1291 module_put(pf->owner);
1292 err = security_socket_post_create(sock, family, type, protocol, kern);
1293 if (err)
1294 goto out_sock_release;
1295 *res = sock;
1296
1297 return 0;
1298
1299 out_module_busy:
1300 err = -EAFNOSUPPORT;
1301 out_module_put:
1302 sock->ops = NULL;
1303 module_put(pf->owner);
1304 out_sock_release:
1305 sock_release(sock);
1306 return err;
1307
1308 out_release:
1309 rcu_read_unlock();
1310 goto out_sock_release;
1311 }
1312 EXPORT_SYMBOL(__sock_create);
1313
1314 int sock_create(int family, int type, int protocol, struct socket **res)
1315 {
1316 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1317 }
1318 EXPORT_SYMBOL(sock_create);
1319
1320 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1321 {
1322 return __sock_create(net, family, type, protocol, res, 1);
1323 }
1324 EXPORT_SYMBOL(sock_create_kern);
1325
1326 int __sys_socket(int family, int type, int protocol)
1327 {
1328 int retval;
1329 struct socket *sock;
1330 int flags;
1331
1332 /* Check the SOCK_* constants for consistency. */
1333 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1334 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1335 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1336 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1337
1338 flags = type & ~SOCK_TYPE_MASK;
1339 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1340 return -EINVAL;
1341 type &= SOCK_TYPE_MASK;
1342
1343 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1344 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1345
1346 retval = sock_create(family, type, protocol, &sock);
1347 if (retval < 0)
1348 return retval;
1349
1350 return sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1351 }
1352
1353 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1354 {
1355 return __sys_socket(family, type, protocol);
1356 }
1357
1358 /*
1359 * Create a pair of connected sockets.
1360 */
1361
1362 int __sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1363 {
1364 struct socket *sock1, *sock2;
1365 int fd1, fd2, err;
1366 struct file *newfile1, *newfile2;
1367 int flags;
1368
1369 flags = type & ~SOCK_TYPE_MASK;
1370 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1371 return -EINVAL;
1372 type &= SOCK_TYPE_MASK;
1373
1374 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1375 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1376
1377 /*
1378 * reserve descriptors and make sure we won't fail
1379 * to return them to userland.
1380 */
1381 fd1 = get_unused_fd_flags(flags);
1382 if (unlikely(fd1 < 0))
1383 return fd1;
1384
1385 fd2 = get_unused_fd_flags(flags);
1386 if (unlikely(fd2 < 0)) {
1387 put_unused_fd(fd1);
1388 return fd2;
1389 }
1390
1391 err = put_user(fd1, &usockvec[0]);
1392 if (err)
1393 goto out;
1394
1395 err = put_user(fd2, &usockvec[1]);
1396 if (err)
1397 goto out;
1398
1399 /*
1400 * Obtain the first socket and check if the underlying protocol
1401 * supports the socketpair call.
1402 */
1403
1404 err = sock_create(family, type, protocol, &sock1);
1405 if (unlikely(err < 0))
1406 goto out;
1407
1408 err = sock_create(family, type, protocol, &sock2);
1409 if (unlikely(err < 0)) {
1410 sock_release(sock1);
1411 goto out;
1412 }
1413
1414 err = security_socket_socketpair(sock1, sock2);
1415 if (unlikely(err)) {
1416 sock_release(sock2);
1417 sock_release(sock1);
1418 goto out;
1419 }
1420
1421 err = sock1->ops->socketpair(sock1, sock2);
1422 if (unlikely(err < 0)) {
1423 sock_release(sock2);
1424 sock_release(sock1);
1425 goto out;
1426 }
1427
1428 newfile1 = sock_alloc_file(sock1, flags, NULL);
1429 if (IS_ERR(newfile1)) {
1430 err = PTR_ERR(newfile1);
1431 sock_release(sock2);
1432 goto out;
1433 }
1434
1435 newfile2 = sock_alloc_file(sock2, flags, NULL);
1436 if (IS_ERR(newfile2)) {
1437 err = PTR_ERR(newfile2);
1438 fput(newfile1);
1439 goto out;
1440 }
1441
1442 audit_fd_pair(fd1, fd2);
1443
1444 fd_install(fd1, newfile1);
1445 fd_install(fd2, newfile2);
1446 return 0;
1447
1448 out:
1449 put_unused_fd(fd2);
1450 put_unused_fd(fd1);
1451 return err;
1452 }
1453
1454 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1455 int __user *, usockvec)
1456 {
1457 return __sys_socketpair(family, type, protocol, usockvec);
1458 }
1459
1460 /*
1461 * Bind a name to a socket. Nothing much to do here since it's
1462 * the protocol's responsibility to handle the local address.
1463 *
1464 * We move the socket address to kernel space before we call
1465 * the protocol layer (having also checked the address is ok).
1466 */
1467
1468 int __sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1469 {
1470 struct socket *sock;
1471 struct sockaddr_storage address;
1472 int err, fput_needed;
1473
1474 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1475 if (sock) {
1476 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1477 if (!err) {
1478 err = security_socket_bind(sock,
1479 (struct sockaddr *)&address,
1480 addrlen);
1481 if (!err)
1482 err = sock->ops->bind(sock,
1483 (struct sockaddr *)
1484 &address, addrlen);
1485 }
1486 fput_light(sock->file, fput_needed);
1487 }
1488 return err;
1489 }
1490
1491 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1492 {
1493 return __sys_bind(fd, umyaddr, addrlen);
1494 }
1495
1496 /*
1497 * Perform a listen. Basically, we allow the protocol to do anything
1498 * necessary for a listen, and if that works, we mark the socket as
1499 * ready for listening.
1500 */
1501
1502 int __sys_listen(int fd, int backlog)
1503 {
1504 struct socket *sock;
1505 int err, fput_needed;
1506 int somaxconn;
1507
1508 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1509 if (sock) {
1510 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1511 if ((unsigned int)backlog > somaxconn)
1512 backlog = somaxconn;
1513
1514 err = security_socket_listen(sock, backlog);
1515 if (!err)
1516 err = sock->ops->listen(sock, backlog);
1517
1518 fput_light(sock->file, fput_needed);
1519 }
1520 return err;
1521 }
1522
1523 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1524 {
1525 return __sys_listen(fd, backlog);
1526 }
1527
1528 /*
1529 * For accept, we attempt to create a new socket, set up the link
1530 * with the client, wake up the client, then return the new
1531 * connected fd. We collect the address of the connector in kernel
1532 * space and move it to user at the very end. This is unclean because
1533 * we open the socket then return an error.
1534 *
1535 * 1003.1g adds the ability to recvmsg() to query connection pending
1536 * status to recvmsg. We need to add that support in a way thats
1537 * clean when we restructure accept also.
1538 */
1539
1540 int __sys_accept4(int fd, struct sockaddr __user *upeer_sockaddr,
1541 int __user *upeer_addrlen, int flags)
1542 {
1543 struct socket *sock, *newsock;
1544 struct file *newfile;
1545 int err, len, newfd, fput_needed;
1546 struct sockaddr_storage address;
1547
1548 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1549 return -EINVAL;
1550
1551 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1552 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1553
1554 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1555 if (!sock)
1556 goto out;
1557
1558 err = -ENFILE;
1559 newsock = sock_alloc();
1560 if (!newsock)
1561 goto out_put;
1562
1563 newsock->type = sock->type;
1564 newsock->ops = sock->ops;
1565
1566 /*
1567 * We don't need try_module_get here, as the listening socket (sock)
1568 * has the protocol module (sock->ops->owner) held.
