fed up with those stupid warnings
[mmotm.git] / net / socket.c
blobfab67c134327a3bb729884bae8adecf957df43c3
1 /*
2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
7 * Ross Biro
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
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
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.
55 * This module is effectively the top level interface to the BSD socket
56 * paradigm.
58 * Based upon Swansea University Computer Society NET3.039
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/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.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>
91 #include <asm/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
95 #include <net/wext.h>
97 #include <net/sock.h>
98 #include <linux/netfilter.h>
100 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
101 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
102 unsigned long nr_segs, loff_t pos);
103 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
104 unsigned long nr_segs, loff_t pos);
105 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
107 static int sock_close(struct inode *inode, struct file *file);
108 static unsigned int sock_poll(struct file *file,
109 struct poll_table_struct *wait);
110 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
111 #ifdef CONFIG_COMPAT
112 static long compat_sock_ioctl(struct file *file,
113 unsigned int cmd, unsigned long arg);
114 #endif
115 static int sock_fasync(int fd, struct file *filp, int on);
116 static ssize_t sock_sendpage(struct file *file, struct page *page,
117 int offset, size_t size, loff_t *ppos, int more);
118 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
119 struct pipe_inode_info *pipe, size_t len,
120 unsigned int flags);
123 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
124 * in the operation structures but are done directly via the socketcall() multiplexor.
127 static const struct file_operations socket_file_ops = {
128 .owner = THIS_MODULE,
129 .llseek = no_llseek,
130 .aio_read = sock_aio_read,
131 .aio_write = sock_aio_write,
132 .poll = sock_poll,
133 .unlocked_ioctl = sock_ioctl,
134 #ifdef CONFIG_COMPAT
135 .compat_ioctl = compat_sock_ioctl,
136 #endif
137 .mmap = sock_mmap,
138 .open = sock_no_open, /* special open code to disallow open via /proc */
139 .release = sock_close,
140 .fasync = sock_fasync,
141 .sendpage = sock_sendpage,
142 .splice_write = generic_splice_sendpage,
143 .splice_read = sock_splice_read,
147 * The protocol list. Each protocol is registered in here.
150 static DEFINE_SPINLOCK(net_family_lock);
151 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
154 * Statistics counters of the socket lists
157 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
160 * Support routines.
161 * Move socket addresses back and forth across the kernel/user
162 * divide and look after the messy bits.
165 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
166 16 for IP, 16 for IPX,
167 24 for IPv6,
168 about 80 for AX.25
169 must be at least one bigger than
170 the AF_UNIX size (see net/unix/af_unix.c
171 :unix_mkname()).
175 * move_addr_to_kernel - copy a socket address into kernel space
176 * @uaddr: Address in user space
177 * @kaddr: Address in kernel space
178 * @ulen: Length in user space
180 * The address is copied into kernel space. If the provided address is
181 * too long an error code of -EINVAL is returned. If the copy gives
182 * invalid addresses -EFAULT is returned. On a success 0 is returned.
185 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
187 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
188 return -EINVAL;
189 if (ulen == 0)
190 return 0;
191 if (copy_from_user(kaddr, uaddr, ulen))
192 return -EFAULT;
193 return audit_sockaddr(ulen, kaddr);
197 * move_addr_to_user - copy an address to user space
198 * @kaddr: kernel space address
199 * @klen: length of address in kernel
200 * @uaddr: user space address
201 * @ulen: pointer to user length field
203 * The value pointed to by ulen on entry is the buffer length available.
204 * This is overwritten with the buffer space used. -EINVAL is returned
205 * if an overlong buffer is specified or a negative buffer size. -EFAULT
206 * is returned if either the buffer or the length field are not
207 * accessible.
208 * After copying the data up to the limit the user specifies, the true
209 * length of the data is written over the length limit the user
210 * specified. Zero is returned for a success.
213 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
214 int __user *ulen)
216 int err;
217 int len;
219 err = get_user(len, ulen);
220 if (err)
221 return err;
222 if (len > klen)
223 len = klen;
224 if (len < 0 || len > sizeof(struct sockaddr_storage))
225 return -EINVAL;
226 if (len) {
227 if (audit_sockaddr(klen, kaddr))
228 return -ENOMEM;
229 if (copy_to_user(uaddr, kaddr, len))
230 return -EFAULT;
233 * "fromlen shall refer to the value before truncation.."
234 * 1003.1g
236 return __put_user(klen, ulen);
239 static struct kmem_cache *sock_inode_cachep __read_mostly;
241 static struct inode *sock_alloc_inode(struct super_block *sb)
243 struct socket_alloc *ei;
245 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
246 if (!ei)
247 return NULL;
248 init_waitqueue_head(&ei->socket.wait);
250 ei->socket.fasync_list = NULL;
251 ei->socket.state = SS_UNCONNECTED;
252 ei->socket.flags = 0;
253 ei->socket.ops = NULL;
254 ei->socket.sk = NULL;
255 ei->socket.file = NULL;
257 return &ei->vfs_inode;
260 static void sock_destroy_inode(struct inode *inode)
262 kmem_cache_free(sock_inode_cachep,
263 container_of(inode, struct socket_alloc, vfs_inode));
266 static void init_once(void *foo)
268 struct socket_alloc *ei = (struct socket_alloc *)foo;
270 inode_init_once(&ei->vfs_inode);
273 static int init_inodecache(void)
275 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
276 sizeof(struct socket_alloc),
278 (SLAB_HWCACHE_ALIGN |
279 SLAB_RECLAIM_ACCOUNT |
280 SLAB_MEM_SPREAD),
281 init_once);
282 if (sock_inode_cachep == NULL)
283 return -ENOMEM;
284 return 0;
287 static const struct super_operations sockfs_ops = {
288 .alloc_inode = sock_alloc_inode,
289 .destroy_inode =sock_destroy_inode,
290 .statfs = simple_statfs,
293 static int sockfs_get_sb(struct file_system_type *fs_type,
294 int flags, const char *dev_name, void *data,
295 struct vfsmount *mnt)
297 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
298 mnt);
301 static struct vfsmount *sock_mnt __read_mostly;
303 static struct file_system_type sock_fs_type = {
304 .name = "sockfs",
305 .get_sb = sockfs_get_sb,
306 .kill_sb = kill_anon_super,
310 * sockfs_dname() is called from d_path().
312 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
314 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
315 dentry->d_inode->i_ino);
318 static const struct dentry_operations sockfs_dentry_operations = {
319 .d_dname = sockfs_dname,
323 * Obtains the first available file descriptor and sets it up for use.
325 * These functions create file structures and maps them to fd space
326 * of the current process. On success it returns file descriptor
327 * and file struct implicitly stored in sock->file.
328 * Note that another thread may close file descriptor before we return
329 * from this function. We use the fact that now we do not refer
330 * to socket after mapping. If one day we will need it, this
331 * function will increment ref. count on file by 1.
333 * In any case returned fd MAY BE not valid!
334 * This race condition is unavoidable
335 * with shared fd spaces, we cannot solve it inside kernel,
336 * but we take care of internal coherence yet.
