Loading...
1/*
2 * NET An implementation of the SOCKET network access protocol.
3 *
4 * Version: @(#)socket.c 1.1.93 18/02/95
5 *
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
9 *
10 * Fixes:
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
12 * shutdown()
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
17 * top level.
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
22 * tty drivers).
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
25 * configurable.
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
34 * stuff.
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
40 * moment.
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
47 *
48 *
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
53 *
54 *
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
57 *
58 * Based upon Swansea University Computer Society NET3.039
59 */
60
61#include <linux/mm.h>
62#include <linux/socket.h>
63#include <linux/file.h>
64#include <linux/net.h>
65#include <linux/interrupt.h>
66#include <linux/thread_info.h>
67#include <linux/rcupdate.h>
68#include <linux/netdevice.h>
69#include <linux/proc_fs.h>
70#include <linux/seq_file.h>
71#include <linux/mutex.h>
72#include <linux/if_bridge.h>
73#include <linux/if_frad.h>
74#include <linux/if_vlan.h>
75#include <linux/ptp_classify.h>
76#include <linux/init.h>
77#include <linux/poll.h>
78#include <linux/cache.h>
79#include <linux/module.h>
80#include <linux/highmem.h>
81#include <linux/mount.h>
82#include <linux/security.h>
83#include <linux/syscalls.h>
84#include <linux/compat.h>
85#include <linux/kmod.h>
86#include <linux/audit.h>
87#include <linux/wireless.h>
88#include <linux/nsproxy.h>
89#include <linux/magic.h>
90#include <linux/slab.h>
91#include <linux/xattr.h>
92
93#include <linux/uaccess.h>
94#include <asm/unistd.h>
95
96#include <net/compat.h>
97#include <net/wext.h>
98#include <net/cls_cgroup.h>
99
100#include <net/sock.h>
101#include <linux/netfilter.h>
102
103#include <linux/if_tun.h>
104#include <linux/ipv6_route.h>
105#include <linux/route.h>
106#include <linux/sockios.h>
107#include <linux/atalk.h>
108#include <net/busy_poll.h>
109#include <linux/errqueue.h>
110
111#ifdef CONFIG_NET_RX_BUSY_POLL
112unsigned int sysctl_net_busy_read __read_mostly;
113unsigned int sysctl_net_busy_poll __read_mostly;
114#endif
115
116static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119
120static int sock_close(struct inode *inode, struct file *file);
121static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124#ifdef CONFIG_COMPAT
125static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
127#endif
128static int sock_fasync(int fd, struct file *filp, int on);
129static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131static 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
140static 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
162static DEFINE_SPINLOCK(net_family_lock);
163static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164
165/*
166 * Statistics counters of the socket lists
167 */
168
169static 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
188int 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
216static 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
243static struct kmem_cache *sock_inode_cachep __read_mostly;
244
245static 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
272static 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
283static 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
290static 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
304static 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 */
313static 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
319static const struct dentry_operations sockfs_dentry_operations = {
320 .d_dname = sockfs_dname,
321};
322
323static 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
339static const struct xattr_handler sockfs_xattr_handler = {
340 .name = XATTR_NAME_SOCKPROTONAME,
341 .get = sockfs_xattr_get,
342};
343
344static int sockfs_security_xattr_set(const struct xattr_handler *handler,
345 struct dentry *dentry, struct inode *inode,
346 const char *suffix, const void *value,
347 size_t size, int flags)
348{
349 /* Handled by LSM. */
350 return -EAGAIN;
351}
352
353static const struct xattr_handler sockfs_security_xattr_handler = {
354 .prefix = XATTR_SECURITY_PREFIX,
355 .set = sockfs_security_xattr_set,
356};
357
358static const struct xattr_handler *sockfs_xattr_handlers[] = {
359 &sockfs_xattr_handler,
360 &sockfs_security_xattr_handler,
361 NULL
362};
363
364static struct dentry *sockfs_mount(struct file_system_type *fs_type,
365 int flags, const char *dev_name, void *data)
366{
367 return mount_pseudo_xattr(fs_type, "socket:", &sockfs_ops,
368 sockfs_xattr_handlers,
369 &sockfs_dentry_operations, SOCKFS_MAGIC);
370}
371
372static struct vfsmount *sock_mnt __read_mostly;
373
374static struct file_system_type sock_fs_type = {
375 .name = "sockfs",
376 .mount = sockfs_mount,
377 .kill_sb = kill_anon_super,
378};
379
380/*
381 * Obtains the first available file descriptor and sets it up for use.
382 *
383 * These functions create file structures and maps them to fd space
384 * of the current process. On success it returns file descriptor
385 * and file struct implicitly stored in sock->file.
386 * Note that another thread may close file descriptor before we return
387 * from this function. We use the fact that now we do not refer
388 * to socket after mapping. If one day we will need it, this
389 * function will increment ref. count on file by 1.
390 *
391 * In any case returned fd MAY BE not valid!
392 * This race condition is unavoidable
393 * with shared fd spaces, we cannot solve it inside kernel,
394 * but we take care of internal coherence yet.
395 */
396
397struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
398{
399 struct qstr name = { .name = "" };
400 struct path path;
401 struct file *file;
402
403 if (dname) {
404 name.name = dname;
405 name.len = strlen(name.name);
406 } else if (sock->sk) {
407 name.name = sock->sk->sk_prot_creator->name;
408 name.len = strlen(name.name);
409 }
410 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
411 if (unlikely(!path.dentry))
412 return ERR_PTR(-ENOMEM);
413 path.mnt = mntget(sock_mnt);
414
415 d_instantiate(path.dentry, SOCK_INODE(sock));
416
417 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
418 &socket_file_ops);
419 if (IS_ERR(file)) {
420 /* drop dentry, keep inode */
421 ihold(d_inode(path.dentry));
422 path_put(&path);
423 return file;
424 }
425
426 sock->file = file;
427 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
428 file->private_data = sock;
429 return file;
430}
431EXPORT_SYMBOL(sock_alloc_file);
432
433static int sock_map_fd(struct socket *sock, int flags)
434{
435 struct file *newfile;
436 int fd = get_unused_fd_flags(flags);
437 if (unlikely(fd < 0))
438 return fd;
439
440 newfile = sock_alloc_file(sock, flags, NULL);
441 if (likely(!IS_ERR(newfile))) {
442 fd_install(fd, newfile);
443 return fd;
444 }
445
446 put_unused_fd(fd);
447 return PTR_ERR(newfile);
448}
449
450struct socket *sock_from_file(struct file *file, int *err)
451{
452 if (file->f_op == &socket_file_ops)
453 return file->private_data; /* set in sock_map_fd */
454
455 *err = -ENOTSOCK;
456 return NULL;
457}
458EXPORT_SYMBOL(sock_from_file);
459
460/**
461 * sockfd_lookup - Go from a file number to its socket slot
462 * @fd: file handle
463 * @err: pointer to an error code return
464 *
465 * The file handle passed in is locked and the socket it is bound
466 * too is returned. If an error occurs the err pointer is overwritten
467 * with a negative errno code and NULL is returned. The function checks
468 * for both invalid handles and passing a handle which is not a socket.
469 *
470 * On a success the socket object pointer is returned.
471 */
472
473struct socket *sockfd_lookup(int fd, int *err)
474{
475 struct file *file;
476 struct socket *sock;
477
478 file = fget(fd);
479 if (!file) {
480 *err = -EBADF;
481 return NULL;
482 }
483
484 sock = sock_from_file(file, err);
485 if (!sock)
486 fput(file);
487 return sock;
488}
489EXPORT_SYMBOL(sockfd_lookup);
490
491static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
492{
493 struct fd f = fdget(fd);
494 struct socket *sock;
495
496 *err = -EBADF;
497 if (f.file) {
498 sock = sock_from_file(f.file, err);
499 if (likely(sock)) {
500 *fput_needed = f.flags;
501 return sock;
502 }
503 fdput(f);
504 }
505 return NULL;
506}
507
508static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
509 size_t size)
510{
511 ssize_t len;
512 ssize_t used = 0;
513
514 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
515 if (len < 0)
516 return len;
517 used += len;
518 if (buffer) {
519 if (size < used)
520 return -ERANGE;
521 buffer += len;
522 }
523
524 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
525 used += len;
526 if (buffer) {
527 if (size < used)
528 return -ERANGE;
529 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
530 buffer += len;
531 }
532
533 return used;
534}
535
536static int sockfs_setattr(struct dentry *dentry, struct iattr *iattr)
537{
538 int err = simple_setattr(dentry, iattr);
539
540 if (!err && (iattr->ia_valid & ATTR_UID)) {
541 struct socket *sock = SOCKET_I(d_inode(dentry));
542
543 sock->sk->sk_uid = iattr->ia_uid;
544 }
545
546 return err;
547}
548
549static const struct inode_operations sockfs_inode_ops = {
550 .listxattr = sockfs_listxattr,
551 .setattr = sockfs_setattr,
552};
553
554/**
555 * sock_alloc - allocate a socket
556 *
557 * Allocate a new inode and socket object. The two are bound together
558 * and initialised. The socket is then returned. If we are out of inodes
559 * NULL is returned.
560 */
561
562struct socket *sock_alloc(void)
563{
564 struct inode *inode;
565 struct socket *sock;
566
567 inode = new_inode_pseudo(sock_mnt->mnt_sb);
568 if (!inode)
569 return NULL;
570
571 sock = SOCKET_I(inode);
572
573 kmemcheck_annotate_bitfield(sock, type);
574 inode->i_ino = get_next_ino();
575 inode->i_mode = S_IFSOCK | S_IRWXUGO;
576 inode->i_uid = current_fsuid();
577 inode->i_gid = current_fsgid();
578 inode->i_op = &sockfs_inode_ops;
579
580 this_cpu_add(sockets_in_use, 1);
581 return sock;
582}
583EXPORT_SYMBOL(sock_alloc);
584
585/**
586 * sock_release - close a socket
587 * @sock: socket to close
588 *
589 * The socket is released from the protocol stack if it has a release
590 * callback, and the inode is then released if the socket is bound to
591 * an inode not a file.
592 */
593
594void sock_release(struct socket *sock)
595{
596 if (sock->ops) {
597 struct module *owner = sock->ops->owner;
598
599 sock->ops->release(sock);
600 sock->ops = NULL;
601 module_put(owner);
602 }
603
604 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
605 pr_err("%s: fasync list not empty!\n", __func__);
606
607 this_cpu_sub(sockets_in_use, 1);
608 if (!sock->file) {
609 iput(SOCK_INODE(sock));
610 return;
611 }
612 sock->file = NULL;
613}
614EXPORT_SYMBOL(sock_release);
615
616void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags)
617{
618 u8 flags = *tx_flags;
619
620 if (tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
621 flags |= SKBTX_HW_TSTAMP;
622
623 if (tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
624 flags |= SKBTX_SW_TSTAMP;
625
626 if (tsflags & SOF_TIMESTAMPING_TX_SCHED)
627 flags |= SKBTX_SCHED_TSTAMP;
628
629 *tx_flags = flags;
630}
631EXPORT_SYMBOL(__sock_tx_timestamp);
632
633static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
634{
635 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
636 BUG_ON(ret == -EIOCBQUEUED);
637 return ret;
638}
639
640int sock_sendmsg(struct socket *sock, struct msghdr *msg)
641{
642 int err = security_socket_sendmsg(sock, msg,
643 msg_data_left(msg));
644
645 return err ?: sock_sendmsg_nosec(sock, msg);
646}
647EXPORT_SYMBOL(sock_sendmsg);
648
649int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
650 struct kvec *vec, size_t num, size_t size)
651{
652 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
653 return sock_sendmsg(sock, msg);
654}
655EXPORT_SYMBOL(kernel_sendmsg);
656
657/*
658 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
659 */
660void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
661 struct sk_buff *skb)
662{
663 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
664 struct scm_timestamping tss;
665 int empty = 1;
666 struct skb_shared_hwtstamps *shhwtstamps =
667 skb_hwtstamps(skb);
668
669 /* Race occurred between timestamp enabling and packet
670 receiving. Fill in the current time for now. */
671 if (need_software_tstamp && skb->tstamp == 0)
672 __net_timestamp(skb);
673
674 if (need_software_tstamp) {
675 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
676 struct timeval tv;
677 skb_get_timestamp(skb, &tv);
678 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
679 sizeof(tv), &tv);
680 } else {
681 struct timespec ts;
682 skb_get_timestampns(skb, &ts);
683 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
684 sizeof(ts), &ts);
685 }
686 }
687
688 memset(&tss, 0, sizeof(tss));
689 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
690 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
691 empty = 0;
692 if (shhwtstamps &&
693 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
694 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
695 empty = 0;
696 if (!empty) {
697 put_cmsg(msg, SOL_SOCKET,
698 SCM_TIMESTAMPING, sizeof(tss), &tss);
699
700 if (skb->len && (sk->sk_tsflags & SOF_TIMESTAMPING_OPT_STATS))
701 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPING_OPT_STATS,
702 skb->len, skb->data);
703 }
704}
705EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
706
707void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
708 struct sk_buff *skb)
709{
710 int ack;
711
712 if (!sock_flag(sk, SOCK_WIFI_STATUS))
713 return;
714 if (!skb->wifi_acked_valid)
715 return;
716
717 ack = skb->wifi_acked;
718
719 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
720}
721EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
722
723static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
724 struct sk_buff *skb)
725{
726 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
727 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
728 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
729}
730
731void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
732 struct sk_buff *skb)
733{
734 sock_recv_timestamp(msg, sk, skb);
735 sock_recv_drops(msg, sk, skb);
736}
737EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
738
739static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
740 int flags)
741{
742 return sock->ops->recvmsg(sock, msg, msg_data_left(msg), flags);
743}
744
745int sock_recvmsg(struct socket *sock, struct msghdr *msg, int flags)
746{
747 int err = security_socket_recvmsg(sock, msg, msg_data_left(msg), flags);
748
749 return err ?: sock_recvmsg_nosec(sock, msg, flags);
750}
751EXPORT_SYMBOL(sock_recvmsg);
752
753/**
754 * kernel_recvmsg - Receive a message from a socket (kernel space)
755 * @sock: The socket to receive the message from
756 * @msg: Received message
757 * @vec: Input s/g array for message data
758 * @num: Size of input s/g array
759 * @size: Number of bytes to read
760 * @flags: Message flags (MSG_DONTWAIT, etc...)
761 *
762 * On return the msg structure contains the scatter/gather array passed in the
763 * vec argument. The array is modified so that it consists of the unfilled
764 * portion of the original array.
765 *
766 * The returned value is the total number of bytes received, or an error.
767 */
768int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
769 struct kvec *vec, size_t num, size_t size, int flags)
770{
771 mm_segment_t oldfs = get_fs();
772 int result;
773
774 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
775 set_fs(KERNEL_DS);
776 result = sock_recvmsg(sock, msg, flags);
777 set_fs(oldfs);
778 return result;
779}
780EXPORT_SYMBOL(kernel_recvmsg);
781
782static ssize_t sock_sendpage(struct file *file, struct page *page,
783 int offset, size_t size, loff_t *ppos, int more)
784{
785 struct socket *sock;
786 int flags;
787
788 sock = file->private_data;
789
790 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
791 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
792 flags |= more;
793
794 return kernel_sendpage(sock, page, offset, size, flags);
795}
796
797static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
798 struct pipe_inode_info *pipe, size_t len,
799 unsigned int flags)
800{
801 struct socket *sock = file->private_data;
802
803 if (unlikely(!sock->ops->splice_read))
804 return -EINVAL;
805
806 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
807}
808
809static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
810{
811 struct file *file = iocb->ki_filp;
812 struct socket *sock = file->private_data;
813 struct msghdr msg = {.msg_iter = *to,
814 .msg_iocb = iocb};
815 ssize_t res;
816
817 if (file->f_flags & O_NONBLOCK)
818 msg.msg_flags = MSG_DONTWAIT;
819
820 if (iocb->ki_pos != 0)
821 return -ESPIPE;
822
823 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
824 return 0;
825
826 res = sock_recvmsg(sock, &msg, msg.msg_flags);
827 *to = msg.msg_iter;
828 return res;
829}
830
831static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
832{
833 struct file *file = iocb->ki_filp;
834 struct socket *sock = file->private_data;
835 struct msghdr msg = {.msg_iter = *from,
836 .msg_iocb = iocb};
837 ssize_t res;
838
839 if (iocb->ki_pos != 0)
840 return -ESPIPE;
841
842 if (file->f_flags & O_NONBLOCK)
843 msg.msg_flags = MSG_DONTWAIT;
844
845 if (sock->type == SOCK_SEQPACKET)
846 msg.msg_flags |= MSG_EOR;
847
848 res = sock_sendmsg(sock, &msg);
849 *from = msg.msg_iter;
850 return res;
851}
852
853/*
854 * Atomic setting of ioctl hooks to avoid race
855 * with module unload.
856 */
857
858static DEFINE_MUTEX(br_ioctl_mutex);
859static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
860
861void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
862{
863 mutex_lock(&br_ioctl_mutex);
864 br_ioctl_hook = hook;
865 mutex_unlock(&br_ioctl_mutex);
866}
867EXPORT_SYMBOL(brioctl_set);
868
869static DEFINE_MUTEX(vlan_ioctl_mutex);
870static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
871
872void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
873{
874 mutex_lock(&vlan_ioctl_mutex);
875 vlan_ioctl_hook = hook;
876 mutex_unlock(&vlan_ioctl_mutex);
877}
878EXPORT_SYMBOL(vlan_ioctl_set);
879
880static DEFINE_MUTEX(dlci_ioctl_mutex);
881static int (*dlci_ioctl_hook) (unsigned int, void __user *);
882
883void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
884{
885 mutex_lock(&dlci_ioctl_mutex);
886 dlci_ioctl_hook = hook;
887 mutex_unlock(&dlci_ioctl_mutex);
888}
889EXPORT_SYMBOL(dlci_ioctl_set);
890
891static long sock_do_ioctl(struct net *net, struct socket *sock,
892 unsigned int cmd, unsigned long arg)
893{
894 int err;
895 void __user *argp = (void __user *)arg;
896
897 err = sock->ops->ioctl(sock, cmd, arg);
898
899 /*
900 * If this ioctl is unknown try to hand it down
901 * to the NIC driver.
902 */
903 if (err == -ENOIOCTLCMD)
904 err = dev_ioctl(net, cmd, argp);
905
906 return err;
907}
908
909/*
910 * With an ioctl, arg may well be a user mode pointer, but we don't know
911 * what to do with it - that's up to the protocol still.
912 */
913
914static struct ns_common *get_net_ns(struct ns_common *ns)
915{
916 return &get_net(container_of(ns, struct net, ns))->ns;
917}
918
919static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
920{
921 struct socket *sock;
922 struct sock *sk;
923 void __user *argp = (void __user *)arg;
924 int pid, err;
925 struct net *net;
926
927 sock = file->private_data;
928 sk = sock->sk;
929 net = sock_net(sk);
930 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
931 err = dev_ioctl(net, cmd, argp);
932 } else
933#ifdef CONFIG_WEXT_CORE
934 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
935 err = dev_ioctl(net, cmd, argp);
936 } else
937#endif
938 switch (cmd) {
939 case FIOSETOWN:
940 case SIOCSPGRP:
941 err = -EFAULT;
942 if (get_user(pid, (int __user *)argp))
943 break;
944 f_setown(sock->file, pid, 1);
945 err = 0;
946 break;
947 case FIOGETOWN:
948 case SIOCGPGRP:
949 err = put_user(f_getown(sock->file),
950 (int __user *)argp);
951 break;
952 case SIOCGIFBR:
953 case SIOCSIFBR:
954 case SIOCBRADDBR:
955 case SIOCBRDELBR:
956 err = -ENOPKG;
957 if (!br_ioctl_hook)
958 request_module("bridge");
959
960 mutex_lock(&br_ioctl_mutex);
961 if (br_ioctl_hook)
962 err = br_ioctl_hook(net, cmd, argp);
963 mutex_unlock(&br_ioctl_mutex);
964 break;
965 case SIOCGIFVLAN:
966 case SIOCSIFVLAN:
967 err = -ENOPKG;
968 if (!vlan_ioctl_hook)
969 request_module("8021q");
970
971 mutex_lock(&vlan_ioctl_mutex);
972 if (vlan_ioctl_hook)
973 err = vlan_ioctl_hook(net, argp);
974 mutex_unlock(&vlan_ioctl_mutex);
975 break;
976 case SIOCADDDLCI:
977 case SIOCDELDLCI:
978 err = -ENOPKG;
979 if (!dlci_ioctl_hook)
980 request_module("dlci");
981
982 mutex_lock(&dlci_ioctl_mutex);
983 if (dlci_ioctl_hook)
984 err = dlci_ioctl_hook(cmd, argp);
985 mutex_unlock(&dlci_ioctl_mutex);
986 break;
987 case SIOCGSKNS:
988 err = -EPERM;
989 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
990 break;
991
992 err = open_related_ns(&net->ns, get_net_ns);
993 break;
994 default:
995 err = sock_do_ioctl(net, sock, cmd, arg);
996 break;
997 }
998 return err;
999}
1000
1001int sock_create_lite(int family, int type, int protocol, struct socket **res)
1002{
1003 int err;
1004 struct socket *sock = NULL;
1005
1006 err = security_socket_create(family, type, protocol, 1);
1007 if (err)
1008 goto out;
1009
1010 sock = sock_alloc();
1011 if (!sock) {
1012 err = -ENOMEM;
1013 goto out;
1014 }
1015
1016 sock->type = type;
1017 err = security_socket_post_create(sock, family, type, protocol, 1);
1018 if (err)
1019 goto out_release;
1020
1021out:
1022 *res = sock;
1023 return err;
1024out_release:
1025 sock_release(sock);
1026 sock = NULL;
1027 goto out;
1028}
1029EXPORT_SYMBOL(sock_create_lite);
1030
1031/* No kernel lock held - perfect */
1032static unsigned int sock_poll(struct file *file, poll_table *wait)
1033{
1034 unsigned int busy_flag = 0;
1035 struct socket *sock;
1036
1037 /*
1038 * We can't return errors to poll, so it's either yes or no.
1039 */
1040 sock = file->private_data;
1041
1042 if (sk_can_busy_loop(sock->sk)) {
1043 /* this socket can poll_ll so tell the system call */
1044 busy_flag = POLL_BUSY_LOOP;
1045
1046 /* once, only if requested by syscall */
1047 if (wait && (wait->_key & POLL_BUSY_LOOP))
1048 sk_busy_loop(sock->sk, 1);
1049 }
1050
1051 return busy_flag | sock->ops->poll(file, sock, wait);
1052}
1053
1054static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1055{
1056 struct socket *sock = file->private_data;
1057
1058 return sock->ops->mmap(file, sock, vma);
1059}
1060
1061static int sock_close(struct inode *inode, struct file *filp)
1062{
1063 sock_release(SOCKET_I(inode));
1064 return 0;
1065}
1066
1067/*
1068 * Update the socket async list
1069 *
1070 * Fasync_list locking strategy.