1569 */
1570 __module_get(newsock->ops->owner);
1571
1572 newfd = get_unused_fd_flags(flags);
1573 if (unlikely(newfd < 0)) {
1574 err = newfd;
1575 sock_release(newsock);
1576 goto out_put;
1577 }
1578 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1579 if (IS_ERR(newfile)) {
1580 err = PTR_ERR(newfile);
1581 put_unused_fd(newfd);
1582 goto out_put;
1583 }
1584
1585 err = security_socket_accept(sock, newsock);
1586 if (err)
1587 goto out_fd;
1588
1589 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1590 if (err < 0)
1591 goto out_fd;
1592
1593 if (ccs_socket_post_accept_permission(sock, newsock)) {
1594 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1595 goto out_fd;
1596 }
1597 if (upeer_sockaddr) {
1598 len = newsock->ops->getname(newsock,
1599 (struct sockaddr *)&address, 2);
1600 if (len < 0) {
1601 err = -ECONNABORTED;
1602 goto out_fd;
1603 }
1604 err = move_addr_to_user(&address,
1605 len, upeer_sockaddr, upeer_addrlen);
1606 if (err < 0)
1607 goto out_fd;
1608 }
1609
1610 /* File flags are not inherited via accept() unlike another OSes. */
1611
1612 fd_install(newfd, newfile);
1613 err = newfd;
1614
1615 out_put:
1616 fput_light(sock->file, fput_needed);
1617 out:
1618 return err;
1619 out_fd:
1620 fput(newfile);
1621 put_unused_fd(newfd);
1622 goto out_put;
1623 }
1624
1625 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1626 int __user *, upeer_addrlen, int, flags)
1627 {
1628 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, flags);
1629 }
1630
1631 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1632 int __user *, upeer_addrlen)
1633 {
1634 return __sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1635 }
1636
1637 /*
1638 * Attempt to connect to a socket with the server address. The address
1639 * is in user space so we verify it is OK and move it to kernel space.
1640 *
1641 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1642 * break bindings
1643 *
1644 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1645 * other SEQPACKET protocols that take time to connect() as it doesn't
1646 * include the -EINPROGRESS status for such sockets.
1647 */
1648
1649 int __sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1650 {
1651 struct socket *sock;
1652 struct sockaddr_storage address;
1653 int err, fput_needed;
1654
1655 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1656 if (!sock)
1657 goto out;
1658 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1659 if (err < 0)
1660 goto out_put;
1661
1662 err =
1663 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1664 if (err)
1665 goto out_put;
1666
1667 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1668 sock->file->f_flags);
1669 out_put:
1670 fput_light(sock->file, fput_needed);
1671 out:
1672 return err;
1673 }
1674
1675 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1676 int, addrlen)
1677 {
1678 return __sys_connect(fd, uservaddr, addrlen);
1679 }
1680
1681 /*
1682 * Get the local address ('name') of a socket object. Move the obtained
1683 * name to user space.
1684 */
1685
1686 int __sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1687 int __user *usockaddr_len)
1688 {
1689 struct socket *sock;
1690 struct sockaddr_storage address;
1691 int err, fput_needed;
1692
1693 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1694 if (!sock)
1695 goto out;
1696
1697 err = security_socket_getsockname(sock);
1698 if (err)
1699 goto out_put;
1700
1701 err = sock->ops->getname(sock, (struct sockaddr *)&address, 0);
1702 if (err < 0)
1703 goto out_put;
1704 /* "err" is actually length in this case */
1705 err = move_addr_to_user(&address, err, usockaddr, usockaddr_len);
1706
1707 out_put:
1708 fput_light(sock->file, fput_needed);
1709 out:
1710 return err;
1711 }
1712
1713 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1714 int __user *, usockaddr_len)
1715 {
1716 return __sys_getsockname(fd, usockaddr, usockaddr_len);
1717 }
1718
1719 /*
1720 * Get the remote address ('name') of a socket object. Move the obtained
1721 * name to user space.
1722 */
1723
1724 int __sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1725 int __user *usockaddr_len)
1726 {
1727 struct socket *sock;
1728 struct sockaddr_storage address;
1729 int err, fput_needed;
1730
1731 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1732 if (sock != NULL) {
1733 err = security_socket_getpeername(sock);
1734 if (err) {
1735 fput_light(sock->file, fput_needed);
1736 return err;
1737 }
1738
1739 err = sock->ops->getname(sock, (struct sockaddr *)&address, 1);
1740 if (err >= 0)
1741 /* "err" is actually length in this case */
1742 err = move_addr_to_user(&address, err, usockaddr,
1743 usockaddr_len);
1744 fput_light(sock->file, fput_needed);
1745 }
1746 return err;
1747 }
1748
1749 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1750 int __user *, usockaddr_len)
1751 {
1752 return __sys_getpeername(fd, usockaddr, usockaddr_len);
1753 }
1754
1755 /*
1756 * Send a datagram to a given address. We move the address into kernel
1757 * space and check the user space data area is readable before invoking
1758 * the protocol.
1759 */
1760 int __sys_sendto(int fd, void __user *buff, size_t len, unsigned int flags,
1761 struct sockaddr __user *addr, int addr_len)
1762 {
1763 struct socket *sock;
1764 struct sockaddr_storage address;
1765 int err;
1766 struct msghdr msg;
1767 struct iovec iov;
1768 int fput_needed;
1769
1770 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1771 if (unlikely(err))
1772 return err;
1773 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1774 if (!sock)
1775 goto out;
1776
1777 msg.msg_name = NULL;
1778 msg.msg_control = NULL;
1779 msg.msg_controllen = 0;
1780 msg.msg_namelen = 0;
1781 if (addr) {
1782 err = move_addr_to_kernel(addr, addr_len, &address);
1783 if (err < 0)
1784 goto out_put;
1785 msg.msg_name = (struct sockaddr *)&address;
1786 msg.msg_namelen = addr_len;
1787 }
1788 if (sock->file->f_flags & O_NONBLOCK)
1789 flags |= MSG_DONTWAIT;
1790 msg.msg_flags = flags;
1791 err = sock_sendmsg(sock, &msg);
1792
1793 out_put:
1794 fput_light(sock->file, fput_needed);
1795 out:
1796 return err;
1797 }
1798
1799 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1800 unsigned int, flags, struct sockaddr __user *, addr,
1801 int, addr_len)
1802 {
1803 return __sys_sendto(fd, buff, len, flags, addr, addr_len);
1804 }
1805
1806 /*
1807 * Send a datagram down a socket.
1808 */
1809
1810 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1811 unsigned int, flags)
1812 {
1813 return __sys_sendto(fd, buff, len, flags, NULL, 0);
1814 }
1815
1816 /*
1817 * Receive a frame from the socket and optionally record the address of the
1818 * sender. We verify the buffers are writable and if needed move the
1819 * sender address from kernel to user space.
1820 */
1821 int __sys_recvfrom(int fd, void __user *ubuf, size_t size, unsigned int flags,
1822 struct sockaddr __user *addr, int __user *addr_len)
1823 {
1824 struct socket *sock;
1825 struct iovec iov;
1826 struct msghdr msg;
1827 struct sockaddr_storage address;
1828 int err, err2;
1829 int fput_needed;
1830
1831 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1832 if (unlikely(err))
1833 return err;
1834 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1835 if (!sock)
1836 goto out;
1837
1838 msg.msg_control = NULL;
1839 msg.msg_controllen = 0;
1840 /* Save some cycles and don't copy the address if not needed */
1841 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1842 /* We assume all kernel code knows the size of sockaddr_storage */
1843 msg.msg_namelen = 0;
1844 msg.msg_iocb = NULL;
1845 msg.msg_flags = 0;
1846 if (sock->file->f_flags & O_NONBLOCK)
1847 flags |= MSG_DONTWAIT;
1848 err = sock_recvmsg(sock, &msg, flags);
1849
1850 if (err >= 0 && addr != NULL) {
1851 err2 = move_addr_to_user(&address,
1852 msg.msg_namelen, addr, addr_len);
1853 if (err2 < 0)
1854 err = err2;
1855 }
1856
1857 fput_light(sock->file, fput_needed);
1858 out:
1859 return err;
1860 }
1861
1862 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1863 unsigned int, flags, struct sockaddr __user *, addr,
1864 int __user *, addr_len)
1865 {
1866 return __sys_recvfrom(fd, ubuf, size, flags, addr, addr_len);
1867 }
1868
1869 /*
1870 * Receive a datagram from a socket.
1871 */
1872
1873 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1874 unsigned int, flags)
1875 {
1876 return __sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1877 }
1878
1879 /*
1880 * Set a socket option. Because we don't know the option lengths we have
1881 * to pass the user mode parameter for the protocols to sort out.
