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, <bir7@leland.Stanford.Edu>
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
61 #include <linux/config.h>
63 #include <linux/smp_lock.h>
64 #include <linux/socket.h>
65 #include <linux/file.h>
66 #include <linux/net.h>
67 #include <linux/interrupt.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
72 #include <linux/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/init.h>
75 #include <linux/poll.h>
76 #include <linux/cache.h>
77 #include <linux/module.h>
78 #include <linux/highmem.h>
79 #include <linux/divert.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>
86 #ifdef CONFIG_NET_RADIO
87 #include <linux/wireless.h> /* Note : will define WIRELESS_EXT */
88 #endif /* CONFIG_NET_RADIO */
90 #include <asm/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
96 #include <linux/netfilter.h>
97 #include <linux/vs_base.h>
98 #include <linux/vs_socket.h>
100 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
101 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
102 size_t size, loff_t pos);
103 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
104 size_t size, 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 int sock_ioctl(struct inode *inode, struct file *file,
111 unsigned int cmd, unsigned long arg);
112 static int sock_fasync(int fd, struct file *filp, int on);
113 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
114 unsigned long count, loff_t *ppos);
115 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
116 unsigned long count, loff_t *ppos);
117 static ssize_t sock_sendpage(struct file *file, struct page *page,
118 int offset, size_t size, loff_t *ppos, int more);
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
126 struct file_operations socket_file_ops = {
127 .owner = THIS_MODULE,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
134 .open = sock_no_open, /* special open code to disallow open via /proc */
135 .release = sock_close,
136 .fasync = sock_fasync,
138 .writev = sock_writev,
139 .sendpage = sock_sendpage
143 * The protocol list. Each protocol is registered in here.
146 static struct net_proto_family *net_families[NPROTO];
148 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
149 static atomic_t net_family_lockct = ATOMIC_INIT(0);
150 static spinlock_t net_family_lock = SPIN_LOCK_UNLOCKED;
152 /* The strategy is: modifications net_family vector are short, do not
153 sleep and veeery rare, but read access should be free of any exclusive
157 static void net_family_write_lock(void)
159 spin_lock(&net_family_lock);
160 while (atomic_read(&net_family_lockct) != 0) {
161 spin_unlock(&net_family_lock);
165 spin_lock(&net_family_lock);
169 static __inline__ void net_family_write_unlock(void)
171 spin_unlock(&net_family_lock);
174 static __inline__ void net_family_read_lock(void)
176 atomic_inc(&net_family_lockct);
177 spin_unlock_wait(&net_family_lock);
180 static __inline__ void net_family_read_unlock(void)
182 atomic_dec(&net_family_lockct);
186 #define net_family_write_lock() do { } while(0)
187 #define net_family_write_unlock() do { } while(0)
188 #define net_family_read_lock() do { } while(0)
189 #define net_family_read_unlock() do { } while(0)
194 * Statistics counters of the socket lists
197 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
200 * Support routines. Move socket addresses back and forth across the kernel/user
201 * divide and look after the messy bits.
204 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
205 16 for IP, 16 for IPX,
208 must be at least one bigger than
209 the AF_UNIX size (see net/unix/af_unix.c
214 * move_addr_to_kernel - copy a socket address into kernel space
215 * @uaddr: Address in user space
216 * @kaddr: Address in kernel space
217 * @ulen: Length in user space
219 * The address is copied into kernel space. If the provided address is
220 * too long an error code of -EINVAL is returned. If the copy gives
221 * invalid addresses -EFAULT is returned. On a success 0 is returned.
224 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
226 if(ulen<0||ulen>MAX_SOCK_ADDR)
230 if(copy_from_user(kaddr,uaddr,ulen))
236 * move_addr_to_user - copy an address to user space
237 * @kaddr: kernel space address
238 * @klen: length of address in kernel
239 * @uaddr: user space address
240 * @ulen: pointer to user length field
242 * The value pointed to by ulen on entry is the buffer length available.
243 * This is overwritten with the buffer space used. -EINVAL is returned
244 * if an overlong buffer is specified or a negative buffer size. -EFAULT
245 * is returned if either the buffer or the length field are not
247 * After copying the data up to the limit the user specifies, the true
248 * length of the data is written over the length limit the user
249 * specified. Zero is returned for a success.
252 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
257 if((err=get_user(len, ulen)))
261 if(len<0 || len> MAX_SOCK_ADDR)
265 if(copy_to_user(uaddr,kaddr,len))
269 * "fromlen shall refer to the value before truncation.."
