]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/socket.c
sctp: define the member stream as an object instead of pointer in asoc
[mirror_ubuntu-artful-kernel.git] / net / socket.c
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
93 #include <linux/uaccess.h>
94 #include <asm/unistd.h>
95
96 #include <net/compat.h>
97 #include <net/wext.h>
98 #include <net/cls_cgroup.h>
99
100 #include <net/sock.h>
101 #include <linux/netfilter.h>
102
103 #include <linux/if_tun.h>
104 #include <linux/ipv6_route.h>
105 #include <linux/route.h>
106 #include <linux/sockios.h>
107 #include <linux/atalk.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 unsigned int 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 * Statistics counters of the socket lists
167 */
168
169 static DEFINE_PER_CPU(int, sockets_in_use);
170
171 /*
172 * Support routines.
173 * Move socket addresses back and forth across the kernel/user
174 * divide and look after the messy bits.
175 */
176
177 /**
178 * move_addr_to_kernel - copy a socket address into kernel space
179 * @uaddr: Address in user space
180 * @kaddr: Address in kernel space
181 * @ulen: Length in user space
182 *
183 * The address is copied into kernel space. If the provided address is
184 * too long an error code of -EINVAL is returned. If the copy gives
185 * invalid addresses -EFAULT is returned. On a success 0 is returned.
186 */
187
188 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr_storage *kaddr)
189 {
190 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191 return -EINVAL;
192 if (ulen == 0)
193 return 0;
194 if (copy_from_user(kaddr, uaddr, ulen))
195 return -EFAULT;
196 return audit_sockaddr(ulen, kaddr);
197 }
198
199 /**
200 * move_addr_to_user - copy an address to user space
201 * @kaddr: kernel space address
202 * @klen: length of address in kernel
203 * @uaddr: user space address
204 * @ulen: pointer to user length field
205 *
206 * The value pointed to by ulen on entry is the buffer length available.
207 * This is overwritten with the buffer space used. -EINVAL is returned
208 * if an overlong buffer is specified or a negative buffer size. -EFAULT
209 * is returned if either the buffer or the length field are not
210 * accessible.
211 * After copying the data up to the limit the user specifies, the true
212 * length of the data is written over the length limit the user
213 * specified. Zero is returned for a success.
214 */
215
216 static int move_addr_to_user(struct sockaddr_storage *kaddr, int klen,
217 void __user *uaddr, int __user *ulen)
218 {
219 int err;
220 int len;
221
222 BUG_ON(klen > sizeof(struct sockaddr_storage));
223 err = get_user(len, ulen);
224 if (err)
225 return err;
226 if (len > klen)
227 len = klen;
228 if (len < 0)
229 return -EINVAL;
230 if (len) {
231 if (audit_sockaddr(klen, kaddr))
232 return -ENOMEM;
233 if (copy_to_user(uaddr, kaddr, len))
234 return -EFAULT;
235 }
236 /*
237 * "fromlen shall refer to the value before truncation.."
238 * 1003.1g
239 */
240 return __put_user(klen, ulen);
241 }
242
243 static struct kmem_cache *sock_inode_cachep __read_mostly;
244
245 static struct inode *sock_alloc_inode(struct super_block *sb)
246 {
247 struct socket_alloc *ei;
248 struct socket_wq *wq;
249
250 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
251 if (!ei)
252 return NULL;
253 wq = kmalloc(sizeof(*wq), GFP_KERNEL);
254 if (!wq) {
255 kmem_cache_free(sock_inode_cachep, ei);
256 return NULL;
257 }
258 init_waitqueue_head(&wq->wait);
259 wq->fasync_list = NULL;
260 wq->flags = 0;
261 RCU_INIT_POINTER(ei->socket.wq, wq);
262
263 ei->socket.state = SS_UNCONNECTED;
264 ei->socket.flags = 0;
265 ei->socket.ops = NULL;
266 ei->socket.sk = NULL;
267 ei->socket.file = NULL;
268
269 return &ei->vfs_inode;
270 }
271
272 static void sock_destroy_inode(struct inode *inode)
273 {
274 struct socket_alloc *ei;
275 struct socket_wq *wq;
276
277 ei = container_of(inode, struct socket_alloc, vfs_inode);
278 wq = rcu_dereference_protected(ei->socket.wq, 1);
279 kfree_rcu(wq, rcu);
280 kmem_cache_free(sock_inode_cachep, ei);
281 }
282
283 static void init_once(void *foo)
284 {
285 struct socket_alloc *ei = (struct socket_alloc *)foo;
286
287 inode_init_once(&ei->vfs_inode);
288 }
289
290 static void init_inodecache(void)
291 {
292 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
293 sizeof(struct socket_alloc),
294 0,
295 (SLAB_HWCACHE_ALIGN |
296 SLAB_RECLAIM_ACCOUNT |
297 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
298 init_once);
299 BUG_ON(sock_inode_cachep == NULL);
300 }
301
302 static const struct super_operations sockfs_ops = {
303 .alloc_inode = sock_alloc_inode,
304 .destroy_inode = sock_destroy_inode,
305 .statfs = simple_statfs,
306 };
307
308 /*
309 * sockfs_dname() is called from d_path().
310 */
311 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
312 {
313 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
314 d_inode(dentry)->i_ino);
315 }
316
317 static const struct dentry_operations sockfs_dentry_operations = {
318 .d_dname = sockfs_dname,
319 };
320
321 static int sockfs_xattr_get(const struct xattr_handler *handler,
322 struct dentry *dentry, struct inode *inode,
323 const char *suffix, void *value, size_t size)
324 {
325 if (value) {
326 if (dentry->d_name.len + 1 > size)
327 return -ERANGE;
328 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
329 }
330 return dentry->d_name.len + 1;
331 }
332
333 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
334 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
335 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
336
337 static const struct xattr_handler sockfs_xattr_handler = {
338 .name = XATTR_NAME_SOCKPROTONAME,
339 .get = sockfs_xattr_get,
340 };
341
342 static int sockfs_security_xattr_set(const struct xattr_handler *handler,
343 struct dentry *dentry, struct inode *inode,
344 const char *suffix, const void *value,
345 size_t size, int flags)
346 {
347 /* Handled by LSM. */
348 return -EAGAIN;
349 }
350
351 static const struct xattr_handler sockfs_security_xattr_handler = {
352 .prefix = XATTR_SECURITY_PREFIX,
353 .set = sockfs_security_xattr_set,
354 };
355
356 static const struct xattr_handler *sockfs_xattr_handlers[] = {
357 &sockfs_xattr_handler,
358 &sockfs_security_xattr_handler,
359 NULL
360 };
361
362 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
363 int flags, const char *dev_name, void *data)
364 {
365 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
366 sockfs_xattr_handlers,
367 &sockfs_dentry_operations, SOCKFS_MAGIC);
368 }
369
370 static struct vfsmount *sock_mnt __read_mostly;
371
372 static struct file_system_type sock_fs_type = {
373 .name = "sockfs",
374 .mount = sockfs_mount,
375 .kill_sb = kill_anon_super,
376 };
377
378 /*
379 * Obtains the first available file descriptor and sets it up for use.
380 *
381 * These functions create file structures and maps them to fd space
382 * of the current process. On success it returns file descriptor
383 * and file struct implicitly stored in sock->file.
384 * Note that another thread may close file descriptor before we return
385 * from this function. We use the fact that now we do not refer
386 * to socket after mapping. If one day we will need it, this
387 * function will increment ref. count on file by 1.
388 *
389 * In any case returned fd MAY BE not valid!
390 * This race condition is unavoidable
391 * with shared fd spaces, we cannot solve it inside kernel,
392 * but we take care of internal coherence yet.
393 */
394
395 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
396 {
397 struct qstr name = { .name = "" };
398 struct path path;
399 struct file *file;
400
401 if (dname) {
402 name.name = dname;
403 name.len = strlen(name.name);
404 } else if (sock->sk) {
405 name.name = sock->sk->sk_prot_creator->name;
406 name.len = strlen(name.name);
407 }
408 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
409 if (unlikely(!path.dentry))
410 return ERR_PTR(-ENOMEM);
411 path.mnt = mntget(sock_mnt);
412
413 d_instantiate(path.dentry, SOCK_INODE(sock));
414
415 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
416 &socket_file_ops);
417 if (IS_ERR(file)) {
418 /* drop dentry, keep inode */
419 ihold(d_inode(path.dentry));
420 path_put(&path);
421 return file;
422 }
423
424 sock->file = file;
425 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
426 file->private_data = sock;
427 return file;
428 }
429 EXPORT_SYMBOL(sock_alloc_file);
430
431 static int sock_map_fd(struct socket *sock, int flags)
432 {
433 struct file *newfile;
434 int fd = get_unused_fd_flags(flags);
435 if (unlikely(fd < 0))
436 return fd;
437
438 newfile = sock_alloc_file(sock, flags, NULL);
439 if (likely(!IS_ERR(newfile))) {
440 fd_install(fd, newfile);
441 return fd;
442 }
443
444 put_unused_fd(fd);
445 return PTR_ERR(newfile);
446 }
447
448 struct socket *sock_from_file(struct file *file, int *err)
449 {
450 if (file->f_op == &socket_file_ops)
451 return file->private_data; /* set in sock_map_fd */
452
453 *err = -ENOTSOCK;
454 return NULL;
455 }
456 EXPORT_SYMBOL(sock_from_file);
457
458 /**
459 * sockfd_lookup - Go from a file number to its socket slot
460 * @fd: file handle
461 * @err: pointer to an error code return
462 *
463 * The file handle passed in is locked and the socket it is bound
464 * to is returned. If an error occurs the err pointer is overwritten
465 * with a negative errno code and NULL is returned. The function checks
466 * for both invalid handles and passing a handle which is not a socket.
467 *
468 * On a success the socket object pointer is returned.
469 */
470
471 struct socket *sockfd_lookup(int fd, int *err)
472 {
473 struct file *file;
474 struct socket *sock;
475
476 file = fget(fd);
477 if (!file) {
478 *err = -EBADF;
479 return NULL;
480 }
481
482 sock = sock_from_file(file, err);
483 if (!sock)
484 fput(file);
485 return sock;
486 }
487 EXPORT_SYMBOL(sockfd_lookup);
488
489 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
490 {
491 struct fd f = fdget(fd);
492 struct socket *sock;
493
494 *err = -EBADF;
495 if (f.file) {
496 sock = sock_from_file(f.file, err);
497 if (likely(sock)) {
498 *fput_needed = f.flags;
499 return sock;
500 }
501 fdput(f);
502 }
503 return NULL;
504 }
505
506 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
507 size_t size)
508 {
509 ssize_t len;
510 ssize_t used = 0;
511
512 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
513 if (len < 0)
514 return len;
515 used += len;
516 if (buffer) {
517 if (size < used)
518 return -ERANGE;
519 buffer += len;
520 }
521
522 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
523 used += len;
524 if (buffer) {
525 if (size < used)
526 return -ERANGE;
527 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
528 buffer += len;
529 }
530
531 return used;
532 }
533
534 static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
535 {
536 int err = simple_setattr(dentry, iattr);
537
538 if (!err && (iattr->ia_valid & ATTR_UID)) {
539 struct socket *sock = SOCKET_I(d_inode(dentry));
540
541 sock->sk->sk_uid = iattr->ia_uid;
542 }
543
544 return err;
545 }
546
547 static const struct inode_operations sockfs_inode_ops = {
548 .listxattr = sockfs_listxattr,
549 .setattr = sockfs_setattr,
550 };
551
552 /**
553 * sock_alloc - allocate a socket
554 *
555 * Allocate a new inode and socket object. The two are bound together
556 * and initialised. The socket is then returned. If we are out of inodes
557 * NULL is returned.