339 static int sock_alloc_fd(struct file **filep, int flags)
341 int fd;
343 fd = get_unused_fd_flags(flags);
344 if (likely(fd >= 0)) {
345 struct file *file = get_empty_filp();
347 *filep = file;
348 if (unlikely(!file)) {
349 put_unused_fd(fd);
350 return -ENFILE;
352 } else
353 *filep = NULL;
354 return fd;
357 static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
359 struct dentry *dentry;
360 struct qstr name = { .name = "" };
362 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
363 if (unlikely(!dentry))
364 return -ENOMEM;
366 dentry->d_op = &sockfs_dentry_operations;
367 d_instantiate(dentry, SOCK_INODE(sock));
369 sock->file = file;
370 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
371 &socket_file_ops);
372 SOCK_INODE(sock)->i_fop = &socket_file_ops;
373 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
374 file->f_pos = 0;
375 file->private_data = sock;
377 return 0;
380 int sock_map_fd(struct socket *sock, int flags)
382 struct file *newfile;
383 int fd = sock_alloc_fd(&newfile, flags);
385 if (likely(fd >= 0)) {
386 int err = sock_attach_fd(sock, newfile, flags);
388 if (unlikely(err < 0)) {
389 put_filp(newfile);
390 put_unused_fd(fd);
391 return err;
393 fd_install(fd, newfile);
395 return fd;
398 static struct socket *sock_from_file(struct file *file, int *err)
400 if (file->f_op == &socket_file_ops)
401 return file->private_data; /* set in sock_map_fd */
403 *err = -ENOTSOCK;
404 return NULL;
408 * sockfd_lookup - Go from a file number to its socket slot
409 * @fd: file handle
410 * @err: pointer to an error code return
412 * The file handle passed in is locked and the socket it is bound
413 * too is returned. If an error occurs the err pointer is overwritten
414 * with a negative errno code and NULL is returned. The function checks
415 * for both invalid handles and passing a handle which is not a socket.
417 * On a success the socket object pointer is returned.
420 struct socket *sockfd_lookup(int fd, int *err)
422 struct file *file;
423 struct socket *sock;
425 file = fget(fd);
426 if (!file) {
427 *err = -EBADF;
428 return NULL;
431 sock = sock_from_file(file, err);
432 if (!sock)
433 fput(file);
434 return sock;
437 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
439 struct file *file;
440 struct socket *sock;
442 *err = -EBADF;
443 file = fget_light(fd, fput_needed);
444 if (file) {
445 sock = sock_from_file(file, err);
446 if (sock)
447 return sock;
448 fput_light(file, *fput_needed);
450 return NULL;
454 * sock_alloc - allocate a socket
456 * Allocate a new inode and socket object. The two are bound together
457 * and initialised. The socket is then returned. If we are out of inodes
458 * NULL is returned.
461 static struct socket *sock_alloc(void)
463 struct inode *inode;
464 struct socket *sock;
466 inode = new_inode(sock_mnt->mnt_sb);
467 if (!inode)
468 return NULL;
470 sock = SOCKET_I(inode);
472 kmemcheck_annotate_bitfield(sock, type);
473 inode->i_mode = S_IFSOCK | S_IRWXUGO;
474 inode->i_uid = current_fsuid();
475 inode->i_gid = current_fsgid();
477 percpu_add(sockets_in_use, 1);
478 return sock;
482 * In theory you can't get an open on this inode, but /proc provides
483 * a back door. Remember to keep it shut otherwise you'll let the
484 * creepy crawlies in.
487 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
489 return -ENXIO;
492 const struct file_operations bad_sock_fops = {
493 .owner = THIS_MODULE,
494 .open = sock_no_open,
498 * sock_release - close a socket
499 * @sock: socket to close
501 * The socket is released from the protocol stack if it has a release
502 * callback, and the inode is then released if the socket is bound to
503 * an inode not a file.
506 void sock_release(struct socket *sock)
508 if (sock->ops) {
509 struct module *owner = sock->ops->owner;
511 sock->ops->release(sock);
512 sock->ops = NULL;
513 module_put(owner);
516 if (sock->fasync_list)
517 printk(KERN_ERR "sock_release: fasync list not empty!\n");
519 percpu_sub(sockets_in_use, 1);
520 if (!sock->file) {
521 iput(SOCK_INODE(sock));
522 return;
524 sock->file = NULL;
527 int sock_tx_timestamp(struct msghdr *msg, struct sock *sk,
528 union skb_shared_tx *shtx)
530 shtx->flags = 0;
531 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
532 shtx->hardware = 1;
533 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
534 shtx->software = 1;
535 return 0;
537 EXPORT_SYMBOL(sock_tx_timestamp);
539 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
540 struct msghdr *msg, size_t size)
542 struct sock_iocb *si = kiocb_to_siocb(iocb);
543 int err;
545 si->sock = sock;
546 si->scm = NULL;
547 si->msg = msg;
548 si->size = size;
550 err = security_socket_sendmsg(sock, msg, size);
551 if (err)
552 return err;
554 return sock->ops->sendmsg(iocb, sock, msg, size);
557 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
559 struct kiocb iocb;
560 struct sock_iocb siocb;
561 int ret;
563 init_sync_kiocb(&iocb, NULL);
564 iocb.private = &siocb;
565 ret = __sock_sendmsg(&iocb, sock, msg, size);
566 if (-EIOCBQUEUED == ret)
567 ret = wait_on_sync_kiocb(&iocb);
568 return ret;
571 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
572 struct kvec *vec, size_t num, size_t size)
574 mm_segment_t oldfs = get_fs();
575 int result;
577 set_fs(KERNEL_DS);
579 * the following is safe, since for compiler definitions of kvec and
580 * iovec are identical, yielding the same in-core layout and alignment
582 msg->msg_iov = (struct iovec *)vec;
583 msg->msg_iovlen = num;
584 result = sock_sendmsg(sock, msg, size);
585 set_fs(oldfs);
586 return result;
589 static int ktime2ts(ktime_t kt, struct timespec *ts)
591 if (kt.tv64) {
592 *ts = ktime_to_timespec(kt);
593 return 1;
594 } else {
595 return 0;
600 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
602 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
603 struct sk_buff *skb)
605 int need_software_tstamp = sock_flag(sk, SOCK_RCVTSTAMP);
606 struct timespec ts[3];
607 int empty = 1;
608 struct skb_shared_hwtstamps *shhwtstamps =
609 skb_hwtstamps(skb);
611 /* Race occurred between timestamp enabling and packet
612 receiving. Fill in the current time for now. */
613 if (need_software_tstamp && skb->tstamp.tv64 == 0)
614 __net_timestamp(skb);
616 if (need_software_tstamp) {
617 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
618 struct timeval tv;
619 skb_get_timestamp(skb, &tv);
620 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP,
621 sizeof(tv), &tv);
622 } else {
623 struct timespec ts;
624 skb_get_timestampns(skb, &ts);
625 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS,
626 sizeof(ts), &ts);
631 memset(ts, 0, sizeof(ts));
632 if (skb->tstamp.tv64 &&
633 sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) {
634 skb_get_timestampns(skb, ts + 0);
635 empty = 0;
637 if (shhwtstamps) {
638 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE) &&
639 ktime2ts(shhwtstamps->syststamp, ts + 1))
640 empty = 0;
641 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE) &&
642 ktime2ts(shhwtstamps->hwtstamp, ts + 2))
643 empty = 0;
645 if (!empty)
646 put_cmsg(msg, SOL_SOCKET,
647 SCM_TIMESTAMPING, sizeof(ts), &ts);
650 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
652 inline void sock_recv_drops(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
