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>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
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
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
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
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
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
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
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/rcupdate.h>
67 #include <linux/netdevice.h>
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70 #include <linux/mutex.h>
71 #include <linux/wanrouter.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/mount.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/compat.h>
84 #include <linux/kmod.h>
85 #include <linux/audit.h>
86 #include <linux/wireless.h>
87 #include <linux/nsproxy.h>
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
95 #include <linux/netfilter.h>
97 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
98 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
99 unsigned long nr_segs, loff_t pos);
100 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
101 unsigned long nr_segs, loff_t pos);
102 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
104 static int sock_close(struct inode *inode, struct file *file);
105 static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
107 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
109 static long compat_sock_ioctl(struct file *file,
110 unsigned int cmd, unsigned long arg);
112 static int sock_fasync(int fd, struct file *filp, int on);
113 static ssize_t sock_sendpage(struct file *file, struct page *page,
114 int offset, size_t size, loff_t *ppos, int more);
117 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
118 * in the operation structures but are done directly via the socketcall() multiplexor.
121 static const struct file_operations socket_file_ops = {
122 .owner = THIS_MODULE,
124 .aio_read = sock_aio_read,
125 .aio_write = sock_aio_write,
127 .unlocked_ioctl = sock_ioctl,
129 .compat_ioctl = compat_sock_ioctl,
132 .open = sock_no_open, /* special open code to disallow open via /proc */
133 .release = sock_close,
134 .fasync = sock_fasync,
135 .sendpage = sock_sendpage,
136 .splice_write = generic_splice_sendpage,
140 * The protocol list. Each protocol is registered in here.
143 static DEFINE_SPINLOCK(net_family_lock);
144 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
147 * Statistics counters of the socket lists
150 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
154 * Move socket addresses back and forth across the kernel/user
155 * divide and look after the messy bits.
158 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
159 16 for IP, 16 for IPX,
162 must be at least one bigger than
163 the AF_UNIX size (see net/unix/af_unix.c
168 * move_addr_to_kernel - copy a socket address into kernel space
169 * @uaddr: Address in user space
170 * @kaddr: Address in kernel space
171 * @ulen: Length in user space
173 * The address is copied into kernel space. If the provided address is
174 * too long an error code of -EINVAL is returned. If the copy gives
175 * invalid addresses -EFAULT is returned. On a success 0 is returned.
178 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
180 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
184 if (copy_from_user(kaddr, uaddr, ulen))
186 return audit_sockaddr(ulen, kaddr);
190 * move_addr_to_user - copy an address to user space
191 * @kaddr: kernel space address
192 * @klen: length of address in kernel
193 * @uaddr: user space address
194 * @ulen: pointer to user length field
196 * The value pointed to by ulen on entry is the buffer length available.
197 * This is overwritten with the buffer space used. -EINVAL is returned
198 * if an overlong buffer is specified or a negative buffer size. -EFAULT
199 * is returned if either the buffer or the length field are not
201 * After copying the data up to the limit the user specifies, the true
202 * length of the data is written over the length limit the user
203 * specified. Zero is returned for a success.
206 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
212 err = get_user(len, ulen);
217 if (len < 0 || len > MAX_SOCK_ADDR)
220 if (audit_sockaddr(klen, kaddr))
222 if (copy_to_user(uaddr, kaddr, len))
226 * "fromlen shall refer to the value before truncation.."
229 return __put_user(klen, ulen);
232 #define SOCKFS_MAGIC 0x534F434B
234 static struct kmem_cache *sock_inode_cachep __read_mostly;
236 static struct inode *sock_alloc_inode(struct super_block *sb)
238 struct socket_alloc *ei;
240 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
243 init_waitqueue_head(&ei->socket.wait);
245 ei->socket.fasync_list = NULL;
246 ei->socket.state = SS_UNCONNECTED;
247 ei->socket.flags = 0;
248 ei->socket.ops = NULL;
249 ei->socket.sk = NULL;
250 ei->socket.file = NULL;
252 return &ei->vfs_inode;
255 static void sock_destroy_inode(struct inode *inode)
257 kmem_cache_free(sock_inode_cachep,
258 container_of(inode, struct socket_alloc, vfs_inode));
261 static void init_once(struct kmem_cache *cachep, void *foo)
263 struct socket_alloc *ei = (struct socket_alloc *)foo;
265 inode_init_once(&ei->vfs_inode);
268 static int init_inodecache(void)
270 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
271 sizeof(struct socket_alloc),
273 (SLAB_HWCACHE_ALIGN |
274 SLAB_RECLAIM_ACCOUNT |
277 if (sock_inode_cachep == NULL)
282 static struct super_operations sockfs_ops = {
283 .alloc_inode = sock_alloc_inode,
284 .destroy_inode =sock_destroy_inode,
285 .statfs = simple_statfs,
288 static int sockfs_get_sb(struct file_system_type *fs_type,
289 int flags, const char *dev_name, void *data,
290 struct vfsmount *mnt)
292 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
296 static struct vfsmount *sock_mnt __read_mostly;
298 static struct file_system_type sock_fs_type = {
300 .get_sb = sockfs_get_sb,
301 .kill_sb = kill_anon_super,
304 static int sockfs_delete_dentry(struct dentry *dentry)
307 * At creation time, we pretended this dentry was hashed
308 * (by clearing DCACHE_UNHASHED bit in d_flags)
309 * At delete time, we restore the truth : not hashed.
