ACPI: thinkpad-acpi: add development version tag
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / net / socket.c
blobd1460304b94bca4feb414567e4d3b128f177a129
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/thread_info.h>
73 #include <linux/wanrouter.h>
74 #include <linux/if_bridge.h>
75 #include <linux/if_frad.h>
76 #include <linux/if_vlan.h>
77 #include <linux/init.h>
78 #include <linux/poll.h>
79 #include <linux/cache.h>
80 #include <linux/module.h>
81 #include <linux/highmem.h>
82 #include <linux/mount.h>
83 #include <linux/security.h>
84 #include <linux/syscalls.h>
85 #include <linux/compat.h>
86 #include <linux/kmod.h>
87 #include <linux/audit.h>
88 #include <linux/wireless.h>
89 #include <linux/nsproxy.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 #define SOCKFS_MAGIC 0x534F434B
241 static struct kmem_cache *sock_inode_cachep __read_mostly;
243 static struct inode *sock_alloc_inode(struct super_block *sb)
245 struct socket_alloc *ei;
247 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
248 if (!ei)
249 return NULL;
250 init_waitqueue_head(&ei->socket.wait);
252 ei->socket.fasync_list = NULL;
253 ei->socket.state = SS_UNCONNECTED;
254 ei->socket.flags = 0;
255 ei->socket.ops = NULL;
256 ei->socket.sk = NULL;
257 ei->socket.file = NULL;
259 return &ei->vfs_inode;
262 static void sock_destroy_inode(struct inode *inode)
264 kmem_cache_free(sock_inode_cachep,
265 container_of(inode, struct socket_alloc, vfs_inode));
268 static void init_once(void *foo)
270 struct socket_alloc *ei = (struct socket_alloc *)foo;
272 inode_init_once(&ei->vfs_inode);
275 static int init_inodecache(void)
277 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
278 sizeof(struct socket_alloc),
280 (SLAB_HWCACHE_ALIGN |
281 SLAB_RECLAIM_ACCOUNT |
282 SLAB_MEM_SPREAD),
283 init_once);
284 if (sock_inode_cachep == NULL)
285 return -ENOMEM;
286 return 0;
289 static struct super_operations sockfs_ops = {
290 .alloc_inode = sock_alloc_inode,
291 .destroy_inode =sock_destroy_inode,
292 .statfs = simple_statfs,
295 static int sockfs_get_sb(struct file_system_type *fs_type,
296 int flags, const char *dev_name, void *data,
297 struct vfsmount *mnt)
299 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
300 mnt);
303 static struct vfsmount *sock_mnt __read_mostly;
305 static struct file_system_type sock_fs_type = {
306 .name = "sockfs",
307 .get_sb = sockfs_get_sb,
308 .kill_sb = kill_anon_super,
311 static int sockfs_delete_dentry(struct dentry *dentry)
314 * At creation time, we pretended this dentry was hashed
315 * (by clearing DCACHE_UNHASHED bit in d_flags)
316 * At delete time, we restore the truth : not hashed.
317 * (so that dput() can proceed correctly)
319 dentry->d_flags |= DCACHE_UNHASHED;
320 return 0;
324 * sockfs_dname() is called from d_path().
326 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
328 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
329 dentry->d_inode->i_ino);
332 static struct dentry_operations sockfs_dentry_operations = {
333 .d_delete = sockfs_delete_dentry,
334 .d_dname = sockfs_dname,
338 * Obtains the first available file descriptor and sets it up for use.
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
354 static int sock_alloc_fd(struct file **filep, int flags)
356 int fd;
358 fd = get_unused_fd_flags(flags);
359 if (likely(fd >= 0)) {
360 struct file *file = get_empty_filp();
362 *filep = file;
363 if (unlikely(!file)) {
364 put_unused_fd(fd);
365 return -ENFILE;
367 } else
368 *filep = NULL;
369 return fd;
372 static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
374 struct dentry *dentry;
375 struct qstr name = { .name = "" };
377 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
378 if (unlikely(!dentry))
379 return -ENOMEM;
381 dentry->d_op = &sockfs_dentry_operations;
383 * We dont want to push this dentry into global dentry hash table.
384 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
385 * This permits a working /proc/$pid/fd/XXX on sockets
387 dentry->d_flags &= ~DCACHE_UNHASHED;
388 d_instantiate(dentry, SOCK_INODE(sock));
390 sock->file = file;
391 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
392 &socket_file_ops);
393 SOCK_INODE(sock)->i_fop = &socket_file_ops;
394 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
395 file->f_pos = 0;
396 file->private_data = sock;
398 return 0;
401 int sock_map_fd(struct socket *sock, int flags)
403 struct file *newfile;
404 int fd = sock_alloc_fd(&newfile, flags);
406 if (likely(fd >= 0)) {
407 int err = sock_attach_fd(sock, newfile, flags);
409 if (unlikely(err < 0)) {
410 put_filp(newfile);
411 put_unused_fd(fd);
412 return err;
414 fd_install(fd, newfile);
416 return fd;
419 static struct socket *sock_from_file(struct file *file, int *err)
421 if (file->f_op == &socket_file_ops)
422 return file->private_data; /* set in sock_map_fd */
424 *err = -ENOTSOCK;
425 return NULL;
429 * sockfd_lookup - Go from a file number to its socket slot
430 * @fd: file handle
431 * @err: pointer to an error code return
433 * The file handle passed in is locked and the socket it is bound
434 * too is returned. If an error occurs the err pointer is overwritten
435 * with a negative errno code and NULL is returned. The function checks
436 * for both invalid handles and passing a handle which is not a socket.
438 * On a success the socket object pointer is returned.
441 struct socket *sockfd_lookup(int fd, int *err)
443 struct file *file;
444 struct socket *sock;
446 file = fget(fd);
447 if (!file) {
448 *err = -EBADF;
449 return NULL;
452 sock = sock_from_file(file, err);
453 if (!sock)
454 fput(file);
455 return sock;
458 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
460 struct file *file;
461 struct socket *sock;
463 *err = -EBADF;
464 file = fget_light(fd, fput_needed);
465 if (file) {
466 sock = sock_from_file(file, err);
467 if (sock)
468 return sock;
469 fput_light(file, *fput_needed);
471 return NULL;
475 * sock_alloc - allocate a socket
477 * Allocate a new inode and socket object. The two are bound together
478 * and initialised. The socket is then returned. If we are out of inodes
479 * NULL is returned.
482 static struct socket *sock_alloc(void)
484 struct inode *inode;
485 struct socket *sock;
487 inode = new_inode(sock_mnt->mnt_sb);
488 if (!inode)
489 return NULL;
491 sock = SOCKET_I(inode);
493 inode->i_mode = S_IFSOCK | S_IRWXUGO;
494 inode->i_uid = current->fsuid;
495 inode->i_gid = current->fsgid;
497 get_cpu_var(sockets_in_use)++;
498 put_cpu_var(sockets_in_use);
499 return sock;
503 * In theory you can't get an open on this inode, but /proc provides
504 * a back door. Remember to keep it shut otherwise you'll let the
505 * creepy crawlies in.