654 if (sock_flag(sk, SOCK_RXQ_OVFL) && skb && skb->dropcount)
655 put_cmsg(msg, SOL_SOCKET, SO_RXQ_OVFL,
656 sizeof(__u32), &skb->dropcount);
659 void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
660 struct sk_buff *skb)
662 sock_recv_timestamp(msg, sk, skb);
663 sock_recv_drops(msg, sk, skb);
665 EXPORT_SYMBOL_GPL(sock_recv_ts_and_drops);
667 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
668 struct msghdr *msg, size_t size, int flags)
670 int err;
671 struct sock_iocb *si = kiocb_to_siocb(iocb);
673 si->sock = sock;
674 si->scm = NULL;
675 si->msg = msg;
676 si->size = size;
677 si->flags = flags;
679 err = security_socket_recvmsg(sock, msg, size, flags);
680 if (err)
681 return err;
683 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
686 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
687 size_t size, int flags)
689 struct kiocb iocb;
690 struct sock_iocb siocb;
691 int ret;
693 init_sync_kiocb(&iocb, NULL);
694 iocb.private = &siocb;
695 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
696 if (-EIOCBQUEUED == ret)
697 ret = wait_on_sync_kiocb(&iocb);
698 return ret;
701 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
702 struct kvec *vec, size_t num, size_t size, int flags)
704 mm_segment_t oldfs = get_fs();
705 int result;
707 set_fs(KERNEL_DS);
709 * the following is safe, since for compiler definitions of kvec and
710 * iovec are identical, yielding the same in-core layout and alignment
712 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
713 result = sock_recvmsg(sock, msg, size, flags);
714 set_fs(oldfs);
715 return result;
718 static void sock_aio_dtor(struct kiocb *iocb)
720 kfree(iocb->private);
723 static ssize_t sock_sendpage(struct file *file, struct page *page,
724 int offset, size_t size, loff_t *ppos, int more)
726 struct socket *sock;
727 int flags;
729 sock = file->private_data;
731 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
732 if (more)
733 flags |= MSG_MORE;
735 return kernel_sendpage(sock, page, offset, size, flags);
738 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
739 struct pipe_inode_info *pipe, size_t len,
740 unsigned int flags)
742 struct socket *sock = file->private_data;
744 if (unlikely(!sock->ops->splice_read))
745 return -EINVAL;
747 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
750 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
751 struct sock_iocb *siocb)
753 if (!is_sync_kiocb(iocb)) {
754 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
755 if (!siocb)
756 return NULL;
757 iocb->ki_dtor = sock_aio_dtor;
760 siocb->kiocb = iocb;
761 iocb->private = siocb;
762 return siocb;
765 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
766 struct file *file, const struct iovec *iov,
767 unsigned long nr_segs)
769 struct socket *sock = file->private_data;
770 size_t size = 0;
771 int i;
773 for (i = 0; i < nr_segs; i++)
774 size += iov[i].iov_len;
776 msg->msg_name = NULL;
777 msg->msg_namelen = 0;
778 msg->msg_control = NULL;
779 msg->msg_controllen = 0;
780 msg->msg_iov = (struct iovec *)iov;
781 msg->msg_iovlen = nr_segs;
782 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
784 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
787 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
788 unsigned long nr_segs, loff_t pos)
790 struct sock_iocb siocb, *x;
792 if (pos != 0)
793 return -ESPIPE;
795 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
796 return 0;
799 x = alloc_sock_iocb(iocb, &siocb);
800 if (!x)
801 return -ENOMEM;
802 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
805 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
806 struct file *file, const struct iovec *iov,
807 unsigned long nr_segs)
809 struct socket *sock = file->private_data;
810 size_t size = 0;
811 int i;
813 for (i = 0; i < nr_segs; i++)
814 size += iov[i].iov_len;
816 msg->msg_name = NULL;
817 msg->msg_namelen = 0;
818 msg->msg_control = NULL;
819 msg->msg_controllen = 0;
820 msg->msg_iov = (struct iovec *)iov;
821 msg->msg_iovlen = nr_segs;
822 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
823 if (sock->type == SOCK_SEQPACKET)
824 msg->msg_flags |= MSG_EOR;
826 return __sock_sendmsg(iocb, sock, msg, size);
829 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
830 unsigned long nr_segs, loff_t pos)
832 struct sock_iocb siocb, *x;
834 if (pos != 0)
835 return -ESPIPE;
837 x = alloc_sock_iocb(iocb, &siocb);
838 if (!x)
839 return -ENOMEM;
841 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
845 * Atomic setting of ioctl hooks to avoid race
846 * with module unload.
849 static DEFINE_MUTEX(br_ioctl_mutex);
850 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
852 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
854 mutex_lock(&br_ioctl_mutex);
855 br_ioctl_hook = hook;
856 mutex_unlock(&br_ioctl_mutex);
859 EXPORT_SYMBOL(brioctl_set);
861 static DEFINE_MUTEX(vlan_ioctl_mutex);
862 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
864 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
866 mutex_lock(&vlan_ioctl_mutex);
867 vlan_ioctl_hook = hook;
868 mutex_unlock(&vlan_ioctl_mutex);
871 EXPORT_SYMBOL(vlan_ioctl_set);
873 static DEFINE_MUTEX(dlci_ioctl_mutex);
874 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
876 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
878 mutex_lock(&dlci_ioctl_mutex);
879 dlci_ioctl_hook = hook;
880 mutex_unlock(&dlci_ioctl_mutex);
883 EXPORT_SYMBOL(dlci_ioctl_set);
886 * With an ioctl, arg may well be a user mode pointer, but we don't know
887 * what to do with it - that's up to the protocol still.
890 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
892 struct socket *sock;
893 struct sock *sk;
894 void __user *argp = (void __user *)arg;
895 int pid, err;
896 struct net *net;
898 sock = file->private_data;
899 sk = sock->sk;
900 net = sock_net(sk);
901 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
902 err = dev_ioctl(net, cmd, argp);
903 } else
904 #ifdef CONFIG_WEXT_CORE
905 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
906 err = dev_ioctl(net, cmd, argp);
907 } else
908 #endif
909 switch (cmd) {
910 case FIOSETOWN:
911 case SIOCSPGRP:
912 err = -EFAULT;
913 if (get_user(pid, (int __user *)argp))
914 break;
915 err = f_setown(sock->file, pid, 1);
916 break;
917 case FIOGETOWN:
918 case SIOCGPGRP:
919 err = put_user(f_getown(sock->file),
920 (int __user *)argp);
921 break;
922 case SIOCGIFBR:
923 case SIOCSIFBR:
924 case SIOCBRADDBR:
925 case SIOCBRDELBR:
926 err = -ENOPKG;
927 if (!br_ioctl_hook)
928 request_module("bridge");
930 mutex_lock(&br_ioctl_mutex);
931 if (br_ioctl_hook)
932 err = br_ioctl_hook(net, cmd, argp);
933 mutex_unlock(&br_ioctl_mutex);
934 break;
935 case SIOCGIFVLAN:
936 case SIOCSIFVLAN:
937 err = -ENOPKG;
938 if (!vlan_ioctl_hook)
939 request_module("8021q");
941 mutex_lock(&vlan_ioctl_mutex);
942 if (vlan_ioctl_hook)
943 err = vlan_ioctl_hook(net, argp);
944 mutex_unlock(&vlan_ioctl_mutex);
945 break;
946 case SIOCADDDLCI:
947 case SIOCDELDLCI:
948 err = -ENOPKG;
949 if (!dlci_ioctl_hook)
950 request_module("dlci");
952 mutex_lock(&dlci_ioctl_mutex);
953 if (dlci_ioctl_hook)
954 err = dlci_ioctl_hook(cmd, argp);
955 mutex_unlock(&dlci_ioctl_mutex);
956 break;
957 default:
958 err = sock->ops->ioctl(sock, cmd, arg);
961 * If this ioctl is unknown try to hand it down
962 * to the NIC driver.