558 */
559
560 struct socket *sock_alloc(void)
561 {
562 struct inode *inode;
563 struct socket *sock;
564
565 inode = new_inode_pseudo(sock_mnt->mnt_sb);
566 if (!inode)
567 return NULL;
568
569 sock = SOCKET_I(inode);
570
571 kmemcheck_annotate_bitfield(sock, type);
572 inode->i_ino = get_next_ino();
573 inode->i_mode = S_IFSOCK | S_IRWXUGO;
574 inode->i_uid = current_fsuid();
575 inode->i_gid = current_fsgid();
576 inode->i_op = &sockfs_inode_ops;
577
578 this_cpu_add(sockets_in_use, 1);
579 return sock;
580 }
581 EXPORT_SYMBOL(sock_alloc);
582
583 /**
584 * sock_release - close a socket
585 * @sock: socket to close
586 *
587 * The socket is released from the protocol stack if it has a release
588 * callback, and the inode is then released if the socket is bound to
589 * an inode not a file.
590 */
591
592 void sock_release(struct socket *sock)
593 {
594 if (sock->ops) {
595 struct module *owner = sock->ops->owner;
596
597 sock->ops->release(sock);
598 sock->ops = NULL;
599 module_put(owner);
600 }
601
602 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
603 pr_err("%s: fasync list not empty!\n", __func__);
604
605 this_cpu_sub(sockets_in_use, 1);
606 if (!sock->file) {
607 iput(SOCK_INODE(sock));
608 return;
609 }
610 sock->file = NULL;
611 }
612 EXPORT_SYMBOL(sock_release);
613
614 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
615 {
616 u8 flags = *tx_flags;
617
618 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
619 flags |= SKBTX_HW_TSTAMP;
620
621 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
622 flags |= SKBTX_SW_TSTAMP;
623
624 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
625 flags |= SKBTX_SCHED_TSTAMP;
626
627 *tx_flags = flags;
628 }
629 EXPORT_SYMBOL(__sock_tx_timestamp);
630
631 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
632 {
633 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
634 BUG_ON(ret == -EIOCBQUEUED);
635 return ret;
636 }
637
638 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
639 {
640 int err = security_socket_sendmsg(sock, msg,
641 msg_data_left(msg));
642
643 return err ?: sock_sendmsg_nosec(sock, msg);
644 }
645 EXPORT_SYMBOL(sock_sendmsg);
646
647 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
648 struct kvec *vec, size_t num, size_t size)
649 {
650 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
651 return sock_sendmsg(sock, msg);
652 }
653 EXPORT_SYMBOL(kernel_sendmsg);
654
655 static bool skb_is_err_queue(const struct sk_buff *skb)
656 {
657 /* pkt_type of skbs enqueued on the error queue are set to
658 * PACKET_OUTGOING in skb_set_err_queue(). This is only safe to do
659 * in recvmsg, since skbs received on a local socket will never
660 * have a pkt_type of PACKET_OUTGOING.
661 */
662 return skb->pkt_type == PACKET_OUTGOING;
663 }
664
665 /* On transmit, software and hardware timestamps are returned independently.
666 * As the two skb clones share the hardware timestamp, which may be updated
667 * before the software timestamp is received, a hardware TX timestamp may be
668 * returned only if there is no software TX timestamp. Ignore false software
669 * timestamps, which may be made in the __sock_recv_timestamp() call when the
670 * option SO_TIMESTAMP(NS) is enabled on the socket, even when the skb has a
671 * hardware timestamp.
672 */
673 static bool skb_is_swtx_tstamp(const struct sk_buff *skb, int false_tstamp)
674 {
675 return skb->tstamp && !false_tstamp && skb_is_err_queue(skb);
676 }
677
678 static void put_ts_pktinfo(struct msghdr *msg, struct sk_buff *skb)
679 {
680 struct scm_ts_pktinfo ts_pktinfo;
681 struct net_device *orig_dev;
682
683 if (!skb_mac_header_was_set(skb))
684 return;
685
686 memset(&ts_pktinfo, 0, sizeof(ts_pktinfo));
687
688 rcu_read_lock();
689 orig_dev = dev_get_by_napi_id(skb_napi_id(skb));
690 if (orig_dev)
691 ts_pktinfo.if_index = orig_dev->ifindex;
692 rcu_read_unlock();
693
694 ts_pktinfo.pkt_length = skb->len - skb_mac_offset(skb);
695 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_PKTINFO,
696 sizeof(ts_pktinfo), &ts_pktinfo);
697 }
698
699 /*
700 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
701 */
702 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
703 struct sk_buff *skb)
704 {
705 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
706 struct scm_timestamping tss;
707 int empty = 1, false_tstamp = 0;
708 struct skb_shared_hwtstamps *shhwtstamps =
709 skb_hwtstamps(skb);
710
711 /* Race occurred between timestamp enabling and packet
712 receiving. Fill in the current time for now. */
713 if (need_software_tstamp && skb->tstamp == 0) {
714 __net_timestamp(skb);
715 false_tstamp = 1;
716 }
717
718 if (need_software_tstamp) {
719 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
720 struct timeval tv;
721 skb_get_timestamp(skb, &tv);
722 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
723 sizeof(tv), &tv);
724 } else {
725 struct timespec ts;
726 skb_get_timestampns(skb, &ts);
727 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
728 sizeof(ts), &ts);
729 }
730 }
731
732 memset(&tss, 0, sizeof(tss));
733 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
734 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
735 empty = 0;
736 if (shhwtstamps &&
737 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
738 !skb_is_swtx_tstamp(skb, false_tstamp) &&
739 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2)) {
740 empty = 0;
741 if ((sk->sk_tsflags & SOF_TIMESTAMPING_OPT_PKTINFO) &&
742 !skb_is_err_queue(skb))
743 put_ts_pktinfo(msg, skb);
744 }
745 if (!empty) {
746 put_cmsg(msg, SOL_SOCKET,
747 SCM_TIMESTAMPING, sizeof(tss), &tss);
748
749 if (skb_is_err_queue(skb) && skb->len &&
750 SKB_EXT_ERR(skb)->opt_stats)
751 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
752 skb->len, skb->data);
753 }
754 }
755 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
756
757 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
758 struct sk_buff *skb)
759 {
760 int ack;
761
762 if (!sock_flag(sk, SOCK_WIFI_STATUS))
763 return;
764 if (!skb->wifi_acked_valid)
765 return;
766
767 ack = skb->wifi_acked;
768
769 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
770 }
771 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
772
773 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
774 struct sk_buff *skb)
775 {
776 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
777 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
778 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
779 }
780
781 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
782 struct sk_buff *skb)
783 {
784 sock_recv_timestamp(msg, sk, skb);
785 sock_recv_drops(msg, sk, skb);
786 }
787 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
788
789 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
790 int flags)
791 {
792 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
793 }
794
795 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
796 {
797 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
798
799 return err ?: sock_recvmsg_nosec(sock, msg, flags);
800 }
801 EXPORT_SYMBOL(sock_recvmsg);
802
803 /**
804 * kernel_recvmsg - Receive a message from a socket (kernel space)
805 * @sock: The socket to receive the message from
806 * @msg: Received message
807 * @vec: Input s/g array for message data
808 * @num: Size of input s/g array
809 * @size: Number of bytes to read
810 * @flags: Message flags (MSG_DONTWAIT, etc...)
811 *
812 * On return the msg structure contains the scatter/gather array passed in the
813 * vec argument. The array is modified so that it consists of the unfilled
814 * portion of the original array.
815 *
816 * The returned value is the total number of bytes received, or an error.
817 */
818 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
819 struct kvec *vec, size_t num, size_t size, int flags)
820 {
821 mm_segment_t oldfs = get_fs();
822 int result;
823
824 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
825 set_fs(KERNEL_DS);
826 result = sock_recvmsg(sock, msg, flags);
827 set_fs(oldfs);
828 return result;
829 }
830 EXPORT_SYMBOL(kernel_recvmsg);
831
832 static ssize_t sock_sendpage(struct file *file, struct page *page,
833 int offset, size_t size, loff_t *ppos, int more)
834 {
835 struct socket *sock;
836 int flags;
837
838 sock = file->private_data;
839
840 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
841 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
842 flags |= more;
843
844 return kernel_sendpage(sock, page, offset, size, flags);
845 }
846
847 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
848 struct pipe_inode_info *pipe, size_t len,
849 unsigned int flags)
850 {
851 struct socket *sock = file->private_data;
852
853 if (unlikely(!sock->ops->splice_read))
854 return -EINVAL;
855
856 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
857 }
858
859 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
860 {
861 struct file *file = iocb->ki_filp;
862 struct socket *sock = file->private_data;
863 struct msghdr msg = {.msg_iter = *to,
864 .msg_iocb = iocb};
865 ssize_t res;
866
867 if (file->f_flags & O_NONBLOCK)
868 msg.msg_flags = MSG_DONTWAIT;
869
870 if (iocb->ki_pos != 0)
871 return -ESPIPE;
872
873 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
874 return 0;
875
876 res = sock_recvmsg(sock, &msg, msg.msg_flags);
877 *to = msg.msg_iter;
878 return res;
879 }
880
881 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
882 {
883 struct file *file = iocb->ki_filp;
884 struct socket *sock = file->private_data;
885 struct msghdr msg = {.msg_iter = *from,
886 .msg_iocb = iocb};
887 ssize_t res;
888
889 if (iocb->ki_pos != 0)
890 return -ESPIPE;
891
892 if (file->f_flags & O_NONBLOCK)
893 msg.msg_flags = MSG_DONTWAIT;
894
895 if (sock->type == SOCK_SEQPACKET)
896 msg.msg_flags |= MSG_EOR;
897
898 res = sock_sendmsg(sock, &msg);
899 *from = msg.msg_iter;
900 return res;
901 }
902
903 /*
904 * Atomic setting of ioctl hooks to avoid race
905 * with module unload.
906 */
907
908 static DEFINE_MUTEX(br_ioctl_mutex);
909 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
910
911 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
912 {
913 mutex_lock(&br_ioctl_mutex);
914 br_ioctl_hook = hook;
915 mutex_unlock(&br_ioctl_mutex);
916 }
917 EXPORT_SYMBOL(brioctl_set);
918
919 static DEFINE_MUTEX(vlan_ioctl_mutex);
920 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
921
922 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
923 {
924 mutex_lock(&vlan_ioctl_mutex);
925 vlan_ioctl_hook = hook;
926 mutex_unlock(&vlan_ioctl_mutex);
927 }
928 EXPORT_SYMBOL(vlan_ioctl_set);
929
930 static DEFINE_MUTEX(dlci_ioctl_mutex);
931 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
932
933 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
934 {
935 mutex_lock(&dlci_ioctl_mutex);
936 dlci_ioctl_hook = hook;
937 mutex_unlock(&dlci_ioctl_mutex);
938 }
939 EXPORT_SYMBOL(dlci_ioctl_set);
940
941 static long sock_do_ioctl(struct net *net, struct socket *sock,
942 unsigned int cmd, unsigned long arg)
943 {
944 int err;
945 void __user *argp = (void __user *)arg;
946
947 err = sock->ops->ioctl(sock, cmd, arg);
948
949 /*
950 * If this ioctl is unknown try to hand it down
951 * to the NIC driver.