272 return __put_user(klen, ulen);
275 #define SOCKFS_MAGIC 0x534F434B
277 static kmem_cache_t * sock_inode_cachep;
279 static struct inode *sock_alloc_inode(struct super_block *sb)
281 struct socket_alloc *ei;
282 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
285 init_waitqueue_head(&ei->socket.wait);
287 ei->socket.fasync_list = NULL;
288 ei->socket.state = SS_UNCONNECTED;
289 ei->socket.flags = 0;
290 ei->socket.ops = NULL;
291 ei->socket.sk = NULL;
292 ei->socket.file = NULL;
293 ei->socket.flags = 0;
295 return &ei->vfs_inode;
298 static void sock_destroy_inode(struct inode *inode)
300 kmem_cache_free(sock_inode_cachep,
301 container_of(inode, struct socket_alloc, vfs_inode));
304 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
306 struct socket_alloc *ei = (struct socket_alloc *) foo;
308 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
309 SLAB_CTOR_CONSTRUCTOR)
310 inode_init_once(&ei->vfs_inode);
313 static int init_inodecache(void)
315 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
316 sizeof(struct socket_alloc),
317 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
319 if (sock_inode_cachep == NULL)
324 static struct super_operations sockfs_ops = {
325 .alloc_inode = sock_alloc_inode,
326 .destroy_inode =sock_destroy_inode,
327 .statfs = simple_statfs,
330 static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
331 int flags, const char *dev_name, void *data)
333 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
336 static struct vfsmount *sock_mnt;
338 static struct file_system_type sock_fs_type = {
340 .get_sb = sockfs_get_sb,
341 .kill_sb = kill_anon_super,
343 static int sockfs_delete_dentry(struct dentry *dentry)
347 static struct dentry_operations sockfs_dentry_operations = {
348 .d_delete = sockfs_delete_dentry,
352 * Obtains the first available file descriptor and sets it up for use.
354 * This function creates file structure and maps it to fd space
355 * of current process. On success it returns file descriptor
356 * and file struct implicitly stored in sock->file.
357 * Note that another thread may close file descriptor before we return
358 * from this function. We use the fact that now we do not refer
359 * to socket after mapping. If one day we will need it, this
360 * function will increment ref. count on file by 1.
362 * In any case returned fd MAY BE not valid!
363 * This race condition is unavoidable
364 * with shared fd spaces, we cannot solve it inside kernel,
365 * but we take care of internal coherence yet.
368 struct file * sock_map_file(struct socket *sock)
374 file = get_empty_filp();
377 return ERR_PTR(-ENFILE);
379 sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
381 this.len = strlen(name);
382 this.hash = SOCK_INODE(sock)->i_ino;
384 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
385 if (!file->f_dentry) {
387 return ERR_PTR(-ENOMEM);
389 file->f_dentry->d_op = &sockfs_dentry_operations;
390 d_add(file->f_dentry, SOCK_INODE(sock));
391 file->f_vfsmnt = mntget(sock_mnt);
392 file->f_mapping = file->f_dentry->d_inode->i_mapping;
397 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
398 file->f_mode = FMODE_READ | FMODE_WRITE;
399 file->f_flags = O_RDWR;
405 int sock_map_fd(struct socket *sock)
411 * Find a file descriptor suitable for return to the user.
414 fd = get_unused_fd();
418 file = sock_map_file(sock);
421 return PTR_ERR(file);
423 fd_install(fd, file);
429 * sockfd_lookup - Go from a file number to its socket slot
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)
447 if (!(file = fget(fd)))
453 inode = file->f_dentry->d_inode;
454 if (!inode->i_sock || !(sock = SOCKET_I(inode)))
461 if (sock->file != file) {
462 printk(KERN_ERR "socki_lookup: socket file changed!\n");
469 * sock_alloc - allocate a socket
471 * Allocate a new inode and socket object. The two are bound together
472 * and initialised. The socket is then returned. If we are out of inodes
476 struct socket *sock_alloc(void)
478 struct inode * inode;
479 struct socket * sock;
481 inode = new_inode(sock_mnt->mnt_sb);
485 sock = SOCKET_I(inode);
487 inode->i_mode = S_IFSOCK|S_IRWXUGO;
489 inode->i_uid = current->fsuid;
490 inode->i_gid = current->fsgid;
492 get_cpu_var(sockets_in_use)++;
493 put_cpu_var(sockets_in_use);
497 EXPORT_SYMBOL_GPL(sock_alloc);
500 * In theory you can't get an open on this inode, but /proc provides
501 * a back door. Remember to keep it shut otherwise you'll let the
502 * creepy crawlies in.
505 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
510 struct file_operations bad_sock_fops = {
511 .owner = THIS_MODULE,
512 .open = sock_no_open,
516 * sock_release - close a socket
517 * @sock: socket to close
519 * The socket is released from the protocol stack if it has a release
520 * callback, and the inode is then released if the socket is bound to
521 * an inode not a file.