964 if (err == -ENOIOCTLCMD)
965 err = dev_ioctl(net, cmd, argp);
966 break;
968 return err;
971 int sock_create_lite(int family, int type, int protocol, struct socket **res)
973 int err;
974 struct socket *sock = NULL;
976 err = security_socket_create(family, type, protocol, 1);
977 if (err)
978 goto out;
980 sock = sock_alloc();
981 if (!sock) {
982 err = -ENOMEM;
983 goto out;
986 sock->type = type;
987 err = security_socket_post_create(sock, family, type, protocol, 1);
988 if (err)
989 goto out_release;
991 out:
992 *res = sock;
993 return err;
994 out_release:
995 sock_release(sock);
996 sock = NULL;
997 goto out;
1000 /* No kernel lock held - perfect */
1001 static unsigned int sock_poll(struct file *file, poll_table *wait)
1003 struct socket *sock;
1006 * We can't return errors to poll, so it's either yes or no.
1008 sock = file->private_data;
1009 return sock->ops->poll(file, sock, wait);
1012 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
1014 struct socket *sock = file->private_data;
1016 return sock->ops->mmap(file, sock, vma);
1019 static int sock_close(struct inode *inode, struct file *filp)
1022 * It was possible the inode is NULL we were
1023 * closing an unfinished socket.
1026 if (!inode) {
1027 printk(KERN_DEBUG "sock_close: NULL inode\n");
1028 return 0;
1030 sock_release(SOCKET_I(inode));
1031 return 0;
1035 * Update the socket async list
1037 * Fasync_list locking strategy.
1039 * 1. fasync_list is modified only under process context socket lock
1040 * i.e. under semaphore.
1041 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1042 * or under socket lock.
1043 * 3. fasync_list can be used from softirq context, so that
1044 * modification under socket lock have to be enhanced with
1045 * write_lock_bh(&sk->sk_callback_lock).
1046 * --ANK (990710)
1049 static int sock_fasync(int fd, struct file *filp, int on)
1051 struct fasync_struct *fa, *fna = NULL, **prev;
1052 struct socket *sock;
1053 struct sock *sk;
1055 if (on) {
1056 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1057 if (fna == NULL)
1058 return -ENOMEM;
1061 sock = filp->private_data;
1063 sk = sock->sk;
1064 if (sk == NULL) {
1065 kfree(fna);
1066 return -EINVAL;
1069 lock_sock(sk);
1071 spin_lock(&filp->f_lock);
1072 if (on)
1073 filp->f_flags |= FASYNC;
1074 else
1075 filp->f_flags &= ~FASYNC;
1076 spin_unlock(&filp->f_lock);
1078 prev = &(sock->fasync_list);
1080 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1081 if (fa->fa_file == filp)
1082 break;
1084 if (on) {
1085 if (fa != NULL) {
1086 write_lock_bh(&sk->sk_callback_lock);
1087 fa->fa_fd = fd;
1088 write_unlock_bh(&sk->sk_callback_lock);
1090 kfree(fna);
1091 goto out;
1093 fna->fa_file = filp;
1094 fna->fa_fd = fd;
1095 fna->magic = FASYNC_MAGIC;
1096 fna->fa_next = sock->fasync_list;
1097 write_lock_bh(&sk->sk_callback_lock);
1098 sock->fasync_list = fna;
1099 sock_set_flag(sk, SOCK_FASYNC);
1100 write_unlock_bh(&sk->sk_callback_lock);
1101 } else {
1102 if (fa != NULL) {
1103 write_lock_bh(&sk->sk_callback_lock);
1104 *prev = fa->fa_next;
1105 if (!sock->fasync_list)
1106 sock_reset_flag(sk, SOCK_FASYNC);
1107 write_unlock_bh(&sk->sk_callback_lock);
1108 kfree(fa);
1112 out:
1113 release_sock(sock->sk);
1114 return 0;
1117 /* This function may be called only under socket lock or callback_lock */
1119 int sock_wake_async(struct socket *sock, int how, int band)
1121 if (!sock || !sock->fasync_list)
1122 return -1;
1123 switch (how) {
1124 case SOCK_WAKE_WAITD:
1125 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1126 break;
1127 goto call_kill;
1128 case SOCK_WAKE_SPACE:
1129 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1130 break;
1131 /* fall through */
1132 case SOCK_WAKE_IO:
1133 call_kill:
1134 __kill_fasync(sock->fasync_list, SIGIO, band);
1135 break;
1136 case SOCK_WAKE_URG:
1137 __kill_fasync(sock->fasync_list, SIGURG, band);
1139 return 0;
1142 static int __sock_create(struct net *net, int family, int type, int protocol,
1143 struct socket **res, int kern)
1145 int err;
1146 struct socket *sock;
1147 const struct net_proto_family *pf;
1150 * Check protocol is in range
1152 if (family < 0 || family >= NPROTO)
1153 return -EAFNOSUPPORT;
1154 if (type < 0 || type >= SOCK_MAX)
1155 return -EINVAL;
1157 /* Compatibility.
1159 This uglymoron is moved from INET layer to here to avoid
1160 deadlock in module load.
1162 if (family == PF_INET && type == SOCK_PACKET) {
1163 static int warned;
1164 if (!warned) {
1165 warned = 1;
1166 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1167 current->comm);
1169 family = PF_PACKET;
1172 err = security_socket_create(family, type, protocol, kern);
1173 if (err)
1174 return err;
1177 * Allocate the socket and allow the family to set things up. if
1178 * the protocol is 0, the family is instructed to select an appropriate
1179 * default.
1181 sock = sock_alloc();
1182 if (!sock) {
1183 if (net_ratelimit())
1184 printk(KERN_WARNING "socket: no more sockets\n");
1185 return -ENFILE; /* Not exactly a match, but its the
1186 closest posix thing */
1189 sock->type = type;
1191 #ifdef CONFIG_MODULES
1192 /* Attempt to load a protocol module if the find failed.
1194 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1195 * requested real, full-featured networking support upon configuration.
1196 * Otherwise module support will break!
1198 if (net_families[family] == NULL)
1199 request_module("net-pf-%d", family);
1200 #endif
1202 rcu_read_lock();
1203 pf = rcu_dereference(net_families[family]);
1204 err = -EAFNOSUPPORT;
1205 if (!pf)
1206 goto out_release;
1209 * We will call the ->create function, that possibly is in a loadable
1210 * module, so we have to bump that loadable module refcnt first.
1212 if (!try_module_get(pf->owner))
1213 goto out_release;
1215 /* Now protected by module ref count */
1216 rcu_read_unlock();
1218 err = pf->create(net, sock, protocol);
1219 if (err < 0)
1220 goto out_module_put;
1223 * Now to bump the refcnt of the [loadable] module that owns this
1224 * socket at sock_release time we decrement its refcnt.
