]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - net/socket.c
Merge tag 'for-linus-4.9-rc5-ofs-1' of git://git.kernel.org/pub/scm/linux/kernel...
[mirror_ubuntu-bionic-kernel.git] / net / socket.c
1 /*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61 #include <linux/mm.h>
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/ptp_classify.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
89 #include <linux/magic.h>
90 #include <linux/slab.h>
91 #include <linux/xattr.h>
92
93 #include <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 int sockfs_xattr_get(const struct xattr_handler *handler,
324 struct dentry *dentry, struct inode *inode,
325 const char *suffix, void *value, size_t size)
326 {
327 if (value) {
328 if (dentry->d_name.len + 1 > size)
329 return -ERANGE;
330 memcpy(value, dentry->d_name.name, dentry->d_name.len + 1);
331 }
332 return dentry->d_name.len + 1;
333 }
334
335 #define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
336 #define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
337 #define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
338
339 static const struct xattr_handler sockfs_xattr_handler = {
340 .name = XATTR_NAME_SOCKPROTONAME,
341 .get = sockfs_xattr_get,
342 };
343
344 static const struct xattr_handler *sockfs_xattr_handlers[] = {
345 &sockfs_xattr_handler,
346 NULL
347 };
348
349 static struct dentry *sockfs_mount(struct file_system_type *fs_type,
350 int flags, const char *dev_name, void *data)
351 {
352 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
353 sockfs_xattr_handlers,
354 &sockfs_dentry_operations, SOCKFS_MAGIC);
355 }
356
357 static struct vfsmount *sock_mnt __read_mostly;
358
359 static struct file_system_type sock_fs_type = {
360 .name = "sockfs",
361 .mount = sockfs_mount,
362 .kill_sb = kill_anon_super,
363 };
364
365 /*
366 * Obtains the first available file descriptor and sets it up for use.
367 *
368 * These functions create file structures and maps them to fd space
369 * of the current process. On success it returns file descriptor
370 * and file struct implicitly stored in sock->file.
371 * Note that another thread may close file descriptor before we return
372 * from this function. We use the fact that now we do not refer
373 * to socket after mapping. If one day we will need it, this
374 * function will increment ref. count on file by 1.
375 *
376 * In any case returned fd MAY BE not valid!
377 * This race condition is unavoidable
378 * with shared fd spaces, we cannot solve it inside kernel,
379 * but we take care of internal coherence yet.
380 */
381
382 struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
383 {
384 struct qstr name = { .name = "" };
385 struct path path;
386 struct file *file;
387
388 if (dname) {
389 name.name = dname;
390 name.len = strlen(name.name);
391 } else if (sock->sk) {
392 name.name = sock->sk->sk_prot_creator->name;
393 name.len = strlen(name.name);
394 }
395 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
396 if (unlikely(!path.dentry))
397 return ERR_PTR(-ENOMEM);
398 path.mnt = mntget(sock_mnt);
399
400 d_instantiate(path.dentry, SOCK_INODE(sock));
401
402 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
403 &socket_file_ops);
404 if (IS_ERR(file)) {
405 /* drop dentry, keep inode */
406 ihold(d_inode(path.dentry));
407 path_put(&path);
408 return file;
409 }
410
411 sock->file = file;
412 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
413 file->private_data = sock;
414 return file;
415 }
416 EXPORT_SYMBOL(sock_alloc_file);
417
418 static int sock_map_fd(struct socket *sock, int flags)
419 {
420 struct file *newfile;
421 int fd = get_unused_fd_flags(flags);
422 if (unlikely(fd < 0))
423 return fd;
424
425 newfile = sock_alloc_file(sock, flags, NULL);
426 if (likely(!IS_ERR(newfile))) {
427 fd_install(fd, newfile);
428 return fd;
429 }
430
431 put_unused_fd(fd);
432 return PTR_ERR(newfile);
433 }
434
435 struct socket *sock_from_file(struct file *file, int *err)
436 {
437 if (file->f_op == &socket_file_ops)
438 return file->private_data; /* set in sock_map_fd */
439
440 *err = -ENOTSOCK;
441 return NULL;
442 }
443 EXPORT_SYMBOL(sock_from_file);
444
445 /**
446 * sockfd_lookup - Go from a file number to its socket slot
447 * @fd: file handle
448 * @err: pointer to an error code return
449 *
450 * The file handle passed in is locked and the socket it is bound
451 * too is returned. If an error occurs the err pointer is overwritten
452 * with a negative errno code and NULL is returned. The function checks
453 * for both invalid handles and passing a handle which is not a socket.
454 *
455 * On a success the socket object pointer is returned.
456 */
457
458 struct socket *sockfd_lookup(int fd, int *err)
459 {
460 struct file *file;
461 struct socket *sock;
462
463 file = fget(fd);
464 if (!file) {
465 *err = -EBADF;
466 return NULL;
467 }
468
469 sock = sock_from_file(file, err);
470 if (!sock)
471 fput(file);
472 return sock;
473 }
474 EXPORT_SYMBOL(sockfd_lookup);
475
476 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
477 {
478 struct fd f = fdget(fd);
479 struct socket *sock;
480
481 *err = -EBADF;
482 if (f.file) {
483 sock = sock_from_file(f.file, err);
484 if (likely(sock)) {
485 *fput_needed = f.flags;
486 return sock;
487 }
488 fdput(f);
489 }
490 return NULL;
491 }
492
493 static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
494 size_t size)
495 {
496 ssize_t len;
497 ssize_t used = 0;
498
499 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
500 if (len < 0)
501 return len;
502 used += len;
503 if (buffer) {
504 if (size < used)
505 return -ERANGE;
506 buffer += len;
507 }
508
509 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
510 used += len;
511 if (buffer) {
512 if (size < used)
513 return -ERANGE;
514 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
515 buffer += len;
516 }
517
518 return used;
519 }
520
521 static const struct inode_operations sockfs_inode_ops = {
522 .listxattr = sockfs_listxattr,
523 };
524
525 /**
526 * sock_alloc - allocate a socket
527 *
528 * Allocate a new inode and socket object. The two are bound together
529 * and initialised. The socket is then returned. If we are out of inodes
530 * NULL is returned.
531 */
532
533 struct socket *sock_alloc(void)
534 {
535 struct inode *inode;
536 struct socket *sock;
537
538 inode = new_inode_pseudo(sock_mnt->mnt_sb);
539 if (!inode)
540 return NULL;
541
542 sock = SOCKET_I(inode);
543
544 kmemcheck_annotate_bitfield(sock, type);
545 inode->i_ino = get_next_ino();
546 inode->i_mode = S_IFSOCK | S_IRWXUGO;
547 inode->i_uid = current_fsuid();
548 inode->i_gid = current_fsgid();
549 inode->i_op = &sockfs_inode_ops;
550
551 this_cpu_add(sockets_in_use, 1);
552 return sock;
553 }
554 EXPORT_SYMBOL(sock_alloc);
555
556 /**
557 * sock_release - close a socket
558 * @sock: socket to close
559 *
560 * The socket is released from the protocol stack if it has a release
561 * callback, and the inode is then released if the socket is bound to
562 * an inode not a file.
563 */
564
565 void sock_release(struct socket *sock)
566 {
567 if (sock->ops) {
568 struct module *owner = sock->ops->owner;
569
570 sock->ops->release(sock);
571 sock->ops = NULL;
572 module_put(owner);
573 }
574
575 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
576 pr_err("%s: fasync list not empty!\n", __func__);
577
578 this_cpu_sub(sockets_in_use, 1);
579 if (!sock->file) {
580 iput(SOCK_INODE(sock));
581 return;
582 }
583 sock->file = NULL;
584 }
585 EXPORT_SYMBOL(sock_release);
586
587 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
588 {
589 u8 flags = *tx_flags;
590
591 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
592 flags |= SKBTX_HW_TSTAMP;
593
594 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
595 flags |= SKBTX_SW_TSTAMP;
596
597 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
598 flags |= SKBTX_SCHED_TSTAMP;
599
600 *tx_flags = flags;
601 }
602 EXPORT_SYMBOL(__sock_tx_timestamp);
603
604 static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
605 {
606 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
607 BUG_ON(ret == -EIOCBQUEUED);
608 return ret;
609 }
610
611 int sock_sendmsg(struct socket *sock, struct msghdr *msg)
612 {
613 int err = security_socket_sendmsg(sock, msg,
614 msg_data_left(msg));
615
616 return err ?: sock_sendmsg_nosec(sock, msg);
617 }
618 EXPORT_SYMBOL(sock_sendmsg);
619
620 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
621 struct kvec *vec, size_t num, size_t size)
622 {
623 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
624 return sock_sendmsg(sock, msg);
625 }
626 EXPORT_SYMBOL(kernel_sendmsg);
627
628 /*
629 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
630 */
631 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
632 struct sk_buff *skb)
633 {
634 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
635 struct scm_timestamping tss;
636 int empty = 1;
637 struct skb_shared_hwtstamps *shhwtstamps =
638 skb_hwtstamps(skb);
639
640 /* Race occurred between timestamp enabling and packet
641 receiving. Fill in the current time for now. */
642 if (need_software_tstamp && skb->tstamp.tv64 == 0)
643 __net_timestamp(skb);
644
645 if (need_software_tstamp) {
646 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
647 struct timeval tv;
648 skb_get_timestamp(skb, &tv);
649 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
650 sizeof(tv), &tv);
651 } else {
652 struct timespec ts;
653 skb_get_timestampns(skb, &ts);
654 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
655 sizeof(ts), &ts);
656 }
657 }
658
659 memset(&tss, 0, sizeof(tss));
660 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
661 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
662 empty = 0;
663 if (shhwtstamps &&
664 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
665 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
666 empty = 0;
667 if (!empty)
668 put_cmsg(msg, SOL_SOCKET,
669 SCM_TIMESTAMPING, sizeof(tss), &tss);
670 }
671 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
672
673 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
674 struct sk_buff *skb)
675 {
676 int ack;
677
678 if (!sock_flag(sk, SOCK_WIFI_STATUS))
679 return;
680 if (!skb->wifi_acked_valid)
681 return;
682
683 ack = skb->wifi_acked;
684
685 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
686 }
687 EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
688
689 static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
690 struct sk_buff *skb)
691 {
692 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
693 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
694 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
695 }
696
697 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
698 struct sk_buff *skb)
699 {
700 sock_recv_timestamp(msg, sk, skb);
701 sock_recv_drops(msg, sk, skb);
702 }
703 EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
704
705 static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
706 int flags)
707 {
708 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
709 }
710
711 int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
712 {
713 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
714
715 return err ?: sock_recvmsg_nosec(sock, msg, flags);
716 }
717 EXPORT_SYMBOL(sock_recvmsg);
718
719 /**
720 * kernel_recvmsg - Receive a message from a socket (kernel space)
721 * @sock: The socket to receive the message from
722 * @msg: Received message
723 * @vec: Input s/g array for message data
724 * @num: Size of input s/g array
725 * @size: Number of bytes to read
726 * @flags: Message flags (MSG_DONTWAIT, etc...)
