]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/socket.c
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
[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 * too 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 /*
656 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
657 */
658 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
659 struct sk_buff *skb)
660 {
661 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
662 struct scm_timestamping tss;
663 int empty = 1;
664 struct skb_shared_hwtstamps *shhwtstamps =
665 skb_hwtstamps(skb);
666
667 /* Race occurred between timestamp enabling and packet
668 receiving. Fill in the current time for now. */
669 if (need_software_tstamp && skb->tstamp == 0)
670 __net_timestamp(skb);
671
672 if (need_software_tstamp) {
673 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
674 struct timeval tv;
675 skb_get_timestamp(skb, &tv);
676 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
677 sizeof(tv), &tv);
678 } else {
679 struct timespec ts;
680 skb_get_timestampns(skb, &ts);
681 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
682 sizeof(ts), &ts);
683 }
684 }
685
686 memset(&tss, 0, sizeof(tss));
687 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
688 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
689 empty = 0;
690 if (shhwtstamps &&
691 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
692 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
693 empty = 0;
694 if (!empty) {
695 put_cmsg(msg, SOL_SOCKET,
696 SCM_TIMESTAMPING, sizeof(tss), &tss);
697
698 if (skb->len && (sk->sk_tsflags & SOF_TIMESTAMPING_OPT_STATS))
699 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
700 skb->len, skb->data);
701 }
702 }
703 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
704
705 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
706 struct sk_buff *skb)
707 {
708 int ack;
709
710 if (!sock_flag(sk, SOCK_WIFI_STATUS))
711 return;
712 if (!skb->wifi_acked_valid)
713 return;
714
715 ack = skb->wifi_acked;
716
717 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
718 }
719 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
720
721 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
722 struct sk_buff *skb)
723 {
724 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
725 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
726 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
727 }
728
729 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
730 struct sk_buff *skb)
731 {
732 sock_recv_timestamp(msg, sk, skb);
733 sock_recv_drops(msg, sk, skb);
734 }
735 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
736
737 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
738 int flags)
739 {
740 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
741 }
742
743 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
744 {
745 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
746
747 return err ?: sock_recvmsg_nosec(sock, msg, flags);
748 }
749 EXPORT_SYMBOL(sock_recvmsg);
750
751 /**
752 * kernel_recvmsg - Receive a message from a socket (kernel space)
753 * @sock: The socket to receive the message from
754 * @msg: Received message
755 * @vec: Input s/g array for message data
756 * @num: Size of input s/g array
757 * @size: Number of bytes to read
758 * @flags: Message flags (MSG_DONTWAIT, etc...)
759 *
760 * On return the msg structure contains the scatter/gather array passed in the
761 * vec argument. The array is modified so that it consists of the unfilled
762 * portion of the original array.
763 *
764 * The returned value is the total number of bytes received, or an error.
765 */
766 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
767 struct kvec *vec, size_t num, size_t size, int flags)
768 {
769 mm_segment_t oldfs = get_fs();
770 int result;
771
772 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
773 set_fs(KERNEL_DS);
774 result = sock_recvmsg(sock, msg, flags);
775 set_fs(oldfs);
776 return result;
777 }
778 EXPORT_SYMBOL(kernel_recvmsg);
779
780 static ssize_t sock_sendpage(struct file *file, struct page *page,
781 int offset, size_t size, loff_t *ppos, int more)
782 {
783 struct socket *sock;
784 int flags;
785
786 sock = file->private_data;
787
788 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
789 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
790 flags |= more;
791
792 return kernel_sendpage(sock, page, offset, size, flags);
793 }
794
795 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
796 struct pipe_inode_info *pipe, size_t len,
797 unsigned int flags)
798 {
799 struct socket *sock = file->private_data;
800
801 if (unlikely(!sock->ops->splice_read))
802 return -EINVAL;
803
804 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
805 }
806
807 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
808 {
809 struct file *file = iocb->ki_filp;
810 struct socket *sock = file->private_data;
811 struct msghdr msg = {.msg_iter = *to,
812 .msg_iocb = iocb};
813 ssize_t res;
814
815 if (file->f_flags & O_NONBLOCK)
816 msg.msg_flags = MSG_DONTWAIT;
817
818 if (iocb->ki_pos != 0)
819 return -ESPIPE;
820
821 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
822 return 0;
823
824 res = sock_recvmsg(sock, &msg, msg.msg_flags);
825 *to = msg.msg_iter;
826 return res;
827 }
828
829 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
830 {
831 struct file *file = iocb->ki_filp;
832 struct socket *sock = file->private_data;
833 struct msghdr msg = {.msg_iter = *from,
834 .msg_iocb = iocb};
835 ssize_t res;
836
837 if (iocb->ki_pos != 0)
838 return -ESPIPE;
839
840 if (file->f_flags & O_NONBLOCK)
841 msg.msg_flags = MSG_DONTWAIT;
842
843 if (sock->type == SOCK_SEQPACKET)
844 msg.msg_flags |= MSG_EOR;
845
846 res = sock_sendmsg(sock, &msg);
847 *from = msg.msg_iter;
848 return res;
849 }
850
851 /*
852 * Atomic setting of ioctl hooks to avoid race
853 * with module unload.
854 */
855
856 static DEFINE_MUTEX(br_ioctl_mutex);
857 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
858
859 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
860 {
861 mutex_lock(&br_ioctl_mutex);
862 br_ioctl_hook = hook;
863 mutex_unlock(&br_ioctl_mutex);
864 }
865 EXPORT_SYMBOL(brioctl_set);
866
867 static DEFINE_MUTEX(vlan_ioctl_mutex);
868 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
869
870 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
871 {
872 mutex_lock(&vlan_ioctl_mutex);
873 vlan_ioctl_hook = hook;
874 mutex_unlock(&vlan_ioctl_mutex);
875 }
876 EXPORT_SYMBOL(vlan_ioctl_set);
877
878 static DEFINE_MUTEX(dlci_ioctl_mutex);
879 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
880
881 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
882 {
883 mutex_lock(&dlci_ioctl_mutex);
884 dlci_ioctl_hook = hook;
885 mutex_unlock(&dlci_ioctl_mutex);
886 }
887 EXPORT_SYMBOL(dlci_ioctl_set);
888
889 static long sock_do_ioctl(struct net *net, struct socket *sock,
890 unsigned int cmd, unsigned long arg)
891 {
892 int err;
893 void __user *argp = (void __user *)arg;
894
895 err = sock->ops->ioctl(sock, cmd, arg);
896
897 /*
898 * If this ioctl is unknown try to hand it down
899 * to the NIC driver.
900 */
901 if (err == -ENOIOCTLCMD)
902 err = dev_ioctl(net, cmd, argp);
903
904 return err;
905 }
906
907 /*
908 * With an ioctl, arg may well be a user mode pointer, but we don't know
909 * what to do with it - that's up to the protocol still.
910 */
911
912 static struct ns_common *get_net_ns(struct ns_common *ns)
913 {
914 return &get_net(container_of(ns, struct net, ns))->ns;
915 }
916
917 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
918 {
919 struct socket *sock;
920 struct sock *sk;
921 void __user *argp = (void __user *)arg;
922 int pid, err;
923 struct net *net;
924
925 sock = file->private_data;
926 sk = sock->sk;
927 net = sock_net(sk);
928 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
929 err = dev_ioctl(net, cmd, argp);
930 } else
931 #ifdef CONFIG_WEXT_CORE
932 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
933 err = dev_ioctl(net, cmd, argp);
934 } else
935 #endif
936 switch (cmd) {
937 case FIOSETOWN:
938 case SIOCSPGRP:
939 err = -EFAULT;
940 if (get_user(pid, (int __user *)argp))
941 break;
942 f_setown(sock->file, pid, 1);
943 err = 0;
944 break;
945 case FIOGETOWN:
946 case SIOCGPGRP:
947 err = put_user(f_getown(sock->file),
948 (int __user *)argp);
949 break;
950 case SIOCGIFBR:
951 case SIOCSIFBR:
952 case SIOCBRADDBR:
953 case SIOCBRDELBR:
954 err = -ENOPKG;
955 if (!br_ioctl_hook)
956 request_module("bridge");
957
958 mutex_lock(&br_ioctl_mutex);
959 if (br_ioctl_hook)
960 err = br_ioctl_hook(net, cmd, argp);
961 mutex_unlock(&br_ioctl_mutex);
962 break;
963 case SIOCGIFVLAN:
964 case SIOCSIFVLAN:
965 err = -ENOPKG;
966 if (!vlan_ioctl_hook)
967 request_module("8021q");
968
969 mutex_lock(&vlan_ioctl_mutex);
970 if (vlan_ioctl_hook)
971 err = vlan_ioctl_hook(net, argp);
972 mutex_unlock(&vlan_ioctl_mutex);
973 break;
974 case SIOCADDDLCI:
975 case SIOCDELDLCI:
976 err = -ENOPKG;
977 if (!dlci_ioctl_hook)
978 request_module("dlci");
979
980 mutex_lock(&dlci_ioctl_mutex);
981 if (dlci_ioctl_hook)
982 err = dlci_ioctl_hook(cmd, argp);
983 mutex_unlock(&dlci_ioctl_mutex);
984 break;
985 case SIOCGSKNS:
986 err = -EPERM;
987 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
988 break;
989
990 err = open_related_ns(&net->ns, get_net_ns);
991 break;
992 default:
993 err = sock_do_ioctl(net, sock, cmd, arg);
994 break;
995 }
996 return err;
997 }
998
999 int sock_create_lite(int family, int type, int protocol, struct socket **res)
1000 {
1001 int err;
1002 struct socket *sock = NULL;
1003
1004 err = security_socket_create(family, type, protocol, 1);
1005 if (err)
1006 goto out;
1007
1008 sock = sock_alloc();
1009 if (!sock) {
1010 err = -ENOMEM;
1011 goto out;
1012 }
1013
1014 sock->type = type;
1015 err = security_socket_post_create(sock, family, type, protocol, 1);
1016 if (err)
1017 goto out_release;
1018
1019 out:
1020 *res = sock;
1021 return err;
1022 out_release:
1023 sock_release(sock);
1024 sock = NULL;
1025 goto out;
1026 }
1027 EXPORT_SYMBOL(sock_create_lite);
1028
1029 /* No kernel lock held - perfect */
1030 static unsigned int sock_poll(struct file *file, poll_table *wait)
1031 {
1032 unsigned int busy_flag = 0;
1033 struct socket *sock;
1034
1035 /*
1036 * We can't return errors to poll, so it's either yes or no.