1226 if (!try_module_get(sock->ops->owner))
1227 goto out_module_busy;
1230 * Now that we're done with the ->create function, the [loadable]
1231 * module can have its refcnt decremented
1233 module_put(pf->owner);
1234 err = security_socket_post_create(sock, family, type, protocol, kern);
1235 if (err)
1236 goto out_sock_release;
1237 *res = sock;
1239 return 0;
1241 out_module_busy:
1242 err = -EAFNOSUPPORT;
1243 out_module_put:
1244 sock->ops = NULL;
1245 module_put(pf->owner);
1246 out_sock_release:
1247 sock_release(sock);
1248 return err;
1250 out_release:
1251 rcu_read_unlock();
1252 goto out_sock_release;
1255 int sock_create(int family, int type, int protocol, struct socket **res)
1257 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1260 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1262 return __sock_create(&init_net, family, type, protocol, res, 1);
1265 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1267 int retval;
1268 struct socket *sock;
1269 int flags;
1271 /* Check the SOCK_* constants for consistency. */
1272 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1273 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1274 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1275 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1277 flags = type & ~SOCK_TYPE_MASK;
1278 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1279 return -EINVAL;
1280 type &= SOCK_TYPE_MASK;
1282 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1283 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1285 retval = sock_create(family, type, protocol, &sock);
1286 if (retval < 0)
1287 goto out;
1289 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1290 if (retval < 0)
1291 goto out_release;
1293 out:
1294 /* It may be already another descriptor 8) Not kernel problem. */
1295 return retval;
1297 out_release:
1298 sock_release(sock);
1299 return retval;
1303 * Create a pair of connected sockets.
1306 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1307 int __user *, usockvec)
1309 struct socket *sock1, *sock2;
1310 int fd1, fd2, err;
1311 struct file *newfile1, *newfile2;
1312 int flags;
1314 flags = type & ~SOCK_TYPE_MASK;
1315 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1316 return -EINVAL;
1317 type &= SOCK_TYPE_MASK;
1319 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1320 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1323 * Obtain the first socket and check if the underlying protocol
1324 * supports the socketpair call.
1327 err = sock_create(family, type, protocol, &sock1);
1328 if (err < 0)
1329 goto out;
1331 err = sock_create(family, type, protocol, &sock2);
1332 if (err < 0)
1333 goto out_release_1;
1335 err = sock1->ops->socketpair(sock1, sock2);
1336 if (err < 0)
1337 goto out_release_both;
1339 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
1340 if (unlikely(fd1 < 0)) {
1341 err = fd1;
1342 goto out_release_both;
1345 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
1346 if (unlikely(fd2 < 0)) {
1347 err = fd2;
1348 put_filp(newfile1);
1349 put_unused_fd(fd1);
1350 goto out_release_both;
1353 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
1354 if (unlikely(err < 0)) {
1355 goto out_fd2;
1358 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
1359 if (unlikely(err < 0)) {
1360 fput(newfile1);
1361 goto out_fd1;
1364 audit_fd_pair(fd1, fd2);
1365 fd_install(fd1, newfile1);
1366 fd_install(fd2, newfile2);
1367 /* fd1 and fd2 may be already another descriptors.
1368 * Not kernel problem.
1371 err = put_user(fd1, &usockvec[0]);
1372 if (!err)
1373 err = put_user(fd2, &usockvec[1]);
1374 if (!err)
1375 return 0;
1377 sys_close(fd2);
1378 sys_close(fd1);
1379 return err;
1381 out_release_both:
1382 sock_release(sock2);
1383 out_release_1:
1384 sock_release(sock1);
1385 out:
1386 return err;
1388 out_fd2:
1389 put_filp(newfile1);
1390 sock_release(sock1);
1391 out_fd1:
1392 put_filp(newfile2);
1393 sock_release(sock2);
1394 put_unused_fd(fd1);
1395 put_unused_fd(fd2);
1396 goto out;
1400 * Bind a name to a socket. Nothing much to do here since it's
1401 * the protocol's responsibility to handle the local address.
1403 * We move the socket address to kernel space before we call
1404 * the protocol layer (having also checked the address is ok).
1407 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1409 struct socket *sock;
1410 struct sockaddr_storage address;
1411 int err, fput_needed;
1413 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1414 if (sock) {
1415 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1416 if (err >= 0) {
1417 err = security_socket_bind(sock,
1418 (struct sockaddr *)&address,
1419 addrlen);
1420 if (!err)
1421 err = sock->ops->bind(sock,
1422 (struct sockaddr *)
1423 &address, addrlen);
1425 fput_light(sock->file, fput_needed);
1427 return err;
1431 * Perform a listen. Basically, we allow the protocol to do anything
1432 * necessary for a listen, and if that works, we mark the socket as
1433 * ready for listening.
1436 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1438 struct socket *sock;
1439 int err, fput_needed;
1440 int somaxconn;
1442 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1443 if (sock) {
1444 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1445 if ((unsigned)backlog > somaxconn)
1446 backlog = somaxconn;
1448 err = security_socket_listen(sock, backlog);
1449 if (!err)
1450 err = sock->ops->listen(sock, backlog);
1452 fput_light(sock->file, fput_needed);
1454 return err;
1458 * For accept, we attempt to create a new socket, set up the link
1459 * with the client, wake up the client, then return the new
1460 * connected fd. We collect the address of the connector in kernel
1461 * space and move it to user at the very end. This is unclean because
1462 * we open the socket then return an error.
1464 * 1003.1g adds the ability to recvmsg() to query connection pending
1465 * status to recvmsg. We need to add that support in a way thats
1466 * clean when we restucture accept also.
1469 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1470 int __user *, upeer_addrlen, int, flags)
1472 struct socket *sock, *newsock;
1473 struct file *newfile;
1474 int err, len, newfd, fput_needed;
1475 struct sockaddr_storage address;
1477 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1478 return -EINVAL;
1480 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1481 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1483 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1484 if (!sock)
1485 goto out;
1487 err = -ENFILE;
1488 if (!(newsock = sock_alloc()))
1489 goto out_put;
1491 newsock->type = sock->type;
1492 newsock->ops = sock->ops;
1495 * We don't need try_module_get here, as the listening socket (sock)
1496 * has the protocol module (sock->ops->owner) held.
1498 __module_get(newsock->ops->owner);
1500 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
1501 if (unlikely(newfd < 0)) {
1502 err = newfd;
1503 sock_release(newsock);
1504 goto out_put;
1507 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
1508 if (err < 0)
1509 goto out_fd_simple;
1511 err = security_socket_accept(sock, newsock);
1512 if (err)
1513 goto out_fd;
1515 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1516 if (err < 0)
1517 goto out_fd;
1519 if (upeer_sockaddr) {
1520 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1521 &len, 2) < 0) {
1522 err = -ECONNABORTED;
1523 goto out_fd;
1525 err = move_addr_to_user((struct sockaddr *)&address,
1526 len, upeer_sockaddr, upeer_addrlen);
1527 if (err < 0)
1528 goto out_fd;
1531 /* File flags are not inherited via accept() unlike another OSes. */
1533 fd_install(newfd, newfile);
1534 err = newfd;
1536 out_put:
1537 fput_light(sock->file, fput_needed);
1538 out:
1539 return err;
1540 out_fd_simple:
1541 sock_release(newsock);
1542 put_filp(newfile);
1543 put_unused_fd(newfd);
1544 goto out_put;
1545 out_fd:
1546 fput(newfile);
1547 put_unused_fd(newfd);
1548 goto out_put;
1551 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1552 int __user *, upeer_addrlen)
1554 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1558 * Attempt to connect to a socket with the server address. The address
1559 * is in user space so we verify it is OK and move it to kernel space.
1561 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1562 * break bindings
1564 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1565 * other SEQPACKET protocols that take time to connect() as it doesn't
1566 * include the -EINPROGRESS status for such sockets.
