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 <asm/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 int 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 if (sock_inode_cachep == NULL)
300 return -ENOMEM;
301 return 0;
302 }
303
304 static const struct super_operations sockfs_ops = {
305 .alloc_inode = sock_alloc_inode,
306 .destroy_inode = sock_destroy_inode,
307 .statfs = simple_statfs,
308 };
309
310 /*
311 * sockfs_dname() is called from d_path().
312 */
313 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314 {
315 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
316 d_inode(dentry)->i_ino);
317 }
318
319 static const struct dentry_operations sockfs_dentry_operations = {
320 .d_dname = sockfs_dname,
321 };
322
323 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324 int flags, const char *dev_name, void *data)
325 {
326 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327 &sockfs_dentry_operations, SOCKFS_MAGIC);
328 }
329
330 static struct vfsmount *sock_mnt __read_mostly;
331
332 static struct file_system_type sock_fs_type = {
333 .name = "sockfs",
334 .mount = sockfs_mount,
335 .kill_sb = kill_anon_super,
336 };
337
338 /*
339 * Obtains the first available file descriptor and sets it up for use.
340 *
341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
348 *
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
353 */
354
355 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
356 {
357 struct qstr name = { .name = "" };
358 struct path path;
359 struct file *file;
360
361 if (dname) {
362 name.name = dname;
363 name.len = strlen(name.name);
364 } else if (sock->sk) {
365 name.name = sock->sk->sk_prot_creator->name;
366 name.len = strlen(name.name);
367 }
368 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
369 if (unlikely(!path.dentry))
370 return ERR_PTR(-ENOMEM);
371 path.mnt = mntget(sock_mnt);
372
373 d_instantiate(path.dentry, SOCK_INODE(sock));
374
375 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
376 &socket_file_ops);
377 if (IS_ERR(file)) {
378 /* drop dentry, keep inode */
379 ihold(d_inode(path.dentry));
380 path_put(&path);
381 return file;
382 }
383
384 sock->file = file;
385 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
386 file->private_data = sock;
387 return file;
388 }
389 EXPORT_SYMBOL(sock_alloc_file);
390
391 static int sock_map_fd(struct socket *sock, int flags)
392 {
393 struct file *newfile;
394 int fd = get_unused_fd_flags(flags);
395 if (unlikely(fd < 0))
396 return fd;
397
398 newfile = sock_alloc_file(sock, flags, NULL);
399 if (likely(!IS_ERR(newfile))) {
400 fd_install(fd, newfile);
401 return fd;
402 }
403
404 put_unused_fd(fd);
405 return PTR_ERR(newfile);
406 }
407
408 struct socket *sock_from_file(struct file *file, int *err)
409 {
410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
412
413 *err = -ENOTSOCK;
414 return NULL;
415 }
416 EXPORT_SYMBOL(sock_from_file);
417
418 /**
419 * sockfd_lookup - Go from a file number to its socket slot
420 * @fd: file handle
421 * @err: pointer to an error code return
422 *
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
427 *
428 * On a success the socket object pointer is returned.
429 */
430
431 struct socket *sockfd_lookup(int fd, int *err)
432 {
433 struct file *file;
434 struct socket *sock;
435
436 file = fget(fd);
437 if (!file) {
438 *err = -EBADF;
439 return NULL;
440 }
441
442 sock = sock_from_file(file, err);
443 if (!sock)
444 fput(file);
445 return sock;
446 }
447 EXPORT_SYMBOL(sockfd_lookup);
448
449 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
450 {
451 struct fd f = fdget(fd);
452 struct socket *sock;
453
454 *err = -EBADF;
455 if (f.file) {
456 sock = sock_from_file(f.file, err);
457 if (likely(sock)) {
458 *fput_needed = f.flags;
459 return sock;
460 }
461 fdput(f);
462 }
463 return NULL;
464 }
465
466 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469 static ssize_t sockfs_getxattr(struct dentry *dentry,
470 const char *name, void *value, size_t size)
471 {
472 const char *proto_name;
473 size_t proto_size;
474 int error;
475
476 error = -ENODATA;
477 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478 proto_name = dentry->d_name.name;
479 proto_size = strlen(proto_name);
480
481 if (value) {
482 error = -ERANGE;
483 if (proto_size + 1 > size)
484 goto out;
485
486 strncpy(value, proto_name, proto_size + 1);
487 }
488 error = proto_size + 1;
489 }
490
491 out:
492 return error;
493 }
494
495 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496 size_t size)
497 {
498 ssize_t len;
499 ssize_t used = 0;
500
501 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
502 if (len < 0)
503 return len;
504 used += len;
505 if (buffer) {
506 if (size < used)
507 return -ERANGE;
508 buffer += len;
509 }
510
511 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512 used += len;
513 if (buffer) {
514 if (size < used)
515 return -ERANGE;
516 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517 buffer += len;
518 }
519
520 return used;
521 }
522
523 static const struct inode_operations sockfs_inode_ops = {
524 .getxattr = sockfs_getxattr,
525 .listxattr = sockfs_listxattr,
526 };
527
528 /**
529 * sock_alloc - allocate a socket
530 *
531 * Allocate a new inode and socket object. The two are bound together
532 * and initialised. The socket is then returned. If we are out of inodes
533 * NULL is returned.
534 */
535
536 struct socket *sock_alloc(void)
537 {
538 struct inode *inode;
539 struct socket *sock;
540
541 inode = new_inode_pseudo(sock_mnt->mnt_sb);
542 if (!inode)
543 return NULL;
544
545 sock = SOCKET_I(inode);
546
547 kmemcheck_annotate_bitfield(sock, type);
548 inode->i_ino = get_next_ino();
549 inode->i_mode = S_IFSOCK | S_IRWXUGO;
550 inode->i_uid = current_fsuid();
551 inode->i_gid = current_fsgid();
552 inode->i_op = &sockfs_inode_ops;
553
554 this_cpu_add(sockets_in_use, 1);
555 return sock;
556 }
557 EXPORT_SYMBOL(sock_alloc);
558
559 /**
560 * sock_release - close a socket
561 * @sock: socket to close
562 *
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
565 * an inode not a file.
566 */
567
568 void sock_release(struct socket *sock)
569 {
570 if (sock->ops) {
571 struct module *owner = sock->ops->owner;
572
573 sock->ops->release(sock);
574 sock->ops = NULL;
575 module_put(owner);
576 }
577
578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579 pr_err("%s: fasync list not empty!\n", __func__);
580
581 this_cpu_sub(sockets_in_use, 1);
582 if (!sock->file) {
583 iput(SOCK_INODE(sock));
584 return;
585 }
586 sock->file = NULL;
587 }
588 EXPORT_SYMBOL(sock_release);
589
590 void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
591 {
592 u8 flags = *tx_flags;
593
594 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
595 flags |= SKBTX_HW_TSTAMP;
596
597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
598 flags |= SKBTX_SW_TSTAMP;
599
600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
601 flags |= SKBTX_SCHED_TSTAMP;
602
603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
604 flags |= SKBTX_ACK_TSTAMP;
605
606 *tx_flags = flags;
607 }
608 EXPORT_SYMBOL(__sock_tx_timestamp);
609
610 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
611 {
612 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
613 BUG_ON(ret == -EIOCBQUEUED);
614 return ret;
615 }
616
617 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
618 {
619 int err = security_socket_sendmsg(sock, msg,
620 msg_data_left(msg));
621
622 return err ?: sock_sendmsg_nosec(sock, msg);
623 }
624 EXPORT_SYMBOL(sock_sendmsg);
625
626 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
627 struct kvec *vec, size_t num, size_t size)
628 {
629 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
630 return sock_sendmsg(sock, msg);
631 }
632 EXPORT_SYMBOL(kernel_sendmsg);
633
634 /*
635 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
636 */
637 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
638 struct sk_buff *skb)
639 {
640 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
641 struct scm_timestamping tss;
642 int empty = 1;
643 struct skb_shared_hwtstamps *shhwtstamps =
644 skb_hwtstamps(skb);
645
646 /* Race occurred between timestamp enabling and packet
647 receiving. Fill in the current time for now. */
648 if (need_software_tstamp && skb->tstamp.tv64 == 0)
649 __net_timestamp(skb);
650
651 if (need_software_tstamp) {
652 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
653 struct timeval tv;
654 skb_get_timestamp(skb, &tv);
655 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
656 sizeof(tv), &tv);
657 } else {
658 struct timespec ts;
659 skb_get_timestampns(skb, &ts);
660 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
661 sizeof(ts), &ts);
662 }
663 }
664
665 memset(&tss, 0, sizeof(tss));
666 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
667 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
668 empty = 0;
669 if (shhwtstamps &&
670 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
671 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
672 empty = 0;
673 if (!empty)
674 put_cmsg(msg, SOL_SOCKET,
675 SCM_TIMESTAMPING, sizeof(tss), &tss);
676 }
677 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
678
679 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
680 struct sk_buff *skb)
681 {
682 int ack;
683
684 if (!sock_flag(sk, SOCK_WIFI_STATUS))
685 return;
686 if (!skb->wifi_acked_valid)
687 return;
688
689 ack = skb->wifi_acked;
690
691 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
692 }
693 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
694
695 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
696 struct sk_buff *skb)
697 {
698 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
699 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
700 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
701 }
702
703 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
704 struct sk_buff *skb)
705 {
706 sock_recv_timestamp(msg, sk, skb);
707 sock_recv_drops(msg, sk, skb);
708 }
709 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
710
711 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
712 size_t size, int flags)
713 {
714 return sock->ops->recvmsg(sock, msg, size, flags);
715 }
716
717 int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
718 int flags)
719 {
720 int err = security_socket_recvmsg(sock, msg, size, flags);
721
722 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
723 }
724 EXPORT_SYMBOL(sock_recvmsg);
725
726 /**
727 * kernel_recvmsg - Receive a message from a socket (kernel space)
728 * @sock: The socket to receive the message from
729 * @msg: Received message
730 * @vec: Input s/g array for message data
731 * @num: Size of input s/g array
732 * @size: Number of bytes to read
733 * @flags: Message flags (MSG_DONTWAIT, etc...)
734 *
735 * On return the msg structure contains the scatter/gather array passed in the
736 * vec argument. The array is modified so that it consists of the unfilled
737 * portion of the original array.