1071 *
1072 * 1. fasync_list is modified only under process context socket lock
1073 * i.e. under semaphore.
1074 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1075 * or under socket lock
1076 */
1077
1078static int sock_fasync(int fd, struct file *filp, int on)
1079{
1080 struct socket *sock = filp->private_data;
1081 struct sock *sk = sock->sk;
1082 struct socket_wq *wq;
1083
1084 if (sk == NULL)
1085 return -EINVAL;
1086
1087 lock_sock(sk);
1088 wq = rcu_dereference_protected(sock->wq, lockdep_sock_is_held(sk));
1089 fasync_helper(fd, filp, on, &wq->fasync_list);
1090
1091 if (!wq->fasync_list)
1092 sock_reset_flag(sk, SOCK_FASYNC);
1093 else
1094 sock_set_flag(sk, SOCK_FASYNC);
1095
1096 release_sock(sk);
1097 return 0;
1098}
1099
1100/* This function may be called only under rcu_lock */
1101
1102int sock_wake_async(struct socket_wq *wq, int how, int band)
1103{
1104 if (!wq || !wq->fasync_list)
1105 return -1;
1106
1107 switch (how) {
1108 case SOCK_WAKE_WAITD:
1109 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1110 break;
1111 goto call_kill;
1112 case SOCK_WAKE_SPACE:
1113 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1114 break;
1115 /* fall through */
1116 case SOCK_WAKE_IO:
1117call_kill:
1118 kill_fasync(&wq->fasync_list, SIGIO, band);
1119 break;
1120 case SOCK_WAKE_URG:
1121 kill_fasync(&wq->fasync_list, SIGURG, band);
1122 }
1123
1124 return 0;
1125}
1126EXPORT_SYMBOL(sock_wake_async);
1127
1128int __sock_create(struct net *net, int family, int type, int protocol,
1129 struct socket **res, int kern)
1130{
1131 int err;
1132 struct socket *sock;
1133 const struct net_proto_family *pf;
1134
1135 /*
1136 * Check protocol is in range
1137 */
1138 if (family < 0 || family >= NPROTO)
1139 return -EAFNOSUPPORT;
1140 if (type < 0 || type >= SOCK_MAX)
1141 return -EINVAL;
1142
1143 /* Compatibility.
1144
1145 This uglymoron is moved from INET layer to here to avoid
1146 deadlock in module load.
1147 */
1148 if (family == PF_INET && type == SOCK_PACKET) {
1149 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1150 current->comm);
1151 family = PF_PACKET;
1152 }
1153
1154 err = security_socket_create(family, type, protocol, kern);
1155 if (err)
1156 return err;
1157
1158 /*
1159 * Allocate the socket and allow the family to set things up. if
1160 * the protocol is 0, the family is instructed to select an appropriate
1161 * default.
1162 */
1163 sock = sock_alloc();
1164 if (!sock) {
1165 net_warn_ratelimited("socket: no more sockets\n");
1166 return -ENFILE; /* Not exactly a match, but its the
1167 closest posix thing */
1168 }
1169
1170 sock->type = type;
1171
1172#ifdef CONFIG_MODULES
1173 /* Attempt to load a protocol module if the find failed.
1174 *
1175 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1176 * requested real, full-featured networking support upon configuration.
1177 * Otherwise module support will break!
1178 */
1179 if (rcu_access_pointer(net_families[family]) == NULL)
1180 request_module("net-pf-%d", family);
1181#endif
1182
1183 rcu_read_lock();
1184 pf = rcu_dereference(net_families[family]);
1185 err = -EAFNOSUPPORT;
1186 if (!pf)
1187 goto out_release;
1188
1189 /*
1190 * We will call the ->create function, that possibly is in a loadable
1191 * module, so we have to bump that loadable module refcnt first.
1192 */
1193 if (!try_module_get(pf->owner))
1194 goto out_release;
1195
1196 /* Now protected by module ref count */
1197 rcu_read_unlock();
1198
1199 err = pf->create(net, sock, protocol, kern);
1200 if (err < 0)
1201 goto out_module_put;
1202
1203 /*
1204 * Now to bump the refcnt of the [loadable] module that owns this
1205 * socket at sock_release time we decrement its refcnt.
1206 */
1207 if (!try_module_get(sock->ops->owner))
1208 goto out_module_busy;
1209
1210 /*
1211 * Now that we're done with the ->create function, the [loadable]
1212 * module can have its refcnt decremented
1213 */
1214 module_put(pf->owner);
1215 err = security_socket_post_create(sock, family, type, protocol, kern);
1216 if (err)
1217 goto out_sock_release;
1218 *res = sock;
1219
1220 return 0;
1221
1222out_module_busy:
1223 err = -EAFNOSUPPORT;
1224out_module_put:
1225 sock->ops = NULL;
1226 module_put(pf->owner);
1227out_sock_release:
1228 sock_release(sock);
1229 return err;
1230
1231out_release:
1232 rcu_read_unlock();
1233 goto out_sock_release;
1234}
1235EXPORT_SYMBOL(__sock_create);
1236
1237int sock_create(int family, int type, int protocol, struct socket **res)
1238{
1239 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1240}
1241EXPORT_SYMBOL(sock_create);
1242
1243int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1244{
1245 return __sock_create(net, family, type, protocol, res, 1);
1246}
1247EXPORT_SYMBOL(sock_create_kern);
1248
1249SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1250{
1251 int retval;
1252 struct socket *sock;
1253 int flags;
1254
1255 /* Check the SOCK_* constants for consistency. */
1256 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1257 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1258 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1259 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1260
1261 flags = type & ~SOCK_TYPE_MASK;
1262 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1263 return -EINVAL;
1264 type &= SOCK_TYPE_MASK;
1265
1266 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1267 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1268
1269 retval = sock_create(family, type, protocol, &sock);
1270 if (retval < 0)
1271 goto out;
1272
1273 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1274 if (retval < 0)
1275 goto out_release;
1276
1277out:
1278 /* It may be already another descriptor 8) Not kernel problem. */
1279 return retval;
1280
1281out_release:
1282 sock_release(sock);
1283 return retval;
1284}
1285
1286/*
1287 * Create a pair of connected sockets.
1288 */
1289
1290SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1291 int __user *, usockvec)
1292{
1293 struct socket *sock1, *sock2;
1294 int fd1, fd2, err;
1295 struct file *newfile1, *newfile2;
1296 int flags;
1297
1298 flags = type & ~SOCK_TYPE_MASK;
1299 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1300 return -EINVAL;
1301 type &= SOCK_TYPE_MASK;
1302
1303 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1304 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1305
1306 /*
1307 * Obtain the first socket and check if the underlying protocol
1308 * supports the socketpair call.
1309 */
1310
1311 err = sock_create(family, type, protocol, &sock1);
1312 if (err < 0)
1313 goto out;
1314
1315 err = sock_create(family, type, protocol, &sock2);
1316 if (err < 0)
1317 goto out_release_1;
1318
1319 err = sock1->ops->socketpair(sock1, sock2);
1320 if (err < 0)
1321 goto out_release_both;
1322
1323 fd1 = get_unused_fd_flags(flags);
1324 if (unlikely(fd1 < 0)) {
1325 err = fd1;
1326 goto out_release_both;
1327 }
1328
1329 fd2 = get_unused_fd_flags(flags);
1330 if (unlikely(fd2 < 0)) {
1331 err = fd2;
1332 goto out_put_unused_1;
1333 }
1334
1335 newfile1 = sock_alloc_file(sock1, flags, NULL);
1336 if (IS_ERR(newfile1)) {
1337 err = PTR_ERR(newfile1);
1338 goto out_put_unused_both;
1339 }
1340
1341 newfile2 = sock_alloc_file(sock2, flags, NULL);
1342 if (IS_ERR(newfile2)) {
1343 err = PTR_ERR(newfile2);
1344 goto out_fput_1;
1345 }
1346
1347 err = put_user(fd1, &usockvec[0]);
1348 if (err)
1349 goto out_fput_both;
1350
1351 err = put_user(fd2, &usockvec[1]);
1352 if (err)
1353 goto out_fput_both;
1354
1355 audit_fd_pair(fd1, fd2);
1356
1357 fd_install(fd1, newfile1);
1358 fd_install(fd2, newfile2);
1359 /* fd1 and fd2 may be already another descriptors.
1360 * Not kernel problem.
1361 */
1362
1363 return 0;
1364
1365out_fput_both:
1366 fput(newfile2);
1367 fput(newfile1);
1368 put_unused_fd(fd2);
1369 put_unused_fd(fd1);
1370 goto out;
1371
1372out_fput_1:
1373 fput(newfile1);
1374 put_unused_fd(fd2);
1375 put_unused_fd(fd1);
1376 sock_release(sock2);
1377 goto out;
1378
1379out_put_unused_both:
1380 put_unused_fd(fd2);
1381out_put_unused_1:
1382 put_unused_fd(fd1);
1383out_release_both:
1384 sock_release(sock2);
1385out_release_1:
1386 sock_release(sock1);
1387out:
1388 return err;
1389}
1390
1391/*
1392 * Bind a name to a socket. Nothing much to do here since it's
1393 * the protocol's responsibility to handle the local address.
1394 *
1395 * We move the socket address to kernel space before we call
1396 * the protocol layer (having also checked the address is ok).
1397 */
1398
1399SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1400{
1401 struct socket *sock;
1402 struct sockaddr_storage address;
1403 int err, fput_needed;
1404
1405 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1406 if (sock) {
1407 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1408 if (err >= 0) {
1409 err = security_socket_bind(sock,
1410 (struct sockaddr *)&address,
1411 addrlen);
1412 if (!err)
1413 err = sock->ops->bind(sock,
1414 (struct sockaddr *)
1415 &address, addrlen);
1416 }
1417 fput_light(sock->file, fput_needed);
1418 }
1419 return err;
1420}
1421
1422/*
1423 * Perform a listen. Basically, we allow the protocol to do anything
1424 * necessary for a listen, and if that works, we mark the socket as
1425 * ready for listening.
1426 */
1427
1428SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1429{
1430 struct socket *sock;
1431 int err, fput_needed;
1432 int somaxconn;
1433
1434 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1435 if (sock) {
1436 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1437 if ((unsigned int)backlog > somaxconn)
1438 backlog = somaxconn;
1439
1440 err = security_socket_listen(sock, backlog);
1441 if (!err)
1442 err = sock->ops->listen(sock, backlog);
1443
1444 fput_light(sock->file, fput_needed);
1445 }
1446 return err;
1447}
1448
1449/*
1450 * For accept, we attempt to create a new socket, set up the link
1451 * with the client, wake up the client, then return the new
1452 * connected fd. We collect the address of the connector in kernel
1453 * space and move it to user at the very end. This is unclean because
1454 * we open the socket then return an error.
1455 *
1456 * 1003.1g adds the ability to recvmsg() to query connection pending
1457 * status to recvmsg. We need to add that support in a way thats
1458 * clean when we restucture accept also.
1459 */
1460
1461SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1462 int __user *, upeer_addrlen, int, flags)
1463{
1464 struct socket *sock, *newsock;
1465 struct file *newfile;
1466 int err, len, newfd, fput_needed;
1467 struct sockaddr_storage address;
1468
1469 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1470 return -EINVAL;
1471
1472 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1473 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1474
1475 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1476 if (!sock)
1477 goto out;
1478
1479 err = -ENFILE;
1480 newsock = sock_alloc();
1481 if (!newsock)
1482 goto out_put;
1483
1484 newsock->type = sock->type;
1485 newsock->ops = sock->ops;
1486
1487 /*
1488 * We don't need try_module_get here, as the listening socket (sock)
1489 * has the protocol module (sock->ops->owner) held.
1490 */
1491 __module_get(newsock->ops->owner);
1492
1493 newfd = get_unused_fd_flags(flags);
1494 if (unlikely(newfd < 0)) {
1495 err = newfd;
1496 sock_release(newsock);
1497 goto out_put;
1498 }
1499 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1500 if (IS_ERR(newfile)) {
1501 err = PTR_ERR(newfile);
1502 put_unused_fd(newfd);
1503 sock_release(newsock);
1504 goto out_put;
1505 }
1506
1507 err = security_socket_accept(sock, newsock);
1508 if (err)
1509 goto out_fd;
1510
1511 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1512 if (err < 0)
1513 goto out_fd;
1514
1515 if (upeer_sockaddr) {
1516 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1517 &len, 2) < 0) {
1518 err = -ECONNABORTED;
1519 goto out_fd;
1520 }
1521 err = move_addr_to_user(&address,
1522 len, upeer_sockaddr, upeer_addrlen);
1523 if (err < 0)
1524 goto out_fd;
1525 }
1526
1527 /* File flags are not inherited via accept() unlike another OSes. */
1528
1529 fd_install(newfd, newfile);
1530 err = newfd;
1531
1532out_put:
1533 fput_light(sock->file, fput_needed);
1534out:
1535 return err;
1536out_fd:
1537 fput(newfile);
1538 put_unused_fd(newfd);
1539 goto out_put;
1540}
1541
1542SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1543 int __user *, upeer_addrlen)
1544{
1545 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1546}
1547
1548/*
1549 * Attempt to connect to a socket with the server address. The address
1550 * is in user space so we verify it is OK and move it to kernel space.
1551 *
1552 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1553 * break bindings
1554 *
1555 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1556 * other SEQPACKET protocols that take time to connect() as it doesn't
1557 * include the -EINPROGRESS status for such sockets.
1558 */
1559
1560SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1561 int, addrlen)
1562{
1563 struct socket *sock;
1564 struct sockaddr_storage address;
1565 int err, fput_needed;
1566
1567 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1568 if (!sock)
1569 goto out;
1570 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1571 if (err < 0)
1572 goto out_put;
1573
1574 err =
1575 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1576 if (err)
1577 goto out_put;
1578
1579 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1580 sock->file->f_flags);
1581out_put:
1582 fput_light(sock->file, fput_needed);
1583out:
1584 return err;
1585}
1586
1587/*
1588 * Get the local address ('name') of a socket object. Move the obtained
1589 * name to user space.
1590 */
1591
1592SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1593 int __user *, usockaddr_len)
1594{
1595 struct socket *sock;
1596 struct sockaddr_storage address;
1597 int len, err, fput_needed;
1598
1599 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1600 if (!sock)
1601 goto out;
1602
1603 err = security_socket_getsockname(sock);
1604 if (err)
1605 goto out_put;
1606
1607 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1608 if (err)
1609 goto out_put;
1610 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1611
1612out_put:
1613 fput_light(sock->file, fput_needed);
1614out:
1615 return err;
1616}
1617
1618/*
1619 * Get the remote address ('name') of a socket object. Move the obtained
1620 * name to user space.
1621 */
1622
1623SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1624 int __user *, usockaddr_len)
1625{
1626 struct socket *sock;
1627 struct sockaddr_storage address;
1628 int len, err, fput_needed;
1629
1630 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1631 if (sock != NULL) {
1632 err = security_socket_getpeername(sock);
1633 if (err) {
1634 fput_light(sock->file, fput_needed);
1635 return err;
1636 }
1637
1638 err =
1639 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1640 1);
1641 if (!err)
1642 err = move_addr_to_user(&address, len, usockaddr,
1643 usockaddr_len);
1644 fput_light(sock->file, fput_needed);
1645 }
1646 return err;
1647}
1648
1649/*
1650 * Send a datagram to a given address. We move the address into kernel
1651 * space and check the user space data area is readable before invoking
1652 * the protocol.
1653 */
1654
1655SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1656 unsigned int, flags, struct sockaddr __user *, addr,
1657 int, addr_len)
1658{
1659 struct socket *sock;
1660 struct sockaddr_storage address;
1661 int err;
1662 struct msghdr msg;
1663 struct iovec iov;
1664 int fput_needed;
1665
1666 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1667 if (unlikely(err))
1668 return err;
1669 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1670 if (!sock)
1671 goto out;
1672
1673 msg.msg_name = NULL;
1674 msg.msg_control = NULL;
1675 msg.msg_controllen = 0;
1676 msg.msg_namelen = 0;
1677 if (addr) {
1678 err = move_addr_to_kernel(addr, addr_len, &address);
1679 if (err < 0)
1680 goto out_put;
1681 msg.msg_name = (struct sockaddr *)&address;
1682 msg.msg_namelen = addr_len;
1683 }
1684 if (sock->file->f_flags & O_NONBLOCK)
1685 flags |= MSG_DONTWAIT;
1686 msg.msg_flags = flags;
1687 err = sock_sendmsg(sock, &msg);
1688
1689out_put:
1690 fput_light(sock->file, fput_needed);
1691out:
1692 return err;
1693}
1694
1695/*
1696 * Send a datagram down a socket.
1697 */
1698
1699SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1700 unsigned int, flags)
1701{
1702 return sys_sendto(fd, buff, len, flags, NULL, 0);
1703}
1704
1705/*
1706 * Receive a frame from the socket and optionally record the address of the
1707 * sender. We verify the buffers are writable and if needed move the
1708 * sender address from kernel to user space.
1709 */
1710
1711SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1712 unsigned int, flags, struct sockaddr __user *, addr,
1713 int __user *, addr_len)
1714{
1715 struct socket *sock;
1716 struct iovec iov;
1717 struct msghdr msg;
1718 struct sockaddr_storage address;
1719 int err, err2;
1720 int fput_needed;
1721
1722 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1723 if (unlikely(err))
1724 return err;
1725 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1726 if (!sock)
1727 goto out;
1728
1729 msg.msg_control = NULL;
1730 msg.msg_controllen = 0;
1731 /* Save some cycles and don't copy the address if not needed */
1732 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1733 /* We assume all kernel code knows the size of sockaddr_storage */
1734 msg.msg_namelen = 0;
1735 msg.msg_iocb = NULL;
1736 if (sock->file->f_flags & O_NONBLOCK)
1737 flags |= MSG_DONTWAIT;
1738 err = sock_recvmsg(sock, &msg, flags);
1739
1740 if (err >= 0 && addr != NULL) {
1741 err2 = move_addr_to_user(&address,
1742 msg.msg_namelen, addr, addr_len);
1743 if (err2 < 0)
1744 err = err2;
1745 }
1746
1747 fput_light(sock->file, fput_needed);
1748out:
1749 return err;
1750}
1751
1752/*
1753 * Receive a datagram from a socket.
1754 */
1755
1756SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1757 unsigned int, flags)
1758{
1759 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1760}
1761
1762/*
1763 * Set a socket option. Because we don't know the option lengths we have
1764 * to pass the user mode parameter for the protocols to sort out.
1765 */
1766
1767SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1768 char __user *, optval, int, optlen)
1769{
1770 int err, fput_needed;
1771 struct socket *sock;
1772
1773 if (optlen < 0)
1774 return -EINVAL;
1775
1776 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1777 if (sock != NULL) {
1778 err = security_socket_setsockopt(sock, level, optname);
1779 if (err)
1780 goto out_put;
1781
1782 if (level == SOL_SOCKET)
1783 err =
1784 sock_setsockopt(sock, level, optname, optval,
1785 optlen);
1786 else
1787 err =
1788 sock->ops->setsockopt(sock, level, optname, optval,
1789 optlen);
1790out_put:
1791 fput_light(sock->file, fput_needed);
1792 }
1793 return err;
1794}
1795
1796/*
1797 * Get a socket option. Because we don't know the option lengths we have
1798 * to pass a user mode parameter for the protocols to sort out.
1799 */
1800
1801SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1802 char __user *, optval, int __user *, optlen)
1803{
1804 int err, fput_needed;
1805 struct socket *sock;
1806
1807 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1808 if (sock != NULL) {
1809 err = security_socket_getsockopt(sock, level, optname);
1810 if (err)
1811 goto out_put;
1812
1813 if (level == SOL_SOCKET)
1814 err =
1815 sock_getsockopt(sock, level, optname, optval,
1816 optlen);
1817 else
1818 err =
1819 sock->ops->getsockopt(sock, level, optname, optval,
1820 optlen);
1821out_put:
1822 fput_light(sock->file, fput_needed);
1823 }
1824 return err;
1825}
1826
1827/*
1828 * Shutdown a socket.
1829 */
1830
1831SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1832{
1833 int err, fput_needed;
1834 struct socket *sock;
1835
1836 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1837 if (sock != NULL) {
1838 err = security_socket_shutdown(sock, how);
1839 if (!err)
1840 err = sock->ops->shutdown(sock, how);
1841 fput_light(sock->file, fput_needed);
1842 }
1843 return err;
1844}
1845
1846/* A couple of helpful macros for getting the address of the 32/64 bit
1847 * fields which are the same type (int / unsigned) on our platforms.
1848 */
1849#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1850#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1851#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1852
1853struct used_address {
1854 struct sockaddr_storage name;
1855 unsigned int name_len;
1856};
1857
1858static int copy_msghdr_from_user(struct msghdr *kmsg,
1859 struct user_msghdr __user *umsg,
1860 struct sockaddr __user **save_addr,
1861 struct iovec **iov)
1862{
1863 struct sockaddr __user *uaddr;
1864 struct iovec __user *uiov;
1865 size_t nr_segs;
1866 ssize_t err;
1867
1868 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1869 __get_user(uaddr, &umsg->msg_name) ||
1870 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1871 __get_user(uiov, &umsg->msg_iov) ||
1872 __get_user(nr_segs, &umsg->msg_iovlen) ||
1873 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1874 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1875 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1876 return -EFAULT;
1877
1878 if (!uaddr)
1879 kmsg->msg_namelen = 0;
1880
1881 if (kmsg->msg_namelen < 0)
1882 return -EINVAL;
1883
1884 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1885 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1886
1887 if (save_addr)
1888 *save_addr = uaddr;
1889
1890 if (uaddr && kmsg->msg_namelen) {
1891 if (!save_addr) {
1892 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1893 kmsg->msg_name);
1894 if (err < 0)
1895 return err;
1896 }
1897 } else {
1898 kmsg->msg_name = NULL;
1899 kmsg->msg_namelen = 0;
1900 }
1901
1902 if (nr_segs > UIO_MAXIOV)
1903 return -EMSGSIZE;
1904
1905 kmsg->msg_iocb = NULL;
1906
1907 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1908 UIO_FASTIOV, iov, &kmsg->msg_iter);
1909}
1910
1911static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1912 struct msghdr *msg_sys, unsigned int flags,
1913 struct used_address *used_address,
1914 unsigned int allowed_msghdr_flags)
1915{
1916 struct compat_msghdr __user *msg_compat =
1917 (struct compat_msghdr __user *)msg;
1918 struct sockaddr_storage address;
1919 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1920 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1921 __aligned(sizeof(__kernel_size_t));
1922 /* 20 is size of ipv6_pktinfo */
1923 unsigned char *ctl_buf = ctl;
1924 int ctl_len;
1925 ssize_t err;
1926
1927 msg_sys->msg_name = &address;
1928
1929 if (MSG_CMSG_COMPAT & flags)
1930 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1931 else
1932 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1933 if (err < 0)
1934 return err;
1935
1936 err = -ENOBUFS;
1937
1938 if (msg_sys->msg_controllen > INT_MAX)
1939 goto out_freeiov;
1940 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1941 ctl_len = msg_sys->msg_controllen;
1942 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1943 err =
1944 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1945 sizeof(ctl));
1946 if (err)
1947 goto out_freeiov;
1948 ctl_buf = msg_sys->msg_control;
1949 ctl_len = msg_sys->msg_controllen;
1950 } else if (ctl_len) {
1951 if (ctl_len > sizeof(ctl)) {
1952 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1953 if (ctl_buf == NULL)
1954 goto out_freeiov;
1955 }
1956 err = -EFAULT;
1957 /*
1958 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1959 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1960 * checking falls down on this.