1882 */
1883
1884 static int __sys_setsockopt(int fd, int level, int optname,
1885 char __user *optval, int optlen)
1886 {
1887 int err, fput_needed;
1888 struct socket *sock;
1889
1890 if (optlen < 0)
1891 return -EINVAL;
1892
1893 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1894 if (sock != NULL) {
1895 err = security_socket_setsockopt(sock, level, optname);
1896 if (err)
1897 goto out_put;
1898
1899 if (level == SOL_SOCKET)
1900 err =
1901 sock_setsockopt(sock, level, optname, optval,
1902 optlen);
1903 else
1904 err =
1905 sock->ops->setsockopt(sock, level, optname, optval,
1906 optlen);
1907 out_put:
1908 fput_light(sock->file, fput_needed);
1909 }
1910 return err;
1911 }
1912
1913 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1914 char __user *, optval, int, optlen)
1915 {
1916 return __sys_setsockopt(fd, level, optname, optval, optlen);
1917 }
1918
1919 /*
1920 * Get a socket option. Because we don't know the option lengths we have
1921 * to pass a user mode parameter for the protocols to sort out.
1922 */
1923
1924 static int __sys_getsockopt(int fd, int level, int optname,
1925 char __user *optval, int __user *optlen)
1926 {
1927 int err, fput_needed;
1928 struct socket *sock;
1929
1930 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1931 if (sock != NULL) {
1932 err = security_socket_getsockopt(sock, level, optname);
1933 if (err)
1934 goto out_put;
1935
1936 if (level == SOL_SOCKET)
1937 err =
1938 sock_getsockopt(sock, level, optname, optval,
1939 optlen);
1940 else
1941 err =
1942 sock->ops->getsockopt(sock, level, optname, optval,
1943 optlen);
1944 out_put:
1945 fput_light(sock->file, fput_needed);
1946 }
1947 return err;
1948 }
1949
1950 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1951 char __user *, optval, int __user *, optlen)
1952 {
1953 return __sys_getsockopt(fd, level, optname, optval, optlen);
1954 }
1955
1956 /*
1957 * Shutdown a socket.
1958 */
1959
1960 int __sys_shutdown(int fd, int how)
1961 {
1962 int err, fput_needed;
1963 struct socket *sock;
1964
1965 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1966 if (sock != NULL) {
1967 err = security_socket_shutdown(sock, how);
1968 if (!err)
1969 err = sock->ops->shutdown(sock, how);
1970 fput_light(sock->file, fput_needed);
1971 }
1972 return err;
1973 }
1974
1975 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1976 {
1977 return __sys_shutdown(fd, how);
1978 }
1979
1980 /* A couple of helpful macros for getting the address of the 32/64 bit
1981 * fields which are the same type (int / unsigned) on our platforms.
1982 */
1983 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1984 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1985 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1986
1987 struct used_address {
1988 struct sockaddr_storage name;
1989 unsigned int name_len;
1990 };
1991
1992 static int copy_msghdr_from_user(struct msghdr *kmsg,
1993 struct user_msghdr __user *umsg,
1994 struct sockaddr __user **save_addr,
1995 struct iovec **iov)
1996 {
1997 struct user_msghdr msg;
1998 ssize_t err;
1999
2000 if (copy_from_user(&msg, umsg, sizeof(*umsg)))
2001 return -EFAULT;
2002
2003 kmsg->msg_control = (void __force *)msg.msg_control;
2004 kmsg->msg_controllen = msg.msg_controllen;
2005 kmsg->msg_flags = msg.msg_flags;
2006
2007 kmsg->msg_namelen = msg.msg_namelen;
2008 if (!msg.msg_name)
2009 kmsg->msg_namelen = 0;
2010
2011 if (kmsg->msg_namelen < 0)
2012 return -EINVAL;
2013
2014 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
2015 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
2016
2017 if (save_addr)
2018 *save_addr = msg.msg_name;
2019
2020 if (msg.msg_name && kmsg->msg_namelen) {
2021 if (!save_addr) {
2022 err = move_addr_to_kernel(msg.msg_name,
2023 kmsg->msg_namelen,
2024 kmsg->msg_name);
2025 if (err < 0)
2026 return err;
2027 }
2028 } else {
2029 kmsg->msg_name = NULL;
2030 kmsg->msg_namelen = 0;
2031 }
2032
2033 if (msg.msg_iovlen > UIO_MAXIOV)
2034 return -EMSGSIZE;
2035
2036 kmsg->msg_iocb = NULL;
2037
2038 return import_iovec(save_addr ? READ : WRITE,
2039 msg.msg_iov, msg.msg_iovlen,
2040 UIO_FASTIOV, iov, &kmsg->msg_iter);
2041 }
2042
2043 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
2044 struct msghdr *msg_sys, unsigned int flags,
2045 struct used_address *used_address,
2046 unsigned int allowed_msghdr_flags)
2047 {
2048 struct compat_msghdr __user *msg_compat =
2049 (struct compat_msghdr __user *)msg;
2050 struct sockaddr_storage address;
2051 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
2052 unsigned char ctl[sizeof(struct cmsghdr) + 20]
2053 __aligned(sizeof(__kernel_size_t));
2054 /* 20 is size of ipv6_pktinfo */
2055 unsigned char *ctl_buf = ctl;
2056 int ctl_len;
2057 ssize_t err;
2058
2059 msg_sys->msg_name = &address;
2060
2061 if (MSG_CMSG_COMPAT & flags)
2062 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
2063 else
2064 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
2065 if (err < 0)
2066 return err;
2067
2068 err = -ENOBUFS;
2069
2070 if (msg_sys->msg_controllen > INT_MAX)
2071 goto out_freeiov;
2072 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
2073 ctl_len = msg_sys->msg_controllen;
2074 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
2075 err =
2076 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
2077 sizeof(ctl));
2078 if (err)
2079 goto out_freeiov;
2080 ctl_buf = msg_sys->msg_control;
2081 ctl_len = msg_sys->msg_controllen;
2082 } else if (ctl_len) {
2083 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2084 CMSG_ALIGN(sizeof(struct cmsghdr)));
2085 if (ctl_len > sizeof(ctl)) {
2086 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2087 if (ctl_buf == NULL)
2088 goto out_freeiov;
2089 }
2090 err = -EFAULT;
2091 /*
2092 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2093 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2094 * checking falls down on this.
2095 */
2096 if (copy_from_user(ctl_buf,
2097 (void __user __force *)msg_sys->msg_control,
2098 ctl_len))
2099 goto out_freectl;
2100 msg_sys->msg_control = ctl_buf;
2101 }
2102 msg_sys->msg_flags = flags;
2103
2104 if (sock->file->f_flags & O_NONBLOCK)
2105 msg_sys->msg_flags |= MSG_DONTWAIT;
2106 /*
2107 * If this is sendmmsg() and current destination address is same as
2108 * previously succeeded address, omit asking LSM's decision.
2109 * used_address->name_len is initialized to UINT_MAX so that the first
2110 * destination address never matches.
2111 */
2112 if (used_address && msg_sys->msg_name &&
2113 used_address->name_len == msg_sys->msg_namelen &&
2114 !memcmp(&used_address->name, msg_sys->msg_name,
2115 used_address->name_len)) {
2116 err = sock_sendmsg_nosec(sock, msg_sys);
2117 goto out_freectl;
2118 }
2119 err = sock_sendmsg(sock, msg_sys);
2120 /*
2121 * If this is sendmmsg() and sending to current destination address was
2122 * successful, remember it.