1037 */
1038 sock = file->private_data;
1039
1040 if (sk_can_busy_loop(sock->sk)) {
1041 /* this socket can poll_ll so tell the system call */
1042 busy_flag = POLL_BUSY_LOOP;
1043
1044 /* once, only if requested by syscall */
1045 if (wait && (wait->_key & POLL_BUSY_LOOP))
1046 sk_busy_loop(sock->sk, 1);
1047 }
1048
1049 return busy_flag | sock->ops->poll(file, sock, wait);
1050 }
1051
1052 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1053 {
1054 struct socket *sock = file->private_data;
1055
1056 return sock->ops->mmap(file, sock, vma);
1057 }
1058
1059 static int sock_close(struct inode *inode, struct file *filp)
1060 {
1061 sock_release(SOCKET_I(inode));
1062 return 0;
1063 }
1064
1065 /*
1066 * Update the socket async list
1067 *
1068 * Fasync_list locking strategy.
1069 *
1070 * 1. fasync_list is modified only under process context socket lock
1071 * i.e. under semaphore.
1072 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1073 * or under socket lock
1074 */
1075
1076 static int sock_fasync(int fd, struct file *filp, int on)
1077 {
1078 struct socket *sock = filp->private_data;
1079 struct sock *sk = sock->sk;
1080 struct socket_wq *wq;
1081
1082 if (sk == NULL)
1083 return -EINVAL;
1084
1085 lock_sock(sk);
1086 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1087 fasync_helper(fd, filp, on, &wq->fasync_list);
1088
1089 if (!wq->fasync_list)
1090 sock_reset_flag(sk, SOCK_FASYNC);
1091 else
1092 sock_set_flag(sk, SOCK_FASYNC);
1093
1094 release_sock(sk);
1095 return 0;
1096 }
1097
1098 /* This function may be called only under rcu_lock */
1099
1100 int sock_wake_async(struct socket_wq *wq, int how, int band)
1101 {
1102 if (!wq || !wq->fasync_list)
1103 return -1;
1104
1105 switch (how) {
1106 case SOCK_WAKE_WAITD:
1107 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1108 break;
1109 goto call_kill;
1110 case SOCK_WAKE_SPACE:
1111 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1112 break;
1113 /* fall through */
1114 case SOCK_WAKE_IO:
1115 call_kill:
1116 kill_fasync(&wq->fasync_list, SIGIO, band);
1117 break;
1118 case SOCK_WAKE_URG:
1119 kill_fasync(&wq->fasync_list, SIGURG, band);
1120 }
1121
1122 return 0;
1123 }
1124 EXPORT_SYMBOL(sock_wake_async);
1125
1126 int __sock_create(struct net *net, int family, int type, int protocol,
1127 struct socket **res, int kern)
1128 {
1129 int err;
1130 struct socket *sock;
1131 const struct net_proto_family *pf;
1132
1133 /*
1134 * Check protocol is in range
1135 */
1136 if (family < 0 || family >= NPROTO)
1137 return -EAFNOSUPPORT;
1138 if (type < 0 || type >= SOCK_MAX)
1139 return -EINVAL;
1140
1141 /* Compatibility.
1142
1143 This uglymoron is moved from INET layer to here to avoid
1144 deadlock in module load.
1145 */
1146 if (family == PF_INET && type == SOCK_PACKET) {
1147 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1148 current->comm);
1149 family = PF_PACKET;
1150 }
1151
1152 err = security_socket_create(family, type, protocol, kern);
1153 if (err)
1154 return err;
1155
1156 /*
1157 * Allocate the socket and allow the family to set things up. if
1158 * the protocol is 0, the family is instructed to select an appropriate
1159 * default.
1160 */
1161 sock = sock_alloc();
1162 if (!sock) {
1163 net_warn_ratelimited("socket: no more sockets\n");
1164 return -ENFILE; /* Not exactly a match, but its the
1165 closest posix thing */
1166 }
1167
1168 sock->type = type;
1169
1170 #ifdef CONFIG_MODULES
1171 /* Attempt to load a protocol module if the find failed.
1172 *
1173 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1174 * requested real, full-featured networking support upon configuration.
1175 * Otherwise module support will break!
1176 */
1177 if (rcu_access_pointer(net_families[family]) == NULL)
1178 request_module("net-pf-%d", family);
1179 #endif
1180
1181 rcu_read_lock();
1182 pf = rcu_dereference(net_families[family]);
1183 err = -EAFNOSUPPORT;
1184 if (!pf)
1185 goto out_release;
1186
1187 /*
1188 * We will call the ->create function, that possibly is in a loadable
1189 * module, so we have to bump that loadable module refcnt first.
1190 */
1191 if (!try_module_get(pf->owner))
1192 goto out_release;
1193
1194 /* Now protected by module ref count */
1195 rcu_read_unlock();
1196
1197 err = pf->create(net, sock, protocol, kern);
1198 if (err < 0)
1199 goto out_module_put;
1200
1201 /*
1202 * Now to bump the refcnt of the [loadable] module that owns this
1203 * socket at sock_release time we decrement its refcnt.
1204 */
1205 if (!try_module_get(sock->ops->owner))
1206 goto out_module_busy;
1207
1208 /*
1209 * Now that we're done with the ->create function, the [loadable]
1210 * module can have its refcnt decremented
1211 */
1212 module_put(pf->owner);
1213 err = security_socket_post_create(sock, family, type, protocol, kern);
1214 if (err)
1215 goto out_sock_release;
1216 *res = sock;
1217
1218 return 0;
1219
1220 out_module_busy:
1221 err = -EAFNOSUPPORT;
1222 out_module_put:
1223 sock->ops = NULL;
1224 module_put(pf->owner);
1225 out_sock_release:
1226 sock_release(sock);
1227 return err;
1228
1229 out_release:
1230 rcu_read_unlock();
1231 goto out_sock_release;
1232 }
1233 EXPORT_SYMBOL(__sock_create);
1234
1235 int sock_create(int family, int type, int protocol, struct socket **res)
1236 {
1237 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1238 }
1239 EXPORT_SYMBOL(sock_create);
1240
1241 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1242 {
1243 return __sock_create(net, family, type, protocol, res, 1);
1244 }
1245 EXPORT_SYMBOL(sock_create_kern);
1246
1247 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1248 {
1249 int retval;
1250 struct socket *sock;
1251 int flags;
1252
1253 /* Check the SOCK_* constants for consistency. */
1254 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1255 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1256 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1257 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1258
1259 flags = type & ~SOCK_TYPE_MASK;
1260 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1261 return -EINVAL;
1262 type &= SOCK_TYPE_MASK;
1263
1264 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1265 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1266
1267 retval = sock_create(family, type, protocol, &sock);
1268 if (retval < 0)
1269 goto out;
1270
1271 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1272 if (retval < 0)
1273 goto out_release;
1274
1275 out:
1276 /* It may be already another descriptor 8) Not kernel problem. */
1277 return retval;
1278
1279 out_release:
1280 sock_release(sock);
1281 return retval;
1282 }
1283
1284 /*
1285 * Create a pair of connected sockets.
1286 */
1287
1288 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1289 int __user *, usockvec)
1290 {
1291 struct socket *sock1, *sock2;
1292 int fd1, fd2, err;
1293 struct file *newfile1, *newfile2;
1294 int flags;
1295
1296 flags = type & ~SOCK_TYPE_MASK;
1297 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1298 return -EINVAL;
1299 type &= SOCK_TYPE_MASK;
1300
1301 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1302 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1303
1304 /*
1305 * Obtain the first socket and check if the underlying protocol
1306 * supports the socketpair call.