952 */
953 if (err == -ENOIOCTLCMD)
954 err = dev_ioctl(net, cmd, argp);
955
956 return err;
957 }
958
959 /*
960 * With an ioctl, arg may well be a user mode pointer, but we don't know
961 * what to do with it - that's up to the protocol still.
962 */
963
964 static struct ns_common *get_net_ns(struct ns_common *ns)
965 {
966 return &get_net(container_of(ns, struct net, ns))->ns;
967 }
968
969 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
970 {
971 struct socket *sock;
972 struct sock *sk;
973 void __user *argp = (void __user *)arg;
974 int pid, err;
975 struct net *net;
976
977 sock = file->private_data;
978 sk = sock->sk;
979 net = sock_net(sk);
980 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
981 err = dev_ioctl(net, cmd, argp);
982 } else
983 #ifdef CONFIG_WEXT_CORE
984 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
985 err = dev_ioctl(net, cmd, argp);
986 } else
987 #endif
988 switch (cmd) {
989 case FIOSETOWN:
990 case SIOCSPGRP:
991 err = -EFAULT;
992 if (get_user(pid, (int __user *)argp))
993 break;
994 f_setown(sock->file, pid, 1);
995 err = 0;
996 break;
997 case FIOGETOWN:
998 case SIOCGPGRP:
999 err = put_user(f_getown(sock->file),
1000 (int __user *)argp);
1001 break;
1002 case SIOCGIFBR:
1003 case SIOCSIFBR:
1004 case SIOCBRADDBR:
1005 case SIOCBRDELBR:
1006 err = -ENOPKG;
1007 if (!br_ioctl_hook)
1008 request_module("bridge");
1009
1010 mutex_lock(&br_ioctl_mutex);
1011 if (br_ioctl_hook)
1012 err = br_ioctl_hook(net, cmd, argp);
1013 mutex_unlock(&br_ioctl_mutex);
1014 break;
1015 case SIOCGIFVLAN:
1016 case SIOCSIFVLAN:
1017 err = -ENOPKG;
1018 if (!vlan_ioctl_hook)
1019 request_module("8021q");
1020
1021 mutex_lock(&vlan_ioctl_mutex);
1022 if (vlan_ioctl_hook)
1023 err = vlan_ioctl_hook(net, argp);
1024 mutex_unlock(&vlan_ioctl_mutex);
1025 break;
1026 case SIOCADDDLCI:
1027 case SIOCDELDLCI:
1028 err = -ENOPKG;
1029 if (!dlci_ioctl_hook)
1030 request_module("dlci");
1031
1032 mutex_lock(&dlci_ioctl_mutex);
1033 if (dlci_ioctl_hook)
1034 err = dlci_ioctl_hook(cmd, argp);
1035 mutex_unlock(&dlci_ioctl_mutex);
1036 break;
1037 case SIOCGSKNS:
1038 err = -EPERM;
1039 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
1040 break;
1041
1042 err = open_related_ns(&net->ns, get_net_ns);
1043 break;
1044 default:
1045 err = sock_do_ioctl(net, sock, cmd, arg);
1046 break;
1047 }
1048 return err;
1049 }
1050
1051 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1052 {
1053 int err;
1054 struct socket *sock = NULL;
1055
1056 err = security_socket_create(family, type, protocol, 1);
1057 if (err)
1058 goto out;
1059
1060 sock = sock_alloc();
1061 if (!sock) {
1062 err = -ENOMEM;
1063 goto out;
1064 }
1065
1066 sock->type = type;
1067 err = security_socket_post_create(sock, family, type, protocol, 1);
1068 if (err)
1069 goto out_release;
1070
1071 out:
1072 *res = sock;
1073 return err;
1074 out_release:
1075 sock_release(sock);
1076 sock = NULL;
1077 goto out;
1078 }
1079 EXPORT_SYMBOL(sock_create_lite);
1080
1081 /* No kernel lock held - perfect */
1082 static unsigned int sock_poll(struct file *file, poll_table *wait)
1083 {
1084 unsigned int busy_flag = 0;
1085 struct socket *sock;
1086
1087 /*
1088 * We can't return errors to poll, so it's either yes or no.
1089 */
1090 sock = file->private_data;
1091
1092 if (sk_can_busy_loop(sock->sk)) {
1093 /* this socket can poll_ll so tell the system call */
1094 busy_flag = POLL_BUSY_LOOP;
1095
1096 /* once, only if requested by syscall */
1097 if (wait && (wait->_key & POLL_BUSY_LOOP))
1098 sk_busy_loop(sock->sk, 1);
1099 }
1100
1101 return busy_flag | sock->ops->poll(file, sock, wait);
1102 }
1103
1104 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1105 {
1106 struct socket *sock = file->private_data;
1107
1108 return sock->ops->mmap(file, sock, vma);
1109 }
1110
1111 static int sock_close(struct inode *inode, struct file *filp)
1112 {
1113 sock_release(SOCKET_I(inode));
1114 return 0;
1115 }
1116
1117 /*
1118 * Update the socket async list
1119 *
1120 * Fasync_list locking strategy.
1121 *
1122 * 1. fasync_list is modified only under process context socket lock
1123 * i.e. under semaphore.
1124 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1125 * or under socket lock
1126 */
1127
1128 static int sock_fasync(int fd, struct file *filp, int on)
1129 {
1130 struct socket *sock = filp->private_data;
1131 struct sock *sk = sock->sk;
1132 struct socket_wq *wq;
1133
1134 if (sk == NULL)
1135 return -EINVAL;
1136
1137 lock_sock(sk);
1138 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1139 fasync_helper(fd, filp, on, &wq->fasync_list);
1140
1141 if (!wq->fasync_list)
1142 sock_reset_flag(sk, SOCK_FASYNC);
1143 else
1144 sock_set_flag(sk, SOCK_FASYNC);
1145
1146 release_sock(sk);
1147 return 0;
1148 }
1149
1150 /* This function may be called only under rcu_lock */
1151
1152 int sock_wake_async(struct socket_wq *wq, int how, int band)
1153 {
1154 if (!wq || !wq->fasync_list)
1155 return -1;
1156
1157 switch (how) {
1158 case SOCK_WAKE_WAITD:
1159 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1160 break;
1161 goto call_kill;
1162 case SOCK_WAKE_SPACE:
1163 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1164 break;
1165 /* fall through */
1166 case SOCK_WAKE_IO:
1167 call_kill:
1168 kill_fasync(&wq->fasync_list, SIGIO, band);
1169 break;
1170 case SOCK_WAKE_URG:
1171 kill_fasync(&wq->fasync_list, SIGURG, band);
1172 }
1173
1174 return 0;
1175 }
1176 EXPORT_SYMBOL(sock_wake_async);
1177
1178 int __sock_create(struct net *net, int family, int type, int protocol,
1179 struct socket **res, int kern)
1180 {
1181 int err;
1182 struct socket *sock;
1183 const struct net_proto_family *pf;
1184
1185 /*
1186 * Check protocol is in range
1187 */
1188 if (family < 0 || family >= NPROTO)
1189 return -EAFNOSUPPORT;
1190 if (type < 0 || type >= SOCK_MAX)
1191 return -EINVAL;
1192
1193 /* Compatibility.
1194
1195 This uglymoron is moved from INET layer to here to avoid
1196 deadlock in module load.
1197 */
1198 if (family == PF_INET && type == SOCK_PACKET) {
1199 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1200 current->comm);
1201 family = PF_PACKET;
1202 }
1203
1204 err = security_socket_create(family, type, protocol, kern);
1205 if (err)
1206 return err;
1207
1208 /*
1209 * Allocate the socket and allow the family to set things up. if
1210 * the protocol is 0, the family is instructed to select an appropriate
1211 * default.
1212 */
1213 sock = sock_alloc();
1214 if (!sock) {
1215 net_warn_ratelimited("socket: no more sockets\n");
1216 return -ENFILE; /* Not exactly a match, but its the
1217 closest posix thing */
1218 }
1219
1220 sock->type = type;
1221
1222 #ifdef CONFIG_MODULES
1223 /* Attempt to load a protocol module if the find failed.
1224 *
1225 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1226 * requested real, full-featured networking support upon configuration.
1227 * Otherwise module support will break!
1228 */
1229 if (rcu_access_pointer(net_families[family]) == NULL)
1230 request_module("net-pf-%d", family);
1231 #endif
1232
1233 rcu_read_lock();
1234 pf = rcu_dereference(net_families[family]);
1235 err = -EAFNOSUPPORT;
1236 if (!pf)
1237 goto out_release;
1238
1239 /*
1240 * We will call the ->create function, that possibly is in a loadable
1241 * module, so we have to bump that loadable module refcnt first.
1242 */
1243 if (!try_module_get(pf->owner))
1244 goto out_release;
1245
1246 /* Now protected by module ref count */
1247 rcu_read_unlock();
1248
1249 err = pf->create(net, sock, protocol, kern);
1250 if (err < 0)
1251 goto out_module_put;
1252
1253 /*
1254 * Now to bump the refcnt of the [loadable] module that owns this
1255 * socket at sock_release time we decrement its refcnt.
1256 */
1257 if (!try_module_get(sock->ops->owner))
1258 goto out_module_busy;
1259
1260 /*
1261 * Now that we're done with the ->create function, the [loadable]
1262 * module can have its refcnt decremented
1263 */
1264 module_put(pf->owner);
1265 err = security_socket_post_create(sock, family, type, protocol, kern);
1266 if (err)
1267 goto out_sock_release;
1268 *res = sock;
1269
1270 return 0;
1271
1272 out_module_busy:
1273 err = -EAFNOSUPPORT;
1274 out_module_put:
1275 sock->ops = NULL;
1276 module_put(pf->owner);
1277 out_sock_release:
1278 sock_release(sock);
1279 return err;
1280
1281 out_release:
1282 rcu_read_unlock();
1283 goto out_sock_release;
1284 }
1285 EXPORT_SYMBOL(__sock_create);
1286
1287 int sock_create(int family, int type, int protocol, struct socket **res)
1288 {
1289 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1290 }
1291 EXPORT_SYMBOL(sock_create);
1292
1293 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1294 {
1295 return __sock_create(net, family, type, protocol, res, 1);
1296 }
1297 EXPORT_SYMBOL(sock_create_kern);
1298
1299 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1300 {
1301 int retval;
1302 struct socket *sock;
1303 int flags;
1304
1305 /* Check the SOCK_* constants for consistency. */
1306 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1307 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1308 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1309 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1310
1311 flags = type & ~SOCK_TYPE_MASK;
1312 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1313 return -EINVAL;
1314 type &= SOCK_TYPE_MASK;
1315
1316 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1317 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1318
1319 retval = sock_create(family, type, protocol, &sock);
1320 if (retval < 0)
1321 goto out;
1322
1323 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1324 if (retval < 0)
1325 goto out_release;
1326
1327 out:
1328 /* It may be already another descriptor 8) Not kernel problem. */
1329 return retval;
1330
1331 out_release:
1332 sock_release(sock);
1333 return retval;
1334 }
1335
1336 /*
1337 * Create a pair of connected sockets.