1961 */
1962 if (copy_from_user(ctl_buf,
1963 (void __user __force *)msg_sys->msg_control,
1964 ctl_len))
1965 goto out_freectl;
1966 msg_sys->msg_control = ctl_buf;
1967 }
1968 msg_sys->msg_flags = flags;
1969
1970 if (sock->file->f_flags & O_NONBLOCK)
1971 msg_sys->msg_flags |= MSG_DONTWAIT;
1972 /*
1973 * If this is sendmmsg() and current destination address is same as
1974 * previously succeeded address, omit asking LSM's decision.
1975 * used_address->name_len is initialized to UINT_MAX so that the first
1976 * destination address never matches.
1977 */
1978 if (used_address && msg_sys->msg_name &&
1979 used_address->name_len == msg_sys->msg_namelen &&
1980 !memcmp(&used_address->name, msg_sys->msg_name,
1981 used_address->name_len)) {
1982 err = sock_sendmsg_nosec(sock, msg_sys);
1983 goto out_freectl;
1984 }
1985 err = sock_sendmsg(sock, msg_sys);
1986 /*
1987 * If this is sendmmsg() and sending to current destination address was
1988 * successful, remember it.
1989 */
1990 if (used_address && err >= 0) {
1991 used_address->name_len = msg_sys->msg_namelen;
1992 if (msg_sys->msg_name)
1993 memcpy(&used_address->name, msg_sys->msg_name,
1994 used_address->name_len);
1995 }
1996
1997out_freectl:
1998 if (ctl_buf != ctl)
1999 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
2000out_freeiov:
2001 kfree(iov);
2002 return err;
2003}
2004
2005/*
2006 * BSD sendmsg interface
2007 */
2008
2009long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2010{
2011 int fput_needed, err;
2012 struct msghdr msg_sys;
2013 struct socket *sock;
2014
2015 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2016 if (!sock)
2017 goto out;
2018
2019 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
2020
2021 fput_light(sock->file, fput_needed);
2022out:
2023 return err;
2024}
2025
2026SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
2027{
2028 if (flags & MSG_CMSG_COMPAT)
2029 return -EINVAL;
2030 return __sys_sendmsg(fd, msg, flags);
2031}
2032
2033/*
2034 * Linux sendmmsg interface
2035 */
2036
2037int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2038 unsigned int flags)
2039{
2040 int fput_needed, err, datagrams;
2041 struct socket *sock;
2042 struct mmsghdr __user *entry;
2043 struct compat_mmsghdr __user *compat_entry;
2044 struct msghdr msg_sys;
2045 struct used_address used_address;
2046 unsigned int oflags = flags;
2047
2048 if (vlen > UIO_MAXIOV)
2049 vlen = UIO_MAXIOV;
2050
2051 datagrams = 0;
2052
2053 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2054 if (!sock)
2055 return err;
2056
2057 used_address.name_len = UINT_MAX;
2058 entry = mmsg;
2059 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2060 err = 0;
2061 flags |= MSG_BATCH;
2062
2063 while (datagrams < vlen) {
2064 if (datagrams == vlen - 1)
2065 flags = oflags;
2066
2067 if (MSG_CMSG_COMPAT & flags) {
2068 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2069 &msg_sys, flags, &used_address, MSG_EOR);
2070 if (err < 0)
2071 break;
2072 err = __put_user(err, &compat_entry->msg_len);
2073 ++compat_entry;
2074 } else {
2075 err = ___sys_sendmsg(sock,
2076 (struct user_msghdr __user *)entry,
2077 &msg_sys, flags, &used_address, MSG_EOR);
2078 if (err < 0)
2079 break;
2080 err = put_user(err, &entry->msg_len);
2081 ++entry;
2082 }
2083
2084 if (err)
2085 break;
2086 ++datagrams;
2087 if (msg_data_left(&msg_sys))
2088 break;
2089 cond_resched();
2090 }
2091
2092 fput_light(sock->file, fput_needed);
2093
2094 /* We only return an error if no datagrams were able to be sent */
2095 if (datagrams != 0)
2096 return datagrams;
2097
2098 return err;
2099}
2100
2101SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2102 unsigned int, vlen, unsigned int, flags)
2103{
2104 if (flags & MSG_CMSG_COMPAT)
2105 return -EINVAL;
2106 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2107}
2108
2109static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2110 struct msghdr *msg_sys, unsigned int flags, int nosec)
2111{
2112 struct compat_msghdr __user *msg_compat =
2113 (struct compat_msghdr __user *)msg;
2114 struct iovec iovstack[UIO_FASTIOV];
2115 struct iovec *iov = iovstack;
2116 unsigned long cmsg_ptr;
2117 int len;
2118 ssize_t err;
2119
2120 /* kernel mode address */
2121 struct sockaddr_storage addr;
2122
2123 /* user mode address pointers */
2124 struct sockaddr __user *uaddr;
2125 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2126
2127 msg_sys->msg_name = &addr;
2128
2129 if (MSG_CMSG_COMPAT & flags)
2130 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2131 else
2132 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2133 if (err < 0)
2134 return err;
2135
2136 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2137 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2138
2139 /* We assume all kernel code knows the size of sockaddr_storage */
2140 msg_sys->msg_namelen = 0;
2141
2142 if (sock->file->f_flags & O_NONBLOCK)
2143 flags |= MSG_DONTWAIT;
2144 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys, flags);
2145 if (err < 0)
2146 goto out_freeiov;
2147 len = err;
2148
2149 if (uaddr != NULL) {
2150 err = move_addr_to_user(&addr,
2151 msg_sys->msg_namelen, uaddr,
2152 uaddr_len);
2153 if (err < 0)
2154 goto out_freeiov;
2155 }
2156 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2157 COMPAT_FLAGS(msg));
2158 if (err)
2159 goto out_freeiov;
2160 if (MSG_CMSG_COMPAT & flags)
2161 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2162 &msg_compat->msg_controllen);
2163 else
2164 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2165 &msg->msg_controllen);
2166 if (err)
2167 goto out_freeiov;
2168 err = len;
2169
2170out_freeiov:
2171 kfree(iov);
2172 return err;
2173}
2174
2175/*
2176 * BSD recvmsg interface
2177 */
2178
2179long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2180{
2181 int fput_needed, err;
2182 struct msghdr msg_sys;
2183 struct socket *sock;
2184
2185 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2186 if (!sock)
2187 goto out;
2188
2189 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2190
2191 fput_light(sock->file, fput_needed);
2192out:
2193 return err;
2194}
2195
2196SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2197 unsigned int, flags)
2198{
2199 if (flags & MSG_CMSG_COMPAT)
2200 return -EINVAL;
2201 return __sys_recvmsg(fd, msg, flags);
2202}
2203
2204/*
2205 * Linux recvmmsg interface
2206 */
2207
2208int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2209 unsigned int flags, struct timespec *timeout)
2210{
2211 int fput_needed, err, datagrams;
2212 struct socket *sock;
2213 struct mmsghdr __user *entry;
2214 struct compat_mmsghdr __user *compat_entry;
2215 struct msghdr msg_sys;
2216 struct timespec64 end_time;
2217 struct timespec64 timeout64;
2218
2219 if (timeout &&
2220 poll_select_set_timeout(&end_time, timeout->tv_sec,
2221 timeout->tv_nsec))
2222 return -EINVAL;
2223
2224 datagrams = 0;
2225
2226 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2227 if (!sock)
2228 return err;
2229
2230 err = sock_error(sock->sk);
2231 if (err) {
2232 datagrams = err;
2233 goto out_put;
2234 }
2235
2236 entry = mmsg;
2237 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2238
2239 while (datagrams < vlen) {
2240 /*
2241 * No need to ask LSM for more than the first datagram.
2242 */
2243 if (MSG_CMSG_COMPAT & flags) {
2244 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2245 &msg_sys, flags & ~MSG_WAITFORONE,
2246 datagrams);
2247 if (err < 0)
2248 break;
2249 err = __put_user(err, &compat_entry->msg_len);
2250 ++compat_entry;
2251 } else {
2252 err = ___sys_recvmsg(sock,
2253 (struct user_msghdr __user *)entry,
2254 &msg_sys, flags & ~MSG_WAITFORONE,
2255 datagrams);
2256 if (err < 0)
2257 break;
2258 err = put_user(err, &entry->msg_len);
2259 ++entry;
2260 }
2261
2262 if (err)
2263 break;
2264 ++datagrams;
2265
2266 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2267 if (flags & MSG_WAITFORONE)
2268 flags |= MSG_DONTWAIT;
2269
2270 if (timeout) {
2271 ktime_get_ts64(&timeout64);
2272 *timeout = timespec64_to_timespec(
2273 timespec64_sub(end_time, timeout64));
2274 if (timeout->tv_sec < 0) {
2275 timeout->tv_sec = timeout->tv_nsec = 0;
2276 break;
2277 }
2278
2279 /* Timeout, return less than vlen datagrams */
2280 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2281 break;
2282 }
2283
2284 /* Out of band data, return right away */
2285 if (msg_sys.msg_flags & MSG_OOB)
2286 break;
2287 cond_resched();
2288 }
2289
2290 if (err == 0)
2291 goto out_put;
2292
2293 if (datagrams == 0) {
2294 datagrams = err;
2295 goto out_put;
2296 }
2297
2298 /*
2299 * We may return less entries than requested (vlen) if the
2300 * sock is non block and there aren't enough datagrams...
2301 */
2302 if (err != -EAGAIN) {
2303 /*
2304 * ... or if recvmsg returns an error after we
2305 * received some datagrams, where we record the
2306 * error to return on the next call or if the
2307 * app asks about it using getsockopt(SO_ERROR).
2308 */
2309 sock->sk->sk_err = -err;
2310 }
2311out_put:
2312 fput_light(sock->file, fput_needed);
2313
2314 return datagrams;
2315}
2316
2317SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2318 unsigned int, vlen, unsigned int, flags,
2319 struct timespec __user *, timeout)
2320{
2321 int datagrams;
2322 struct timespec timeout_sys;
2323
2324 if (flags & MSG_CMSG_COMPAT)
2325 return -EINVAL;
2326
2327 if (!timeout)
2328 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2329
2330 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2331 return -EFAULT;
2332
2333 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2334
2335 if (datagrams > 0 &&
2336 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2337 datagrams = -EFAULT;
2338
2339 return datagrams;
2340}
2341
2342#ifdef __ARCH_WANT_SYS_SOCKETCALL
2343/* Argument list sizes for sys_socketcall */
2344#define AL(x) ((x) * sizeof(unsigned long))
2345static const unsigned char nargs[21] = {
2346 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2347 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2348 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2349 AL(4), AL(5), AL(4)
2350};
2351
2352#undef AL
2353
2354/*
2355 * System call vectors.
2356 *
2357 * Argument checking cleaned up. Saved 20% in size.
2358 * This function doesn't need to set the kernel lock because
2359 * it is set by the callees.
2360 */
2361
2362SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2363{
2364 unsigned long a[AUDITSC_ARGS];
2365 unsigned long a0, a1;
2366 int err;
2367 unsigned int len;
2368
2369 if (call < 1 || call > SYS_SENDMMSG)
2370 return -EINVAL;
2371
2372 len = nargs[call];
2373 if (len > sizeof(a))
2374 return -EINVAL;
2375
2376 /* copy_from_user should be SMP safe. */
2377 if (copy_from_user(a, args, len))
2378 return -EFAULT;
2379
2380 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2381 if (err)
2382 return err;
2383
2384 a0 = a[0];
2385 a1 = a[1];
2386
2387 switch (call) {
2388 case SYS_SOCKET:
2389 err = sys_socket(a0, a1, a[2]);
2390 break;
2391 case SYS_BIND:
2392 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2393 break;
2394 case SYS_CONNECT:
2395 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2396 break;
2397 case SYS_LISTEN:
2398 err = sys_listen(a0, a1);
2399 break;
2400 case SYS_ACCEPT:
2401 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2402 (int __user *)a[2], 0);
2403 break;
2404 case SYS_GETSOCKNAME:
2405 err =
2406 sys_getsockname(a0, (struct sockaddr __user *)a1,
2407 (int __user *)a[2]);
2408 break;
2409 case SYS_GETPEERNAME:
2410 err =
2411 sys_getpeername(a0, (struct sockaddr __user *)a1,
2412 (int __user *)a[2]);
2413 break;
2414 case SYS_SOCKETPAIR:
2415 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2416 break;
2417 case SYS_SEND:
2418 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2419 break;
2420 case SYS_SENDTO:
2421 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2422 (struct sockaddr __user *)a[4], a[5]);
2423 break;
2424 case SYS_RECV:
2425 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2426 break;
2427 case SYS_RECVFROM:
2428 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2429 (struct sockaddr __user *)a[4],
2430 (int __user *)a[5]);
2431 break;
2432 case SYS_SHUTDOWN:
2433 err = sys_shutdown(a0, a1);
2434 break;
2435 case SYS_SETSOCKOPT:
2436 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2437 break;
2438 case SYS_GETSOCKOPT:
2439 err =
2440 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2441 (int __user *)a[4]);
2442 break;
2443 case SYS_SENDMSG:
2444 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2445 break;
2446 case SYS_SENDMMSG:
2447 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2448 break;
2449 case SYS_RECVMSG:
2450 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2451 break;
2452 case SYS_RECVMMSG:
2453 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2454 (struct timespec __user *)a[4]);
2455 break;
2456 case SYS_ACCEPT4:
2457 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2458 (int __user *)a[2], a[3]);
2459 break;
2460 default:
2461 err = -EINVAL;
2462 break;
2463 }
2464 return err;
2465}
2466
2467#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2468
2469/**
2470 * sock_register - add a socket protocol handler
2471 * @ops: description of protocol
2472 *
2473 * This function is called by a protocol handler that wants to
2474 * advertise its address family, and have it linked into the
2475 * socket interface. The value ops->family corresponds to the
2476 * socket system call protocol family.
2477 */
2478int sock_register(const struct net_proto_family *ops)
2479{
2480 int err;
2481
2482 if (ops->family >= NPROTO) {
2483 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2484 return -ENOBUFS;
2485 }
2486
2487 spin_lock(&net_family_lock);
2488 if (rcu_dereference_protected(net_families[ops->family],
2489 lockdep_is_held(&net_family_lock)))
2490 err = -EEXIST;
2491 else {
2492 rcu_assign_pointer(net_families[ops->family], ops);
2493 err = 0;
2494 }
2495 spin_unlock(&net_family_lock);
2496
2497 pr_info("NET: Registered protocol family %d\n", ops->family);
2498 return err;
2499}
2500EXPORT_SYMBOL(sock_register);
2501
2502/**
2503 * sock_unregister - remove a protocol handler
2504 * @family: protocol family to remove
2505 *
2506 * This function is called by a protocol handler that wants to
2507 * remove its address family, and have it unlinked from the
2508 * new socket creation.
2509 *
2510 * If protocol handler is a module, then it can use module reference
2511 * counts to protect against new references. If protocol handler is not
2512 * a module then it needs to provide its own protection in
2513 * the ops->create routine.
2514 */
2515void sock_unregister(int family)
2516{
2517 BUG_ON(family < 0 || family >= NPROTO);
2518
2519 spin_lock(&net_family_lock);
2520 RCU_INIT_POINTER(net_families[family], NULL);
2521 spin_unlock(&net_family_lock);
2522
2523 synchronize_rcu();
2524
2525 pr_info("NET: Unregistered protocol family %d\n", family);
2526}
2527EXPORT_SYMBOL(sock_unregister);
2528
2529static int __init sock_init(void)
2530{
2531 int err;
2532 /*
2533 * Initialize the network sysctl infrastructure.
2534 */
2535 err = net_sysctl_init();
2536 if (err)
2537 goto out;
2538
2539 /*
2540 * Initialize skbuff SLAB cache
2541 */
2542 skb_init();
2543
2544 /*
2545 * Initialize the protocols module.
2546 */
2547
2548 init_inodecache();
2549
2550 err = register_filesystem(&sock_fs_type);
2551 if (err)
2552 goto out_fs;
2553 sock_mnt = kern_mount(&sock_fs_type);
2554 if (IS_ERR(sock_mnt)) {
2555 err = PTR_ERR(sock_mnt);
2556 goto out_mount;
2557 }
2558
2559 /* The real protocol initialization is performed in later initcalls.
2560 */
2561
2562#ifdef CONFIG_NETFILTER
2563 err = netfilter_init();
2564 if (err)
2565 goto out;
2566#endif
2567
2568 ptp_classifier_init();
2569
2570out:
2571 return err;
2572
2573out_mount:
2574 unregister_filesystem(&sock_fs_type);
2575out_fs:
2576 goto out;
2577}
2578
2579core_initcall(sock_init); /* early initcall */
2580
2581#ifdef CONFIG_PROC_FS
2582void socket_seq_show(struct seq_file *seq)
2583{
2584 int cpu;
2585 int counter = 0;
2586
2587 for_each_possible_cpu(cpu)
2588 counter += per_cpu(sockets_in_use, cpu);
2589
2590 /* It can be negative, by the way. 8) */
2591 if (counter < 0)
2592 counter = 0;
2593
2594 seq_printf(seq, "sockets: used %d\n", counter);
2595}
2596#endif /* CONFIG_PROC_FS */
2597
2598#ifdef CONFIG_COMPAT
2599static int do_siocgstamp(struct net *net, struct socket *sock,
2600 unsigned int cmd, void __user *up)
2601{
2602 mm_segment_t old_fs = get_fs();
2603 struct timeval ktv;
2604 int err;
2605
2606 set_fs(KERNEL_DS);
2607 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2608 set_fs(old_fs);
2609 if (!err)
2610 err = compat_put_timeval(&ktv, up);
2611
2612 return err;
2613}
2614
2615static int do_siocgstampns(struct net *net, struct socket *sock,
2616 unsigned int cmd, void __user *up)
2617{
2618 mm_segment_t old_fs = get_fs();
2619 struct timespec kts;
2620 int err;
2621
2622 set_fs(KERNEL_DS);
2623 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2624 set_fs(old_fs);
2625 if (!err)
2626 err = compat_put_timespec(&kts, up);
2627
2628 return err;
2629}
2630
2631static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2632{
2633 struct ifreq __user *uifr;
2634 int err;
2635
2636 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2637 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2638 return -EFAULT;
2639
2640 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2641 if (err)
2642 return err;
2643
2644 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2645 return -EFAULT;
2646
2647 return 0;
2648}
2649
2650static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2651{
2652 struct compat_ifconf ifc32;
2653 struct ifconf ifc;
2654 struct ifconf __user *uifc;
2655 struct compat_ifreq __user *ifr32;
2656 struct ifreq __user *ifr;
2657 unsigned int i, j;
2658 int err;
2659
2660 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2661 return -EFAULT;
2662
2663 memset(&ifc, 0, sizeof(ifc));
2664 if (ifc32.ifcbuf == 0) {
2665 ifc32.ifc_len = 0;
2666 ifc.ifc_len = 0;
2667 ifc.ifc_req = NULL;
2668 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2669 } else {
2670 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2671 sizeof(struct ifreq);
2672 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2673 ifc.ifc_len = len;
2674 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2675 ifr32 = compat_ptr(ifc32.ifcbuf);
2676 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2677 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2678 return -EFAULT;
2679 ifr++;
2680 ifr32++;
2681 }
2682 }
2683 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2684 return -EFAULT;
2685
2686 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2687 if (err)
2688 return err;
2689
2690 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2691 return -EFAULT;
2692
2693 ifr = ifc.ifc_req;
2694 ifr32 = compat_ptr(ifc32.ifcbuf);
2695 for (i = 0, j = 0;
2696 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2697 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2698 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2699 return -EFAULT;
2700 ifr32++;
2701 ifr++;
2702 }
2703
2704 if (ifc32.ifcbuf == 0) {
2705 /* Translate from 64-bit structure multiple to
2706 * a 32-bit one.
2707 */
2708 i = ifc.ifc_len;
2709 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2710 ifc32.ifc_len = i;
2711 } else {
2712 ifc32.ifc_len = i;
2713 }
2714 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2715 return -EFAULT;
2716
2717 return 0;
2718}
2719
2720static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2721{
2722 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2723 bool convert_in = false, convert_out = false;
2724 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2725 struct ethtool_rxnfc __user *rxnfc;
2726 struct ifreq __user *ifr;
2727 u32 rule_cnt = 0, actual_rule_cnt;
2728 u32 ethcmd;
2729 u32 data;
2730 int ret;
2731
2732 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2733 return -EFAULT;
2734
2735 compat_rxnfc = compat_ptr(data);
2736
2737 if (get_user(ethcmd, &compat_rxnfc->cmd))
2738 return -EFAULT;
2739
2740 /* Most ethtool structures are defined without padding.
2741 * Unfortunately struct ethtool_rxnfc is an exception.
2742 */
2743 switch (ethcmd) {
2744 default:
2745 break;
2746 case ETHTOOL_GRXCLSRLALL:
2747 /* Buffer size is variable */
2748 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2749 return -EFAULT;
2750 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2751 return -ENOMEM;
2752 buf_size += rule_cnt * sizeof(u32);
2753 /* fall through */
2754 case ETHTOOL_GRXRINGS:
2755 case ETHTOOL_GRXCLSRLCNT:
2756 case ETHTOOL_GRXCLSRULE:
2757 case ETHTOOL_SRXCLSRLINS:
2758 convert_out = true;
2759 /* fall through */
2760 case ETHTOOL_SRXCLSRLDEL:
2761 buf_size += sizeof(struct ethtool_rxnfc);
2762 convert_in = true;
2763 break;
2764 }
2765
2766 ifr = compat_alloc_user_space(buf_size);
2767 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2768
2769 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2770 return -EFAULT;
2771
2772 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2773 &ifr->ifr_ifru.ifru_data))
2774 return -EFAULT;
2775
2776 if (convert_in) {
2777 /* We expect there to be holes between fs.m_ext and
2778 * fs.ring_cookie and at the end of fs, but nowhere else.