727 *
728 * On return the msg structure contains the scatter/gather array passed in the
729 * vec argument. The array is modified so that it consists of the unfilled
730 * portion of the original array.
731 *
732 * The returned value is the total number of bytes received, or an error.
733 */
734 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
735 struct kvec *vec, size_t num, size_t size, int flags)
736 {
737 mm_segment_t oldfs = get_fs();
738 int result;
739
740 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
741 set_fs(KERNEL_DS);
742 result = sock_recvmsg(sock, msg, flags);
743 set_fs(oldfs);
744 return result;
745 }
746 EXPORT_SYMBOL(kernel_recvmsg);
747
748 static ssize_t sock_sendpage(struct file *file, struct page *page,
749 int offset, size_t size, loff_t *ppos, int more)
750 {
751 struct socket *sock;
752 int flags;
753
754 sock = file->private_data;
755
756 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
757 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
758 flags |= more;
759
760 return kernel_sendpage(sock, page, offset, size, flags);
761 }
762
763 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
764 struct pipe_inode_info *pipe, size_t len,
765 unsigned int flags)
766 {
767 struct socket *sock = file->private_data;
768
769 if (unlikely(!sock->ops->splice_read))
770 return -EINVAL;
771
772 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
773 }
774
775 static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
776 {
777 struct file *file = iocb->ki_filp;
778 struct socket *sock = file->private_data;
779 struct msghdr msg = {.msg_iter = *to,
780 .msg_iocb = iocb};
781 ssize_t res;
782
783 if (file->f_flags & O_NONBLOCK)
784 msg.msg_flags = MSG_DONTWAIT;
785
786 if (iocb->ki_pos != 0)
787 return -ESPIPE;
788
789 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
790 return 0;
791
792 res = sock_recvmsg(sock, &msg, msg.msg_flags);
793 *to = msg.msg_iter;
794 return res;
795 }
796
797 static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
798 {
799 struct file *file = iocb->ki_filp;
800 struct socket *sock = file->private_data;
801 struct msghdr msg = {.msg_iter = *from,
802 .msg_iocb = iocb};
803 ssize_t res;
804
805 if (iocb->ki_pos != 0)
806 return -ESPIPE;
807
808 if (file->f_flags & O_NONBLOCK)
809 msg.msg_flags = MSG_DONTWAIT;
810
811 if (sock->type == SOCK_SEQPACKET)
812 msg.msg_flags |= MSG_EOR;
813
814 res = sock_sendmsg(sock, &msg);
815 *from = msg.msg_iter;
816 return res;
817 }
818
819 /*
820 * Atomic setting of ioctl hooks to avoid race
821 * with module unload.
822 */
823
824 static DEFINE_MUTEX(br_ioctl_mutex);
825 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
826
827 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
828 {
829 mutex_lock(&br_ioctl_mutex);
830 br_ioctl_hook = hook;
831 mutex_unlock(&br_ioctl_mutex);
832 }
833 EXPORT_SYMBOL(brioctl_set);
834
835 static DEFINE_MUTEX(vlan_ioctl_mutex);
836 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
837
838 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
839 {
840 mutex_lock(&vlan_ioctl_mutex);
841 vlan_ioctl_hook = hook;
842 mutex_unlock(&vlan_ioctl_mutex);
843 }
844 EXPORT_SYMBOL(vlan_ioctl_set);
845
846 static DEFINE_MUTEX(dlci_ioctl_mutex);
847 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
848
849 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
850 {
851 mutex_lock(&dlci_ioctl_mutex);
852 dlci_ioctl_hook = hook;
853 mutex_unlock(&dlci_ioctl_mutex);
854 }
855 EXPORT_SYMBOL(dlci_ioctl_set);
856
857 static long sock_do_ioctl(struct net *net, struct socket *sock,
858 unsigned int cmd, unsigned long arg)
859 {
860 int err;
861 void __user *argp = (void __user *)arg;
862
863 err = sock->ops->ioctl(sock, cmd, arg);
864
865 /*
866 * If this ioctl is unknown try to hand it down
867 * to the NIC driver.
868 */
869 if (err == -ENOIOCTLCMD)
870 err = dev_ioctl(net, cmd, argp);
871
872 return err;
873 }
874
875 /*
876 * With an ioctl, arg may well be a user mode pointer, but we don't know
877 * what to do with it - that's up to the protocol still.
878 */
879
880 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
881 {
882 struct socket *sock;
883 struct sock *sk;
884 void __user *argp = (void __user *)arg;
885 int pid, err;
886 struct net *net;
887
888 sock = file->private_data;
889 sk = sock->sk;
890 net = sock_net(sk);
891 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
892 err = dev_ioctl(net, cmd, argp);
893 } else
894 #ifdef CONFIG_WEXT_CORE
895 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
896 err = dev_ioctl(net, cmd, argp);
897 } else
898 #endif
899 switch (cmd) {
900 case FIOSETOWN:
901 case SIOCSPGRP:
902 err = -EFAULT;
903 if (get_user(pid, (int __user *)argp))
904 break;
905 f_setown(sock->file, pid, 1);
906 err = 0;
907 break;
908 case FIOGETOWN:
909 case SIOCGPGRP:
910 err = put_user(f_getown(sock->file),
911 (int __user *)argp);
912 break;
913 case SIOCGIFBR:
914 case SIOCSIFBR:
915 case SIOCBRADDBR:
916 case SIOCBRDELBR:
917 err = -ENOPKG;
918 if (!br_ioctl_hook)
919 request_module("bridge");
920
921 mutex_lock(&br_ioctl_mutex);
922 if (br_ioctl_hook)
923 err = br_ioctl_hook(net, cmd, argp);
924 mutex_unlock(&br_ioctl_mutex);
925 break;
926 case SIOCGIFVLAN:
927 case SIOCSIFVLAN:
928 err = -ENOPKG;
929 if (!vlan_ioctl_hook)
930 request_module("8021q");
931
932 mutex_lock(&vlan_ioctl_mutex);
933 if (vlan_ioctl_hook)
934 err = vlan_ioctl_hook(net, argp);
935 mutex_unlock(&vlan_ioctl_mutex);
936 break;
937 case SIOCADDDLCI:
938 case SIOCDELDLCI:
939 err = -ENOPKG;
940 if (!dlci_ioctl_hook)
941 request_module("dlci");
942
943 mutex_lock(&dlci_ioctl_mutex);
944 if (dlci_ioctl_hook)
945 err = dlci_ioctl_hook(cmd, argp);
946 mutex_unlock(&dlci_ioctl_mutex);
947 break;
948 default:
949 err = sock_do_ioctl(net, sock, cmd, arg);
950 break;
951 }
952 return err;
953 }
954
955 int sock_create_lite(int family, int type, int protocol, struct socket **res)
956 {
957 int err;
958 struct socket *sock = NULL;
959
960 err = security_socket_create(family, type, protocol, 1);
961 if (err)
962 goto out;
963
964 sock = sock_alloc();
965 if (!sock) {
966 err = -ENOMEM;
967 goto out;
968 }
969
970 sock->type = type;
971 err = security_socket_post_create(sock, family, type, protocol, 1);
972 if (err)
973 goto out_release;
974
975 out:
976 *res = sock;
977 return err;
978 out_release:
979 sock_release(sock);
980 sock = NULL;
981 goto out;
982 }
983 EXPORT_SYMBOL(sock_create_lite);
984
985 /* No kernel lock held - perfect */
986 static unsigned int sock_poll(struct file *file, poll_table *wait)
987 {
988 unsigned int busy_flag = 0;
989 struct socket *sock;
990
991 /*
992 * We can't return errors to poll, so it's either yes or no.
993 */
994 sock = file->private_data;
995
996 if (sk_can_busy_loop(sock->sk)) {
997 /* this socket can poll_ll so tell the system call */
998 busy_flag = POLL_BUSY_LOOP;
999
1000 /* once, only if requested by syscall */
1001 if (wait && (wait->_key & POLL_BUSY_LOOP))
1002 sk_busy_loop(sock->sk, 1);
1003 }
1004
1005 return busy_flag | sock->ops->poll(file, sock, wait);
1006 }
1007
1008 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1009 {
1010 struct socket *sock = file->private_data;
1011
1012 return sock->ops->mmap(file, sock, vma);
1013 }
1014
1015 static int sock_close(struct inode *inode, struct file *filp)
1016 {
1017 sock_release(SOCKET_I(inode));
1018 return 0;
1019 }
1020
1021 /*
1022 * Update the socket async list
1023 *
1024 * Fasync_list locking strategy.
1025 *
1026 * 1. fasync_list is modified only under process context socket lock
1027 * i.e. under semaphore.
1028 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1029 * or under socket lock
1030 */
1031
1032 static int sock_fasync(int fd, struct file *filp, int on)
1033 {
1034 struct socket *sock = filp->private_data;
1035 struct sock *sk = sock->sk;
1036 struct socket_wq *wq;
1037
1038 if (sk == NULL)
1039 return -EINVAL;
1040
1041 lock_sock(sk);
1042 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1043 fasync_helper(fd, filp, on, &wq->fasync_list);
1044
1045 if (!wq->fasync_list)
1046 sock_reset_flag(sk, SOCK_FASYNC);
1047 else
1048 sock_set_flag(sk, SOCK_FASYNC);
1049
1050 release_sock(sk);
1051 return 0;
1052 }
1053
1054 /* This function may be called only under rcu_lock */
1055
1056 int sock_wake_async(struct socket_wq *wq, int how, int band)
1057 {
1058 if (!wq || !wq->fasync_list)
1059 return -1;
1060
1061 switch (how) {
1062 case SOCK_WAKE_WAITD:
1063 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1064 break;
1065 goto call_kill;
1066 case SOCK_WAKE_SPACE:
1067 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1068 break;
1069 /* fall through */
1070 case SOCK_WAKE_IO:
1071 call_kill:
1072 kill_fasync(&wq->fasync_list, SIGIO, band);
1073 break;
1074 case SOCK_WAKE_URG:
1075 kill_fasync(&wq->fasync_list, SIGURG, band);
1076 }
1077
1078 return 0;
1079 }
1080 EXPORT_SYMBOL(sock_wake_async);
1081
1082 int __sock_create(struct net *net, int family, int type, int protocol,
1083 struct socket **res, int kern)
1084 {
1085 int err;
1086 struct socket *sock;
1087 const struct net_proto_family *pf;
1088
1089 /*
1090 * Check protocol is in range
1091 */
1092 if (family < 0 || family >= NPROTO)
1093 return -EAFNOSUPPORT;
1094 if (type < 0 || type >= SOCK_MAX)
1095 return -EINVAL;
1096
1097 /* Compatibility.