1338 */
1339
1340 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1341 int __user *, usockvec)
1342 {
1343 struct socket *sock1, *sock2;
1344 int fd1, fd2, err;
1345 struct file *newfile1, *newfile2;
1346 int flags;
1347
1348 flags = type & ~SOCK_TYPE_MASK;
1349 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1350 return -EINVAL;
1351 type &= SOCK_TYPE_MASK;
1352
1353 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1354 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1355
1356 /*
1357 * Obtain the first socket and check if the underlying protocol
1358 * supports the socketpair call.
1359 */
1360
1361 err = sock_create(family, type, protocol, &sock1);
1362 if (err < 0)
1363 goto out;
1364
1365 err = sock_create(family, type, protocol, &sock2);
1366 if (err < 0)
1367 goto out_release_1;
1368
1369 err = sock1->ops->socketpair(sock1, sock2);
1370 if (err < 0)
1371 goto out_release_both;
1372
1373 fd1 = get_unused_fd_flags(flags);
1374 if (unlikely(fd1 < 0)) {
1375 err = fd1;
1376 goto out_release_both;
1377 }
1378
1379 fd2 = get_unused_fd_flags(flags);
1380 if (unlikely(fd2 < 0)) {
1381 err = fd2;
1382 goto out_put_unused_1;
1383 }
1384
1385 newfile1 = sock_alloc_file(sock1, flags, NULL);
1386 if (IS_ERR(newfile1)) {
1387 err = PTR_ERR(newfile1);
1388 goto out_put_unused_both;
1389 }
1390
1391 newfile2 = sock_alloc_file(sock2, flags, NULL);
1392 if (IS_ERR(newfile2)) {
1393 err = PTR_ERR(newfile2);
1394 goto out_fput_1;
1395 }
1396
1397 err = put_user(fd1, &usockvec[0]);
1398 if (err)
1399 goto out_fput_both;
1400
1401 err = put_user(fd2, &usockvec[1]);
1402 if (err)
1403 goto out_fput_both;
1404
1405 audit_fd_pair(fd1, fd2);
1406
1407 fd_install(fd1, newfile1);
1408 fd_install(fd2, newfile2);
1409 /* fd1 and fd2 may be already another descriptors.
1410 * Not kernel problem.
1411 */
1412
1413 return 0;
1414
1415 out_fput_both:
1416 fput(newfile2);
1417 fput(newfile1);
1418 put_unused_fd(fd2);
1419 put_unused_fd(fd1);
1420 goto out;
1421
1422 out_fput_1:
1423 fput(newfile1);
1424 put_unused_fd(fd2);
1425 put_unused_fd(fd1);
1426 sock_release(sock2);
1427 goto out;
1428
1429 out_put_unused_both:
1430 put_unused_fd(fd2);
1431 out_put_unused_1:
1432 put_unused_fd(fd1);
1433 out_release_both:
1434 sock_release(sock2);
1435 out_release_1:
1436 sock_release(sock1);
1437 out:
1438 return err;
1439 }
1440
1441 /*
1442 * Bind a name to a socket. Nothing much to do here since it's
1443 * the protocol's responsibility to handle the local address.
1444 *
1445 * We move the socket address to kernel space before we call
1446 * the protocol layer (having also checked the address is ok).
1447 */
1448
1449 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1450 {
1451 struct socket *sock;
1452 struct sockaddr_storage address;
1453 int err, fput_needed;
1454
1455 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1456 if (sock) {
1457 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1458 if (err >= 0) {
1459 err = security_socket_bind(sock,
1460 (struct sockaddr *)&address,
1461 addrlen);
1462 if (!err)
1463 err = sock->ops->bind(sock,
1464 (struct sockaddr *)
1465 &address, addrlen);
1466 }
1467 fput_light(sock->file, fput_needed);
1468 }
1469 return err;
1470 }
1471
1472 /*
1473 * Perform a listen. Basically, we allow the protocol to do anything
1474 * necessary for a listen, and if that works, we mark the socket as
1475 * ready for listening.
1476 */
1477
1478 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1479 {
1480 struct socket *sock;
1481 int err, fput_needed;
1482 int somaxconn;
1483
1484 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1485 if (sock) {
1486 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1487 if ((unsigned int)backlog > somaxconn)
1488 backlog = somaxconn;
1489
1490 err = security_socket_listen(sock, backlog);
1491 if (!err)
1492 err = sock->ops->listen(sock, backlog);
1493
1494 fput_light(sock->file, fput_needed);
1495 }
1496 return err;
1497 }
1498
1499 /*
1500 * For accept, we attempt to create a new socket, set up the link
1501 * with the client, wake up the client, then return the new
1502 * connected fd. We collect the address of the connector in kernel
1503 * space and move it to user at the very end. This is unclean because
1504 * we open the socket then return an error.
1505 *
1506 * 1003.1g adds the ability to recvmsg() to query connection pending
1507 * status to recvmsg. We need to add that support in a way thats
1508 * clean when we restucture accept also.
1509 */
1510
1511 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1512 int __user *, upeer_addrlen, int, flags)
1513 {
1514 struct socket *sock, *newsock;
1515 struct file *newfile;
1516 int err, len, newfd, fput_needed;
1517 struct sockaddr_storage address;
1518
1519 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1520 return -EINVAL;
1521
1522 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1523 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1524
1525 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1526 if (!sock)
1527 goto out;
1528
1529 err = -ENFILE;
1530 newsock = sock_alloc();
1531 if (!newsock)
1532 goto out_put;
1533
1534 newsock->type = sock->type;
1535 newsock->ops = sock->ops;
1536
1537 /*
1538 * We don't need try_module_get here, as the listening socket (sock)
1539 * has the protocol module (sock->ops->owner) held.
1540 */
1541 __module_get(newsock->ops->owner);
1542
1543 newfd = get_unused_fd_flags(flags);
1544 if (unlikely(newfd < 0)) {
1545 err = newfd;
1546 sock_release(newsock);
1547 goto out_put;
1548 }
1549 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1550 if (IS_ERR(newfile)) {
1551 err = PTR_ERR(newfile);
1552 put_unused_fd(newfd);
1553 sock_release(newsock);
1554 goto out_put;
1555 }
1556
1557 err = security_socket_accept(sock, newsock);
1558 if (err)
1559 goto out_fd;
1560
1561 err = sock->ops->accept(sock, newsock, sock->file->f_flags, false);
1562 if (err < 0)
1563 goto out_fd;
1564
1565 if (upeer_sockaddr) {
1566 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1567 &len, 2) < 0) {
1568 err = -ECONNABORTED;
1569 goto out_fd;
1570 }
1571 err = move_addr_to_user(&address,
1572 len, upeer_sockaddr, upeer_addrlen);
1573 if (err < 0)
1574 goto out_fd;
1575 }
1576
1577 /* File flags are not inherited via accept() unlike another OSes. */
1578
1579 fd_install(newfd, newfile);
1580 err = newfd;
1581
1582 out_put:
1583 fput_light(sock->file, fput_needed);
1584 out:
1585 return err;
1586 out_fd:
1587 fput(newfile);
1588 put_unused_fd(newfd);
1589 goto out_put;
1590 }
1591
1592 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1593 int __user *, upeer_addrlen)
1594 {
1595 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1596 }
1597
1598 /*
1599 * Attempt to connect to a socket with the server address. The address
1600 * is in user space so we verify it is OK and move it to kernel space.
1601 *
1602 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1603 * break bindings
1604 *
1605 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1606 * other SEQPACKET protocols that take time to connect() as it doesn't
1607 * include the -EINPROGRESS status for such sockets.
1608 */
1609
1610 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1611 int, addrlen)
1612 {
1613 struct socket *sock;
1614 struct sockaddr_storage address;
1615 int err, fput_needed;
1616
1617 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1618 if (!sock)
1619 goto out;
1620 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1621 if (err < 0)
1622 goto out_put;
1623
1624 err =
1625 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1626 if (err)
1627 goto out_put;
1628
1629 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1630 sock->file->f_flags);
1631 out_put:
1632 fput_light(sock->file, fput_needed);
1633 out:
1634 return err;
1635 }
1636
1637 /*
1638 * Get the local address ('name') of a socket object. Move the obtained
1639 * name to user space.
1640 */
1641
1642 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1643 int __user *, usockaddr_len)
1644 {
1645 struct socket *sock;
1646 struct sockaddr_storage address;
1647 int len, err, fput_needed;
1648
1649 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1650 if (!sock)
1651 goto out;
1652
1653 err = security_socket_getsockname(sock);
1654 if (err)
1655 goto out_put;
1656
1657 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1658 if (err)
1659 goto out_put;
1660 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1661
1662 out_put:
1663 fput_light(sock->file, fput_needed);
1664 out:
1665 return err;
1666 }
1667
1668 /*
1669 * Get the remote address ('name') of a socket object. Move the obtained
1670 * name to user space.
1671 */
1672
1673 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1674 int __user *, usockaddr_len)
1675 {
1676 struct socket *sock;
1677 struct sockaddr_storage address;
1678 int len, err, fput_needed;
1679
1680 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1681 if (sock != NULL) {
1682 err = security_socket_getpeername(sock);
1683 if (err) {
1684 fput_light(sock->file, fput_needed);
1685 return err;
1686 }
1687
1688 err =
1689 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1690 1);
1691 if (!err)
1692 err = move_addr_to_user(&address, len, usockaddr,
1693 usockaddr_len);
1694 fput_light(sock->file, fput_needed);
1695 }
1696 return err;
1697 }
1698
1699 /*
1700 * Send a datagram to a given address. We move the address into kernel
1701 * space and check the user space data area is readable before invoking
1702 * the protocol.
1703 */
1704
1705 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1706 unsigned int, flags, struct sockaddr __user *, addr,
1707 int, addr_len)
1708 {
1709 struct socket *sock;
1710 struct sockaddr_storage address;
1711 int err;
1712 struct msghdr msg;
1713 struct iovec iov;
1714 int fput_needed;
1715
1716 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1717 if (unlikely(err))
1718 return err;
1719 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1720 if (!sock)
1721 goto out;
1722
1723 msg.msg_name = NULL;
1724 msg.msg_control = NULL;
1725 msg.msg_controllen = 0;
1726 msg.msg_namelen = 0;
1727 if (addr) {
1728 err = move_addr_to_kernel(addr, addr_len, &address);
1729 if (err < 0)
1730 goto out_put;
1731 msg.msg_name = (struct sockaddr *)&address;
1732 msg.msg_namelen = addr_len;
1733 }
1734 if (sock->file->f_flags & O_NONBLOCK)
1735 flags |= MSG_DONTWAIT;
1736 msg.msg_flags = flags;
1737 err = sock_sendmsg(sock, &msg);
1738
1739 out_put:
1740 fput_light(sock->file, fput_needed);
1741 out:
1742 return err;
1743 }
1744
1745 /*
1746 * Send a datagram down a socket.
1747 */
1748
1749 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1750 unsigned int, flags)
1751 {
1752 return sys_sendto(fd, buff, len, flags, NULL, 0);
1753 }
1754
1755 /*
1756 * Receive a frame from the socket and optionally record the address of the
1757 * sender. We verify the buffers are writable and if needed move the
1758 * sender address from kernel to user space.