508 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
510 return -ENXIO;
513 const struct file_operations bad_sock_fops = {
514 .owner = THIS_MODULE,
515 .open = sock_no_open,
519 * sock_release - close a socket
520 * @sock: socket to close
522 * The socket is released from the protocol stack if it has a release
523 * callback, and the inode is then released if the socket is bound to
524 * an inode not a file.
527 void sock_release(struct socket *sock)
529 if (sock->ops) {
530 struct module *owner = sock->ops->owner;
532 sock->ops->release(sock);
533 sock->ops = NULL;
534 module_put(owner);
537 if (sock->fasync_list)
538 printk(KERN_ERR "sock_release: fasync list not empty!\n");
540 get_cpu_var(sockets_in_use)--;
541 put_cpu_var(sockets_in_use);
542 if (!sock->file) {
543 iput(SOCK_INODE(sock));
544 return;
546 sock->file = NULL;
549 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
550 struct msghdr *msg, size_t size)
552 struct sock_iocb *si = kiocb_to_siocb(iocb);
553 int err;
555 si->sock = sock;
556 si->scm = NULL;
557 si->msg = msg;
558 si->size = size;
560 err = security_socket_sendmsg(sock, msg, size);
561 if (err)
562 return err;
564 return sock->ops->sendmsg(iocb, sock, msg, size);
567 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
569 struct kiocb iocb;
570 struct sock_iocb siocb;
571 int ret;
573 init_sync_kiocb(&iocb, NULL);
574 iocb.private = &siocb;
575 ret = __sock_sendmsg(&iocb, sock, msg, size);
576 if (-EIOCBQUEUED == ret)
577 ret = wait_on_sync_kiocb(&iocb);
578 return ret;
581 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
582 struct kvec *vec, size_t num, size_t size)
584 mm_segment_t oldfs = get_fs();
585 int result;
587 set_fs(KERNEL_DS);
589 * the following is safe, since for compiler definitions of kvec and
590 * iovec are identical, yielding the same in-core layout and alignment
592 msg->msg_iov = (struct iovec *)vec;
593 msg->msg_iovlen = num;
594 result = sock_sendmsg(sock, msg, size);
595 set_fs(oldfs);
596 return result;
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 ktime_t kt = skb->tstamp;
607 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
608 struct timeval tv;
609 /* Race occurred between timestamp enabling and packet
610 receiving. Fill in the current time for now. */
611 if (kt.tv64 == 0)
612 kt = ktime_get_real();
613 skb->tstamp = kt;
614 tv = ktime_to_timeval(kt);
615 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
616 } else {
617 struct timespec ts;
618 /* Race occurred between timestamp enabling and packet
619 receiving. Fill in the current time for now. */
620 if (kt.tv64 == 0)
621 kt = ktime_get_real();
622 skb->tstamp = kt;
623 ts = ktime_to_timespec(kt);
624 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
628 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
630 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
631 struct msghdr *msg, size_t size, int flags)
633 int err;
634 struct sock_iocb *si = kiocb_to_siocb(iocb);
636 si->sock = sock;
637 si->scm = NULL;
638 si->msg = msg;
639 si->size = size;
640 si->flags = flags;
642 err = security_socket_recvmsg(sock, msg, size, flags);
643 if (err)
644 return err;
646 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
649 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
650 size_t size, int flags)
652 struct kiocb iocb;
653 struct sock_iocb siocb;
654 int ret;
656 init_sync_kiocb(&iocb, NULL);
657 iocb.private = &siocb;
658 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
659 if (-EIOCBQUEUED == ret)
660 ret = wait_on_sync_kiocb(&iocb);
661 return ret;
664 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
665 struct kvec *vec, size_t num, size_t size, int flags)
667 mm_segment_t oldfs = get_fs();
668 int result;
670 set_fs(KERNEL_DS);
672 * the following is safe, since for compiler definitions of kvec and
673 * iovec are identical, yielding the same in-core layout and alignment
675 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
676 result = sock_recvmsg(sock, msg, size, flags);
677 set_fs(oldfs);
678 return result;
681 static void sock_aio_dtor(struct kiocb *iocb)
683 kfree(iocb->private);
686 static ssize_t sock_sendpage(struct file *file, struct page *page,
687 int offset, size_t size, loff_t *ppos, int more)
689 struct socket *sock;
690 int flags;
692 sock = file->private_data;
694 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
695 if (more)
696 flags |= MSG_MORE;
698 return kernel_sendpage(sock, page, offset, size, flags);
701 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
702 struct pipe_inode_info *pipe, size_t len,
703 unsigned int flags)
705 struct socket *sock = file->private_data;
707 if (unlikely(!sock->ops->splice_read))
708 return -EINVAL;
710 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
713 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
714 struct sock_iocb *siocb)
716 if (!is_sync_kiocb(iocb)) {
717 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
718 if (!siocb)
719 return NULL;
720 iocb->ki_dtor = sock_aio_dtor;
723 siocb->kiocb = iocb;
724 iocb->private = siocb;
725 return siocb;
728 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
729 struct file *file, const struct iovec *iov,
730 unsigned long nr_segs)
732 struct socket *sock = file->private_data;
733 size_t size = 0;
734 int i;
736 for (i = 0; i < nr_segs; i++)
737 size += iov[i].iov_len;
739 msg->msg_name = NULL;
740 msg->msg_namelen = 0;
741 msg->msg_control = NULL;
742 msg->msg_controllen = 0;
743 msg->msg_iov = (struct iovec *)iov;
744 msg->msg_iovlen = nr_segs;
745 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
747 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
750 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
751 unsigned long nr_segs, loff_t pos)
753 struct sock_iocb siocb, *x;
755 if (pos != 0)
756 return -ESPIPE;
758 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
759 return 0;
762 x = alloc_sock_iocb(iocb, &siocb);
763 if (!x)
764 return -ENOMEM;
765 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
768 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
769 struct file *file, const struct iovec *iov,
770 unsigned long nr_segs)
772 struct socket *sock = file->private_data;
773 size_t size = 0;
774 int i;
776 for (i = 0; i < nr_segs; i++)
777 size += iov[i].iov_len;
779 msg->msg_name = NULL;
780 msg->msg_namelen = 0;
781 msg->msg_control = NULL;
782 msg->msg_controllen = 0;
783 msg->msg_iov = (struct iovec *)iov;
784 msg->msg_iovlen = nr_segs;
785 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
786 if (sock->type == SOCK_SEQPACKET)
787 msg->msg_flags |= MSG_EOR;
789 return __sock_sendmsg(iocb, sock, msg, size);
792 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
793 unsigned long nr_segs, loff_t pos)
795 struct sock_iocb siocb, *x;
797 if (pos != 0)
798 return -ESPIPE;
800 x = alloc_sock_iocb(iocb, &siocb);
801 if (!x)
802 return -ENOMEM;
804 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
808 * Atomic setting of ioctl hooks to avoid race
809 * with module unload.