524 void sock_release(struct socket *sock)
527 struct module *owner = sock->ops->owner;
529 sock->ops->release(sock);
534 if (sock->fasync_list)
535 printk(KERN_ERR "sock_release: fasync list not empty!\n");
537 get_cpu_var(sockets_in_use)--;
538 put_cpu_var(sockets_in_use);
540 iput(SOCK_INODE(sock));
546 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
547 struct msghdr *msg, size_t size)
549 struct sock_iocb *si = kiocb_to_siocb(iocb);
557 err = security_socket_sendmsg(sock, msg, size);
561 len = sock->ops->sendmsg(iocb, sock, msg, size);
564 vx_sock_send(sock->sk, size);
566 vx_sock_fail(sock->sk, size);
568 vxdprintk(VXD_CBIT(net, 7),
569 "__sock_sendmsg: %p[%p,%p,%p;%d]:%d/%d",
571 (sock->sk)?sock->sk->sk_nx_info:0,
572 (sock->sk)?sock->sk->sk_vx_info:0,
573 (sock->sk)?sock->sk->sk_xid:0,
574 (unsigned int)size, len);
578 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
581 struct sock_iocb siocb;
584 init_sync_kiocb(&iocb, NULL);
585 iocb.private = &siocb;
586 ret = __sock_sendmsg(&iocb, sock, msg, size);
587 if (-EIOCBQUEUED == ret)
588 ret = wait_on_sync_kiocb(&iocb);
592 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
593 struct kvec *vec, size_t num, size_t size)
595 mm_segment_t oldfs = get_fs();
600 * the following is safe, since for compiler definitions of kvec and
601 * iovec are identical, yielding the same in-core layout and alignment
603 msg->msg_iov = (struct iovec *)vec,
604 msg->msg_iovlen = num;
605 result = sock_sendmsg(sock, msg, size);
610 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
611 struct msghdr *msg, size_t size, int flags)
614 struct sock_iocb *si = kiocb_to_siocb(iocb);
622 err = security_socket_recvmsg(sock, msg, size, flags);
626 len = sock->ops->recvmsg(iocb, sock, msg, size, flags);
627 if ((len >= 0) && sock->sk)
628 vx_sock_recv(sock->sk, len);
629 vxdprintk(VXD_CBIT(net, 7),
630 "__sock_recvmsg: %p[%p,%p,%p;%d]:%d/%d",
632 (sock->sk)?sock->sk->sk_nx_info:0,
633 (sock->sk)?sock->sk->sk_vx_info:0,
634 (sock->sk)?sock->sk->sk_xid:0,
635 (unsigned int)size, len);
639 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
640 size_t size, int flags)
643 struct sock_iocb siocb;
646 init_sync_kiocb(&iocb, NULL);
647 iocb.private = &siocb;
648 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
649 if (-EIOCBQUEUED == ret)
650 ret = wait_on_sync_kiocb(&iocb);
654 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
655 struct kvec *vec, size_t num,
656 size_t size, int flags)
658 mm_segment_t oldfs = get_fs();
663 * the following is safe, since for compiler definitions of kvec and
664 * iovec are identical, yielding the same in-core layout and alignment
666 msg->msg_iov = (struct iovec *)vec,
667 msg->msg_iovlen = num;
668 result = sock_recvmsg(sock, msg, size, flags);
673 static void sock_aio_dtor(struct kiocb *iocb)
675 kfree(iocb->private);
679 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
680 * area ubuf...ubuf+size-1 is writable before asking the protocol.
683 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
684 size_t size, loff_t pos)
686 struct sock_iocb *x, siocb;
692 if (size==0) /* Match SYS5 behaviour */
695 if (is_sync_kiocb(iocb))
698 x = kmalloc(sizeof(struct sock_iocb), GFP_KERNEL);
701 iocb->ki_dtor = sock_aio_dtor;
705 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
707 x->async_msg.msg_name = NULL;
708 x->async_msg.msg_namelen = 0;
709 x->async_msg.msg_iov = &x->async_iov;
710 x->async_msg.msg_iovlen = 1;
711 x->async_msg.msg_control = NULL;
712 x->async_msg.msg_controllen = 0;
713 x->async_iov.iov_base = ubuf;
714 x->async_iov.iov_len = size;
715 flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
717 return __sock_recvmsg(iocb, sock, &x->async_msg, size, flags);
722 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
723 * is readable by the user process.