738 *
739 * The returned value is the total number of bytes received, or an error.
740 */
741 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
742 struct kvec *vec, size_t num, size_t size, int flags)
743 {
744 mm_segment_t oldfs = get_fs();
745 int result;
746
747 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
748 set_fs(KERNEL_DS);
749 result = sock_recvmsg(sock, msg, size, flags);
750 set_fs(oldfs);
751 return result;
752 }
753 EXPORT_SYMBOL(kernel_recvmsg);
754
755 static ssize_t sock_sendpage(struct file *file, struct page *page,
756 int offset, size_t size, loff_t *ppos, int more)
757 {
758 struct socket *sock;
759 int flags;
760
761 sock = file->private_data;
762
763 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
764 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
765 flags |= more;
766
767 return kernel_sendpage(sock, page, offset, size, flags);
768 }
769
770 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
771 struct pipe_inode_info *pipe, size_t len,
772 unsigned int flags)
773 {
774 struct socket *sock = file->private_data;
775
776 if (unlikely(!sock->ops->splice_read))
777 return -EINVAL;
778
779 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
780 }
781
782 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
783 {
784 struct file *file = iocb->ki_filp;
785 struct socket *sock = file->private_data;
786 struct msghdr msg = {.msg_iter = *to,
787 .msg_iocb = iocb};
788 ssize_t res;
789
790 if (file->f_flags & O_NONBLOCK)
791 msg.msg_flags = MSG_DONTWAIT;
792
793 if (iocb->ki_pos != 0)
794 return -ESPIPE;
795
796 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
797 return 0;
798
799 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
800 *to = msg.msg_iter;
801 return res;
802 }
803
804 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
805 {
806 struct file *file = iocb->ki_filp;
807 struct socket *sock = file->private_data;
808 struct msghdr msg = {.msg_iter = *from,
809 .msg_iocb = iocb};
810 ssize_t res;
811
812 if (iocb->ki_pos != 0)
813 return -ESPIPE;
814
815 if (file->f_flags & O_NONBLOCK)
816 msg.msg_flags = MSG_DONTWAIT;
817
818 if (sock->type == SOCK_SEQPACKET)
819 msg.msg_flags |= MSG_EOR;
820
821 res = sock_sendmsg(sock, &msg);
822 *from = msg.msg_iter;
823 return res;
824 }
825
826 /*
827 * Atomic setting of ioctl hooks to avoid race
828 * with module unload.
829 */
830
831 static DEFINE_MUTEX(br_ioctl_mutex);
832 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
833
834 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
835 {
836 mutex_lock(&br_ioctl_mutex);
837 br_ioctl_hook = hook;
838 mutex_unlock(&br_ioctl_mutex);
839 }
840 EXPORT_SYMBOL(brioctl_set);
841
842 static DEFINE_MUTEX(vlan_ioctl_mutex);
843 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
844
845 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
846 {
847 mutex_lock(&vlan_ioctl_mutex);
848 vlan_ioctl_hook = hook;
849 mutex_unlock(&vlan_ioctl_mutex);
850 }
851 EXPORT_SYMBOL(vlan_ioctl_set);
852
853 static DEFINE_MUTEX(dlci_ioctl_mutex);
854 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
855
856 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
857 {
858 mutex_lock(&dlci_ioctl_mutex);
859 dlci_ioctl_hook = hook;
860 mutex_unlock(&dlci_ioctl_mutex);
861 }
862 EXPORT_SYMBOL(dlci_ioctl_set);
863
864 static long sock_do_ioctl(struct net *net, struct socket *sock,
865 unsigned int cmd, unsigned long arg)
866 {
867 int err;
868 void __user *argp = (void __user *)arg;
869
870 err = sock->ops->ioctl(sock, cmd, arg);
871
872 /*
873 * If this ioctl is unknown try to hand it down
874 * to the NIC driver.
875 */
876 if (err == -ENOIOCTLCMD)
877 err = dev_ioctl(net, cmd, argp);
878
879 return err;
880 }
881
882 /*
883 * With an ioctl, arg may well be a user mode pointer, but we don't know
884 * what to do with it - that's up to the protocol still.
885 */
886
887 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
888 {
889 struct socket *sock;
890 struct sock *sk;
891 void __user *argp = (void __user *)arg;
892 int pid, err;
893 struct net *net;
894
895 sock = file->private_data;
896 sk = sock->sk;
897 net = sock_net(sk);
898 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
899 err = dev_ioctl(net, cmd, argp);
900 } else
901 #ifdef CONFIG_WEXT_CORE
902 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
903 err = dev_ioctl(net, cmd, argp);
904 } else
905 #endif
906 switch (cmd) {
907 case FIOSETOWN:
908 case SIOCSPGRP:
909 err = -EFAULT;
910 if (get_user(pid, (int __user *)argp))
911 break;
912 f_setown(sock->file, pid, 1);
913 err = 0;
914 break;
915 case FIOGETOWN:
916 case SIOCGPGRP:
917 err = put_user(f_getown(sock->file),
918 (int __user *)argp);
919 break;
920 case SIOCGIFBR:
921 case SIOCSIFBR:
922 case SIOCBRADDBR:
923 case SIOCBRDELBR:
924 err = -ENOPKG;
925 if (!br_ioctl_hook)
926 request_module("bridge");
927
928 mutex_lock(&br_ioctl_mutex);
929 if (br_ioctl_hook)
930 err = br_ioctl_hook(net, cmd, argp);
931 mutex_unlock(&br_ioctl_mutex);
932 break;
933 case SIOCGIFVLAN:
934 case SIOCSIFVLAN:
935 err = -ENOPKG;
936 if (!vlan_ioctl_hook)
937 request_module("8021q");
938
939 mutex_lock(&vlan_ioctl_mutex);
940 if (vlan_ioctl_hook)
941 err = vlan_ioctl_hook(net, argp);
942 mutex_unlock(&vlan_ioctl_mutex);
943 break;
944 case SIOCADDDLCI:
945 case SIOCDELDLCI:
946 err = -ENOPKG;
947 if (!dlci_ioctl_hook)
948 request_module("dlci");
949
950 mutex_lock(&dlci_ioctl_mutex);
951 if (dlci_ioctl_hook)
952 err = dlci_ioctl_hook(cmd, argp);
953 mutex_unlock(&dlci_ioctl_mutex);
954 break;
955 default:
956 err = sock_do_ioctl(net, sock, cmd, arg);
957 break;
958 }
959 return err;
960 }
961
962 int sock_create_lite(int family, int type, int protocol, struct socket **res)
963 {
964 int err;
965 struct socket *sock = NULL;
966
967 err = security_socket_create(family, type, protocol, 1);
968 if (err)
969 goto out;
970
971 sock = sock_alloc();
972 if (!sock) {
973 err = -ENOMEM;
974 goto out;
975 }
976
977 sock->type = type;
978 err = security_socket_post_create(sock, family, type, protocol, 1);
979 if (err)
980 goto out_release;
981
982 out:
983 *res = sock;
984 return err;
985 out_release:
986 sock_release(sock);
987 sock = NULL;
988 goto out;
989 }
990 EXPORT_SYMBOL(sock_create_lite);
991
992 /* No kernel lock held - perfect */
993 static unsigned int sock_poll(struct file *file, poll_table *wait)
994 {
995 unsigned int busy_flag = 0;
996 struct socket *sock;
997
998 /*
999 * We can't return errors to poll, so it's either yes or no.
1000 */
1001 sock = file->private_data;
1002
1003 if (sk_can_busy_loop(sock->sk)) {
1004 /* this socket can poll_ll so tell the system call */
1005 busy_flag = POLL_BUSY_LOOP;
1006
1007 /* once, only if requested by syscall */
1008 if (wait && (wait->_key & POLL_BUSY_LOOP))
1009 sk_busy_loop(sock->sk, 1);
1010 }
1011
1012 return busy_flag | sock->ops->poll(file, sock, wait);
1013 }
1014
1015 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1016 {
1017 struct socket *sock = file->private_data;
1018
1019 return sock->ops->mmap(file, sock, vma);
1020 }
1021
1022 static int sock_close(struct inode *inode, struct file *filp)
1023 {
1024 sock_release(SOCKET_I(inode));
1025 return 0;
1026 }
1027
1028 /*
1029 * Update the socket async list
1030 *
1031 * Fasync_list locking strategy.
1032 *
1033 * 1. fasync_list is modified only under process context socket lock
1034 * i.e. under semaphore.
1035 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1036 * or under socket lock
1037 */
1038
1039 static int sock_fasync(int fd, struct file *filp, int on)
1040 {
1041 struct socket *sock = filp->private_data;
1042 struct sock *sk = sock->sk;
1043 struct socket_wq *wq;
1044
1045 if (sk == NULL)
1046 return -EINVAL;
1047
1048 lock_sock(sk);
1049 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1050 fasync_helper(fd, filp, on, &wq->fasync_list);
1051
1052 if (!wq->fasync_list)
1053 sock_reset_flag(sk, SOCK_FASYNC);
1054 else
1055 sock_set_flag(sk, SOCK_FASYNC);
1056
1057 release_sock(sk);
1058 return 0;
1059 }
1060
1061 /* This function may be called only under rcu_lock */
1062
1063 int sock_wake_async(struct socket_wq *wq, int how, int band)
1064 {
1065 if (!wq || !wq->fasync_list)
1066 return -1;
1067
1068 switch (how) {
1069 case SOCK_WAKE_WAITD:
1070 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1071 break;
1072 goto call_kill;
1073 case SOCK_WAKE_SPACE:
1074 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1075 break;
1076 /* fall through */
1077 case SOCK_WAKE_IO:
1078 call_kill:
1079 kill_fasync(&wq->fasync_list, SIGIO, band);
1080 break;
1081 case SOCK_WAKE_URG:
1082 kill_fasync(&wq->fasync_list, SIGURG, band);
1083 }
1084
1085 return 0;
1086 }
1087 EXPORT_SYMBOL(sock_wake_async);
1088
1089 int __sock_create(struct net *net, int family, int type, int protocol,
1090 struct socket **res, int kern)
1091 {
1092 int err;
1093 struct socket *sock;
1094 const struct net_proto_family *pf;
1095
1096 /*
1097 * Check protocol is in range
1098 */
1099 if (family < 0 || family >= NPROTO)
1100 return -EAFNOSUPPORT;
1101 if (type < 0 || type >= SOCK_MAX)
1102 return -EINVAL;
1103
1104 /* Compatibility.