310 * (so that dput() can proceed correctly)
312 dentry->d_flags |= DCACHE_UNHASHED;
317 * sockfs_dname() is called from d_path().
319 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
321 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
322 dentry->d_inode->i_ino);
325 static struct dentry_operations sockfs_dentry_operations = {
326 .d_delete = sockfs_delete_dentry,
327 .d_dname = sockfs_dname,
331 * Obtains the first available file descriptor and sets it up for use.
333 * These functions create file structures and maps them to fd space
334 * of the current process. On success it returns file descriptor
335 * and file struct implicitly stored in sock->file.
336 * Note that another thread may close file descriptor before we return
337 * from this function. We use the fact that now we do not refer
338 * to socket after mapping. If one day we will need it, this
339 * function will increment ref. count on file by 1.
341 * In any case returned fd MAY BE not valid!
342 * This race condition is unavoidable
343 * with shared fd spaces, we cannot solve it inside kernel,
344 * but we take care of internal coherence yet.
347 static int sock_alloc_fd(struct file **filep)
351 fd = get_unused_fd();
352 if (likely(fd >= 0)) {
353 struct file *file = get_empty_filp();
356 if (unlikely(!file)) {
365 static int sock_attach_fd(struct socket *sock, struct file *file)
367 struct dentry *dentry;
368 struct qstr name = { .name = "" };
370 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
371 if (unlikely(!dentry))
374 dentry->d_op = &sockfs_dentry_operations;
376 * We dont want to push this dentry into global dentry hash table.
377 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
378 * This permits a working /proc/$pid/fd/XXX on sockets
380 dentry->d_flags &= ~DCACHE_UNHASHED;
381 d_instantiate(dentry, SOCK_INODE(sock));
384 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
386 SOCK_INODE(sock)->i_fop = &socket_file_ops;
387 file->f_flags = O_RDWR;
389 file->private_data = sock;
394 int sock_map_fd(struct socket *sock)
396 struct file *newfile;
397 int fd = sock_alloc_fd(&newfile);
399 if (likely(fd >= 0)) {
400 int err = sock_attach_fd(sock, newfile);
402 if (unlikely(err < 0)) {
407 fd_install(fd, newfile);
412 static struct socket *sock_from_file(struct file *file, int *err)
414 if (file->f_op == &socket_file_ops)
415 return file->private_data; /* set in sock_map_fd */
422 * sockfd_lookup - Go from a file number to its socket slot
424 * @err: pointer to an error code return
426 * The file handle passed in is locked and the socket it is bound
427 * too is returned. If an error occurs the err pointer is overwritten
428 * with a negative errno code and NULL is returned. The function checks
429 * for both invalid handles and passing a handle which is not a socket.
431 * On a success the socket object pointer is returned.
434 struct socket *sockfd_lookup(int fd, int *err)
445 sock = sock_from_file(file, err);
451 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
457 file = fget_light(fd, fput_needed);
459 sock = sock_from_file(file, err);
462 fput_light(file, *fput_needed);
468 * sock_alloc - allocate a socket
470 * Allocate a new inode and socket object. The two are bound together
471 * and initialised. The socket is then returned. If we are out of inodes
475 static struct socket *sock_alloc(void)
480 inode = new_inode(sock_mnt->mnt_sb);
484 sock = SOCKET_I(inode);
486 inode->i_mode = S_IFSOCK | S_IRWXUGO;
487 inode->i_uid = current->fsuid;
488 inode->i_gid = current->fsgid;
490 get_cpu_var(sockets_in_use)++;
491 put_cpu_var(sockets_in_use);
496 * In theory you can't get an open on this inode, but /proc provides
497 * a back door. Remember to keep it shut otherwise you'll let the
498 * creepy crawlies in.
501 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
506 const struct file_operations bad_sock_fops = {
507 .owner = THIS_MODULE,
508 .open = sock_no_open,
512 * sock_release - close a socket
513 * @sock: socket to close
515 * The socket is released from the protocol stack if it has a release
516 * callback, and the inode is then released if the socket is bound to
517 * an inode not a file.