2123 */
2124 if (used_address && err >= 0) {
2125 used_address->name_len = msg_sys->msg_namelen;
2126 if (msg_sys->msg_name)
2127 memcpy(&used_address->name, msg_sys->msg_name,
2128 used_address->name_len);
2129 }
2130
2131 out_freectl:
2132 if (ctl_buf != ctl)
2133 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2134 out_freeiov:
2135 kfree(iov);
2136 return err;
2137 }
2138
2139 /*
2140 * BSD sendmsg interface
2141 */
2142
2143 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2144 bool forbid_cmsg_compat)
2145 {
2146 int fput_needed, err;
2147 struct msghdr msg_sys;
2148 struct socket *sock;
2149
2150 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2151 return -EINVAL;
2152
2153 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2154 if (!sock)
2155 goto out;
2156
2157 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2158
2159 fput_light(sock->file, fput_needed);
2160 out:
2161 return err;
2162 }
2163
2164 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2165 {
2166 return __sys_sendmsg(fd, msg, flags, true);
2167 }
2168
2169 /*
2170 * Linux sendmmsg interface
2171 */
2172
2173 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2174 unsigned int flags, bool forbid_cmsg_compat)
2175 {
2176 int fput_needed, err, datagrams;
2177 struct socket *sock;
2178 struct mmsghdr __user *entry;
2179 struct compat_mmsghdr __user *compat_entry;
2180 struct msghdr msg_sys;
2181 struct used_address used_address;
2182 unsigned int oflags = flags;
2183
2184 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2185 return -EINVAL;
2186
2187 if (vlen > UIO_MAXIOV)
2188 vlen = UIO_MAXIOV;
2189
2190 datagrams = 0;
2191
2192 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2193 if (!sock)
2194 return err;
2195
2196 used_address.name_len = UINT_MAX;
2197 entry = mmsg;
2198 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2199 err = 0;
2200 flags |= MSG_BATCH;
2201
2202 while (datagrams < vlen) {
2203 if (datagrams == vlen - 1)
2204 flags = oflags;
2205
2206 if (MSG_CMSG_COMPAT & flags) {
2207 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2208 &msg_sys, flags, &used_address, MSG_EOR);
2209 if (err < 0)
2210 break;
2211 err = __put_user(err, &compat_entry->msg_len);
2212 ++compat_entry;
2213 } else {
2214 err = ___sys_sendmsg(sock,
2215 (struct user_msghdr __user *)entry,
2216 &msg_sys, flags, &used_address, MSG_EOR);
2217 if (err < 0)
2218 break;
2219 err = put_user(err, &entry->msg_len);
2220 ++entry;
2221 }
2222
2223 if (err)
2224 break;
2225 ++datagrams;
2226 if (msg_data_left(&msg_sys))
2227 break;
2228 cond_resched();
2229 }
2230
2231 fput_light(sock->file, fput_needed);
2232
2233 /* We only return an error if no datagrams were able to be sent */
2234 if (datagrams != 0)
2235 return datagrams;
2236
2237 return err;
2238 }
2239
2240 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2241 unsigned int, vlen, unsigned int, flags)
2242 {
2243 return __sys_sendmmsg(fd, mmsg, vlen, flags, true);
2244 }
2245
2246 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2247 struct msghdr *msg_sys, unsigned int flags, int nosec)
2248 {
2249 struct compat_msghdr __user *msg_compat =
2250 (struct compat_msghdr __user *)msg;
2251 struct iovec iovstack[UIO_FASTIOV];
2252 struct iovec *iov = iovstack;
2253 unsigned long cmsg_ptr;
2254 int len;
2255 ssize_t err;
2256
2257 /* kernel mode address */
2258 struct sockaddr_storage addr;
2259
2260 /* user mode address pointers */
2261 struct sockaddr __user *uaddr;
2262 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2263
2264 msg_sys->msg_name = &addr;
2265
2266 if (MSG_CMSG_COMPAT & flags)
2267 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2268 else
2269 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2270 if (err < 0)
2271 return err;
2272
2273 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2274 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2275
2276 /* We assume all kernel code knows the size of sockaddr_storage */
2277 msg_sys->msg_namelen = 0;
2278
2279 if (sock->file->f_flags & O_NONBLOCK)
2280 flags |= MSG_DONTWAIT;
2281 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2282 if (err < 0)
2283 goto out_freeiov;
2284 len = err;
2285
2286 if (uaddr != NULL) {
2287 err = move_addr_to_user(&addr,
2288 msg_sys->msg_namelen, uaddr,
2289 uaddr_len);
2290 if (err < 0)
2291 goto out_freeiov;
2292 }
2293 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2294 COMPAT_FLAGS(msg));
2295 if (err)
2296 goto out_freeiov;
2297 if (MSG_CMSG_COMPAT & flags)
2298 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2299 &msg_compat->msg_controllen);
2300 else
2301 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2302 &msg->msg_controllen);
2303 if (err)
2304 goto out_freeiov;
2305 err = len;
2306
2307 out_freeiov:
2308 kfree(iov);
2309 return err;
2310 }
2311
2312 /*
2313 * BSD recvmsg interface
2314 */
2315
2316 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned int flags,
2317 bool forbid_cmsg_compat)
2318 {
2319 int fput_needed, err;
2320 struct msghdr msg_sys;
2321 struct socket *sock;
2322
2323 if (forbid_cmsg_compat && (flags & MSG_CMSG_COMPAT))
2324 return -EINVAL;
2325
2326 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2327 if (!sock)
2328 goto out;
2329
2330 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2331
2332 fput_light(sock->file, fput_needed);
2333 out:
2334 return err;
2335 }
2336
2337 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2338 unsigned int, flags)
2339 {
2340 return __sys_recvmsg(fd, msg, flags, true);
2341 }
2342
2343 /*
2344 * Linux recvmmsg interface
2345 */
2346
2347 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2348 unsigned int flags, struct timespec64 *timeout)
2349 {
2350 int fput_needed, err, datagrams;
2351 struct socket *sock;
2352 struct mmsghdr __user *entry;
2353 struct compat_mmsghdr __user *compat_entry;
2354 struct msghdr msg_sys;
2355 struct timespec64 end_time;
2356 struct timespec64 timeout64;
2357
2358 if (timeout &&
2359 poll_select_set_timeout(&end_time, timeout->tv_sec,
2360 timeout->tv_nsec))
2361 return -EINVAL;
2362
2363 datagrams = 0;
2364
2365 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2366 if (!sock)
2367 return err;
2368
2369 if (likely(!(flags & MSG_ERRQUEUE))) {
2370 err = sock_error(sock->sk);
2371 if (err) {
2372 datagrams = err;
2373 goto out_put;
2374 }
2375 }
2376
2377 entry = mmsg;
2378 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2379
2380 while (datagrams < vlen) {
2381 /*
2382 * No need to ask LSM for more than the first datagram.
2383 */
2384 if (MSG_CMSG_COMPAT & flags) {
2385 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2386 &msg_sys, flags & ~MSG_WAITFORONE,
2387 datagrams);
2388 if (err < 0)
2389 break;
2390 err = __put_user(err, &compat_entry->msg_len);
2391 ++compat_entry;
2392 } else {
2393 err = ___sys_recvmsg(sock,
2394 (struct user_msghdr __user *)entry,
2395 &msg_sys, flags & ~MSG_WAITFORONE,
2396 datagrams);
2397 if (err < 0)
2398 break;
2399 err = put_user(err, &entry->msg_len);
2400 ++entry;
2401 }
2402
2403 if (err)
2404 break;
2405 ++datagrams;
2406
2407 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2408 if (flags & MSG_WAITFORONE)
2409 flags |= MSG_DONTWAIT;
2410
2411 if (timeout) {
2412 ktime_get_ts64(&timeout64);
2413 *timeout = timespec64_sub(end_time, timeout64);
2414 if (timeout->tv_sec < 0) {
2415 timeout->tv_sec = timeout->tv_nsec = 0;
2416 break;
2417 }
2418
2419 /* Timeout, return less than vlen datagrams */
2420 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2421 break;
2422 }
2423
2424 /* Out of band data, return right away */
2425 if (msg_sys.msg_flags & MSG_OOB)
2426 break;
2427 cond_resched();
2428 }
2429
2430 if (err == 0)
2431 goto out_put;
2432
2433 if (datagrams == 0) {
2434 datagrams = err;
2435 goto out_put;
2436 }
2437
2438 /*
2439 * We may return less entries than requested (vlen) if the
2440 * sock is non block and there aren't enough datagrams...