1307 */
1308
1309 err = sock_create(family, type, protocol, &sock1);
1310 if (err < 0)
1311 goto out;
1312
1313 err = sock_create(family, type, protocol, &sock2);
1314 if (err < 0)
1315 goto out_release_1;
1316
1317 err = sock1->ops->socketpair(sock1, sock2);
1318 if (err < 0)
1319 goto out_release_both;
1320
1321 fd1 = get_unused_fd_flags(flags);
1322 if (unlikely(fd1 < 0)) {
1323 err = fd1;
1324 goto out_release_both;
1325 }
1326
1327 fd2 = get_unused_fd_flags(flags);
1328 if (unlikely(fd2 < 0)) {
1329 err = fd2;
1330 goto out_put_unused_1;
1331 }
1332
1333 newfile1 = sock_alloc_file(sock1, flags, NULL);
1334 if (IS_ERR(newfile1)) {
1335 err = PTR_ERR(newfile1);
1336 goto out_put_unused_both;
1337 }
1338
1339 newfile2 = sock_alloc_file(sock2, flags, NULL);
1340 if (IS_ERR(newfile2)) {
1341 err = PTR_ERR(newfile2);
1342 goto out_fput_1;
1343 }
1344
1345 err = put_user(fd1, &usockvec[0]);
1346 if (err)
1347 goto out_fput_both;
1348
1349 err = put_user(fd2, &usockvec[1]);
1350 if (err)
1351 goto out_fput_both;
1352
1353 audit_fd_pair(fd1, fd2);
1354
1355 fd_install(fd1, newfile1);
1356 fd_install(fd2, newfile2);
1357 /* fd1 and fd2 may be already another descriptors.
1358 * Not kernel problem.
1359 */
1360
1361 return 0;
1362
1363 out_fput_both:
1364 fput(newfile2);
1365 fput(newfile1);
1366 put_unused_fd(fd2);
1367 put_unused_fd(fd1);
1368 goto out;
1369
1370 out_fput_1:
1371 fput(newfile1);
1372 put_unused_fd(fd2);
1373 put_unused_fd(fd1);
1374 sock_release(sock2);
1375 goto out;
1376
1377 out_put_unused_both:
1378 put_unused_fd(fd2);
1379 out_put_unused_1:
1380 put_unused_fd(fd1);
1381 out_release_both:
1382 sock_release(sock2);
1383 out_release_1:
1384 sock_release(sock1);
1385 out:
1386 return err;
1387 }
1388
1389 /*
1390 * Bind a name to a socket. Nothing much to do here since it's
1391 * the protocol's responsibility to handle the local address.
1392 *
1393 * We move the socket address to kernel space before we call
1394 * the protocol layer (having also checked the address is ok).
1395 */
1396
1397 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1398 {
1399 struct socket *sock;
1400 struct sockaddr_storage address;
1401 int err, fput_needed;
1402
1403 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1404 if (sock) {
1405 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1406 if (err >= 0) {
1407 err = security_socket_bind(sock,
1408 (struct sockaddr *)&address,
1409 addrlen);
1410 if (!err)
1411 err = sock->ops->bind(sock,
1412 (struct sockaddr *)
1413 &address, addrlen);
1414 }
1415 fput_light(sock->file, fput_needed);
1416 }
1417 return err;
1418 }
1419
1420 /*
1421 * Perform a listen. Basically, we allow the protocol to do anything
1422 * necessary for a listen, and if that works, we mark the socket as
1423 * ready for listening.
1424 */
1425
1426 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1427 {
1428 struct socket *sock;
1429 int err, fput_needed;
1430 int somaxconn;
1431
1432 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1433 if (sock) {
1434 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1435 if ((unsigned int)backlog > somaxconn)
1436 backlog = somaxconn;
1437
1438 err = security_socket_listen(sock, backlog);
1439 if (!err)
1440 err = sock->ops->listen(sock, backlog);
1441
1442 fput_light(sock->file, fput_needed);
1443 }
1444 return err;
1445 }
1446
1447 /*
1448 * For accept, we attempt to create a new socket, set up the link
1449 * with the client, wake up the client, then return the new
1450 * connected fd. We collect the address of the connector in kernel
1451 * space and move it to user at the very end. This is unclean because
1452 * we open the socket then return an error.
1453 *
1454 * 1003.1g adds the ability to recvmsg() to query connection pending
1455 * status to recvmsg. We need to add that support in a way thats
1456 * clean when we restucture accept also.
1457 */
1458
1459 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1460 int __user *, upeer_addrlen, int, flags)
1461 {
1462 struct socket *sock, *newsock;
1463 struct file *newfile;
1464 int err, len, newfd, fput_needed;
1465 struct sockaddr_storage address;
1466
1467 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1468 return -EINVAL;
1469
1470 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1471 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1472
1473 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1474 if (!sock)
1475 goto out;
1476
1477 err = -ENFILE;
1478 newsock = sock_alloc();
1479 if (!newsock)
1480 goto out_put;
1481
1482 newsock->type = sock->type;
1483 newsock->ops = sock->ops;
1484
1485 /*
1486 * We don't need try_module_get here, as the listening socket (sock)
1487 * has the protocol module (sock->ops->owner) held.
1488 */
1489 __module_get(newsock->ops->owner);
1490
1491 newfd = get_unused_fd_flags(flags);
1492 if (unlikely(newfd < 0)) {
1493 err = newfd;
1494 sock_release(newsock);
1495 goto out_put;
1496 }
1497 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1498 if (IS_ERR(newfile)) {
1499 err = PTR_ERR(newfile);
1500 put_unused_fd(newfd);
1501 sock_release(newsock);
1502 goto out_put;
1503 }
1504
1505 err = security_socket_accept(sock, newsock);
1506 if (err)
1507 goto out_fd;
1508
1509 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1510 if (err < 0)
1511 goto out_fd;
1512
1513 if (upeer_sockaddr) {
1514 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1515 &len, 2) < 0) {
1516 err = -ECONNABORTED;
1517 goto out_fd;
1518 }
1519 err = move_addr_to_user(&address,
1520 len, upeer_sockaddr, upeer_addrlen);
1521 if (err < 0)
1522 goto out_fd;
1523 }
1524
1525 /* File flags are not inherited via accept() unlike another OSes. */
1526
1527 fd_install(newfd, newfile);
1528 err = newfd;
1529
1530 out_put:
1531 fput_light(sock->file, fput_needed);
1532 out:
1533 return err;
1534 out_fd:
1535 fput(newfile);
1536 put_unused_fd(newfd);
1537 goto out_put;
1538 }
1539
1540 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1541 int __user *, upeer_addrlen)
1542 {
1543 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1544 }
1545
1546 /*
1547 * Attempt to connect to a socket with the server address. The address
1548 * is in user space so we verify it is OK and move it to kernel space.
1549 *
1550 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1551 * break bindings
1552 *
1553 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1554 * other SEQPACKET protocols that take time to connect() as it doesn't
1555 * include the -EINPROGRESS status for such sockets.
1556 */
1557
1558 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1559 int, addrlen)
1560 {
1561 struct socket *sock;
1562 struct sockaddr_storage address;
1563 int err, fput_needed;
1564
1565 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1566 if (!sock)
1567 goto out;
1568 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1569 if (err < 0)
1570 goto out_put;
1571
1572 err =
1573 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1574 if (err)
1575 goto out_put;
1576
1577 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1578 sock->file->f_flags);
1579 out_put:
1580 fput_light(sock->file, fput_needed);
1581 out:
1582 return err;
1583 }
1584
1585 /*
1586 * Get the local address ('name') of a socket object. Move the obtained
1587 * name to user space.
1588 */
1589
1590 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1591 int __user *, usockaddr_len)
1592 {
1593 struct socket *sock;
1594 struct sockaddr_storage address;
1595 int len, err, fput_needed;
1596
1597 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1598 if (!sock)
1599 goto out;
1600
1601 err = security_socket_getsockname(sock);
1602 if (err)
1603 goto out_put;
1604
1605 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1606 if (err)
1607 goto out_put;
1608 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1609
1610 out_put:
1611 fput_light(sock->file, fput_needed);
1612 out:
1613 return err;
1614 }
1615
1616 /*
1617 * Get the remote address ('name') of a socket object. Move the obtained
1618 * name to user space.
1619 */
1620
1621 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1622 int __user *, usockaddr_len)
1623 {
1624 struct socket *sock;
1625 struct sockaddr_storage address;
1626 int len, err, fput_needed;
1627
1628 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1629 if (sock != NULL) {
1630 err = security_socket_getpeername(sock);
1631 if (err) {
1632 fput_light(sock->file, fput_needed);
1633 return err;
1634 }
1635
1636 err =
1637 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1638 1);
1639 if (!err)
1640 err = move_addr_to_user(&address, len, usockaddr,
1641 usockaddr_len);
1642 fput_light(sock->file, fput_needed);
1643 }
1644 return err;
1645 }
1646
1647 /*
1648 * Send a datagram to a given address. We move the address into kernel
1649 * space and check the user space data area is readable before invoking
1650 * the protocol.
1651 */
1652
1653 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1654 unsigned int, flags, struct sockaddr __user *, addr,
1655 int, addr_len)
1656 {
1657 struct socket *sock;
1658 struct sockaddr_storage address;
1659 int err;
1660 struct msghdr msg;
1661 struct iovec iov;
1662 int fput_needed;
1663
1664 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1665 if (unlikely(err))
1666 return err;
1667 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1668 if (!sock)
1669 goto out;
1670
1671 msg.msg_name = NULL;
1672 msg.msg_control = NULL;
1673 msg.msg_controllen = 0;
1674 msg.msg_namelen = 0;
1675 if (addr) {
1676 err = move_addr_to_kernel(addr, addr_len, &address);
1677 if (err < 0)
1678 goto out_put;
1679 msg.msg_name = (struct sockaddr *)&address;
1680 msg.msg_namelen = addr_len;
1681 }
1682 if (sock->file->f_flags & O_NONBLOCK)
1683 flags |= MSG_DONTWAIT;
1684 msg.msg_flags = flags;
1685 err = sock_sendmsg(sock, &msg);
1686
1687 out_put:
1688 fput_light(sock->file, fput_needed);
1689 out:
1690 return err;
1691 }
1692
1693 /*
1694 * Send a datagram down a socket.