520 void sock_release(struct socket *sock)
523 struct module *owner = sock->ops->owner;
525 sock->ops->release(sock);
530 if (sock->fasync_list)
531 printk(KERN_ERR "sock_release: fasync list not empty!\n");
533 get_cpu_var(sockets_in_use)--;
534 put_cpu_var(sockets_in_use);
536 iput(SOCK_INODE(sock));
542 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
543 struct msghdr *msg, size_t size)
545 struct sock_iocb *si = kiocb_to_siocb(iocb);
553 err = security_socket_sendmsg(sock, msg, size);
557 return sock->ops->sendmsg(iocb, sock, msg, size);
560 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
563 struct sock_iocb siocb;
566 init_sync_kiocb(&iocb, NULL);
567 iocb.private = &siocb;
568 ret = __sock_sendmsg(&iocb, sock, msg, size);
569 if (-EIOCBQUEUED == ret)
570 ret = wait_on_sync_kiocb(&iocb);
574 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
575 struct kvec *vec, size_t num, size_t size)
577 mm_segment_t oldfs = get_fs();
582 * the following is safe, since for compiler definitions of kvec and
583 * iovec are identical, yielding the same in-core layout and alignment
585 msg->msg_iov = (struct iovec *)vec;
586 msg->msg_iovlen = num;
587 result = sock_sendmsg(sock, msg, size);
593 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
595 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
598 ktime_t kt = skb->tstamp;
600 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
602 /* Race occurred between timestamp enabling and packet
603 receiving. Fill in the current time for now. */
605 kt = ktime_get_real();
607 tv = ktime_to_timeval(kt);
608 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
611 /* Race occurred between timestamp enabling and packet
612 receiving. Fill in the current time for now. */
614 kt = ktime_get_real();
616 ts = ktime_to_timespec(kt);
617 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
621 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
623 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
624 struct msghdr *msg, size_t size, int flags)
627 struct sock_iocb *si = kiocb_to_siocb(iocb);
635 err = security_socket_recvmsg(sock, msg, size, flags);
639 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
642 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
643 size_t size, int flags)
646 struct sock_iocb siocb;
649 init_sync_kiocb(&iocb, NULL);
650 iocb.private = &siocb;
651 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
652 if (-EIOCBQUEUED == ret)
653 ret = wait_on_sync_kiocb(&iocb);
657 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
658 struct kvec *vec, size_t num, size_t size, int flags)
660 mm_segment_t oldfs = get_fs();
665 * the following is safe, since for compiler definitions of kvec and
666 * iovec are identical, yielding the same in-core layout and alignment
668 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
669 result = sock_recvmsg(sock, msg, size, flags);
674 static void sock_aio_dtor(struct kiocb *iocb)
676 kfree(iocb->private);
679 static ssize_t sock_sendpage(struct file *file, struct page *page,
680 int offset, size_t size, loff_t *ppos, int more)
685 sock = file->private_data;
687 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
691 return sock->ops->sendpage(sock, page, offset, size, flags);
694 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
695 struct sock_iocb *siocb)
697 if (!is_sync_kiocb(iocb)) {
698 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
701 iocb->ki_dtor = sock_aio_dtor;
705 iocb->private = siocb;
709 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
710 struct file *file, const struct iovec *iov,
711 unsigned long nr_segs)
713 struct socket *sock = file->private_data;
717 for (i = 0; i < nr_segs; i++)
718 size += iov[i].iov_len;
720 msg->msg_name = NULL;
721 msg->msg_namelen = 0;
722 msg->msg_control = NULL;
723 msg->msg_controllen = 0;
724 msg->msg_iov = (struct iovec *)iov;
725 msg->msg_iovlen = nr_segs;
726 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
728 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
731 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
732 unsigned long nr_segs, loff_t pos)
734 struct sock_iocb siocb, *x;
739 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
743 x = alloc_sock_iocb(iocb, &siocb);
746 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
749 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
750 struct file *file, const struct iovec *iov,
751 unsigned long nr_segs)
753 struct socket *sock = file->private_data;
757 for (i = 0; i < nr_segs; i++)
758 size += iov[i].iov_len;
760 msg->msg_name = NULL;
761 msg->msg_namelen = 0;
762 msg->msg_control = NULL;
763 msg->msg_controllen = 0;
764 msg->msg_iov = (struct iovec *)iov;
765 msg->msg_iovlen = nr_segs;
766 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
767 if (sock->type == SOCK_SEQPACKET)
768 msg->msg_flags |= MSG_EOR;
770 return __sock_sendmsg(iocb, sock, msg, size);
773 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
774 unsigned long nr_segs, loff_t pos)
776 struct sock_iocb siocb, *x;
781 x = alloc_sock_iocb(iocb, &siocb);
785 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
789 * Atomic setting of ioctl hooks to avoid race
790 * with module unload.
793 static DEFINE_MUTEX(br_ioctl_mutex);
794 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
796 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
798 mutex_lock(&br_ioctl_mutex);
799 br_ioctl_hook = hook;
800 mutex_unlock(&br_ioctl_mutex);
803 EXPORT_SYMBOL(brioctl_set);
805 static DEFINE_MUTEX(vlan_ioctl_mutex);
806 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
808 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
810 mutex_lock(&vlan_ioctl_mutex);
811 vlan_ioctl_hook = hook;
812 mutex_unlock(&vlan_ioctl_mutex);
815 EXPORT_SYMBOL(vlan_ioctl_set);
817 static DEFINE_MUTEX(dlci_ioctl_mutex);
818 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
820 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
822 mutex_lock(&dlci_ioctl_mutex);
823 dlci_ioctl_hook = hook;
824 mutex_unlock(&dlci_ioctl_mutex);
827 EXPORT_SYMBOL(dlci_ioctl_set);
830 * With an ioctl, arg may well be a user mode pointer, but we don't know
831 * what to do with it - that's up to the protocol still.