2441 */
2442 if (err != -EAGAIN) {
2443 /*
2444 * ... or if recvmsg returns an error after we
2445 * received some datagrams, where we record the
2446 * error to return on the next call or if the
2447 * app asks about it using getsockopt(SO_ERROR).
2448 */
2449 sock->sk->sk_err = -err;
2450 }
2451 out_put:
2452 fput_light(sock->file, fput_needed);
2453
2454 return datagrams;
2455 }
2456
2457 static int do_sys_recvmmsg(int fd, struct mmsghdr __user *mmsg,
2458 unsigned int vlen, unsigned int flags,
2459 struct __kernel_timespec __user *timeout)
2460 {
2461 int datagrams;
2462 struct timespec64 timeout_sys;
2463
2464 if (flags & MSG_CMSG_COMPAT)
2465 return -EINVAL;
2466
2467 if (!timeout)
2468 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2469
2470 if (get_timespec64(&timeout_sys, timeout))
2471 return -EFAULT;
2472
2473 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2474
2475 if (datagrams > 0 && put_timespec64(&timeout_sys, timeout))
2476 datagrams = -EFAULT;
2477
2478 return datagrams;
2479 }
2480
2481 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2482 unsigned int, vlen, unsigned int, flags,
2483 struct __kernel_timespec __user *, timeout)
2484 {
2485 return do_sys_recvmmsg(fd, mmsg, vlen, flags, timeout);
2486 }
2487
2488 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2489 /* Argument list sizes for sys_socketcall */
2490 #define AL(x) ((x) * sizeof(unsigned long))
2491 static const unsigned char nargs[21] = {
2492 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2493 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2494 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2495 AL(4), AL(5), AL(4)
2496 };
2497
2498 #undef AL
2499
2500 /*
2501 * System call vectors.
2502 *
2503 * Argument checking cleaned up. Saved 20% in size.
2504 * This function doesn't need to set the kernel lock because
2505 * it is set by the callees.
2506 */
2507
2508 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2509 {
2510 unsigned long a[AUDITSC_ARGS];
2511 unsigned long a0, a1;
2512 int err;
2513 unsigned int len;
2514
2515 if (call < 1 || call > SYS_SENDMMSG)
2516 return -EINVAL;
2517 call = array_index_nospec(call, SYS_SENDMMSG + 1);
2518
2519 len = nargs[call];
2520 if (len > sizeof(a))
2521 return -EINVAL;
2522
2523 /* copy_from_user should be SMP safe. */
2524 if (copy_from_user(a, args, len))
2525 return -EFAULT;
2526
2527 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2528 if (err)
2529 return err;
2530
2531 a0 = a[0];
2532 a1 = a[1];
2533
2534 switch (call) {
2535 case SYS_SOCKET:
2536 err = __sys_socket(a0, a1, a[2]);
2537 break;
2538 case SYS_BIND:
2539 err = __sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2540 break;
2541 case SYS_CONNECT:
2542 err = __sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2543 break;
2544 case SYS_LISTEN:
2545 err = __sys_listen(a0, a1);
2546 break;
2547 case SYS_ACCEPT:
2548 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2549 (int __user *)a[2], 0);
2550 break;
2551 case SYS_GETSOCKNAME:
2552 err =
2553 __sys_getsockname(a0, (struct sockaddr __user *)a1,
2554 (int __user *)a[2]);
2555 break;
2556 case SYS_GETPEERNAME:
2557 err =
2558 __sys_getpeername(a0, (struct sockaddr __user *)a1,
2559 (int __user *)a[2]);
2560 break;
2561 case SYS_SOCKETPAIR:
2562 err = __sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2563 break;
2564 case SYS_SEND:
2565 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2566 NULL, 0);
2567 break;
2568 case SYS_SENDTO:
2569 err = __sys_sendto(a0, (void __user *)a1, a[2], a[3],
2570 (struct sockaddr __user *)a[4], a[5]);
2571 break;
2572 case SYS_RECV:
2573 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2574 NULL, NULL);
2575 break;
2576 case SYS_RECVFROM:
2577 err = __sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2578 (struct sockaddr __user *)a[4],
2579 (int __user *)a[5]);
2580 break;
2581 case SYS_SHUTDOWN:
2582 err = __sys_shutdown(a0, a1);
2583 break;
2584 case SYS_SETSOCKOPT:
2585 err = __sys_setsockopt(a0, a1, a[2], (char __user *)a[3],
2586 a[4]);
2587 break;
2588 case SYS_GETSOCKOPT:
2589 err =
2590 __sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2591 (int __user *)a[4]);
2592 break;
2593 case SYS_SENDMSG:
2594 err = __sys_sendmsg(a0, (struct user_msghdr __user *)a1,
2595 a[2], true);
2596 break;
2597 case SYS_SENDMMSG:
2598 err = __sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2599 a[3], true);
2600 break;
2601 case SYS_RECVMSG:
2602 err = __sys_recvmsg(a0, (struct user_msghdr __user *)a1,
2603 a[2], true);
2604 break;
2605 case SYS_RECVMMSG:
2606 err = do_sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2],
2607 a[3], (struct __kernel_timespec __user *)a[4]);
2608 break;
2609 case SYS_ACCEPT4:
2610 err = __sys_accept4(a0, (struct sockaddr __user *)a1,
2611 (int __user *)a[2], a[3]);
2612 break;
2613 default:
2614 err = -EINVAL;
2615 break;
2616 }
2617 return err;
2618 }
2619
2620 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2621
2622 /**
2623 * sock_register - add a socket protocol handler
2624 * @ops: description of protocol
2625 *
2626 * This function is called by a protocol handler that wants to
2627 * advertise its address family, and have it linked into the
2628 * socket interface. The value ops->family corresponds to the
2629 * socket system call protocol family.
2630 */
2631 int sock_register(const struct net_proto_family *ops)
2632 {
2633 int err;
2634
2635 if (ops->family >= NPROTO) {
2636 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2637 return -ENOBUFS;
2638 }
2639
2640 spin_lock(&net_family_lock);
2641 if (rcu_dereference_protected(net_families[ops->family],
2642 lockdep_is_held(&net_family_lock)))
2643 err = -EEXIST;
2644 else {
2645 rcu_assign_pointer(net_families[ops->family], ops);
2646 err = 0;
2647 }
2648 spin_unlock(&net_family_lock);
2649
2650 pr_info("NET: Registered protocol family %d\n", ops->family);
2651 return err;
2652 }
2653 EXPORT_SYMBOL(sock_register);
2654
2655 /**
2656 * sock_unregister - remove a protocol handler
2657 * @family: protocol family to remove
2658 *
2659 * This function is called by a protocol handler that wants to
2660 * remove its address family, and have it unlinked from the
2661 * new socket creation.
2662 *
2663 * If protocol handler is a module, then it can use module reference
2664 * counts to protect against new references. If protocol handler is not
2665 * a module then it needs to provide its own protection in
2666 * the ops->create routine.
2667 */
2668 void sock_unregister(int family)
2669 {
2670 BUG_ON(family < 0 || family >= NPROTO);
2671
2672 spin_lock(&net_family_lock);
2673 RCU_INIT_POINTER(net_families[family], NULL);
2674 spin_unlock(&net_family_lock);
2675
2676 synchronize_rcu();
2677
2678 pr_info("NET: Unregistered protocol family %d\n", family);
2679 }
2680 EXPORT_SYMBOL(sock_unregister);
2681
2682 bool sock_is_registered(int family)
2683 {
2684 return family < NPROTO && rcu_access_pointer(net_families[family]);
2685 }
2686
2687 static int __init sock_init(void)
2688 {
2689 int err;
2690 /*
2691 * Initialize the network sysctl infrastructure.
2692 */
2693 err = net_sysctl_init();
2694 if (err)
2695 goto out;
2696
2697 /*
2698 * Initialize skbuff SLAB cache
2699 */
2700 skb_init();
2701
2702 /*
2703 * Initialize the protocols module.