1695 */
1696
1697 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1698 unsigned int, flags)
1699 {
1700 return sys_sendto(fd, buff, len, flags, NULL, 0);
1701 }
1702
1703 /*
1704 * Receive a frame from the socket and optionally record the address of the
1705 * sender. We verify the buffers are writable and if needed move the
1706 * sender address from kernel to user space.
1707 */
1708
1709 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1710 unsigned int, flags, struct sockaddr __user *, addr,
1711 int __user *, addr_len)
1712 {
1713 struct socket *sock;
1714 struct iovec iov;
1715 struct msghdr msg;
1716 struct sockaddr_storage address;
1717 int err, err2;
1718 int fput_needed;
1719
1720 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1721 if (unlikely(err))
1722 return err;
1723 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1724 if (!sock)
1725 goto out;
1726
1727 msg.msg_control = NULL;
1728 msg.msg_controllen = 0;
1729 /* Save some cycles and don't copy the address if not needed */
1730 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1731 /* We assume all kernel code knows the size of sockaddr_storage */
1732 msg.msg_namelen = 0;
1733 msg.msg_iocb = NULL;
1734 if (sock->file->f_flags & O_NONBLOCK)
1735 flags |= MSG_DONTWAIT;
1736 err = sock_recvmsg(sock, &msg, flags);
1737
1738 if (err >= 0 && addr != NULL) {
1739 err2 = move_addr_to_user(&address,
1740 msg.msg_namelen, addr, addr_len);
1741 if (err2 < 0)
1742 err = err2;
1743 }
1744
1745 fput_light(sock->file, fput_needed);
1746 out:
1747 return err;
1748 }
1749
1750 /*
1751 * Receive a datagram from a socket.
1752 */
1753
1754 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1755 unsigned int, flags)
1756 {
1757 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1758 }
1759
1760 /*
1761 * Set a socket option. Because we don't know the option lengths we have
1762 * to pass the user mode parameter for the protocols to sort out.
1763 */
1764
1765 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1766 char __user *, optval, int, optlen)
1767 {
1768 int err, fput_needed;
1769 struct socket *sock;
1770
1771 if (optlen < 0)
1772 return -EINVAL;
1773
1774 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1775 if (sock != NULL) {
1776 err = security_socket_setsockopt(sock, level, optname);
1777 if (err)
1778 goto out_put;
1779
1780 if (level == SOL_SOCKET)
1781 err =
1782 sock_setsockopt(sock, level, optname, optval,
1783 optlen);
1784 else
1785 err =
1786 sock->ops->setsockopt(sock, level, optname, optval,
1787 optlen);
1788 out_put:
1789 fput_light(sock->file, fput_needed);
1790 }
1791 return err;
1792 }
1793
1794 /*
1795 * Get a socket option. Because we don't know the option lengths we have
1796 * to pass a user mode parameter for the protocols to sort out.
1797 */
1798
1799 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1800 char __user *, optval, int __user *, optlen)
1801 {
1802 int err, fput_needed;
1803 struct socket *sock;
1804
1805 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1806 if (sock != NULL) {
1807 err = security_socket_getsockopt(sock, level, optname);
1808 if (err)
1809 goto out_put;
1810
1811 if (level == SOL_SOCKET)
1812 err =
1813 sock_getsockopt(sock, level, optname, optval,
1814 optlen);
1815 else
1816 err =
1817 sock->ops->getsockopt(sock, level, optname, optval,
1818 optlen);
1819 out_put:
1820 fput_light(sock->file, fput_needed);
1821 }
1822 return err;
1823 }
1824
1825 /*
1826 * Shutdown a socket.
1827 */
1828
1829 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1830 {
1831 int err, fput_needed;
1832 struct socket *sock;
1833
1834 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1835 if (sock != NULL) {
1836 err = security_socket_shutdown(sock, how);
1837 if (!err)
1838 err = sock->ops->shutdown(sock, how);
1839 fput_light(sock->file, fput_needed);
1840 }
1841 return err;
1842 }
1843
1844 /* A couple of helpful macros for getting the address of the 32/64 bit
1845 * fields which are the same type (int / unsigned) on our platforms.
1846 */
1847 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1848 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1849 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1850
1851 struct used_address {
1852 struct sockaddr_storage name;
1853 unsigned int name_len;
1854 };
1855
1856 static int copy_msghdr_from_user(struct msghdr *kmsg,
1857 struct user_msghdr __user *umsg,
1858 struct sockaddr __user **save_addr,
1859 struct iovec **iov)
1860 {
1861 struct sockaddr __user *uaddr;
1862 struct iovec __user *uiov;
1863 size_t nr_segs;
1864 ssize_t err;
1865
1866 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1867 __get_user(uaddr, &umsg->msg_name) ||
1868 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1869 __get_user(uiov, &umsg->msg_iov) ||
1870 __get_user(nr_segs, &umsg->msg_iovlen) ||
1871 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1872 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1873 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1874 return -EFAULT;
1875
1876 if (!uaddr)
1877 kmsg->msg_namelen = 0;
1878
1879 if (kmsg->msg_namelen < 0)
1880 return -EINVAL;
1881
1882 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1883 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1884
1885 if (save_addr)
1886 *save_addr = uaddr;
1887
1888 if (uaddr && kmsg->msg_namelen) {
1889 if (!save_addr) {
1890 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1891 kmsg->msg_name);
1892 if (err < 0)
1893 return err;
1894 }
1895 } else {
1896 kmsg->msg_name = NULL;
1897 kmsg->msg_namelen = 0;
1898 }
1899
1900 if (nr_segs > UIO_MAXIOV)
1901 return -EMSGSIZE;
1902
1903 kmsg->msg_iocb = NULL;
1904
1905 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1906 UIO_FASTIOV, iov, &kmsg->msg_iter);
1907 }
1908
1909 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1910 struct msghdr *msg_sys, unsigned int flags,
1911 struct used_address *used_address,
1912 unsigned int allowed_msghdr_flags)
1913 {
1914 struct compat_msghdr __user *msg_compat =
1915 (struct compat_msghdr __user *)msg;
1916 struct sockaddr_storage address;
1917 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1918 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1919 __aligned(sizeof(__kernel_size_t));
1920 /* 20 is size of ipv6_pktinfo */
1921 unsigned char *ctl_buf = ctl;
1922 int ctl_len;
1923 ssize_t err;
1924
1925 msg_sys->msg_name = &address;
1926
1927 if (MSG_CMSG_COMPAT & flags)
1928 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1929 else
1930 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1931 if (err < 0)
1932 return err;
1933
1934 err = -ENOBUFS;
1935
1936 if (msg_sys->msg_controllen > INT_MAX)
1937 goto out_freeiov;
1938 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1939 ctl_len = msg_sys->msg_controllen;
1940 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1941 err =
1942 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1943 sizeof(ctl));
1944 if (err)
1945 goto out_freeiov;
1946 ctl_buf = msg_sys->msg_control;
1947 ctl_len = msg_sys->msg_controllen;
1948 } else if (ctl_len) {
1949 BUILD_BUG_ON(sizeof(struct cmsghdr) !=
1950 CMSG_ALIGN(sizeof(struct cmsghdr)));
1951 if (ctl_len > sizeof(ctl)) {
1952 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1953 if (ctl_buf == NULL)
1954 goto out_freeiov;
1955 }
1956 err = -EFAULT;
1957 /*
1958 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1959 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1960 * checking falls down on this.
1961 */
1962 if (copy_from_user(ctl_buf,
1963 (void __user __force *)msg_sys->msg_control,
1964 ctl_len))
1965 goto out_freectl;
1966 msg_sys->msg_control = ctl_buf;
1967 }
1968 msg_sys->msg_flags = flags;
1969
1970 if (sock->file->f_flags & O_NONBLOCK)
1971 msg_sys->msg_flags |= MSG_DONTWAIT;
1972 /*
1973 * If this is sendmmsg() and current destination address is same as
1974 * previously succeeded address, omit asking LSM's decision.
1975 * used_address->name_len is initialized to UINT_MAX so that the first
1976 * destination address never matches.
1977 */
1978 if (used_address && msg_sys->msg_name &&
1979 used_address->name_len == msg_sys->msg_namelen &&
1980 !memcmp(&used_address->name, msg_sys->msg_name,
1981 used_address->name_len)) {
1982 err = sock_sendmsg_nosec(sock, msg_sys);
1983 goto out_freectl;
1984 }
1985 err = sock_sendmsg(sock, msg_sys);
1986 /*
1987 * If this is sendmmsg() and sending to current destination address was
1988 * successful, remember it.