812 static DEFINE_MUTEX(br_ioctl_mutex);
813 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
815 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
817 mutex_lock(&br_ioctl_mutex);
818 br_ioctl_hook = hook;
819 mutex_unlock(&br_ioctl_mutex);
822 EXPORT_SYMBOL(brioctl_set);
824 static DEFINE_MUTEX(vlan_ioctl_mutex);
825 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
827 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
829 mutex_lock(&vlan_ioctl_mutex);
830 vlan_ioctl_hook = hook;
831 mutex_unlock(&vlan_ioctl_mutex);
834 EXPORT_SYMBOL(vlan_ioctl_set);
836 static DEFINE_MUTEX(dlci_ioctl_mutex);
837 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
839 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
841 mutex_lock(&dlci_ioctl_mutex);
842 dlci_ioctl_hook = hook;
843 mutex_unlock(&dlci_ioctl_mutex);
846 EXPORT_SYMBOL(dlci_ioctl_set);
849 * With an ioctl, arg may well be a user mode pointer, but we don't know
850 * what to do with it - that's up to the protocol still.
853 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
855 struct socket *sock;
856 struct sock *sk;
857 void __user *argp = (void __user *)arg;
858 int pid, err;
859 struct net *net;
861 sock = file->private_data;
862 sk = sock->sk;
863 net = sock_net(sk);
864 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
865 err = dev_ioctl(net, cmd, argp);
866 } else
867 #ifdef CONFIG_WIRELESS_EXT
868 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
869 err = dev_ioctl(net, cmd, argp);
870 } else
871 #endif /* CONFIG_WIRELESS_EXT */
872 switch (cmd) {
873 case FIOSETOWN:
874 case SIOCSPGRP:
875 err = -EFAULT;
876 if (get_user(pid, (int __user *)argp))
877 break;
878 err = f_setown(sock->file, pid, 1);
879 break;
880 case FIOGETOWN:
881 case SIOCGPGRP:
882 err = put_user(f_getown(sock->file),
883 (int __user *)argp);
884 break;
885 case SIOCGIFBR:
886 case SIOCSIFBR:
887 case SIOCBRADDBR:
888 case SIOCBRDELBR:
889 err = -ENOPKG;
890 if (!br_ioctl_hook)
891 request_module("bridge");
893 mutex_lock(&br_ioctl_mutex);
894 if (br_ioctl_hook)
895 err = br_ioctl_hook(net, cmd, argp);
896 mutex_unlock(&br_ioctl_mutex);
897 break;
898 case SIOCGIFVLAN:
899 case SIOCSIFVLAN:
900 err = -ENOPKG;
901 if (!vlan_ioctl_hook)
902 request_module("8021q");
904 mutex_lock(&vlan_ioctl_mutex);
905 if (vlan_ioctl_hook)
906 err = vlan_ioctl_hook(net, argp);
907 mutex_unlock(&vlan_ioctl_mutex);
908 break;
909 case SIOCADDDLCI:
910 case SIOCDELDLCI:
911 err = -ENOPKG;
912 if (!dlci_ioctl_hook)
913 request_module("dlci");
915 mutex_lock(&dlci_ioctl_mutex);
916 if (dlci_ioctl_hook)
917 err = dlci_ioctl_hook(cmd, argp);
918 mutex_unlock(&dlci_ioctl_mutex);
919 break;
920 default:
921 err = sock->ops->ioctl(sock, cmd, arg);
924 * If this ioctl is unknown try to hand it down
925 * to the NIC driver.
927 if (err == -ENOIOCTLCMD)
928 err = dev_ioctl(net, cmd, argp);
929 break;
931 return err;
934 int sock_create_lite(int family, int type, int protocol, struct socket **res)
936 int err;
937 struct socket *sock = NULL;
939 err = security_socket_create(family, type, protocol, 1);
940 if (err)
941 goto out;
943 sock = sock_alloc();
944 if (!sock) {
945 err = -ENOMEM;
946 goto out;
949 sock->type = type;
950 err = security_socket_post_create(sock, family, type, protocol, 1);
951 if (err)
952 goto out_release;
954 out:
955 *res = sock;
956 return err;
957 out_release:
958 sock_release(sock);
959 sock = NULL;
960 goto out;
963 /* No kernel lock held - perfect */
964 static unsigned int sock_poll(struct file *file, poll_table *wait)
966 struct socket *sock;
969 * We can't return errors to poll, so it's either yes or no.
971 sock = file->private_data;
972 return sock->ops->poll(file, sock, wait);
975 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
977 struct socket *sock = file->private_data;
979 return sock->ops->mmap(file, sock, vma);
982 static int sock_close(struct inode *inode, struct file *filp)
985 * It was possible the inode is NULL we were
986 * closing an unfinished socket.
989 if (!inode) {
990 printk(KERN_DEBUG "sock_close: NULL inode\n");
991 return 0;
993 sock_fasync(-1, filp, 0);
994 sock_release(SOCKET_I(inode));
995 return 0;
999 * Update the socket async list
1001 * Fasync_list locking strategy.
1003 * 1. fasync_list is modified only under process context socket lock
1004 * i.e. under semaphore.
1005 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1006 * or under socket lock.
1007 * 3. fasync_list can be used from softirq context, so that
1008 * modification under socket lock have to be enhanced with
1009 * write_lock_bh(&sk->sk_callback_lock).
1010 * --ANK (990710)
1013 static int sock_fasync(int fd, struct file *filp, int on)
1015 struct fasync_struct *fa, *fna = NULL, **prev;
1016 struct socket *sock;
1017 struct sock *sk;
1019 if (on) {
1020 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1021 if (fna == NULL)
1022 return -ENOMEM;
1025 sock = filp->private_data;
1027 sk = sock->sk;
1028 if (sk == NULL) {
1029 kfree(fna);
1030 return -EINVAL;
1033 lock_sock(sk);
1035 prev = &(sock->fasync_list);
1037 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1038 if (fa->fa_file == filp)
1039 break;
1041 if (on) {
1042 if (fa != NULL) {
1043 write_lock_bh(&sk->sk_callback_lock);
1044 fa->fa_fd = fd;
1045 write_unlock_bh(&sk->sk_callback_lock);
1047 kfree(fna);
1048 goto out;
1050 fna->fa_file = filp;
1051 fna->fa_fd = fd;
1052 fna->magic = FASYNC_MAGIC;
1053 fna->fa_next = sock->fasync_list;
1054 write_lock_bh(&sk->sk_callback_lock);
1055 sock->fasync_list = fna;
1056 write_unlock_bh(&sk->sk_callback_lock);
1057 } else {
1058 if (fa != NULL) {
1059 write_lock_bh(&sk->sk_callback_lock);
1060 *prev = fa->fa_next;
1061 write_unlock_bh(&sk->sk_callback_lock);
1062 kfree(fa);
1066 out:
1067 release_sock(sock->sk);
1068 return 0;
1071 /* This function may be called only under socket lock or callback_lock */
1073 int sock_wake_async(struct socket *sock, int how, int band)
1075 if (!sock || !sock->fasync_list)
1076 return -1;
1077 switch (how) {
1078 case SOCK_WAKE_WAITD:
1079 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1080 break;
1081 goto call_kill;
1082 case SOCK_WAKE_SPACE:
1083 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1084 break;
1085 /* fall through */
1086 case SOCK_WAKE_IO:
1087 call_kill:
1088 __kill_fasync(sock->fasync_list, SIGIO, band);
1089 break;
1090 case SOCK_WAKE_URG:
1091 __kill_fasync(sock->fasync_list, SIGURG, band);
1093 return 0;
1096 static int __sock_create(struct net *net, int family, int type, int protocol,
1097 struct socket **res, int kern)
1099 int err;
1100 struct socket *sock;
1101 const struct net_proto_family *pf;
1104 * Check protocol is in range
1106 if (family < 0 || family >= NPROTO)
1107 return -EAFNOSUPPORT;
1108 if (type < 0 || type >= SOCK_MAX)
1109 return -EINVAL;
1111 /* Compatibility.