1569 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1570 int, addrlen)
1572 struct socket *sock;
1573 struct sockaddr_storage address;
1574 int err, fput_needed;
1576 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1577 if (!sock)
1578 goto out;
1579 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1580 if (err < 0)
1581 goto out_put;
1583 err =
1584 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1585 if (err)
1586 goto out_put;
1588 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1589 sock->file->f_flags);
1590 out_put:
1591 fput_light(sock->file, fput_needed);
1592 out:
1593 return err;
1597 * Get the local address ('name') of a socket object. Move the obtained
1598 * name to user space.
1601 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1602 int __user *, usockaddr_len)
1604 struct socket *sock;
1605 struct sockaddr_storage address;
1606 int len, err, fput_needed;
1608 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1609 if (!sock)
1610 goto out;
1612 err = security_socket_getsockname(sock);
1613 if (err)
1614 goto out_put;
1616 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1617 if (err)
1618 goto out_put;
1619 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1621 out_put:
1622 fput_light(sock->file, fput_needed);
1623 out:
1624 return err;
1628 * Get the remote address ('name') of a socket object. Move the obtained
1629 * name to user space.
1632 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1633 int __user *, usockaddr_len)
1635 struct socket *sock;
1636 struct sockaddr_storage address;
1637 int len, err, fput_needed;
1639 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1640 if (sock != NULL) {
1641 err = security_socket_getpeername(sock);
1642 if (err) {
1643 fput_light(sock->file, fput_needed);
1644 return err;
1647 err =
1648 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1650 if (!err)
1651 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1652 usockaddr_len);
1653 fput_light(sock->file, fput_needed);
1655 return err;
1659 * Send a datagram to a given address. We move the address into kernel
1660 * space and check the user space data area is readable before invoking
1661 * the protocol.
1664 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1665 unsigned, flags, struct sockaddr __user *, addr,
1666 int, addr_len)
1668 struct socket *sock;
1669 struct sockaddr_storage address;
1670 int err;
1671 struct msghdr msg;
1672 struct iovec iov;
1673 int fput_needed;
1675 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1676 if (!sock)
1677 goto out;
1679 iov.iov_base = buff;
1680 iov.iov_len = len;
1681 msg.msg_name = NULL;
1682 msg.msg_iov = &iov;
1683 msg.msg_iovlen = 1;
1684 msg.msg_control = NULL;
1685 msg.msg_controllen = 0;
1686 msg.msg_namelen = 0;
1687 if (addr) {
1688 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1689 if (err < 0)
1690 goto out_put;
1691 msg.msg_name = (struct sockaddr *)&address;
1692 msg.msg_namelen = addr_len;
1694 if (sock->file->f_flags & O_NONBLOCK)
1695 flags |= MSG_DONTWAIT;
1696 msg.msg_flags = flags;
1697 err = sock_sendmsg(sock, &msg, len);
1699 out_put:
1700 fput_light(sock->file, fput_needed);
1701 out:
1702 return err;
1706 * Send a datagram down a socket.
1709 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1710 unsigned, flags)
1712 return sys_sendto(fd, buff, len, flags, NULL, 0);
1716 * Receive a frame from the socket and optionally record the address of the
1717 * sender. We verify the buffers are writable and if needed move the
1718 * sender address from kernel to user space.
1721 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1722 unsigned, flags, struct sockaddr __user *, addr,
1723 int __user *, addr_len)
1725 struct socket *sock;
1726 struct iovec iov;
1727 struct msghdr msg;
1728 struct sockaddr_storage address;
1729 int err, err2;
1730 int fput_needed;
1732 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1733 if (!sock)
1734 goto out;
1736 msg.msg_control = NULL;
1737 msg.msg_controllen = 0;
1738 msg.msg_iovlen = 1;
1739 msg.msg_iov = &iov;
1740 iov.iov_len = size;
1741 iov.iov_base = ubuf;
1742 msg.msg_name = (struct sockaddr *)&address;
1743 msg.msg_namelen = sizeof(address);
1744 if (sock->file->f_flags & O_NONBLOCK)
1745 flags |= MSG_DONTWAIT;
1746 err = sock_recvmsg(sock, &msg, size, flags);
1748 if (err >= 0 && addr != NULL) {
1749 err2 = move_addr_to_user((struct sockaddr *)&address,
1750 msg.msg_namelen, addr, addr_len);
1751 if (err2 < 0)
1752 err = err2;
1755 fput_light(sock->file, fput_needed);
1756 out:
1757 return err;
1761 * Receive a datagram from a socket.
1764 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1765 unsigned flags)
1767 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1771 * Set a socket option. Because we don't know the option lengths we have
1772 * to pass the user mode parameter for the protocols to sort out.
1775 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1776 char __user *, optval, int, optlen)
1778 int err, fput_needed;
1779 struct socket *sock;
1781 if (optlen < 0)
1782 return -EINVAL;
1784 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1785 if (sock != NULL) {
1786 err = security_socket_setsockopt(sock, level, optname);
1787 if (err)
1788 goto out_put;
1790 if (level == SOL_SOCKET)
1791 err =
1792 sock_setsockopt(sock, level, optname, optval,
1793 optlen);
1794 else
1795 err =
1796 sock->ops->setsockopt(sock, level, optname, optval,
1797 optlen);
1798 out_put:
1799 fput_light(sock->file, fput_needed);
1801 return err;
1805 * Get a socket option. Because we don't know the option lengths we have
1806 * to pass a user mode parameter for the protocols to sort out.
1809 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1810 char __user *, optval, int __user *, optlen)
1812 int err, fput_needed;
1813 struct socket *sock;
1815 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1816 if (sock != NULL) {
1817 err = security_socket_getsockopt(sock, level, optname);
1818 if (err)
1819 goto out_put;
1821 if (level == SOL_SOCKET)
1822 err =
1823 sock_getsockopt(sock, level, optname, optval,
1824 optlen);
1825 else
1826 err =
1827 sock->ops->getsockopt(sock, level, optname, optval,
1828 optlen);
1829 out_put:
1830 fput_light(sock->file, fput_needed);
1832 return err;
1836 * Shutdown a socket.
1839 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1841 int err, fput_needed;
1842 struct socket *sock;
1844 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1845 if (sock != NULL) {
1846 err = security_socket_shutdown(sock, how);
1847 if (!err)
1848 err = sock->ops->shutdown(sock, how);
1849 fput_light(sock->file, fput_needed);
1851 return err;
1854 /* A couple of helpful macros for getting the address of the 32/64 bit
1855 * fields which are the same type (int / unsigned) on our platforms.