1989 */
1990 if (used_address && err >= 0) {
1991 used_address->name_len = msg_sys->msg_namelen;
1992 if (msg_sys->msg_name)
1993 memcpy(&used_address->name, msg_sys->msg_name,
1994 used_address->name_len);
1995 }
1996
1997 out_freectl:
1998 if (ctl_buf != ctl)
1999 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2000 out_freeiov:
2001 kfree(iov);
2002 return err;
2003 }
2004
2005 /*
2006 * BSD sendmsg interface
2007 */
2008
2009 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2010 {
2011 int fput_needed, err;
2012 struct msghdr msg_sys;
2013 struct socket *sock;
2014
2015 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2016 if (!sock)
2017 goto out;
2018
2019 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2020
2021 fput_light(sock->file, fput_needed);
2022 out:
2023 return err;
2024 }
2025
2026 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2027 {
2028 if (flags & MSG_CMSG_COMPAT)
2029 return -EINVAL;
2030 return __sys_sendmsg(fd, msg, flags);
2031 }
2032
2033 /*
2034 * Linux sendmmsg interface
2035 */
2036
2037 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2038 unsigned int flags)
2039 {
2040 int fput_needed, err, datagrams;
2041 struct socket *sock;
2042 struct mmsghdr __user *entry;
2043 struct compat_mmsghdr __user *compat_entry;
2044 struct msghdr msg_sys;
2045 struct used_address used_address;
2046 unsigned int oflags = flags;
2047
2048 if (vlen > UIO_MAXIOV)
2049 vlen = UIO_MAXIOV;
2050
2051 datagrams = 0;
2052
2053 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2054 if (!sock)
2055 return err;
2056
2057 used_address.name_len = UINT_MAX;
2058 entry = mmsg;
2059 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2060 err = 0;
2061 flags |= MSG_BATCH;
2062
2063 while (datagrams < vlen) {
2064 if (datagrams == vlen - 1)
2065 flags = oflags;
2066
2067 if (MSG_CMSG_COMPAT & flags) {
2068 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2069 &msg_sys, flags, &used_address, MSG_EOR);
2070 if (err < 0)
2071 break;
2072 err = __put_user(err, &compat_entry->msg_len);
2073 ++compat_entry;
2074 } else {
2075 err = ___sys_sendmsg(sock,
2076 (struct user_msghdr __user *)entry,
2077 &msg_sys, flags, &used_address, MSG_EOR);
2078 if (err < 0)
2079 break;
2080 err = put_user(err, &entry->msg_len);
2081 ++entry;
2082 }
2083
2084 if (err)
2085 break;
2086 ++datagrams;
2087 if (msg_data_left(&msg_sys))
2088 break;
2089 cond_resched();
2090 }
2091
2092 fput_light(sock->file, fput_needed);
2093
2094 /* We only return an error if no datagrams were able to be sent */
2095 if (datagrams != 0)
2096 return datagrams;
2097
2098 return err;
2099 }
2100
2101 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2102 unsigned int, vlen, unsigned int, flags)
2103 {
2104 if (flags & MSG_CMSG_COMPAT)
2105 return -EINVAL;
2106 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2107 }
2108
2109 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2110 struct msghdr *msg_sys, unsigned int flags, int nosec)
2111 {
2112 struct compat_msghdr __user *msg_compat =
2113 (struct compat_msghdr __user *)msg;
2114 struct iovec iovstack[UIO_FASTIOV];
2115 struct iovec *iov = iovstack;
2116 unsigned long cmsg_ptr;
2117 int len;
2118 ssize_t err;
2119
2120 /* kernel mode address */
2121 struct sockaddr_storage addr;
2122
2123 /* user mode address pointers */
2124 struct sockaddr __user *uaddr;
2125 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2126
2127 msg_sys->msg_name = &addr;
2128
2129 if (MSG_CMSG_COMPAT & flags)
2130 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2131 else
2132 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2133 if (err < 0)
2134 return err;
2135
2136 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2137 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2138
2139 /* We assume all kernel code knows the size of sockaddr_storage */
2140 msg_sys->msg_namelen = 0;
2141
2142 if (sock->file->f_flags & O_NONBLOCK)
2143 flags |= MSG_DONTWAIT;
2144 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2145 if (err < 0)
2146 goto out_freeiov;
2147 len = err;
2148
2149 if (uaddr != NULL) {
2150 err = move_addr_to_user(&addr,
2151 msg_sys->msg_namelen, uaddr,
2152 uaddr_len);
2153 if (err < 0)
2154 goto out_freeiov;
2155 }
2156 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2157 COMPAT_FLAGS(msg));
2158 if (err)
2159 goto out_freeiov;
2160 if (MSG_CMSG_COMPAT & flags)
2161 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2162 &msg_compat->msg_controllen);
2163 else
2164 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2165 &msg->msg_controllen);
2166 if (err)
2167 goto out_freeiov;
2168 err = len;
2169
2170 out_freeiov:
2171 kfree(iov);
2172 return err;
2173 }
2174
2175 /*
2176 * BSD recvmsg interface
2177 */
2178
2179 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2180 {
2181 int fput_needed, err;
2182 struct msghdr msg_sys;
2183 struct socket *sock;
2184
2185 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2186 if (!sock)
2187 goto out;
2188
2189 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2190
2191 fput_light(sock->file, fput_needed);
2192 out:
2193 return err;
2194 }
2195
2196 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2197 unsigned int, flags)
2198 {
2199 if (flags & MSG_CMSG_COMPAT)
2200 return -EINVAL;
2201 return __sys_recvmsg(fd, msg, flags);
2202 }
2203
2204 /*
2205 * Linux recvmmsg interface
2206 */
2207
2208 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2209 unsigned int flags, struct timespec *timeout)
2210 {
2211 int fput_needed, err, datagrams;
2212 struct socket *sock;
2213 struct mmsghdr __user *entry;
2214 struct compat_mmsghdr __user *compat_entry;
2215 struct msghdr msg_sys;
2216 struct timespec64 end_time;
2217 struct timespec64 timeout64;
2218
2219 if (timeout &&
2220 poll_select_set_timeout(&end_time, timeout->tv_sec,
2221 timeout->tv_nsec))
2222 return -EINVAL;
2223
2224 datagrams = 0;
2225
2226 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2227 if (!sock)
2228 return err;
2229
2230 err = sock_error(sock->sk);
2231 if (err) {
2232 datagrams = err;
2233 goto out_put;
2234 }
2235
2236 entry = mmsg;
2237 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2238
2239 while (datagrams < vlen) {
2240 /*
2241 * No need to ask LSM for more than the first datagram.
2242 */
2243 if (MSG_CMSG_COMPAT & flags) {
2244 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2245 &msg_sys, flags & ~MSG_WAITFORONE,
2246 datagrams);
2247 if (err < 0)
2248 break;
2249 err = __put_user(err, &compat_entry->msg_len);
2250 ++compat_entry;
2251 } else {
2252 err = ___sys_recvmsg(sock,
2253 (struct user_msghdr __user *)entry,
2254 &msg_sys, flags & ~MSG_WAITFORONE,
2255 datagrams);
2256 if (err < 0)
2257 break;
2258 err = put_user(err, &entry->msg_len);
2259 ++entry;
2260 }
2261
2262 if (err)
2263 break;
2264 ++datagrams;
2265
2266 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2267 if (flags & MSG_WAITFORONE)
2268 flags |= MSG_DONTWAIT;
2269
2270 if (timeout) {
2271 ktime_get_ts64(&timeout64);
2272 *timeout = timespec64_to_timespec(
2273 timespec64_sub(end_time, timeout64));
2274 if (timeout->tv_sec < 0) {
2275 timeout->tv_sec = timeout->tv_nsec = 0;
2276 break;
2277 }
2278
2279 /* Timeout, return less than vlen datagrams */
2280 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2281 break;
2282 }
2283
2284 /* Out of band data, return right away */
2285 if (msg_sys.msg_flags & MSG_OOB)
2286 break;
2287 cond_resched();
2288 }
2289
2290 if (err == 0)
2291 goto out_put;
2292
2293 if (datagrams == 0) {
2294 datagrams = err;
2295 goto out_put;
2296 }
2297
2298 /*
2299 * We may return less entries than requested (vlen) if the
2300 * sock is non block and there aren't enough datagrams...
2301 */
2302 if (err != -EAGAIN) {
2303 /*
2304 * ... or if recvmsg returns an error after we
2305 * received some datagrams, where we record the
2306 * error to return on the next call or if the
2307 * app asks about it using getsockopt(SO_ERROR).
2308 */
2309 sock->sk->sk_err = -err;
2310 }
2311 out_put:
2312 fput_light(sock->file, fput_needed);
2313
2314 return datagrams;
2315 }
2316
2317 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2318 unsigned int, vlen, unsigned int, flags,
2319 struct timespec __user *, timeout)
2320 {
2321 int datagrams;
2322 struct timespec timeout_sys;
2323
2324 if (flags & MSG_CMSG_COMPAT)
2325 return -EINVAL;
2326
2327 if (!timeout)
2328 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2329
2330 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2331 return -EFAULT;
2332
2333 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2334
2335 if (datagrams > 0 &&
2336 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2337 datagrams = -EFAULT;
2338
2339 return datagrams;
2340 }
2341
2342 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2343 /* Argument list sizes for sys_socketcall */
2344 #define AL(x) ((x) * sizeof(unsigned long))
2345 static const unsigned char nargs[21] = {
2346 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2347 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2348 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2349 AL(4), AL(5), AL(4)
2350 };
2351
2352 #undef AL
2353
2354 /*
2355 * System call vectors.
2356 *
2357 * Argument checking cleaned up. Saved 20% in size.
2358 * This function doesn't need to set the kernel lock because
2359 * it is set by the callees.