726 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
727 size_t size, loff_t pos)
729 struct sock_iocb *x, siocb;
734 if(size==0) /* Match SYS5 behaviour */
737 if (is_sync_kiocb(iocb))
740 x = kmalloc(sizeof(struct sock_iocb), GFP_KERNEL);
743 iocb->ki_dtor = sock_aio_dtor;
747 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
749 x->async_msg.msg_name = NULL;
750 x->async_msg.msg_namelen = 0;
751 x->async_msg.msg_iov = &x->async_iov;
752 x->async_msg.msg_iovlen = 1;
753 x->async_msg.msg_control = NULL;
754 x->async_msg.msg_controllen = 0;
755 x->async_msg.msg_flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
756 if (sock->type == SOCK_SEQPACKET)
757 x->async_msg.msg_flags |= MSG_EOR;
758 x->async_iov.iov_base = (void __user *)ubuf;
759 x->async_iov.iov_len = size;
761 return __sock_sendmsg(iocb, sock, &x->async_msg, size);
764 ssize_t sock_sendpage(struct file *file, struct page *page,
765 int offset, size_t size, loff_t *ppos, int more)
770 sock = SOCKET_I(file->f_dentry->d_inode);
772 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
776 return sock->ops->sendpage(sock, page, offset, size, flags);
779 int sock_readv_writev(int type, struct inode * inode, struct file * file,
780 const struct iovec * iov, long count, size_t size)
785 sock = SOCKET_I(inode);
789 msg.msg_control = NULL;
790 msg.msg_controllen = 0;
791 msg.msg_iov = (struct iovec *) iov;
792 msg.msg_iovlen = count;
793 msg.msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
795 /* read() does a VERIFY_WRITE */
796 if (type == VERIFY_WRITE)
797 return sock_recvmsg(sock, &msg, size, msg.msg_flags);
799 if (sock->type == SOCK_SEQPACKET)
800 msg.msg_flags |= MSG_EOR;
802 return sock_sendmsg(sock, &msg, size);
805 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
806 unsigned long count, loff_t *ppos)
810 for (i = 0 ; i < count ; i++)
811 tot_len += vector[i].iov_len;
812 return sock_readv_writev(VERIFY_WRITE, file->f_dentry->d_inode,
813 file, vector, count, tot_len);
816 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
817 unsigned long count, loff_t *ppos)
821 for (i = 0 ; i < count ; i++)
822 tot_len += vector[i].iov_len;
823 return sock_readv_writev(VERIFY_READ, file->f_dentry->d_inode,
824 file, vector, count, tot_len);
829 * Atomic setting of ioctl hooks to avoid race
830 * with module unload.
833 static DECLARE_MUTEX(br_ioctl_mutex);
834 static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
836 void brioctl_set(int (*hook)(unsigned int, void __user *))
838 down(&br_ioctl_mutex);
839 br_ioctl_hook = hook;
842 EXPORT_SYMBOL(brioctl_set);
844 static DECLARE_MUTEX(vlan_ioctl_mutex);
845 static int (*vlan_ioctl_hook)(void __user *arg);
847 void vlan_ioctl_set(int (*hook)(void __user *))
849 down(&vlan_ioctl_mutex);
850 vlan_ioctl_hook = hook;
851 up(&vlan_ioctl_mutex);
853 EXPORT_SYMBOL(vlan_ioctl_set);
855 static DECLARE_MUTEX(dlci_ioctl_mutex);
856 static int (*dlci_ioctl_hook)(unsigned int, void __user *);
858 void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
860 down(&dlci_ioctl_mutex);
861 dlci_ioctl_hook = hook;
862 up(&dlci_ioctl_mutex);
864 EXPORT_SYMBOL(dlci_ioctl_set);
867 * With an ioctl, arg may well be a user mode pointer, but we don't know
868 * what to do with it - that's up to the protocol still.
871 static int sock_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
875 void __user *argp = (void __user *)arg;
879 sock = SOCKET_I(inode);
880 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
881 err = dev_ioctl(cmd, argp);
884 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
885 err = dev_ioctl(cmd, argp);
887 #endif /* WIRELESS_EXT */
892 if (get_user(pid, (int __user *)argp))
894 err = f_setown(sock->file, pid, 1);
898 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
906 request_module("bridge");
908 down(&br_ioctl_mutex);
910 err = br_ioctl_hook(cmd, argp);
916 if (!vlan_ioctl_hook)
917 request_module("8021q");
919 down(&vlan_ioctl_mutex);
921 err = vlan_ioctl_hook(argp);
922 up(&vlan_ioctl_mutex);
926 /* Convert this to call through a hook */
927 err = divert_ioctl(cmd, argp);
932 if (!dlci_ioctl_hook)
933 request_module("dlci");
935 if (dlci_ioctl_hook) {
936 down(&dlci_ioctl_mutex);
937 err = dlci_ioctl_hook(cmd, argp);
938 up(&dlci_ioctl_mutex);
942 err = sock->ops->ioctl(sock, cmd, arg);
950 int sock_create_lite(int family, int type, int protocol, struct socket **res)
953 struct socket *sock = NULL;
955 err = security_socket_create(family, type, protocol, 1);
965 security_socket_post_create(sock, family, type, protocol, 1);
972 /* No kernel lock held - perfect */
973 static unsigned int sock_poll(struct file *file, poll_table * wait)
978 * We can't return errors to poll, so it's either yes or no.
980 sock = SOCKET_I(file->f_dentry->d_inode);
981 return sock->ops->poll(file, sock, wait);
984 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
986 struct socket *sock = SOCKET_I(file->f_dentry->d_inode);
988 return sock->ops->mmap(file, sock, vma);
991 int sock_close(struct inode *inode, struct file *filp)
994 * It was possible the inode is NULL we were
995 * closing an unfinished socket.
1000 printk(KERN_DEBUG "sock_close: NULL inode\n");
1003 sock_fasync(-1, filp, 0);
1004 sock_release(SOCKET_I(inode));
1009 * Update the socket async list
1011 * Fasync_list locking strategy.
1013 * 1. fasync_list is modified only under process context socket lock
1014 * i.e. under semaphore.
1015 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1016 * or under socket lock.