834 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
838 void __user *argp = (void __user *)arg;
842 sock = file->private_data;
845 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
846 err = dev_ioctl(net, cmd, argp);
848 #ifdef CONFIG_WIRELESS_EXT
849 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
850 err = dev_ioctl(net, cmd, argp);
852 #endif /* CONFIG_WIRELESS_EXT */
857 if (get_user(pid, (int __user *)argp))
859 err = f_setown(sock->file, pid, 1);
863 err = put_user(f_getown(sock->file),
872 request_module("bridge");
874 mutex_lock(&br_ioctl_mutex);
876 err = br_ioctl_hook(net, cmd, argp);
877 mutex_unlock(&br_ioctl_mutex);
882 if (!vlan_ioctl_hook)
883 request_module("8021q");
885 mutex_lock(&vlan_ioctl_mutex);
887 err = vlan_ioctl_hook(net, argp);
888 mutex_unlock(&vlan_ioctl_mutex);
893 if (!dlci_ioctl_hook)
894 request_module("dlci");
896 if (dlci_ioctl_hook) {
897 mutex_lock(&dlci_ioctl_mutex);
898 err = dlci_ioctl_hook(cmd, argp);
899 mutex_unlock(&dlci_ioctl_mutex);
903 err = sock->ops->ioctl(sock, cmd, arg);
906 * If this ioctl is unknown try to hand it down
909 if (err == -ENOIOCTLCMD)
910 err = dev_ioctl(net, cmd, argp);
916 int sock_create_lite(int family, int type, int protocol, struct socket **res)
919 struct socket *sock = NULL;
921 err = security_socket_create(family, type, protocol, 1);
932 err = security_socket_post_create(sock, family, type, protocol, 1);
945 /* No kernel lock held - perfect */
946 static unsigned int sock_poll(struct file *file, poll_table *wait)
951 * We can't return errors to poll, so it's either yes or no.
953 sock = file->private_data;
954 return sock->ops->poll(file, sock, wait);
957 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
959 struct socket *sock = file->private_data;
961 return sock->ops->mmap(file, sock, vma);
964 static int sock_close(struct inode *inode, struct file *filp)
967 * It was possible the inode is NULL we were
968 * closing an unfinished socket.
972 printk(KERN_DEBUG "sock_close: NULL inode\n");
975 sock_fasync(-1, filp, 0);
976 sock_release(SOCKET_I(inode));
981 * Update the socket async list
983 * Fasync_list locking strategy.
985 * 1. fasync_list is modified only under process context socket lock
986 * i.e. under semaphore.
987 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
988 * or under socket lock.
989 * 3. fasync_list can be used from softirq context, so that
990 * modification under socket lock have to be enhanced with
991 * write_lock_bh(&sk->sk_callback_lock).
995 static int sock_fasync(int fd, struct file *filp, int on)
997 struct fasync_struct *fa, *fna = NULL, **prev;
1002 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1007 sock = filp->private_data;
1017 prev = &(sock->fasync_list);
1019 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1020 if (fa->fa_file == filp)
1025 write_lock_bh(&sk->sk_callback_lock);
1027 write_unlock_bh(&sk->sk_callback_lock);
1032 fna->fa_file = filp;
1034 fna->magic = FASYNC_MAGIC;
1035 fna->fa_next = sock->fasync_list;
1036 write_lock_bh(&sk->sk_callback_lock);
1037 sock->fasync_list = fna;
1038 write_unlock_bh(&sk->sk_callback_lock);
1041 write_lock_bh(&sk->sk_callback_lock);
1042 *prev = fa->fa_next;
1043 write_unlock_bh(&sk->sk_callback_lock);
1049 release_sock(sock->sk);
1053 /* This function may be called only under socket lock or callback_lock */
1055 int sock_wake_async(struct socket *sock, int how, int band)
1057 if (!sock || !sock->fasync_list)
1062 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1066 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1071 __kill_fasync(sock->fasync_list, SIGIO, band);
1074 __kill_fasync(sock->fasync_list, SIGURG, band);
1079 static int __sock_create(struct net *net, int family, int type, int protocol,
1080 struct socket **res, int kern)
1083 struct socket *sock;
1084 const struct net_proto_family *pf;
1087 * Check protocol is in range
1089 if (family < 0 || family >= NPROTO)
1090 return -EAFNOSUPPORT;
1091 if (type < 0 || type >= SOCK_MAX)
1096 This uglymoron is moved from INET layer to here to avoid
1097 deadlock in module load.
1099 if (family == PF_INET && type == SOCK_PACKET) {
1103 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1109 err = security_socket_create(family, type, protocol, kern);
1114 * Allocate the socket and allow the family to set things up. if
1115 * the protocol is 0, the family is instructed to select an appropriate
1118 sock = sock_alloc();
1120 if (net_ratelimit())
1121 printk(KERN_WARNING "socket: no more sockets\n");
1122 return -ENFILE; /* Not exactly a match, but its the
1123 closest posix thing */
1128 #if defined(CONFIG_KMOD)
1129 /* Attempt to load a protocol module if the find failed.
1131 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1132 * requested real, full-featured networking support upon configuration.
1133 * Otherwise module support will break!
1135 if (net_families[family] == NULL)
1136 request_module("net-pf-%d", family);
1140 pf = rcu_dereference(net_families[family]);
1141 err = -EAFNOSUPPORT;
1146 * We will call the ->create function, that possibly is in a loadable
1147 * module, so we have to bump that loadable module refcnt first.
1149 if (!try_module_get(pf->owner))
1152 /* Now protected by module ref count */
1155 err = pf->create(net, sock, protocol);
1157 goto out_module_put;
1160 * Now to bump the refcnt of the [loadable] module that owns this
1161 * socket at sock_release time we decrement its refcnt.