1105
1106 This uglymoron is moved from INET layer to here to avoid
1107 deadlock in module load.
1108 */
1109 if (family == PF_INET && type == SOCK_PACKET) {
1110 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1111 current->comm);
1112 family = PF_PACKET;
1113 }
1114
1115 err = security_socket_create(family, type, protocol, kern);
1116 if (err)
1117 return err;
1118
1119 /*
1120 * Allocate the socket and allow the family to set things up. if
1121 * the protocol is 0, the family is instructed to select an appropriate
1122 * default.
1123 */
1124 sock = sock_alloc();
1125 if (!sock) {
1126 net_warn_ratelimited("socket: no more sockets\n");
1127 return -ENFILE; /* Not exactly a match, but its the
1128 closest posix thing */
1129 }
1130
1131 sock->type = type;
1132
1133 #ifdef CONFIG_MODULES
1134 /* Attempt to load a protocol module if the find failed.
1135 *
1136 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1137 * requested real, full-featured networking support upon configuration.
1138 * Otherwise module support will break!
1139 */
1140 if (rcu_access_pointer(net_families[family]) == NULL)
1141 request_module("net-pf-%d", family);
1142 #endif
1143
1144 rcu_read_lock();
1145 pf = rcu_dereference(net_families[family]);
1146 err = -EAFNOSUPPORT;
1147 if (!pf)
1148 goto out_release;
1149
1150 /*
1151 * We will call the ->create function, that possibly is in a loadable
1152 * module, so we have to bump that loadable module refcnt first.
1153 */
1154 if (!try_module_get(pf->owner))
1155 goto out_release;
1156
1157 /* Now protected by module ref count */
1158 rcu_read_unlock();
1159
1160 err = pf->create(net, sock, protocol, kern);
1161 if (err < 0)
1162 goto out_module_put;
1163
1164 /*
1165 * Now to bump the refcnt of the [loadable] module that owns this
1166 * socket at sock_release time we decrement its refcnt.
1167 */
1168 if (!try_module_get(sock->ops->owner))
1169 goto out_module_busy;
1170
1171 /*
1172 * Now that we're done with the ->create function, the [loadable]
1173 * module can have its refcnt decremented
1174 */
1175 module_put(pf->owner);
1176 err = security_socket_post_create(sock, family, type, protocol, kern);
1177 if (err)
1178 goto out_sock_release;
1179 *res = sock;
1180
1181 return 0;
1182
1183 out_module_busy:
1184 err = -EAFNOSUPPORT;
1185 out_module_put:
1186 sock->ops = NULL;
1187 module_put(pf->owner);
1188 out_sock_release:
1189 sock_release(sock);
1190 return err;
1191
1192 out_release:
1193 rcu_read_unlock();
1194 goto out_sock_release;
1195 }
1196 EXPORT_SYMBOL(__sock_create);
1197
1198 int sock_create(int family, int type, int protocol, struct socket **res)
1199 {
1200 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1201 }
1202 EXPORT_SYMBOL(sock_create);
1203
1204 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1205 {
1206 return __sock_create(net, family, type, protocol, res, 1);
1207 }
1208 EXPORT_SYMBOL(sock_create_kern);
1209
1210 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1211 {
1212 int retval;
1213 struct socket *sock;
1214 int flags;
1215
1216 /* Check the SOCK_* constants for consistency. */
1217 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1218 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1219 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1220 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1221
1222 flags = type & ~SOCK_TYPE_MASK;
1223 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1224 return -EINVAL;
1225 type &= SOCK_TYPE_MASK;
1226
1227 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1228 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1229
1230 retval = sock_create(family, type, protocol, &sock);
1231 if (retval < 0)
1232 goto out;
1233
1234 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1235 if (retval < 0)
1236 goto out_release;
1237
1238 out:
1239 /* It may be already another descriptor 8) Not kernel problem. */
1240 return retval;
1241
1242 out_release:
1243 sock_release(sock);
1244 return retval;
1245 }
1246
1247 /*
1248 * Create a pair of connected sockets.
1249 */
1250
1251 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1252 int __user *, usockvec)
1253 {
1254 struct socket *sock1, *sock2;
1255 int fd1, fd2, err;
1256 struct file *newfile1, *newfile2;
1257 int flags;
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 /*
1268 * Obtain the first socket and check if the underlying protocol
1269 * supports the socketpair call.
1270 */
1271
1272 err = sock_create(family, type, protocol, &sock1);
1273 if (err < 0)
1274 goto out;
1275
1276 err = sock_create(family, type, protocol, &sock2);
1277 if (err < 0)
1278 goto out_release_1;
1279
1280 err = sock1->ops->socketpair(sock1, sock2);
1281 if (err < 0)
1282 goto out_release_both;
1283
1284 fd1 = get_unused_fd_flags(flags);
1285 if (unlikely(fd1 < 0)) {
1286 err = fd1;
1287 goto out_release_both;
1288 }
1289
1290 fd2 = get_unused_fd_flags(flags);
1291 if (unlikely(fd2 < 0)) {
1292 err = fd2;
1293 goto out_put_unused_1;
1294 }
1295
1296 newfile1 = sock_alloc_file(sock1, flags, NULL);
1297 if (IS_ERR(newfile1)) {
1298 err = PTR_ERR(newfile1);
1299 goto out_put_unused_both;
1300 }
1301
1302 newfile2 = sock_alloc_file(sock2, flags, NULL);
1303 if (IS_ERR(newfile2)) {
1304 err = PTR_ERR(newfile2);
1305 goto out_fput_1;
1306 }
1307
1308 err = put_user(fd1, &usockvec[0]);
1309 if (err)
1310 goto out_fput_both;
1311
1312 err = put_user(fd2, &usockvec[1]);
1313 if (err)
1314 goto out_fput_both;
1315
1316 audit_fd_pair(fd1, fd2);
1317
1318 fd_install(fd1, newfile1);
1319 fd_install(fd2, newfile2);
1320 /* fd1 and fd2 may be already another descriptors.
1321 * Not kernel problem.
1322 */
1323
1324 return 0;
1325
1326 out_fput_both:
1327 fput(newfile2);
1328 fput(newfile1);
1329 put_unused_fd(fd2);
1330 put_unused_fd(fd1);
1331 goto out;
1332
1333 out_fput_1:
1334 fput(newfile1);
1335 put_unused_fd(fd2);
1336 put_unused_fd(fd1);
1337 sock_release(sock2);
1338 goto out;
1339
1340 out_put_unused_both:
1341 put_unused_fd(fd2);
1342 out_put_unused_1:
1343 put_unused_fd(fd1);
1344 out_release_both:
1345 sock_release(sock2);
1346 out_release_1:
1347 sock_release(sock1);
1348 out:
1349 return err;
1350 }
1351
1352 /*
1353 * Bind a name to a socket. Nothing much to do here since it's
1354 * the protocol's responsibility to handle the local address.
1355 *
1356 * We move the socket address to kernel space before we call
1357 * the protocol layer (having also checked the address is ok).
1358 */
1359
1360 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1361 {
1362 struct socket *sock;
1363 struct sockaddr_storage address;
1364 int err, fput_needed;
1365
1366 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1367 if (sock) {
1368 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1369 if (err >= 0) {
1370 err = security_socket_bind(sock,
1371 (struct sockaddr *)&address,
1372 addrlen);
1373 if (!err)
1374 err = sock->ops->bind(sock,
1375 (struct sockaddr *)
1376 &address, addrlen);
1377 }
1378 fput_light(sock->file, fput_needed);
1379 }
1380 return err;
1381 }
1382
1383 /*
1384 * Perform a listen. Basically, we allow the protocol to do anything
1385 * necessary for a listen, and if that works, we mark the socket as
1386 * ready for listening.
1387 */
1388
1389 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1390 {
1391 struct socket *sock;
1392 int err, fput_needed;
1393 int somaxconn;
1394
1395 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1396 if (sock) {
1397 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1398 if ((unsigned int)backlog > somaxconn)
1399 backlog = somaxconn;
1400
1401 err = security_socket_listen(sock, backlog);
1402 if (!err)
1403 err = sock->ops->listen(sock, backlog);
1404
1405 fput_light(sock->file, fput_needed);
1406 }
1407 return err;
1408 }
1409
1410 /*
1411 * For accept, we attempt to create a new socket, set up the link
1412 * with the client, wake up the client, then return the new
1413 * connected fd. We collect the address of the connector in kernel
1414 * space and move it to user at the very end. This is unclean because
1415 * we open the socket then return an error.
1416 *
1417 * 1003.1g adds the ability to recvmsg() to query connection pending
1418 * status to recvmsg. We need to add that support in a way thats
1419 * clean when we restucture accept also.
1420 */
1421
1422 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1423 int __user *, upeer_addrlen, int, flags)
1424 {
1425 struct socket *sock, *newsock;
1426 struct file *newfile;
1427 int err, len, newfd, fput_needed;
1428 struct sockaddr_storage address;
1429
1430 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1431 return -EINVAL;
1432
1433 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1434 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1435
1436 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1437 if (!sock)
1438 goto out;
1439
1440 err = -ENFILE;
1441 newsock = sock_alloc();
1442 if (!newsock)
1443 goto out_put;
1444
1445 newsock->type = sock->type;
1446 newsock->ops = sock->ops;
1447
1448 /*
1449 * We don't need try_module_get here, as the listening socket (sock)
1450 * has the protocol module (sock->ops->owner) held.
