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_socket.h>
99 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
100 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
101 size_t size, loff_t pos);
102 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
103 size_t size, loff_t pos);
104 static int sock_mmap(struct file *file, struct vm_area_struct * vma);
106 static int sock_close(struct inode *inode, struct file *file);
107 static unsigned int sock_poll(struct file *file,
108 struct poll_table_struct *wait);
109 static int sock_ioctl(struct inode *inode, struct file *file,
110 unsigned int cmd, unsigned long arg);
111 static int sock_fasync(int fd, struct file *filp, int on);
112 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
113 unsigned long count, loff_t *ppos);
114 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
115 unsigned long count, loff_t *ppos);
116 static ssize_t sock_sendpage(struct file *file, struct page *page,
117 int offset, size_t size, loff_t *ppos, int more);
121 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
122 * in the operation structures but are done directly via the socketcall() multiplexor.
125 struct file_operations socket_file_ops = {
126 .owner = THIS_MODULE,
128 .aio_read = sock_aio_read,
129 .aio_write = sock_aio_write,
133 .open = sock_no_open, /* special open code to disallow open via /proc */
134 .release = sock_close,
135 .fasync = sock_fasync,
137 .writev = sock_writev,
138 .sendpage = sock_sendpage
142 * The protocol list. Each protocol is registered in here.
145 static struct net_proto_family *net_families[NPROTO];
147 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
148 static atomic_t net_family_lockct = ATOMIC_INIT(0);
149 static spinlock_t net_family_lock = SPIN_LOCK_UNLOCKED;
151 /* The strategy is: modifications net_family vector are short, do not
152 sleep and veeery rare, but read access should be free of any exclusive
156 static void net_family_write_lock(void)
158 spin_lock(&net_family_lock);
159 while (atomic_read(&net_family_lockct) != 0) {
160 spin_unlock(&net_family_lock);
164 spin_lock(&net_family_lock);
168 static __inline__ void net_family_write_unlock(void)
170 spin_unlock(&net_family_lock);
173 static __inline__ void net_family_read_lock(void)
175 atomic_inc(&net_family_lockct);
176 spin_unlock_wait(&net_family_lock);
179 static __inline__ void net_family_read_unlock(void)
181 atomic_dec(&net_family_lockct);
185 #define net_family_write_lock() do { } while(0)
186 #define net_family_write_unlock() do { } while(0)
187 #define net_family_read_lock() do { } while(0)
188 #define net_family_read_unlock() do { } while(0)
193 * Statistics counters of the socket lists
196 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
199 * Support routines. Move socket addresses back and forth across the kernel/user
200 * divide and look after the messy bits.
203 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
204 16 for IP, 16 for IPX,
207 must be at least one bigger than
208 the AF_UNIX size (see net/unix/af_unix.c
213 * move_addr_to_kernel - copy a socket address into kernel space
214 * @uaddr: Address in user space
215 * @kaddr: Address in kernel space
216 * @ulen: Length in user space
218 * The address is copied into kernel space. If the provided address is
219 * too long an error code of -EINVAL is returned. If the copy gives
220 * invalid addresses -EFAULT is returned. On a success 0 is returned.
223 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
225 if(ulen<0||ulen>MAX_SOCK_ADDR)
229 if(copy_from_user(kaddr,uaddr,ulen))
235 * move_addr_to_user - copy an address to user space
236 * @kaddr: kernel space address
237 * @klen: length of address in kernel
238 * @uaddr: user space address
239 * @ulen: pointer to user length field
241 * The value pointed to by ulen on entry is the buffer length available.
242 * This is overwritten with the buffer space used. -EINVAL is returned
243 * if an overlong buffer is specified or a negative buffer size. -EFAULT
244 * is returned if either the buffer or the length field are not
246 * After copying the data up to the limit the user specifies, the true
247 * length of the data is written over the length limit the user
248 * specified. Zero is returned for a success.
251 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
256 if((err=get_user(len, ulen)))
260 if(len<0 || len> MAX_SOCK_ADDR)
264 if(copy_to_user(uaddr,kaddr,len))
268 * "fromlen shall refer to the value before truncation.."