1017 * 3. fasync_list can be used from softirq context, so that
1018 * modification under socket lock have to be enhanced with
1019 * write_lock_bh(&sk->sk_callback_lock).
1023 static int sock_fasync(int fd, struct file *filp, int on)
1025 struct fasync_struct *fa, *fna=NULL, **prev;
1026 struct socket *sock;
1031 fna=(struct fasync_struct *)kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1036 sock = SOCKET_I(filp->f_dentry->d_inode);
1038 if ((sk=sock->sk) == NULL) {
1046 prev=&(sock->fasync_list);
1048 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1049 if (fa->fa_file==filp)
1056 write_lock_bh(&sk->sk_callback_lock);
1058 write_unlock_bh(&sk->sk_callback_lock);
1065 fna->magic=FASYNC_MAGIC;
1066 fna->fa_next=sock->fasync_list;
1067 write_lock_bh(&sk->sk_callback_lock);
1068 sock->fasync_list=fna;
1069 write_unlock_bh(&sk->sk_callback_lock);
1075 write_lock_bh(&sk->sk_callback_lock);
1077 write_unlock_bh(&sk->sk_callback_lock);
1085 release_sock(sock->sk);
1089 /* This function may be called only under socket lock or callback_lock */
1091 int sock_wake_async(struct socket *sock, int how, int band)
1093 if (!sock || !sock->fasync_list)
1099 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1103 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1108 __kill_fasync(sock->fasync_list, SIGIO, band);
1111 __kill_fasync(sock->fasync_list, SIGURG, band);
1116 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1120 struct socket *sock;
1123 * Check protocol is in range
1125 if (family < 0 || family >= NPROTO)
1126 return -EAFNOSUPPORT;
1127 if (type < 0 || type >= SOCK_MAX)
1130 /* disable IPv6 inside vservers for now */
1131 if (family == PF_INET6 && !vx_check(0, VX_ADMIN))
1132 return -EAFNOSUPPORT;
1136 This uglymoron is moved from INET layer to here to avoid
1137 deadlock in module load.
1139 if (family == PF_INET && type == SOCK_PACKET) {
1143 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1148 err = security_socket_create(family, type, protocol, kern);
1152 #if defined(CONFIG_KMOD)
1153 /* Attempt to load a protocol module if the find failed.
1155 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1156 * requested real, full-featured networking support upon configuration.
1157 * Otherwise module support will break!
1159 if (net_families[family]==NULL)
1161 request_module("net-pf-%d",family);
1165 net_family_read_lock();
1166 if (net_families[family] == NULL) {
1172 * Allocate the socket and allow the family to set things up. if
1173 * the protocol is 0, the family is instructed to select an appropriate
1177 if (!(sock = sock_alloc()))
1179 printk(KERN_WARNING "socket: no more sockets\n");
1180 i = -ENFILE; /* Not exactly a match, but its the
1181 closest posix thing */
1188 * We will call the ->create function, that possibly is in a loadable
1189 * module, so we have to bump that loadable module refcnt first.
1192 if (!try_module_get(net_families[family]->owner))
1195 if ((i = net_families[family]->create(sock, protocol)) < 0)
1196 goto out_module_put;
1198 * Now to bump the refcnt of the [loadable] module that owns this
1199 * socket at sock_release time we decrement its refcnt.
1201 if (!try_module_get(sock->ops->owner)) {
1203 goto out_module_put;
1206 * Now that we're done with the ->create function, the [loadable]
1207 * module can have its refcnt decremented
1209 module_put(net_families[family]->owner);
1211 security_socket_post_create(sock, family, type, protocol, kern);
1214 net_family_read_unlock();
1217 module_put(net_families[family]->owner);
1223 int sock_create(int family, int type, int protocol, struct socket **res)
1225 return __sock_create(family, type, protocol, res, 0);
1228 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1230 return __sock_create(family, type, protocol, res, 1);
1233 asmlinkage long sys_socket(int family, int type, int protocol)
1236 struct socket *sock;
1238 retval = sock_create(family, type, protocol, &sock);
1242 set_bit(SOCK_USER_SOCKET, &sock->flags);
1243 retval = sock_map_fd(sock);
1248 /* It may be already another descriptor 8) Not kernel problem. */
1257 * Create a pair of connected sockets.
1260 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1262 struct socket *sock1, *sock2;
1266 * Obtain the first socket and check if the underlying protocol
1267 * supports the socketpair call.
1270 err = sock_create(family, type, protocol, &sock1);
1273 set_bit(SOCK_USER_SOCKET, &sock1->flags);
1275 err = sock_create(family, type, protocol, &sock2);
1278 set_bit(SOCK_USER_SOCKET, &sock2->flags);
1280 err = sock1->ops->socketpair(sock1, sock2);
1282 goto out_release_both;
1286 err = sock_map_fd(sock1);
1288 goto out_release_both;
1291 err = sock_map_fd(sock2);
1296 /* fd1 and fd2 may be already another descriptors.
1297 * Not kernel problem.