1163 if (!try_module_get(sock->ops->owner))
1164 goto out_module_busy;
1167 * Now that we're done with the ->create function, the [loadable]
1168 * module can have its refcnt decremented
1170 module_put(pf->owner);
1171 err = security_socket_post_create(sock, family, type, protocol, kern);
1173 goto out_sock_release;
1179 err = -EAFNOSUPPORT;
1182 module_put(pf->owner);
1189 goto out_sock_release;
1192 int sock_create(int family, int type, int protocol, struct socket **res)
1194 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1197 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1199 return __sock_create(&init_net, family, type, protocol, res, 1);
1202 asmlinkage long sys_socket(int family, int type, int protocol)
1205 struct socket *sock;
1207 retval = sock_create(family, type, protocol, &sock);
1211 retval = sock_map_fd(sock);
1216 /* It may be already another descriptor 8) Not kernel problem. */
1225 * Create a pair of connected sockets.
1228 asmlinkage long sys_socketpair(int family, int type, int protocol,
1229 int __user *usockvec)
1231 struct socket *sock1, *sock2;
1233 struct file *newfile1, *newfile2;
1236 * Obtain the first socket and check if the underlying protocol
1237 * supports the socketpair call.
1240 err = sock_create(family, type, protocol, &sock1);
1244 err = sock_create(family, type, protocol, &sock2);
1248 err = sock1->ops->socketpair(sock1, sock2);
1250 goto out_release_both;
1252 fd1 = sock_alloc_fd(&newfile1);
1253 if (unlikely(fd1 < 0)) {
1255 goto out_release_both;
1258 fd2 = sock_alloc_fd(&newfile2);
1259 if (unlikely(fd2 < 0)) {
1263 goto out_release_both;
1266 err = sock_attach_fd(sock1, newfile1);
1267 if (unlikely(err < 0)) {
1271 err = sock_attach_fd(sock2, newfile2);
1272 if (unlikely(err < 0)) {
1277 err = audit_fd_pair(fd1, fd2);
1284 fd_install(fd1, newfile1);
1285 fd_install(fd2, newfile2);
1286 /* fd1 and fd2 may be already another descriptors.
1287 * Not kernel problem.
1290 err = put_user(fd1, &usockvec[0]);
1292 err = put_user(fd2, &usockvec[1]);
1301 sock_release(sock2);
1303 sock_release(sock1);
1309 sock_release(sock1);
1312 sock_release(sock2);
1320 * Bind a name to a socket. Nothing much to do here since it's
1321 * the protocol's responsibility to handle the local address.
1323 * We move the socket address to kernel space before we call
1324 * the protocol layer (having also checked the address is ok).
1327 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1329 struct socket *sock;
1330 char address[MAX_SOCK_ADDR];
1331 int err, fput_needed;
1333 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1335 err = move_addr_to_kernel(umyaddr, addrlen, address);
1337 err = security_socket_bind(sock,
1338 (struct sockaddr *)address,
1341 err = sock->ops->bind(sock,
1345 fput_light(sock->file, fput_needed);
1351 * Perform a listen. Basically, we allow the protocol to do anything
1352 * necessary for a listen, and if that works, we mark the socket as
1353 * ready for listening.
1356 int sysctl_somaxconn __read_mostly = SOMAXCONN;
1358 asmlinkage long sys_listen(int fd, int backlog)
1360 struct socket *sock;
1361 int err, fput_needed;
1363 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1365 if ((unsigned)backlog > sysctl_somaxconn)
1366 backlog = sysctl_somaxconn;
1368 err = security_socket_listen(sock, backlog);
1370 err = sock->ops->listen(sock, backlog);
1372 fput_light(sock->file, fput_needed);
1378 * For accept, we attempt to create a new socket, set up the link
1379 * with the client, wake up the client, then return the new
1380 * connected fd. We collect the address of the connector in kernel
1381 * space and move it to user at the very end. This is unclean because
1382 * we open the socket then return an error.
1384 * 1003.1g adds the ability to recvmsg() to query connection pending
1385 * status to recvmsg. We need to add that support in a way thats
1386 * clean when we restucture accept also.
1389 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1390 int __user *upeer_addrlen)
1392 struct socket *sock, *newsock;
1393 struct file *newfile;
1394 int err, len, newfd, fput_needed;
1395 char address[MAX_SOCK_ADDR];
1397 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1402 if (!(newsock = sock_alloc()))
1405 newsock->type = sock->type;
1406 newsock->ops = sock->ops;
1409 * We don't need try_module_get here, as the listening socket (sock)
1410 * has the protocol module (sock->ops->owner) held.
1412 __module_get(newsock->ops->owner);
1414 newfd = sock_alloc_fd(&newfile);
1415 if (unlikely(newfd < 0)) {
1417 sock_release(newsock);
1421 err = sock_attach_fd(newsock, newfile);
1425 err = security_socket_accept(sock, newsock);
1429 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1433 if (upeer_sockaddr) {
1434 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1436 err = -ECONNABORTED;
1439 err = move_addr_to_user(address, len, upeer_sockaddr,
1445 /* File flags are not inherited via accept() unlike another OSes. */
1447 fd_install(newfd, newfile);
1450 security_socket_post_accept(sock, newsock);
1453 fput_light(sock->file, fput_needed);
1457 sock_release(newsock);
1459 put_unused_fd(newfd);
1463 put_unused_fd(newfd);
1468 * Attempt to connect to a socket with the server address. The address
1469 * is in user space so we verify it is OK and move it to kernel space.