2360 */
2361
2362 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2363 {
2364 unsigned long a[AUDITSC_ARGS];
2365 unsigned long a0, a1;
2366 int err;
2367 unsigned int len;
2368
2369 if (call < 1 || call > SYS_SENDMMSG)
2370 return -EINVAL;
2371
2372 len = nargs[call];
2373 if (len > sizeof(a))
2374 return -EINVAL;
2375
2376 /* copy_from_user should be SMP safe. */
2377 if (copy_from_user(a, args, len))
2378 return -EFAULT;
2379
2380 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2381 if (err)
2382 return err;
2383
2384 a0 = a[0];
2385 a1 = a[1];
2386
2387 switch (call) {
2388 case SYS_SOCKET:
2389 err = sys_socket(a0, a1, a[2]);
2390 break;
2391 case SYS_BIND:
2392 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2393 break;
2394 case SYS_CONNECT:
2395 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2396 break;
2397 case SYS_LISTEN:
2398 err = sys_listen(a0, a1);
2399 break;
2400 case SYS_ACCEPT:
2401 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2402 (int __user *)a[2], 0);
2403 break;
2404 case SYS_GETSOCKNAME:
2405 err =
2406 sys_getsockname(a0, (struct sockaddr __user *)a1,
2407 (int __user *)a[2]);
2408 break;
2409 case SYS_GETPEERNAME:
2410 err =
2411 sys_getpeername(a0, (struct sockaddr __user *)a1,
2412 (int __user *)a[2]);
2413 break;
2414 case SYS_SOCKETPAIR:
2415 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2416 break;
2417 case SYS_SEND:
2418 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2419 break;
2420 case SYS_SENDTO:
2421 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2422 (struct sockaddr __user *)a[4], a[5]);
2423 break;
2424 case SYS_RECV:
2425 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2426 break;
2427 case SYS_RECVFROM:
2428 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2429 (struct sockaddr __user *)a[4],
2430 (int __user *)a[5]);
2431 break;
2432 case SYS_SHUTDOWN:
2433 err = sys_shutdown(a0, a1);
2434 break;
2435 case SYS_SETSOCKOPT:
2436 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2437 break;
2438 case SYS_GETSOCKOPT:
2439 err =
2440 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2441 (int __user *)a[4]);
2442 break;
2443 case SYS_SENDMSG:
2444 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2445 break;
2446 case SYS_SENDMMSG:
2447 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2448 break;
2449 case SYS_RECVMSG:
2450 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2451 break;
2452 case SYS_RECVMMSG:
2453 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2454 (struct timespec __user *)a[4]);
2455 break;
2456 case SYS_ACCEPT4:
2457 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2458 (int __user *)a[2], a[3]);
2459 break;
2460 default:
2461 err = -EINVAL;
2462 break;
2463 }
2464 return err;
2465 }
2466
2467 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2468
2469 /**
2470 * sock_register - add a socket protocol handler
2471 * @ops: description of protocol
2472 *
2473 * This function is called by a protocol handler that wants to
2474 * advertise its address family, and have it linked into the
2475 * socket interface. The value ops->family corresponds to the
2476 * socket system call protocol family.
2477 */
2478 int sock_register(const struct net_proto_family *ops)
2479 {
2480 int err;
2481
2482 if (ops->family >= NPROTO) {
2483 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2484 return -ENOBUFS;
2485 }
2486
2487 spin_lock(&net_family_lock);
2488 if (rcu_dereference_protected(net_families[ops->family],
2489 lockdep_is_held(&net_family_lock)))
2490 err = -EEXIST;
2491 else {
2492 rcu_assign_pointer(net_families[ops->family], ops);
2493 err = 0;
2494 }
2495 spin_unlock(&net_family_lock);
2496
2497 pr_info("NET: Registered protocol family %d\n", ops->family);
2498 return err;
2499 }
2500 EXPORT_SYMBOL(sock_register);
2501
2502 /**
2503 * sock_unregister - remove a protocol handler
2504 * @family: protocol family to remove
2505 *
2506 * This function is called by a protocol handler that wants to
2507 * remove its address family, and have it unlinked from the
2508 * new socket creation.
2509 *
2510 * If protocol handler is a module, then it can use module reference
2511 * counts to protect against new references. If protocol handler is not
2512 * a module then it needs to provide its own protection in
2513 * the ops->create routine.
2514 */
2515 void sock_unregister(int family)
2516 {
2517 BUG_ON(family < 0 || family >= NPROTO);
2518
2519 spin_lock(&net_family_lock);
2520 RCU_INIT_POINTER(net_families[family], NULL);
2521 spin_unlock(&net_family_lock);
2522
2523 synchronize_rcu();
2524
2525 pr_info("NET: Unregistered protocol family %d\n", family);
2526 }
2527 EXPORT_SYMBOL(sock_unregister);
2528
2529 static int __init sock_init(void)
2530 {
2531 int err;
2532 /*
2533 * Initialize the network sysctl infrastructure.
2534 */
2535 err = net_sysctl_init();
2536 if (err)
2537 goto out;
2538
2539 /*
2540 * Initialize skbuff SLAB cache
2541 */
2542 skb_init();
2543
2544 /*
2545 * Initialize the protocols module.
2546 */
2547
2548 init_inodecache();
2549
2550 err = register_filesystem(&sock_fs_type);
2551 if (err)
2552 goto out_fs;
2553 sock_mnt = kern_mount(&sock_fs_type);
2554 if (IS_ERR(sock_mnt)) {
2555 err = PTR_ERR(sock_mnt);
2556 goto out_mount;
2557 }
2558
2559 /* The real protocol initialization is performed in later initcalls.
2560 */
2561
2562 #ifdef CONFIG_NETFILTER
2563 err = netfilter_init();
2564 if (err)
2565 goto out;
2566 #endif
2567
2568 ptp_classifier_init();
2569
2570 out:
2571 return err;
2572
2573 out_mount:
2574 unregister_filesystem(&sock_fs_type);
2575 out_fs:
2576 goto out;
2577 }
2578
2579 core_initcall(sock_init); /* early initcall */
2580
2581 #ifdef CONFIG_PROC_FS
2582 void socket_seq_show(struct seq_file *seq)
2583 {
2584 int cpu;
2585 int counter = 0;
2586
2587 for_each_possible_cpu(cpu)
2588 counter += per_cpu(sockets_in_use, cpu);
2589
2590 /* It can be negative, by the way. 8) */
2591 if (counter < 0)
2592 counter = 0;
2593
2594 seq_printf(seq, "sockets: used %d\n", counter);
2595 }
2596 #endif /* CONFIG_PROC_FS */
2597
2598 #ifdef CONFIG_COMPAT
2599 static int do_siocgstamp(struct net *net, struct socket *sock,
2600 unsigned int cmd, void __user *up)
2601 {
2602 mm_segment_t old_fs = get_fs();
2603 struct timeval ktv;
2604 int err;
2605
2606 set_fs(KERNEL_DS);
2607 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2608 set_fs(old_fs);
2609 if (!err)
2610 err = compat_put_timeval(&ktv, up);
2611
2612 return err;
2613 }
2614
2615 static int do_siocgstampns(struct net *net, struct socket *sock,
2616 unsigned int cmd, void __user *up)
2617 {
2618 mm_segment_t old_fs = get_fs();
2619 struct timespec kts;
2620 int err;
2621
2622 set_fs(KERNEL_DS);
2623 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2624 set_fs(old_fs);
2625 if (!err)
2626 err = compat_put_timespec(&kts, up);
2627
2628 return err;
2629 }
2630
2631 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2632 {
2633 struct ifreq __user *uifr;
2634 int err;
2635
2636 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2637 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2638 return -EFAULT;
2639
2640 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2641 if (err)
2642 return err;
2643
2644 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2645 return -EFAULT;
2646
2647 return 0;
2648 }
2649
2650 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2651 {
2652 struct compat_ifconf ifc32;
2653 struct ifconf ifc;
2654 struct ifconf __user *uifc;
2655 struct compat_ifreq __user *ifr32;
2656 struct ifreq __user *ifr;
2657 unsigned int i, j;
2658 int err;
2659
2660 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2661 return -EFAULT;
2662
2663 memset(&ifc, 0, sizeof(ifc));
2664 if (ifc32.ifcbuf == 0) {
2665 ifc32.ifc_len = 0;
2666 ifc.ifc_len = 0;
2667 ifc.ifc_req = NULL;
2668 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2669 } else {
2670 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2671 sizeof(struct ifreq);
2672 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2673 ifc.ifc_len = len;
2674 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2675 ifr32 = compat_ptr(ifc32.ifcbuf);
2676 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2677 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2678 return -EFAULT;
2679 ifr++;
2680 ifr32++;
2681 }
2682 }
2683 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2684 return -EFAULT;
2685
2686 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2687 if (err)
2688 return err;
2689
2690 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2691 return -EFAULT;
2692
2693 ifr = ifc.ifc_req;
2694 ifr32 = compat_ptr(ifc32.ifcbuf);
2695 for (i = 0, j = 0;
2696 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2697 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2698 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2699 return -EFAULT;
2700 ifr32++;
2701 ifr++;
2702 }
2703
2704 if (ifc32.ifcbuf == 0) {
2705 /* Translate from 64-bit structure multiple to
2706 * a 32-bit one.
2707 */
2708 i = ifc.ifc_len;
2709 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2710 ifc32.ifc_len = i;
2711 } else {
2712 ifc32.ifc_len = i;
2713 }
2714 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2715 return -EFAULT;
2716
2717 return 0;
2718 }
2719
2720 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2721 {
2722 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2723 bool convert_in = false, convert_out = false;
2724 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2725 struct ethtool_rxnfc __user *rxnfc;
2726 struct ifreq __user *ifr;
2727 u32 rule_cnt = 0, actual_rule_cnt;
2728 u32 ethcmd;
2729 u32 data;
2730 int ret;
2731
2732 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2733 return -EFAULT;
2734
2735 compat_rxnfc = compat_ptr(data);
2736
2737 if (get_user(ethcmd, &compat_rxnfc->cmd))
2738 return -EFAULT;
2739
2740 /* Most ethtool structures are defined without padding.