1451 */
1452 __module_get(newsock->ops->owner);
1453
1454 newfd = get_unused_fd_flags(flags);
1455 if (unlikely(newfd < 0)) {
1456 err = newfd;
1457 sock_release(newsock);
1458 goto out_put;
1459 }
1460 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1461 if (IS_ERR(newfile)) {
1462 err = PTR_ERR(newfile);
1463 put_unused_fd(newfd);
1464 sock_release(newsock);
1465 goto out_put;
1466 }
1467
1468 err = security_socket_accept(sock, newsock);
1469 if (err)
1470 goto out_fd;
1471
1472 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1473 if (err < 0)
1474 goto out_fd;
1475
1476 if (ccs_socket_post_accept_permission(sock, newsock)) {
1477 err = -EAGAIN; /* Hope less harmful than -EPERM. */
1478 goto out_fd;
1479 }
1480 if (upeer_sockaddr) {
1481 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1482 &len, 2) < 0) {
1483 err = -ECONNABORTED;
1484 goto out_fd;
1485 }
1486 err = move_addr_to_user(&address,
1487 len, upeer_sockaddr, upeer_addrlen);
1488 if (err < 0)
1489 goto out_fd;
1490 }
1491
1492 /* File flags are not inherited via accept() unlike another OSes. */
1493
1494 fd_install(newfd, newfile);
1495 err = newfd;
1496
1497 out_put:
1498 fput_light(sock->file, fput_needed);
1499 out:
1500 return err;
1501 out_fd:
1502 fput(newfile);
1503 put_unused_fd(newfd);
1504 goto out_put;
1505 }
1506
1507 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1508 int __user *, upeer_addrlen)
1509 {
1510 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1511 }
1512
1513 /*
1514 * Attempt to connect to a socket with the server address. The address
1515 * is in user space so we verify it is OK and move it to kernel space.
1516 *
1517 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1518 * break bindings
1519 *
1520 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1521 * other SEQPACKET protocols that take time to connect() as it doesn't
1522 * include the -EINPROGRESS status for such sockets.
1523 */
1524
1525 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1526 int, addrlen)
1527 {
1528 struct socket *sock;
1529 struct sockaddr_storage address;
1530 int err, fput_needed;
1531
1532 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1533 if (!sock)
1534 goto out;
1535 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1536 if (err < 0)
1537 goto out_put;
1538
1539 err =
1540 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1541 if (err)
1542 goto out_put;
1543
1544 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1545 sock->file->f_flags);
1546 out_put:
1547 fput_light(sock->file, fput_needed);
1548 out:
1549 return err;
1550 }
1551
1552 /*
1553 * Get the local address ('name') of a socket object. Move the obtained
1554 * name to user space.
1555 */
1556
1557 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1558 int __user *, usockaddr_len)
1559 {
1560 struct socket *sock;
1561 struct sockaddr_storage address;
1562 int len, err, fput_needed;
1563
1564 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1565 if (!sock)
1566 goto out;
1567
1568 err = security_socket_getsockname(sock);
1569 if (err)
1570 goto out_put;
1571
1572 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1573 if (err)
1574 goto out_put;
1575 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1576
1577 out_put:
1578 fput_light(sock->file, fput_needed);
1579 out:
1580 return err;
1581 }
1582
1583 /*
1584 * Get the remote address ('name') of a socket object. Move the obtained
1585 * name to user space.
1586 */
1587
1588 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1589 int __user *, usockaddr_len)
1590 {
1591 struct socket *sock;
1592 struct sockaddr_storage address;
1593 int len, err, fput_needed;
1594
1595 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1596 if (sock != NULL) {
1597 err = security_socket_getpeername(sock);
1598 if (err) {
1599 fput_light(sock->file, fput_needed);
1600 return err;
1601 }
1602
1603 err =
1604 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1605 1);
1606 if (!err)
1607 err = move_addr_to_user(&address, len, usockaddr,
1608 usockaddr_len);
1609 fput_light(sock->file, fput_needed);
1610 }
1611 return err;
1612 }
1613
1614 /*
1615 * Send a datagram to a given address. We move the address into kernel
1616 * space and check the user space data area is readable before invoking
1617 * the protocol.
1618 */
1619
1620 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1621 unsigned int, flags, struct sockaddr __user *, addr,
1622 int, addr_len)
1623 {
1624 struct socket *sock;
1625 struct sockaddr_storage address;
1626 int err;
1627 struct msghdr msg;
1628 struct iovec iov;
1629 int fput_needed;
1630
1631 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1632 if (unlikely(err))
1633 return err;
1634 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1635 if (!sock)
1636 goto out;
1637
1638 msg.msg_name = NULL;
1639 msg.msg_control = NULL;
1640 msg.msg_controllen = 0;
1641 msg.msg_namelen = 0;
1642 if (addr) {
1643 err = move_addr_to_kernel(addr, addr_len, &address);
1644 if (err < 0)
1645 goto out_put;
1646 msg.msg_name = (struct sockaddr *)&address;
1647 msg.msg_namelen = addr_len;
1648 }
1649 if (sock->file->f_flags & O_NONBLOCK)
1650 flags |= MSG_DONTWAIT;
1651 msg.msg_flags = flags;
1652 err = sock_sendmsg(sock, &msg);
1653
1654 out_put:
1655 fput_light(sock->file, fput_needed);
1656 out:
1657 return err;
1658 }
1659
1660 /*
1661 * Send a datagram down a socket.
1662 */
1663
1664 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1665 unsigned int, flags)
1666 {
1667 return sys_sendto(fd, buff, len, flags, NULL, 0);
1668 }
1669
1670 /*
1671 * Receive a frame from the socket and optionally record the address of the
1672 * sender. We verify the buffers are writable and if needed move the
1673 * sender address from kernel to user space.
1674 */
1675
1676 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1677 unsigned int, flags, struct sockaddr __user *, addr,
1678 int __user *, addr_len)
1679 {
1680 struct socket *sock;
1681 struct iovec iov;
1682 struct msghdr msg;
1683 struct sockaddr_storage address;
1684 int err, err2;
1685 int fput_needed;
1686
1687 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1688 if (unlikely(err))
1689 return err;
1690 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1691 if (!sock)
1692 goto out;
1693
1694 msg.msg_control = NULL;
1695 msg.msg_controllen = 0;
1696 /* Save some cycles and don't copy the address if not needed */
1697 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1698 /* We assume all kernel code knows the size of sockaddr_storage */
1699 msg.msg_namelen = 0;
1700 msg.msg_iocb = NULL;
1701 if (sock->file->f_flags & O_NONBLOCK)
1702 flags |= MSG_DONTWAIT;
1703 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1704
1705 if (err >= 0 && addr != NULL) {
1706 err2 = move_addr_to_user(&address,
1707 msg.msg_namelen, addr, addr_len);
1708 if (err2 < 0)
1709 err = err2;
1710 }
1711
1712 fput_light(sock->file, fput_needed);
1713 out:
1714 return err;
1715 }
1716
1717 /*
1718 * Receive a datagram from a socket.
1719 */
1720
1721 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1722 unsigned int, flags)
1723 {
1724 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1725 }
1726
1727 /*
1728 * Set a socket option. Because we don't know the option lengths we have
1729 * to pass the user mode parameter for the protocols to sort out.
1730 */
1731
1732 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1733 char __user *, optval, int, optlen)
1734 {
1735 int err, fput_needed;
1736 struct socket *sock;
1737
1738 if (optlen < 0)
1739 return -EINVAL;
1740
1741 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1742 if (sock != NULL) {
1743 err = security_socket_setsockopt(sock, level, optname);
1744 if (err)
1745 goto out_put;
1746
1747 if (level == SOL_SOCKET)
1748 err =
1749 sock_setsockopt(sock, level, optname, optval,
1750 optlen);
1751 else
1752 err =
1753 sock->ops->setsockopt(sock, level, optname, optval,
1754 optlen);
1755 out_put:
1756 fput_light(sock->file, fput_needed);
1757 }
1758 return err;
1759 }
1760
1761 /*
1762 * Get a socket option. Because we don't know the option lengths we have
1763 * to pass a user mode parameter for the protocols to sort out.
1764 */
1765
1766 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1767 char __user *, optval, int __user *, optlen)
1768 {
1769 int err, fput_needed;
1770 struct socket *sock;
1771
1772 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1773 if (sock != NULL) {
1774 err = security_socket_getsockopt(sock, level, optname);
1775 if (err)
1776 goto out_put;
1777
1778 if (level == SOL_SOCKET)
1779 err =
1780 sock_getsockopt(sock, level, optname, optval,
1781 optlen);
1782 else
1783 err =
1784 sock->ops->getsockopt(sock, level, optname, optval,
1785 optlen);
1786 out_put:
1787 fput_light(sock->file, fput_needed);
1788 }
1789 return err;
1790 }
1791
1792 /*
1793 * Shutdown a socket.
1794 */
1795
1796 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1797 {
1798 int err, fput_needed;
1799 struct socket *sock;
1800
1801 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1802 if (sock != NULL) {
1803 err = security_socket_shutdown(sock, how);
1804 if (!err)
1805 err = sock->ops->shutdown(sock, how);
1806 fput_light(sock->file, fput_needed);
1807 }
1808 return err;
1809 }
1810
1811 /* A couple of helpful macros for getting the address of the 32/64 bit
1812 * fields which are the same type (int / unsigned) on our platforms.