1098
1099 This uglymoron is moved from INET layer to here to avoid
1100 deadlock in module load.
1101 */
1102 if (family == PF_INET && type == SOCK_PACKET) {
1103 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1104 current->comm);
1105 family = PF_PACKET;
1106 }
1107
1108 err = security_socket_create(family, type, protocol, kern);
1109 if (err)
1110 return err;
1111
1112 /*
1113 * Allocate the socket and allow the family to set things up. if
1114 * the protocol is 0, the family is instructed to select an appropriate
1115 * default.
1116 */
1117 sock = sock_alloc();
1118 if (!sock) {
1119 net_warn_ratelimited("socket: no more sockets\n");
1120 return -ENFILE; /* Not exactly a match, but its the
1121 closest posix thing */
1122 }
1123
1124 sock->type = type;
1125
1126 #ifdef CONFIG_MODULES
1127 /* Attempt to load a protocol module if the find failed.
1128 *
1129 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1130 * requested real, full-featured networking support upon configuration.
1131 * Otherwise module support will break!
1132 */
1133 if (rcu_access_pointer(net_families[family]) == NULL)
1134 request_module("net-pf-%d", family);
1135 #endif
1136
1137 rcu_read_lock();
1138 pf = rcu_dereference(net_families[family]);
1139 err = -EAFNOSUPPORT;
1140 if (!pf)
1141 goto out_release;
1142
1143 /*
1144 * We will call the ->create function, that possibly is in a loadable
1145 * module, so we have to bump that loadable module refcnt first.
1146 */
1147 if (!try_module_get(pf->owner))
1148 goto out_release;
1149
1150 /* Now protected by module ref count */
1151 rcu_read_unlock();
1152
1153 err = pf->create(net, sock, protocol, kern);
1154 if (err < 0)
1155 goto out_module_put;
1156
1157 /*
1158 * Now to bump the refcnt of the [loadable] module that owns this
1159 * socket at sock_release time we decrement its refcnt.
1160 */
1161 if (!try_module_get(sock->ops->owner))
1162 goto out_module_busy;
1163
1164 /*
1165 * Now that we're done with the ->create function, the [loadable]
1166 * module can have its refcnt decremented
1167 */
1168 module_put(pf->owner);
1169 err = security_socket_post_create(sock, family, type, protocol, kern);
1170 if (err)
1171 goto out_sock_release;
1172 *res = sock;
1173
1174 return 0;
1175
1176 out_module_busy:
1177 err = -EAFNOSUPPORT;
1178 out_module_put:
1179 sock->ops = NULL;
1180 module_put(pf->owner);
1181 out_sock_release:
1182 sock_release(sock);
1183 return err;
1184
1185 out_release:
1186 rcu_read_unlock();
1187 goto out_sock_release;
1188 }
1189 EXPORT_SYMBOL(__sock_create);
1190
1191 int sock_create(int family, int type, int protocol, struct socket **res)
1192 {
1193 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1194 }
1195 EXPORT_SYMBOL(sock_create);
1196
1197 int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1198 {
1199 return __sock_create(net, family, type, protocol, res, 1);
1200 }
1201 EXPORT_SYMBOL(sock_create_kern);
1202
1203 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1204 {
1205 int retval;
1206 struct socket *sock;
1207 int flags;
1208
1209 /* Check the SOCK_* constants for consistency. */
1210 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1211 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1212 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1213 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1214
1215 flags = type & ~SOCK_TYPE_MASK;
1216 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1217 return -EINVAL;
1218 type &= SOCK_TYPE_MASK;
1219
1220 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1221 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1222
1223 retval = sock_create(family, type, protocol, &sock);
1224 if (retval < 0)
1225 goto out;
1226
1227 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1228 if (retval < 0)
1229 goto out_release;
1230
1231 out:
1232 /* It may be already another descriptor 8) Not kernel problem. */
1233 return retval;
1234
1235 out_release:
1236 sock_release(sock);
1237 return retval;
1238 }
1239
1240 /*
1241 * Create a pair of connected sockets.
1242 */
1243
1244 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1245 int __user *, usockvec)
1246 {
1247 struct socket *sock1, *sock2;
1248 int fd1, fd2, err;
1249 struct file *newfile1, *newfile2;
1250 int flags;
1251
1252 flags = type & ~SOCK_TYPE_MASK;
1253 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1254 return -EINVAL;
1255 type &= SOCK_TYPE_MASK;
1256
1257 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1258 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1259
1260 /*
1261 * Obtain the first socket and check if the underlying protocol
1262 * supports the socketpair call.
1263 */
1264
1265 err = sock_create(family, type, protocol, &sock1);
1266 if (err < 0)
1267 goto out;
1268
1269 err = sock_create(family, type, protocol, &sock2);
1270 if (err < 0)
1271 goto out_release_1;
1272
1273 err = sock1->ops->socketpair(sock1, sock2);
1274 if (err < 0)
1275 goto out_release_both;
1276
1277 fd1 = get_unused_fd_flags(flags);
1278 if (unlikely(fd1 < 0)) {
1279 err = fd1;
1280 goto out_release_both;
1281 }
1282
1283 fd2 = get_unused_fd_flags(flags);
1284 if (unlikely(fd2 < 0)) {
1285 err = fd2;
1286 goto out_put_unused_1;
1287 }
1288
1289 newfile1 = sock_alloc_file(sock1, flags, NULL);
1290 if (IS_ERR(newfile1)) {
1291 err = PTR_ERR(newfile1);
1292 goto out_put_unused_both;
1293 }
1294
1295 newfile2 = sock_alloc_file(sock2, flags, NULL);
1296 if (IS_ERR(newfile2)) {
1297 err = PTR_ERR(newfile2);
1298 goto out_fput_1;
1299 }
1300
1301 err = put_user(fd1, &usockvec[0]);
1302 if (err)
1303 goto out_fput_both;
1304
1305 err = put_user(fd2, &usockvec[1]);
1306 if (err)
1307 goto out_fput_both;
1308
1309 audit_fd_pair(fd1, fd2);
1310
1311 fd_install(fd1, newfile1);
1312 fd_install(fd2, newfile2);
1313 /* fd1 and fd2 may be already another descriptors.
1314 * Not kernel problem.
1315 */
1316
1317 return 0;
1318
1319 out_fput_both:
1320 fput(newfile2);
1321 fput(newfile1);
1322 put_unused_fd(fd2);
1323 put_unused_fd(fd1);
1324 goto out;
1325
1326 out_fput_1:
1327 fput(newfile1);
1328 put_unused_fd(fd2);
1329 put_unused_fd(fd1);
1330 sock_release(sock2);
1331 goto out;
1332
1333 out_put_unused_both:
1334 put_unused_fd(fd2);
1335 out_put_unused_1:
1336 put_unused_fd(fd1);
1337 out_release_both:
1338 sock_release(sock2);
1339 out_release_1:
1340 sock_release(sock1);
1341 out:
1342 return err;
1343 }
1344
1345 /*
1346 * Bind a name to a socket. Nothing much to do here since it's
1347 * the protocol's responsibility to handle the local address.
1348 *
1349 * We move the socket address to kernel space before we call
1350 * the protocol layer (having also checked the address is ok).
1351 */
1352
1353 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1354 {
1355 struct socket *sock;
1356 struct sockaddr_storage address;
1357 int err, fput_needed;
1358
1359 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1360 if (sock) {
1361 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1362 if (err >= 0) {
1363 err = security_socket_bind(sock,
1364 (struct sockaddr *)&address,
1365 addrlen);
1366 if (!err)
1367 err = sock->ops->bind(sock,
1368 (struct sockaddr *)
1369 &address, addrlen);
1370 }
1371 fput_light(sock->file, fput_needed);
1372 }
1373 return err;
1374 }
1375
1376 /*
1377 * Perform a listen. Basically, we allow the protocol to do anything
1378 * necessary for a listen, and if that works, we mark the socket as
1379 * ready for listening.
1380 */
1381
1382 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1383 {
1384 struct socket *sock;
1385 int err, fput_needed;
1386 int somaxconn;
1387
1388 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1389 if (sock) {
1390 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1391 if ((unsigned int)backlog > somaxconn)
1392 backlog = somaxconn;
1393
1394 err = security_socket_listen(sock, backlog);
1395 if (!err)
1396 err = sock->ops->listen(sock, backlog);
1397
1398 fput_light(sock->file, fput_needed);
1399 }
1400 return err;
1401 }
1402
1403 /*
1404 * For accept, we attempt to create a new socket, set up the link
1405 * with the client, wake up the client, then return the new
1406 * connected fd. We collect the address of the connector in kernel
1407 * space and move it to user at the very end. This is unclean because
1408 * we open the socket then return an error.
1409 *
1410 * 1003.1g adds the ability to recvmsg() to query connection pending
1411 * status to recvmsg. We need to add that support in a way thats
1412 * clean when we restucture accept also.
1413 */
1414
1415 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1416 int __user *, upeer_addrlen, int, flags)
1417 {
1418 struct socket *sock, *newsock;
1419 struct file *newfile;
1420 int err, len, newfd, fput_needed;
1421 struct sockaddr_storage address;
1422
1423 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1424 return -EINVAL;
1425
1426 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1427 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1428
1429 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1430 if (!sock)
1431 goto out;
1432
1433 err = -ENFILE;
1434 newsock = sock_alloc();
1435 if (!newsock)
1436 goto out_put;
1437
1438 newsock->type = sock->type;
1439 newsock->ops = sock->ops;
1440
1441 /*
1442 * We don't need try_module_get here, as the listening socket (sock)
1443 * has the protocol module (sock->ops->owner) held.