2779 */
2780 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2781 sizeof(compat_rxnfc->fs.m_ext) !=
2782 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2783 sizeof(rxnfc->fs.m_ext));
2784 BUILD_BUG_ON(
2785 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2786 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2787 offsetof(struct ethtool_rxnfc, fs.location) -
2788 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2789
2790 if (copy_in_user(rxnfc, compat_rxnfc,
2791 (void __user *)(&rxnfc->fs.m_ext + 1) -
2792 (void __user *)rxnfc) ||
2793 copy_in_user(&rxnfc->fs.ring_cookie,
2794 &compat_rxnfc->fs.ring_cookie,
2795 (void __user *)(&rxnfc->fs.location + 1) -
2796 (void __user *)&rxnfc->fs.ring_cookie) ||
2797 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2798 sizeof(rxnfc->rule_cnt)))
2799 return -EFAULT;
2800 }
2801
2802 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2803 if (ret)
2804 return ret;
2805
2806 if (convert_out) {
2807 if (copy_in_user(compat_rxnfc, rxnfc,
2808 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2809 (const void __user *)rxnfc) ||
2810 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2811 &rxnfc->fs.ring_cookie,
2812 (const void __user *)(&rxnfc->fs.location + 1) -
2813 (const void __user *)&rxnfc->fs.ring_cookie) ||
2814 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2815 sizeof(rxnfc->rule_cnt)))
2816 return -EFAULT;
2817
2818 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2819 /* As an optimisation, we only copy the actual
2820 * number of rules that the underlying
2821 * function returned. Since Mallory might
2822 * change the rule count in user memory, we
2823 * check that it is less than the rule count
2824 * originally given (as the user buffer size),
2825 * which has been range-checked.
2826 */
2827 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2828 return -EFAULT;
2829 if (actual_rule_cnt < rule_cnt)
2830 rule_cnt = actual_rule_cnt;
2831 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2832 &rxnfc->rule_locs[0],
2833 rule_cnt * sizeof(u32)))
2834 return -EFAULT;
2835 }
2836 }
2837
2838 return 0;
2839}
2840
2841static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2842{
2843 void __user *uptr;
2844 compat_uptr_t uptr32;
2845 struct ifreq __user *uifr;
2846
2847 uifr = compat_alloc_user_space(sizeof(*uifr));
2848 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2849 return -EFAULT;
2850
2851 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2852 return -EFAULT;
2853
2854 uptr = compat_ptr(uptr32);
2855
2856 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2857 return -EFAULT;
2858
2859 return dev_ioctl(net, SIOCWANDEV, uifr);
2860}
2861
2862static int bond_ioctl(struct net *net, unsigned int cmd,
2863 struct compat_ifreq __user *ifr32)
2864{
2865 struct ifreq kifr;
2866 mm_segment_t old_fs;
2867 int err;
2868
2869 switch (cmd) {
2870 case SIOCBONDENSLAVE:
2871 case SIOCBONDRELEASE:
2872 case SIOCBONDSETHWADDR:
2873 case SIOCBONDCHANGEACTIVE:
2874 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2875 return -EFAULT;
2876
2877 old_fs = get_fs();
2878 set_fs(KERNEL_DS);
2879 err = dev_ioctl(net, cmd,
2880 (struct ifreq __user __force *) &kifr);
2881 set_fs(old_fs);
2882
2883 return err;
2884 default:
2885 return -ENOIOCTLCMD;
2886 }
2887}
2888
2889/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2890static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2891 struct compat_ifreq __user *u_ifreq32)
2892{
2893 struct ifreq __user *u_ifreq64;
2894 char tmp_buf[IFNAMSIZ];
2895 void __user *data64;
2896 u32 data32;
2897
2898 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2899 IFNAMSIZ))
2900 return -EFAULT;
2901 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2902 return -EFAULT;
2903 data64 = compat_ptr(data32);
2904
2905 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2906
2907 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2908 IFNAMSIZ))
2909 return -EFAULT;
2910 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2911 return -EFAULT;
2912
2913 return dev_ioctl(net, cmd, u_ifreq64);
2914}
2915
2916static int dev_ifsioc(struct net *net, struct socket *sock,
2917 unsigned int cmd, struct compat_ifreq __user *uifr32)
2918{
2919 struct ifreq __user *uifr;
2920 int err;
2921
2922 uifr = compat_alloc_user_space(sizeof(*uifr));
2923 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2924 return -EFAULT;
2925
2926 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2927
2928 if (!err) {
2929 switch (cmd) {
2930 case SIOCGIFFLAGS:
2931 case SIOCGIFMETRIC:
2932 case SIOCGIFMTU:
2933 case SIOCGIFMEM:
2934 case SIOCGIFHWADDR:
2935 case SIOCGIFINDEX:
2936 case SIOCGIFADDR:
2937 case SIOCGIFBRDADDR:
2938 case SIOCGIFDSTADDR:
2939 case SIOCGIFNETMASK:
2940 case SIOCGIFPFLAGS:
2941 case SIOCGIFTXQLEN:
2942 case SIOCGMIIPHY:
2943 case SIOCGMIIREG:
2944 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2945 err = -EFAULT;
2946 break;
2947 }
2948 }
2949 return err;
2950}
2951
2952static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2953 struct compat_ifreq __user *uifr32)
2954{
2955 struct ifreq ifr;
2956 struct compat_ifmap __user *uifmap32;
2957 mm_segment_t old_fs;
2958 int err;
2959
2960 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2961 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2962 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2963 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2964 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2965 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2966 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2967 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2968 if (err)
2969 return -EFAULT;
2970
2971 old_fs = get_fs();
2972 set_fs(KERNEL_DS);
2973 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2974 set_fs(old_fs);
2975
2976 if (cmd == SIOCGIFMAP && !err) {
2977 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2978 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2979 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2980 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2981 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2982 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2983 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2984 if (err)
2985 err = -EFAULT;
2986 }
2987 return err;
2988}
2989
2990struct rtentry32 {
2991 u32 rt_pad1;
2992 struct sockaddr rt_dst; /* target address */
2993 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2994 struct sockaddr rt_genmask; /* target network mask (IP) */
2995 unsigned short rt_flags;
2996 short rt_pad2;
2997 u32 rt_pad3;
2998 unsigned char rt_tos;
2999 unsigned char rt_class;
3000 short rt_pad4;
3001 short rt_metric; /* +1 for binary compatibility! */
3002 /* char * */ u32 rt_dev; /* forcing the device at add */
3003 u32 rt_mtu; /* per route MTU/Window */
3004 u32 rt_window; /* Window clamping */
3005 unsigned short rt_irtt; /* Initial RTT */
3006};
3007
3008struct in6_rtmsg32 {
3009 struct in6_addr rtmsg_dst;
3010 struct in6_addr rtmsg_src;
3011 struct in6_addr rtmsg_gateway;
3012 u32 rtmsg_type;
3013 u16 rtmsg_dst_len;
3014 u16 rtmsg_src_len;
3015 u32 rtmsg_metric;
3016 u32 rtmsg_info;
3017 u32 rtmsg_flags;
3018 s32 rtmsg_ifindex;
3019};
3020
3021static int routing_ioctl(struct net *net, struct socket *sock,
3022 unsigned int cmd, void __user *argp)
3023{
3024 int ret;
3025 void *r = NULL;
3026 struct in6_rtmsg r6;
3027 struct rtentry r4;
3028 char devname[16];
3029 u32 rtdev;
3030 mm_segment_t old_fs = get_fs();
3031
3032 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
3033 struct in6_rtmsg32 __user *ur6 = argp;
3034 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
3035 3 * sizeof(struct in6_addr));
3036 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
3037 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
3038 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
3039 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
3040 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
3041 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
3042 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3043
3044 r = (void *) &r6;
3045 } else { /* ipv4 */
3046 struct rtentry32 __user *ur4 = argp;
3047 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3048 3 * sizeof(struct sockaddr));
3049 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3050 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3051 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3052 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3053 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3054 ret |= get_user(rtdev, &(ur4->rt_dev));
3055 if (rtdev) {
3056 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3057 r4.rt_dev = (char __user __force *)devname;
3058 devname[15] = 0;
3059 } else
3060 r4.rt_dev = NULL;
3061
3062 r = (void *) &r4;
3063 }
3064
3065 if (ret) {
3066 ret = -EFAULT;
3067 goto out;
3068 }
3069
3070 set_fs(KERNEL_DS);
3071 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3072 set_fs(old_fs);
3073
3074out:
3075 return ret;
3076}
3077
3078/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3079 * for some operations; this forces use of the newer bridge-utils that
3080 * use compatible ioctls
3081 */
3082static int old_bridge_ioctl(compat_ulong_t __user *argp)
3083{
3084 compat_ulong_t tmp;
3085
3086 if (get_user(tmp, argp))
3087 return -EFAULT;
3088 if (tmp == BRCTL_GET_VERSION)
3089 return BRCTL_VERSION + 1;
3090 return -EINVAL;
3091}
3092
3093static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3094 unsigned int cmd, unsigned long arg)
3095{
3096 void __user *argp = compat_ptr(arg);
3097 struct sock *sk = sock->sk;
3098 struct net *net = sock_net(sk);
3099
3100 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3101 return compat_ifr_data_ioctl(net, cmd, argp);
3102
3103 switch (cmd) {
3104 case SIOCSIFBR:
3105 case SIOCGIFBR:
3106 return old_bridge_ioctl(argp);
3107 case SIOCGIFNAME:
3108 return dev_ifname32(net, argp);
3109 case SIOCGIFCONF:
3110 return dev_ifconf(net, argp);
3111 case SIOCETHTOOL:
3112 return ethtool_ioctl(net, argp);
3113 case SIOCWANDEV:
3114 return compat_siocwandev(net, argp);
3115 case SIOCGIFMAP:
3116 case SIOCSIFMAP:
3117 return compat_sioc_ifmap(net, cmd, argp);
3118 case SIOCBONDENSLAVE:
3119 case SIOCBONDRELEASE:
3120 case SIOCBONDSETHWADDR:
3121 case SIOCBONDCHANGEACTIVE:
3122 return bond_ioctl(net, cmd, argp);
3123 case SIOCADDRT:
3124 case SIOCDELRT:
3125 return routing_ioctl(net, sock, cmd, argp);
3126 case SIOCGSTAMP:
3127 return do_siocgstamp(net, sock, cmd, argp);
3128 case SIOCGSTAMPNS:
3129 return do_siocgstampns(net, sock, cmd, argp);
3130 case SIOCBONDSLAVEINFOQUERY:
3131 case SIOCBONDINFOQUERY:
3132 case SIOCSHWTSTAMP:
3133 case SIOCGHWTSTAMP:
3134 return compat_ifr_data_ioctl(net, cmd, argp);
3135
3136 case FIOSETOWN:
3137 case SIOCSPGRP:
3138 case FIOGETOWN:
3139 case SIOCGPGRP:
3140 case SIOCBRADDBR:
3141 case SIOCBRDELBR:
3142 case SIOCGIFVLAN:
3143 case SIOCSIFVLAN:
3144 case SIOCADDDLCI:
3145 case SIOCDELDLCI:
3146 case SIOCGSKNS:
3147 return sock_ioctl(file, cmd, arg);
3148
3149 case SIOCGIFFLAGS:
3150 case SIOCSIFFLAGS:
3151 case SIOCGIFMETRIC:
3152 case SIOCSIFMETRIC:
3153 case SIOCGIFMTU:
3154 case SIOCSIFMTU:
3155 case SIOCGIFMEM:
3156 case SIOCSIFMEM:
3157 case SIOCGIFHWADDR:
3158 case SIOCSIFHWADDR:
3159 case SIOCADDMULTI:
3160 case SIOCDELMULTI:
3161 case SIOCGIFINDEX:
3162 case SIOCGIFADDR:
3163 case SIOCSIFADDR:
3164 case SIOCSIFHWBROADCAST:
3165 case SIOCDIFADDR:
3166 case SIOCGIFBRDADDR:
3167 case SIOCSIFBRDADDR:
3168 case SIOCGIFDSTADDR:
3169 case SIOCSIFDSTADDR:
3170 case SIOCGIFNETMASK:
3171 case SIOCSIFNETMASK:
3172 case SIOCSIFPFLAGS:
3173 case SIOCGIFPFLAGS:
3174 case SIOCGIFTXQLEN:
3175 case SIOCSIFTXQLEN:
3176 case SIOCBRADDIF:
3177 case SIOCBRDELIF:
3178 case SIOCSIFNAME:
3179 case SIOCGMIIPHY:
3180 case SIOCGMIIREG:
3181 case SIOCSMIIREG:
3182 return dev_ifsioc(net, sock, cmd, argp);
3183
3184 case SIOCSARP:
3185 case SIOCGARP:
3186 case SIOCDARP:
3187 case SIOCATMARK:
3188 return sock_do_ioctl(net, sock, cmd, arg);
3189 }
3190
3191 return -ENOIOCTLCMD;
3192}
3193
3194static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3195 unsigned long arg)
3196{
3197 struct socket *sock = file->private_data;
3198 int ret = -ENOIOCTLCMD;
3199 struct sock *sk;
3200 struct net *net;
3201
3202 sk = sock->sk;
3203 net = sock_net(sk);
3204
3205 if (sock->ops->compat_ioctl)
3206 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3207
3208 if (ret == -ENOIOCTLCMD &&
3209 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3210 ret = compat_wext_handle_ioctl(net, cmd, arg);
3211
3212 if (ret == -ENOIOCTLCMD)
3213 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3214
3215 return ret;
3216}
3217#endif
3218
3219int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3220{
3221 return sock->ops->bind(sock, addr, addrlen);
3222}
3223EXPORT_SYMBOL(kernel_bind);
3224
3225int kernel_listen(struct socket *sock, int backlog)
3226{
3227 return sock->ops->listen(sock, backlog);
3228}
3229EXPORT_SYMBOL(kernel_listen);
3230
3231int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3232{
3233 struct sock *sk = sock->sk;
3234 int err;
3235
3236 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3237 newsock);
3238 if (err < 0)
3239 goto done;
3240
3241 err = sock->ops->accept(sock, *newsock, flags);
3242 if (err < 0) {
3243 sock_release(*newsock);
3244 *newsock = NULL;
3245 goto done;
3246 }
3247
3248 (*newsock)->ops = sock->ops;
3249 __module_get((*newsock)->ops->owner);
3250
3251done:
3252 return err;
3253}
3254EXPORT_SYMBOL(kernel_accept);
3255
3256int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3257 int flags)
3258{
3259 return sock->ops->connect(sock, addr, addrlen, flags);
3260}
3261EXPORT_SYMBOL(kernel_connect);
3262
3263int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3264 int *addrlen)
3265{
3266 return sock->ops->getname(sock, addr, addrlen, 0);
3267}
3268EXPORT_SYMBOL(kernel_getsockname);
3269
3270int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3271 int *addrlen)
3272{
3273 return sock->ops->getname(sock, addr, addrlen, 1);
3274}
3275EXPORT_SYMBOL(kernel_getpeername);
3276
3277int kernel_getsockopt(struct socket *sock, int level, int optname,
3278 char *optval, int *optlen)
3279{
3280 mm_segment_t oldfs = get_fs();
3281 char __user *uoptval;
3282 int __user *uoptlen;
3283 int err;
3284
3285 uoptval = (char __user __force *) optval;
3286 uoptlen = (int __user __force *) optlen;
3287
3288 set_fs(KERNEL_DS);
3289 if (level == SOL_SOCKET)
3290 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3291 else
3292 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3293 uoptlen);
3294 set_fs(oldfs);
3295 return err;
3296}
3297EXPORT_SYMBOL(kernel_getsockopt);
3298
3299int kernel_setsockopt(struct socket *sock, int level, int optname,
3300 char *optval, unsigned int optlen)
3301{
3302 mm_segment_t oldfs = get_fs();
3303 char __user *uoptval;
3304 int err;
3305
3306 uoptval = (char __user __force *) optval;
3307
3308 set_fs(KERNEL_DS);
3309 if (level == SOL_SOCKET)
3310 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3311 else
3312 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3313 optlen);
3314 set_fs(oldfs);
3315 return err;
3316}
3317EXPORT_SYMBOL(kernel_setsockopt);
3318
3319int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3320 size_t size, int flags)
3321{
3322 if (sock->ops->sendpage)
3323 return sock->ops->sendpage(sock, page, offset, size, flags);
3324
3325 return sock_no_sendpage(sock, page, offset, size, flags);
3326}
3327EXPORT_SYMBOL(kernel_sendpage);
3328
3329int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3330{
3331 mm_segment_t oldfs = get_fs();
3332 int err;
3333
3334 set_fs(KERNEL_DS);
3335 err = sock->ops->ioctl(sock, cmd, arg);
3336 set_fs(oldfs);
3337
3338 return err;
3339}
3340EXPORT_SYMBOL(kernel_sock_ioctl);
3341
3342int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3343{
3344 return sock->ops->shutdown(sock, how);
3345}
3346EXPORT_SYMBOL(kernel_sock_shutdown);
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
112unsigned int sysctl_net_busy_read __read_mostly;
113unsigned int sysctl_net_busy_poll __read_mostly;
114#endif
115
116static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to);
117static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from);
118static int sock_mmap(struct file *file, struct vm_area_struct *vma);
119
120static int sock_close(struct inode *inode, struct file *file);
121static unsigned int sock_poll(struct file *file,
122 struct poll_table_struct *wait);
123static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
124#ifdef CONFIG_COMPAT
125static long compat_sock_ioctl(struct file *file,
126 unsigned int cmd, unsigned long arg);
127#endif
128static int sock_fasync(int fd, struct file *filp, int on);
129static ssize_t sock_sendpage(struct file *file, struct page *page,
130 int offset, size_t size, loff_t *ppos, int more);
131static 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
140static 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
162static DEFINE_SPINLOCK(net_family_lock);
163static const struct net_proto_family __rcu *net_families[NPROTO] __read_mostly;
164
165/*
166 * Statistics counters of the socket lists
167 */
168
169static 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
188int 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
216static 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
243static struct kmem_cache *sock_inode_cachep __read_mostly;
244
245static 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
272static 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
283static 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
290static 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
304static 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 */
313static 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
319static const struct dentry_operations sockfs_dentry_operations = {
320 .d_dname = sockfs_dname,
321};
322
323static struct dentry *sockfs_mount(struct file_system_type *fs_type,
324 int flags, const char *dev_name, void *data)
325{
326 return mount_pseudo(fs_type, "socket:", &sockfs_ops,
327 &sockfs_dentry_operations, SOCKFS_MAGIC);
328}
329
330static struct vfsmount *sock_mnt __read_mostly;
331
332static struct file_system_type sock_fs_type = {
333 .name = "sockfs",
334 .mount = sockfs_mount,
335 .kill_sb = kill_anon_super,
336};
337
338/*
339 * Obtains the first available file descriptor and sets it up for use.
340 *
341 * These functions create file structures and maps them to fd space
342 * of the current process. On success it returns file descriptor
343 * and file struct implicitly stored in sock->file.
344 * Note that another thread may close file descriptor before we return
345 * from this function. We use the fact that now we do not refer
346 * to socket after mapping. If one day we will need it, this
347 * function will increment ref. count on file by 1.
348 *
349 * In any case returned fd MAY BE not valid!
350 * This race condition is unavoidable
351 * with shared fd spaces, we cannot solve it inside kernel,
352 * but we take care of internal coherence yet.
353 */
354
355struct file *sock_alloc_file(struct socket *sock, int flags, const char *dname)
356{
357 struct qstr name = { .name = "" };
358 struct path path;
359 struct file *file;
360
361 if (dname) {
362 name.name = dname;
363 name.len = strlen(name.name);
364 } else if (sock->sk) {
365 name.name = sock->sk->sk_prot_creator->name;
366 name.len = strlen(name.name);
367 }
368 path.dentry = d_alloc_pseudo(sock_mnt->mnt_sb, &name);
369 if (unlikely(!path.dentry))
370 return ERR_PTR(-ENOMEM);
371 path.mnt = mntget(sock_mnt);
372
373 d_instantiate(path.dentry, SOCK_INODE(sock));
374
375 file = alloc_file(&path, FMODE_READ | FMODE_WRITE,
376 &socket_file_ops);
377 if (IS_ERR(file)) {
378 /* drop dentry, keep inode */
379 ihold(d_inode(path.dentry));
380 path_put(&path);
381 return file;
382 }
383
384 sock->file = file;
385 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
386 file->private_data = sock;
387 return file;
388}
389EXPORT_SYMBOL(sock_alloc_file);
390
391static int sock_map_fd(struct socket *sock, int flags)
392{
393 struct file *newfile;
394 int fd = get_unused_fd_flags(flags);
395 if (unlikely(fd < 0))
396 return fd;
397
398 newfile = sock_alloc_file(sock, flags, NULL);
399 if (likely(!IS_ERR(newfile))) {
400 fd_install(fd, newfile);
401 return fd;
402 }
403
404 put_unused_fd(fd);
405 return PTR_ERR(newfile);
406}
407
408struct socket *sock_from_file(struct file *file, int *err)
409{
410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
412
413 *err = -ENOTSOCK;
414 return NULL;
415}
416EXPORT_SYMBOL(sock_from_file);
417
418/**
419 * sockfd_lookup - Go from a file number to its socket slot
420 * @fd: file handle
421 * @err: pointer to an error code return
422 *
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
427 *
428 * On a success the socket object pointer is returned.