1300 err = put_user(fd1, &usockvec[0]);
1302 err = put_user(fd2, &usockvec[1]);
1311 sock_release(sock2);
1316 sock_release(sock2);
1318 sock_release(sock1);
1325 * Bind a name to a socket. Nothing much to do here since it's
1326 * the protocol's responsibility to handle the local address.
1328 * We move the socket address to kernel space before we call
1329 * the protocol layer (having also checked the address is ok).
1332 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1334 struct socket *sock;
1335 char address[MAX_SOCK_ADDR];
1338 if((sock = sockfd_lookup(fd,&err))!=NULL)
1340 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1341 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1346 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1355 * Perform a listen. Basically, we allow the protocol to do anything
1356 * necessary for a listen, and if that works, we mark the socket as
1357 * ready for listening.
1360 int sysctl_somaxconn = SOMAXCONN;
1362 asmlinkage long sys_listen(int fd, int backlog)
1364 struct socket *sock;
1367 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1368 if ((unsigned) backlog > sysctl_somaxconn)
1369 backlog = sysctl_somaxconn;
1371 err = security_socket_listen(sock, backlog);
1377 err=sock->ops->listen(sock, backlog);
1385 * For accept, we attempt to create a new socket, set up the link
1386 * with the client, wake up the client, then return the new
1387 * connected fd. We collect the address of the connector in kernel
1388 * space and move it to user at the very end. This is unclean because
1389 * we open the socket then return an error.
1391 * 1003.1g adds the ability to recvmsg() to query connection pending
1392 * status to recvmsg. We need to add that support in a way thats
1393 * clean when we restucture accept also.
1396 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1398 struct socket *sock, *newsock;
1400 char address[MAX_SOCK_ADDR];
1402 sock = sockfd_lookup(fd, &err);
1407 if (!(newsock = sock_alloc()))
1410 newsock->type = sock->type;
1411 newsock->ops = sock->ops;
1413 err = security_socket_accept(sock, newsock);
1418 * We don't need try_module_get here, as the listening socket (sock)
1419 * has the protocol module (sock->ops->owner) held.
1421 __module_get(newsock->ops->owner);
1423 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1427 if (upeer_sockaddr) {
1428 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1429 err = -ECONNABORTED;
1432 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1437 /* File flags are not inherited via accept() unlike another OSes. */
1439 if ((err = sock_map_fd(newsock)) < 0)
1442 security_socket_post_accept(sock, newsock);
1449 sock_release(newsock);
1455 * Attempt to connect to a socket with the server address. The address
1456 * is in user space so we verify it is OK and move it to kernel space.
1458 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1461 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1462 * other SEQPACKET protocols that take time to connect() as it doesn't
1463 * include the -EINPROGRESS status for such sockets.
1466 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1468 struct socket *sock;
1469 char address[MAX_SOCK_ADDR];
1472 sock = sockfd_lookup(fd, &err);
1475 err = move_addr_to_kernel(uservaddr, addrlen, address);
1479 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1483 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1484 sock->file->f_flags);
1492 * Get the local address ('name') of a socket object. Move the obtained
1493 * name to user space.
1496 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1498 struct socket *sock;
1499 char address[MAX_SOCK_ADDR];
1502 sock = sockfd_lookup(fd, &err);
1506 err = security_socket_getsockname(sock);
1510 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1513 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1522 * Get the remote address ('name') of a socket object. Move the obtained
1523 * name to user space.
1526 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1528 struct socket *sock;
1529 char address[MAX_SOCK_ADDR];
1532 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1534 err = security_socket_getpeername(sock);
1540 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1542 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1549 * Send a datagram to a given address. We move the address into kernel
1550 * space and check the user space data area is readable before invoking
1554 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1555 struct sockaddr __user *addr, int addr_len)
1557 struct socket *sock;
1558 char address[MAX_SOCK_ADDR];
1563 sock = sockfd_lookup(fd, &err);
1571 msg.msg_control=NULL;
1572 msg.msg_controllen=0;
1576 err = move_addr_to_kernel(addr, addr_len, address);
1579 msg.msg_name=address;
1580 msg.msg_namelen=addr_len;
1582 if (sock->file->f_flags & O_NONBLOCK)
1583 flags |= MSG_DONTWAIT;
1584 msg.msg_flags = flags;
1585 err = sock_sendmsg(sock, &msg, len);
1594 * Send a datagram down a socket.
1597 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1599 return sys_sendto(fd, buff, len, flags, NULL, 0);
1603 * Receive a frame from the socket and optionally record the address of the
1604 * sender. We verify the buffers are writable and if needed move the
1605 * sender address from kernel to user space.
1608 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1609 struct sockaddr __user *addr, int __user *addr_len)
1611 struct socket *sock;
1614 char address[MAX_SOCK_ADDR];
1617 sock = sockfd_lookup(fd, &err);
1621 msg.msg_control=NULL;
1622 msg.msg_controllen=0;
1627 msg.msg_name=address;
1628 msg.msg_namelen=MAX_SOCK_ADDR;
1629 if (sock->file->f_flags & O_NONBLOCK)
1630 flags |= MSG_DONTWAIT;
1631 err=sock_recvmsg(sock, &msg, size, flags);
1633 if(err >= 0 && addr != NULL)
1635 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1645 * Receive a datagram from a socket.