1444 */
1445 __module_get(newsock->ops->owner);
1446
1447 newfd = get_unused_fd_flags(flags);
1448 if (unlikely(newfd < 0)) {
1449 err = newfd;
1450 sock_release(newsock);
1451 goto out_put;
1452 }
1453 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1454 if (IS_ERR(newfile)) {
1455 err = PTR_ERR(newfile);
1456 put_unused_fd(newfd);
1457 sock_release(newsock);
1458 goto out_put;
1459 }
1460
1461 err = security_socket_accept(sock, newsock);
1462 if (err)
1463 goto out_fd;
1464
1465 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1466 if (err < 0)
1467 goto out_fd;
1468
1469 if (upeer_sockaddr) {
1470 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1471 &len, 2) < 0) {
1472 err = -ECONNABORTED;
1473 goto out_fd;
1474 }
1475 err = move_addr_to_user(&address,
1476 len, upeer_sockaddr, upeer_addrlen);
1477 if (err < 0)
1478 goto out_fd;
1479 }
1480
1481 /* File flags are not inherited via accept() unlike another OSes. */
1482
1483 fd_install(newfd, newfile);
1484 err = newfd;
1485
1486 out_put:
1487 fput_light(sock->file, fput_needed);
1488 out:
1489 return err;
1490 out_fd:
1491 fput(newfile);
1492 put_unused_fd(newfd);
1493 goto out_put;
1494 }
1495
1496 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1497 int __user *, upeer_addrlen)
1498 {
1499 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1500 }
1501
1502 /*
1503 * Attempt to connect to a socket with the server address. The address
1504 * is in user space so we verify it is OK and move it to kernel space.
1505 *
1506 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1507 * break bindings
1508 *
1509 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1510 * other SEQPACKET protocols that take time to connect() as it doesn't
1511 * include the -EINPROGRESS status for such sockets.
1512 */
1513
1514 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1515 int, addrlen)
1516 {
1517 struct socket *sock;
1518 struct sockaddr_storage address;
1519 int err, fput_needed;
1520
1521 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1522 if (!sock)
1523 goto out;
1524 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1525 if (err < 0)
1526 goto out_put;
1527
1528 err =
1529 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1530 if (err)
1531 goto out_put;
1532
1533 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1534 sock->file->f_flags);
1535 out_put:
1536 fput_light(sock->file, fput_needed);
1537 out:
1538 return err;
1539 }
1540
1541 /*
1542 * Get the local address ('name') of a socket object. Move the obtained
1543 * name to user space.
1544 */
1545
1546 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1547 int __user *, usockaddr_len)
1548 {
1549 struct socket *sock;
1550 struct sockaddr_storage address;
1551 int len, err, fput_needed;
1552
1553 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1554 if (!sock)
1555 goto out;
1556
1557 err = security_socket_getsockname(sock);
1558 if (err)
1559 goto out_put;
1560
1561 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1562 if (err)
1563 goto out_put;
1564 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1565
1566 out_put:
1567 fput_light(sock->file, fput_needed);
1568 out:
1569 return err;
1570 }
1571
1572 /*
1573 * Get the remote address ('name') of a socket object. Move the obtained
1574 * name to user space.
1575 */
1576
1577 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1578 int __user *, usockaddr_len)
1579 {
1580 struct socket *sock;
1581 struct sockaddr_storage address;
1582 int len, err, fput_needed;
1583
1584 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1585 if (sock != NULL) {
1586 err = security_socket_getpeername(sock);
1587 if (err) {
1588 fput_light(sock->file, fput_needed);
1589 return err;
1590 }
1591
1592 err =
1593 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1594 1);
1595 if (!err)
1596 err = move_addr_to_user(&address, len, usockaddr,
1597 usockaddr_len);
1598 fput_light(sock->file, fput_needed);
1599 }
1600 return err;
1601 }
1602
1603 /*
1604 * Send a datagram to a given address. We move the address into kernel
1605 * space and check the user space data area is readable before invoking
1606 * the protocol.
1607 */
1608
1609 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1610 unsigned int, flags, struct sockaddr __user *, addr,
1611 int, addr_len)
1612 {
1613 struct socket *sock;
1614 struct sockaddr_storage address;
1615 int err;
1616 struct msghdr msg;
1617 struct iovec iov;
1618 int fput_needed;
1619
1620 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1621 if (unlikely(err))
1622 return err;
1623 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1624 if (!sock)
1625 goto out;
1626
1627 msg.msg_name = NULL;
1628 msg.msg_control = NULL;
1629 msg.msg_controllen = 0;
1630 msg.msg_namelen = 0;
1631 if (addr) {
1632 err = move_addr_to_kernel(addr, addr_len, &address);
1633 if (err < 0)
1634 goto out_put;
1635 msg.msg_name = (struct sockaddr *)&address;
1636 msg.msg_namelen = addr_len;
1637 }
1638 if (sock->file->f_flags & O_NONBLOCK)
1639 flags |= MSG_DONTWAIT;
1640 msg.msg_flags = flags;
1641 err = sock_sendmsg(sock, &msg);
1642
1643 out_put:
1644 fput_light(sock->file, fput_needed);
1645 out:
1646 return err;
1647 }
1648
1649 /*
1650 * Send a datagram down a socket.
1651 */
1652
1653 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1654 unsigned int, flags)
1655 {
1656 return sys_sendto(fd, buff, len, flags, NULL, 0);
1657 }
1658
1659 /*
1660 * Receive a frame from the socket and optionally record the address of the
1661 * sender. We verify the buffers are writable and if needed move the
1662 * sender address from kernel to user space.
1663 */
1664
1665 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1666 unsigned int, flags, struct sockaddr __user *, addr,
1667 int __user *, addr_len)
1668 {
1669 struct socket *sock;
1670 struct iovec iov;
1671 struct msghdr msg;
1672 struct sockaddr_storage address;
1673 int err, err2;
1674 int fput_needed;
1675
1676 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1677 if (unlikely(err))
1678 return err;
1679 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1680 if (!sock)
1681 goto out;
1682
1683 msg.msg_control = NULL;
1684 msg.msg_controllen = 0;
1685 /* Save some cycles and don't copy the address if not needed */
1686 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1687 /* We assume all kernel code knows the size of sockaddr_storage */
1688 msg.msg_namelen = 0;
1689 msg.msg_iocb = NULL;
1690 if (sock->file->f_flags & O_NONBLOCK)
1691 flags |= MSG_DONTWAIT;
1692 err = sock_recvmsg(sock, &msg, flags);
1693
1694 if (err >= 0 && addr != NULL) {
1695 err2 = move_addr_to_user(&address,
1696 msg.msg_namelen, addr, addr_len);
1697 if (err2 < 0)
1698 err = err2;
1699 }
1700
1701 fput_light(sock->file, fput_needed);
1702 out:
1703 return err;
1704 }
1705
1706 /*
1707 * Receive a datagram from a socket.
1708 */
1709
1710 SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1711 unsigned int, flags)
1712 {
1713 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1714 }
1715
1716 /*
1717 * Set a socket option. Because we don't know the option lengths we have
1718 * to pass the user mode parameter for the protocols to sort out.
1719 */
1720
1721 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1722 char __user *, optval, int, optlen)
1723 {
1724 int err, fput_needed;
1725 struct socket *sock;
1726
1727 if (optlen < 0)
1728 return -EINVAL;
1729
1730 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1731 if (sock != NULL) {
1732 err = security_socket_setsockopt(sock, level, optname);
1733 if (err)
1734 goto out_put;
1735
1736 if (level == SOL_SOCKET)
1737 err =
1738 sock_setsockopt(sock, level, optname, optval,
1739 optlen);
1740 else
1741 err =
1742 sock->ops->setsockopt(sock, level, optname, optval,
1743 optlen);
1744 out_put:
1745 fput_light(sock->file, fput_needed);
1746 }
1747 return err;
1748 }
1749
1750 /*
1751 * Get a socket option. Because we don't know the option lengths we have
1752 * to pass a user mode parameter for the protocols to sort out.
1753 */
1754
1755 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1756 char __user *, optval, int __user *, optlen)
1757 {
1758 int err, fput_needed;
1759 struct socket *sock;
1760
1761 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1762 if (sock != NULL) {
1763 err = security_socket_getsockopt(sock, level, optname);
1764 if (err)
1765 goto out_put;
1766
1767 if (level == SOL_SOCKET)
1768 err =
1769 sock_getsockopt(sock, level, optname, optval,
1770 optlen);
1771 else
1772 err =
1773 sock->ops->getsockopt(sock, level, optname, optval,
1774 optlen);
1775 out_put:
1776 fput_light(sock->file, fput_needed);
1777 }
1778 return err;
1779 }
1780
1781 /*
1782 * Shutdown a socket.
1783 */
1784
1785 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1786 {
1787 int err, fput_needed;
1788 struct socket *sock;
1789
1790 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1791 if (sock != NULL) {
1792 err = security_socket_shutdown(sock, how);
1793 if (!err)
1794 err = sock->ops->shutdown(sock, how);
1795 fput_light(sock->file, fput_needed);
1796 }
1797 return err;
1798 }
1799
1800 /* A couple of helpful macros for getting the address of the 32/64 bit
1801 * fields which are the same type (int / unsigned) on our platforms.