1759 */
1760
1761 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1762 unsigned int, flags, struct sockaddr __user *, addr,
1763 int __user *, addr_len)
1764 {
1765 struct socket *sock;
1766 struct iovec iov;
1767 struct msghdr msg;
1768 struct sockaddr_storage address;
1769 int err, err2;
1770 int fput_needed;
1771
1772 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1773 if (unlikely(err))
1774 return err;
1775 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1776 if (!sock)
1777 goto out;
1778
1779 msg.msg_control = NULL;
1780 msg.msg_controllen = 0;
1781 /* Save some cycles and don't copy the address if not needed */
1782 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1783 /* We assume all kernel code knows the size of sockaddr_storage */
1784 msg.msg_namelen = 0;
1785 msg.msg_iocb = NULL;
1786 msg.msg_flags = 0;
1787 if (sock->file->f_flags & O_NONBLOCK)
1788 flags |= MSG_DONTWAIT;
1789 err = sock_recvmsg(sock, &msg, flags);
1790
1791 if (err >= 0 && addr != NULL) {
1792 err2 = move_addr_to_user(&address,
1793 msg.msg_namelen, addr, addr_len);
1794 if (err2 < 0)
1795 err = err2;
1796 }
1797
1798 fput_light(sock->file, fput_needed);
1799 out:
1800 return err;
1801 }
1802
1803 /*
1804 * Receive a datagram from a socket.
1805 */
1806
1807 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1808 unsigned int, flags)
1809 {
1810 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1811 }
1812
1813 /*
1814 * Set a socket option. Because we don't know the option lengths we have
1815 * to pass the user mode parameter for the protocols to sort out.
1816 */
1817
1818 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1819 char __user *, optval, int, optlen)
1820 {
1821 int err, fput_needed;
1822 struct socket *sock;
1823
1824 if (optlen < 0)
1825 return -EINVAL;
1826
1827 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1828 if (sock != NULL) {
1829 err = security_socket_setsockopt(sock, level, optname);
1830 if (err)
1831 goto out_put;
1832
1833 if (level == SOL_SOCKET)
1834 err =
1835 sock_setsockopt(sock, level, optname, optval,
1836 optlen);
1837 else
1838 err =
1839 sock->ops->setsockopt(sock, level, optname, optval,
1840 optlen);
1841 out_put:
1842 fput_light(sock->file, fput_needed);
1843 }
1844 return err;
1845 }
1846
1847 /*
1848 * Get a socket option. Because we don't know the option lengths we have
1849 * to pass a user mode parameter for the protocols to sort out.
1850 */
1851
1852 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1853 char __user *, optval, int __user *, optlen)
1854 {
1855 int err, fput_needed;
1856 struct socket *sock;
1857
1858 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1859 if (sock != NULL) {
1860 err = security_socket_getsockopt(sock, level, optname);
1861 if (err)
1862 goto out_put;
1863
1864 if (level == SOL_SOCKET)
1865 err =
1866 sock_getsockopt(sock, level, optname, optval,
1867 optlen);
1868 else
1869 err =
1870 sock->ops->getsockopt(sock, level, optname, optval,
1871 optlen);
1872 out_put:
1873 fput_light(sock->file, fput_needed);
1874 }
1875 return err;
1876 }
1877
1878 /*
1879 * Shutdown a socket.
1880 */
1881
1882 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1883 {
1884 int err, fput_needed;
1885 struct socket *sock;
1886
1887 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1888 if (sock != NULL) {
1889 err = security_socket_shutdown(sock, how);
1890 if (!err)
1891 err = sock->ops->shutdown(sock, how);
1892 fput_light(sock->file, fput_needed);
1893 }
1894 return err;
1895 }
1896
1897 /* A couple of helpful macros for getting the address of the 32/64 bit
1898 * fields which are the same type (int / unsigned) on our platforms.
1899 */
1900 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1901 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1902 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1903
1904 struct used_address {
1905 struct sockaddr_storage name;
1906 unsigned int name_len;
1907 };
1908
1909 static int copy_msghdr_from_user(struct msghdr *kmsg,
1910 struct user_msghdr __user *umsg,
1911 struct sockaddr __user **save_addr,
1912 struct iovec **iov)
1913 {
1914 struct sockaddr __user *uaddr;
1915 struct iovec __user *uiov;
1916 size_t nr_segs;
1917 ssize_t err;
1918
1919 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1920 __get_user(uaddr, &umsg->msg_name) ||
1921 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1922 __get_user(uiov, &umsg->msg_iov) ||
1923 __get_user(nr_segs, &umsg->msg_iovlen) ||
1924 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1925 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1926 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1927 return -EFAULT;
1928
1929 if (!uaddr)
1930 kmsg->msg_namelen = 0;
1931
1932 if (kmsg->msg_namelen < 0)
1933 return -EINVAL;
1934
1935 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1936 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1937
1938 if (save_addr)
1939 *save_addr = uaddr;
1940
1941 if (uaddr && kmsg->msg_namelen) {
1942 if (!save_addr) {
1943 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1944 kmsg->msg_name);
1945 if (err < 0)
1946 return err;
1947 }
1948 } else {
1949 kmsg->msg_name = NULL;
1950 kmsg->msg_namelen = 0;
1951 }
1952
1953 if (nr_segs > UIO_MAXIOV)
1954 return -EMSGSIZE;
1955
1956 kmsg->msg_iocb = NULL;
1957
1958 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1959 UIO_FASTIOV, iov, &kmsg->msg_iter);
1960 }
1961
1962 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1963 struct msghdr *msg_sys, unsigned int flags,
1964 struct used_address *used_address,
1965 unsigned int allowed_msghdr_flags)
1966 {
1967 struct compat_msghdr __user *msg_compat =
1968 (struct compat_msghdr __user *)msg;
1969 struct sockaddr_storage address;
1970 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1971 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1972 __aligned(sizeof(__kernel_size_t));
1973 /* 20 is size of ipv6_pktinfo */
1974 unsigned char *ctl_buf = ctl;
1975 int ctl_len;
1976 ssize_t err;
1977
1978 msg_sys->msg_name = &address;
1979
1980 if (MSG_CMSG_COMPAT & flags)
1981 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1982 else
1983 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1984 if (err < 0)
1985 return err;
1986
1987 err = -ENOBUFS;
1988
1989 if (msg_sys->msg_controllen > INT_MAX)
1990 goto out_freeiov;
1991 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1992 ctl_len = msg_sys->msg_controllen;
1993 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1994 err =
1995 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1996 sizeof(ctl));
1997 if (err)
1998 goto out_freeiov;
1999 ctl_buf = msg_sys->msg_control;
2000 ctl_len = msg_sys->msg_controllen;
2001 } else if (ctl_len) {
2002 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
2003 CMSG_ALIGN(sizeof(struct cmsghdr)));
2004 if (ctl_len > sizeof(ctl)) {
2005 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
2006 if (ctl_buf == NULL)
2007 goto out_freeiov;
2008 }
2009 err = -EFAULT;
2010 /*
2011 * Careful! Before this, msg_sys->msg_control contains a user pointer.
2012 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
2013 * checking falls down on this.
2014 */
2015 if (copy_from_user(ctl_buf,
2016 (void __user __force *)msg_sys->msg_control,
2017 ctl_len))
2018 goto out_freectl;
2019 msg_sys->msg_control = ctl_buf;
2020 }
2021 msg_sys->msg_flags = flags;
2022
2023 if (sock->file->f_flags & O_NONBLOCK)
2024 msg_sys->msg_flags |= MSG_DONTWAIT;
2025 /*
2026 * If this is sendmmsg() and current destination address is same as
2027 * previously succeeded address, omit asking LSM's decision.
2028 * used_address->name_len is initialized to UINT_MAX so that the first
2029 * destination address never matches.
2030 */
2031 if (used_address && msg_sys->msg_name &&
2032 used_address->name_len == msg_sys->msg_namelen &&
2033 !memcmp(&used_address->name, msg_sys->msg_name,
2034 used_address->name_len)) {
2035 err = sock_sendmsg_nosec(sock, msg_sys);
2036 goto out_freectl;
2037 }
2038 err = sock_sendmsg(sock, msg_sys);
2039 /*
2040 * If this is sendmmsg() and sending to current destination address was
2041 * successful, remember it.
2042 */
2043 if (used_address && err >= 0) {
2044 used_address->name_len = msg_sys->msg_namelen;
2045 if (msg_sys->msg_name)
2046 memcpy(&used_address->name, msg_sys->msg_name,
2047 used_address->name_len);
2048 }
2049
2050 out_freectl:
2051 if (ctl_buf != ctl)
2052 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2053 out_freeiov:
2054 kfree(iov);
2055 return err;
2056 }
2057
2058 /*
2059 * BSD sendmsg interface
2060 */
2061
2062 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2063 {
2064 int fput_needed, err;
2065 struct msghdr msg_sys;
2066 struct socket *sock;
2067
2068 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2069 if (!sock)
2070 goto out;
2071
2072 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2073
2074 fput_light(sock->file, fput_needed);
2075 out:
2076 return err;
2077 }
2078
2079 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2080 {
2081 if (flags & MSG_CMSG_COMPAT)
2082 return -EINVAL;
2083 return __sys_sendmsg(fd, msg, flags);
2084 }
2085
2086 /*
2087 * Linux sendmmsg interface
2088 */
2089
2090 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2091 unsigned int flags)
2092 {
2093 int fput_needed, err, datagrams;
2094 struct socket *sock;
2095 struct mmsghdr __user *entry;
2096 struct compat_mmsghdr __user *compat_entry;
2097 struct msghdr msg_sys;
2098 struct used_address used_address;
2099 unsigned int oflags = flags;
2100
2101 if (vlen > UIO_MAXIOV)
2102 vlen = UIO_MAXIOV;
2103
2104 datagrams = 0;
2105
2106 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2107 if (!sock)
2108 return err;
2109
2110 used_address.name_len = UINT_MAX;
2111 entry = mmsg;
2112 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2113 err = 0;
2114 flags |= MSG_BATCH;
2115
2116 while (datagrams < vlen) {
2117 if (datagrams == vlen - 1)
2118 flags = oflags;
2119
2120 if (MSG_CMSG_COMPAT & flags) {
2121 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2122 &msg_sys, flags, &used_address, MSG_EOR);
2123 if (err < 0)
2124 break;
2125 err = __put_user(err, &compat_entry->msg_len);
2126 ++compat_entry;
2127 } else {
2128 err = ___sys_sendmsg(sock,
2129 (struct user_msghdr __user *)entry,
2130 &msg_sys, flags, &used_address, MSG_EOR);
2131 if (err < 0)
2132 break;
2133 err = put_user(err, &entry->msg_len);
2134 ++entry;
2135 }
2136
2137 if (err)
2138 break;
2139 ++datagrams;
2140 if (msg_data_left(&msg_sys))
2141 break;
2142 cond_resched();
2143 }
2144
2145 fput_light(sock->file, fput_needed);
2146
2147 /* We only return an error if no datagrams were able to be sent */
2148 if (datagrams != 0)
2149 return datagrams;
2150
2151 return err;
2152 }
2153
2154 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2155 unsigned int, vlen, unsigned int, flags)
2156 {
2157 if (flags & MSG_CMSG_COMPAT)
2158 return -EINVAL;
2159 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2160 }
2161
2162 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2163 struct msghdr *msg_sys, unsigned int flags, int nosec)
2164 {
2165 struct compat_msghdr __user *msg_compat =
2166 (struct compat_msghdr __user *)msg;
2167 struct iovec iovstack[UIO_FASTIOV];
2168 struct iovec *iov = iovstack;
2169 unsigned long cmsg_ptr;
2170 int len;
2171 ssize_t err;
2172
2173 /* kernel mode address */
2174 struct sockaddr_storage addr;
2175
2176 /* user mode address pointers */
2177 struct sockaddr __user *uaddr;
2178 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2179
2180 msg_sys->msg_name = &addr;
2181
2182 if (MSG_CMSG_COMPAT & flags)
2183 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2184 else
2185 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2186 if (err < 0)
2187 return err;
2188
2189 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2190 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2191
2192 /* We assume all kernel code knows the size of sockaddr_storage */
2193 msg_sys->msg_namelen = 0;
2194
2195 if (sock->file->f_flags & O_NONBLOCK)
2196 flags |= MSG_DONTWAIT;
2197 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2198 if (err < 0)
2199 goto out_freeiov;
2200 len = err;
2201
2202 if (uaddr != NULL) {
2203 err = move_addr_to_user(&addr,
2204 msg_sys->msg_namelen, uaddr,
2205 uaddr_len);
2206 if (err < 0)
2207 goto out_freeiov;
2208 }
2209 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2210 COMPAT_FLAGS(msg));
2211 if (err)
2212 goto out_freeiov;
2213 if (MSG_CMSG_COMPAT & flags)
2214 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2215 &msg_compat->msg_controllen);
2216 else
2217 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2218 &msg->msg_controllen);
2219 if (err)
2220 goto out_freeiov;
2221 err = len;
2222
2223 out_freeiov:
2224 kfree(iov);
2225 return err;
2226 }
2227
2228 /*
2229 * BSD recvmsg interface
2230 */
2231
2232 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2233 {
2234 int fput_needed, err;
2235 struct msghdr msg_sys;
2236 struct socket *sock;
2237
2238 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2239 if (!sock)
2240 goto out;
2241
2242 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2243
2244 fput_light(sock->file, fput_needed);
2245 out:
2246 return err;
2247 }
2248
2249 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2250 unsigned int, flags)
2251 {
2252 if (flags & MSG_CMSG_COMPAT)
2253 return -EINVAL;
2254 return __sys_recvmsg(fd, msg, flags);
2255 }
2256
2257 /*
2258 * Linux recvmmsg interface
2259 */
2260
2261 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2262 unsigned int flags, struct timespec *timeout)
2263 {
2264 int fput_needed, err, datagrams;
2265 struct socket *sock;
2266 struct mmsghdr __user *entry;
2267 struct compat_mmsghdr __user *compat_entry;
2268 struct msghdr msg_sys;
2269 struct timespec64 end_time;
2270 struct timespec64 timeout64;
2271
2272 if (timeout &&
2273 poll_select_set_timeout(&end_time, timeout->tv_sec,
2274 timeout->tv_nsec))
2275 return -EINVAL;
2276
2277 datagrams = 0;
2278
2279 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2280 if (!sock)
2281 return err;
2282
2283 err = sock_error(sock->sk);
2284 if (err) {
2285 datagrams = err;
2286 goto out_put;
2287 }
2288
2289 entry = mmsg;
2290 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2291
2292 while (datagrams < vlen) {
2293 /*
2294 * No need to ask LSM for more than the first datagram.