1648 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1650 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1654 * Set a socket option. Because we don't know the option lengths we have
1655 * to pass the user mode parameter for the protocols to sort out.
1658 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1661 struct socket *sock;
1666 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1668 err = security_socket_setsockopt(sock,level,optname);
1674 if (level == SOL_SOCKET)
1675 err=sock_setsockopt(sock,level,optname,optval,optlen);
1677 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1684 * Get a socket option. Because we don't know the option lengths we have
1685 * to pass a user mode parameter for the protocols to sort out.
1688 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1691 struct socket *sock;
1693 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1695 err = security_socket_getsockopt(sock, level,
1702 if (level == SOL_SOCKET)
1703 err=sock_getsockopt(sock,level,optname,optval,optlen);
1705 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1713 * Shutdown a socket.
1716 asmlinkage long sys_shutdown(int fd, int how)
1719 struct socket *sock;
1721 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1723 err = security_socket_shutdown(sock, how);
1729 err=sock->ops->shutdown(sock, how);
1735 /* A couple of helpful macros for getting the address of the 32/64 bit
1736 * fields which are the same type (int / unsigned) on our platforms.
1738 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1739 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1740 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1744 * BSD sendmsg interface
1747 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1749 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1750 struct socket *sock;
1751 char address[MAX_SOCK_ADDR];
1752 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1753 unsigned char ctl[sizeof(struct cmsghdr) + 20]; /* 20 is size of ipv6_pktinfo */
1754 unsigned char *ctl_buf = ctl;
1755 struct msghdr msg_sys;
1756 int err, ctl_len, iov_size, total_len;
1759 if (MSG_CMSG_COMPAT & flags) {
1760 if (get_compat_msghdr(&msg_sys, msg_compat))
1762 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1765 sock = sockfd_lookup(fd, &err);
1769 /* do not move before msg_sys is valid */
1771 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1774 /* Check whether to allocate the iovec area*/
1776 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1777 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1778 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1783 /* This will also move the address data into kernel space */
1784 if (MSG_CMSG_COMPAT & flags) {
1785 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1787 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1794 if (msg_sys.msg_controllen > INT_MAX)
1796 ctl_len = msg_sys.msg_controllen;
1797 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1798 err = cmsghdr_from_user_compat_to_kern(&msg_sys, ctl, sizeof(ctl));
1801 ctl_buf = msg_sys.msg_control;
1802 } else if (ctl_len) {
1803 if (ctl_len > sizeof(ctl))
1805 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1806 if (ctl_buf == NULL)
1811 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1812 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1813 * checking falls down on this.
1815 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1817 msg_sys.msg_control = ctl_buf;
1819 msg_sys.msg_flags = flags;
1821 if (sock->file->f_flags & O_NONBLOCK)
1822 msg_sys.msg_flags |= MSG_DONTWAIT;
1823 err = sock_sendmsg(sock, &msg_sys, total_len);
1827 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1829 if (iov != iovstack)
1830 sock_kfree_s(sock->sk, iov, iov_size);
1838 * BSD recvmsg interface
1841 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1843 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1844 struct socket *sock;
1845 struct iovec iovstack[UIO_FASTIOV];
1846 struct iovec *iov=iovstack;
1847 struct msghdr msg_sys;
1848 unsigned long cmsg_ptr;
1849 int err, iov_size, total_len, len;
1851 /* kernel mode address */
1852 char addr[MAX_SOCK_ADDR];
1854 /* user mode address pointers */
1855 struct sockaddr __user *uaddr;
1856 int __user *uaddr_len;
1858 if (MSG_CMSG_COMPAT & flags) {
1859 if (get_compat_msghdr(&msg_sys, msg_compat))
1862 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1865 sock = sockfd_lookup(fd, &err);
1870 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1873 /* Check whether to allocate the iovec area*/
1875 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1876 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1877 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1883 * Save the user-mode address (verify_iovec will change the
1884 * kernel msghdr to use the kernel address space)
1887 uaddr = (void __user *) msg_sys.msg_name;
1888 uaddr_len = COMPAT_NAMELEN(msg);
1889 if (MSG_CMSG_COMPAT & flags) {
1890 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1892 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1897 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1898 msg_sys.msg_flags = 0;
1899 if (MSG_CMSG_COMPAT & flags)
1900 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1902 if (sock->file->f_flags & O_NONBLOCK)
1903 flags |= MSG_DONTWAIT;
1904 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1909 if (uaddr != NULL) {
1910 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1914 err = __put_user(msg_sys.msg_flags, COMPAT_FLAGS(msg));
1917 if (MSG_CMSG_COMPAT & flags)
1918 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1919 &msg_compat->msg_controllen);
1921 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1922 &msg->msg_controllen);
1928 if (iov != iovstack)
1929 sock_kfree_s(sock->sk, iov, iov_size);
1936 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1938 /* Argument list sizes for sys_socketcall */
1939 #define AL(x) ((x) * sizeof(unsigned long))
1940 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1941 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1942 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1946 * System call vectors.