1802 */
1803 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1804 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1805 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1806
1807 struct used_address {
1808 struct sockaddr_storage name;
1809 unsigned int name_len;
1810 };
1811
1812 static int copy_msghdr_from_user(struct msghdr *kmsg,
1813 struct user_msghdr __user *umsg,
1814 struct sockaddr __user **save_addr,
1815 struct iovec **iov)
1816 {
1817 struct sockaddr __user *uaddr;
1818 struct iovec __user *uiov;
1819 size_t nr_segs;
1820 ssize_t err;
1821
1822 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1823 __get_user(uaddr, &umsg->msg_name) ||
1824 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1825 __get_user(uiov, &umsg->msg_iov) ||
1826 __get_user(nr_segs, &umsg->msg_iovlen) ||
1827 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1828 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1829 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1830 return -EFAULT;
1831
1832 if (!uaddr)
1833 kmsg->msg_namelen = 0;
1834
1835 if (kmsg->msg_namelen < 0)
1836 return -EINVAL;
1837
1838 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1839 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1840
1841 if (save_addr)
1842 *save_addr = uaddr;
1843
1844 if (uaddr && kmsg->msg_namelen) {
1845 if (!save_addr) {
1846 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1847 kmsg->msg_name);
1848 if (err < 0)
1849 return err;
1850 }
1851 } else {
1852 kmsg->msg_name = NULL;
1853 kmsg->msg_namelen = 0;
1854 }
1855
1856 if (nr_segs > UIO_MAXIOV)
1857 return -EMSGSIZE;
1858
1859 kmsg->msg_iocb = NULL;
1860
1861 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1862 UIO_FASTIOV, iov, &kmsg->msg_iter);
1863 }
1864
1865 static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1866 struct msghdr *msg_sys, unsigned int flags,
1867 struct used_address *used_address,
1868 unsigned int allowed_msghdr_flags)
1869 {
1870 struct compat_msghdr __user *msg_compat =
1871 (struct compat_msghdr __user *)msg;
1872 struct sockaddr_storage address;
1873 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1874 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1875 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1876 /* 20 is size of ipv6_pktinfo */
1877 unsigned char *ctl_buf = ctl;
1878 int ctl_len;
1879 ssize_t err;
1880
1881 msg_sys->msg_name = &address;
1882
1883 if (MSG_CMSG_COMPAT & flags)
1884 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1885 else
1886 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1887 if (err < 0)
1888 return err;
1889
1890 err = -ENOBUFS;
1891
1892 if (msg_sys->msg_controllen > INT_MAX)
1893 goto out_freeiov;
1894 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1895 ctl_len = msg_sys->msg_controllen;
1896 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1897 err =
1898 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1899 sizeof(ctl));
1900 if (err)
1901 goto out_freeiov;
1902 ctl_buf = msg_sys->msg_control;
1903 ctl_len = msg_sys->msg_controllen;
1904 } else if (ctl_len) {
1905 if (ctl_len > sizeof(ctl)) {
1906 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1907 if (ctl_buf == NULL)
1908 goto out_freeiov;
1909 }
1910 err = -EFAULT;
1911 /*
1912 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1913 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1914 * checking falls down on this.
1915 */
1916 if (copy_from_user(ctl_buf,
1917 (void __user __force *)msg_sys->msg_control,
1918 ctl_len))
1919 goto out_freectl;
1920 msg_sys->msg_control = ctl_buf;
1921 }
1922 msg_sys->msg_flags = flags;
1923
1924 if (sock->file->f_flags & O_NONBLOCK)
1925 msg_sys->msg_flags |= MSG_DONTWAIT;
1926 /*
1927 * If this is sendmmsg() and current destination address is same as
1928 * previously succeeded address, omit asking LSM's decision.
1929 * used_address->name_len is initialized to UINT_MAX so that the first
1930 * destination address never matches.
1931 */
1932 if (used_address && msg_sys->msg_name &&
1933 used_address->name_len == msg_sys->msg_namelen &&
1934 !memcmp(&used_address->name, msg_sys->msg_name,
1935 used_address->name_len)) {
1936 err = sock_sendmsg_nosec(sock, msg_sys);
1937 goto out_freectl;
1938 }
1939 err = sock_sendmsg(sock, msg_sys);
1940 /*
1941 * If this is sendmmsg() and sending to current destination address was
1942 * successful, remember it.
1943 */
1944 if (used_address && err >= 0) {
1945 used_address->name_len = msg_sys->msg_namelen;
1946 if (msg_sys->msg_name)
1947 memcpy(&used_address->name, msg_sys->msg_name,
1948 used_address->name_len);
1949 }
1950
1951 out_freectl:
1952 if (ctl_buf != ctl)
1953 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1954 out_freeiov:
1955 kfree(iov);
1956 return err;
1957 }
1958
1959 /*
1960 * BSD sendmsg interface
1961 */
1962
1963 long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1964 {
1965 int fput_needed, err;
1966 struct msghdr msg_sys;
1967 struct socket *sock;
1968
1969 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1970 if (!sock)
1971 goto out;
1972
1973 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
1974
1975 fput_light(sock->file, fput_needed);
1976 out:
1977 return err;
1978 }
1979
1980 SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1981 {
1982 if (flags & MSG_CMSG_COMPAT)
1983 return -EINVAL;
1984 return __sys_sendmsg(fd, msg, flags);
1985 }
1986
1987 /*
1988 * Linux sendmmsg interface
1989 */
1990
1991 int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1992 unsigned int flags)
1993 {
1994 int fput_needed, err, datagrams;
1995 struct socket *sock;
1996 struct mmsghdr __user *entry;
1997 struct compat_mmsghdr __user *compat_entry;
1998 struct msghdr msg_sys;
1999 struct used_address used_address;
2000 unsigned int oflags = flags;
2001
2002 if (vlen > UIO_MAXIOV)
2003 vlen = UIO_MAXIOV;
2004
2005 datagrams = 0;
2006
2007 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2008 if (!sock)
2009 return err;
2010
2011 used_address.name_len = UINT_MAX;
2012 entry = mmsg;
2013 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2014 err = 0;
2015 flags |= MSG_BATCH;
2016
2017 while (datagrams < vlen) {
2018 if (datagrams == vlen - 1)
2019 flags = oflags;
2020
2021 if (MSG_CMSG_COMPAT & flags) {
2022 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2023 &msg_sys, flags, &used_address, MSG_EOR);
2024 if (err < 0)
2025 break;
2026 err = __put_user(err, &compat_entry->msg_len);
2027 ++compat_entry;
2028 } else {
2029 err = ___sys_sendmsg(sock,
2030 (struct user_msghdr __user *)entry,
2031 &msg_sys, flags, &used_address, MSG_EOR);
2032 if (err < 0)
2033 break;
2034 err = put_user(err, &entry->msg_len);
2035 ++entry;
2036 }
2037
2038 if (err)
2039 break;
2040 ++datagrams;
2041 if (msg_data_left(&msg_sys))
2042 break;
2043 cond_resched();
2044 }
2045
2046 fput_light(sock->file, fput_needed);
2047
2048 /* We only return an error if no datagrams were able to be sent */
2049 if (datagrams != 0)
2050 return datagrams;
2051
2052 return err;
2053 }
2054
2055 SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2056 unsigned int, vlen, unsigned int, flags)
2057 {
2058 if (flags & MSG_CMSG_COMPAT)
2059 return -EINVAL;
2060 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2061 }
2062
2063 static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2064 struct msghdr *msg_sys, unsigned int flags, int nosec)
2065 {
2066 struct compat_msghdr __user *msg_compat =
2067 (struct compat_msghdr __user *)msg;
2068 struct iovec iovstack[UIO_FASTIOV];
2069 struct iovec *iov = iovstack;
2070 unsigned long cmsg_ptr;
2071 int len;
2072 ssize_t err;
2073
2074 /* kernel mode address */
2075 struct sockaddr_storage addr;
2076
2077 /* user mode address pointers */
2078 struct sockaddr __user *uaddr;
2079 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2080
2081 msg_sys->msg_name = &addr;
2082
2083 if (MSG_CMSG_COMPAT & flags)
2084 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2085 else
2086 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2087 if (err < 0)
2088 return err;
2089
2090 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2091 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2092
2093 /* We assume all kernel code knows the size of sockaddr_storage */
2094 msg_sys->msg_namelen = 0;
2095
2096 if (sock->file->f_flags & O_NONBLOCK)
2097 flags |= MSG_DONTWAIT;
2098 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2099 if (err < 0)
2100 goto out_freeiov;
2101 len = err;
2102
2103 if (uaddr != NULL) {
2104 err = move_addr_to_user(&addr,
2105 msg_sys->msg_namelen, uaddr,
2106 uaddr_len);
2107 if (err < 0)
2108 goto out_freeiov;
2109 }
2110 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2111 COMPAT_FLAGS(msg));
2112 if (err)
2113 goto out_freeiov;
2114 if (MSG_CMSG_COMPAT & flags)
2115 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2116 &msg_compat->msg_controllen);
2117 else
2118 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2119 &msg->msg_controllen);
2120 if (err)
2121 goto out_freeiov;
2122 err = len;
2123
2124 out_freeiov:
2125 kfree(iov);
2126 return err;
2127 }
2128
2129 /*
2130 * BSD recvmsg interface
2131 */
2132
2133 long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2134 {
2135 int fput_needed, err;
2136 struct msghdr msg_sys;
2137 struct socket *sock;
2138
2139 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2140 if (!sock)
2141 goto out;
2142
2143 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2144
2145 fput_light(sock->file, fput_needed);
2146 out:
2147 return err;
2148 }
2149
2150 SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2151 unsigned int, flags)
2152 {
2153 if (flags & MSG_CMSG_COMPAT)
2154 return -EINVAL;
2155 return __sys_recvmsg(fd, msg, flags);
2156 }
2157
2158 /*
2159 * Linux recvmmsg interface
2160 */
2161
2162 int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2163 unsigned int flags, struct timespec *timeout)
2164 {
2165 int fput_needed, err, datagrams;
2166 struct socket *sock;
2167 struct mmsghdr __user *entry;
2168 struct compat_mmsghdr __user *compat_entry;
2169 struct msghdr msg_sys;
2170 struct timespec64 end_time;
2171 struct timespec64 timeout64;
2172
2173 if (timeout &&
2174 poll_select_set_timeout(&end_time, timeout->tv_sec,
2175 timeout->tv_nsec))
2176 return -EINVAL;
2177
2178 datagrams = 0;
2179
2180 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2181 if (!sock)
2182 return err;
2183
2184 err = sock_error(sock->sk);
2185 if (err)
2186 goto out_put;
2187
2188 entry = mmsg;
2189 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2190
2191 while (datagrams < vlen) {
2192 /*
2193 * No need to ask LSM for more than the first datagram.