1813 */
1814 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1815 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1816 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1817
1818 struct used_address {
1819 struct sockaddr_storage name;
1820 unsigned int name_len;
1821 };
1822
1823 static int copy_msghdr_from_user(struct msghdr *kmsg,
1824 struct user_msghdr __user *umsg,
1825 struct sockaddr __user **save_addr,
1826 struct iovec **iov)
1827 {
1828 struct sockaddr __user *uaddr;
1829 struct iovec __user *uiov;
1830 size_t nr_segs;
1831 ssize_t err;
1832
1833 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1834 __get_user(uaddr, &umsg->msg_name) ||
1835 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1836 __get_user(uiov, &umsg->msg_iov) ||
1837 __get_user(nr_segs, &umsg->msg_iovlen) ||
1838 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1839 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1840 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1841 return -EFAULT;
1842
1843 if (!uaddr)
1844 kmsg->msg_namelen = 0;
1845
1846 if (kmsg->msg_namelen < 0)
1847 return -EINVAL;
1848
1849 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1850 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1851
1852 if (save_addr)
1853 *save_addr = uaddr;
1854
1855 if (uaddr && kmsg->msg_namelen) {
1856 if (!save_addr) {
1857 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1858 kmsg->msg_name);
1859 if (err < 0)
1860 return err;
1861 }
1862 } else {
1863 kmsg->msg_name = NULL;
1864 kmsg->msg_namelen = 0;
1865 }
1866
1867 if (nr_segs > UIO_MAXIOV)
1868 return -EMSGSIZE;
1869
1870 kmsg->msg_iocb = NULL;
1871
1872 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1873 UIO_FASTIOV, iov, &kmsg->msg_iter);
1874 }
1875
1876 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1877 struct msghdr *msg_sys, unsigned int flags,
1878 struct used_address *used_address,
1879 unsigned int allowed_msghdr_flags)
1880 {
1881 struct compat_msghdr __user *msg_compat =
1882 (struct compat_msghdr __user *)msg;
1883 struct sockaddr_storage address;
1884 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1885 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1886 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1887 /* 20 is size of ipv6_pktinfo */
1888 unsigned char *ctl_buf = ctl;
1889 int ctl_len;
1890 ssize_t err;
1891
1892 msg_sys->msg_name = &address;
1893
1894 if (MSG_CMSG_COMPAT & flags)
1895 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1896 else
1897 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1898 if (err < 0)
1899 return err;
1900
1901 err = -ENOBUFS;
1902
1903 if (msg_sys->msg_controllen > INT_MAX)
1904 goto out_freeiov;
1905 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1906 ctl_len = msg_sys->msg_controllen;
1907 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1908 err =
1909 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1910 sizeof(ctl));
1911 if (err)
1912 goto out_freeiov;
1913 ctl_buf = msg_sys->msg_control;
1914 ctl_len = msg_sys->msg_controllen;
1915 } else if (ctl_len) {
1916 if (ctl_len > sizeof(ctl)) {
1917 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1918 if (ctl_buf == NULL)
1919 goto out_freeiov;
1920 }
1921 err = -EFAULT;
1922 /*
1923 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1924 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1925 * checking falls down on this.
1926 */
1927 if (copy_from_user(ctl_buf,
1928 (void __user __force *)msg_sys->msg_control,
1929 ctl_len))
1930 goto out_freectl;
1931 msg_sys->msg_control = ctl_buf;
1932 }
1933 msg_sys->msg_flags = flags;
1934
1935 if (sock->file->f_flags & O_NONBLOCK)
1936 msg_sys->msg_flags |= MSG_DONTWAIT;
1937 /*
1938 * If this is sendmmsg() and current destination address is same as
1939 * previously succeeded address, omit asking LSM's decision.
1940 * used_address->name_len is initialized to UINT_MAX so that the first
1941 * destination address never matches.
1942 */
1943 if (used_address && msg_sys->msg_name &&
1944 used_address->name_len == msg_sys->msg_namelen &&
1945 !memcmp(&used_address->name, msg_sys->msg_name,
1946 used_address->name_len)) {
1947 err = sock_sendmsg_nosec(sock, msg_sys);
1948 goto out_freectl;
1949 }
1950 err = sock_sendmsg(sock, msg_sys);
1951 /*
1952 * If this is sendmmsg() and sending to current destination address was
1953 * successful, remember it.
1954 */
1955 if (used_address && err >= 0) {
1956 used_address->name_len = msg_sys->msg_namelen;
1957 if (msg_sys->msg_name)
1958 memcpy(&used_address->name, msg_sys->msg_name,
1959 used_address->name_len);
1960 }
1961
1962 out_freectl:
1963 if (ctl_buf != ctl)
1964 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1965 out_freeiov:
1966 kfree(iov);
1967 return err;
1968 }
1969
1970 /*
1971 * BSD sendmsg interface
1972 */
1973
1974 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1975 {
1976 int fput_needed, err;
1977 struct msghdr msg_sys;
1978 struct socket *sock;
1979
1980 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1981 if (!sock)
1982 goto out;
1983
1984 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
1985
1986 fput_light(sock->file, fput_needed);
1987 out:
1988 return err;
1989 }
1990
1991 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1992 {
1993 if (flags & MSG_CMSG_COMPAT)
1994 return -EINVAL;
1995 return __sys_sendmsg(fd, msg, flags);
1996 }
1997
1998 /*
1999 * Linux sendmmsg interface
2000 */
2001
2002 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2003 unsigned int flags)
2004 {
2005 int fput_needed, err, datagrams;
2006 struct socket *sock;
2007 struct mmsghdr __user *entry;
2008 struct compat_mmsghdr __user *compat_entry;
2009 struct msghdr msg_sys;
2010 struct used_address used_address;
2011 unsigned int oflags = flags;
2012
2013 if (vlen > UIO_MAXIOV)
2014 vlen = UIO_MAXIOV;
2015
2016 datagrams = 0;
2017
2018 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2019 if (!sock)
2020 return err;
2021
2022 used_address.name_len = UINT_MAX;
2023 entry = mmsg;
2024 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2025 err = 0;
2026 flags |= MSG_BATCH;
2027
2028 while (datagrams < vlen) {
2029 if (datagrams == vlen - 1)
2030 flags = oflags;
2031
2032 if (MSG_CMSG_COMPAT & flags) {
2033 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2034 &msg_sys, flags, &used_address, MSG_EOR);
2035 if (err < 0)
2036 break;
2037 err = __put_user(err, &compat_entry->msg_len);
2038 ++compat_entry;
2039 } else {
2040 err = ___sys_sendmsg(sock,
2041 (struct user_msghdr __user *)entry,
2042 &msg_sys, flags, &used_address, MSG_EOR);
2043 if (err < 0)
2044 break;
2045 err = put_user(err, &entry->msg_len);
2046 ++entry;
2047 }
2048
2049 if (err)
2050 break;
2051 ++datagrams;
2052 cond_resched();
2053 }
2054
2055 fput_light(sock->file, fput_needed);
2056
2057 /* We only return an error if no datagrams were able to be sent */
2058 if (datagrams != 0)
2059 return datagrams;
2060
2061 return err;
2062 }
2063
2064 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2065 unsigned int, vlen, unsigned int, flags)
2066 {
2067 if (flags & MSG_CMSG_COMPAT)
2068 return -EINVAL;
2069 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2070 }
2071
2072 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2073 struct msghdr *msg_sys, unsigned int flags, int nosec)
2074 {
2075 struct compat_msghdr __user *msg_compat =
2076 (struct compat_msghdr __user *)msg;
2077 struct iovec iovstack[UIO_FASTIOV];
2078 struct iovec *iov = iovstack;
2079 unsigned long cmsg_ptr;
2080 int total_len, len;
2081 ssize_t err;
2082
2083 /* kernel mode address */
2084 struct sockaddr_storage addr;
2085
2086 /* user mode address pointers */
2087 struct sockaddr __user *uaddr;
2088 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2089
2090 msg_sys->msg_name = &addr;
2091
2092 if (MSG_CMSG_COMPAT & flags)
2093 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2094 else
2095 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2096 if (err < 0)
2097 return err;
2098 total_len = iov_iter_count(&msg_sys->msg_iter);
2099
2100 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2101 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2102
2103 /* We assume all kernel code knows the size of sockaddr_storage */
2104 msg_sys->msg_namelen = 0;
2105
2106 if (sock->file->f_flags & O_NONBLOCK)
2107 flags |= MSG_DONTWAIT;
2108 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2109 total_len, flags);
2110 if (err < 0)
2111 goto out_freeiov;
2112 len = err;
2113
2114 if (uaddr != NULL) {
2115 err = move_addr_to_user(&addr,
2116 msg_sys->msg_namelen, uaddr,
2117 uaddr_len);
2118 if (err < 0)
2119 goto out_freeiov;
2120 }
2121 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2122 COMPAT_FLAGS(msg));
2123 if (err)
2124 goto out_freeiov;
2125 if (MSG_CMSG_COMPAT & flags)
2126 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2127 &msg_compat->msg_controllen);
2128 else
2129 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2130 &msg->msg_controllen);
2131 if (err)
2132 goto out_freeiov;
2133 err = len;
2134
2135 out_freeiov:
2136 kfree(iov);
2137 return err;
2138 }
2139
2140 /*
2141 * BSD recvmsg interface
2142 */
2143
2144 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2145 {
2146 int fput_needed, err;
2147 struct msghdr msg_sys;
2148 struct socket *sock;
2149
2150 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2151 if (!sock)
2152 goto out;
2153
2154 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2155
2156 fput_light(sock->file, fput_needed);
2157 out:
2158 return err;
2159 }
2160
2161 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2162 unsigned int, flags)
2163 {
2164 if (flags & MSG_CMSG_COMPAT)
2165 return -EINVAL;
2166 return __sys_recvmsg(fd, msg, flags);
2167 }
2168
2169 /*
2170 * Linux recvmmsg interface
2171 */
2172
2173 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2174 unsigned int flags, struct timespec *timeout)
2175 {
2176 int fput_needed, err, datagrams;
2177 struct socket *sock;
2178 struct mmsghdr __user *entry;
2179 struct compat_mmsghdr __user *compat_entry;
2180 struct msghdr msg_sys;
2181 struct timespec end_time;
2182
2183 if (timeout &&
2184 poll_select_set_timeout(&end_time, timeout->tv_sec,
2185 timeout->tv_nsec))
2186 return -EINVAL;
2187
2188 datagrams = 0;
2189
2190 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2191 if (!sock)
2192 return err;
2193
2194 err = sock_error(sock->sk);
2195 if (err)
2196 goto out_put;
2197
2198 entry = mmsg;
2199 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2200
2201 while (datagrams < vlen) {
2202 /*
2203 * No need to ask LSM for more than the first datagram.
2204 */
2205 if (MSG_CMSG_COMPAT & flags) {
2206 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2207 &msg_sys, flags & ~MSG_WAITFORONE,
2208 datagrams);
2209 if (err < 0)
2210 break;
2211 err = __put_user(err, &compat_entry->msg_len);
2212 ++compat_entry;
2213 } else {
2214 err = ___sys_recvmsg(sock,
2215 (struct user_msghdr __user *)entry,
2216 &msg_sys, flags & ~MSG_WAITFORONE,
2217 datagrams);
2218 if (err < 0)
2219 break;
2220 err = put_user(err, &entry->msg_len);
2221 ++entry;
2222 }
2223
2224 if (err)
2225 break;
2226 ++datagrams;
2227
2228 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2229 if (flags & MSG_WAITFORONE)
2230 flags |= MSG_DONTWAIT;
2231
2232 if (timeout) {
2233 ktime_get_ts(timeout);
2234 *timeout = timespec_sub(end_time, *timeout);
2235 if (timeout->tv_sec < 0) {
2236 timeout->tv_sec = timeout->tv_nsec = 0;
2237 break;
2238 }
2239
2240 /* Timeout, return less than vlen datagrams */
2241 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2242 break;
2243 }
2244
2245 /* Out of band data, return right away */
2246 if (msg_sys.msg_flags & MSG_OOB)
2247 break;
2248 cond_resched();
2249 }
2250
2251 if (err == 0)
2252 goto out_put;
2253
2254 if (datagrams == 0) {
2255 datagrams = err;
2256 goto out_put;
2257 }
2258
2259 /*
2260 * We may return less entries than requested (vlen) if the
2261 * sock is non block and there aren't enough datagrams...