2741 * Unfortunately struct ethtool_rxnfc is an exception.
2742 */
2743 switch (ethcmd) {
2744 default:
2745 break;
2746 case ETHTOOL_GRXCLSRLALL:
2747 /* Buffer size is variable */
2748 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2749 return -EFAULT;
2750 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2751 return -ENOMEM;
2752 buf_size += rule_cnt * sizeof(u32);
2753 /* fall through */
2754 case ETHTOOL_GRXRINGS:
2755 case ETHTOOL_GRXCLSRLCNT:
2756 case ETHTOOL_GRXCLSRULE:
2757 case ETHTOOL_SRXCLSRLINS:
2758 convert_out = true;
2759 /* fall through */
2760 case ETHTOOL_SRXCLSRLDEL:
2761 buf_size += sizeof(struct ethtool_rxnfc);
2762 convert_in = true;
2763 break;
2764 }
2765
2766 ifr = compat_alloc_user_space(buf_size);
2767 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2768
2769 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2770 return -EFAULT;
2771
2772 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2773 &ifr->ifr_ifru.ifru_data))
2774 return -EFAULT;
2775
2776 if (convert_in) {
2777 /* We expect there to be holes between fs.m_ext and
2778 * fs.ring_cookie and at the end of fs, but nowhere else.
2779 */
2780 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2781 sizeof(compat_rxnfc->fs.m_ext) !=
2782 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2783 sizeof(rxnfc->fs.m_ext));
2784 BUILD_BUG_ON(
2785 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2786 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2787 offsetof(struct ethtool_rxnfc, fs.location) -
2788 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2789
2790 if (copy_in_user(rxnfc, compat_rxnfc,
2791 (void __user *)(&rxnfc->fs.m_ext + 1) -
2792 (void __user *)rxnfc) ||
2793 copy_in_user(&rxnfc->fs.ring_cookie,
2794 &compat_rxnfc->fs.ring_cookie,
2795 (void __user *)(&rxnfc->fs.location + 1) -
2796 (void __user *)&rxnfc->fs.ring_cookie) ||
2797 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2798 sizeof(rxnfc->rule_cnt)))
2799 return -EFAULT;
2800 }
2801
2802 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2803 if (ret)
2804 return ret;
2805
2806 if (convert_out) {
2807 if (copy_in_user(compat_rxnfc, rxnfc,
2808 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2809 (const void __user *)rxnfc) ||
2810 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2811 &rxnfc->fs.ring_cookie,
2812 (const void __user *)(&rxnfc->fs.location + 1) -
2813 (const void __user *)&rxnfc->fs.ring_cookie) ||
2814 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2815 sizeof(rxnfc->rule_cnt)))
2816 return -EFAULT;
2817
2818 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2819 /* As an optimisation, we only copy the actual
2820 * number of rules that the underlying
2821 * function returned. Since Mallory might
2822 * change the rule count in user memory, we
2823 * check that it is less than the rule count
2824 * originally given (as the user buffer size),
2825 * which has been range-checked.
2826 */
2827 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2828 return -EFAULT;
2829 if (actual_rule_cnt < rule_cnt)
2830 rule_cnt = actual_rule_cnt;
2831 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2832 &rxnfc->rule_locs[0],
2833 rule_cnt * sizeof(u32)))
2834 return -EFAULT;
2835 }
2836 }
2837
2838 return 0;
2839 }
2840
2841 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2842 {
2843 void __user *uptr;
2844 compat_uptr_t uptr32;
2845 struct ifreq __user *uifr;
2846
2847 uifr = compat_alloc_user_space(sizeof(*uifr));
2848 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2849 return -EFAULT;
2850
2851 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2852 return -EFAULT;
2853
2854 uptr = compat_ptr(uptr32);
2855
2856 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2857 return -EFAULT;
2858
2859 return dev_ioctl(net, SIOCWANDEV, uifr);
2860 }
2861
2862 static int bond_ioctl(struct net *net, unsigned int cmd,
2863 struct compat_ifreq __user *ifr32)
2864 {
2865 struct ifreq kifr;
2866 mm_segment_t old_fs;
2867 int err;
2868
2869 switch (cmd) {
2870 case SIOCBONDENSLAVE:
2871 case SIOCBONDRELEASE:
2872 case SIOCBONDSETHWADDR:
2873 case SIOCBONDCHANGEACTIVE:
2874 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2875 return -EFAULT;
2876
2877 old_fs = get_fs();
2878 set_fs(KERNEL_DS);
2879 err = dev_ioctl(net, cmd,
2880 (struct ifreq __user __force *) &kifr);
2881 set_fs(old_fs);
2882
2883 return err;
2884 default:
2885 return -ENOIOCTLCMD;
2886 }
2887 }
2888
2889 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2890 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2891 struct compat_ifreq __user *u_ifreq32)
2892 {
2893 struct ifreq __user *u_ifreq64;
2894 char tmp_buf[IFNAMSIZ];
2895 void __user *data64;
2896 u32 data32;
2897
2898 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2899 IFNAMSIZ))
2900 return -EFAULT;
2901 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2902 return -EFAULT;
2903 data64 = compat_ptr(data32);
2904
2905 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2906
2907 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2908 IFNAMSIZ))
2909 return -EFAULT;
2910 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2911 return -EFAULT;
2912
2913 return dev_ioctl(net, cmd, u_ifreq64);
2914 }
2915
2916 static int dev_ifsioc(struct net *net, struct socket *sock,
2917 unsigned int cmd, struct compat_ifreq __user *uifr32)
2918 {
2919 struct ifreq __user *uifr;
2920 int err;
2921
2922 uifr = compat_alloc_user_space(sizeof(*uifr));
2923 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2924 return -EFAULT;
2925
2926 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2927
2928 if (!err) {
2929 switch (cmd) {
2930 case SIOCGIFFLAGS:
2931 case SIOCGIFMETRIC:
2932 case SIOCGIFMTU:
2933 case SIOCGIFMEM:
2934 case SIOCGIFHWADDR:
2935 case SIOCGIFINDEX:
2936 case SIOCGIFADDR:
2937 case SIOCGIFBRDADDR:
2938 case SIOCGIFDSTADDR:
2939 case SIOCGIFNETMASK:
2940 case SIOCGIFPFLAGS:
2941 case SIOCGIFTXQLEN:
2942 case SIOCGMIIPHY:
2943 case SIOCGMIIREG:
2944 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2945 err = -EFAULT;
2946 break;
2947 }
2948 }
2949 return err;
2950 }
2951
2952 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2953 struct compat_ifreq __user *uifr32)
2954 {
2955 struct ifreq ifr;
2956 struct compat_ifmap __user *uifmap32;
2957 mm_segment_t old_fs;
2958 int err;
2959
2960 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2961 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2962 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2963 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2964 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2965 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2966 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2967 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2968 if (err)
2969 return -EFAULT;
2970
2971 old_fs = get_fs();
2972 set_fs(KERNEL_DS);
2973 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2974 set_fs(old_fs);
2975
2976 if (cmd == SIOCGIFMAP && !err) {
2977 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2978 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2979 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2980 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2981 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2982 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2983 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2984 if (err)
2985 err = -EFAULT;
2986 }
2987 return err;
2988 }
2989
2990 struct rtentry32 {
2991 u32 rt_pad1;
2992 struct sockaddr rt_dst; /* target address */
2993 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2994 struct sockaddr rt_genmask; /* target network mask (IP) */
2995 unsigned short rt_flags;
2996 short rt_pad2;
2997 u32 rt_pad3;
2998 unsigned char rt_tos;
2999 unsigned char rt_class;
3000 short rt_pad4;
3001 short rt_metric; /* +1 for binary compatibility! */
3002 /* char * */ u32 rt_dev; /* forcing the device at add */
3003 u32 rt_mtu; /* per route MTU/Window */
3004 u32 rt_window; /* Window clamping */
3005 unsigned short rt_irtt; /* Initial RTT */
3006 };
3007
3008 struct in6_rtmsg32 {
3009 struct in6_addr rtmsg_dst;
3010 struct in6_addr rtmsg_src;
3011 struct in6_addr rtmsg_gateway;
3012 u32 rtmsg_type;
3013 u16 rtmsg_dst_len;
3014 u16 rtmsg_src_len;
3015 u32 rtmsg_metric;
3016 u32 rtmsg_info;
3017 u32 rtmsg_flags;
3018 s32 rtmsg_ifindex;
3019 };
3020
3021 static int routing_ioctl(struct net *net, struct socket *sock,
3022 unsigned int cmd, void __user *argp)
3023 {
3024 int ret;
3025 void *r = NULL;
3026 struct in6_rtmsg r6;
3027 struct rtentry r4;