1857 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1858 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1859 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1862 * BSD sendmsg interface
1865 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1867 struct compat_msghdr __user *msg_compat =
1868 (struct compat_msghdr __user *)msg;
1869 struct socket *sock;
1870 struct sockaddr_storage address;
1871 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1872 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1873 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1874 /* 20 is size of ipv6_pktinfo */
1875 unsigned char *ctl_buf = ctl;
1876 struct msghdr msg_sys;
1877 int err, ctl_len, iov_size, total_len;
1878 int fput_needed;
1880 err = -EFAULT;
1881 if (MSG_CMSG_COMPAT & flags) {
1882 if (get_compat_msghdr(&msg_sys, msg_compat))
1883 return -EFAULT;
1885 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1886 return -EFAULT;
1888 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1889 if (!sock)
1890 goto out;
1892 /* do not move before msg_sys is valid */
1893 err = -EMSGSIZE;
1894 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1895 goto out_put;
1897 /* Check whether to allocate the iovec area */
1898 err = -ENOMEM;
1899 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1900 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1901 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1902 if (!iov)
1903 goto out_put;
1906 /* This will also move the address data into kernel space */
1907 if (MSG_CMSG_COMPAT & flags) {
1908 err = verify_compat_iovec(&msg_sys, iov,
1909 (struct sockaddr *)&address,
1910 VERIFY_READ);
1911 } else
1912 err = verify_iovec(&msg_sys, iov,
1913 (struct sockaddr *)&address,
1914 VERIFY_READ);
1915 if (err < 0)
1916 goto out_freeiov;
1917 total_len = err;
1919 err = -ENOBUFS;
1921 if (msg_sys.msg_controllen > INT_MAX)
1922 goto out_freeiov;
1923 ctl_len = msg_sys.msg_controllen;
1924 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1925 err =
1926 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1927 sizeof(ctl));
1928 if (err)
1929 goto out_freeiov;
1930 ctl_buf = msg_sys.msg_control;
1931 ctl_len = msg_sys.msg_controllen;
1932 } else if (ctl_len) {
1933 if (ctl_len > sizeof(ctl)) {
1934 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1935 if (ctl_buf == NULL)
1936 goto out_freeiov;
1938 err = -EFAULT;
1940 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1941 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1942 * checking falls down on this.
1944 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1945 ctl_len))
1946 goto out_freectl;
1947 msg_sys.msg_control = ctl_buf;
1949 msg_sys.msg_flags = flags;
1951 if (sock->file->f_flags & O_NONBLOCK)
1952 msg_sys.msg_flags |= MSG_DONTWAIT;
1953 err = sock_sendmsg(sock, &msg_sys, total_len);
1955 out_freectl:
1956 if (ctl_buf != ctl)
1957 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1958 out_freeiov:
1959 if (iov != iovstack)
1960 sock_kfree_s(sock->sk, iov, iov_size);
1961 out_put:
1962 fput_light(sock->file, fput_needed);
1963 out:
1964 return err;
1968 * BSD recvmsg interface
1971 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
1972 unsigned int, flags)
1974 struct compat_msghdr __user *msg_compat =
1975 (struct compat_msghdr __user *)msg;
1976 struct socket *sock;
1977 struct iovec iovstack[UIO_FASTIOV];
1978 struct iovec *iov = iovstack;
1979 struct msghdr msg_sys;
1980 unsigned long cmsg_ptr;
1981 int err, iov_size, total_len, len;
1982 int fput_needed;
1984 /* kernel mode address */
1985 struct sockaddr_storage addr;
1987 /* user mode address pointers */
1988 struct sockaddr __user *uaddr;
1989 int __user *uaddr_len;
1991 if (MSG_CMSG_COMPAT & flags) {
1992 if (get_compat_msghdr(&msg_sys, msg_compat))
1993 return -EFAULT;
1995 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1996 return -EFAULT;
1998 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1999 if (!sock)
2000 goto out;
2002 err = -EMSGSIZE;
2003 if (msg_sys.msg_iovlen > UIO_MAXIOV)
2004 goto out_put;
2006 /* Check whether to allocate the iovec area */
2007 err = -ENOMEM;
2008 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2009 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2010 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2011 if (!iov)
2012 goto out_put;
2016 * Save the user-mode address (verify_iovec will change the
2017 * kernel msghdr to use the kernel address space)
2020 uaddr = (__force void __user *)msg_sys.msg_name;
2021 uaddr_len = COMPAT_NAMELEN(msg);
2022 if (MSG_CMSG_COMPAT & flags) {
2023 err = verify_compat_iovec(&msg_sys, iov,
2024 (struct sockaddr *)&addr,
2025 VERIFY_WRITE);
2026 } else
2027 err = verify_iovec(&msg_sys, iov,
2028 (struct sockaddr *)&addr,
2029 VERIFY_WRITE);
2030 if (err < 0)
2031 goto out_freeiov;
2032 total_len = err;
2034 cmsg_ptr = (unsigned long)msg_sys.msg_control;
2035 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2037 if (sock->file->f_flags & O_NONBLOCK)
2038 flags |= MSG_DONTWAIT;
2039 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2040 if (err < 0)
2041 goto out_freeiov;
2042 len = err;
2044 if (uaddr != NULL) {
2045 err = move_addr_to_user((struct sockaddr *)&addr,
2046 msg_sys.msg_namelen, uaddr,
2047 uaddr_len);
2048 if (err < 0)
2049 goto out_freeiov;
2051 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2052 COMPAT_FLAGS(msg));
2053 if (err)
2054 goto out_freeiov;
2055 if (MSG_CMSG_COMPAT & flags)
2056 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2057 &msg_compat->msg_controllen);
2058 else
2059 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2060 &msg->msg_controllen);
2061 if (err)
2062 goto out_freeiov;
2063 err = len;
2065 out_freeiov:
2066 if (iov != iovstack)
2067 sock_kfree_s(sock->sk, iov, iov_size);
2068 out_put:
2069 fput_light(sock->file, fput_needed);
2070 out:
2071 return err;
2074 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2076 /* Argument list sizes for sys_socketcall */
2077 #define AL(x) ((x) * sizeof(unsigned long))
2078 static const unsigned char nargs[19]={
2079 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2080 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2081 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2082 AL(4)
2085 #undef AL
2088 * System call vectors.
2090 * Argument checking cleaned up. Saved 20% in size.
2091 * This function doesn't need to set the kernel lock because
2092 * it is set by the callees.
2095 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2097 unsigned long a[6];
2098 unsigned long a0, a1;
2099 int err;
2100 unsigned int len;
2102 if (call < 1 || call > SYS_ACCEPT4)
2103 return -EINVAL;
2105 len = nargs[call];
2106 if (len > sizeof(a))
2107 return -EINVAL;
2109 /* copy_from_user should be SMP safe. */
2110 if (copy_from_user(a, args, len))
2111 return -EFAULT;
2113 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2115 a0 = a[0];
2116 a1 = a[1];
2118 switch (call) {
2119 case SYS_SOCKET:
2120 err = sys_socket(a0, a1, a[2]);
2121 break;
2122 case SYS_BIND:
2123 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2124 break;
2125 case SYS_CONNECT:
2126 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2127 break;
2128 case SYS_LISTEN:
2129 err = sys_listen(a0, a1);
2130 break;
2131 case SYS_ACCEPT:
2132 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2133 (int __user *)a[2], 0);
2134 break;
2135 case SYS_GETSOCKNAME:
2136 err =
2137 sys_getsockname(a0, (struct sockaddr __user *)a1,
2138 (int __user *)a[2]);
2139 break;
2140 case SYS_GETPEERNAME:
2141 err =
2142 sys_getpeername(a0, (struct sockaddr __user *)a1,
2143 (int __user *)a[2]);
2144 break;
2145 case SYS_SOCKETPAIR:
2146 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2147 break;
2148 case SYS_SEND:
2149 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2150 break;
2151 case SYS_SENDTO:
2152 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2153 (struct sockaddr __user *)a[4], a[5]);
2154 break;
2155 case SYS_RECV:
2156 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2157 break;
2158 case SYS_RECVFROM:
2159 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2160 (struct sockaddr __user *)a[4],
2161 (int __user *)a[5]);
2162 break;
2163 case SYS_SHUTDOWN:
2164 err = sys_shutdown(a0, a1);
2165 break;
2166 case SYS_SETSOCKOPT:
2167 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2168 break;
2169 case SYS_GETSOCKOPT:
2170 err =
2171 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2172 (int __user *)a[4]);
2173 break;
2174 case SYS_SENDMSG:
2175 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2176 break;
2177 case SYS_RECVMSG:
2178 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2179 break;
2180 case SYS_ACCEPT4:
2181 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2182 (int __user *)a[2], a[3]);
2183 break;
2184 default:
2185 err = -EINVAL;
2186 break;
2188 return err;
2191 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2194 * sock_register - add a socket protocol handler
2195 * @ops: description of protocol
2197 * This function is called by a protocol handler that wants to
2198 * advertise its address family, and have it linked into the
2199 * socket interface. The value ops->family coresponds to the
2200 * socket system call protocol family.