2262 */
2263 if (err != -EAGAIN) {
2264 /*
2265 * ... or if recvmsg returns an error after we
2266 * received some datagrams, where we record the
2267 * error to return on the next call or if the
2268 * app asks about it using getsockopt(SO_ERROR).
2269 */
2270 sock->sk->sk_err = -err;
2271 }
2272 out_put:
2273 fput_light(sock->file, fput_needed);
2274
2275 return datagrams;
2276 }
2277
2278 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2279 unsigned int, vlen, unsigned int, flags,
2280 struct timespec __user *, timeout)
2281 {
2282 int datagrams;
2283 struct timespec timeout_sys;
2284
2285 if (flags & MSG_CMSG_COMPAT)
2286 return -EINVAL;
2287
2288 if (!timeout)
2289 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2290
2291 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2292 return -EFAULT;
2293
2294 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2295
2296 if (datagrams > 0 &&
2297 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2298 datagrams = -EFAULT;
2299
2300 return datagrams;
2301 }
2302
2303 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2304 /* Argument list sizes for sys_socketcall */
2305 #define AL(x) ((x) * sizeof(unsigned long))
2306 static const unsigned char nargs[21] = {
2307 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2308 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2309 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2310 AL(4), AL(5), AL(4)
2311 };
2312
2313 #undef AL
2314
2315 /*
2316 * System call vectors.
2317 *
2318 * Argument checking cleaned up. Saved 20% in size.
2319 * This function doesn't need to set the kernel lock because
2320 * it is set by the callees.
2321 */
2322
2323 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2324 {
2325 unsigned long a[AUDITSC_ARGS];
2326 unsigned long a0, a1;
2327 int err;
2328 unsigned int len;
2329
2330 if (call < 1 || call > SYS_SENDMMSG)
2331 return -EINVAL;
2332
2333 len = nargs[call];
2334 if (len > sizeof(a))
2335 return -EINVAL;
2336
2337 /* copy_from_user should be SMP safe. */
2338 if (copy_from_user(a, args, len))
2339 return -EFAULT;
2340
2341 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2342 if (err)
2343 return err;
2344
2345 a0 = a[0];
2346 a1 = a[1];
2347
2348 switch (call) {
2349 case SYS_SOCKET:
2350 err = sys_socket(a0, a1, a[2]);
2351 break;
2352 case SYS_BIND:
2353 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2354 break;
2355 case SYS_CONNECT:
2356 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2357 break;
2358 case SYS_LISTEN:
2359 err = sys_listen(a0, a1);
2360 break;
2361 case SYS_ACCEPT:
2362 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2363 (int __user *)a[2], 0);
2364 break;
2365 case SYS_GETSOCKNAME:
2366 err =
2367 sys_getsockname(a0, (struct sockaddr __user *)a1,
2368 (int __user *)a[2]);
2369 break;
2370 case SYS_GETPEERNAME:
2371 err =
2372 sys_getpeername(a0, (struct sockaddr __user *)a1,
2373 (int __user *)a[2]);
2374 break;
2375 case SYS_SOCKETPAIR:
2376 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2377 break;
2378 case SYS_SEND:
2379 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2380 break;
2381 case SYS_SENDTO:
2382 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2383 (struct sockaddr __user *)a[4], a[5]);
2384 break;
2385 case SYS_RECV:
2386 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2387 break;
2388 case SYS_RECVFROM:
2389 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2390 (struct sockaddr __user *)a[4],
2391 (int __user *)a[5]);
2392 break;
2393 case SYS_SHUTDOWN:
2394 err = sys_shutdown(a0, a1);
2395 break;
2396 case SYS_SETSOCKOPT:
2397 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2398 break;
2399 case SYS_GETSOCKOPT:
2400 err =
2401 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2402 (int __user *)a[4]);
2403 break;
2404 case SYS_SENDMSG:
2405 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2406 break;
2407 case SYS_SENDMMSG:
2408 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2409 break;
2410 case SYS_RECVMSG:
2411 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2412 break;
2413 case SYS_RECVMMSG:
2414 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2415 (struct timespec __user *)a[4]);
2416 break;
2417 case SYS_ACCEPT4:
2418 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2419 (int __user *)a[2], a[3]);
2420 break;
2421 default:
2422 err = -EINVAL;
2423 break;
2424 }
2425 return err;
2426 }
2427
2428 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2429
2430 /**
2431 * sock_register - add a socket protocol handler
2432 * @ops: description of protocol
2433 *
2434 * This function is called by a protocol handler that wants to
2435 * advertise its address family, and have it linked into the
2436 * socket interface. The value ops->family corresponds to the
2437 * socket system call protocol family.
2438 */
2439 int sock_register(const struct net_proto_family *ops)
2440 {
2441 int err;
2442
2443 if (ops->family >= NPROTO) {
2444 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2445 return -ENOBUFS;
2446 }
2447
2448 spin_lock(&net_family_lock);
2449 if (rcu_dereference_protected(net_families[ops->family],
2450 lockdep_is_held(&net_family_lock)))
2451 err = -EEXIST;
2452 else {
2453 rcu_assign_pointer(net_families[ops->family], ops);
2454 err = 0;
2455 }
2456 spin_unlock(&net_family_lock);
2457
2458 pr_info("NET: Registered protocol family %d\n", ops->family);
2459 return err;
2460 }
2461 EXPORT_SYMBOL(sock_register);
2462
2463 /**
2464 * sock_unregister - remove a protocol handler
2465 * @family: protocol family to remove
2466 *
2467 * This function is called by a protocol handler that wants to
2468 * remove its address family, and have it unlinked from the
2469 * new socket creation.
2470 *
2471 * If protocol handler is a module, then it can use module reference
2472 * counts to protect against new references. If protocol handler is not
2473 * a module then it needs to provide its own protection in
2474 * the ops->create routine.
2475 */
2476 void sock_unregister(int family)
2477 {
2478 BUG_ON(family < 0 || family >= NPROTO);
2479
2480 spin_lock(&net_family_lock);
2481 RCU_INIT_POINTER(net_families[family], NULL);
2482 spin_unlock(&net_family_lock);
2483
2484 synchronize_rcu();
2485
2486 pr_info("NET: Unregistered protocol family %d\n", family);
2487 }
2488 EXPORT_SYMBOL(sock_unregister);
2489
2490 static int __init sock_init(void)
2491 {
2492 int err;
2493 /*
2494 * Initialize the network sysctl infrastructure.
2495 */
2496 err = net_sysctl_init();
2497 if (err)
2498 goto out;
2499
2500 /*
2501 * Initialize skbuff SLAB cache
2502 */
2503 skb_init();
2504
2505 /*
2506 * Initialize the protocols module.
2507 */
2508
2509 init_inodecache();
2510
2511 err = register_filesystem(&sock_fs_type);
2512 if (err)
2513 goto out_fs;
2514 sock_mnt = kern_mount(&sock_fs_type);
2515 if (IS_ERR(sock_mnt)) {
2516 err = PTR_ERR(sock_mnt);
2517 goto out_mount;
2518 }
2519
2520 /* The real protocol initialization is performed in later initcalls.
2521 */
2522
2523 #ifdef CONFIG_NETFILTER
2524 err = netfilter_init();
2525 if (err)
2526 goto out;
2527 #endif
2528
2529 ptp_classifier_init();
2530
2531 out:
2532 return err;
2533
2534 out_mount:
2535 unregister_filesystem(&sock_fs_type);
2536 out_fs:
2537 goto out;
2538 }
2539
2540 core_initcall(sock_init); /* early initcall */
2541
2542 #ifdef CONFIG_PROC_FS
2543 void socket_seq_show(struct seq_file *seq)
2544 {
2545 int cpu;
2546 int counter = 0;
2547
2548 for_each_possible_cpu(cpu)
2549 counter += per_cpu(sockets_in_use, cpu);
2550
2551 /* It can be negative, by the way. 8) */
2552 if (counter < 0)
2553 counter = 0;
2554
2555 seq_printf(seq, "sockets: used %d\n", counter);
2556 }
2557 #endif /* CONFIG_PROC_FS */
2558
2559 #ifdef CONFIG_COMPAT
2560 static int do_siocgstamp(struct net *net, struct socket *sock,
2561 unsigned int cmd, void __user *up)
2562 {
2563 mm_segment_t old_fs = get_fs();
2564 struct timeval ktv;
2565 int err;
2566
2567 set_fs(KERNEL_DS);
2568 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2569 set_fs(old_fs);
2570 if (!err)
2571 err = compat_put_timeval(&ktv, up);
2572
2573 return err;
2574 }
2575
2576 static int do_siocgstampns(struct net *net, struct socket *sock,
2577 unsigned int cmd, void __user *up)
2578 {
2579 mm_segment_t old_fs = get_fs();
2580 struct timespec kts;
2581 int err;
2582
2583 set_fs(KERNEL_DS);
2584 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2585 set_fs(old_fs);
2586 if (!err)
2587 err = compat_put_timespec(&kts, up);
2588
2589 return err;
2590 }
2591
2592 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2593 {
2594 struct ifreq __user *uifr;
2595 int err;
2596
2597 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2598 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2599 return -EFAULT;
2600
2601 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2602 if (err)
2603 return err;
2604
2605 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2606 return -EFAULT;
2607
2608 return 0;
2609 }
2610
2611 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2612 {
2613 struct compat_ifconf ifc32;
2614 struct ifconf ifc;
2615 struct ifconf __user *uifc;
2616 struct compat_ifreq __user *ifr32;
2617 struct ifreq __user *ifr;
2618 unsigned int i, j;
2619 int err;
2620
2621 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2622 return -EFAULT;
2623
2624 memset(&ifc, 0, sizeof(ifc));
2625 if (ifc32.ifcbuf == 0) {
2626 ifc32.ifc_len = 0;
2627 ifc.ifc_len = 0;
2628 ifc.ifc_req = NULL;
2629 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2630 } else {
2631 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2632 sizeof(struct ifreq);
2633 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2634 ifc.ifc_len = len;
2635 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2636 ifr32 = compat_ptr(ifc32.ifcbuf);
2637 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2638 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2639 return -EFAULT;
2640 ifr++;
2641 ifr32++;
2642 }
2643 }
2644 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2645 return -EFAULT;
2646
2647 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2648 if (err)
2649 return err;
2650
2651 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2652 return -EFAULT;
2653
2654 ifr = ifc.ifc_req;
2655 ifr32 = compat_ptr(ifc32.ifcbuf);
2656 for (i = 0, j = 0;
2657 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2658 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2659 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2660 return -EFAULT;
2661 ifr32++;
2662 ifr++;
2663 }
2664
2665 if (ifc32.ifcbuf == 0) {
2666 /* Translate from 64-bit structure multiple to
2667 * a 32-bit one.