2194 */
2195 if (MSG_CMSG_COMPAT & flags) {
2196 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2197 &msg_sys, flags & ~MSG_WAITFORONE,
2198 datagrams);
2199 if (err < 0)
2200 break;
2201 err = __put_user(err, &compat_entry->msg_len);
2202 ++compat_entry;
2203 } else {
2204 err = ___sys_recvmsg(sock,
2205 (struct user_msghdr __user *)entry,
2206 &msg_sys, flags & ~MSG_WAITFORONE,
2207 datagrams);
2208 if (err < 0)
2209 break;
2210 err = put_user(err, &entry->msg_len);
2211 ++entry;
2212 }
2213
2214 if (err)
2215 break;
2216 ++datagrams;
2217
2218 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2219 if (flags & MSG_WAITFORONE)
2220 flags |= MSG_DONTWAIT;
2221
2222 if (timeout) {
2223 ktime_get_ts64(&timeout64);
2224 *timeout = timespec64_to_timespec(
2225 timespec64_sub(end_time, timeout64));
2226 if (timeout->tv_sec < 0) {
2227 timeout->tv_sec = timeout->tv_nsec = 0;
2228 break;
2229 }
2230
2231 /* Timeout, return less than vlen datagrams */
2232 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2233 break;
2234 }
2235
2236 /* Out of band data, return right away */
2237 if (msg_sys.msg_flags & MSG_OOB)
2238 break;
2239 cond_resched();
2240 }
2241
2242 if (err == 0)
2243 goto out_put;
2244
2245 if (datagrams == 0) {
2246 datagrams = err;
2247 goto out_put;
2248 }
2249
2250 /*
2251 * We may return less entries than requested (vlen) if the
2252 * sock is non block and there aren't enough datagrams...
2253 */
2254 if (err != -EAGAIN) {
2255 /*
2256 * ... or if recvmsg returns an error after we
2257 * received some datagrams, where we record the
2258 * error to return on the next call or if the
2259 * app asks about it using getsockopt(SO_ERROR).
2260 */
2261 sock->sk->sk_err = -err;
2262 }
2263 out_put:
2264 fput_light(sock->file, fput_needed);
2265
2266 return datagrams;
2267 }
2268
2269 SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2270 unsigned int, vlen, unsigned int, flags,
2271 struct timespec __user *, timeout)
2272 {
2273 int datagrams;
2274 struct timespec timeout_sys;
2275
2276 if (flags & MSG_CMSG_COMPAT)
2277 return -EINVAL;
2278
2279 if (!timeout)
2280 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2281
2282 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2283 return -EFAULT;
2284
2285 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2286
2287 if (datagrams > 0 &&
2288 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2289 datagrams = -EFAULT;
2290
2291 return datagrams;
2292 }
2293
2294 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2295 /* Argument list sizes for sys_socketcall */
2296 #define AL(x) ((x) * sizeof(unsigned long))
2297 static const unsigned char nargs[21] = {
2298 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2299 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2300 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2301 AL(4), AL(5), AL(4)
2302 };
2303
2304 #undef AL
2305
2306 /*
2307 * System call vectors.
2308 *
2309 * Argument checking cleaned up. Saved 20% in size.
2310 * This function doesn't need to set the kernel lock because
2311 * it is set by the callees.
2312 */
2313
2314 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2315 {
2316 unsigned long a[AUDITSC_ARGS];
2317 unsigned long a0, a1;
2318 int err;
2319 unsigned int len;
2320
2321 if (call < 1 || call > SYS_SENDMMSG)
2322 return -EINVAL;
2323
2324 len = nargs[call];
2325 if (len > sizeof(a))
2326 return -EINVAL;
2327
2328 /* copy_from_user should be SMP safe. */
2329 if (copy_from_user(a, args, len))
2330 return -EFAULT;
2331
2332 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2333 if (err)
2334 return err;
2335
2336 a0 = a[0];
2337 a1 = a[1];
2338
2339 switch (call) {
2340 case SYS_SOCKET:
2341 err = sys_socket(a0, a1, a[2]);
2342 break;
2343 case SYS_BIND:
2344 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2345 break;
2346 case SYS_CONNECT:
2347 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2348 break;
2349 case SYS_LISTEN:
2350 err = sys_listen(a0, a1);
2351 break;
2352 case SYS_ACCEPT:
2353 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2354 (int __user *)a[2], 0);
2355 break;
2356 case SYS_GETSOCKNAME:
2357 err =
2358 sys_getsockname(a0, (struct sockaddr __user *)a1,
2359 (int __user *)a[2]);
2360 break;
2361 case SYS_GETPEERNAME:
2362 err =
2363 sys_getpeername(a0, (struct sockaddr __user *)a1,
2364 (int __user *)a[2]);
2365 break;
2366 case SYS_SOCKETPAIR:
2367 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2368 break;
2369 case SYS_SEND:
2370 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2371 break;
2372 case SYS_SENDTO:
2373 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2374 (struct sockaddr __user *)a[4], a[5]);
2375 break;
2376 case SYS_RECV:
2377 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2378 break;
2379 case SYS_RECVFROM:
2380 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2381 (struct sockaddr __user *)a[4],
2382 (int __user *)a[5]);
2383 break;
2384 case SYS_SHUTDOWN:
2385 err = sys_shutdown(a0, a1);
2386 break;
2387 case SYS_SETSOCKOPT:
2388 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2389 break;
2390 case SYS_GETSOCKOPT:
2391 err =
2392 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2393 (int __user *)a[4]);
2394 break;
2395 case SYS_SENDMSG:
2396 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2397 break;
2398 case SYS_SENDMMSG:
2399 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2400 break;
2401 case SYS_RECVMSG:
2402 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2403 break;
2404 case SYS_RECVMMSG:
2405 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2406 (struct timespec __user *)a[4]);
2407 break;
2408 case SYS_ACCEPT4:
2409 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2410 (int __user *)a[2], a[3]);
2411 break;
2412 default:
2413 err = -EINVAL;
2414 break;
2415 }
2416 return err;
2417 }
2418
2419 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2420
2421 /**
2422 * sock_register - add a socket protocol handler
2423 * @ops: description of protocol
2424 *
2425 * This function is called by a protocol handler that wants to
2426 * advertise its address family, and have it linked into the
2427 * socket interface. The value ops->family corresponds to the
2428 * socket system call protocol family.
2429 */
2430 int sock_register(const struct net_proto_family *ops)
2431 {
2432 int err;
2433
2434 if (ops->family >= NPROTO) {
2435 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2436 return -ENOBUFS;
2437 }
2438
2439 spin_lock(&net_family_lock);
2440 if (rcu_dereference_protected(net_families[ops->family],
2441 lockdep_is_held(&net_family_lock)))
2442 err = -EEXIST;
2443 else {
2444 rcu_assign_pointer(net_families[ops->family], ops);
2445 err = 0;
2446 }
2447 spin_unlock(&net_family_lock);
2448
2449 pr_info("NET: Registered protocol family %d\n", ops->family);
2450 return err;
2451 }
2452 EXPORT_SYMBOL(sock_register);
2453
2454 /**
2455 * sock_unregister - remove a protocol handler
2456 * @family: protocol family to remove
2457 *
2458 * This function is called by a protocol handler that wants to
2459 * remove its address family, and have it unlinked from the
2460 * new socket creation.
2461 *
2462 * If protocol handler is a module, then it can use module reference
2463 * counts to protect against new references. If protocol handler is not
2464 * a module then it needs to provide its own protection in
2465 * the ops->create routine.
2466 */
2467 void sock_unregister(int family)
2468 {
2469 BUG_ON(family < 0 || family >= NPROTO);
2470
2471 spin_lock(&net_family_lock);
2472 RCU_INIT_POINTER(net_families[family], NULL);
2473 spin_unlock(&net_family_lock);
2474
2475 synchronize_rcu();
2476
2477 pr_info("NET: Unregistered protocol family %d\n", family);
2478 }
2479 EXPORT_SYMBOL(sock_unregister);
2480
2481 static int __init sock_init(void)
2482 {
2483 int err;
2484 /*
2485 * Initialize the network sysctl infrastructure.
2486 */
2487 err = net_sysctl_init();
2488 if (err)
2489 goto out;
2490
2491 /*
2492 * Initialize skbuff SLAB cache
2493 */
2494 skb_init();
2495
2496 /*
2497 * Initialize the protocols module.
2498 */
2499
2500 init_inodecache();
2501
2502 err = register_filesystem(&sock_fs_type);
2503 if (err)
2504 goto out_fs;
2505 sock_mnt = kern_mount(&sock_fs_type);
2506 if (IS_ERR(sock_mnt)) {
2507 err = PTR_ERR(sock_mnt);
2508 goto out_mount;
2509 }
2510
2511 /* The real protocol initialization is performed in later initcalls.
2512 */
2513
2514 #ifdef CONFIG_NETFILTER
2515 err = netfilter_init();
2516 if (err)
2517 goto out;
2518 #endif
2519
2520 ptp_classifier_init();
2521
2522 out:
2523 return err;
2524
2525 out_mount:
2526 unregister_filesystem(&sock_fs_type);
2527 out_fs:
2528 goto out;
2529 }
2530
2531 core_initcall(sock_init); /* early initcall */
2532
2533 #ifdef CONFIG_PROC_FS
2534 void socket_seq_show(struct seq_file *seq)
2535 {
2536 int cpu;
2537 int counter = 0;
2538
2539 for_each_possible_cpu(cpu)
2540 counter += per_cpu(sockets_in_use, cpu);
2541
2542 /* It can be negative, by the way. 8) */
2543 if (counter < 0)
2544 counter = 0;
2545
2546 seq_printf(seq, "sockets: used %d\n", counter);
2547 }
2548 #endif /* CONFIG_PROC_FS */
2549
2550 #ifdef CONFIG_COMPAT
2551 static int do_siocgstamp(struct net *net, struct socket *sock,
2552 unsigned int cmd, void __user *up)
2553 {
2554 mm_segment_t old_fs = get_fs();
2555 struct timeval ktv;
2556 int err;
2557
2558 set_fs(KERNEL_DS);
2559 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2560 set_fs(old_fs);
2561 if (!err)
2562 err = compat_put_timeval(&ktv, up);
2563
2564 return err;
2565 }
2566
2567 static int do_siocgstampns(struct net *net, struct socket *sock,
2568 unsigned int cmd, void __user *up)
2569 {
2570 mm_segment_t old_fs = get_fs();
2571 struct timespec kts;
2572 int err;
2573
2574 set_fs(KERNEL_DS);
2575 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2576 set_fs(old_fs);
2577 if (!err)
2578 err = compat_put_timespec(&kts, up);
2579
2580 return err;
2581 }
2582
2583 static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2584 {
2585 struct ifreq __user *uifr;
2586 int err;
2587
2588 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2589 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2590 return -EFAULT;
2591
2592 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2593 if (err)
2594 return err;
2595
2596 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2597 return -EFAULT;
2598
2599 return 0;
2600 }
2601
2602 static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2603 {
2604 struct compat_ifconf ifc32;
2605 struct ifconf ifc;
2606 struct ifconf __user *uifc;
2607 struct compat_ifreq __user *ifr32;
2608 struct ifreq __user *ifr;
2609 unsigned int i, j;
2610 int err;
2611
2612 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2613 return -EFAULT;
2614
2615 memset(&ifc, 0, sizeof(ifc));
2616 if (ifc32.ifcbuf == 0) {
2617 ifc32.ifc_len = 0;
2618 ifc.ifc_len = 0;
2619 ifc.ifc_req = NULL;
2620 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2621 } else {
2622 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2623 sizeof(struct ifreq);
2624 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2625 ifc.ifc_len = len;
2626 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2627 ifr32 = compat_ptr(ifc32.ifcbuf);
2628 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2629 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2630 return -EFAULT;
2631 ifr++;
2632 ifr32++;
2633 }
2634 }
2635 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2636 return -EFAULT;
2637
2638 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2639 if (err)
2640 return err;
2641
2642 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2643 return -EFAULT;
2644
2645 ifr = ifc.ifc_req;
2646 ifr32 = compat_ptr(ifc32.ifcbuf);
2647 for (i = 0, j = 0;
2648 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2649 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2650 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2651 return -EFAULT;
2652 ifr32++;
2653 ifr++;
2654 }
2655
2656 if (ifc32.ifcbuf == 0) {
2657 /* Translate from 64-bit structure multiple to
2658 * a 32-bit one.