3028 char devname[16];
3029 u32 rtdev;
3030 mm_segment_t old_fs = get_fs();
3031
3032 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3033 struct in6_rtmsg32 __user *ur6 = argp;
3034 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3035 3 * sizeof(struct in6_addr));
3036 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3037 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3038 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3039 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3040 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3041 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3042 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3043
3044 r = (void *) &r6;
3045 } else { /* ipv4 */
3046 struct rtentry32 __user *ur4 = argp;
3047 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3048 3 * sizeof(struct sockaddr));
3049 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3050 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3051 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3052 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3053 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3054 ret |= get_user(rtdev, &(ur4->rt_dev));
3055 if (rtdev) {
3056 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3057 r4.rt_dev = (char __user __force *)devname;
3058 devname[15] = 0;
3059 } else
3060 r4.rt_dev = NULL;
3061
3062 r = (void *) &r4;
3063 }
3064
3065 if (ret) {
3066 ret = -EFAULT;
3067 goto out;
3068 }
3069
3070 set_fs(KERNEL_DS);
3071 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3072 set_fs(old_fs);
3073
3074 out:
3075 return ret;
3076 }
3077
3078 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3079 * for some operations; this forces use of the newer bridge-utils that
3080 * use compatible ioctls
3081 */
3082 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3083 {
3084 compat_ulong_t tmp;
3085
3086 if (get_user(tmp, argp))
3087 return -EFAULT;
3088 if (tmp == BRCTL_GET_VERSION)
3089 return BRCTL_VERSION + 1;
3090 return -EINVAL;
3091 }
3092
3093 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3094 unsigned int cmd, unsigned long arg)
3095 {
3096 void __user *argp = compat_ptr(arg);
3097 struct sock *sk = sock->sk;
3098 struct net *net = sock_net(sk);
3099
3100 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3101 return compat_ifr_data_ioctl(net, cmd, argp);
3102
3103 switch (cmd) {
3104 case SIOCSIFBR:
3105 case SIOCGIFBR:
3106 return old_bridge_ioctl(argp);
3107 case SIOCGIFNAME:
3108 return dev_ifname32(net, argp);
3109 case SIOCGIFCONF:
3110 return dev_ifconf(net, argp);
3111 case SIOCETHTOOL:
3112 return ethtool_ioctl(net, argp);
3113 case SIOCWANDEV:
3114 return compat_siocwandev(net, argp);
3115 case SIOCGIFMAP:
3116 case SIOCSIFMAP:
3117 return compat_sioc_ifmap(net, cmd, argp);
3118 case SIOCBONDENSLAVE:
3119 case SIOCBONDRELEASE:
3120 case SIOCBONDSETHWADDR:
3121 case SIOCBONDCHANGEACTIVE:
3122 return bond_ioctl(net, cmd, argp);
3123 case SIOCADDRT:
3124 case SIOCDELRT:
3125 return routing_ioctl(net, sock, cmd, argp);
3126 case SIOCGSTAMP:
3127 return do_siocgstamp(net, sock, cmd, argp);
3128 case SIOCGSTAMPNS:
3129 return do_siocgstampns(net, sock, cmd, argp);
3130 case SIOCBONDSLAVEINFOQUERY:
3131 case SIOCBONDINFOQUERY:
3132 case SIOCSHWTSTAMP:
3133 case SIOCGHWTSTAMP:
3134 return compat_ifr_data_ioctl(net, cmd, argp);
3135
3136 case FIOSETOWN:
3137 case SIOCSPGRP:
3138 case FIOGETOWN:
3139 case SIOCGPGRP:
3140 case SIOCBRADDBR:
3141 case SIOCBRDELBR:
3142 case SIOCGIFVLAN:
3143 case SIOCSIFVLAN:
3144 case SIOCADDDLCI:
3145 case SIOCDELDLCI:
3146 case SIOCGSKNS:
3147 return sock_ioctl(file, cmd, arg);
3148
3149 case SIOCGIFFLAGS:
3150 case SIOCSIFFLAGS:
3151 case SIOCGIFMETRIC:
3152 case SIOCSIFMETRIC:
3153 case SIOCGIFMTU:
3154 case SIOCSIFMTU:
3155 case SIOCGIFMEM:
3156 case SIOCSIFMEM:
3157 case SIOCGIFHWADDR:
3158 case SIOCSIFHWADDR:
3159 case SIOCADDMULTI:
3160 case SIOCDELMULTI:
3161 case SIOCGIFINDEX:
3162 case SIOCGIFADDR:
3163 case SIOCSIFADDR:
3164 case SIOCSIFHWBROADCAST:
3165 case SIOCDIFADDR:
3166 case SIOCGIFBRDADDR:
3167 case SIOCSIFBRDADDR:
3168 case SIOCGIFDSTADDR:
3169 case SIOCSIFDSTADDR:
3170 case SIOCGIFNETMASK:
3171 case SIOCSIFNETMASK:
3172 case SIOCSIFPFLAGS:
3173 case SIOCGIFPFLAGS:
3174 case SIOCGIFTXQLEN:
3175 case SIOCSIFTXQLEN:
3176 case SIOCBRADDIF:
3177 case SIOCBRDELIF:
3178 case SIOCSIFNAME:
3179 case SIOCGMIIPHY:
3180 case SIOCGMIIREG:
3181 case SIOCSMIIREG:
3182 return dev_ifsioc(net, sock, cmd, argp);
3183
3184 case SIOCSARP:
3185 case SIOCGARP:
3186 case SIOCDARP:
3187 case SIOCATMARK:
3188 return sock_do_ioctl(net, sock, cmd, arg);
3189 }
3190
3191 return -ENOIOCTLCMD;
3192 }
3193
3194 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3195 unsigned long arg)
3196 {
3197 struct socket *sock = file->private_data;
3198 int ret = -ENOIOCTLCMD;
3199 struct sock *sk;
3200 struct net *net;
3201
3202 sk = sock->sk;
3203 net = sock_net(sk);
3204
3205 if (sock->ops->compat_ioctl)
3206 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3207
3208 if (ret == -ENOIOCTLCMD &&
3209 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3210 ret = compat_wext_handle_ioctl(net, cmd, arg);
3211
3212 if (ret == -ENOIOCTLCMD)
3213 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3214
3215 return ret;
3216 }
3217 #endif
3218
3219 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3220 {
3221 return sock->ops->bind(sock, addr, addrlen);
3222 }
3223 EXPORT_SYMBOL(kernel_bind);
3224
3225 int kernel_listen(struct socket *sock, int backlog)
3226 {
3227 return sock->ops->listen(sock, backlog);
3228 }
3229 EXPORT_SYMBOL(kernel_listen);
3230
3231 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3232 {
3233 struct sock *sk = sock->sk;
3234 int err;
3235
3236 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3237 newsock);
3238 if (err < 0)
3239 goto done;
3240
3241 err = sock->ops->accept(sock, *newsock, flags);
3242 if (err < 0) {
3243 sock_release(*newsock);
3244 *newsock = NULL;
3245 goto done;
3246 }
3247
3248 (*newsock)->ops = sock->ops;
3249 __module_get((*newsock)->ops->owner);
3250
3251 done:
3252 return err;
3253 }
3254 EXPORT_SYMBOL(kernel_accept);
3255
3256 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3257 int flags)
3258 {
3259 return sock->ops->connect(sock, addr, addrlen, flags);
3260 }
3261 EXPORT_SYMBOL(kernel_connect);
3262
3263 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3264 int *addrlen)
3265 {
3266 return sock->ops->getname(sock, addr, addrlen, 0);
3267 }
3268 EXPORT_SYMBOL(kernel_getsockname);
3269
3270 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3271 int *addrlen)
3272 {
3273 return sock->ops->getname(sock, addr, addrlen, 1);
3274 }
3275 EXPORT_SYMBOL(kernel_getpeername);
3276
3277 int kernel_getsockopt(struct socket *sock, int level, int optname,
3278 char *optval, int *optlen)
3279 {
3280 mm_segment_t oldfs = get_fs();
3281 char __user *uoptval;
3282 int __user *uoptlen;
3283 int err;
3284
3285 uoptval = (char __user __force *) optval;
3286 uoptlen = (int __user __force *) optlen;
3287
3288 set_fs(KERNEL_DS);
3289 if (level == SOL_SOCKET)
3290 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3291 else
3292 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3293 uoptlen);
3294 set_fs(oldfs);
3295 return err;
3296 }
3297 EXPORT_SYMBOL(kernel_getsockopt);
3298
3299 int kernel_setsockopt(struct socket *sock, int level, int optname,
3300 char *optval, unsigned int optlen)
3301 {
3302 mm_segment_t oldfs = get_fs();
3303 char __user *uoptval;
3304 int err;
3305
3306 uoptval = (char __user __force *) optval;
3307
3308 set_fs(KERNEL_DS);
3309 if (level == SOL_SOCKET)
3310 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3311 else
3312 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3313 optlen);
3314 set_fs(oldfs);
3315 return err;
3316 }
3317 EXPORT_SYMBOL(kernel_setsockopt);
3318
3319 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3320 size_t size, int flags)
3321 {
3322 if (sock->ops->sendpage)
3323 return sock->ops->sendpage(sock, page, offset, size, flags);
3324
3325 return sock_no_sendpage(sock, page, offset, size, flags);
3326 }
3327 EXPORT_SYMBOL(kernel_sendpage);
3328
3329 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3330 {
3331 mm_segment_t oldfs = get_fs();
3332 int err;
3333
3334 set_fs(KERNEL_DS);
3335 err = sock->ops->ioctl(sock, cmd, arg);
3336 set_fs(oldfs);
3337
3338 return err;
3339 }
3340 EXPORT_SYMBOL(kernel_sock_ioctl);
3341
3342 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3343 {
3344 return sock->ops->shutdown(sock, how);
3345 }
3346 EXPORT_SYMBOL(kernel_sock_shutdown);