429 */
430
431struct socket *sockfd_lookup(int fd, int *err)
432{
433 struct file *file;
434 struct socket *sock;
435
436 file = fget(fd);
437 if (!file) {
438 *err = -EBADF;
439 return NULL;
440 }
441
442 sock = sock_from_file(file, err);
443 if (!sock)
444 fput(file);
445 return sock;
446}
447EXPORT_SYMBOL(sockfd_lookup);
448
449static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
450{
451 struct fd f = fdget(fd);
452 struct socket *sock;
453
454 *err = -EBADF;
455 if (f.file) {
456 sock = sock_from_file(f.file, err);
457 if (likely(sock)) {
458 *fput_needed = f.flags;
459 return sock;
460 }
461 fdput(f);
462 }
463 return NULL;
464}
465
466#define XATTR_SOCKPROTONAME_SUFFIX "sockprotoname"
467#define XATTR_NAME_SOCKPROTONAME (XATTR_SYSTEM_PREFIX XATTR_SOCKPROTONAME_SUFFIX)
468#define XATTR_NAME_SOCKPROTONAME_LEN (sizeof(XATTR_NAME_SOCKPROTONAME)-1)
469static ssize_t sockfs_getxattr(struct dentry *dentry,
470 const char *name, void *value, size_t size)
471{
472 const char *proto_name;
473 size_t proto_size;
474 int error;
475
476 error = -ENODATA;
477 if (!strncmp(name, XATTR_NAME_SOCKPROTONAME, XATTR_NAME_SOCKPROTONAME_LEN)) {
478 proto_name = dentry->d_name.name;
479 proto_size = strlen(proto_name);
480
481 if (value) {
482 error = -ERANGE;
483 if (proto_size + 1 > size)
484 goto out;
485
486 strncpy(value, proto_name, proto_size + 1);
487 }
488 error = proto_size + 1;
489 }
490
491out:
492 return error;
493}
494
495static ssize_t sockfs_listxattr(struct dentry *dentry, char *buffer,
496 size_t size)
497{
498 ssize_t len;
499 ssize_t used = 0;
500
501 len = security_inode_listsecurity(d_inode(dentry), buffer, size);
502 if (len < 0)
503 return len;
504 used += len;
505 if (buffer) {
506 if (size < used)
507 return -ERANGE;
508 buffer += len;
509 }
510
511 len = (XATTR_NAME_SOCKPROTONAME_LEN + 1);
512 used += len;
513 if (buffer) {
514 if (size < used)
515 return -ERANGE;
516 memcpy(buffer, XATTR_NAME_SOCKPROTONAME, len);
517 buffer += len;
518 }
519
520 return used;
521}
522
523static const struct inode_operations sockfs_inode_ops = {
524 .getxattr = sockfs_getxattr,
525 .listxattr = sockfs_listxattr,
526};
527
528/**
529 * sock_alloc - allocate a socket
530 *
531 * Allocate a new inode and socket object. The two are bound together
532 * and initialised. The socket is then returned. If we are out of inodes
533 * NULL is returned.
534 */
535
536struct socket *sock_alloc(void)
537{
538 struct inode *inode;
539 struct socket *sock;
540
541 inode = new_inode_pseudo(sock_mnt->mnt_sb);
542 if (!inode)
543 return NULL;
544
545 sock = SOCKET_I(inode);
546
547 kmemcheck_annotate_bitfield(sock, type);
548 inode->i_ino = get_next_ino();
549 inode->i_mode = S_IFSOCK | S_IRWXUGO;
550 inode->i_uid = current_fsuid();
551 inode->i_gid = current_fsgid();
552 inode->i_op = &sockfs_inode_ops;
553
554 this_cpu_add(sockets_in_use, 1);
555 return sock;
556}
557EXPORT_SYMBOL(sock_alloc);
558
559/**
560 * sock_release - close a socket
561 * @sock: socket to close
562 *
563 * The socket is released from the protocol stack if it has a release
564 * callback, and the inode is then released if the socket is bound to
565 * an inode not a file.
566 */
567
568void sock_release(struct socket *sock)
569{
570 if (sock->ops) {
571 struct module *owner = sock->ops->owner;
572
573 sock->ops->release(sock);
574 sock->ops = NULL;
575 module_put(owner);
576 }
577
578 if (rcu_dereference_protected(sock->wq, 1)->fasync_list)
579 pr_err("%s: fasync list not empty!\n", __func__);
580
581 this_cpu_sub(sockets_in_use, 1);
582 if (!sock->file) {
583 iput(SOCK_INODE(sock));
584 return;
585 }
586 sock->file = NULL;
587}
588EXPORT_SYMBOL(sock_release);
589
590void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
591{
592 u8 flags = *tx_flags;
593
594 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_HARDWARE)
595 flags |= SKBTX_HW_TSTAMP;
596
597 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SOFTWARE)
598 flags |= SKBTX_SW_TSTAMP;
599
600 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_SCHED)
601 flags |= SKBTX_SCHED_TSTAMP;
602
603 if (sk->sk_tsflags & SOF_TIMESTAMPING_TX_ACK)
604 flags |= SKBTX_ACK_TSTAMP;
605
606 *tx_flags = flags;
607}
608EXPORT_SYMBOL(__sock_tx_timestamp);
609
610static inline int sock_sendmsg_nosec(struct socket *sock, struct msghdr *msg)
611{
612 int ret = sock->ops->sendmsg(sock, msg, msg_data_left(msg));
613 BUG_ON(ret == -EIOCBQUEUED);
614 return ret;
615}
616
617int sock_sendmsg(struct socket *sock, struct msghdr *msg)
618{
619 int err = security_socket_sendmsg(sock, msg,
620 msg_data_left(msg));
621
622 return err ?: sock_sendmsg_nosec(sock, msg);
623}
624EXPORT_SYMBOL(sock_sendmsg);
625
626int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
627 struct kvec *vec, size_t num, size_t size)
628{
629 iov_iter_kvec(&msg->msg_iter, WRITE | ITER_KVEC, vec, num, size);
630 return sock_sendmsg(sock, msg);
631}
632EXPORT_SYMBOL(kernel_sendmsg);
633
634/*
635 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
636 */
637void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
638 struct sk_buff *skb)
639{
640 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
641 struct scm_timestamping tss;
642 int empty = 1;
643 struct skb_shared_hwtstamps *shhwtstamps =
644 skb_hwtstamps(skb);
645
646 /* Race occurred between timestamp enabling and packet
647 receiving. Fill in the current time for now. */
648 if (need_software_tstamp && skb->tstamp.tv64 == 0)
649 __net_timestamp(skb);
650
651 if (need_software_tstamp) {
652 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
653 struct timeval tv;
654 skb_get_timestamp(skb, &tv);
655 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
656 sizeof(tv), &tv);
657 } else {
658 struct timespec ts;
659 skb_get_timestampns(skb, &ts);
660 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
661 sizeof(ts), &ts);
662 }
663 }
664
665 memset(&tss, 0, sizeof(tss));
666 if ((sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) &&
667 ktime_to_timespec_cond(skb->tstamp, tss.ts + 0))
668 empty = 0;
669 if (shhwtstamps &&
670 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE) &&
671 ktime_to_timespec_cond(shhwtstamps->hwtstamp, tss.ts + 2))
672 empty = 0;
673 if (!empty)
674 put_cmsg(msg, SOL_SOCKET,
675 SCM_TIMESTAMPING, sizeof(tss), &tss);
676}
677EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
678
679void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
680 struct sk_buff *skb)
681{
682 int ack;
683
684 if (!sock_flag(sk, SOCK_WIFI_STATUS))
685 return;
686 if (!skb->wifi_acked_valid)
687 return;
688
689 ack = skb->wifi_acked;
690
691 put_cmsg(msg, SOL_SOCKET, SCM_WIFI_STATUS, sizeof(ack), &ack);
692}
693EXPORT_SYMBOL_GPL(__sock_recv_wifi_status);
694
695static inline void sock_recv_drops(struct msghdr *msg, struct sock *sk,
696 struct sk_buff *skb)
697{
698 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && SOCK_SKB_CB(skb)->dropcount)
699 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
700 sizeof(__u32), &SOCK_SKB_CB(skb)->dropcount);
701}
702
703void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
704 struct sk_buff *skb)
705{
706 sock_recv_timestamp(msg, sk, skb);
707 sock_recv_drops(msg, sk, skb);
708}
709EXPORT_SYMBOL_GPL(__sock_recv_ts_and_drops);
710
711static inline int sock_recvmsg_nosec(struct socket *sock, struct msghdr *msg,
712 size_t size, int flags)
713{
714 return sock->ops->recvmsg(sock, msg, size, flags);
715}
716
717int sock_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
718 int flags)
719{
720 int err = security_socket_recvmsg(sock, msg, size, flags);
721
722 return err ?: sock_recvmsg_nosec(sock, msg, size, flags);
723}
724EXPORT_SYMBOL(sock_recvmsg);
725
726/**
727 * kernel_recvmsg - Receive a message from a socket (kernel space)
728 * @sock: The socket to receive the message from
729 * @msg: Received message
730 * @vec: Input s/g array for message data
731 * @num: Size of input s/g array
732 * @size: Number of bytes to read
733 * @flags: Message flags (MSG_DONTWAIT, etc...)
734 *
735 * On return the msg structure contains the scatter/gather array passed in the
736 * vec argument. The array is modified so that it consists of the unfilled
737 * portion of the original array.
738 *
739 * The returned value is the total number of bytes received, or an error.
740 */
741int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
742 struct kvec *vec, size_t num, size_t size, int flags)
743{
744 mm_segment_t oldfs = get_fs();
745 int result;
746
747 iov_iter_kvec(&msg->msg_iter, READ | ITER_KVEC, vec, num, size);
748 set_fs(KERNEL_DS);
749 result = sock_recvmsg(sock, msg, size, flags);
750 set_fs(oldfs);
751 return result;
752}
753EXPORT_SYMBOL(kernel_recvmsg);
754
755static ssize_t sock_sendpage(struct file *file, struct page *page,
756 int offset, size_t size, loff_t *ppos, int more)
757{
758 struct socket *sock;
759 int flags;
760
761 sock = file->private_data;
762
763 flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
764 /* more is a combination of MSG_MORE and MSG_SENDPAGE_NOTLAST */
765 flags |= more;
766
767 return kernel_sendpage(sock, page, offset, size, flags);
768}
769
770static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
771 struct pipe_inode_info *pipe, size_t len,
772 unsigned int flags)
773{
774 struct socket *sock = file->private_data;
775
776 if (unlikely(!sock->ops->splice_read))
777 return -EINVAL;
778
779 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
780}
781
782static ssize_t sock_read_iter(struct kiocb *iocb, struct iov_iter *to)
783{
784 struct file *file = iocb->ki_filp;
785 struct socket *sock = file->private_data;
786 struct msghdr msg = {.msg_iter = *to,
787 .msg_iocb = iocb};
788 ssize_t res;
789
790 if (file->f_flags & O_NONBLOCK)
791 msg.msg_flags = MSG_DONTWAIT;
792
793 if (iocb->ki_pos != 0)
794 return -ESPIPE;
795
796 if (!iov_iter_count(to)) /* Match SYS5 behaviour */
797 return 0;
798
799 res = sock_recvmsg(sock, &msg, iov_iter_count(to), msg.msg_flags);
800 *to = msg.msg_iter;
801 return res;
802}
803
804static ssize_t sock_write_iter(struct kiocb *iocb, struct iov_iter *from)
805{
806 struct file *file = iocb->ki_filp;
807 struct socket *sock = file->private_data;
808 struct msghdr msg = {.msg_iter = *from,
809 .msg_iocb = iocb};
810 ssize_t res;
811
812 if (iocb->ki_pos != 0)
813 return -ESPIPE;
814
815 if (file->f_flags & O_NONBLOCK)
816 msg.msg_flags = MSG_DONTWAIT;
817
818 if (sock->type == SOCK_SEQPACKET)
819 msg.msg_flags |= MSG_EOR;
820
821 res = sock_sendmsg(sock, &msg);
822 *from = msg.msg_iter;
823 return res;
824}
825
826/*
827 * Atomic setting of ioctl hooks to avoid race
828 * with module unload.
829 */
830
831static DEFINE_MUTEX(br_ioctl_mutex);
832static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg);
833
834void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
835{
836 mutex_lock(&br_ioctl_mutex);
837 br_ioctl_hook = hook;
838 mutex_unlock(&br_ioctl_mutex);
839}
840EXPORT_SYMBOL(brioctl_set);
841
842static DEFINE_MUTEX(vlan_ioctl_mutex);
843static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
844
845void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
846{
847 mutex_lock(&vlan_ioctl_mutex);
848 vlan_ioctl_hook = hook;
849 mutex_unlock(&vlan_ioctl_mutex);
850}
851EXPORT_SYMBOL(vlan_ioctl_set);
852
853static DEFINE_MUTEX(dlci_ioctl_mutex);
854static int (*dlci_ioctl_hook) (unsigned int, void __user *);
855
856void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
857{
858 mutex_lock(&dlci_ioctl_mutex);
859 dlci_ioctl_hook = hook;
860 mutex_unlock(&dlci_ioctl_mutex);
861}
862EXPORT_SYMBOL(dlci_ioctl_set);
863
864static long sock_do_ioctl(struct net *net, struct socket *sock,
865 unsigned int cmd, unsigned long arg)
866{
867 int err;
868 void __user *argp = (void __user *)arg;
869
870 err = sock->ops->ioctl(sock, cmd, arg);
871
872 /*
873 * If this ioctl is unknown try to hand it down
874 * to the NIC driver.
875 */
876 if (err == -ENOIOCTLCMD)
877 err = dev_ioctl(net, cmd, argp);
878
879 return err;
880}
881
882/*
883 * With an ioctl, arg may well be a user mode pointer, but we don't know
884 * what to do with it - that's up to the protocol still.
885 */
886
887static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
888{
889 struct socket *sock;
890 struct sock *sk;
891 void __user *argp = (void __user *)arg;
892 int pid, err;
893 struct net *net;
894
895 sock = file->private_data;
896 sk = sock->sk;
897 net = sock_net(sk);
898 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
899 err = dev_ioctl(net, cmd, argp);
900 } else
901#ifdef CONFIG_WEXT_CORE
902 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
903 err = dev_ioctl(net, cmd, argp);
904 } else
905#endif
906 switch (cmd) {
907 case FIOSETOWN:
908 case SIOCSPGRP:
909 err = -EFAULT;
910 if (get_user(pid, (int __user *)argp))
911 break;
912 f_setown(sock->file, pid, 1);
913 err = 0;
914 break;
915 case FIOGETOWN:
916 case SIOCGPGRP:
917 err = put_user(f_getown(sock->file),
918 (int __user *)argp);
919 break;
920 case SIOCGIFBR:
921 case SIOCSIFBR:
922 case SIOCBRADDBR:
923 case SIOCBRDELBR:
924 err = -ENOPKG;
925 if (!br_ioctl_hook)
926 request_module("bridge");
927
928 mutex_lock(&br_ioctl_mutex);
929 if (br_ioctl_hook)
930 err = br_ioctl_hook(net, cmd, argp);
931 mutex_unlock(&br_ioctl_mutex);
932 break;
933 case SIOCGIFVLAN:
934 case SIOCSIFVLAN:
935 err = -ENOPKG;
936 if (!vlan_ioctl_hook)
937 request_module("8021q");
938
939 mutex_lock(&vlan_ioctl_mutex);
940 if (vlan_ioctl_hook)
941 err = vlan_ioctl_hook(net, argp);
942 mutex_unlock(&vlan_ioctl_mutex);
943 break;
944 case SIOCADDDLCI:
945 case SIOCDELDLCI:
946 err = -ENOPKG;
947 if (!dlci_ioctl_hook)
948 request_module("dlci");
949
950 mutex_lock(&dlci_ioctl_mutex);
951 if (dlci_ioctl_hook)
952 err = dlci_ioctl_hook(cmd, argp);
953 mutex_unlock(&dlci_ioctl_mutex);
954 break;
955 default:
956 err = sock_do_ioctl(net, sock, cmd, arg);
957 break;
958 }
959 return err;
960}
961
962int sock_create_lite(int family, int type, int protocol, struct socket **res)
963{
964 int err;
965 struct socket *sock = NULL;
966
967 err = security_socket_create(family, type, protocol, 1);
968 if (err)
969 goto out;
970
971 sock = sock_alloc();
972 if (!sock) {
973 err = -ENOMEM;
974 goto out;
975 }
976
977 sock->type = type;
978 err = security_socket_post_create(sock, family, type, protocol, 1);
979 if (err)
980 goto out_release;
981
982out:
983 *res = sock;
984 return err;
985out_release:
986 sock_release(sock);
987 sock = NULL;
988 goto out;
989}
990EXPORT_SYMBOL(sock_create_lite);
991
992/* No kernel lock held - perfect */
993static unsigned int sock_poll(struct file *file, poll_table *wait)
994{
995 unsigned int busy_flag = 0;
996 struct socket *sock;
997
998 /*
999 * We can't return errors to poll, so it's either yes or no.
1000 */
1001 sock = file->private_data;
1002
1003 if (sk_can_busy_loop(sock->sk)) {
1004 /* this socket can poll_ll so tell the system call */
1005 busy_flag = POLL_BUSY_LOOP;
1006
1007 /* once, only if requested by syscall */
1008 if (wait && (wait->_key & POLL_BUSY_LOOP))
1009 sk_busy_loop(sock->sk, 1);
1010 }
1011
1012 return busy_flag | sock->ops->poll(file, sock, wait);
1013}
1014
1015static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1016{
1017 struct socket *sock = file->private_data;
1018
1019 return sock->ops->mmap(file, sock, vma);
1020}
1021
1022static int sock_close(struct inode *inode, struct file *filp)
1023{
1024 sock_release(SOCKET_I(inode));
1025 return 0;
1026}
1027
1028/*
1029 * Update the socket async list
1030 *
1031 * Fasync_list locking strategy.
1032 *
1033 * 1. fasync_list is modified only under process context socket lock
1034 * i.e. under semaphore.
1035 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1036 * or under socket lock
1037 */
1038
1039static int sock_fasync(int fd, struct file *filp, int on)
1040{
1041 struct socket *sock = filp->private_data;
1042 struct sock *sk = sock->sk;
1043 struct socket_wq *wq;
1044
1045 if (sk == NULL)
1046 return -EINVAL;
1047
1048 lock_sock(sk);
1049 wq = rcu_dereference_protected(sock->wq, sock_owned_by_user(sk));
1050 fasync_helper(fd, filp, on, &wq->fasync_list);
1051
1052 if (!wq->fasync_list)
1053 sock_reset_flag(sk, SOCK_FASYNC);
1054 else
1055 sock_set_flag(sk, SOCK_FASYNC);
1056
1057 release_sock(sk);
1058 return 0;
1059}
1060
1061/* This function may be called only under rcu_lock */
1062
1063int sock_wake_async(struct socket_wq *wq, int how, int band)
1064{
1065 if (!wq || !wq->fasync_list)
1066 return -1;
1067
1068 switch (how) {
1069 case SOCK_WAKE_WAITD:
1070 if (test_bit(SOCKWQ_ASYNC_WAITDATA, &wq->flags))
1071 break;
1072 goto call_kill;
1073 case SOCK_WAKE_SPACE:
1074 if (!test_and_clear_bit(SOCKWQ_ASYNC_NOSPACE, &wq->flags))
1075 break;
1076 /* fall through */
1077 case SOCK_WAKE_IO:
1078call_kill:
1079 kill_fasync(&wq->fasync_list, SIGIO, band);
1080 break;
1081 case SOCK_WAKE_URG:
1082 kill_fasync(&wq->fasync_list, SIGURG, band);
1083 }
1084
1085 return 0;
1086}
1087EXPORT_SYMBOL(sock_wake_async);
1088
1089int __sock_create(struct net *net, int family, int type, int protocol,
1090 struct socket **res, int kern)
1091{
1092 int err;
1093 struct socket *sock;
1094 const struct net_proto_family *pf;
1095
1096 /*
1097 * Check protocol is in range
1098 */
1099 if (family < 0 || family >= NPROTO)
1100 return -EAFNOSUPPORT;
1101 if (type < 0 || type >= SOCK_MAX)
1102 return -EINVAL;
1103
1104 /* Compatibility.
1105
1106 This uglymoron is moved from INET layer to here to avoid
1107 deadlock in module load.
1108 */
1109 if (family == PF_INET && type == SOCK_PACKET) {
1110 pr_info_once("%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1111 current->comm);
1112 family = PF_PACKET;
1113 }
1114
1115 err = security_socket_create(family, type, protocol, kern);
1116 if (err)
1117 return err;
1118
1119 /*
1120 * Allocate the socket and allow the family to set things up. if
1121 * the protocol is 0, the family is instructed to select an appropriate
1122 * default.
1123 */
1124 sock = sock_alloc();
1125 if (!sock) {
1126 net_warn_ratelimited("socket: no more sockets\n");
1127 return -ENFILE; /* Not exactly a match, but its the
1128 closest posix thing */
1129 }
1130
1131 sock->type = type;
1132
1133#ifdef CONFIG_MODULES
1134 /* Attempt to load a protocol module if the find failed.
1135 *
1136 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1137 * requested real, full-featured networking support upon configuration.
1138 * Otherwise module support will break!
1139 */
1140 if (rcu_access_pointer(net_families[family]) == NULL)
1141 request_module("net-pf-%d", family);
1142#endif
1143
1144 rcu_read_lock();
1145 pf = rcu_dereference(net_families[family]);
1146 err = -EAFNOSUPPORT;
1147 if (!pf)
1148 goto out_release;
1149
1150 /*
1151 * We will call the ->create function, that possibly is in a loadable
1152 * module, so we have to bump that loadable module refcnt first.
1153 */
1154 if (!try_module_get(pf->owner))
1155 goto out_release;
1156
1157 /* Now protected by module ref count */
1158 rcu_read_unlock();
1159
1160 err = pf->create(net, sock, protocol, kern);
1161 if (err < 0)
1162 goto out_module_put;
1163
1164 /*
1165 * Now to bump the refcnt of the [loadable] module that owns this
1166 * socket at sock_release time we decrement its refcnt.
1167 */
1168 if (!try_module_get(sock->ops->owner))
1169 goto out_module_busy;
1170
1171 /*
1172 * Now that we're done with the ->create function, the [loadable]
1173 * module can have its refcnt decremented
1174 */
1175 module_put(pf->owner);
1176 err = security_socket_post_create(sock, family, type, protocol, kern);
1177 if (err)
1178 goto out_sock_release;
1179 *res = sock;
1180
1181 return 0;
1182
1183out_module_busy:
1184 err = -EAFNOSUPPORT;
1185out_module_put:
1186 sock->ops = NULL;
1187 module_put(pf->owner);
1188out_sock_release:
1189 sock_release(sock);
1190 return err;
1191
1192out_release:
1193 rcu_read_unlock();
1194 goto out_sock_release;
1195}
1196EXPORT_SYMBOL(__sock_create);
1197
1198int sock_create(int family, int type, int protocol, struct socket **res)
1199{
1200 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1201}
1202EXPORT_SYMBOL(sock_create);
1203
1204int sock_create_kern(struct net *net, int family, int type, int protocol, struct socket **res)
1205{
1206 return __sock_create(net, family, type, protocol, res, 1);
1207}
1208EXPORT_SYMBOL(sock_create_kern);
1209
1210SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1211{
1212 int retval;
1213 struct socket *sock;
1214 int flags;
1215
1216 /* Check the SOCK_* constants for consistency. */
1217 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1218 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1219 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1220 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1221
1222 flags = type & ~SOCK_TYPE_MASK;
1223 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1224 return -EINVAL;
1225 type &= SOCK_TYPE_MASK;
1226
1227 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1228 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1229
1230 retval = sock_create(family, type, protocol, &sock);
1231 if (retval < 0)
1232 goto out;
1233
1234 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1235 if (retval < 0)
1236 goto out_release;
1237
1238out:
1239 /* It may be already another descriptor 8) Not kernel problem. */
1240 return retval;
1241
1242out_release:
1243 sock_release(sock);
1244 return retval;
1245}
1246
1247/*
1248 * Create a pair of connected sockets.