1113 This uglymoron is moved from INET layer to here to avoid
1114 deadlock in module load.
1116 if (family == PF_INET && type == SOCK_PACKET) {
1117 static int warned;
1118 if (!warned) {
1119 warned = 1;
1120 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1121 current->comm);
1123 family = PF_PACKET;
1126 err = security_socket_create(family, type, protocol, kern);
1127 if (err)
1128 return err;
1131 * Allocate the socket and allow the family to set things up. if
1132 * the protocol is 0, the family is instructed to select an appropriate
1133 * default.
1135 sock = sock_alloc();
1136 if (!sock) {
1137 if (net_ratelimit())
1138 printk(KERN_WARNING "socket: no more sockets\n");
1139 return -ENFILE; /* Not exactly a match, but its the
1140 closest posix thing */
1143 sock->type = type;
1145 #if defined(CONFIG_KMOD)
1146 /* Attempt to load a protocol module if the find failed.
1148 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1149 * requested real, full-featured networking support upon configuration.
1150 * Otherwise module support will break!
1152 if (net_families[family] == NULL)
1153 request_module("net-pf-%d", family);
1154 #endif
1156 rcu_read_lock();
1157 pf = rcu_dereference(net_families[family]);
1158 err = -EAFNOSUPPORT;
1159 if (!pf)
1160 goto out_release;
1163 * We will call the ->create function, that possibly is in a loadable
1164 * module, so we have to bump that loadable module refcnt first.
1166 if (!try_module_get(pf->owner))
1167 goto out_release;
1169 /* Now protected by module ref count */
1170 rcu_read_unlock();
1172 err = pf->create(net, sock, protocol);
1173 if (err < 0)
1174 goto out_module_put;
1177 * Now to bump the refcnt of the [loadable] module that owns this
1178 * socket at sock_release time we decrement its refcnt.
1180 if (!try_module_get(sock->ops->owner))
1181 goto out_module_busy;
1184 * Now that we're done with the ->create function, the [loadable]
1185 * module can have its refcnt decremented
1187 module_put(pf->owner);
1188 err = security_socket_post_create(sock, family, type, protocol, kern);
1189 if (err)
1190 goto out_sock_release;
1191 *res = sock;
1193 return 0;
1195 out_module_busy:
1196 err = -EAFNOSUPPORT;
1197 out_module_put:
1198 sock->ops = NULL;
1199 module_put(pf->owner);
1200 out_sock_release:
1201 sock_release(sock);
1202 return err;
1204 out_release:
1205 rcu_read_unlock();
1206 goto out_sock_release;
1209 int sock_create(int family, int type, int protocol, struct socket **res)
1211 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1214 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1216 return __sock_create(&init_net, family, type, protocol, res, 1);
1219 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1221 int retval;
1222 struct socket *sock;
1223 int flags;
1225 /* Check the SOCK_* constants for consistency. */
1226 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1227 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1228 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1229 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1231 flags = type & ~SOCK_TYPE_MASK;
1232 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1233 return -EINVAL;
1234 type &= SOCK_TYPE_MASK;
1236 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1237 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1239 retval = sock_create(family, type, protocol, &sock);
1240 if (retval < 0)
1241 goto out;
1243 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1244 if (retval < 0)
1245 goto out_release;
1247 out:
1248 /* It may be already another descriptor 8) Not kernel problem. */
1249 return retval;
1251 out_release:
1252 sock_release(sock);
1253 return retval;
1257 * Create a pair of connected sockets.
1260 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1261 int __user *, usockvec)
1263 struct socket *sock1, *sock2;
1264 int fd1, fd2, err;
1265 struct file *newfile1, *newfile2;
1266 int flags;
1268 flags = type & ~SOCK_TYPE_MASK;
1269 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1270 return -EINVAL;
1271 type &= SOCK_TYPE_MASK;
1273 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1274 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1277 * Obtain the first socket and check if the underlying protocol
1278 * supports the socketpair call.
1281 err = sock_create(family, type, protocol, &sock1);
1282 if (err < 0)
1283 goto out;
1285 err = sock_create(family, type, protocol, &sock2);
1286 if (err < 0)
1287 goto out_release_1;
1289 err = sock1->ops->socketpair(sock1, sock2);
1290 if (err < 0)
1291 goto out_release_both;
1293 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
1294 if (unlikely(fd1 < 0)) {
1295 err = fd1;
1296 goto out_release_both;
1299 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
1300 if (unlikely(fd2 < 0)) {
1301 err = fd2;
1302 put_filp(newfile1);
1303 put_unused_fd(fd1);
1304 goto out_release_both;
1307 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
1308 if (unlikely(err < 0)) {
1309 goto out_fd2;
1312 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
1313 if (unlikely(err < 0)) {
1314 fput(newfile1);
1315 goto out_fd1;
1318 err = audit_fd_pair(fd1, fd2);
1319 if (err < 0) {
1320 fput(newfile1);
1321 fput(newfile2);
1322 goto out_fd;
1325 fd_install(fd1, newfile1);
1326 fd_install(fd2, newfile2);
1327 /* fd1 and fd2 may be already another descriptors.
1328 * Not kernel problem.
1331 err = put_user(fd1, &usockvec[0]);
1332 if (!err)
1333 err = put_user(fd2, &usockvec[1]);
1334 if (!err)
1335 return 0;
1337 sys_close(fd2);
1338 sys_close(fd1);
1339 return err;
1341 out_release_both:
1342 sock_release(sock2);
1343 out_release_1:
1344 sock_release(sock1);
1345 out:
1346 return err;
1348 out_fd2:
1349 put_filp(newfile1);
1350 sock_release(sock1);
1351 out_fd1:
1352 put_filp(newfile2);
1353 sock_release(sock2);
1354 out_fd:
1355 put_unused_fd(fd1);
1356 put_unused_fd(fd2);
1357 goto out;
1361 * Bind a name to a socket. Nothing much to do here since it's
1362 * the protocol's responsibility to handle the local address.
1364 * We move the socket address to kernel space before we call
1365 * the protocol layer (having also checked the address is ok).