2704 */
2705
2706 init_inodecache();
2707
2708 err = register_filesystem(&sock_fs_type);
2709 if (err)
2710 goto out_fs;
2711 sock_mnt = kern_mount(&sock_fs_type);
2712 if (IS_ERR(sock_mnt)) {
2713 err = PTR_ERR(sock_mnt);
2714 goto out_mount;
2715 }
2716
2717 /* The real protocol initialization is performed in later initcalls.
2718 */
2719
2720 #ifdef CONFIG_NETFILTER
2721 err = netfilter_init();
2722 if (err)
2723 goto out;
2724 #endif
2725
2726 ptp_classifier_init();
2727
2728 out:
2729 return err;
2730
2731 out_mount:
2732 unregister_filesystem(&sock_fs_type);
2733 out_fs:
2734 goto out;
2735 }
2736
2737 core_initcall(sock_init); /* early initcall */
2738
2739 #ifdef CONFIG_PROC_FS
2740 void socket_seq_show(struct seq_file *seq)
2741 {
2742 seq_printf(seq, "sockets: used %d\n",
2743 sock_inuse_get(seq->private));
2744 }
2745 #endif /* CONFIG_PROC_FS */
2746
2747 #ifdef CONFIG_COMPAT
2748 static int do_siocgstamp(struct net *net, struct socket *sock,
2749 unsigned int cmd, void __user *up)
2750 {
2751 mm_segment_t old_fs = get_fs();
2752 struct timeval ktv;
2753 int err;
2754
2755 set_fs(KERNEL_DS);
2756 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2757 set_fs(old_fs);
2758 if (!err)
2759 err = compat_put_timeval(&ktv, up);
2760
2761 return err;
2762 }
2763
2764 static int do_siocgstampns(struct net *net, struct socket *sock,
2765 unsigned int cmd, void __user *up)
2766 {
2767 mm_segment_t old_fs = get_fs();
2768 struct timespec kts;
2769 int err;
2770
2771 set_fs(KERNEL_DS);
2772 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2773 set_fs(old_fs);
2774 if (!err)
2775 err = compat_put_timespec(&kts, up);
2776
2777 return err;
2778 }
2779
2780 static int compat_dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2781 {
2782 struct compat_ifconf ifc32;
2783 struct ifconf ifc;
2784 int err;
2785
2786 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2787 return -EFAULT;
2788
2789 ifc.ifc_len = ifc32.ifc_len;
2790 ifc.ifc_req = compat_ptr(ifc32.ifcbuf);
2791
2792 rtnl_lock();
2793 err = dev_ifconf(net, &ifc, sizeof(struct compat_ifreq));
2794 rtnl_unlock();
2795 if (err)
2796 return err;
2797
2798 ifc32.ifc_len = ifc.ifc_len;
2799 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2800 return -EFAULT;
2801
2802 return 0;
2803 }
2804
2805 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2806 {
2807 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2808 bool convert_in = false, convert_out = false;
2809 size_t buf_size = 0;
2810 struct ethtool_rxnfc __user *rxnfc = NULL;
2811 struct ifreq ifr;
2812 u32 rule_cnt = 0, actual_rule_cnt;
2813 u32 ethcmd;
2814 u32 data;
2815 int ret;
2816
2817 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2818 return -EFAULT;
2819
2820 compat_rxnfc = compat_ptr(data);
2821
2822 if (get_user(ethcmd, &compat_rxnfc->cmd))
2823 return -EFAULT;
2824
2825 /* Most ethtool structures are defined without padding.
2826 * Unfortunately struct ethtool_rxnfc is an exception.
2827 */
2828 switch (ethcmd) {
2829 default:
2830 break;
2831 case ETHTOOL_GRXCLSRLALL:
2832 /* Buffer size is variable */
2833 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2834 return -EFAULT;
2835 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2836 return -ENOMEM;
2837 buf_size += rule_cnt * sizeof(u32);
2838 /* fall through */
2839 case ETHTOOL_GRXRINGS:
2840 case ETHTOOL_GRXCLSRLCNT:
2841 case ETHTOOL_GRXCLSRULE:
2842 case ETHTOOL_SRXCLSRLINS:
2843 convert_out = true;
2844 /* fall through */
2845 case ETHTOOL_SRXCLSRLDEL:
2846 buf_size += sizeof(struct ethtool_rxnfc);
2847 convert_in = true;
2848 rxnfc = compat_alloc_user_space(buf_size);
2849 break;
2850 }
2851
2852 if (copy_from_user(&ifr.ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2853 return -EFAULT;
2854
2855 ifr.ifr_data = convert_in ? rxnfc : (void __user *)compat_rxnfc;
2856
2857 if (convert_in) {
2858 /* We expect there to be holes between fs.m_ext and
2859 * fs.ring_cookie and at the end of fs, but nowhere else.
2860 */
2861 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2862 sizeof(compat_rxnfc->fs.m_ext) !=
2863 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2864 sizeof(rxnfc->fs.m_ext));
2865 BUILD_BUG_ON(
2866 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2867 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2868 offsetof(struct ethtool_rxnfc, fs.location) -
2869 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2870
2871 if (copy_in_user(rxnfc, compat_rxnfc,
2872 (void __user *)(&rxnfc->fs.m_ext + 1) -
2873 (void __user *)rxnfc) ||
2874 copy_in_user(&rxnfc->fs.ring_cookie,
2875 &compat_rxnfc->fs.ring_cookie,
2876 (void __user *)(&rxnfc->fs.location + 1) -
2877 (void __user *)&rxnfc->fs.ring_cookie))
2878 return -EFAULT;
2879 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2880 if (put_user(rule_cnt, &rxnfc->rule_cnt))
2881 return -EFAULT;
2882 } else if (copy_in_user(&rxnfc->rule_cnt,
2883 &compat_rxnfc->rule_cnt,
2884 sizeof(rxnfc->rule_cnt)))
2885 return -EFAULT;
2886 }
2887
2888 ret = dev_ioctl(net, SIOCETHTOOL, &ifr, NULL);
2889 if (ret)
2890 return ret;
2891
2892 if (convert_out) {
2893 if (copy_in_user(compat_rxnfc, rxnfc,
2894 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2895 (const void __user *)rxnfc) ||
2896 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2897 &rxnfc->fs.ring_cookie,
2898 (const void __user *)(&rxnfc->fs.location + 1) -
2899 (const void __user *)&rxnfc->fs.ring_cookie) ||
2900 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2901 sizeof(rxnfc->rule_cnt)))
2902 return -EFAULT;
2903
2904 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2905 /* As an optimisation, we only copy the actual
2906 * number of rules that the underlying
2907 * function returned. Since Mallory might
2908 * change the rule count in user memory, we
2909 * check that it is less than the rule count
2910 * originally given (as the user buffer size),
2911 * which has been range-checked.
2912 */
2913 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2914 return -EFAULT;
2915 if (actual_rule_cnt < rule_cnt)
2916 rule_cnt = actual_rule_cnt;
2917 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2918 &rxnfc->rule_locs[0],
2919 rule_cnt * sizeof(u32)))
2920 return -EFAULT;
2921 }
2922 }
2923
2924 return 0;
2925 }
2926
2927 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2928 {
2929 compat_uptr_t uptr32;
2930 struct ifreq ifr;
2931 void __user *saved;
2932 int err;
2933
2934 if (copy_from_user(&ifr, uifr32, sizeof(struct compat_ifreq)))
2935 return -EFAULT;
2936
2937 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2938 return -EFAULT;
2939
2940 saved = ifr.ifr_settings.ifs_ifsu.raw_hdlc;
2941 ifr.ifr_settings.ifs_ifsu.raw_hdlc = compat_ptr(uptr32);
2942
2943 err = dev_ioctl(net, SIOCWANDEV, &ifr, NULL);
2944 if (!err) {
2945 ifr.ifr_settings.ifs_ifsu.raw_hdlc = saved;
2946 if (copy_to_user(uifr32, &ifr, sizeof(struct compat_ifreq)))
2947 err = -EFAULT;
2948 }
2949 return err;
2950 }
2951
2952 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2953 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2954 struct compat_ifreq __user *u_ifreq32)
2955 {
2956 struct ifreq ifreq;
2957 u32 data32;
2958
2959 if (copy_from_user(ifreq.ifr_name, u_ifreq32->ifr_name, IFNAMSIZ))
2960 return -EFAULT;
2961 if (get_user(data32, &u_ifreq32->ifr_data))
2962 return -EFAULT;
2963 ifreq.ifr_data = compat_ptr(data32);
2964
2965 return dev_ioctl(net, cmd, &ifreq, NULL);
2966 }
2967
2968 static int compat_ifreq_ioctl(struct net *net, struct socket *sock,
2969 unsigned int cmd,
2970 struct compat_ifreq __user *uifr32)
2971 {
2972 struct ifreq __user *uifr;
2973 int err;
2974
2975 /* Handle the fact that while struct ifreq has the same *layout* on
2976 * 32/64 for everything but ifreq::ifru_ifmap and ifreq::ifru_data,
2977 * which are handled elsewhere, it still has different *size* due to
2978 * ifreq::ifru_ifmap (which is 16 bytes on 32 bit, 24 bytes on 64-bit,
2979 * resulting in struct ifreq being 32 and 40 bytes respectively).