2295 */
2296 if (MSG_CMSG_COMPAT & flags) {
2297 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2298 &msg_sys, flags & ~MSG_WAITFORONE,
2299 datagrams);
2300 if (err < 0)
2301 break;
2302 err = __put_user(err, &compat_entry->msg_len);
2303 ++compat_entry;
2304 } else {
2305 err = ___sys_recvmsg(sock,
2306 (struct user_msghdr __user *)entry,
2307 &msg_sys, flags & ~MSG_WAITFORONE,
2308 datagrams);
2309 if (err < 0)
2310 break;
2311 err = put_user(err, &entry->msg_len);
2312 ++entry;
2313 }
2314
2315 if (err)
2316 break;
2317 ++datagrams;
2318
2319 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2320 if (flags & MSG_WAITFORONE)
2321 flags |= MSG_DONTWAIT;
2322
2323 if (timeout) {
2324 ktime_get_ts64(&timeout64);
2325 *timeout = timespec64_to_timespec(
2326 timespec64_sub(end_time, timeout64));
2327 if (timeout->tv_sec < 0) {
2328 timeout->tv_sec = timeout->tv_nsec = 0;
2329 break;
2330 }
2331
2332 /* Timeout, return less than vlen datagrams */
2333 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2334 break;
2335 }
2336
2337 /* Out of band data, return right away */
2338 if (msg_sys.msg_flags & MSG_OOB)
2339 break;
2340 cond_resched();
2341 }
2342
2343 if (err == 0)
2344 goto out_put;
2345
2346 if (datagrams == 0) {
2347 datagrams = err;
2348 goto out_put;
2349 }
2350
2351 /*
2352 * We may return less entries than requested (vlen) if the
2353 * sock is non block and there aren't enough datagrams...
2354 */
2355 if (err != -EAGAIN) {
2356 /*
2357 * ... or if recvmsg returns an error after we
2358 * received some datagrams, where we record the
2359 * error to return on the next call or if the
2360 * app asks about it using getsockopt(SO_ERROR).
2361 */
2362 sock->sk->sk_err = -err;
2363 }
2364 out_put:
2365 fput_light(sock->file, fput_needed);
2366
2367 return datagrams;
2368 }
2369
2370 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2371 unsigned int, vlen, unsigned int, flags,
2372 struct timespec __user *, timeout)
2373 {
2374 int datagrams;
2375 struct timespec timeout_sys;
2376
2377 if (flags & MSG_CMSG_COMPAT)
2378 return -EINVAL;
2379
2380 if (!timeout)
2381 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2382
2383 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2384 return -EFAULT;
2385
2386 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2387
2388 if (datagrams > 0 &&
2389 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2390 datagrams = -EFAULT;
2391
2392 return datagrams;
2393 }
2394
2395 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2396 /* Argument list sizes for sys_socketcall */
2397 #define AL(x) ((x) * sizeof(unsigned long))
2398 static const unsigned char nargs[21] = {
2399 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2400 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2401 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2402 AL(4), AL(5), AL(4)
2403 };
2404
2405 #undef AL
2406
2407 /*
2408 * System call vectors.
2409 *
2410 * Argument checking cleaned up. Saved 20% in size.
2411 * This function doesn't need to set the kernel lock because
2412 * it is set by the callees.
2413 */
2414
2415 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2416 {
2417 unsigned long a[AUDITSC_ARGS];
2418 unsigned long a0, a1;
2419 int err;
2420 unsigned int len;
2421
2422 if (call < 1 || call > SYS_SENDMMSG)
2423 return -EINVAL;
2424
2425 len = nargs[call];
2426 if (len > sizeof(a))
2427 return -EINVAL;
2428
2429 /* copy_from_user should be SMP safe. */
2430 if (copy_from_user(a, args, len))
2431 return -EFAULT;
2432
2433 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2434 if (err)
2435 return err;
2436
2437 a0 = a[0];
2438 a1 = a[1];
2439
2440 switch (call) {
2441 case SYS_SOCKET:
2442 err = sys_socket(a0, a1, a[2]);
2443 break;
2444 case SYS_BIND:
2445 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2446 break;
2447 case SYS_CONNECT:
2448 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2449 break;
2450 case SYS_LISTEN:
2451 err = sys_listen(a0, a1);
2452 break;
2453 case SYS_ACCEPT:
2454 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2455 (int __user *)a[2], 0);
2456 break;
2457 case SYS_GETSOCKNAME:
2458 err =
2459 sys_getsockname(a0, (struct sockaddr __user *)a1,
2460 (int __user *)a[2]);
2461 break;
2462 case SYS_GETPEERNAME:
2463 err =
2464 sys_getpeername(a0, (struct sockaddr __user *)a1,
2465 (int __user *)a[2]);
2466 break;
2467 case SYS_SOCKETPAIR:
2468 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2469 break;
2470 case SYS_SEND:
2471 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2472 break;
2473 case SYS_SENDTO:
2474 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2475 (struct sockaddr __user *)a[4], a[5]);
2476 break;
2477 case SYS_RECV:
2478 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2479 break;
2480 case SYS_RECVFROM:
2481 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2482 (struct sockaddr __user *)a[4],
2483 (int __user *)a[5]);
2484 break;
2485 case SYS_SHUTDOWN:
2486 err = sys_shutdown(a0, a1);
2487 break;
2488 case SYS_SETSOCKOPT:
2489 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2490 break;
2491 case SYS_GETSOCKOPT:
2492 err =
2493 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2494 (int __user *)a[4]);
2495 break;
2496 case SYS_SENDMSG:
2497 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2498 break;
2499 case SYS_SENDMMSG:
2500 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2501 break;
2502 case SYS_RECVMSG:
2503 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2504 break;
2505 case SYS_RECVMMSG:
2506 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2507 (struct timespec __user *)a[4]);
2508 break;
2509 case SYS_ACCEPT4:
2510 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2511 (int __user *)a[2], a[3]);
2512 break;
2513 default:
2514 err = -EINVAL;
2515 break;
2516 }
2517 return err;
2518 }
2519
2520 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2521
2522 /**
2523 * sock_register - add a socket protocol handler
2524 * @ops: description of protocol
2525 *
2526 * This function is called by a protocol handler that wants to
2527 * advertise its address family, and have it linked into the
2528 * socket interface. The value ops->family corresponds to the
2529 * socket system call protocol family.
2530 */
2531 int sock_register(const struct net_proto_family *ops)
2532 {
2533 int err;
2534
2535 if (ops->family >= NPROTO) {
2536 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2537 return -ENOBUFS;
2538 }
2539
2540 spin_lock(&net_family_lock);
2541 if (rcu_dereference_protected(net_families[ops->family],
2542 lockdep_is_held(&net_family_lock)))
2543 err = -EEXIST;
2544 else {
2545 rcu_assign_pointer(net_families[ops->family], ops);
2546 err = 0;
2547 }
2548 spin_unlock(&net_family_lock);
2549
2550 pr_info("NET: Registered protocol family %d\n", ops->family);
2551 return err;
2552 }
2553 EXPORT_SYMBOL(sock_register);
2554
2555 /**
2556 * sock_unregister - remove a protocol handler
2557 * @family: protocol family to remove
2558 *
2559 * This function is called by a protocol handler that wants to
2560 * remove its address family, and have it unlinked from the
2561 * new socket creation.
2562 *
2563 * If protocol handler is a module, then it can use module reference
2564 * counts to protect against new references. If protocol handler is not
2565 * a module then it needs to provide its own protection in
2566 * the ops->create routine.
2567 */
2568 void sock_unregister(int family)
2569 {
2570 BUG_ON(family < 0 || family >= NPROTO);
2571
2572 spin_lock(&net_family_lock);
2573 RCU_INIT_POINTER(net_families[family], NULL);
2574 spin_unlock(&net_family_lock);
2575
2576 synchronize_rcu();
2577
2578 pr_info("NET: Unregistered protocol family %d\n", family);
2579 }
2580 EXPORT_SYMBOL(sock_unregister);
2581
2582 static int __init sock_init(void)
2583 {
2584 int err;
2585 /*
2586 * Initialize the network sysctl infrastructure.
2587 */
2588 err = net_sysctl_init();
2589 if (err)
2590 goto out;
2591
2592 /*
2593 * Initialize skbuff SLAB cache
2594 */
2595 skb_init();
2596
2597 /*
2598 * Initialize the protocols module.
2599 */
2600
2601 init_inodecache();
2602
2603 err = register_filesystem(&sock_fs_type);
2604 if (err)
2605 goto out_fs;
2606 sock_mnt = kern_mount(&sock_fs_type);
2607 if (IS_ERR(sock_mnt)) {
2608 err = PTR_ERR(sock_mnt);
2609 goto out_mount;
2610 }
2611
2612 /* The real protocol initialization is performed in later initcalls.