1249 */
1250
1251SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1252 int __user *, usockvec)
1253{
1254 struct socket *sock1, *sock2;
1255 int fd1, fd2, err;
1256 struct file *newfile1, *newfile2;
1257 int flags;
1258
1259 flags = type & ~SOCK_TYPE_MASK;
1260 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1261 return -EINVAL;
1262 type &= SOCK_TYPE_MASK;
1263
1264 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1265 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1266
1267 /*
1268 * Obtain the first socket and check if the underlying protocol
1269 * supports the socketpair call.
1270 */
1271
1272 err = sock_create(family, type, protocol, &sock1);
1273 if (err < 0)
1274 goto out;
1275
1276 err = sock_create(family, type, protocol, &sock2);
1277 if (err < 0)
1278 goto out_release_1;
1279
1280 err = sock1->ops->socketpair(sock1, sock2);
1281 if (err < 0)
1282 goto out_release_both;
1283
1284 fd1 = get_unused_fd_flags(flags);
1285 if (unlikely(fd1 < 0)) {
1286 err = fd1;
1287 goto out_release_both;
1288 }
1289
1290 fd2 = get_unused_fd_flags(flags);
1291 if (unlikely(fd2 < 0)) {
1292 err = fd2;
1293 goto out_put_unused_1;
1294 }
1295
1296 newfile1 = sock_alloc_file(sock1, flags, NULL);
1297 if (IS_ERR(newfile1)) {
1298 err = PTR_ERR(newfile1);
1299 goto out_put_unused_both;
1300 }
1301
1302 newfile2 = sock_alloc_file(sock2, flags, NULL);
1303 if (IS_ERR(newfile2)) {
1304 err = PTR_ERR(newfile2);
1305 goto out_fput_1;
1306 }
1307
1308 err = put_user(fd1, &usockvec[0]);
1309 if (err)
1310 goto out_fput_both;
1311
1312 err = put_user(fd2, &usockvec[1]);
1313 if (err)
1314 goto out_fput_both;
1315
1316 audit_fd_pair(fd1, fd2);
1317
1318 fd_install(fd1, newfile1);
1319 fd_install(fd2, newfile2);
1320 /* fd1 and fd2 may be already another descriptors.
1321 * Not kernel problem.
1322 */
1323
1324 return 0;
1325
1326out_fput_both:
1327 fput(newfile2);
1328 fput(newfile1);
1329 put_unused_fd(fd2);
1330 put_unused_fd(fd1);
1331 goto out;
1332
1333out_fput_1:
1334 fput(newfile1);
1335 put_unused_fd(fd2);
1336 put_unused_fd(fd1);
1337 sock_release(sock2);
1338 goto out;
1339
1340out_put_unused_both:
1341 put_unused_fd(fd2);
1342out_put_unused_1:
1343 put_unused_fd(fd1);
1344out_release_both:
1345 sock_release(sock2);
1346out_release_1:
1347 sock_release(sock1);
1348out:
1349 return err;
1350}
1351
1352/*
1353 * Bind a name to a socket. Nothing much to do here since it's
1354 * the protocol's responsibility to handle the local address.
1355 *
1356 * We move the socket address to kernel space before we call
1357 * the protocol layer (having also checked the address is ok).
1358 */
1359
1360SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1361{
1362 struct socket *sock;
1363 struct sockaddr_storage address;
1364 int err, fput_needed;
1365
1366 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1367 if (sock) {
1368 err = move_addr_to_kernel(umyaddr, addrlen, &address);
1369 if (err >= 0) {
1370 err = security_socket_bind(sock,
1371 (struct sockaddr *)&address,
1372 addrlen);
1373 if (!err)
1374 err = sock->ops->bind(sock,
1375 (struct sockaddr *)
1376 &address, addrlen);
1377 }
1378 fput_light(sock->file, fput_needed);
1379 }
1380 return err;
1381}
1382
1383/*
1384 * Perform a listen. Basically, we allow the protocol to do anything
1385 * necessary for a listen, and if that works, we mark the socket as
1386 * ready for listening.
1387 */
1388
1389SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1390{
1391 struct socket *sock;
1392 int err, fput_needed;
1393 int somaxconn;
1394
1395 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1396 if (sock) {
1397 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1398 if ((unsigned int)backlog > somaxconn)
1399 backlog = somaxconn;
1400
1401 err = security_socket_listen(sock, backlog);
1402 if (!err)
1403 err = sock->ops->listen(sock, backlog);
1404
1405 fput_light(sock->file, fput_needed);
1406 }
1407 return err;
1408}
1409
1410/*
1411 * For accept, we attempt to create a new socket, set up the link
1412 * with the client, wake up the client, then return the new
1413 * connected fd. We collect the address of the connector in kernel
1414 * space and move it to user at the very end. This is unclean because
1415 * we open the socket then return an error.
1416 *
1417 * 1003.1g adds the ability to recvmsg() to query connection pending
1418 * status to recvmsg. We need to add that support in a way thats
1419 * clean when we restucture accept also.
1420 */
1421
1422SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1423 int __user *, upeer_addrlen, int, flags)
1424{
1425 struct socket *sock, *newsock;
1426 struct file *newfile;
1427 int err, len, newfd, fput_needed;
1428 struct sockaddr_storage address;
1429
1430 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1431 return -EINVAL;
1432
1433 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1434 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1435
1436 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1437 if (!sock)
1438 goto out;
1439
1440 err = -ENFILE;
1441 newsock = sock_alloc();
1442 if (!newsock)
1443 goto out_put;
1444
1445 newsock->type = sock->type;
1446 newsock->ops = sock->ops;
1447
1448 /*
1449 * We don't need try_module_get here, as the listening socket (sock)
1450 * has the protocol module (sock->ops->owner) held.
1451 */
1452 __module_get(newsock->ops->owner);
1453
1454 newfd = get_unused_fd_flags(flags);
1455 if (unlikely(newfd < 0)) {
1456 err = newfd;
1457 sock_release(newsock);
1458 goto out_put;
1459 }
1460 newfile = sock_alloc_file(newsock, flags, sock->sk->sk_prot_creator->name);
1461 if (IS_ERR(newfile)) {
1462 err = PTR_ERR(newfile);
1463 put_unused_fd(newfd);
1464 sock_release(newsock);
1465 goto out_put;
1466 }
1467
1468 err = security_socket_accept(sock, newsock);
1469 if (err)
1470 goto out_fd;
1471
1472 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1473 if (err < 0)
1474 goto out_fd;
1475
1476 if (upeer_sockaddr) {
1477 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1478 &len, 2) < 0) {
1479 err = -ECONNABORTED;
1480 goto out_fd;
1481 }
1482 err = move_addr_to_user(&address,
1483 len, upeer_sockaddr, upeer_addrlen);
1484 if (err < 0)
1485 goto out_fd;
1486 }
1487
1488 /* File flags are not inherited via accept() unlike another OSes. */
1489
1490 fd_install(newfd, newfile);
1491 err = newfd;
1492
1493out_put:
1494 fput_light(sock->file, fput_needed);
1495out:
1496 return err;
1497out_fd:
1498 fput(newfile);
1499 put_unused_fd(newfd);
1500 goto out_put;
1501}
1502
1503SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1504 int __user *, upeer_addrlen)
1505{
1506 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1507}
1508
1509/*
1510 * Attempt to connect to a socket with the server address. The address
1511 * is in user space so we verify it is OK and move it to kernel space.
1512 *
1513 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1514 * break bindings
1515 *
1516 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1517 * other SEQPACKET protocols that take time to connect() as it doesn't
1518 * include the -EINPROGRESS status for such sockets.
1519 */
1520
1521SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1522 int, addrlen)
1523{
1524 struct socket *sock;
1525 struct sockaddr_storage address;
1526 int err, fput_needed;
1527
1528 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1529 if (!sock)
1530 goto out;
1531 err = move_addr_to_kernel(uservaddr, addrlen, &address);
1532 if (err < 0)
1533 goto out_put;
1534
1535 err =
1536 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1537 if (err)
1538 goto out_put;
1539
1540 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1541 sock->file->f_flags);
1542out_put:
1543 fput_light(sock->file, fput_needed);
1544out:
1545 return err;
1546}
1547
1548/*
1549 * Get the local address ('name') of a socket object. Move the obtained
1550 * name to user space.
1551 */
1552
1553SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1554 int __user *, usockaddr_len)
1555{
1556 struct socket *sock;
1557 struct sockaddr_storage address;
1558 int len, err, fput_needed;
1559
1560 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1561 if (!sock)
1562 goto out;
1563
1564 err = security_socket_getsockname(sock);
1565 if (err)
1566 goto out_put;
1567
1568 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1569 if (err)
1570 goto out_put;
1571 err = move_addr_to_user(&address, len, usockaddr, usockaddr_len);
1572
1573out_put:
1574 fput_light(sock->file, fput_needed);
1575out:
1576 return err;
1577}
1578
1579/*
1580 * Get the remote address ('name') of a socket object. Move the obtained
1581 * name to user space.
1582 */
1583
1584SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1585 int __user *, usockaddr_len)
1586{
1587 struct socket *sock;
1588 struct sockaddr_storage address;
1589 int len, err, fput_needed;
1590
1591 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1592 if (sock != NULL) {
1593 err = security_socket_getpeername(sock);
1594 if (err) {
1595 fput_light(sock->file, fput_needed);
1596 return err;
1597 }
1598
1599 err =
1600 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1601 1);
1602 if (!err)
1603 err = move_addr_to_user(&address, len, usockaddr,
1604 usockaddr_len);
1605 fput_light(sock->file, fput_needed);
1606 }
1607 return err;
1608}
1609
1610/*
1611 * Send a datagram to a given address. We move the address into kernel
1612 * space and check the user space data area is readable before invoking
1613 * the protocol.
1614 */
1615
1616SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1617 unsigned int, flags, struct sockaddr __user *, addr,
1618 int, addr_len)
1619{
1620 struct socket *sock;
1621 struct sockaddr_storage address;
1622 int err;
1623 struct msghdr msg;
1624 struct iovec iov;
1625 int fput_needed;
1626
1627 err = import_single_range(WRITE, buff, len, &iov, &msg.msg_iter);
1628 if (unlikely(err))
1629 return err;
1630 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1631 if (!sock)
1632 goto out;
1633
1634 msg.msg_name = NULL;
1635 msg.msg_control = NULL;
1636 msg.msg_controllen = 0;
1637 msg.msg_namelen = 0;
1638 if (addr) {
1639 err = move_addr_to_kernel(addr, addr_len, &address);
1640 if (err < 0)
1641 goto out_put;
1642 msg.msg_name = (struct sockaddr *)&address;
1643 msg.msg_namelen = addr_len;
1644 }
1645 if (sock->file->f_flags & O_NONBLOCK)
1646 flags |= MSG_DONTWAIT;
1647 msg.msg_flags = flags;
1648 err = sock_sendmsg(sock, &msg);
1649
1650out_put:
1651 fput_light(sock->file, fput_needed);
1652out:
1653 return err;
1654}
1655
1656/*
1657 * Send a datagram down a socket.
1658 */
1659
1660SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1661 unsigned int, flags)
1662{
1663 return sys_sendto(fd, buff, len, flags, NULL, 0);
1664}
1665
1666/*
1667 * Receive a frame from the socket and optionally record the address of the
1668 * sender. We verify the buffers are writable and if needed move the
1669 * sender address from kernel to user space.
1670 */
1671
1672SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1673 unsigned int, flags, struct sockaddr __user *, addr,
1674 int __user *, addr_len)
1675{
1676 struct socket *sock;
1677 struct iovec iov;
1678 struct msghdr msg;
1679 struct sockaddr_storage address;
1680 int err, err2;
1681 int fput_needed;
1682
1683 err = import_single_range(READ, ubuf, size, &iov, &msg.msg_iter);
1684 if (unlikely(err))
1685 return err;
1686 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1687 if (!sock)
1688 goto out;
1689
1690 msg.msg_control = NULL;
1691 msg.msg_controllen = 0;
1692 /* Save some cycles and don't copy the address if not needed */
1693 msg.msg_name = addr ? (struct sockaddr *)&address : NULL;
1694 /* We assume all kernel code knows the size of sockaddr_storage */
1695 msg.msg_namelen = 0;
1696 msg.msg_iocb = NULL;
1697 if (sock->file->f_flags & O_NONBLOCK)
1698 flags |= MSG_DONTWAIT;
1699 err = sock_recvmsg(sock, &msg, iov_iter_count(&msg.msg_iter), flags);
1700
1701 if (err >= 0 && addr != NULL) {
1702 err2 = move_addr_to_user(&address,
1703 msg.msg_namelen, addr, addr_len);
1704 if (err2 < 0)
1705 err = err2;
1706 }
1707
1708 fput_light(sock->file, fput_needed);
1709out:
1710 return err;
1711}
1712
1713/*
1714 * Receive a datagram from a socket.
1715 */
1716
1717SYSCALL_DEFINE4(recv, int, fd, void __user *, ubuf, size_t, size,
1718 unsigned int, flags)
1719{
1720 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1721}
1722
1723/*
1724 * Set a socket option. Because we don't know the option lengths we have
1725 * to pass the user mode parameter for the protocols to sort out.
1726 */
1727
1728SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1729 char __user *, optval, int, optlen)
1730{
1731 int err, fput_needed;
1732 struct socket *sock;
1733
1734 if (optlen < 0)
1735 return -EINVAL;
1736
1737 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1738 if (sock != NULL) {
1739 err = security_socket_setsockopt(sock, level, optname);
1740 if (err)
1741 goto out_put;
1742
1743 if (level == SOL_SOCKET)
1744 err =
1745 sock_setsockopt(sock, level, optname, optval,
1746 optlen);
1747 else
1748 err =
1749 sock->ops->setsockopt(sock, level, optname, optval,
1750 optlen);
1751out_put:
1752 fput_light(sock->file, fput_needed);
1753 }
1754 return err;
1755}
1756
1757/*
1758 * Get a socket option. Because we don't know the option lengths we have
1759 * to pass a user mode parameter for the protocols to sort out.
1760 */
1761
1762SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1763 char __user *, optval, int __user *, optlen)
1764{
1765 int err, fput_needed;
1766 struct socket *sock;
1767
1768 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1769 if (sock != NULL) {
1770 err = security_socket_getsockopt(sock, level, optname);
1771 if (err)
1772 goto out_put;
1773
1774 if (level == SOL_SOCKET)
1775 err =
1776 sock_getsockopt(sock, level, optname, optval,
1777 optlen);
1778 else
1779 err =
1780 sock->ops->getsockopt(sock, level, optname, optval,
1781 optlen);
1782out_put:
1783 fput_light(sock->file, fput_needed);
1784 }
1785 return err;
1786}
1787
1788/*
1789 * Shutdown a socket.
1790 */
1791
1792SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1793{
1794 int err, fput_needed;
1795 struct socket *sock;
1796
1797 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1798 if (sock != NULL) {
1799 err = security_socket_shutdown(sock, how);
1800 if (!err)
1801 err = sock->ops->shutdown(sock, how);
1802 fput_light(sock->file, fput_needed);
1803 }
1804 return err;
1805}
1806
1807/* A couple of helpful macros for getting the address of the 32/64 bit
1808 * fields which are the same type (int / unsigned) on our platforms.
1809 */
1810#define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1811#define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1812#define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1813
1814struct used_address {
1815 struct sockaddr_storage name;
1816 unsigned int name_len;
1817};
1818
1819static int copy_msghdr_from_user(struct msghdr *kmsg,
1820 struct user_msghdr __user *umsg,
1821 struct sockaddr __user **save_addr,
1822 struct iovec **iov)
1823{
1824 struct sockaddr __user *uaddr;
1825 struct iovec __user *uiov;
1826 size_t nr_segs;
1827 ssize_t err;
1828
1829 if (!access_ok(VERIFY_READ, umsg, sizeof(*umsg)) ||
1830 __get_user(uaddr, &umsg->msg_name) ||
1831 __get_user(kmsg->msg_namelen, &umsg->msg_namelen) ||
1832 __get_user(uiov, &umsg->msg_iov) ||
1833 __get_user(nr_segs, &umsg->msg_iovlen) ||
1834 __get_user(kmsg->msg_control, &umsg->msg_control) ||
1835 __get_user(kmsg->msg_controllen, &umsg->msg_controllen) ||
1836 __get_user(kmsg->msg_flags, &umsg->msg_flags))
1837 return -EFAULT;
1838
1839 if (!uaddr)
1840 kmsg->msg_namelen = 0;
1841
1842 if (kmsg->msg_namelen < 0)
1843 return -EINVAL;
1844
1845 if (kmsg->msg_namelen > sizeof(struct sockaddr_storage))
1846 kmsg->msg_namelen = sizeof(struct sockaddr_storage);
1847
1848 if (save_addr)
1849 *save_addr = uaddr;
1850
1851 if (uaddr && kmsg->msg_namelen) {
1852 if (!save_addr) {
1853 err = move_addr_to_kernel(uaddr, kmsg->msg_namelen,
1854 kmsg->msg_name);
1855 if (err < 0)
1856 return err;
1857 }
1858 } else {
1859 kmsg->msg_name = NULL;
1860 kmsg->msg_namelen = 0;
1861 }
1862
1863 if (nr_segs > UIO_MAXIOV)
1864 return -EMSGSIZE;
1865
1866 kmsg->msg_iocb = NULL;
1867
1868 return import_iovec(save_addr ? READ : WRITE, uiov, nr_segs,
1869 UIO_FASTIOV, iov, &kmsg->msg_iter);
1870}
1871
1872static int ___sys_sendmsg(struct socket *sock, struct user_msghdr __user *msg,
1873 struct msghdr *msg_sys, unsigned int flags,
1874 struct used_address *used_address,
1875 unsigned int allowed_msghdr_flags)
1876{
1877 struct compat_msghdr __user *msg_compat =
1878 (struct compat_msghdr __user *)msg;
1879 struct sockaddr_storage address;
1880 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1881 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1882 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1883 /* 20 is size of ipv6_pktinfo */
1884 unsigned char *ctl_buf = ctl;
1885 int ctl_len;
1886 ssize_t err;
1887
1888 msg_sys->msg_name = &address;
1889
1890 if (MSG_CMSG_COMPAT & flags)
1891 err = get_compat_msghdr(msg_sys, msg_compat, NULL, &iov);
1892 else
1893 err = copy_msghdr_from_user(msg_sys, msg, NULL, &iov);
1894 if (err < 0)
1895 return err;
1896
1897 err = -ENOBUFS;
1898
1899 if (msg_sys->msg_controllen > INT_MAX)
1900 goto out_freeiov;
1901 flags |= (msg_sys->msg_flags & allowed_msghdr_flags);
1902 ctl_len = msg_sys->msg_controllen;
1903 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1904 err =
1905 cmsghdr_from_user_compat_to_kern(msg_sys, sock->sk, ctl,
1906 sizeof(ctl));
1907 if (err)
1908 goto out_freeiov;
1909 ctl_buf = msg_sys->msg_control;
1910 ctl_len = msg_sys->msg_controllen;
1911 } else if (ctl_len) {
1912 if (ctl_len > sizeof(ctl)) {
1913 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1914 if (ctl_buf == NULL)
1915 goto out_freeiov;
1916 }
1917 err = -EFAULT;
1918 /*
1919 * Careful! Before this, msg_sys->msg_control contains a user pointer.
1920 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1921 * checking falls down on this.
1922 */
1923 if (copy_from_user(ctl_buf,
1924 (void __user __force *)msg_sys->msg_control,
1925 ctl_len))
1926 goto out_freectl;
1927 msg_sys->msg_control = ctl_buf;
1928 }
1929 msg_sys->msg_flags = flags;
1930
1931 if (sock->file->f_flags & O_NONBLOCK)
1932 msg_sys->msg_flags |= MSG_DONTWAIT;
1933 /*
1934 * If this is sendmmsg() and current destination address is same as
1935 * previously succeeded address, omit asking LSM's decision.
1936 * used_address->name_len is initialized to UINT_MAX so that the first
1937 * destination address never matches.
1938 */
1939 if (used_address && msg_sys->msg_name &&
1940 used_address->name_len == msg_sys->msg_namelen &&
1941 !memcmp(&used_address->name, msg_sys->msg_name,
1942 used_address->name_len)) {
1943 err = sock_sendmsg_nosec(sock, msg_sys);
1944 goto out_freectl;
1945 }
1946 err = sock_sendmsg(sock, msg_sys);
1947 /*
1948 * If this is sendmmsg() and sending to current destination address was
1949 * successful, remember it.