1471 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1474 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1475 * other SEQPACKET protocols that take time to connect() as it doesn't
1476 * include the -EINPROGRESS status for such sockets.
1479 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1482 struct socket *sock;
1483 char address[MAX_SOCK_ADDR];
1484 int err, fput_needed;
1486 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1489 err = move_addr_to_kernel(uservaddr, addrlen, address);
1494 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1498 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1499 sock->file->f_flags);
1501 fput_light(sock->file, fput_needed);
1507 * Get the local address ('name') of a socket object. Move the obtained
1508 * name to user space.
1511 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1512 int __user *usockaddr_len)
1514 struct socket *sock;
1515 char address[MAX_SOCK_ADDR];
1516 int len, err, fput_needed;
1518 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1522 err = security_socket_getsockname(sock);
1526 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1529 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1532 fput_light(sock->file, fput_needed);
1538 * Get the remote address ('name') of a socket object. Move the obtained
1539 * name to user space.
1542 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1543 int __user *usockaddr_len)
1545 struct socket *sock;
1546 char address[MAX_SOCK_ADDR];
1547 int len, err, fput_needed;
1549 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1551 err = security_socket_getpeername(sock);
1553 fput_light(sock->file, fput_needed);
1558 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1561 err = move_addr_to_user(address, len, usockaddr,
1563 fput_light(sock->file, fput_needed);
1569 * Send a datagram to a given address. We move the address into kernel
1570 * space and check the user space data area is readable before invoking
1574 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1575 unsigned flags, struct sockaddr __user *addr,
1578 struct socket *sock;
1579 char address[MAX_SOCK_ADDR];
1584 struct file *sock_file;
1586 sock_file = fget_light(fd, &fput_needed);
1591 sock = sock_from_file(sock_file, &err);
1594 iov.iov_base = buff;
1596 msg.msg_name = NULL;
1599 msg.msg_control = NULL;
1600 msg.msg_controllen = 0;
1601 msg.msg_namelen = 0;
1603 err = move_addr_to_kernel(addr, addr_len, address);
1606 msg.msg_name = address;
1607 msg.msg_namelen = addr_len;
1609 if (sock->file->f_flags & O_NONBLOCK)
1610 flags |= MSG_DONTWAIT;
1611 msg.msg_flags = flags;
1612 err = sock_sendmsg(sock, &msg, len);
1615 fput_light(sock_file, fput_needed);
1621 * Send a datagram down a socket.
1624 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1626 return sys_sendto(fd, buff, len, flags, NULL, 0);
1630 * Receive a frame from the socket and optionally record the address of the
1631 * sender. We verify the buffers are writable and if needed move the
1632 * sender address from kernel to user space.
1635 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1636 unsigned flags, struct sockaddr __user *addr,
1637 int __user *addr_len)
1639 struct socket *sock;
1642 char address[MAX_SOCK_ADDR];
1644 struct file *sock_file;
1647 sock_file = fget_light(fd, &fput_needed);
1652 sock = sock_from_file(sock_file, &err);
1656 msg.msg_control = NULL;
1657 msg.msg_controllen = 0;
1661 iov.iov_base = ubuf;
1662 msg.msg_name = address;
1663 msg.msg_namelen = MAX_SOCK_ADDR;
1664 if (sock->file->f_flags & O_NONBLOCK)
1665 flags |= MSG_DONTWAIT;
1666 err = sock_recvmsg(sock, &msg, size, flags);
1668 if (err >= 0 && addr != NULL) {
1669 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1674 fput_light(sock_file, fput_needed);
1680 * Receive a datagram from a socket.
1683 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1686 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1690 * Set a socket option. Because we don't know the option lengths we have
1691 * to pass the user mode parameter for the protocols to sort out.
1694 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1695 char __user *optval, int optlen)
1697 int err, fput_needed;
1698 struct socket *sock;
1703 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1705 err = security_socket_setsockopt(sock, level, optname);
1709 if (level == SOL_SOCKET)
1711 sock_setsockopt(sock, level, optname, optval,
1715 sock->ops->setsockopt(sock, level, optname, optval,
1718 fput_light(sock->file, fput_needed);
1724 * Get a socket option. Because we don't know the option lengths we have
1725 * to pass a user mode parameter for the protocols to sort out.
1728 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1729 char __user *optval, int __user *optlen)
1731 int err, fput_needed;
1732 struct socket *sock;
1734 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1736 err = security_socket_getsockopt(sock, level, optname);
1740 if (level == SOL_SOCKET)
1742 sock_getsockopt(sock, level, optname, optval,
1746 sock->ops->getsockopt(sock, level, optname, optval,
1749 fput_light(sock->file, fput_needed);
1755 * Shutdown a socket.
1758 asmlinkage long sys_shutdown(int fd, int how)
1760 int err, fput_needed;
1761 struct socket *sock;
1763 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1765 err = security_socket_shutdown(sock, how);
1767 err = sock->ops->shutdown(sock, how);
1768 fput_light(sock->file, fput_needed);
1773 /* A couple of helpful macros for getting the address of the 32/64 bit
1774 * fields which are the same type (int / unsigned) on our platforms.