2668 */
2669 i = ifc.ifc_len;
2670 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2671 ifc32.ifc_len = i;
2672 } else {
2673 ifc32.ifc_len = i;
2674 }
2675 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2676 return -EFAULT;
2677
2678 return 0;
2679 }
2680
2681 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2682 {
2683 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2684 bool convert_in = false, convert_out = false;
2685 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2686 struct ethtool_rxnfc __user *rxnfc;
2687 struct ifreq __user *ifr;
2688 u32 rule_cnt = 0, actual_rule_cnt;
2689 u32 ethcmd;
2690 u32 data;
2691 int ret;
2692
2693 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2694 return -EFAULT;
2695
2696 compat_rxnfc = compat_ptr(data);
2697
2698 if (get_user(ethcmd, &compat_rxnfc->cmd))
2699 return -EFAULT;
2700
2701 /* Most ethtool structures are defined without padding.
2702 * Unfortunately struct ethtool_rxnfc is an exception.
2703 */
2704 switch (ethcmd) {
2705 default:
2706 break;
2707 case ETHTOOL_GRXCLSRLALL:
2708 /* Buffer size is variable */
2709 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2710 return -EFAULT;
2711 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2712 return -ENOMEM;
2713 buf_size += rule_cnt * sizeof(u32);
2714 /* fall through */
2715 case ETHTOOL_GRXRINGS:
2716 case ETHTOOL_GRXCLSRLCNT:
2717 case ETHTOOL_GRXCLSRULE:
2718 case ETHTOOL_SRXCLSRLINS:
2719 convert_out = true;
2720 /* fall through */
2721 case ETHTOOL_SRXCLSRLDEL:
2722 buf_size += sizeof(struct ethtool_rxnfc);
2723 convert_in = true;
2724 break;
2725 }
2726
2727 ifr = compat_alloc_user_space(buf_size);
2728 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2729
2730 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2731 return -EFAULT;
2732
2733 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2734 &ifr->ifr_ifru.ifru_data))
2735 return -EFAULT;
2736
2737 if (convert_in) {
2738 /* We expect there to be holes between fs.m_ext and
2739 * fs.ring_cookie and at the end of fs, but nowhere else.
2740 */
2741 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2742 sizeof(compat_rxnfc->fs.m_ext) !=
2743 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2744 sizeof(rxnfc->fs.m_ext));
2745 BUILD_BUG_ON(
2746 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2747 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2748 offsetof(struct ethtool_rxnfc, fs.location) -
2749 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2750
2751 if (copy_in_user(rxnfc, compat_rxnfc,
2752 (void __user *)(&rxnfc->fs.m_ext + 1) -
2753 (void __user *)rxnfc) ||
2754 copy_in_user(&rxnfc->fs.ring_cookie,
2755 &compat_rxnfc->fs.ring_cookie,
2756 (void __user *)(&rxnfc->fs.location + 1) -
2757 (void __user *)&rxnfc->fs.ring_cookie) ||
2758 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2759 sizeof(rxnfc->rule_cnt)))
2760 return -EFAULT;
2761 }
2762
2763 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2764 if (ret)
2765 return ret;
2766
2767 if (convert_out) {
2768 if (copy_in_user(compat_rxnfc, rxnfc,
2769 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2770 (const void __user *)rxnfc) ||
2771 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2772 &rxnfc->fs.ring_cookie,
2773 (const void __user *)(&rxnfc->fs.location + 1) -
2774 (const void __user *)&rxnfc->fs.ring_cookie) ||
2775 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2776 sizeof(rxnfc->rule_cnt)))
2777 return -EFAULT;
2778
2779 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2780 /* As an optimisation, we only copy the actual
2781 * number of rules that the underlying
2782 * function returned. Since Mallory might
2783 * change the rule count in user memory, we
2784 * check that it is less than the rule count
2785 * originally given (as the user buffer size),
2786 * which has been range-checked.
2787 */
2788 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2789 return -EFAULT;
2790 if (actual_rule_cnt < rule_cnt)
2791 rule_cnt = actual_rule_cnt;
2792 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2793 &rxnfc->rule_locs[0],
2794 rule_cnt * sizeof(u32)))
2795 return -EFAULT;
2796 }
2797 }
2798
2799 return 0;
2800 }
2801
2802 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2803 {
2804 void __user *uptr;
2805 compat_uptr_t uptr32;
2806 struct ifreq __user *uifr;
2807
2808 uifr = compat_alloc_user_space(sizeof(*uifr));
2809 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2810 return -EFAULT;
2811
2812 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2813 return -EFAULT;
2814
2815 uptr = compat_ptr(uptr32);
2816
2817 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2818 return -EFAULT;
2819
2820 return dev_ioctl(net, SIOCWANDEV, uifr);
2821 }
2822
2823 static int bond_ioctl(struct net *net, unsigned int cmd,
2824 struct compat_ifreq __user *ifr32)
2825 {
2826 struct ifreq kifr;
2827 mm_segment_t old_fs;
2828 int err;
2829
2830 switch (cmd) {
2831 case SIOCBONDENSLAVE:
2832 case SIOCBONDRELEASE:
2833 case SIOCBONDSETHWADDR:
2834 case SIOCBONDCHANGEACTIVE:
2835 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2836 return -EFAULT;
2837
2838 old_fs = get_fs();
2839 set_fs(KERNEL_DS);
2840 err = dev_ioctl(net, cmd,
2841 (struct ifreq __user __force *) &kifr);
2842 set_fs(old_fs);
2843
2844 return err;
2845 default:
2846 return -ENOIOCTLCMD;
2847 }
2848 }
2849
2850 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2851 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2852 struct compat_ifreq __user *u_ifreq32)
2853 {
2854 struct ifreq __user *u_ifreq64;
2855 char tmp_buf[IFNAMSIZ];
2856 void __user *data64;
2857 u32 data32;
2858
2859 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2860 IFNAMSIZ))
2861 return -EFAULT;
2862 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2863 return -EFAULT;
2864 data64 = compat_ptr(data32);
2865
2866 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2867
2868 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2869 IFNAMSIZ))
2870 return -EFAULT;
2871 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2872 return -EFAULT;
2873
2874 return dev_ioctl(net, cmd, u_ifreq64);
2875 }
2876
2877 static int dev_ifsioc(struct net *net, struct socket *sock,
2878 unsigned int cmd, struct compat_ifreq __user *uifr32)
2879 {
2880 struct ifreq __user *uifr;
2881 int err;
2882
2883 uifr = compat_alloc_user_space(sizeof(*uifr));
2884 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2885 return -EFAULT;
2886
2887 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2888
2889 if (!err) {
2890 switch (cmd) {
2891 case SIOCGIFFLAGS:
2892 case SIOCGIFMETRIC:
2893 case SIOCGIFMTU:
2894 case SIOCGIFMEM:
2895 case SIOCGIFHWADDR:
2896 case SIOCGIFINDEX:
2897 case SIOCGIFADDR:
2898 case SIOCGIFBRDADDR:
2899 case SIOCGIFDSTADDR:
2900 case SIOCGIFNETMASK:
2901 case SIOCGIFPFLAGS:
2902 case SIOCGIFTXQLEN:
2903 case SIOCGMIIPHY:
2904 case SIOCGMIIREG:
2905 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2906 err = -EFAULT;
2907 break;
2908 }
2909 }
2910 return err;
2911 }
2912
2913 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2914 struct compat_ifreq __user *uifr32)
2915 {
2916 struct ifreq ifr;
2917 struct compat_ifmap __user *uifmap32;
2918 mm_segment_t old_fs;
2919 int err;
2920
2921 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2922 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2923 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2924 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2925 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2926 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2927 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2928 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2929 if (err)
2930 return -EFAULT;
2931
2932 old_fs = get_fs();
2933 set_fs(KERNEL_DS);
2934 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2935 set_fs(old_fs);
2936
2937 if (cmd == SIOCGIFMAP && !err) {
2938 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2939 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2940 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2941 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2942 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2943 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2944 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2945 if (err)
2946 err = -EFAULT;
2947 }
2948 return err;
2949 }
2950
2951 struct rtentry32 {
2952 u32 rt_pad1;
2953 struct sockaddr rt_dst; /* target address */
2954 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2955 struct sockaddr rt_genmask; /* target network mask (IP) */
2956 unsigned short rt_flags;
2957 short rt_pad2;
2958 u32 rt_pad3;
2959 unsigned char rt_tos;
2960 unsigned char rt_class;
2961 short rt_pad4;
2962 short rt_metric; /* +1 for binary compatibility! */
2963 /* char * */ u32 rt_dev; /* forcing the device at add */
2964 u32 rt_mtu; /* per route MTU/Window */
2965 u32 rt_window; /* Window clamping */
2966 unsigned short rt_irtt; /* Initial RTT */
2967 };
2968
2969 struct in6_rtmsg32 {
2970 struct in6_addr rtmsg_dst;