1950 */
1951 if (used_address && err >= 0) {
1952 used_address->name_len = msg_sys->msg_namelen;
1953 if (msg_sys->msg_name)
1954 memcpy(&used_address->name, msg_sys->msg_name,
1955 used_address->name_len);
1956 }
1957
1958out_freectl:
1959 if (ctl_buf != ctl)
1960 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1961out_freeiov:
1962 kfree(iov);
1963 return err;
1964}
1965
1966/*
1967 * BSD sendmsg interface
1968 */
1969
1970long __sys_sendmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
1971{
1972 int fput_needed, err;
1973 struct msghdr msg_sys;
1974 struct socket *sock;
1975
1976 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1977 if (!sock)
1978 goto out;
1979
1980 err = ___sys_sendmsg(sock, msg, &msg_sys, flags, NULL, 0);
1981
1982 fput_light(sock->file, fput_needed);
1983out:
1984 return err;
1985}
1986
1987SYSCALL_DEFINE3(sendmsg, int, fd, struct user_msghdr __user *, msg, unsigned int, flags)
1988{
1989 if (flags & MSG_CMSG_COMPAT)
1990 return -EINVAL;
1991 return __sys_sendmsg(fd, msg, flags);
1992}
1993
1994/*
1995 * Linux sendmmsg interface
1996 */
1997
1998int __sys_sendmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
1999 unsigned int flags)
2000{
2001 int fput_needed, err, datagrams;
2002 struct socket *sock;
2003 struct mmsghdr __user *entry;
2004 struct compat_mmsghdr __user *compat_entry;
2005 struct msghdr msg_sys;
2006 struct used_address used_address;
2007 unsigned int oflags = flags;
2008
2009 if (vlen > UIO_MAXIOV)
2010 vlen = UIO_MAXIOV;
2011
2012 datagrams = 0;
2013
2014 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2015 if (!sock)
2016 return err;
2017
2018 used_address.name_len = UINT_MAX;
2019 entry = mmsg;
2020 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2021 err = 0;
2022 flags |= MSG_BATCH;
2023
2024 while (datagrams < vlen) {
2025 if (datagrams == vlen - 1)
2026 flags = oflags;
2027
2028 if (MSG_CMSG_COMPAT & flags) {
2029 err = ___sys_sendmsg(sock, (struct user_msghdr __user *)compat_entry,
2030 &msg_sys, flags, &used_address, MSG_EOR);
2031 if (err < 0)
2032 break;
2033 err = __put_user(err, &compat_entry->msg_len);
2034 ++compat_entry;
2035 } else {
2036 err = ___sys_sendmsg(sock,
2037 (struct user_msghdr __user *)entry,
2038 &msg_sys, flags, &used_address, MSG_EOR);
2039 if (err < 0)
2040 break;
2041 err = put_user(err, &entry->msg_len);
2042 ++entry;
2043 }
2044
2045 if (err)
2046 break;
2047 ++datagrams;
2048 cond_resched();
2049 }
2050
2051 fput_light(sock->file, fput_needed);
2052
2053 /* We only return an error if no datagrams were able to be sent */
2054 if (datagrams != 0)
2055 return datagrams;
2056
2057 return err;
2058}
2059
2060SYSCALL_DEFINE4(sendmmsg, int, fd, struct mmsghdr __user *, mmsg,
2061 unsigned int, vlen, unsigned int, flags)
2062{
2063 if (flags & MSG_CMSG_COMPAT)
2064 return -EINVAL;
2065 return __sys_sendmmsg(fd, mmsg, vlen, flags);
2066}
2067
2068static int ___sys_recvmsg(struct socket *sock, struct user_msghdr __user *msg,
2069 struct msghdr *msg_sys, unsigned int flags, int nosec)
2070{
2071 struct compat_msghdr __user *msg_compat =
2072 (struct compat_msghdr __user *)msg;
2073 struct iovec iovstack[UIO_FASTIOV];
2074 struct iovec *iov = iovstack;
2075 unsigned long cmsg_ptr;
2076 int total_len, len;
2077 ssize_t err;
2078
2079 /* kernel mode address */
2080 struct sockaddr_storage addr;
2081
2082 /* user mode address pointers */
2083 struct sockaddr __user *uaddr;
2084 int __user *uaddr_len = COMPAT_NAMELEN(msg);
2085
2086 msg_sys->msg_name = &addr;
2087
2088 if (MSG_CMSG_COMPAT & flags)
2089 err = get_compat_msghdr(msg_sys, msg_compat, &uaddr, &iov);
2090 else
2091 err = copy_msghdr_from_user(msg_sys, msg, &uaddr, &iov);
2092 if (err < 0)
2093 return err;
2094 total_len = iov_iter_count(&msg_sys->msg_iter);
2095
2096 cmsg_ptr = (unsigned long)msg_sys->msg_control;
2097 msg_sys->msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2098
2099 /* We assume all kernel code knows the size of sockaddr_storage */
2100 msg_sys->msg_namelen = 0;
2101
2102 if (sock->file->f_flags & O_NONBLOCK)
2103 flags |= MSG_DONTWAIT;
2104 err = (nosec ? sock_recvmsg_nosec : sock_recvmsg)(sock, msg_sys,
2105 total_len, flags);
2106 if (err < 0)
2107 goto out_freeiov;
2108 len = err;
2109
2110 if (uaddr != NULL) {
2111 err = move_addr_to_user(&addr,
2112 msg_sys->msg_namelen, uaddr,
2113 uaddr_len);
2114 if (err < 0)
2115 goto out_freeiov;
2116 }
2117 err = __put_user((msg_sys->msg_flags & ~MSG_CMSG_COMPAT),
2118 COMPAT_FLAGS(msg));
2119 if (err)
2120 goto out_freeiov;
2121 if (MSG_CMSG_COMPAT & flags)
2122 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2123 &msg_compat->msg_controllen);
2124 else
2125 err = __put_user((unsigned long)msg_sys->msg_control - cmsg_ptr,
2126 &msg->msg_controllen);
2127 if (err)
2128 goto out_freeiov;
2129 err = len;
2130
2131out_freeiov:
2132 kfree(iov);
2133 return err;
2134}
2135
2136/*
2137 * BSD recvmsg interface
2138 */
2139
2140long __sys_recvmsg(int fd, struct user_msghdr __user *msg, unsigned flags)
2141{
2142 int fput_needed, err;
2143 struct msghdr msg_sys;
2144 struct socket *sock;
2145
2146 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2147 if (!sock)
2148 goto out;
2149
2150 err = ___sys_recvmsg(sock, msg, &msg_sys, flags, 0);
2151
2152 fput_light(sock->file, fput_needed);
2153out:
2154 return err;
2155}
2156
2157SYSCALL_DEFINE3(recvmsg, int, fd, struct user_msghdr __user *, msg,
2158 unsigned int, flags)
2159{
2160 if (flags & MSG_CMSG_COMPAT)
2161 return -EINVAL;
2162 return __sys_recvmsg(fd, msg, flags);
2163}
2164
2165/*
2166 * Linux recvmmsg interface
2167 */
2168
2169int __sys_recvmmsg(int fd, struct mmsghdr __user *mmsg, unsigned int vlen,
2170 unsigned int flags, struct timespec *timeout)
2171{
2172 int fput_needed, err, datagrams;
2173 struct socket *sock;
2174 struct mmsghdr __user *entry;
2175 struct compat_mmsghdr __user *compat_entry;
2176 struct msghdr msg_sys;
2177 struct timespec end_time;
2178
2179 if (timeout &&
2180 poll_select_set_timeout(&end_time, timeout->tv_sec,
2181 timeout->tv_nsec))
2182 return -EINVAL;
2183
2184 datagrams = 0;
2185
2186 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2187 if (!sock)
2188 return err;
2189
2190 err = sock_error(sock->sk);
2191 if (err)
2192 goto out_put;
2193
2194 entry = mmsg;
2195 compat_entry = (struct compat_mmsghdr __user *)mmsg;
2196
2197 while (datagrams < vlen) {
2198 /*
2199 * No need to ask LSM for more than the first datagram.
2200 */
2201 if (MSG_CMSG_COMPAT & flags) {
2202 err = ___sys_recvmsg(sock, (struct user_msghdr __user *)compat_entry,
2203 &msg_sys, flags & ~MSG_WAITFORONE,
2204 datagrams);
2205 if (err < 0)
2206 break;
2207 err = __put_user(err, &compat_entry->msg_len);
2208 ++compat_entry;
2209 } else {
2210 err = ___sys_recvmsg(sock,
2211 (struct user_msghdr __user *)entry,
2212 &msg_sys, flags & ~MSG_WAITFORONE,
2213 datagrams);
2214 if (err < 0)
2215 break;
2216 err = put_user(err, &entry->msg_len);
2217 ++entry;
2218 }
2219
2220 if (err)
2221 break;
2222 ++datagrams;
2223
2224 /* MSG_WAITFORONE turns on MSG_DONTWAIT after one packet */
2225 if (flags & MSG_WAITFORONE)
2226 flags |= MSG_DONTWAIT;
2227
2228 if (timeout) {
2229 ktime_get_ts(timeout);
2230 *timeout = timespec_sub(end_time, *timeout);
2231 if (timeout->tv_sec < 0) {
2232 timeout->tv_sec = timeout->tv_nsec = 0;
2233 break;
2234 }
2235
2236 /* Timeout, return less than vlen datagrams */
2237 if (timeout->tv_nsec == 0 && timeout->tv_sec == 0)
2238 break;
2239 }
2240
2241 /* Out of band data, return right away */
2242 if (msg_sys.msg_flags & MSG_OOB)
2243 break;
2244 cond_resched();
2245 }
2246
2247 if (err == 0)
2248 goto out_put;
2249
2250 if (datagrams == 0) {
2251 datagrams = err;
2252 goto out_put;
2253 }
2254
2255 /*
2256 * We may return less entries than requested (vlen) if the
2257 * sock is non block and there aren't enough datagrams...
2258 */
2259 if (err != -EAGAIN) {
2260 /*
2261 * ... or if recvmsg returns an error after we
2262 * received some datagrams, where we record the
2263 * error to return on the next call or if the
2264 * app asks about it using getsockopt(SO_ERROR).
2265 */
2266 sock->sk->sk_err = -err;
2267 }
2268out_put:
2269 fput_light(sock->file, fput_needed);
2270
2271 return datagrams;
2272}
2273
2274SYSCALL_DEFINE5(recvmmsg, int, fd, struct mmsghdr __user *, mmsg,
2275 unsigned int, vlen, unsigned int, flags,
2276 struct timespec __user *, timeout)
2277{
2278 int datagrams;
2279 struct timespec timeout_sys;
2280
2281 if (flags & MSG_CMSG_COMPAT)
2282 return -EINVAL;
2283
2284 if (!timeout)
2285 return __sys_recvmmsg(fd, mmsg, vlen, flags, NULL);
2286
2287 if (copy_from_user(&timeout_sys, timeout, sizeof(timeout_sys)))
2288 return -EFAULT;
2289
2290 datagrams = __sys_recvmmsg(fd, mmsg, vlen, flags, &timeout_sys);
2291
2292 if (datagrams > 0 &&
2293 copy_to_user(timeout, &timeout_sys, sizeof(timeout_sys)))
2294 datagrams = -EFAULT;
2295
2296 return datagrams;
2297}
2298
2299#ifdef __ARCH_WANT_SYS_SOCKETCALL
2300/* Argument list sizes for sys_socketcall */
2301#define AL(x) ((x) * sizeof(unsigned long))
2302static const unsigned char nargs[21] = {
2303 AL(0), AL(3), AL(3), AL(3), AL(2), AL(3),
2304 AL(3), AL(3), AL(4), AL(4), AL(4), AL(6),
2305 AL(6), AL(2), AL(5), AL(5), AL(3), AL(3),
2306 AL(4), AL(5), AL(4)
2307};
2308
2309#undef AL
2310
2311/*
2312 * System call vectors.
2313 *
2314 * Argument checking cleaned up. Saved 20% in size.
2315 * This function doesn't need to set the kernel lock because
2316 * it is set by the callees.
2317 */
2318
2319SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2320{
2321 unsigned long a[AUDITSC_ARGS];
2322 unsigned long a0, a1;
2323 int err;
2324 unsigned int len;
2325
2326 if (call < 1 || call > SYS_SENDMMSG)
2327 return -EINVAL;
2328
2329 len = nargs[call];
2330 if (len > sizeof(a))
2331 return -EINVAL;
2332
2333 /* copy_from_user should be SMP safe. */
2334 if (copy_from_user(a, args, len))
2335 return -EFAULT;
2336
2337 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2338 if (err)
2339 return err;
2340
2341 a0 = a[0];
2342 a1 = a[1];
2343
2344 switch (call) {
2345 case SYS_SOCKET:
2346 err = sys_socket(a0, a1, a[2]);
2347 break;
2348 case SYS_BIND:
2349 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2350 break;
2351 case SYS_CONNECT:
2352 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2353 break;
2354 case SYS_LISTEN:
2355 err = sys_listen(a0, a1);
2356 break;
2357 case SYS_ACCEPT:
2358 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2359 (int __user *)a[2], 0);
2360 break;
2361 case SYS_GETSOCKNAME:
2362 err =
2363 sys_getsockname(a0, (struct sockaddr __user *)a1,
2364 (int __user *)a[2]);
2365 break;
2366 case SYS_GETPEERNAME:
2367 err =
2368 sys_getpeername(a0, (struct sockaddr __user *)a1,
2369 (int __user *)a[2]);
2370 break;
2371 case SYS_SOCKETPAIR:
2372 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2373 break;
2374 case SYS_SEND:
2375 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2376 break;
2377 case SYS_SENDTO:
2378 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2379 (struct sockaddr __user *)a[4], a[5]);
2380 break;
2381 case SYS_RECV:
2382 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2383 break;
2384 case SYS_RECVFROM:
2385 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2386 (struct sockaddr __user *)a[4],
2387 (int __user *)a[5]);
2388 break;
2389 case SYS_SHUTDOWN:
2390 err = sys_shutdown(a0, a1);
2391 break;
2392 case SYS_SETSOCKOPT:
2393 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2394 break;
2395 case SYS_GETSOCKOPT:
2396 err =
2397 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2398 (int __user *)a[4]);
2399 break;
2400 case SYS_SENDMSG:
2401 err = sys_sendmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2402 break;
2403 case SYS_SENDMMSG:
2404 err = sys_sendmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3]);
2405 break;
2406 case SYS_RECVMSG:
2407 err = sys_recvmsg(a0, (struct user_msghdr __user *)a1, a[2]);
2408 break;
2409 case SYS_RECVMMSG:
2410 err = sys_recvmmsg(a0, (struct mmsghdr __user *)a1, a[2], a[3],
2411 (struct timespec __user *)a[4]);
2412 break;
2413 case SYS_ACCEPT4:
2414 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2415 (int __user *)a[2], a[3]);
2416 break;
2417 default:
2418 err = -EINVAL;
2419 break;
2420 }
2421 return err;
2422}
2423
2424#endif /* __ARCH_WANT_SYS_SOCKETCALL */
2425
2426/**
2427 * sock_register - add a socket protocol handler
2428 * @ops: description of protocol
2429 *
2430 * This function is called by a protocol handler that wants to
2431 * advertise its address family, and have it linked into the
2432 * socket interface. The value ops->family corresponds to the
2433 * socket system call protocol family.
2434 */
2435int sock_register(const struct net_proto_family *ops)
2436{
2437 int err;
2438
2439 if (ops->family >= NPROTO) {
2440 pr_crit("protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2441 return -ENOBUFS;
2442 }
2443
2444 spin_lock(&net_family_lock);
2445 if (rcu_dereference_protected(net_families[ops->family],
2446 lockdep_is_held(&net_family_lock)))
2447 err = -EEXIST;
2448 else {
2449 rcu_assign_pointer(net_families[ops->family], ops);
2450 err = 0;
2451 }
2452 spin_unlock(&net_family_lock);
2453
2454 pr_info("NET: Registered protocol family %d\n", ops->family);
2455 return err;
2456}
2457EXPORT_SYMBOL(sock_register);
2458
2459/**
2460 * sock_unregister - remove a protocol handler
2461 * @family: protocol family to remove
2462 *
2463 * This function is called by a protocol handler that wants to
2464 * remove its address family, and have it unlinked from the
2465 * new socket creation.
2466 *
2467 * If protocol handler is a module, then it can use module reference
2468 * counts to protect against new references. If protocol handler is not
2469 * a module then it needs to provide its own protection in
2470 * the ops->create routine.
2471 */
2472void sock_unregister(int family)
2473{
2474 BUG_ON(family < 0 || family >= NPROTO);
2475
2476 spin_lock(&net_family_lock);
2477 RCU_INIT_POINTER(net_families[family], NULL);
2478 spin_unlock(&net_family_lock);
2479
2480 synchronize_rcu();
2481
2482 pr_info("NET: Unregistered protocol family %d\n", family);
2483}
2484EXPORT_SYMBOL(sock_unregister);
2485
2486static int __init sock_init(void)
2487{
2488 int err;
2489 /*
2490 * Initialize the network sysctl infrastructure.
2491 */
2492 err = net_sysctl_init();
2493 if (err)
2494 goto out;
2495
2496 /*
2497 * Initialize skbuff SLAB cache
2498 */
2499 skb_init();
2500
2501 /*
2502 * Initialize the protocols module.
2503 */
2504
2505 init_inodecache();
2506
2507 err = register_filesystem(&sock_fs_type);
2508 if (err)
2509 goto out_fs;
2510 sock_mnt = kern_mount(&sock_fs_type);
2511 if (IS_ERR(sock_mnt)) {
2512 err = PTR_ERR(sock_mnt);
2513 goto out_mount;
2514 }
2515
2516 /* The real protocol initialization is performed in later initcalls.
2517 */
2518
2519#ifdef CONFIG_NETFILTER
2520 err = netfilter_init();
2521 if (err)
2522 goto out;
2523#endif
2524
2525 ptp_classifier_init();
2526
2527out:
2528 return err;
2529
2530out_mount:
2531 unregister_filesystem(&sock_fs_type);
2532out_fs:
2533 goto out;
2534}
2535
2536core_initcall(sock_init); /* early initcall */
2537
2538#ifdef CONFIG_PROC_FS
2539void socket_seq_show(struct seq_file *seq)
2540{
2541 int cpu;
2542 int counter = 0;
2543
2544 for_each_possible_cpu(cpu)
2545 counter += per_cpu(sockets_in_use, cpu);
2546
2547 /* It can be negative, by the way. 8) */
2548 if (counter < 0)
2549 counter = 0;
2550
2551 seq_printf(seq, "sockets: used %d\n", counter);
2552}
2553#endif /* CONFIG_PROC_FS */
2554
2555#ifdef CONFIG_COMPAT
2556static int do_siocgstamp(struct net *net, struct socket *sock,
2557 unsigned int cmd, void __user *up)
2558{
2559 mm_segment_t old_fs = get_fs();
2560 struct timeval ktv;
2561 int err;
2562
2563 set_fs(KERNEL_DS);
2564 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&ktv);
2565 set_fs(old_fs);
2566 if (!err)
2567 err = compat_put_timeval(&ktv, up);
2568
2569 return err;
2570}
2571
2572static int do_siocgstampns(struct net *net, struct socket *sock,
2573 unsigned int cmd, void __user *up)
2574{
2575 mm_segment_t old_fs = get_fs();
2576 struct timespec kts;
2577 int err;
2578
2579 set_fs(KERNEL_DS);
2580 err = sock_do_ioctl(net, sock, cmd, (unsigned long)&kts);
2581 set_fs(old_fs);
2582 if (!err)
2583 err = compat_put_timespec(&kts, up);
2584
2585 return err;
2586}
2587
2588static int dev_ifname32(struct net *net, struct compat_ifreq __user *uifr32)
2589{
2590 struct ifreq __user *uifr;
2591 int err;
2592
2593 uifr = compat_alloc_user_space(sizeof(struct ifreq));
2594 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2595 return -EFAULT;
2596
2597 err = dev_ioctl(net, SIOCGIFNAME, uifr);
2598 if (err)
2599 return err;
2600
2601 if (copy_in_user(uifr32, uifr, sizeof(struct compat_ifreq)))
2602 return -EFAULT;
2603
2604 return 0;
2605}
2606
2607static int dev_ifconf(struct net *net, struct compat_ifconf __user *uifc32)
2608{
2609 struct compat_ifconf ifc32;
2610 struct ifconf ifc;
2611 struct ifconf __user *uifc;
2612 struct compat_ifreq __user *ifr32;
2613 struct ifreq __user *ifr;
2614 unsigned int i, j;
2615 int err;
2616
2617 if (copy_from_user(&ifc32, uifc32, sizeof(struct compat_ifconf)))
2618 return -EFAULT;
2619
2620 memset(&ifc, 0, sizeof(ifc));
2621 if (ifc32.ifcbuf == 0) {
2622 ifc32.ifc_len = 0;
2623 ifc.ifc_len = 0;
2624 ifc.ifc_req = NULL;
2625 uifc = compat_alloc_user_space(sizeof(struct ifconf));
2626 } else {
2627 size_t len = ((ifc32.ifc_len / sizeof(struct compat_ifreq)) + 1) *
2628 sizeof(struct ifreq);
2629 uifc = compat_alloc_user_space(sizeof(struct ifconf) + len);
2630 ifc.ifc_len = len;
2631 ifr = ifc.ifc_req = (void __user *)(uifc + 1);
2632 ifr32 = compat_ptr(ifc32.ifcbuf);
2633 for (i = 0; i < ifc32.ifc_len; i += sizeof(struct compat_ifreq)) {
2634 if (copy_in_user(ifr, ifr32, sizeof(struct compat_ifreq)))
2635 return -EFAULT;
2636 ifr++;
2637 ifr32++;
2638 }
2639 }
2640 if (copy_to_user(uifc, &ifc, sizeof(struct ifconf)))
2641 return -EFAULT;
2642
2643 err = dev_ioctl(net, SIOCGIFCONF, uifc);
2644 if (err)
2645 return err;
2646
2647 if (copy_from_user(&ifc, uifc, sizeof(struct ifconf)))
2648 return -EFAULT;
2649
2650 ifr = ifc.ifc_req;
2651 ifr32 = compat_ptr(ifc32.ifcbuf);
2652 for (i = 0, j = 0;
2653 i + sizeof(struct compat_ifreq) <= ifc32.ifc_len && j < ifc.ifc_len;
2654 i += sizeof(struct compat_ifreq), j += sizeof(struct ifreq)) {
2655 if (copy_in_user(ifr32, ifr, sizeof(struct compat_ifreq)))
2656 return -EFAULT;
2657 ifr32++;
2658 ifr++;
2659 }
2660
2661 if (ifc32.ifcbuf == 0) {
2662 /* Translate from 64-bit structure multiple to
2663 * a 32-bit one.
2664 */
2665 i = ifc.ifc_len;
2666 i = ((i / sizeof(struct ifreq)) * sizeof(struct compat_ifreq));
2667 ifc32.ifc_len = i;
2668 } else {
2669 ifc32.ifc_len = i;
2670 }
2671 if (copy_to_user(uifc32, &ifc32, sizeof(struct compat_ifconf)))
2672 return -EFAULT;
2673
2674 return 0;
2675}
2676
2677static int ethtool_ioctl(struct net *net, struct compat_ifreq __user *ifr32)
2678{
2679 struct compat_ethtool_rxnfc __user *compat_rxnfc;
2680 bool convert_in = false, convert_out = false;
2681 size_t buf_size = ALIGN(sizeof(struct ifreq), 8);
2682 struct ethtool_rxnfc __user *rxnfc;
2683 struct ifreq __user *ifr;
2684 u32 rule_cnt = 0, actual_rule_cnt;
2685 u32 ethcmd;
2686 u32 data;
2687 int ret;
2688
2689 if (get_user(data, &ifr32->ifr_ifru.ifru_data))
2690 return -EFAULT;
2691
2692 compat_rxnfc = compat_ptr(data);
2693
2694 if (get_user(ethcmd, &compat_rxnfc->cmd))
2695 return -EFAULT;
2696
2697 /* Most ethtool structures are defined without padding.
2698 * Unfortunately struct ethtool_rxnfc is an exception.
2699 */
2700 switch (ethcmd) {
2701 default:
2702 break;
2703 case ETHTOOL_GRXCLSRLALL:
2704 /* Buffer size is variable */
2705 if (get_user(rule_cnt, &compat_rxnfc->rule_cnt))
2706 return -EFAULT;
2707 if (rule_cnt > KMALLOC_MAX_SIZE / sizeof(u32))
2708 return -ENOMEM;
2709 buf_size += rule_cnt * sizeof(u32);
2710 /* fall through */
2711 case ETHTOOL_GRXRINGS:
2712 case ETHTOOL_GRXCLSRLCNT:
2713 case ETHTOOL_GRXCLSRULE:
2714 case ETHTOOL_SRXCLSRLINS:
2715 convert_out = true;
2716 /* fall through */
2717 case ETHTOOL_SRXCLSRLDEL:
2718 buf_size += sizeof(struct ethtool_rxnfc);
2719 convert_in = true;
2720 break;
2721 }
2722
2723 ifr = compat_alloc_user_space(buf_size);
2724 rxnfc = (void __user *)ifr + ALIGN(sizeof(struct ifreq), 8);
2725
2726 if (copy_in_user(&ifr->ifr_name, &ifr32->ifr_name, IFNAMSIZ))
2727 return -EFAULT;
2728
2729 if (put_user(convert_in ? rxnfc : compat_ptr(data),
2730 &ifr->ifr_ifru.ifru_data))
2731 return -EFAULT;
2732
2733 if (convert_in) {
2734 /* We expect there to be holes between fs.m_ext and
2735 * fs.ring_cookie and at the end of fs, but nowhere else.