1948 * Argument checking cleaned up. Saved 20% in size.
1949 * This function doesn't need to set the kernel lock because
1950 * it is set by the callees.
1953 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1956 unsigned long a0,a1;
1959 if(call<1||call>SYS_RECVMSG)
1962 /* copy_from_user should be SMP safe. */
1963 if (copy_from_user(a, args, nargs[call]))
1972 err = sys_socket(a0,a1,a[2]);
1975 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1978 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1981 err = sys_listen(a0,a1);
1984 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1986 case SYS_GETSOCKNAME:
1987 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1989 case SYS_GETPEERNAME:
1990 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1992 case SYS_SOCKETPAIR:
1993 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1996 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1999 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2000 (struct sockaddr __user *)a[4], a[5]);
2003 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2006 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2007 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2010 err = sys_shutdown(a0,a1);
2012 case SYS_SETSOCKOPT:
2013 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2015 case SYS_GETSOCKOPT:
2016 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2019 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2022 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2031 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2034 * This function is called by a protocol handler that wants to
2035 * advertise its address family, and have it linked into the
2039 int sock_register(struct net_proto_family *ops)
2043 if (ops->family >= NPROTO) {
2044 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2047 net_family_write_lock();
2049 if (net_families[ops->family] == NULL) {
2050 net_families[ops->family]=ops;
2053 net_family_write_unlock();
2054 printk(KERN_INFO "NET: Registered protocol family %d\n",
2060 * This function is called by a protocol handler that wants to
2061 * remove its address family, and have it unlinked from the
2065 int sock_unregister(int family)
2067 if (family < 0 || family >= NPROTO)
2070 net_family_write_lock();
2071 net_families[family]=NULL;
2072 net_family_write_unlock();
2073 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2079 extern void sk_init(void);
2081 void __init sock_init(void)
2086 * Initialize all address (protocol) families.
2089 for (i = 0; i < NPROTO; i++)
2090 net_families[i] = NULL;
2093 * Initialize sock SLAB cache.
2100 * Initialize skbuff SLAB cache
2106 * Initialize the protocols module.
2110 register_filesystem(&sock_fs_type);
2111 sock_mnt = kern_mount(&sock_fs_type);
2112 /* The real protocol initialization is performed when
2113 * do_initcalls is run.
2116 #ifdef CONFIG_NETFILTER
2124 struct module *tux_module = NULL;
2126 #ifdef CONFIG_TUX_MODULE
2128 asmlinkage long (*sys_tux_ptr) (unsigned int action, user_req_t *u_info) = NULL;
2129 spinlock_t tux_module_lock = SPIN_LOCK_UNLOCKED;
2131 asmlinkage long sys_tux (unsigned int action, user_req_t *u_info)
2135 if (current->tux_info)
2136 return sys_tux_ptr(action, u_info);
2139 spin_lock(&tux_module_lock);
2142 if (!try_module_get(tux_module))
2144 spin_unlock(&tux_module_lock);
2148 ret = sys_tux_ptr(action, u_info);
2150 spin_lock(&tux_module_lock);
2151 module_put(tux_module);
2153 spin_unlock(&tux_module_lock);
2158 EXPORT_SYMBOL_GPL(tux_module);
2159 EXPORT_SYMBOL_GPL(tux_module_lock);
2160 EXPORT_SYMBOL_GPL(sys_tux_ptr);
2162 EXPORT_SYMBOL_GPL(tux_Dprintk);
2163 EXPORT_SYMBOL_GPL(tux_TDprintk);
2166 #ifdef CONFIG_PROC_FS
2167 void socket_seq_show(struct seq_file *seq)
2172 for (cpu = 0; cpu < NR_CPUS; cpu++)
2173 counter += per_cpu(sockets_in_use, cpu);
2175 /* It can be negative, by the way. 8) */
2179 seq_printf(seq, "sockets: used %d\n", counter);
2181 #endif /* CONFIG_PROC_FS */
2183 /* ABI emulation layers need these two */
2184 EXPORT_SYMBOL(move_addr_to_kernel);
2185 EXPORT_SYMBOL(move_addr_to_user);
2186 EXPORT_SYMBOL(sock_create);
2187 EXPORT_SYMBOL(sock_create_kern);
2188 EXPORT_SYMBOL(sock_create_lite);
2189 EXPORT_SYMBOL(sock_map_fd);
2190 EXPORT_SYMBOL(sock_recvmsg);
2191 EXPORT_SYMBOL(sock_register);
2192 EXPORT_SYMBOL(sock_release);
2193 EXPORT_SYMBOL(sock_sendmsg);
2194 EXPORT_SYMBOL(sock_unregister);
2195 EXPORT_SYMBOL(sock_wake_async);
2196 EXPORT_SYMBOL(sockfd_lookup);
2197 EXPORT_SYMBOL(kernel_sendmsg);
2198 EXPORT_SYMBOL(kernel_recvmsg);