2613 */
2614
2615 #ifdef CONFIG_NETFILTER
2616 err = netfilter_init();
2617 if (err)
2618 goto out;
2619 #endif
2620
2621 ptp_classifier_init();
2622
2623 out:
2624 return err;
2625
2626 out_mount:
2627 unregister_filesystem(&sock_fs_type);
2628 out_fs:
2629 goto out;
2630 }
2631
2632 core_initcall(sock_init); /* early initcall */
2633
2634 #ifdef CONFIG_PROC_FS
2635 void socket_seq_show(struct seq_file *seq)
2636 {
2637 int cpu;
2638 int counter = 0;
2639
2640 for_each_possible_cpu(cpu)
2641 counter += per_cpu(sockets_in_use, cpu);
2642
2643 /* It can be negative, by the way. 8) */
2644 if (counter < 0)
2645 counter = 0;
2646
2647 seq_printf(seq, "sockets: used %d\n", counter);
2648 }
2649 #endif /* CONFIG_PROC_FS */
2650
2651 #ifdef CONFIG_COMPAT
2652 static int do_siocgstamp(struct net *net, struct socket *sock,
2653 unsigned int cmd, void __user *up)
2654 {
2655 mm_segment_t old_fs = get_fs();
2656 struct timeval ktv;
2657 int err;
2658
2659 set_fs(KERNEL_DS);
2660 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2661 set_fs(old_fs);
2662 if (!err)
2663 err = compat_put_timeval(&ktv, up);
2664
2665 return err;
2666 }
2667
2668 static int do_siocgstampns(struct net *net, struct socket *sock,
2669 unsigned int cmd, void __user *up)
2670 {
2671 mm_segment_t old_fs = get_fs();
2672 struct timespec kts;
2673 int err;
2674
2675 set_fs(KERNEL_DS);
2676 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2677 set_fs(old_fs);
2678 if (!err)
2679 err = compat_put_timespec(&kts, up);
2680
2681 return err;
2682 }
2683
2684 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2685 {
2686 struct ifreq __user *uifr;
2687 int err;
2688
2689 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2690 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2691 return -EFAULT;
2692
2693 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2694 if (err)
2695 return err;
2696
2697 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2698 return -EFAULT;
2699
2700 return 0;
2701 }
2702
2703 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2704 {
2705 struct compat_ifconf ifc32;
2706 struct ifconf ifc;
2707 struct ifconf __user *uifc;
2708 struct compat_ifreq __user *ifr32;
2709 struct ifreq __user *ifr;
2710 unsigned int i, j;
2711 int err;
2712
2713 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2714 return -EFAULT;
2715
2716 memset(&ifc, 0, sizeof(ifc));
2717 if (ifc32.ifcbuf == 0) {
2718 ifc32.ifc_len = 0;
2719 ifc.ifc_len = 0;
2720 ifc.ifc_req = NULL;
2721 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2722 } else {
2723 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2724 sizeof(struct ifreq);
2725 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2726 ifc.ifc_len = len;
2727 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2728 ifr32 = compat_ptr(ifc32.ifcbuf);
2729 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2730 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2731 return -EFAULT;
2732 ifr++;
2733 ifr32++;
2734 }
2735 }
2736 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2737 return -EFAULT;
2738
2739 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2740 if (err)
2741 return err;
2742
2743 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2744 return -EFAULT;
2745
2746 ifr = ifc.ifc_req;
2747 ifr32 = compat_ptr(ifc32.ifcbuf);
2748 for (i = 0, j = 0;
2749 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2750 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2751 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2752 return -EFAULT;
2753 ifr32++;
2754 ifr++;
2755 }
2756
2757 if (ifc32.ifcbuf == 0) {
2758 /* Translate from 64-bit structure multiple to
2759 * a 32-bit one.
2760 */
2761 i = ifc.ifc_len;
2762 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2763 ifc32.ifc_len = i;
2764 } else {
2765 ifc32.ifc_len = i;
2766 }
2767 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2768 return -EFAULT;
2769
2770 return 0;
2771 }
2772
2773 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2774 {
2775 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2776 bool convert_in = false, convert_out = false;
2777 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2778 struct ethtool_rxnfc __user *rxnfc;
2779 struct ifreq __user *ifr;
2780 u32 rule_cnt = 0, actual_rule_cnt;
2781 u32 ethcmd;
2782 u32 data;
2783 int ret;
2784
2785 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2786 return -EFAULT;
2787
2788 compat_rxnfc = compat_ptr(data);
2789
2790 if (get_user(ethcmd, &compat_rxnfc->cmd))
2791 return -EFAULT;
2792
2793 /* Most ethtool structures are defined without padding.
2794 * Unfortunately struct ethtool_rxnfc is an exception.
2795 */
2796 switch (ethcmd) {
2797 default:
2798 break;
2799 case ETHTOOL_GRXCLSRLALL:
2800 /* Buffer size is variable */
2801 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2802 return -EFAULT;
2803 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2804 return -ENOMEM;
2805 buf_size += rule_cnt * sizeof(u32);
2806 /* fall through */
2807 case ETHTOOL_GRXRINGS:
2808 case ETHTOOL_GRXCLSRLCNT:
2809 case ETHTOOL_GRXCLSRULE:
2810 case ETHTOOL_SRXCLSRLINS:
2811 convert_out = true;
2812 /* fall through */
2813 case ETHTOOL_SRXCLSRLDEL:
2814 buf_size += sizeof(struct ethtool_rxnfc);
2815 convert_in = true;
2816 break;
2817 }
2818
2819 ifr = compat_alloc_user_space(buf_size);
2820 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2821
2822 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2823 return -EFAULT;
2824
2825 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2826 &ifr->ifr_ifru.ifru_data))
2827 return -EFAULT;
2828
2829 if (convert_in) {
2830 /* We expect there to be holes between fs.m_ext and
2831 * fs.ring_cookie and at the end of fs, but nowhere else.
2832 */
2833 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2834 sizeof(compat_rxnfc->fs.m_ext) !=
2835 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2836 sizeof(rxnfc->fs.m_ext));
2837 BUILD_BUG_ON(
2838 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2839 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2840 offsetof(struct ethtool_rxnfc, fs.location) -
2841 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2842
2843 if (copy_in_user(rxnfc, compat_rxnfc,
2844 (void __user *)(&rxnfc->fs.m_ext + 1) -
2845 (void __user *)rxnfc) ||
2846 copy_in_user(&rxnfc->fs.ring_cookie,
2847 &compat_rxnfc->fs.ring_cookie,
2848 (void __user *)(&rxnfc->fs.location + 1) -
2849 (void __user *)&rxnfc->fs.ring_cookie) ||
2850 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2851 sizeof(rxnfc->rule_cnt)))
2852 return -EFAULT;
2853 }
2854
2855 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2856 if (ret)
2857 return ret;
2858
2859 if (convert_out) {
2860 if (copy_in_user(compat_rxnfc, rxnfc,
2861 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2862 (const void __user *)rxnfc) ||
2863 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2864 &rxnfc->fs.ring_cookie,
2865 (const void __user *)(&rxnfc->fs.location + 1) -
2866 (const void __user *)&rxnfc->fs.ring_cookie) ||
2867 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2868 sizeof(rxnfc->rule_cnt)))
2869 return -EFAULT;
2870
2871 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2872 /* As an optimisation, we only copy the actual
2873 * number of rules that the underlying
2874 * function returned. Since Mallory might
2875 * change the rule count in user memory, we
2876 * check that it is less than the rule count
2877 * originally given (as the user buffer size),
2878 * which has been range-checked.
2879 */
2880 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2881 return -EFAULT;
2882 if (actual_rule_cnt < rule_cnt)
2883 rule_cnt = actual_rule_cnt;
2884 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2885 &rxnfc->rule_locs[0],
2886 rule_cnt * sizeof(u32)))
2887 return -EFAULT;
2888 }
2889 }
2890
2891 return 0;
2892 }
2893
2894 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2895 {
2896 void __user *uptr;
2897 compat_uptr_t uptr32;
2898 struct ifreq __user *uifr;
2899
2900 uifr = compat_alloc_user_space(sizeof(*uifr));
2901 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2902 return -EFAULT;
2903
2904 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2905 return -EFAULT;
2906
2907 uptr = compat_ptr(uptr32);
2908
2909 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2910 return -EFAULT;
2911
2912 return dev_ioctl(net, SIOCWANDEV, uifr);
2913 }
2914
2915 static int bond_ioctl(struct net *net, unsigned int cmd,
2916 struct compat_ifreq __user *ifr32)
2917 {
2918 struct ifreq kifr;
2919 mm_segment_t old_fs;
2920 int err;
2921
2922 switch (cmd) {
2923 case SIOCBONDENSLAVE:
2924 case SIOCBONDRELEASE:
2925 case SIOCBONDSETHWADDR:
2926 case SIOCBONDCHANGEACTIVE:
2927 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2928 return -EFAULT;
2929
2930 old_fs = get_fs();
2931 set_fs(KERNEL_DS);
2932 err = dev_ioctl(net, cmd,
2933 (struct ifreq __user __force *) &kifr);
2934 set_fs(old_fs);
2935
2936 return err;
2937 default:
2938 return -ENOIOCTLCMD;
2939 }
2940 }
2941
2942 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2943 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2944 struct compat_ifreq __user *u_ifreq32)
2945 {
2946 struct ifreq __user *u_ifreq64;
2947 char tmp_buf[IFNAMSIZ];
2948 void __user *data64;
2949 u32 data32;
2950
2951 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2952 IFNAMSIZ))
2953 return -EFAULT;
2954 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2955 return -EFAULT;
2956 data64 = compat_ptr(data32);
2957
2958 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2959
2960 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2961 IFNAMSIZ))
2962 return -EFAULT;
2963 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2964 return -EFAULT;
2965
2966 return dev_ioctl(net, cmd, u_ifreq64);
2967 }
2968
2969 static int dev_ifsioc(struct net *net, struct socket *sock,
2970 unsigned int cmd, struct compat_ifreq __user *uifr32)
2971 {
2972 struct ifreq __user *uifr;
2973 int err;
2974
2975 uifr = compat_alloc_user_space(sizeof(*uifr));
2976 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2977 return -EFAULT;
2978
2979 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2980
2981 if (!err) {
2982 switch (cmd) {
2983 case SIOCGIFFLAGS:
2984 case SIOCGIFMETRIC:
2985 case SIOCGIFMTU:
2986 case SIOCGIFMEM:
2987 case SIOCGIFHWADDR:
2988 case SIOCGIFINDEX:
2989 case SIOCGIFADDR:
2990 case SIOCGIFBRDADDR:
2991 case SIOCGIFDSTADDR:
2992 case SIOCGIFNETMASK:
2993 case SIOCGIFPFLAGS:
2994 case SIOCGIFTXQLEN:
2995 case SIOCGMIIPHY:
2996 case SIOCGMIIREG:
2997 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2998 err = -EFAULT;
2999 break;
3000 }
3001 }
3002 return err;
3003 }
3004
3005 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
3006 struct compat_ifreq __user *uifr32)
3007 {
3008 struct ifreq ifr;
3009 struct compat_ifmap __user *uifmap32;
3010 mm_segment_t old_fs;
3011 int err;
3012