2659 */
2660 i = ifc.ifc_len;
2661 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2662 ifc32.ifc_len = i;
2663 } else {
2664 ifc32.ifc_len = i;
2665 }
2666 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2667 return -EFAULT;
2668
2669 return 0;
2670 }
2671
2672 static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2673 {
2674 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2675 bool convert_in = false, convert_out = false;
2676 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2677 struct ethtool_rxnfc __user *rxnfc;
2678 struct ifreq __user *ifr;
2679 u32 rule_cnt = 0, actual_rule_cnt;
2680 u32 ethcmd;
2681 u32 data;
2682 int ret;
2683
2684 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2685 return -EFAULT;
2686
2687 compat_rxnfc = compat_ptr(data);
2688
2689 if (get_user(ethcmd, &compat_rxnfc->cmd))
2690 return -EFAULT;
2691
2692 /* Most ethtool structures are defined without padding.
2693 * Unfortunately struct ethtool_rxnfc is an exception.
2694 */
2695 switch (ethcmd) {
2696 default:
2697 break;
2698 case ETHTOOL_GRXCLSRLALL:
2699 /* Buffer size is variable */
2700 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2701 return -EFAULT;
2702 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2703 return -ENOMEM;
2704 buf_size += rule_cnt * sizeof(u32);
2705 /* fall through */
2706 case ETHTOOL_GRXRINGS:
2707 case ETHTOOL_GRXCLSRLCNT:
2708 case ETHTOOL_GRXCLSRULE:
2709 case ETHTOOL_SRXCLSRLINS:
2710 convert_out = true;
2711 /* fall through */
2712 case ETHTOOL_SRXCLSRLDEL:
2713 buf_size += sizeof(struct ethtool_rxnfc);
2714 convert_in = true;
2715 break;
2716 }
2717
2718 ifr = compat_alloc_user_space(buf_size);
2719 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2720
2721 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2722 return -EFAULT;
2723
2724 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2725 &ifr->ifr_ifru.ifru_data))
2726 return -EFAULT;
2727
2728 if (convert_in) {
2729 /* We expect there to be holes between fs.m_ext and
2730 * fs.ring_cookie and at the end of fs, but nowhere else.
2731 */
2732 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2733 sizeof(compat_rxnfc->fs.m_ext) !=
2734 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2735 sizeof(rxnfc->fs.m_ext));
2736 BUILD_BUG_ON(
2737 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2738 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2739 offsetof(struct ethtool_rxnfc, fs.location) -
2740 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2741
2742 if (copy_in_user(rxnfc, compat_rxnfc,
2743 (void __user *)(&rxnfc->fs.m_ext + 1) -
2744 (void __user *)rxnfc) ||
2745 copy_in_user(&rxnfc->fs.ring_cookie,
2746 &compat_rxnfc->fs.ring_cookie,
2747 (void __user *)(&rxnfc->fs.location + 1) -
2748 (void __user *)&rxnfc->fs.ring_cookie) ||
2749 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2750 sizeof(rxnfc->rule_cnt)))
2751 return -EFAULT;
2752 }
2753
2754 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2755 if (ret)
2756 return ret;
2757
2758 if (convert_out) {
2759 if (copy_in_user(compat_rxnfc, rxnfc,
2760 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2761 (const void __user *)rxnfc) ||
2762 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2763 &rxnfc->fs.ring_cookie,
2764 (const void __user *)(&rxnfc->fs.location + 1) -
2765 (const void __user *)&rxnfc->fs.ring_cookie) ||
2766 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2767 sizeof(rxnfc->rule_cnt)))
2768 return -EFAULT;
2769
2770 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2771 /* As an optimisation, we only copy the actual
2772 * number of rules that the underlying
2773 * function returned. Since Mallory might
2774 * change the rule count in user memory, we
2775 * check that it is less than the rule count
2776 * originally given (as the user buffer size),
2777 * which has been range-checked.
2778 */
2779 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2780 return -EFAULT;
2781 if (actual_rule_cnt < rule_cnt)
2782 rule_cnt = actual_rule_cnt;
2783 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2784 &rxnfc->rule_locs[0],
2785 rule_cnt * sizeof(u32)))
2786 return -EFAULT;
2787 }
2788 }
2789
2790 return 0;
2791 }
2792
2793 static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2794 {
2795 void __user *uptr;
2796 compat_uptr_t uptr32;
2797 struct ifreq __user *uifr;
2798
2799 uifr = compat_alloc_user_space(sizeof(*uifr));
2800 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2801 return -EFAULT;
2802
2803 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2804 return -EFAULT;
2805
2806 uptr = compat_ptr(uptr32);
2807
2808 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2809 return -EFAULT;
2810
2811 return dev_ioctl(net, SIOCWANDEV, uifr);
2812 }
2813
2814 static int bond_ioctl(struct net *net, unsigned int cmd,
2815 struct compat_ifreq __user *ifr32)
2816 {
2817 struct ifreq kifr;
2818 mm_segment_t old_fs;
2819 int err;
2820
2821 switch (cmd) {
2822 case SIOCBONDENSLAVE:
2823 case SIOCBONDRELEASE:
2824 case SIOCBONDSETHWADDR:
2825 case SIOCBONDCHANGEACTIVE:
2826 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2827 return -EFAULT;
2828
2829 old_fs = get_fs();
2830 set_fs(KERNEL_DS);
2831 err = dev_ioctl(net, cmd,
2832 (struct ifreq __user __force *) &kifr);
2833 set_fs(old_fs);
2834
2835 return err;
2836 default:
2837 return -ENOIOCTLCMD;
2838 }
2839 }
2840
2841 /* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2842 static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2843 struct compat_ifreq __user *u_ifreq32)
2844 {
2845 struct ifreq __user *u_ifreq64;
2846 char tmp_buf[IFNAMSIZ];
2847 void __user *data64;
2848 u32 data32;
2849
2850 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2851 IFNAMSIZ))
2852 return -EFAULT;
2853 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2854 return -EFAULT;
2855 data64 = compat_ptr(data32);
2856
2857 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2858
2859 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2860 IFNAMSIZ))
2861 return -EFAULT;
2862 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2863 return -EFAULT;
2864
2865 return dev_ioctl(net, cmd, u_ifreq64);
2866 }
2867
2868 static int dev_ifsioc(struct net *net, struct socket *sock,
2869 unsigned int cmd, struct compat_ifreq __user *uifr32)
2870 {
2871 struct ifreq __user *uifr;
2872 int err;
2873
2874 uifr = compat_alloc_user_space(sizeof(*uifr));
2875 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2876 return -EFAULT;
2877
2878 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2879
2880 if (!err) {
2881 switch (cmd) {
2882 case SIOCGIFFLAGS:
2883 case SIOCGIFMETRIC:
2884 case SIOCGIFMTU:
2885 case SIOCGIFMEM:
2886 case SIOCGIFHWADDR:
2887 case SIOCGIFINDEX:
2888 case SIOCGIFADDR:
2889 case SIOCGIFBRDADDR:
2890 case SIOCGIFDSTADDR:
2891 case SIOCGIFNETMASK:
2892 case SIOCGIFPFLAGS:
2893 case SIOCGIFTXQLEN:
2894 case SIOCGMIIPHY:
2895 case SIOCGMIIREG:
2896 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2897 err = -EFAULT;
2898 break;
2899 }
2900 }
2901 return err;
2902 }
2903
2904 static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2905 struct compat_ifreq __user *uifr32)
2906 {
2907 struct ifreq ifr;
2908 struct compat_ifmap __user *uifmap32;
2909 mm_segment_t old_fs;
2910 int err;
2911
2912 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2913 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2914 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2915 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2916 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2917 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2918 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2919 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2920 if (err)
2921 return -EFAULT;
2922
2923 old_fs = get_fs();
2924 set_fs(KERNEL_DS);
2925 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2926 set_fs(old_fs);
2927
2928 if (cmd == SIOCGIFMAP && !err) {
2929 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2930 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2931 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2932 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2933 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2934 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2935 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2936 if (err)
2937 err = -EFAULT;
2938 }
2939 return err;
2940 }
2941
2942 struct rtentry32 {
2943 u32 rt_pad1;
2944 struct sockaddr rt_dst; /* target address */
2945 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2946 struct sockaddr rt_genmask; /* target network mask (IP) */
2947 unsigned short rt_flags;
2948 short rt_pad2;
2949 u32 rt_pad3;
2950 unsigned char rt_tos;
2951 unsigned char rt_class;
2952 short rt_pad4;
2953 short rt_metric; /* +1 for binary compatibility! */
2954 /* char * */ u32 rt_dev; /* forcing the device at add */
2955 u32 rt_mtu; /* per route MTU/Window */
2956 u32 rt_window; /* Window clamping */
2957 unsigned short rt_irtt; /* Initial RTT */
2958 };
2959
2960 struct in6_rtmsg32 {