2980 * As a result, if the struct happens to be at the end of a page and
2981 * the next page isn't readable/writable, we get a fault. To prevent
2982 * that, copy back and forth to the full size.
2983 */
2984
2985 uifr = compat_alloc_user_space(sizeof(*uifr));
2986 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2987 return -EFAULT;
2988
2989 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2990
2991 if (!err) {
2992 switch (cmd) {
2993 case SIOCGIFFLAGS:
2994 case SIOCGIFMETRIC:
2995 case SIOCGIFMTU:
2996 case SIOCGIFMEM:
2997 case SIOCGIFHWADDR:
2998 case SIOCGIFINDEX:
2999 case SIOCGIFADDR:
3000 case SIOCGIFBRDADDR:
3001 case SIOCGIFDSTADDR:
3002 case SIOCGIFNETMASK:
3003 case SIOCGIFPFLAGS:
3004 case SIOCGIFTXQLEN:
3005 case SIOCGMIIPHY:
3006 case SIOCGMIIREG:
3007 case SIOCGIFNAME:
3008 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
3009 err = -EFAULT;
3010 break;
3011 }
3012 }
3013 return err;
3014 }
3015
3016 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3017 struct compat_ifreq __user *uifr32)
3018 {
3019 struct ifreq ifr;
3020 struct compat_ifmap __user *uifmap32;
3021 int err;
3022
3023 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3024 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3025 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3026 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3027 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3028 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3029 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3030 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3031 if (err)
3032 return -EFAULT;
3033
3034 err = dev_ioctl(net, cmd, &ifr, NULL);
3035
3036 if (cmd == SIOCGIFMAP && !err) {
3037 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3038 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3039 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3040 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3041 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3042 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3043 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3044 if (err)
3045 err = -EFAULT;
3046 }
3047 return err;
3048 }
3049
3050 struct rtentry32 {
3051 u32 rt_pad1;
3052 struct sockaddr rt_dst; /* target address */
3053 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3054 struct sockaddr rt_genmask; /* target network mask (IP) */
3055 unsigned short rt_flags;
3056 short rt_pad2;
3057 u32 rt_pad3;
3058 unsigned char rt_tos;
3059 unsigned char rt_class;
3060 short rt_pad4;
3061 short rt_metric; /* +1 for binary compatibility! */
3062 /* char * */ u32 rt_dev; /* forcing the device at add */
3063 u32 rt_mtu; /* per route MTU/Window */
3064 u32 rt_window; /* Window clamping */
3065 unsigned short rt_irtt; /* Initial RTT */
3066 };
3067
3068 struct in6_rtmsg32 {
3069 struct in6_addr rtmsg_dst;
3070 struct in6_addr rtmsg_src;
3071 struct in6_addr rtmsg_gateway;
3072 u32 rtmsg_type;
3073 u16 rtmsg_dst_len;
3074 u16 rtmsg_src_len;
3075 u32 rtmsg_metric;
3076 u32 rtmsg_info;
3077 u32 rtmsg_flags;
3078 s32 rtmsg_ifindex;
3079 };
3080
3081 static int routing_ioctl(struct net *net, struct socket *sock,
3082 unsigned int cmd, void __user *argp)
3083 {
3084 int ret;
3085 void *r = NULL;
3086 struct in6_rtmsg r6;
3087 struct rtentry r4;
3088 char devname[16];
3089 u32 rtdev;
3090 mm_segment_t old_fs = get_fs();
3091
3092 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3093 struct in6_rtmsg32 __user *ur6 = argp;
3094 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3095 3 * sizeof(struct in6_addr));
3096 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3097 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3098 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3099 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3100 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3101 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3102 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3103
3104 r = (void *) &r6;
3105 } else { /* ipv4 */
3106 struct rtentry32 __user *ur4 = argp;
3107 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3108 3 * sizeof(struct sockaddr));
3109 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3110 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3111 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3112 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3113 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3114 ret |= get_user(rtdev, &(ur4->rt_dev));
3115 if (rtdev) {
3116 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3117 r4.rt_dev = (char __user __force *)devname;
3118 devname[15] = 0;
3119 } else
3120 r4.rt_dev = NULL;
3121
3122 r = (void *) &r4;
3123 }
3124
3125 if (ret) {
3126 ret = -EFAULT;
3127 goto out;
3128 }
3129
3130 set_fs(KERNEL_DS);
3131 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3132 set_fs(old_fs);
3133
3134 out:
3135 return ret;
3136 }
3137
3138 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3139 * for some operations; this forces use of the newer bridge-utils that
3140 * use compatible ioctls
3141 */
3142 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3143 {
3144 compat_ulong_t tmp;
3145
3146 if (get_user(tmp, argp))
3147 return -EFAULT;
3148 if (tmp == BRCTL_GET_VERSION)
3149 return BRCTL_VERSION + 1;
3150 return -EINVAL;
3151 }
3152
3153 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3154 unsigned int cmd, unsigned long arg)
3155 {
3156 void __user *argp = compat_ptr(arg);
3157 struct sock *sk = sock->sk;
3158 struct net *net = sock_net(sk);
3159
3160 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3161 return compat_ifr_data_ioctl(net, cmd, argp);
3162
3163 switch (cmd) {
3164 case SIOCSIFBR:
3165 case SIOCGIFBR:
3166 return old_bridge_ioctl(argp);
3167 case SIOCGIFCONF:
3168 return compat_dev_ifconf(net, argp);
3169 case SIOCETHTOOL:
3170 return ethtool_ioctl(net, argp);
3171 case SIOCWANDEV:
3172 return compat_siocwandev(net, argp);
3173 case SIOCGIFMAP:
3174 case SIOCSIFMAP:
3175 return compat_sioc_ifmap(net, cmd, argp);
3176 case SIOCADDRT:
3177 case SIOCDELRT:
3178 return routing_ioctl(net, sock, cmd, argp);
3179 case SIOCGSTAMP:
3180 return do_siocgstamp(net, sock, cmd, argp);
3181 case SIOCGSTAMPNS:
3182 return do_siocgstampns(net, sock, cmd, argp);
3183 case SIOCBONDSLAVEINFOQUERY:
3184 case SIOCBONDINFOQUERY:
3185 case SIOCSHWTSTAMP:
3186 case SIOCGHWTSTAMP:
3187 return compat_ifr_data_ioctl(net, cmd, argp);
3188
3189 case FIOSETOWN:
3190 case SIOCSPGRP:
3191 case FIOGETOWN:
3192 case SIOCGPGRP:
3193 case SIOCBRADDBR:
3194 case SIOCBRDELBR:
3195 case SIOCGIFVLAN:
3196 case SIOCSIFVLAN:
3197 case SIOCADDDLCI:
3198 case SIOCDELDLCI:
3199 case SIOCGSKNS:
3200 return sock_ioctl(file, cmd, arg);
3201
3202 case SIOCGIFFLAGS:
3203 case SIOCSIFFLAGS:
3204 case SIOCGIFMETRIC:
3205 case SIOCSIFMETRIC:
3206 case SIOCGIFMTU:
3207 case SIOCSIFMTU:
3208 case SIOCGIFMEM:
3209 case SIOCSIFMEM:
3210 case SIOCGIFHWADDR:
3211 case SIOCSIFHWADDR:
3212 case SIOCADDMULTI:
3213 case SIOCDELMULTI:
3214 case SIOCGIFINDEX:
3215 case SIOCGIFADDR:
3216 case SIOCSIFADDR:
3217 case SIOCSIFHWBROADCAST:
3218 case SIOCDIFADDR:
3219 case SIOCGIFBRDADDR:
3220 case SIOCSIFBRDADDR:
3221 case SIOCGIFDSTADDR:
3222 case SIOCSIFDSTADDR:
3223 case SIOCGIFNETMASK:
3224 case SIOCSIFNETMASK:
3225 case SIOCSIFPFLAGS:
3226 case SIOCGIFPFLAGS:
3227 case SIOCGIFTXQLEN:
3228 case SIOCSIFTXQLEN:
3229 case SIOCBRADDIF:
3230 case SIOCBRDELIF:
3231 case SIOCGIFNAME:
3232 case SIOCSIFNAME:
3233 case SIOCGMIIPHY:
3234 case SIOCGMIIREG:
3235 case SIOCSMIIREG:
3236 case SIOCBONDENSLAVE:
3237 case SIOCBONDRELEASE:
3238 case SIOCBONDSETHWADDR:
3239 case SIOCBONDCHANGEACTIVE:
3240 return compat_ifreq_ioctl(net, sock, cmd, argp);
3241
3242 case SIOCSARP:
3243 case SIOCGARP:
3244 case SIOCDARP:
3245 case SIOCATMARK:
3246 return sock_do_ioctl(net, sock, cmd, arg);
3247 }
3248
3249 return -ENOIOCTLCMD;
3250 }
3251
3252 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3253 unsigned long arg)
3254 {
3255 struct socket *sock = file->private_data;
3256 int ret = -ENOIOCTLCMD;
3257 struct sock *sk;
3258 struct net *net;
3259
3260 sk = sock->sk;
3261 net = sock_net(sk);
3262
3263 if (sock->ops->compat_ioctl)
3264 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3265
3266 if (ret == -ENOIOCTLCMD &&
3267 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3268 ret = compat_wext_handle_ioctl(net, cmd, arg);
3269
3270 if (ret == -ENOIOCTLCMD)
3271 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3272
3273 return ret;
3274 }
3275 #endif
3276
3277 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3278 {
3279 return sock->ops->bind(sock, addr, addrlen);
3280 }
3281 EXPORT_SYMBOL(kernel_bind);
3282
3283 int kernel_listen(struct socket *sock, int backlog)
3284 {
3285 return sock->ops->listen(sock, backlog);
3286 }
3287 EXPORT_SYMBOL(kernel_listen);
3288
3289 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3290 {
3291 struct sock *sk = sock->sk;
3292 int err;
3293
3294 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3295 newsock);
3296 if (err < 0)
3297 goto done;
3298
3299 err = sock->ops->accept(sock, *newsock, flags, true);
3300 if (err < 0) {
3301 sock_release(*newsock);
3302 *newsock = NULL;
3303 goto done;
3304 }
3305
3306 (*newsock)->ops = sock->ops;
3307 __module_get((*newsock)->ops->owner);
3308
3309 done:
3310 return err;
3311 }
3312 EXPORT_SYMBOL(kernel_accept);
3313
3314 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3315 int flags)
3316 {
3317 return sock->ops->connect(sock, addr, addrlen, flags);
3318 }
3319 EXPORT_SYMBOL(kernel_connect);
3320
3321 int kernel_getsockname(struct socket *sock, struct sockaddr *addr)
3322 {
3323 return sock->ops->getname(sock, addr, 0);
3324 }
3325 EXPORT_SYMBOL(kernel_getsockname);
3326
3327 int kernel_getpeername(struct socket *sock, struct sockaddr *addr)
3328 {
3329 return sock->ops->getname(sock, addr, 1);
3330 }
3331 EXPORT_SYMBOL(kernel_getpeername);
3332
3333 int kernel_getsockopt(struct socket *sock, int level, int optname,
3334 char *optval, int *optlen)
3335 {
3336 mm_segment_t oldfs = get_fs();
3337 char __user *uoptval;
3338 int __user *uoptlen;
3339 int err;
3340
3341 uoptval = (char __user __force *) optval;
3342 uoptlen = (int __user __force *) optlen;
3343
3344 set_fs(KERNEL_DS);
3345 if (level == SOL_SOCKET)
3346 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3347 else
3348 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3349 uoptlen);
3350 set_fs(oldfs);
3351 return err;
3352 }
3353 EXPORT_SYMBOL(kernel_getsockopt);
3354
3355 int kernel_setsockopt(struct socket *sock, int level, int optname,
3356 char *optval, unsigned int optlen)
3357 {
3358 mm_segment_t oldfs = get_fs();
3359 char __user *uoptval;
3360 int err;
3361
3362 uoptval = (char __user __force *) optval;
3363
3364 set_fs(KERNEL_DS);
3365 if (level == SOL_SOCKET)
3366 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3367 else
3368 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3369 optlen);
3370 set_fs(oldfs);
3371 return err;
3372 }
3373 EXPORT_SYMBOL(kernel_setsockopt);
3374
3375 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3376 size_t size, int flags)
3377 {
3378 if (sock->ops->sendpage)
3379 return sock->ops->sendpage(sock, page, offset, size, flags);
3380
3381 return sock_no_sendpage(sock, page, offset, size, flags);
3382 }
3383 EXPORT_SYMBOL(kernel_sendpage);
3384
3385 int kernel_sendpage_locked(struct sock *sk, struct page *page, int offset,
3386 size_t size, int flags)
3387 {
3388 struct socket *sock = sk->sk_socket;
3389
3390 if (sock->ops->sendpage_locked)
3391 return sock->ops->sendpage_locked(sk, page, offset, size,
3392 flags);
3393
3394 return sock_no_sendpage_locked(sk, page, offset, size, flags);
3395 }
3396 EXPORT_SYMBOL(kernel_sendpage_locked);
3397
3398 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3399 {
3400 return sock->ops->shutdown(sock, how);
3401 }
3402 EXPORT_SYMBOL(kernel_sock_shutdown);
3403
3404 /* This routine returns the IP overhead imposed by a socket i.e.
3405 * the length of the underlying IP header, depending on whether
3406 * this is an IPv4 or IPv6 socket and the length from IP options turned
3407 * on at the socket. Assumes that the caller has a lock on the socket.
3408 */
3409 u32 kernel_sock_ip_overhead(struct sock *sk)
3410 {
3411 struct inet_sock *inet;
3412 struct ip_options_rcu *opt;
3413 u32 overhead = 0;
3414 #if IS_ENABLED(CONFIG_IPV6)
3415 struct ipv6_pinfo *np;
3416 struct ipv6_txoptions *optv6 = NULL;
3417 #endif /* IS_ENABLED(CONFIG_IPV6) */
3418
3419 if (!sk)
3420 return overhead;
3421
3422 switch (sk->sk_family) {
3423 case AF_INET:
3424 inet = inet_sk(sk);
3425 overhead += sizeof(struct iphdr);
3426 opt = rcu_dereference_protected(inet->inet_opt,
3427 sock_owned_by_user(sk));
3428 if (opt)
3429 overhead += opt->opt.optlen;
3430 return overhead;
3431 #if IS_ENABLED(CONFIG_IPV6)
3432 case AF_INET6:
3433 np = inet6_sk(sk);
3434 overhead += sizeof(struct ipv6hdr);
3435 if (np)
3436 optv6 = rcu_dereference_protected(np->opt,
3437 sock_owned_by_user(sk));
3438 if (optv6)
3439 overhead += (optv6->opt_flen + optv6->opt_nflen);
3440 return overhead;
3441 #endif /* IS_ENABLED(CONFIG_IPV6) */
3442 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3443 return overhead;
3444 }
3445 }
3446 EXPORT_SYMBOL(kernel_sock_ip_overhead);
3447
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.