3013 uifmap32 = &uifr32->ifr_ifru.ifru_map;
3014 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
3015 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3016 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3017 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3018 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
3019 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
3020 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
3021 if (err)
3022 return -EFAULT;
3023
3024 old_fs = get_fs();
3025 set_fs(KERNEL_DS);
3026 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
3027 set_fs(old_fs);
3028
3029 if (cmd == SIOCGIFMAP && !err) {
3030 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
3031 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
3032 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
3033 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
3034 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
3035 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
3036 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
3037 if (err)
3038 err = -EFAULT;
3039 }
3040 return err;
3041 }
3042
3043 struct rtentry32 {
3044 u32 rt_pad1;
3045 struct sockaddr rt_dst; /* target address */
3046 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
3047 struct sockaddr rt_genmask; /* target network mask (IP) */
3048 unsigned short rt_flags;
3049 short rt_pad2;
3050 u32 rt_pad3;
3051 unsigned char rt_tos;
3052 unsigned char rt_class;
3053 short rt_pad4;
3054 short rt_metric; /* +1 for binary compatibility! */
3055 /* char * */ u32 rt_dev; /* forcing the device at add */
3056 u32 rt_mtu; /* per route MTU/Window */
3057 u32 rt_window; /* Window clamping */
3058 unsigned short rt_irtt; /* Initial RTT */
3059 };
3060
3061 struct in6_rtmsg32 {
3062 struct in6_addr rtmsg_dst;
3063 struct in6_addr rtmsg_src;
3064 struct in6_addr rtmsg_gateway;
3065 u32 rtmsg_type;
3066 u16 rtmsg_dst_len;
3067 u16 rtmsg_src_len;
3068 u32 rtmsg_metric;
3069 u32 rtmsg_info;
3070 u32 rtmsg_flags;
3071 s32 rtmsg_ifindex;
3072 };
3073
3074 static int routing_ioctl(struct net *net, struct socket *sock,
3075 unsigned int cmd, void __user *argp)
3076 {
3077 int ret;
3078 void *r = NULL;
3079 struct in6_rtmsg r6;
3080 struct rtentry r4;
3081 char devname[16];
3082 u32 rtdev;
3083 mm_segment_t old_fs = get_fs();
3084
3085 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3086 struct in6_rtmsg32 __user *ur6 = argp;
3087 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3088 3 * sizeof(struct in6_addr));
3089 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3090 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3091 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3092 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3093 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3094 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3095 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3096
3097 r = (void *) &r6;
3098 } else { /* ipv4 */
3099 struct rtentry32 __user *ur4 = argp;
3100 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3101 3 * sizeof(struct sockaddr));
3102 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3103 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3104 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3105 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3106 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3107 ret |= get_user(rtdev, &(ur4->rt_dev));
3108 if (rtdev) {
3109 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3110 r4.rt_dev = (char __user __force *)devname;
3111 devname[15] = 0;
3112 } else
3113 r4.rt_dev = NULL;
3114
3115 r = (void *) &r4;
3116 }
3117
3118 if (ret) {
3119 ret = -EFAULT;
3120 goto out;
3121 }
3122
3123 set_fs(KERNEL_DS);
3124 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3125 set_fs(old_fs);
3126
3127 out:
3128 return ret;
3129 }
3130
3131 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3132 * for some operations; this forces use of the newer bridge-utils that
3133 * use compatible ioctls
3134 */
3135 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3136 {
3137 compat_ulong_t tmp;
3138
3139 if (get_user(tmp, argp))
3140 return -EFAULT;
3141 if (tmp == BRCTL_GET_VERSION)
3142 return BRCTL_VERSION + 1;
3143 return -EINVAL;
3144 }
3145
3146 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3147 unsigned int cmd, unsigned long arg)
3148 {
3149 void __user *argp = compat_ptr(arg);
3150 struct sock *sk = sock->sk;
3151 struct net *net = sock_net(sk);
3152
3153 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3154 return compat_ifr_data_ioctl(net, cmd, argp);
3155
3156 switch (cmd) {
3157 case SIOCSIFBR:
3158 case SIOCGIFBR:
3159 return old_bridge_ioctl(argp);
3160 case SIOCGIFNAME:
3161 return dev_ifname32(net, argp);
3162 case SIOCGIFCONF:
3163 return dev_ifconf(net, argp);
3164 case SIOCETHTOOL:
3165 return ethtool_ioctl(net, argp);
3166 case SIOCWANDEV:
3167 return compat_siocwandev(net, argp);
3168 case SIOCGIFMAP:
3169 case SIOCSIFMAP:
3170 return compat_sioc_ifmap(net, cmd, argp);
3171 case SIOCBONDENSLAVE:
3172 case SIOCBONDRELEASE:
3173 case SIOCBONDSETHWADDR:
3174 case SIOCBONDCHANGEACTIVE:
3175 return bond_ioctl(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 SIOCSIFNAME:
3232 case SIOCGMIIPHY:
3233 case SIOCGMIIREG:
3234 case SIOCSMIIREG:
3235 return dev_ifsioc(net, sock, cmd, argp);
3236
3237 case SIOCSARP:
3238 case SIOCGARP:
3239 case SIOCDARP:
3240 case SIOCATMARK:
3241 return sock_do_ioctl(net, sock, cmd, arg);
3242 }
3243
3244 return -ENOIOCTLCMD;
3245 }
3246
3247 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3248 unsigned long arg)
3249 {
3250 struct socket *sock = file->private_data;
3251 int ret = -ENOIOCTLCMD;
3252 struct sock *sk;
3253 struct net *net;
3254
3255 sk = sock->sk;
3256 net = sock_net(sk);
3257
3258 if (sock->ops->compat_ioctl)
3259 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3260
3261 if (ret == -ENOIOCTLCMD &&
3262 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3263 ret = compat_wext_handle_ioctl(net, cmd, arg);
3264
3265 if (ret == -ENOIOCTLCMD)
3266 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3267
3268 return ret;
3269 }
3270 #endif
3271
3272 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3273 {
3274 return sock->ops->bind(sock, addr, addrlen);
3275 }
3276 EXPORT_SYMBOL(kernel_bind);
3277
3278 int kernel_listen(struct socket *sock, int backlog)
3279 {
3280 return sock->ops->listen(sock, backlog);
3281 }
3282 EXPORT_SYMBOL(kernel_listen);
3283
3284 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3285 {
3286 struct sock *sk = sock->sk;
3287 int err;
3288
3289 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3290 newsock);
3291 if (err < 0)
3292 goto done;
3293
3294 err = sock->ops->accept(sock, *newsock, flags, true);
3295 if (err < 0) {
3296 sock_release(*newsock);
3297 *newsock = NULL;
3298 goto done;
3299 }
3300
3301 (*newsock)->ops = sock->ops;
3302 __module_get((*newsock)->ops->owner);
3303
3304 done:
3305 return err;
3306 }
3307 EXPORT_SYMBOL(kernel_accept);
3308
3309 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3310 int flags)
3311 {
3312 return sock->ops->connect(sock, addr, addrlen, flags);
3313 }
3314 EXPORT_SYMBOL(kernel_connect);
3315
3316 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3317 int *addrlen)
3318 {
3319 return sock->ops->getname(sock, addr, addrlen, 0);
3320 }
3321 EXPORT_SYMBOL(kernel_getsockname);
3322
3323 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3324 int *addrlen)
3325 {
3326 return sock->ops->getname(sock, addr, addrlen, 1);
3327 }
3328 EXPORT_SYMBOL(kernel_getpeername);
3329
3330 int kernel_getsockopt(struct socket *sock, int level, int optname,
3331 char *optval, int *optlen)
3332 {
3333 mm_segment_t oldfs = get_fs();
3334 char __user *uoptval;
3335 int __user *uoptlen;
3336 int err;
3337
3338 uoptval = (char __user __force *) optval;
3339 uoptlen = (int __user __force *) optlen;
3340
3341 set_fs(KERNEL_DS);
3342 if (level == SOL_SOCKET)
3343 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3344 else
3345 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3346 uoptlen);
3347 set_fs(oldfs);
3348 return err;
3349 }
3350 EXPORT_SYMBOL(kernel_getsockopt);
3351
3352 int kernel_setsockopt(struct socket *sock, int level, int optname,
3353 char *optval, unsigned int optlen)
3354 {
3355 mm_segment_t oldfs = get_fs();
3356 char __user *uoptval;
3357 int err;
3358
3359 uoptval = (char __user __force *) optval;
3360
3361 set_fs(KERNEL_DS);
3362 if (level == SOL_SOCKET)
3363 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3364 else
3365 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3366 optlen);
3367 set_fs(oldfs);
3368 return err;
3369 }
3370 EXPORT_SYMBOL(kernel_setsockopt);
3371
3372 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3373 size_t size, int flags)
3374 {
3375 if (sock->ops->sendpage)
3376 return sock->ops->sendpage(sock, page, offset, size, flags);
3377
3378 return sock_no_sendpage(sock, page, offset, size, flags);
3379 }
3380 EXPORT_SYMBOL(kernel_sendpage);
3381
3382 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3383 {
3384 mm_segment_t oldfs = get_fs();
3385 int err;
3386
3387 set_fs(KERNEL_DS);
3388 err = sock->ops->ioctl(sock, cmd, arg);
3389 set_fs(oldfs);
3390
3391 return err;
3392 }
3393 EXPORT_SYMBOL(kernel_sock_ioctl);
3394
3395 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3396 {
3397 return sock->ops->shutdown(sock, how);
3398 }
3399 EXPORT_SYMBOL(kernel_sock_shutdown);
3400
3401 /* This routine returns the IP overhead imposed by a socket i.e.
3402 * the length of the underlying IP header, depending on whether
3403 * this is an IPv4 or IPv6 socket and the length from IP options turned
3404 * on at the socket. Assumes that the caller has a lock on the socket.
3405 */
3406 u32 kernel_sock_ip_overhead(struct sock *sk)
3407 {
3408 struct inet_sock *inet;
3409 struct ip_options_rcu *opt;
3410 u32 overhead = 0;
3411 bool owned_by_user;
3412 #if IS_ENABLED(CONFIG_IPV6)
3413 struct ipv6_pinfo *np;
3414 struct ipv6_txoptions *optv6 = NULL;
3415 #endif /* IS_ENABLED(CONFIG_IPV6) */
3416
3417 if (!sk)
3418 return overhead;
3419
3420 owned_by_user = sock_owned_by_user(sk);
3421 switch (sk->sk_family) {
3422 case AF_INET:
3423 inet = inet_sk(sk);
3424 overhead += sizeof(struct iphdr);
3425 opt = rcu_dereference_protected(inet->inet_opt,
3426 owned_by_user);
3427 if (opt)
3428 overhead += opt->opt.optlen;
3429 return overhead;
3430 #if IS_ENABLED(CONFIG_IPV6)
3431 case AF_INET6:
3432 np = inet6_sk(sk);
3433 overhead += sizeof(struct ipv6hdr);
3434 if (np)
3435 optv6 = rcu_dereference_protected(np->opt,
3436 owned_by_user);
3437 if (optv6)
3438 overhead += (optv6->opt_flen + optv6->opt_nflen);
3439 return overhead;
3440 #endif /* IS_ENABLED(CONFIG_IPV6) */
3441 default: /* Returns 0 overhead if the socket is not ipv4 or ipv6 */
3442 return overhead;
3443 }
3444 }
3445 EXPORT_SYMBOL(kernel_sock_ip_overhead);