1776 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1777 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1778 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1781 * BSD sendmsg interface
1784 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1786 struct compat_msghdr __user *msg_compat =
1787 (struct compat_msghdr __user *)msg;
1788 struct socket *sock;
1789 char address[MAX_SOCK_ADDR];
1790 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1791 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1792 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1793 /* 20 is size of ipv6_pktinfo */
1794 unsigned char *ctl_buf = ctl;
1795 struct msghdr msg_sys;
1796 int err, ctl_len, iov_size, total_len;
1800 if (MSG_CMSG_COMPAT & flags) {
1801 if (get_compat_msghdr(&msg_sys, msg_compat))
1804 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1807 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1811 /* do not move before msg_sys is valid */
1813 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1816 /* Check whether to allocate the iovec area */
1818 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1819 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1820 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1825 /* This will also move the address data into kernel space */
1826 if (MSG_CMSG_COMPAT & flags) {
1827 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1829 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1836 if (msg_sys.msg_controllen > INT_MAX)
1838 ctl_len = msg_sys.msg_controllen;
1839 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1841 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1845 ctl_buf = msg_sys.msg_control;
1846 ctl_len = msg_sys.msg_controllen;
1847 } else if (ctl_len) {
1848 if (ctl_len > sizeof(ctl)) {
1849 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1850 if (ctl_buf == NULL)
1855 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1856 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1857 * checking falls down on this.
1859 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1862 msg_sys.msg_control = ctl_buf;
1864 msg_sys.msg_flags = flags;
1866 if (sock->file->f_flags & O_NONBLOCK)
1867 msg_sys.msg_flags |= MSG_DONTWAIT;
1868 err = sock_sendmsg(sock, &msg_sys, total_len);
1872 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1874 if (iov != iovstack)
1875 sock_kfree_s(sock->sk, iov, iov_size);
1877 fput_light(sock->file, fput_needed);
1883 * BSD recvmsg interface
1886 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1889 struct compat_msghdr __user *msg_compat =
1890 (struct compat_msghdr __user *)msg;
1891 struct socket *sock;
1892 struct iovec iovstack[UIO_FASTIOV];
1893 struct iovec *iov = iovstack;
1894 struct msghdr msg_sys;
1895 unsigned long cmsg_ptr;
1896 int err, iov_size, total_len, len;
1899 /* kernel mode address */
1900 char addr[MAX_SOCK_ADDR];
1902 /* user mode address pointers */
1903 struct sockaddr __user *uaddr;
1904 int __user *uaddr_len;
1906 if (MSG_CMSG_COMPAT & flags) {
1907 if (get_compat_msghdr(&msg_sys, msg_compat))
1910 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1913 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1918 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1921 /* Check whether to allocate the iovec area */
1923 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1924 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1925 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1931 * Save the user-mode address (verify_iovec will change the
1932 * kernel msghdr to use the kernel address space)
1935 uaddr = (__force void __user *)msg_sys.msg_name;
1936 uaddr_len = COMPAT_NAMELEN(msg);
1937 if (MSG_CMSG_COMPAT & flags) {
1938 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1940 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1945 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1946 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
1948 if (sock->file->f_flags & O_NONBLOCK)
1949 flags |= MSG_DONTWAIT;
1950 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1955 if (uaddr != NULL) {
1956 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1961 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1965 if (MSG_CMSG_COMPAT & flags)
1966 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1967 &msg_compat->msg_controllen);
1969 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1970 &msg->msg_controllen);
1976 if (iov != iovstack)
1977 sock_kfree_s(sock->sk, iov, iov_size);
1979 fput_light(sock->file, fput_needed);
1984 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1986 /* Argument list sizes for sys_socketcall */
1987 #define AL(x) ((x) * sizeof(unsigned long))
1988 static const unsigned char nargs[18]={
1989 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1990 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1991 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1997 * System call vectors.
1999 * Argument checking cleaned up. Saved 20% in size.
2000 * This function doesn't need to set the kernel lock because
2001 * it is set by the callees.
2004 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2007 unsigned long a0, a1;
2010 if (call < 1 || call > SYS_RECVMSG)
2013 /* copy_from_user should be SMP safe. */
2014 if (copy_from_user(a, args, nargs[call]))
2017 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2026 err = sys_socket(a0, a1, a[2]);
2029 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2032 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2035 err = sys_listen(a0, a1);
2039 sys_accept(a0, (struct sockaddr __user *)a1,
2040 (int __user *)a[2]);
2042 case SYS_GETSOCKNAME:
2044 sys_getsockname(a0, (struct sockaddr __user *)a1,
2045 (int __user *)a[2]);
2047 case SYS_GETPEERNAME:
2049 sys_getpeername(a0, (struct sockaddr __user *)a1,
2050 (int __user *)a[2]);
2052 case SYS_SOCKETPAIR:
2053 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2056 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2059 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2060 (struct sockaddr __user *)a[4], a[5]);
2063 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2066 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2067 (struct sockaddr __user *)a[4],
2068 (int __user *)a[5]);
2071 err = sys_shutdown(a0, a1);
2073 case SYS_SETSOCKOPT:
2074 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2076 case SYS_GETSOCKOPT:
2078 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2079 (int __user *)a[4]);
2082 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2085 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2094 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2097 * sock_register - add a socket protocol handler
2098 * @ops: description of protocol
2100 * This function is called by a protocol handler that wants to
2101 * advertise its address family, and have it linked into the
2102 * socket interface. The value ops->family coresponds to the
2103 * socket system call protocol family.