1368 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1370 struct socket *sock;
1371 struct sockaddr_storage address;
1372 int err, fput_needed;
1374 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1375 if (sock) {
1376 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1377 if (err >= 0) {
1378 err = security_socket_bind(sock,
1379 (struct sockaddr *)&address,
1380 addrlen);
1381 if (!err)
1382 err = sock->ops->bind(sock,
1383 (struct sockaddr *)
1384 &address, addrlen);
1386 fput_light(sock->file, fput_needed);
1388 return err;
1392 * Perform a listen. Basically, we allow the protocol to do anything
1393 * necessary for a listen, and if that works, we mark the socket as
1394 * ready for listening.
1397 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1399 struct socket *sock;
1400 int err, fput_needed;
1401 int somaxconn;
1403 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1404 if (sock) {
1405 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1406 if ((unsigned)backlog > somaxconn)
1407 backlog = somaxconn;
1409 err = security_socket_listen(sock, backlog);
1410 if (!err)
1411 err = sock->ops->listen(sock, backlog);
1413 fput_light(sock->file, fput_needed);
1415 return err;
1419 * For accept, we attempt to create a new socket, set up the link
1420 * with the client, wake up the client, then return the new
1421 * connected fd. We collect the address of the connector in kernel
1422 * space and move it to user at the very end. This is unclean because
1423 * we open the socket then return an error.
1425 * 1003.1g adds the ability to recvmsg() to query connection pending
1426 * status to recvmsg. We need to add that support in a way thats
1427 * clean when we restucture accept also.
1430 long do_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1431 int __user *upeer_addrlen, int flags)
1433 struct socket *sock, *newsock;
1434 struct file *newfile;
1435 int err, len, newfd, fput_needed;
1436 struct sockaddr_storage address;
1438 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1439 return -EINVAL;
1441 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1442 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1444 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1445 if (!sock)
1446 goto out;
1448 err = -ENFILE;
1449 if (!(newsock = sock_alloc()))
1450 goto out_put;
1452 newsock->type = sock->type;
1453 newsock->ops = sock->ops;
1456 * We don't need try_module_get here, as the listening socket (sock)
1457 * has the protocol module (sock->ops->owner) held.
1459 __module_get(newsock->ops->owner);
1461 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
1462 if (unlikely(newfd < 0)) {
1463 err = newfd;
1464 sock_release(newsock);
1465 goto out_put;
1468 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
1469 if (err < 0)
1470 goto out_fd_simple;
1472 err = security_socket_accept(sock, newsock);
1473 if (err)
1474 goto out_fd;
1476 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1477 if (err < 0)
1478 goto out_fd;
1480 if (upeer_sockaddr) {
1481 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1482 &len, 2) < 0) {
1483 err = -ECONNABORTED;
1484 goto out_fd;
1486 err = move_addr_to_user((struct sockaddr *)&address,
1487 len, upeer_sockaddr, upeer_addrlen);
1488 if (err < 0)
1489 goto out_fd;
1492 /* File flags are not inherited via accept() unlike another OSes. */
1494 fd_install(newfd, newfile);
1495 err = newfd;
1497 security_socket_post_accept(sock, newsock);
1499 out_put:
1500 fput_light(sock->file, fput_needed);
1501 out:
1502 return err;
1503 out_fd_simple:
1504 sock_release(newsock);
1505 put_filp(newfile);
1506 put_unused_fd(newfd);
1507 goto out_put;
1508 out_fd:
1509 fput(newfile);
1510 put_unused_fd(newfd);
1511 goto out_put;
1514 #if 0
1515 #ifdef HAVE_SET_RESTORE_SIGMASK
1516 asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1517 int __user *upeer_addrlen,
1518 const sigset_t __user *sigmask,
1519 size_t sigsetsize, int flags)
1521 sigset_t ksigmask, sigsaved;
1522 int ret;
1524 if (sigmask) {
1525 /* XXX: Don't preclude handling different sized sigset_t's. */
1526 if (sigsetsize != sizeof(sigset_t))
1527 return -EINVAL;
1528 if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
1529 return -EFAULT;
1531 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1532 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1535 ret = do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1537 if (ret < 0 && signal_pending(current)) {
1539 * Don't restore the signal mask yet. Let do_signal() deliver
1540 * the signal on the way back to userspace, before the signal
1541 * mask is restored.
1543 if (sigmask) {
1544 memcpy(&current->saved_sigmask, &sigsaved,
1545 sizeof(sigsaved));
1546 set_restore_sigmask();
1548 } else if (sigmask)
1549 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1551 return ret;
1553 #else
1554 asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1555 int __user *upeer_addrlen,
1556 const sigset_t __user *sigmask,
1557 size_t sigsetsize, int flags)
1559 /* The platform does not support restoring the signal mask in the
1560 * return path. So we do not allow using paccept() with a signal
1561 * mask. */
1562 if (sigmask)
1563 return -EINVAL;
1565 return do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1567 #endif
1568 #endif
1570 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1571 int __user *, upeer_addrlen)
1573 return do_accept(fd, upeer_sockaddr, upeer_addrlen, 0);
1577 * Attempt to connect to a socket with the server address. The address
1578 * is in user space so we verify it is OK and move it to kernel space.
1580 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1581 * break bindings
1583 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1584 * other SEQPACKET protocols that take time to connect() as it doesn't
1585 * include the -EINPROGRESS status for such sockets.
1588 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1589 int, addrlen)
1591 struct socket *sock;
1592 struct sockaddr_storage address;
1593 int err, fput_needed;
1595 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1596 if (!sock)
1597 goto out;
1598 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1599 if (err < 0)
1600 goto out_put;
1602 err =
1603 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1604 if (err)
1605 goto out_put;
1607 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1608 sock->file->f_flags);
1609 out_put:
1610 fput_light(sock->file, fput_needed);
1611 out:
1612 return err;
1616 * Get the local address ('name') of a socket object. Move the obtained
1617 * name to user space.
1620 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1621 int __user *, usockaddr_len)
1623 struct socket *sock;
1624 struct sockaddr_storage address;
1625 int len, err, fput_needed;
1627 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1628 if (!sock)
1629 goto out;
1631 err = security_socket_getsockname(sock);
1632 if (err)
1633 goto out_put;
1635 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1636 if (err)
1637 goto out_put;
1638 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1640 out_put:
1641 fput_light(sock->file, fput_needed);
1642 out:
1643 return err;
1647 * Get the remote address ('name') of a socket object. Move the obtained
1648 * name to user space.
1651 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1652 int __user *, usockaddr_len)
1654 struct socket *sock;
1655 struct sockaddr_storage address;
1656 int len, err, fput_needed;
1658 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1659 if (sock != NULL) {
1660 err = security_socket_getpeername(sock);
1661 if (err) {
1662 fput_light(sock->file, fput_needed);
1663 return err;
1666 err =
1667 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1669 if (!err)
1670 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1671 usockaddr_len);
1672 fput_light(sock->file, fput_needed);
1674 return err;
1678 * Send a datagram to a given address. We move the address into kernel
1679 * space and check the user space data area is readable before invoking
1680 * the protocol.