2961 struct in6_addr rtmsg_dst;
2962 struct in6_addr rtmsg_src;
2963 struct in6_addr rtmsg_gateway;
2964 u32 rtmsg_type;
2965 u16 rtmsg_dst_len;
2966 u16 rtmsg_src_len;
2967 u32 rtmsg_metric;
2968 u32 rtmsg_info;
2969 u32 rtmsg_flags;
2970 s32 rtmsg_ifindex;
2971 };
2972
2973 static int routing_ioctl(struct net *net, struct socket *sock,
2974 unsigned int cmd, void __user *argp)
2975 {
2976 int ret;
2977 void *r = NULL;
2978 struct in6_rtmsg r6;
2979 struct rtentry r4;
2980 char devname[16];
2981 u32 rtdev;
2982 mm_segment_t old_fs = get_fs();
2983
2984 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2985 struct in6_rtmsg32 __user *ur6 = argp;
2986 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2987 3 * sizeof(struct in6_addr));
2988 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2989 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2990 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2991 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2992 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2993 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2994 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
2995
2996 r = (void *) &r6;
2997 } else { /* ipv4 */
2998 struct rtentry32 __user *ur4 = argp;
2999 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3000 3 * sizeof(struct sockaddr));
3001 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3002 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3003 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3004 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3005 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3006 ret |= get_user(rtdev, &(ur4->rt_dev));
3007 if (rtdev) {
3008 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3009 r4.rt_dev = (char __user __force *)devname;
3010 devname[15] = 0;
3011 } else
3012 r4.rt_dev = NULL;
3013
3014 r = (void *) &r4;
3015 }
3016
3017 if (ret) {
3018 ret = -EFAULT;
3019 goto out;
3020 }
3021
3022 set_fs(KERNEL_DS);
3023 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3024 set_fs(old_fs);
3025
3026 out:
3027 return ret;
3028 }
3029
3030 /* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3031 * for some operations; this forces use of the newer bridge-utils that
3032 * use compatible ioctls
3033 */
3034 static int old_bridge_ioctl(compat_ulong_t __user *argp)
3035 {
3036 compat_ulong_t tmp;
3037
3038 if (get_user(tmp, argp))
3039 return -EFAULT;
3040 if (tmp == BRCTL_GET_VERSION)
3041 return BRCTL_VERSION + 1;
3042 return -EINVAL;
3043 }
3044
3045 static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3046 unsigned int cmd, unsigned long arg)
3047 {
3048 void __user *argp = compat_ptr(arg);
3049 struct sock *sk = sock->sk;
3050 struct net *net = sock_net(sk);
3051
3052 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3053 return compat_ifr_data_ioctl(net, cmd, argp);
3054
3055 switch (cmd) {
3056 case SIOCSIFBR:
3057 case SIOCGIFBR:
3058 return old_bridge_ioctl(argp);
3059 case SIOCGIFNAME:
3060 return dev_ifname32(net, argp);
3061 case SIOCGIFCONF:
3062 return dev_ifconf(net, argp);
3063 case SIOCETHTOOL:
3064 return ethtool_ioctl(net, argp);
3065 case SIOCWANDEV:
3066 return compat_siocwandev(net, argp);
3067 case SIOCGIFMAP:
3068 case SIOCSIFMAP:
3069 return compat_sioc_ifmap(net, cmd, argp);
3070 case SIOCBONDENSLAVE:
3071 case SIOCBONDRELEASE:
3072 case SIOCBONDSETHWADDR:
3073 case SIOCBONDCHANGEACTIVE:
3074 return bond_ioctl(net, cmd, argp);
3075 case SIOCADDRT:
3076 case SIOCDELRT:
3077 return routing_ioctl(net, sock, cmd, argp);
3078 case SIOCGSTAMP:
3079 return do_siocgstamp(net, sock, cmd, argp);
3080 case SIOCGSTAMPNS:
3081 return do_siocgstampns(net, sock, cmd, argp);
3082 case SIOCBONDSLAVEINFOQUERY:
3083 case SIOCBONDINFOQUERY:
3084 case SIOCSHWTSTAMP:
3085 case SIOCGHWTSTAMP:
3086 return compat_ifr_data_ioctl(net, cmd, argp);
3087
3088 case FIOSETOWN:
3089 case SIOCSPGRP:
3090 case FIOGETOWN:
3091 case SIOCGPGRP:
3092 case SIOCBRADDBR:
3093 case SIOCBRDELBR:
3094 case SIOCGIFVLAN:
3095 case SIOCSIFVLAN:
3096 case SIOCADDDLCI:
3097 case SIOCDELDLCI:
3098 return sock_ioctl(file, cmd, arg);
3099
3100 case SIOCGIFFLAGS:
3101 case SIOCSIFFLAGS:
3102 case SIOCGIFMETRIC:
3103 case SIOCSIFMETRIC:
3104 case SIOCGIFMTU:
3105 case SIOCSIFMTU:
3106 case SIOCGIFMEM:
3107 case SIOCSIFMEM:
3108 case SIOCGIFHWADDR:
3109 case SIOCSIFHWADDR:
3110 case SIOCADDMULTI:
3111 case SIOCDELMULTI:
3112 case SIOCGIFINDEX:
3113 case SIOCGIFADDR:
3114 case SIOCSIFADDR:
3115 case SIOCSIFHWBROADCAST:
3116 case SIOCDIFADDR:
3117 case SIOCGIFBRDADDR:
3118 case SIOCSIFBRDADDR:
3119 case SIOCGIFDSTADDR:
3120 case SIOCSIFDSTADDR:
3121 case SIOCGIFNETMASK:
3122 case SIOCSIFNETMASK:
3123 case SIOCSIFPFLAGS:
3124 case SIOCGIFPFLAGS:
3125 case SIOCGIFTXQLEN:
3126 case SIOCSIFTXQLEN:
3127 case SIOCBRADDIF:
3128 case SIOCBRDELIF:
3129 case SIOCSIFNAME:
3130 case SIOCGMIIPHY:
3131 case SIOCGMIIREG:
3132 case SIOCSMIIREG:
3133 return dev_ifsioc(net, sock, cmd, argp);
3134
3135 case SIOCSARP:
3136 case SIOCGARP:
3137 case SIOCDARP:
3138 case SIOCATMARK:
3139 return sock_do_ioctl(net, sock, cmd, arg);
3140 }
3141
3142 return -ENOIOCTLCMD;
3143 }
3144
3145 static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3146 unsigned long arg)
3147 {
3148 struct socket *sock = file->private_data;
3149 int ret = -ENOIOCTLCMD;
3150 struct sock *sk;
3151 struct net *net;
3152
3153 sk = sock->sk;
3154 net = sock_net(sk);
3155
3156 if (sock->ops->compat_ioctl)
3157 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3158
3159 if (ret == -ENOIOCTLCMD &&
3160 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3161 ret = compat_wext_handle_ioctl(net, cmd, arg);
3162
3163 if (ret == -ENOIOCTLCMD)
3164 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3165
3166 return ret;
3167 }
3168 #endif
3169
3170 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3171 {
3172 return sock->ops->bind(sock, addr, addrlen);
3173 }
3174 EXPORT_SYMBOL(kernel_bind);
3175
3176 int kernel_listen(struct socket *sock, int backlog)
3177 {
3178 return sock->ops->listen(sock, backlog);
3179 }
3180 EXPORT_SYMBOL(kernel_listen);
3181
3182 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3183 {
3184 struct sock *sk = sock->sk;
3185 int err;
3186
3187 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3188 newsock);
3189 if (err < 0)
3190 goto done;
3191
3192 err = sock->ops->accept(sock, *newsock, flags);
3193 if (err < 0) {
3194 sock_release(*newsock);
3195 *newsock = NULL;
3196 goto done;
3197 }
3198
3199 (*newsock)->ops = sock->ops;
3200 __module_get((*newsock)->ops->owner);
3201
3202 done:
3203 return err;
3204 }
3205 EXPORT_SYMBOL(kernel_accept);
3206
3207 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3208 int flags)
3209 {
3210 return sock->ops->connect(sock, addr, addrlen, flags);
3211 }
3212 EXPORT_SYMBOL(kernel_connect);
3213
3214 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3215 int *addrlen)
3216 {
3217 return sock->ops->getname(sock, addr, addrlen, 0);
3218 }
3219 EXPORT_SYMBOL(kernel_getsockname);
3220
3221 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3222 int *addrlen)
3223 {
3224 return sock->ops->getname(sock, addr, addrlen, 1);
3225 }
3226 EXPORT_SYMBOL(kernel_getpeername);
3227
3228 int kernel_getsockopt(struct socket *sock, int level, int optname,
3229 char *optval, int *optlen)
3230 {
3231 mm_segment_t oldfs = get_fs();
3232 char __user *uoptval;
3233 int __user *uoptlen;
3234 int err;
3235
3236 uoptval = (char __user __force *) optval;
3237 uoptlen = (int __user __force *) optlen;
3238
3239 set_fs(KERNEL_DS);
3240 if (level == SOL_SOCKET)
3241 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3242 else
3243 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3244 uoptlen);
3245 set_fs(oldfs);
3246 return err;
3247 }
3248 EXPORT_SYMBOL(kernel_getsockopt);
3249
3250 int kernel_setsockopt(struct socket *sock, int level, int optname,
3251 char *optval, unsigned int optlen)
3252 {
3253 mm_segment_t oldfs = get_fs();
3254 char __user *uoptval;
3255 int err;
3256
3257 uoptval = (char __user __force *) optval;
3258
3259 set_fs(KERNEL_DS);
3260 if (level == SOL_SOCKET)
3261 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3262 else
3263 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3264 optlen);
3265 set_fs(oldfs);
3266 return err;
3267 }
3268 EXPORT_SYMBOL(kernel_setsockopt);
3269
3270 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3271 size_t size, int flags)
3272 {
3273 if (sock->ops->sendpage)
3274 return sock->ops->sendpage(sock, page, offset, size, flags);
3275
3276 return sock_no_sendpage(sock, page, offset, size, flags);
3277 }
3278 EXPORT_SYMBOL(kernel_sendpage);
3279
3280 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3281 {
3282 mm_segment_t oldfs = get_fs();
3283 int err;
3284
3285 set_fs(KERNEL_DS);
3286 err = sock->ops->ioctl(sock, cmd, arg);
3287 set_fs(oldfs);
3288
3289 return err;
3290 }
3291 EXPORT_SYMBOL(kernel_sock_ioctl);
3292
3293 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3294 {
3295 return sock->ops->shutdown(sock, how);
3296 }
3297 EXPORT_SYMBOL(kernel_sock_shutdown);