2736 */
2737 BUILD_BUG_ON(offsetof(struct compat_ethtool_rxnfc, fs.m_ext) +
2738 sizeof(compat_rxnfc->fs.m_ext) !=
2739 offsetof(struct ethtool_rxnfc, fs.m_ext) +
2740 sizeof(rxnfc->fs.m_ext));
2741 BUILD_BUG_ON(
2742 offsetof(struct compat_ethtool_rxnfc, fs.location) -
2743 offsetof(struct compat_ethtool_rxnfc, fs.ring_cookie) !=
2744 offsetof(struct ethtool_rxnfc, fs.location) -
2745 offsetof(struct ethtool_rxnfc, fs.ring_cookie));
2746
2747 if (copy_in_user(rxnfc, compat_rxnfc,
2748 (void __user *)(&rxnfc->fs.m_ext + 1) -
2749 (void __user *)rxnfc) ||
2750 copy_in_user(&rxnfc->fs.ring_cookie,
2751 &compat_rxnfc->fs.ring_cookie,
2752 (void __user *)(&rxnfc->fs.location + 1) -
2753 (void __user *)&rxnfc->fs.ring_cookie) ||
2754 copy_in_user(&rxnfc->rule_cnt, &compat_rxnfc->rule_cnt,
2755 sizeof(rxnfc->rule_cnt)))
2756 return -EFAULT;
2757 }
2758
2759 ret = dev_ioctl(net, SIOCETHTOOL, ifr);
2760 if (ret)
2761 return ret;
2762
2763 if (convert_out) {
2764 if (copy_in_user(compat_rxnfc, rxnfc,
2765 (const void __user *)(&rxnfc->fs.m_ext + 1) -
2766 (const void __user *)rxnfc) ||
2767 copy_in_user(&compat_rxnfc->fs.ring_cookie,
2768 &rxnfc->fs.ring_cookie,
2769 (const void __user *)(&rxnfc->fs.location + 1) -
2770 (const void __user *)&rxnfc->fs.ring_cookie) ||
2771 copy_in_user(&compat_rxnfc->rule_cnt, &rxnfc->rule_cnt,
2772 sizeof(rxnfc->rule_cnt)))
2773 return -EFAULT;
2774
2775 if (ethcmd == ETHTOOL_GRXCLSRLALL) {
2776 /* As an optimisation, we only copy the actual
2777 * number of rules that the underlying
2778 * function returned. Since Mallory might
2779 * change the rule count in user memory, we
2780 * check that it is less than the rule count
2781 * originally given (as the user buffer size),
2782 * which has been range-checked.
2783 */
2784 if (get_user(actual_rule_cnt, &rxnfc->rule_cnt))
2785 return -EFAULT;
2786 if (actual_rule_cnt < rule_cnt)
2787 rule_cnt = actual_rule_cnt;
2788 if (copy_in_user(&compat_rxnfc->rule_locs[0],
2789 &rxnfc->rule_locs[0],
2790 rule_cnt * sizeof(u32)))
2791 return -EFAULT;
2792 }
2793 }
2794
2795 return 0;
2796}
2797
2798static int compat_siocwandev(struct net *net, struct compat_ifreq __user *uifr32)
2799{
2800 void __user *uptr;
2801 compat_uptr_t uptr32;
2802 struct ifreq __user *uifr;
2803
2804 uifr = compat_alloc_user_space(sizeof(*uifr));
2805 if (copy_in_user(uifr, uifr32, sizeof(struct compat_ifreq)))
2806 return -EFAULT;
2807
2808 if (get_user(uptr32, &uifr32->ifr_settings.ifs_ifsu))
2809 return -EFAULT;
2810
2811 uptr = compat_ptr(uptr32);
2812
2813 if (put_user(uptr, &uifr->ifr_settings.ifs_ifsu.raw_hdlc))
2814 return -EFAULT;
2815
2816 return dev_ioctl(net, SIOCWANDEV, uifr);
2817}
2818
2819static int bond_ioctl(struct net *net, unsigned int cmd,
2820 struct compat_ifreq __user *ifr32)
2821{
2822 struct ifreq kifr;
2823 mm_segment_t old_fs;
2824 int err;
2825
2826 switch (cmd) {
2827 case SIOCBONDENSLAVE:
2828 case SIOCBONDRELEASE:
2829 case SIOCBONDSETHWADDR:
2830 case SIOCBONDCHANGEACTIVE:
2831 if (copy_from_user(&kifr, ifr32, sizeof(struct compat_ifreq)))
2832 return -EFAULT;
2833
2834 old_fs = get_fs();
2835 set_fs(KERNEL_DS);
2836 err = dev_ioctl(net, cmd,
2837 (struct ifreq __user __force *) &kifr);
2838 set_fs(old_fs);
2839
2840 return err;
2841 default:
2842 return -ENOIOCTLCMD;
2843 }
2844}
2845
2846/* Handle ioctls that use ifreq::ifr_data and just need struct ifreq converted */
2847static int compat_ifr_data_ioctl(struct net *net, unsigned int cmd,
2848 struct compat_ifreq __user *u_ifreq32)
2849{
2850 struct ifreq __user *u_ifreq64;
2851 char tmp_buf[IFNAMSIZ];
2852 void __user *data64;
2853 u32 data32;
2854
2855 if (copy_from_user(&tmp_buf[0], &(u_ifreq32->ifr_ifrn.ifrn_name[0]),
2856 IFNAMSIZ))
2857 return -EFAULT;
2858 if (get_user(data32, &u_ifreq32->ifr_ifru.ifru_data))
2859 return -EFAULT;
2860 data64 = compat_ptr(data32);
2861
2862 u_ifreq64 = compat_alloc_user_space(sizeof(*u_ifreq64));
2863
2864 if (copy_to_user(&u_ifreq64->ifr_ifrn.ifrn_name[0], &tmp_buf[0],
2865 IFNAMSIZ))
2866 return -EFAULT;
2867 if (put_user(data64, &u_ifreq64->ifr_ifru.ifru_data))
2868 return -EFAULT;
2869
2870 return dev_ioctl(net, cmd, u_ifreq64);
2871}
2872
2873static int dev_ifsioc(struct net *net, struct socket *sock,
2874 unsigned int cmd, struct compat_ifreq __user *uifr32)
2875{
2876 struct ifreq __user *uifr;
2877 int err;
2878
2879 uifr = compat_alloc_user_space(sizeof(*uifr));
2880 if (copy_in_user(uifr, uifr32, sizeof(*uifr32)))
2881 return -EFAULT;
2882
2883 err = sock_do_ioctl(net, sock, cmd, (unsigned long)uifr);
2884
2885 if (!err) {
2886 switch (cmd) {
2887 case SIOCGIFFLAGS:
2888 case SIOCGIFMETRIC:
2889 case SIOCGIFMTU:
2890 case SIOCGIFMEM:
2891 case SIOCGIFHWADDR:
2892 case SIOCGIFINDEX:
2893 case SIOCGIFADDR:
2894 case SIOCGIFBRDADDR:
2895 case SIOCGIFDSTADDR:
2896 case SIOCGIFNETMASK:
2897 case SIOCGIFPFLAGS:
2898 case SIOCGIFTXQLEN:
2899 case SIOCGMIIPHY:
2900 case SIOCGMIIREG:
2901 if (copy_in_user(uifr32, uifr, sizeof(*uifr32)))
2902 err = -EFAULT;
2903 break;
2904 }
2905 }
2906 return err;
2907}
2908
2909static int compat_sioc_ifmap(struct net *net, unsigned int cmd,
2910 struct compat_ifreq __user *uifr32)
2911{
2912 struct ifreq ifr;
2913 struct compat_ifmap __user *uifmap32;
2914 mm_segment_t old_fs;
2915 int err;
2916
2917 uifmap32 = &uifr32->ifr_ifru.ifru_map;
2918 err = copy_from_user(&ifr, uifr32, sizeof(ifr.ifr_name));
2919 err |= get_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2920 err |= get_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2921 err |= get_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2922 err |= get_user(ifr.ifr_map.irq, &uifmap32->irq);
2923 err |= get_user(ifr.ifr_map.dma, &uifmap32->dma);
2924 err |= get_user(ifr.ifr_map.port, &uifmap32->port);
2925 if (err)
2926 return -EFAULT;
2927
2928 old_fs = get_fs();
2929 set_fs(KERNEL_DS);
2930 err = dev_ioctl(net, cmd, (void __user __force *)&ifr);
2931 set_fs(old_fs);
2932
2933 if (cmd == SIOCGIFMAP && !err) {
2934 err = copy_to_user(uifr32, &ifr, sizeof(ifr.ifr_name));
2935 err |= put_user(ifr.ifr_map.mem_start, &uifmap32->mem_start);
2936 err |= put_user(ifr.ifr_map.mem_end, &uifmap32->mem_end);
2937 err |= put_user(ifr.ifr_map.base_addr, &uifmap32->base_addr);
2938 err |= put_user(ifr.ifr_map.irq, &uifmap32->irq);
2939 err |= put_user(ifr.ifr_map.dma, &uifmap32->dma);
2940 err |= put_user(ifr.ifr_map.port, &uifmap32->port);
2941 if (err)
2942 err = -EFAULT;
2943 }
2944 return err;
2945}
2946
2947struct rtentry32 {
2948 u32 rt_pad1;
2949 struct sockaddr rt_dst; /* target address */
2950 struct sockaddr rt_gateway; /* gateway addr (RTF_GATEWAY) */
2951 struct sockaddr rt_genmask; /* target network mask (IP) */
2952 unsigned short rt_flags;
2953 short rt_pad2;
2954 u32 rt_pad3;
2955 unsigned char rt_tos;
2956 unsigned char rt_class;
2957 short rt_pad4;
2958 short rt_metric; /* +1 for binary compatibility! */
2959 /* char * */ u32 rt_dev; /* forcing the device at add */
2960 u32 rt_mtu; /* per route MTU/Window */
2961 u32 rt_window; /* Window clamping */
2962 unsigned short rt_irtt; /* Initial RTT */
2963};
2964
2965struct in6_rtmsg32 {
2966 struct in6_addr rtmsg_dst;
2967 struct in6_addr rtmsg_src;
2968 struct in6_addr rtmsg_gateway;
2969 u32 rtmsg_type;
2970 u16 rtmsg_dst_len;
2971 u16 rtmsg_src_len;
2972 u32 rtmsg_metric;
2973 u32 rtmsg_info;
2974 u32 rtmsg_flags;
2975 s32 rtmsg_ifindex;
2976};
2977
2978static int routing_ioctl(struct net *net, struct socket *sock,
2979 unsigned int cmd, void __user *argp)
2980{
2981 int ret;
2982 void *r = NULL;
2983 struct in6_rtmsg r6;
2984 struct rtentry r4;
2985 char devname[16];
2986 u32 rtdev;
2987 mm_segment_t old_fs = get_fs();
2988
2989 if (sock && sock->sk && sock->sk->sk_family == AF_INET6) { /* ipv6 */
2990 struct in6_rtmsg32 __user *ur6 = argp;
2991 ret = copy_from_user(&r6.rtmsg_dst, &(ur6->rtmsg_dst),
2992 3 * sizeof(struct in6_addr));
2993 ret |= get_user(r6.rtmsg_type, &(ur6->rtmsg_type));
2994 ret |= get_user(r6.rtmsg_dst_len, &(ur6->rtmsg_dst_len));
2995 ret |= get_user(r6.rtmsg_src_len, &(ur6->rtmsg_src_len));
2996 ret |= get_user(r6.rtmsg_metric, &(ur6->rtmsg_metric));
2997 ret |= get_user(r6.rtmsg_info, &(ur6->rtmsg_info));
2998 ret |= get_user(r6.rtmsg_flags, &(ur6->rtmsg_flags));
2999 ret |= get_user(r6.rtmsg_ifindex, &(ur6->rtmsg_ifindex));
3000
3001 r = (void *) &r6;
3002 } else { /* ipv4 */
3003 struct rtentry32 __user *ur4 = argp;
3004 ret = copy_from_user(&r4.rt_dst, &(ur4->rt_dst),
3005 3 * sizeof(struct sockaddr));
3006 ret |= get_user(r4.rt_flags, &(ur4->rt_flags));
3007 ret |= get_user(r4.rt_metric, &(ur4->rt_metric));
3008 ret |= get_user(r4.rt_mtu, &(ur4->rt_mtu));
3009 ret |= get_user(r4.rt_window, &(ur4->rt_window));
3010 ret |= get_user(r4.rt_irtt, &(ur4->rt_irtt));
3011 ret |= get_user(rtdev, &(ur4->rt_dev));
3012 if (rtdev) {
3013 ret |= copy_from_user(devname, compat_ptr(rtdev), 15);
3014 r4.rt_dev = (char __user __force *)devname;
3015 devname[15] = 0;
3016 } else
3017 r4.rt_dev = NULL;
3018
3019 r = (void *) &r4;
3020 }
3021
3022 if (ret) {
3023 ret = -EFAULT;
3024 goto out;
3025 }
3026
3027 set_fs(KERNEL_DS);
3028 ret = sock_do_ioctl(net, sock, cmd, (unsigned long) r);
3029 set_fs(old_fs);
3030
3031out:
3032 return ret;
3033}
3034
3035/* Since old style bridge ioctl's endup using SIOCDEVPRIVATE
3036 * for some operations; this forces use of the newer bridge-utils that
3037 * use compatible ioctls
3038 */
3039static int old_bridge_ioctl(compat_ulong_t __user *argp)
3040{
3041 compat_ulong_t tmp;
3042
3043 if (get_user(tmp, argp))
3044 return -EFAULT;
3045 if (tmp == BRCTL_GET_VERSION)
3046 return BRCTL_VERSION + 1;
3047 return -EINVAL;
3048}
3049
3050static int compat_sock_ioctl_trans(struct file *file, struct socket *sock,
3051 unsigned int cmd, unsigned long arg)
3052{
3053 void __user *argp = compat_ptr(arg);
3054 struct sock *sk = sock->sk;
3055 struct net *net = sock_net(sk);
3056
3057 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
3058 return compat_ifr_data_ioctl(net, cmd, argp);
3059
3060 switch (cmd) {
3061 case SIOCSIFBR:
3062 case SIOCGIFBR:
3063 return old_bridge_ioctl(argp);
3064 case SIOCGIFNAME:
3065 return dev_ifname32(net, argp);
3066 case SIOCGIFCONF:
3067 return dev_ifconf(net, argp);
3068 case SIOCETHTOOL:
3069 return ethtool_ioctl(net, argp);
3070 case SIOCWANDEV:
3071 return compat_siocwandev(net, argp);
3072 case SIOCGIFMAP:
3073 case SIOCSIFMAP:
3074 return compat_sioc_ifmap(net, cmd, argp);
3075 case SIOCBONDENSLAVE:
3076 case SIOCBONDRELEASE:
3077 case SIOCBONDSETHWADDR:
3078 case SIOCBONDCHANGEACTIVE:
3079 return bond_ioctl(net, cmd, argp);
3080 case SIOCADDRT:
3081 case SIOCDELRT:
3082 return routing_ioctl(net, sock, cmd, argp);
3083 case SIOCGSTAMP:
3084 return do_siocgstamp(net, sock, cmd, argp);
3085 case SIOCGSTAMPNS:
3086 return do_siocgstampns(net, sock, cmd, argp);
3087 case SIOCBONDSLAVEINFOQUERY:
3088 case SIOCBONDINFOQUERY:
3089 case SIOCSHWTSTAMP:
3090 case SIOCGHWTSTAMP:
3091 return compat_ifr_data_ioctl(net, cmd, argp);
3092
3093 case FIOSETOWN:
3094 case SIOCSPGRP:
3095 case FIOGETOWN:
3096 case SIOCGPGRP:
3097 case SIOCBRADDBR:
3098 case SIOCBRDELBR:
3099 case SIOCGIFVLAN:
3100 case SIOCSIFVLAN:
3101 case SIOCADDDLCI:
3102 case SIOCDELDLCI:
3103 return sock_ioctl(file, cmd, arg);
3104
3105 case SIOCGIFFLAGS:
3106 case SIOCSIFFLAGS:
3107 case SIOCGIFMETRIC:
3108 case SIOCSIFMETRIC:
3109 case SIOCGIFMTU:
3110 case SIOCSIFMTU:
3111 case SIOCGIFMEM:
3112 case SIOCSIFMEM:
3113 case SIOCGIFHWADDR:
3114 case SIOCSIFHWADDR:
3115 case SIOCADDMULTI:
3116 case SIOCDELMULTI:
3117 case SIOCGIFINDEX:
3118 case SIOCGIFADDR:
3119 case SIOCSIFADDR:
3120 case SIOCSIFHWBROADCAST:
3121 case SIOCDIFADDR:
3122 case SIOCGIFBRDADDR:
3123 case SIOCSIFBRDADDR:
3124 case SIOCGIFDSTADDR:
3125 case SIOCSIFDSTADDR:
3126 case SIOCGIFNETMASK:
3127 case SIOCSIFNETMASK:
3128 case SIOCSIFPFLAGS:
3129 case SIOCGIFPFLAGS:
3130 case SIOCGIFTXQLEN:
3131 case SIOCSIFTXQLEN:
3132 case SIOCBRADDIF:
3133 case SIOCBRDELIF:
3134 case SIOCSIFNAME:
3135 case SIOCGMIIPHY:
3136 case SIOCGMIIREG:
3137 case SIOCSMIIREG:
3138 return dev_ifsioc(net, sock, cmd, argp);
3139
3140 case SIOCSARP:
3141 case SIOCGARP:
3142 case SIOCDARP:
3143 case SIOCATMARK:
3144 return sock_do_ioctl(net, sock, cmd, arg);
3145 }
3146
3147 return -ENOIOCTLCMD;
3148}
3149
3150static long compat_sock_ioctl(struct file *file, unsigned int cmd,
3151 unsigned long arg)
3152{
3153 struct socket *sock = file->private_data;
3154 int ret = -ENOIOCTLCMD;
3155 struct sock *sk;
3156 struct net *net;
3157
3158 sk = sock->sk;
3159 net = sock_net(sk);
3160
3161 if (sock->ops->compat_ioctl)
3162 ret = sock->ops->compat_ioctl(sock, cmd, arg);
3163
3164 if (ret == -ENOIOCTLCMD &&
3165 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
3166 ret = compat_wext_handle_ioctl(net, cmd, arg);
3167
3168 if (ret == -ENOIOCTLCMD)
3169 ret = compat_sock_ioctl_trans(file, sock, cmd, arg);
3170
3171 return ret;
3172}
3173#endif
3174
3175int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
3176{
3177 return sock->ops->bind(sock, addr, addrlen);
3178}
3179EXPORT_SYMBOL(kernel_bind);
3180
3181int kernel_listen(struct socket *sock, int backlog)
3182{
3183 return sock->ops->listen(sock, backlog);
3184}
3185EXPORT_SYMBOL(kernel_listen);
3186
3187int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
3188{
3189 struct sock *sk = sock->sk;
3190 int err;
3191
3192 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
3193 newsock);
3194 if (err < 0)
3195 goto done;
3196
3197 err = sock->ops->accept(sock, *newsock, flags);
3198 if (err < 0) {
3199 sock_release(*newsock);
3200 *newsock = NULL;
3201 goto done;
3202 }
3203
3204 (*newsock)->ops = sock->ops;
3205 __module_get((*newsock)->ops->owner);
3206
3207done:
3208 return err;
3209}
3210EXPORT_SYMBOL(kernel_accept);
3211
3212int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
3213 int flags)
3214{
3215 return sock->ops->connect(sock, addr, addrlen, flags);
3216}
3217EXPORT_SYMBOL(kernel_connect);
3218
3219int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
3220 int *addrlen)
3221{
3222 return sock->ops->getname(sock, addr, addrlen, 0);
3223}
3224EXPORT_SYMBOL(kernel_getsockname);
3225
3226int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
3227 int *addrlen)
3228{
3229 return sock->ops->getname(sock, addr, addrlen, 1);
3230}
3231EXPORT_SYMBOL(kernel_getpeername);
3232
3233int kernel_getsockopt(struct socket *sock, int level, int optname,
3234 char *optval, int *optlen)
3235{
3236 mm_segment_t oldfs = get_fs();
3237 char __user *uoptval;
3238 int __user *uoptlen;
3239 int err;
3240
3241 uoptval = (char __user __force *) optval;
3242 uoptlen = (int __user __force *) optlen;
3243
3244 set_fs(KERNEL_DS);
3245 if (level == SOL_SOCKET)
3246 err = sock_getsockopt(sock, level, optname, uoptval, uoptlen);
3247 else
3248 err = sock->ops->getsockopt(sock, level, optname, uoptval,
3249 uoptlen);
3250 set_fs(oldfs);
3251 return err;
3252}
3253EXPORT_SYMBOL(kernel_getsockopt);
3254
3255int kernel_setsockopt(struct socket *sock, int level, int optname,
3256 char *optval, unsigned int optlen)
3257{
3258 mm_segment_t oldfs = get_fs();
3259 char __user *uoptval;
3260 int err;
3261
3262 uoptval = (char __user __force *) optval;
3263
3264 set_fs(KERNEL_DS);
3265 if (level == SOL_SOCKET)
3266 err = sock_setsockopt(sock, level, optname, uoptval, optlen);
3267 else
3268 err = sock->ops->setsockopt(sock, level, optname, uoptval,
3269 optlen);
3270 set_fs(oldfs);
3271 return err;
3272}
3273EXPORT_SYMBOL(kernel_setsockopt);
3274
3275int kernel_sendpage(struct socket *sock, struct page *page, int offset,
3276 size_t size, int flags)
3277{
3278 if (sock->ops->sendpage)
3279 return sock->ops->sendpage(sock, page, offset, size, flags);
3280
3281 return sock_no_sendpage(sock, page, offset, size, flags);
3282}
3283EXPORT_SYMBOL(kernel_sendpage);
3284
3285int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
3286{
3287 mm_segment_t oldfs = get_fs();
3288 int err;
3289
3290 set_fs(KERNEL_DS);
3291 err = sock->ops->ioctl(sock, cmd, arg);
3292 set_fs(oldfs);
3293
3294 return err;
3295}
3296EXPORT_SYMBOL(kernel_sock_ioctl);
3297
3298int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
3299{
3300 return sock->ops->shutdown(sock, how);
3301}
3302EXPORT_SYMBOL(kernel_sock_shutdown);