1683 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1684 unsigned, flags, struct sockaddr __user *, addr,
1685 int, addr_len)
1687 struct socket *sock;
1688 struct sockaddr_storage address;
1689 int err;
1690 struct msghdr msg;
1691 struct iovec iov;
1692 int fput_needed;
1694 if (len > INT_MAX)
1695 len = INT_MAX;
1696 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1697 if (!sock)
1698 goto out;
1700 iov.iov_base = buff;
1701 iov.iov_len = len;
1702 msg.msg_name = NULL;
1703 msg.msg_iov = &iov;
1704 msg.msg_iovlen = 1;
1705 msg.msg_control = NULL;
1706 msg.msg_controllen = 0;
1707 msg.msg_namelen = 0;
1708 if (addr) {
1709 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1710 if (err < 0)
1711 goto out_put;
1712 msg.msg_name = (struct sockaddr *)&address;
1713 msg.msg_namelen = addr_len;
1715 if (sock->file->f_flags & O_NONBLOCK)
1716 flags |= MSG_DONTWAIT;
1717 msg.msg_flags = flags;
1718 err = sock_sendmsg(sock, &msg, len);
1720 out_put:
1721 fput_light(sock->file, fput_needed);
1722 out:
1723 return err;
1727 * Send a datagram down a socket.
1730 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1731 unsigned, flags)
1733 return sys_sendto(fd, buff, len, flags, NULL, 0);
1737 * Receive a frame from the socket and optionally record the address of the
1738 * sender. We verify the buffers are writable and if needed move the
1739 * sender address from kernel to user space.
1742 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1743 unsigned, flags, struct sockaddr __user *, addr,
1744 int __user *, addr_len)
1746 struct socket *sock;
1747 struct iovec iov;
1748 struct msghdr msg;
1749 struct sockaddr_storage address;
1750 int err, err2;
1751 int fput_needed;
1753 if (size > INT_MAX)
1754 size = INT_MAX;
1755 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1756 if (!sock)
1757 goto out;
1759 msg.msg_control = NULL;
1760 msg.msg_controllen = 0;
1761 msg.msg_iovlen = 1;
1762 msg.msg_iov = &iov;
1763 iov.iov_len = size;
1764 iov.iov_base = ubuf;
1765 msg.msg_name = (struct sockaddr *)&address;
1766 msg.msg_namelen = sizeof(address);
1767 if (sock->file->f_flags & O_NONBLOCK)
1768 flags |= MSG_DONTWAIT;
1769 err = sock_recvmsg(sock, &msg, size, flags);
1771 if (err >= 0 && addr != NULL) {
1772 err2 = move_addr_to_user((struct sockaddr *)&address,
1773 msg.msg_namelen, addr, addr_len);
1774 if (err2 < 0)
1775 err = err2;
1778 fput_light(sock->file, fput_needed);
1779 out:
1780 return err;
1784 * Receive a datagram from a socket.
1787 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1788 unsigned flags)
1790 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1794 * Set a socket option. Because we don't know the option lengths we have
1795 * to pass the user mode parameter for the protocols to sort out.
1798 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1799 char __user *, optval, int, optlen)
1801 int err, fput_needed;
1802 struct socket *sock;
1804 if (optlen < 0)
1805 return -EINVAL;
1807 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1808 if (sock != NULL) {
1809 err = security_socket_setsockopt(sock, level, optname);
1810 if (err)
1811 goto out_put;
1813 if (level == SOL_SOCKET)
1814 err =
1815 sock_setsockopt(sock, level, optname, optval,
1816 optlen);
1817 else
1818 err =
1819 sock->ops->setsockopt(sock, level, optname, optval,
1820 optlen);
1821 out_put:
1822 fput_light(sock->file, fput_needed);
1824 return err;
1828 * Get a socket option. Because we don't know the option lengths we have
1829 * to pass a user mode parameter for the protocols to sort out.
1832 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1833 char __user *, optval, int __user *, optlen)
1835 int err, fput_needed;
1836 struct socket *sock;
1838 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1839 if (sock != NULL) {
1840 err = security_socket_getsockopt(sock, level, optname);
1841 if (err)
1842 goto out_put;
1844 if (level == SOL_SOCKET)
1845 err =
1846 sock_getsockopt(sock, level, optname, optval,
1847 optlen);
1848 else
1849 err =
1850 sock->ops->getsockopt(sock, level, optname, optval,
1851 optlen);
1852 out_put:
1853 fput_light(sock->file, fput_needed);
1855 return err;
1859 * Shutdown a socket.
1862 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1864 int err, fput_needed;
1865 struct socket *sock;
1867 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1868 if (sock != NULL) {
1869 err = security_socket_shutdown(sock, how);
1870 if (!err)
1871 err = sock->ops->shutdown(sock, how);
1872 fput_light(sock->file, fput_needed);
1874 return err;
1877 /* A couple of helpful macros for getting the address of the 32/64 bit
1878 * fields which are the same type (int / unsigned) on our platforms.
1880 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1881 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1882 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1885 * BSD sendmsg interface
1888 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1890 struct compat_msghdr __user *msg_compat =
1891 (struct compat_msghdr __user *)msg;
1892 struct socket *sock;
1893 struct sockaddr_storage address;
1894 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1895 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1896 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1897 /* 20 is size of ipv6_pktinfo */
1898 unsigned char *ctl_buf = ctl;
1899 struct msghdr msg_sys;
1900 int err, ctl_len, iov_size, total_len;
1901 int fput_needed;
1903 err = -EFAULT;
1904 if (MSG_CMSG_COMPAT & flags) {
1905 if (get_compat_msghdr(&msg_sys, msg_compat))
1906 return -EFAULT;
1908 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1909 return -EFAULT;
1911 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1912 if (!sock)
1913 goto out;
1915 /* do not move before msg_sys is valid */
1916 err = -EMSGSIZE;
1917 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1918 goto out_put;
1920 /* Check whether to allocate the iovec area */
1921 err = -ENOMEM;
1922 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1923 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1924 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1925 if (!iov)
1926 goto out_put;
1929 /* This will also move the address data into kernel space */
1930 if (MSG_CMSG_COMPAT & flags) {
1931 err = verify_compat_iovec(&msg_sys, iov,
1932 (struct sockaddr *)&address,
1933 VERIFY_READ);
1934 } else
1935 err = verify_iovec(&msg_sys, iov,
1936 (struct sockaddr *)&address,
1937 VERIFY_READ);
1938 if (err < 0)
1939 goto out_freeiov;
1940 total_len = err;
1942 err = -ENOBUFS;
1944 if (msg_sys.msg_controllen > INT_MAX)
1945 goto out_freeiov;
1946 ctl_len = msg_sys.msg_controllen;
1947 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1948 err =
1949 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1950 sizeof(ctl));
1951 if (err)
1952 goto out_freeiov;
1953 ctl_buf = msg_sys.msg_control;
1954 ctl_len = msg_sys.msg_controllen;
1955 } else if (ctl_len) {
1956 if (ctl_len > sizeof(ctl)) {
1957 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1958 if (ctl_buf == NULL)
1959 goto out_freeiov;
1961 err = -EFAULT;
1963 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1964 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1965 * checking falls down on this.