2971 struct in6_addr rtmsg_src;
2972 struct in6_addr rtmsg_gateway;
2973 u32 rtmsg_type;
2974 u16 rtmsg_dst_len;
2975 u16 rtmsg_src_len;
2976 u32 rtmsg_metric;
2977 u32 rtmsg_info;
2978 u32 rtmsg_flags;
2979 s32 rtmsg_ifindex;
2980 };
2981
2982 static int routing_ioctl(struct net *net, struct socket *sock,
2983 unsigned int cmd, void __user *argp)
2984 {
2985 int ret;
2986 void *r = NULL;
2987 struct in6_rtmsg r6;
2988 struct rtentry r4;
2989 char devname[16];
2990 u32 rtdev;
2991 mm_segment_t old_fs = get_fs();
2992
2993 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2994 struct in6_rtmsg32 __user *ur6 = argp;
2995 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2996 3 * sizeof(struct in6_addr));
2997 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2998 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2999 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3000 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3001 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3002 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3003 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3004
3005 r = (void *) &r6;
3006 } else { /* ipv4 */
3007 struct rtentry32 __user *ur4 = argp;
3008 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3009 3 * sizeof(struct sockaddr));
3010 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3011 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3012 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3013 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3014 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3015 ret |= get_user(rtdev, &(ur4->rt_dev));
3016 if (rtdev) {
3017 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3018 r4.rt_dev = (char __user __force *)devname;
3019 devname[15] = 0;
3020 } else
3021 r4.rt_dev = NULL;
3022
3023 r = (void *) &r4;
3024 }
3025
3026 if (ret) {
3027 ret = -EFAULT;
3028 goto out;
3029 }
3030
3031 set_fs(KERNEL_DS);
3032 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3033 set_fs(old_fs);
3034
3035 out:
3036 return ret;
3037 }
3038
3039 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3040 * for some operations; this forces use of the newer bridge-utils that
3041 * use compatible ioctls
3042 */
3043 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3044 {
3045 compat_ulong_t tmp;
3046
3047 if (get_user(tmp, argp))
3048 return -EFAULT;
3049 if (tmp == BRCTL_GET_VERSION)
3050 return BRCTL_VERSION + 1;
3051 return -EINVAL;
3052 }
3053
3054 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3055 unsigned int cmd, unsigned long arg)
3056 {
3057 void __user *argp = compat_ptr(arg);
3058 struct sock *sk = sock->sk;
3059 struct net *net = sock_net(sk);
3060
3061 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3062 return compat_ifr_data_ioctl(net, cmd, argp);
3063
3064 switch (cmd) {
3065 case SIOCSIFBR:
3066 case SIOCGIFBR:
3067 return old_bridge_ioctl(argp);
3068 case SIOCGIFNAME:
3069 return dev_ifname32(net, argp);
3070 case SIOCGIFCONF:
3071 return dev_ifconf(net, argp);
3072 case SIOCETHTOOL:
3073 return ethtool_ioctl(net, argp);
3074 case SIOCWANDEV:
3075 return compat_siocwandev(net, argp);
3076 case SIOCGIFMAP:
3077 case SIOCSIFMAP:
3078 return compat_sioc_ifmap(net, cmd, argp);
3079 case SIOCBONDENSLAVE:
3080 case SIOCBONDRELEASE:
3081 case SIOCBONDSETHWADDR:
3082 case SIOCBONDCHANGEACTIVE:
3083 return bond_ioctl(net, cmd, argp);
3084 case SIOCADDRT:
3085 case SIOCDELRT:
3086 return routing_ioctl(net, sock, cmd, argp);
3087 case SIOCGSTAMP:
3088 return do_siocgstamp(net, sock, cmd, argp);
3089 case SIOCGSTAMPNS:
3090 return do_siocgstampns(net, sock, cmd, argp);
3091 case SIOCBONDSLAVEINFOQUERY:
3092 case SIOCBONDINFOQUERY:
3093 case SIOCSHWTSTAMP:
3094 case SIOCGHWTSTAMP:
3095 return compat_ifr_data_ioctl(net, cmd, argp);
3096
3097 case FIOSETOWN:
3098 case SIOCSPGRP:
3099 case FIOGETOWN:
3100 case SIOCGPGRP:
3101 case SIOCBRADDBR:
3102 case SIOCBRDELBR:
3103 case SIOCGIFVLAN:
3104 case SIOCSIFVLAN:
3105 case SIOCADDDLCI:
3106 case SIOCDELDLCI:
3107 return sock_ioctl(file, cmd, arg);
3108
3109 case SIOCGIFFLAGS:
3110 case SIOCSIFFLAGS:
3111 case SIOCGIFMETRIC:
3112 case SIOCSIFMETRIC:
3113 case SIOCGIFMTU:
3114 case SIOCSIFMTU:
3115 case SIOCGIFMEM:
3116 case SIOCSIFMEM:
3117 case SIOCGIFHWADDR:
3118 case SIOCSIFHWADDR:
3119 case SIOCADDMULTI:
3120 case SIOCDELMULTI:
3121 case SIOCGIFINDEX:
3122 case SIOCGIFADDR:
3123 case SIOCSIFADDR:
3124 case SIOCSIFHWBROADCAST:
3125 case SIOCDIFADDR:
3126 case SIOCGIFBRDADDR:
3127 case SIOCSIFBRDADDR:
3128 case SIOCGIFDSTADDR:
3129 case SIOCSIFDSTADDR:
3130 case SIOCGIFNETMASK:
3131 case SIOCSIFNETMASK:
3132 case SIOCSIFPFLAGS:
3133 case SIOCGIFPFLAGS:
3134 case SIOCGIFTXQLEN:
3135 case SIOCSIFTXQLEN:
3136 case SIOCBRADDIF:
3137 case SIOCBRDELIF:
3138 case SIOCSIFNAME:
3139 case SIOCGMIIPHY:
3140 case SIOCGMIIREG:
3141 case SIOCSMIIREG:
3142 return dev_ifsioc(net, sock, cmd, argp);
3143
3144 case SIOCSARP:
3145 case SIOCGARP:
3146 case SIOCDARP:
3147 case SIOCATMARK:
3148 return sock_do_ioctl(net, sock, cmd, arg);
3149 }
3150
3151 return -ENOIOCTLCMD;
3152 }
3153
3154 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3155 unsigned long arg)
3156 {
3157 struct socket *sock = file->private_data;
3158 int ret = -ENOIOCTLCMD;
3159 struct sock *sk;
3160 struct net *net;
3161
3162 sk = sock->sk;
3163 net = sock_net(sk);
3164
3165 if (sock->ops->compat_ioctl)
3166 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3167
3168 if (ret == -ENOIOCTLCMD &&
3169 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3170 ret = compat_wext_handle_ioctl(net, cmd, arg);
3171
3172 if (ret == -ENOIOCTLCMD)
3173 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3174
3175 return ret;
3176 }
3177 #endif
3178
3179 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3180 {
3181 return sock->ops->bind(sock, addr, addrlen);
3182 }
3183 EXPORT_SYMBOL(kernel_bind);
3184
3185 int kernel_listen(struct socket *sock, int backlog)
3186 {
3187 return sock->ops->listen(sock, backlog);
3188 }
3189 EXPORT_SYMBOL(kernel_listen);
3190
3191 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3192 {
3193 struct sock *sk = sock->sk;
3194 int err;
3195
3196 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3197 newsock);
3198 if (err < 0)
3199 goto done;
3200
3201 err = sock->ops->accept(sock, *newsock, flags);
3202 if (err < 0) {
3203 sock_release(*newsock);
3204 *newsock = NULL;
3205 goto done;
3206 }
3207
3208 (*newsock)->ops = sock->ops;
3209 __module_get((*newsock)->ops->owner);
3210
3211 done:
3212 return err;
3213 }
3214 EXPORT_SYMBOL(kernel_accept);
3215
3216 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3217 int flags)
3218 {
3219 return sock->ops->connect(sock, addr, addrlen, flags);
3220 }
3221 EXPORT_SYMBOL(kernel_connect);
3222
3223 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3224 int *addrlen)
3225 {
3226 return sock->ops->getname(sock, addr, addrlen, 0);
3227 }
3228 EXPORT_SYMBOL(kernel_getsockname);
3229
3230 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3231 int *addrlen)
3232 {
3233 return sock->ops->getname(sock, addr, addrlen, 1);
3234 }
3235 EXPORT_SYMBOL(kernel_getpeername);
3236
3237 int kernel_getsockopt(struct socket *sock, int level, int optname,
3238 char *optval, int *optlen)
3239 {
3240 mm_segment_t oldfs = get_fs();
3241 char __user *uoptval;
3242 int __user *uoptlen;
3243 int err;
3244
3245 uoptval = (char __user __force *) optval;
3246 uoptlen = (int __user __force *) optlen;
3247
3248 set_fs(KERNEL_DS);
3249 if (level == SOL_SOCKET)
3250 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3251 else
3252 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3253 uoptlen);
3254 set_fs(oldfs);
3255 return err;
3256 }
3257 EXPORT_SYMBOL(kernel_getsockopt);
3258
3259 int kernel_setsockopt(struct socket *sock, int level, int optname,
3260 char *optval, unsigned int optlen)
3261 {
3262 mm_segment_t oldfs = get_fs();
3263 char __user *uoptval;
3264 int err;
3265
3266 uoptval = (char __user __force *) optval;
3267
3268 set_fs(KERNEL_DS);
3269 if (level == SOL_SOCKET)
3270 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3271 else
3272 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3273 optlen);
3274 set_fs(oldfs);
3275 return err;
3276 }
3277 EXPORT_SYMBOL(kernel_setsockopt);
3278
3279 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3280 size_t size, int flags)
3281 {
3282 if (sock->ops->sendpage)
3283 return sock->ops->sendpage(sock, page, offset, size, flags);
3284
3285 return sock_no_sendpage(sock, page, offset, size, flags);
3286 }
3287 EXPORT_SYMBOL(kernel_sendpage);
3288
3289 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3290 {
3291 mm_segment_t oldfs = get_fs();
3292 int err;
3293
3294 set_fs(KERNEL_DS);
3295 err = sock->ops->ioctl(sock, cmd, arg);
3296 set_fs(oldfs);
3297
3298 return err;
3299 }
3300 EXPORT_SYMBOL(kernel_sock_ioctl);
3301
3302 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3303 {
3304 return sock->ops->shutdown(sock, how);
3305 }
3306 EXPORT_SYMBOL(kernel_sock_shutdown);
3307
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.