2202 int sock_register(const struct net_proto_family *ops)
2204 int err;
2206 if (ops->family >= NPROTO) {
2207 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2208 NPROTO);
2209 return -ENOBUFS;
2212 spin_lock(&net_family_lock);
2213 if (net_families[ops->family])
2214 err = -EEXIST;
2215 else {
2216 net_families[ops->family] = ops;
2217 err = 0;
2219 spin_unlock(&net_family_lock);
2221 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2222 return err;
2226 * sock_unregister - remove a protocol handler
2227 * @family: protocol family to remove
2229 * This function is called by a protocol handler that wants to
2230 * remove its address family, and have it unlinked from the
2231 * new socket creation.
2233 * If protocol handler is a module, then it can use module reference
2234 * counts to protect against new references. If protocol handler is not
2235 * a module then it needs to provide its own protection in
2236 * the ops->create routine.
2238 void sock_unregister(int family)
2240 BUG_ON(family < 0 || family >= NPROTO);
2242 spin_lock(&net_family_lock);
2243 net_families[family] = NULL;
2244 spin_unlock(&net_family_lock);
2246 synchronize_rcu();
2248 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2251 static int __init sock_init(void)
2254 * Initialize sock SLAB cache.
2257 sk_init();
2260 * Initialize skbuff SLAB cache
2262 skb_init();
2265 * Initialize the protocols module.
2268 init_inodecache();
2269 register_filesystem(&sock_fs_type);
2270 sock_mnt = kern_mount(&sock_fs_type);
2272 /* The real protocol initialization is performed in later initcalls.
2275 #ifdef CONFIG_NETFILTER
2276 netfilter_init();
2277 #endif
2279 return 0;
2282 core_initcall(sock_init); /* early initcall */
2284 #ifdef CONFIG_PROC_FS
2285 void socket_seq_show(struct seq_file *seq)
2287 int cpu;
2288 int counter = 0;
2290 for_each_possible_cpu(cpu)
2291 counter += per_cpu(sockets_in_use, cpu);
2293 /* It can be negative, by the way. 8) */
2294 if (counter < 0)
2295 counter = 0;
2297 seq_printf(seq, "sockets: used %d\n", counter);
2299 #endif /* CONFIG_PROC_FS */
2301 #ifdef CONFIG_COMPAT
2302 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2303 unsigned long arg)
2305 struct socket *sock = file->private_data;
2306 int ret = -ENOIOCTLCMD;
2307 struct sock *sk;
2308 struct net *net;
2310 sk = sock->sk;
2311 net = sock_net(sk);
2313 if (sock->ops->compat_ioctl)
2314 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2316 if (ret == -ENOIOCTLCMD &&
2317 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2318 ret = compat_wext_handle_ioctl(net, cmd, arg);
2320 return ret;
2322 #endif
2324 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2326 return sock->ops->bind(sock, addr, addrlen);
2329 int kernel_listen(struct socket *sock, int backlog)
2331 return sock->ops->listen(sock, backlog);
2334 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2336 struct sock *sk = sock->sk;
2337 int err;
2339 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2340 newsock);
2341 if (err < 0)
2342 goto done;
2344 err = sock->ops->accept(sock, *newsock, flags);
2345 if (err < 0) {
2346 sock_release(*newsock);
2347 *newsock = NULL;
2348 goto done;
2351 (*newsock)->ops = sock->ops;
2352 __module_get((*newsock)->ops->owner);
2354 done:
2355 return err;
2358 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2359 int flags)
2361 return sock->ops->connect(sock, addr, addrlen, flags);
2364 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2365 int *addrlen)
2367 return sock->ops->getname(sock, addr, addrlen, 0);
2370 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2371 int *addrlen)
2373 return sock->ops->getname(sock, addr, addrlen, 1);
2376 int kernel_getsockopt(struct socket *sock, int level, int optname,
2377 char *optval, int *optlen)
2379 mm_segment_t oldfs = get_fs();
2380 int err;
2382 set_fs(KERNEL_DS);
2383 if (level == SOL_SOCKET)
2384 err = sock_getsockopt(sock, level, optname, optval, optlen);
2385 else
2386 err = sock->ops->getsockopt(sock, level, optname, optval,
2387 optlen);
2388 set_fs(oldfs);
2389 return err;
2392 int kernel_setsockopt(struct socket *sock, int level, int optname,
2393 char *optval, unsigned int optlen)
2395 mm_segment_t oldfs = get_fs();
2396 int err;
2398 set_fs(KERNEL_DS);
2399 if (level == SOL_SOCKET)
2400 err = sock_setsockopt(sock, level, optname, optval, optlen);
2401 else
2402 err = sock->ops->setsockopt(sock, level, optname, optval,
2403 optlen);
2404 set_fs(oldfs);
2405 return err;
2408 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2409 size_t size, int flags)
2411 if (sock->ops->sendpage)
2412 return sock->ops->sendpage(sock, page, offset, size, flags);
2414 return sock_no_sendpage(sock, page, offset, size, flags);
2417 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2419 mm_segment_t oldfs = get_fs();
2420 int err;
2422 set_fs(KERNEL_DS);
2423 err = sock->ops->ioctl(sock, cmd, arg);
2424 set_fs(oldfs);
2426 return err;
2429 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2431 return sock->ops->shutdown(sock, how);
2434 EXPORT_SYMBOL(sock_create);
2435 EXPORT_SYMBOL(sock_create_kern);
2436 EXPORT_SYMBOL(sock_create_lite);
2437 EXPORT_SYMBOL(sock_map_fd);
2438 EXPORT_SYMBOL(sock_recvmsg);
2439 EXPORT_SYMBOL(sock_register);
2440 EXPORT_SYMBOL(sock_release);
2441 EXPORT_SYMBOL(sock_sendmsg);
2442 EXPORT_SYMBOL(sock_unregister);
2443 EXPORT_SYMBOL(sock_wake_async);
2444 EXPORT_SYMBOL(sockfd_lookup);
2445 EXPORT_SYMBOL(kernel_sendmsg);
2446 EXPORT_SYMBOL(kernel_recvmsg);
2447 EXPORT_SYMBOL(kernel_bind);
2448 EXPORT_SYMBOL(kernel_listen);
2449 EXPORT_SYMBOL(kernel_accept);
2450 EXPORT_SYMBOL(kernel_connect);
2451 EXPORT_SYMBOL(kernel_getsockname);
2452 EXPORT_SYMBOL(kernel_getpeername);
2453 EXPORT_SYMBOL(kernel_getsockopt);
2454 EXPORT_SYMBOL(kernel_setsockopt);
2455 EXPORT_SYMBOL(kernel_sendpage);
2456 EXPORT_SYMBOL(kernel_sock_ioctl);
2457 EXPORT_SYMBOL(kernel_sock_shutdown);