1967 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1968 ctl_len))
1969 goto out_freectl;
1970 msg_sys.msg_control = ctl_buf;
1972 msg_sys.msg_flags = flags;
1974 if (sock->file->f_flags & O_NONBLOCK)
1975 msg_sys.msg_flags |= MSG_DONTWAIT;
1976 err = sock_sendmsg(sock, &msg_sys, total_len);
1978 out_freectl:
1979 if (ctl_buf != ctl)
1980 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1981 out_freeiov:
1982 if (iov != iovstack)
1983 sock_kfree_s(sock->sk, iov, iov_size);
1984 out_put:
1985 fput_light(sock->file, fput_needed);
1986 out:
1987 return err;
1991 * BSD recvmsg interface
1994 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
1995 unsigned int, flags)
1997 struct compat_msghdr __user *msg_compat =
1998 (struct compat_msghdr __user *)msg;
1999 struct socket *sock;
2000 struct iovec iovstack[UIO_FASTIOV];
2001 struct iovec *iov = iovstack;
2002 struct msghdr msg_sys;
2003 unsigned long cmsg_ptr;
2004 int err, iov_size, total_len, len;
2005 int fput_needed;
2007 /* kernel mode address */
2008 struct sockaddr_storage addr;
2010 /* user mode address pointers */
2011 struct sockaddr __user *uaddr;
2012 int __user *uaddr_len;
2014 if (MSG_CMSG_COMPAT & flags) {
2015 if (get_compat_msghdr(&msg_sys, msg_compat))
2016 return -EFAULT;
2018 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
2019 return -EFAULT;
2021 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2022 if (!sock)
2023 goto out;
2025 err = -EMSGSIZE;
2026 if (msg_sys.msg_iovlen > UIO_MAXIOV)
2027 goto out_put;
2029 /* Check whether to allocate the iovec area */
2030 err = -ENOMEM;
2031 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2032 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2033 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2034 if (!iov)
2035 goto out_put;
2039 * Save the user-mode address (verify_iovec will change the
2040 * kernel msghdr to use the kernel address space)
2043 uaddr = (__force void __user *)msg_sys.msg_name;
2044 uaddr_len = COMPAT_NAMELEN(msg);
2045 if (MSG_CMSG_COMPAT & flags) {
2046 err = verify_compat_iovec(&msg_sys, iov,
2047 (struct sockaddr *)&addr,
2048 VERIFY_WRITE);
2049 } else
2050 err = verify_iovec(&msg_sys, iov,
2051 (struct sockaddr *)&addr,
2052 VERIFY_WRITE);
2053 if (err < 0)
2054 goto out_freeiov;
2055 total_len = err;
2057 cmsg_ptr = (unsigned long)msg_sys.msg_control;
2058 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2060 if (sock->file->f_flags & O_NONBLOCK)
2061 flags |= MSG_DONTWAIT;
2062 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2063 if (err < 0)
2064 goto out_freeiov;
2065 len = err;
2067 if (uaddr != NULL) {
2068 err = move_addr_to_user((struct sockaddr *)&addr,
2069 msg_sys.msg_namelen, uaddr,
2070 uaddr_len);
2071 if (err < 0)
2072 goto out_freeiov;
2074 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2075 COMPAT_FLAGS(msg));
2076 if (err)
2077 goto out_freeiov;
2078 if (MSG_CMSG_COMPAT & flags)
2079 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2080 &msg_compat->msg_controllen);
2081 else
2082 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2083 &msg->msg_controllen);
2084 if (err)
2085 goto out_freeiov;
2086 err = len;
2088 out_freeiov:
2089 if (iov != iovstack)
2090 sock_kfree_s(sock->sk, iov, iov_size);
2091 out_put:
2092 fput_light(sock->file, fput_needed);
2093 out:
2094 return err;
2097 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2099 /* Argument list sizes for sys_socketcall */
2100 #define AL(x) ((x) * sizeof(unsigned long))
2101 static const unsigned char nargs[19]={
2102 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2103 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2104 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2105 AL(6)
2108 #undef AL
2111 * System call vectors.
2113 * Argument checking cleaned up. Saved 20% in size.
2114 * This function doesn't need to set the kernel lock because
2115 * it is set by the callees.
2118 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2120 unsigned long a[6];
2121 unsigned long a0, a1;
2122 int err;
2124 if (call < 1 || call > SYS_PACCEPT)
2125 return -EINVAL;
2127 /* copy_from_user should be SMP safe. */
2128 if (copy_from_user(a, args, nargs[call]))
2129 return -EFAULT;
2131 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2132 if (err)
2133 return err;
2135 a0 = a[0];
2136 a1 = a[1];
2138 switch (call) {
2139 case SYS_SOCKET:
2140 err = sys_socket(a0, a1, a[2]);
2141 break;
2142 case SYS_BIND:
2143 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2144 break;
2145 case SYS_CONNECT:
2146 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2147 break;
2148 case SYS_LISTEN:
2149 err = sys_listen(a0, a1);
2150 break;
2151 case SYS_ACCEPT:
2152 err =
2153 do_accept(a0, (struct sockaddr __user *)a1,
2154 (int __user *)a[2], 0);
2155 break;
2156 case SYS_GETSOCKNAME:
2157 err =
2158 sys_getsockname(a0, (struct sockaddr __user *)a1,
2159 (int __user *)a[2]);
2160 break;
2161 case SYS_GETPEERNAME:
2162 err =
2163 sys_getpeername(a0, (struct sockaddr __user *)a1,
2164 (int __user *)a[2]);
2165 break;
2166 case SYS_SOCKETPAIR:
2167 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2168 break;
2169 case SYS_SEND:
2170 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2171 break;
2172 case SYS_SENDTO:
2173 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2174 (struct sockaddr __user *)a[4], a[5]);
2175 break;
2176 case SYS_RECV:
2177 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2178 break;
2179 case SYS_RECVFROM:
2180 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2181 (struct sockaddr __user *)a[4],
2182 (int __user *)a[5]);
2183 break;
2184 case SYS_SHUTDOWN:
2185 err = sys_shutdown(a0, a1);
2186 break;
2187 case SYS_SETSOCKOPT:
2188 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2189 break;
2190 case SYS_GETSOCKOPT:
2191 err =
2192 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2193 (int __user *)a[4]);
2194 break;
2195 case SYS_SENDMSG:
2196 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2197 break;
2198 case SYS_RECVMSG:
2199 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2200 break;
2201 case SYS_PACCEPT:
2202 err =
2203 sys_paccept(a0, (struct sockaddr __user *)a1,
2204 (int __user *)a[2],
2205 (const sigset_t __user *) a[3],
2206 a[4], a[5]);
2207 break;
2208 default:
2209 err = -EINVAL;
2210 break;
2212 return err;
2215 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2218 * sock_register - add a socket protocol handler
2219 * @ops: description of protocol
2221 * This function is called by a protocol handler that wants to
2222 * advertise its address family, and have it linked into the
2223 * socket interface. The value ops->family coresponds to the
2224 * socket system call protocol family.