2105 int sock_register(const struct net_proto_family *ops)
2109 if (ops->family >= NPROTO) {
2110 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2115 spin_lock(&net_family_lock);
2116 if (net_families[ops->family])
2119 net_families[ops->family] = ops;
2122 spin_unlock(&net_family_lock);
2124 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2129 * sock_unregister - remove a protocol handler
2130 * @family: protocol family to remove
2132 * This function is called by a protocol handler that wants to
2133 * remove its address family, and have it unlinked from the
2134 * new socket creation.
2136 * If protocol handler is a module, then it can use module reference
2137 * counts to protect against new references. If protocol handler is not
2138 * a module then it needs to provide its own protection in
2139 * the ops->create routine.
2141 void sock_unregister(int family)
2143 BUG_ON(family < 0 || family >= NPROTO);
2145 spin_lock(&net_family_lock);
2146 net_families[family] = NULL;
2147 spin_unlock(&net_family_lock);
2151 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2154 static int __init sock_init(void)
2157 * Initialize sock SLAB cache.
2163 * Initialize skbuff SLAB cache
2168 * Initialize the protocols module.
2172 register_filesystem(&sock_fs_type);
2173 sock_mnt = kern_mount(&sock_fs_type);
2175 /* The real protocol initialization is performed in later initcalls.
2178 #ifdef CONFIG_NETFILTER
2185 core_initcall(sock_init); /* early initcall */
2187 #ifdef CONFIG_PROC_FS
2188 void socket_seq_show(struct seq_file *seq)
2193 for_each_possible_cpu(cpu)
2194 counter += per_cpu(sockets_in_use, cpu);
2196 /* It can be negative, by the way. 8) */
2200 seq_printf(seq, "sockets: used %d\n", counter);
2202 #endif /* CONFIG_PROC_FS */
2204 #ifdef CONFIG_COMPAT
2205 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2208 struct socket *sock = file->private_data;
2209 int ret = -ENOIOCTLCMD;
2211 if (sock->ops->compat_ioctl)
2212 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2218 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2220 return sock->ops->bind(sock, addr, addrlen);
2223 int kernel_listen(struct socket *sock, int backlog)
2225 return sock->ops->listen(sock, backlog);
2228 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2230 struct sock *sk = sock->sk;
2233 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2238 err = sock->ops->accept(sock, *newsock, flags);
2240 sock_release(*newsock);
2245 (*newsock)->ops = sock->ops;
2251 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2254 return sock->ops->connect(sock, addr, addrlen, flags);
2257 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2260 return sock->ops->getname(sock, addr, addrlen, 0);
2263 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2266 return sock->ops->getname(sock, addr, addrlen, 1);
2269 int kernel_getsockopt(struct socket *sock, int level, int optname,
2270 char *optval, int *optlen)
2272 mm_segment_t oldfs = get_fs();
2276 if (level == SOL_SOCKET)
2277 err = sock_getsockopt(sock, level, optname, optval, optlen);
2279 err = sock->ops->getsockopt(sock, level, optname, optval,
2285 int kernel_setsockopt(struct socket *sock, int level, int optname,
2286 char *optval, int optlen)
2288 mm_segment_t oldfs = get_fs();
2292 if (level == SOL_SOCKET)
2293 err = sock_setsockopt(sock, level, optname, optval, optlen);
2295 err = sock->ops->setsockopt(sock, level, optname, optval,
2301 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2302 size_t size, int flags)
2304 if (sock->ops->sendpage)
2305 return sock->ops->sendpage(sock, page, offset, size, flags);
2307 return sock_no_sendpage(sock, page, offset, size, flags);
2310 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2312 mm_segment_t oldfs = get_fs();
2316 err = sock->ops->ioctl(sock, cmd, arg);
2322 /* ABI emulation layers need these two */
2323 EXPORT_SYMBOL(move_addr_to_kernel);
2324 EXPORT_SYMBOL(move_addr_to_user);
2325 EXPORT_SYMBOL(sock_create);
2326 EXPORT_SYMBOL(sock_create_kern);
2327 EXPORT_SYMBOL(sock_create_lite);
2328 EXPORT_SYMBOL(sock_map_fd);
2329 EXPORT_SYMBOL(sock_recvmsg);
2330 EXPORT_SYMBOL(sock_register);
2331 EXPORT_SYMBOL(sock_release);
2332 EXPORT_SYMBOL(sock_sendmsg);
2333 EXPORT_SYMBOL(sock_unregister);
2334 EXPORT_SYMBOL(sock_wake_async);
2335 EXPORT_SYMBOL(sockfd_lookup);
2336 EXPORT_SYMBOL(kernel_sendmsg);
2337 EXPORT_SYMBOL(kernel_recvmsg);
2338 EXPORT_SYMBOL(kernel_bind);
2339 EXPORT_SYMBOL(kernel_listen);
2340 EXPORT_SYMBOL(kernel_accept);
2341 EXPORT_SYMBOL(kernel_connect);
2342 EXPORT_SYMBOL(kernel_getsockname);
2343 EXPORT_SYMBOL(kernel_getpeername);
2344 EXPORT_SYMBOL(kernel_getsockopt);
2345 EXPORT_SYMBOL(kernel_setsockopt);
2346 EXPORT_SYMBOL(kernel_sendpage);
2347 EXPORT_SYMBOL(kernel_sock_ioctl);