2226 int sock_register(const struct net_proto_family *ops)
2228 int err;
2230 if (ops->family >= NPROTO) {
2231 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2232 NPROTO);
2233 return -ENOBUFS;
2236 spin_lock(&net_family_lock);
2237 if (net_families[ops->family])
2238 err = -EEXIST;
2239 else {
2240 net_families[ops->family] = ops;
2241 err = 0;
2243 spin_unlock(&net_family_lock);
2245 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2246 return err;
2250 * sock_unregister - remove a protocol handler
2251 * @family: protocol family to remove
2253 * This function is called by a protocol handler that wants to
2254 * remove its address family, and have it unlinked from the
2255 * new socket creation.
2257 * If protocol handler is a module, then it can use module reference
2258 * counts to protect against new references. If protocol handler is not
2259 * a module then it needs to provide its own protection in
2260 * the ops->create routine.
2262 void sock_unregister(int family)
2264 BUG_ON(family < 0 || family >= NPROTO);
2266 spin_lock(&net_family_lock);
2267 net_families[family] = NULL;
2268 spin_unlock(&net_family_lock);
2270 synchronize_rcu();
2272 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2275 static int __init sock_init(void)
2278 * Initialize sock SLAB cache.
2281 sk_init();
2284 * Initialize skbuff SLAB cache
2286 skb_init();
2289 * Initialize the protocols module.
2292 init_inodecache();
2293 register_filesystem(&sock_fs_type);
2294 sock_mnt = kern_mount(&sock_fs_type);
2296 /* The real protocol initialization is performed in later initcalls.
2299 #ifdef CONFIG_NETFILTER
2300 netfilter_init();
2301 #endif
2303 return 0;
2306 core_initcall(sock_init); /* early initcall */
2308 #ifdef CONFIG_PROC_FS
2309 void socket_seq_show(struct seq_file *seq)
2311 int cpu;
2312 int counter = 0;
2314 for_each_possible_cpu(cpu)
2315 counter += per_cpu(sockets_in_use, cpu);
2317 /* It can be negative, by the way. 8) */
2318 if (counter < 0)
2319 counter = 0;
2321 seq_printf(seq, "sockets: used %d\n", counter);
2323 #endif /* CONFIG_PROC_FS */
2325 #ifdef CONFIG_COMPAT
2326 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2327 unsigned long arg)
2329 struct socket *sock = file->private_data;
2330 int ret = -ENOIOCTLCMD;
2331 struct sock *sk;
2332 struct net *net;
2334 sk = sock->sk;
2335 net = sock_net(sk);
2337 if (sock->ops->compat_ioctl)
2338 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2340 if (ret == -ENOIOCTLCMD &&
2341 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2342 ret = compat_wext_handle_ioctl(net, cmd, arg);
2344 return ret;
2346 #endif
2348 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2350 return sock->ops->bind(sock, addr, addrlen);
2353 int kernel_listen(struct socket *sock, int backlog)
2355 return sock->ops->listen(sock, backlog);
2358 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2360 struct sock *sk = sock->sk;
2361 int err;
2363 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2364 newsock);
2365 if (err < 0)
2366 goto done;
2368 err = sock->ops->accept(sock, *newsock, flags);
2369 if (err < 0) {
2370 sock_release(*newsock);
2371 *newsock = NULL;
2372 goto done;
2375 (*newsock)->ops = sock->ops;
2377 done:
2378 return err;
2381 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2382 int flags)
2384 return sock->ops->connect(sock, addr, addrlen, flags);
2387 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2388 int *addrlen)
2390 return sock->ops->getname(sock, addr, addrlen, 0);
2393 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2394 int *addrlen)
2396 return sock->ops->getname(sock, addr, addrlen, 1);
2399 int kernel_getsockopt(struct socket *sock, int level, int optname,
2400 char *optval, int *optlen)
2402 mm_segment_t oldfs = get_fs();
2403 int err;
2405 set_fs(KERNEL_DS);
2406 if (level == SOL_SOCKET)
2407 err = sock_getsockopt(sock, level, optname, optval, optlen);
2408 else
2409 err = sock->ops->getsockopt(sock, level, optname, optval,
2410 optlen);
2411 set_fs(oldfs);
2412 return err;
2415 int kernel_setsockopt(struct socket *sock, int level, int optname,
2416 char *optval, int optlen)
2418 mm_segment_t oldfs = get_fs();
2419 int err;
2421 set_fs(KERNEL_DS);
2422 if (level == SOL_SOCKET)
2423 err = sock_setsockopt(sock, level, optname, optval, optlen);
2424 else
2425 err = sock->ops->setsockopt(sock, level, optname, optval,
2426 optlen);
2427 set_fs(oldfs);
2428 return err;
2431 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2432 size_t size, int flags)
2434 if (sock->ops->sendpage)
2435 return sock->ops->sendpage(sock, page, offset, size, flags);
2437 return sock_no_sendpage(sock, page, offset, size, flags);
2440 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2442 mm_segment_t oldfs = get_fs();
2443 int err;
2445 set_fs(KERNEL_DS);
2446 err = sock->ops->ioctl(sock, cmd, arg);
2447 set_fs(oldfs);
2449 return err;
2452 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2454 return sock->ops->shutdown(sock, how);
2457 EXPORT_SYMBOL(sock_create);
2458 EXPORT_SYMBOL(sock_create_kern);
2459 EXPORT_SYMBOL(sock_create_lite);
2460 EXPORT_SYMBOL(sock_map_fd);
2461 EXPORT_SYMBOL(sock_recvmsg);
2462 EXPORT_SYMBOL(sock_register);
2463 EXPORT_SYMBOL(sock_release);
2464 EXPORT_SYMBOL(sock_sendmsg);
2465 EXPORT_SYMBOL(sock_unregister);
2466 EXPORT_SYMBOL(sock_wake_async);
2467 EXPORT_SYMBOL(sockfd_lookup);
2468 EXPORT_SYMBOL(kernel_sendmsg);
2469 EXPORT_SYMBOL(kernel_recvmsg);
2470 EXPORT_SYMBOL(kernel_bind);
2471 EXPORT_SYMBOL(kernel_listen);
2472 EXPORT_SYMBOL(kernel_accept);
2473 EXPORT_SYMBOL(kernel_connect);
2474 EXPORT_SYMBOL(kernel_getsockname);
2475 EXPORT_SYMBOL(kernel_getpeername);
2476 EXPORT_SYMBOL(kernel_getsockopt);
2477 EXPORT_SYMBOL(kernel_setsockopt);
2478 EXPORT_SYMBOL(kernel_sendpage);
2479 EXPORT_SYMBOL(kernel_sock_ioctl);
2480 EXPORT_SYMBOL(kernel_sock_shutdown);