271 return __put_user(klen, ulen);
274 #define SOCKFS_MAGIC 0x534F434B
276 static kmem_cache_t * sock_inode_cachep;
278 static struct inode *sock_alloc_inode(struct super_block *sb)
280 struct socket_alloc *ei;
281 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
284 init_waitqueue_head(&ei->socket.wait);
286 ei->socket.fasync_list = NULL;
287 ei->socket.state = SS_UNCONNECTED;
288 ei->socket.flags = 0;
289 ei->socket.ops = NULL;
290 ei->socket.sk = NULL;
291 ei->socket.file = NULL;
292 ei->socket.flags = 0;
294 return &ei->vfs_inode;
297 static void sock_destroy_inode(struct inode *inode)
299 kmem_cache_free(sock_inode_cachep,
300 container_of(inode, struct socket_alloc, vfs_inode));
303 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
305 struct socket_alloc *ei = (struct socket_alloc *) foo;
307 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
308 SLAB_CTOR_CONSTRUCTOR)
309 inode_init_once(&ei->vfs_inode);
312 static int init_inodecache(void)
314 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
315 sizeof(struct socket_alloc),
316 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
318 if (sock_inode_cachep == NULL)
323 static struct super_operations sockfs_ops = {
324 .alloc_inode = sock_alloc_inode,
325 .destroy_inode =sock_destroy_inode,
326 .statfs = simple_statfs,
329 static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
330 int flags, const char *dev_name, void *data)
332 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
335 static struct vfsmount *sock_mnt;
337 static struct file_system_type sock_fs_type = {
339 .get_sb = sockfs_get_sb,
340 .kill_sb = kill_anon_super,
342 static int sockfs_delete_dentry(struct dentry *dentry)
346 static struct dentry_operations sockfs_dentry_operations = {
347 .d_delete = sockfs_delete_dentry,
351 * Obtains the first available file descriptor and sets it up for use.
353 * This function creates file structure and maps it to fd space
354 * of current process. On success it returns file descriptor
355 * and file struct implicitly stored in sock->file.
356 * Note that another thread may close file descriptor before we return
357 * from this function. We use the fact that now we do not refer
358 * to socket after mapping. If one day we will need it, this
359 * function will increment ref. count on file by 1.
361 * In any case returned fd MAY BE not valid!
362 * This race condition is unavoidable
363 * with shared fd spaces, we cannot solve it inside kernel,
364 * but we take care of internal coherence yet.
367 struct file * sock_map_file(struct socket *sock)
373 file = get_empty_filp();
376 return ERR_PTR(-ENFILE);
378 sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
380 this.len = strlen(name);
381 this.hash = SOCK_INODE(sock)->i_ino;
383 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
384 if (!file->f_dentry) {
386 return ERR_PTR(-ENOMEM);
388 file->f_dentry->d_op = &sockfs_dentry_operations;
389 d_add(file->f_dentry, SOCK_INODE(sock));
390 file->f_vfsmnt = mntget(sock_mnt);
391 file->f_mapping = file->f_dentry->d_inode->i_mapping;
396 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
397 file->f_mode = FMODE_READ | FMODE_WRITE;
398 file->f_flags = O_RDWR;
404 int sock_map_fd(struct socket *sock)
410 * Find a file descriptor suitable for return to the user.
413 fd = get_unused_fd();
417 file = sock_map_file(sock);
420 return PTR_ERR(file);
422 fd_install(fd, file);
428 * sockfd_lookup - Go from a file number to its socket slot
430 * @err: pointer to an error code return
432 * The file handle passed in is locked and the socket it is bound
433 * too is returned. If an error occurs the err pointer is overwritten
434 * with a negative errno code and NULL is returned. The function checks
435 * for both invalid handles and passing a handle which is not a socket.
437 * On a success the socket object pointer is returned.
440 struct socket *sockfd_lookup(int fd, int *err)
446 if (!(file = fget(fd)))
452 inode = file->f_dentry->d_inode;
453 if (!inode->i_sock || !(sock = SOCKET_I(inode)))
460 if (sock->file != file) {
461 printk(KERN_ERR "socki_lookup: socket file changed!\n");
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 struct socket *sock_alloc(void)
477 struct inode * inode;
478 struct socket * sock;
480 inode = new_inode(sock_mnt->mnt_sb);
484 sock = SOCKET_I(inode);
486 inode->i_mode = S_IFSOCK|S_IRWXUGO;
488 inode->i_uid = current->fsuid;
489 inode->i_gid = current->fsgid;
491 get_cpu_var(sockets_in_use)++;
492 put_cpu_var(sockets_in_use);
496 EXPORT_SYMBOL_GPL(sock_alloc);
499 * In theory you can't get an open on this inode, but /proc provides
500 * a back door. Remember to keep it shut otherwise you'll let the
501 * creepy crawlies in.
504 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
509 struct file_operations bad_sock_fops = {
510 .owner = THIS_MODULE,
511 .open = sock_no_open,
515 * sock_release - close a socket
516 * @sock: socket to close
518 * The socket is released from the protocol stack if it has a release
519 * callback, and the inode is then released if the socket is bound to
520 * an inode not a file.
523 void sock_release(struct socket *sock)
526 struct module *owner = sock->ops->owner;
528 sock->ops->release(sock);
533 if (sock->fasync_list)
534 printk(KERN_ERR "sock_release: fasync list not empty!\n");
536 get_cpu_var(sockets_in_use)--;
537 put_cpu_var(sockets_in_use);
539 iput(SOCK_INODE(sock));
545 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
546 struct msghdr *msg, size_t size)
548 struct sock_iocb *si = kiocb_to_siocb(iocb);
556 err = security_socket_sendmsg(sock, msg, size);
560 len = sock->ops->sendmsg(iocb, sock, msg, size);
563 vx_sock_send(sock->sk, size);
565 vx_sock_fail(sock->sk, size);
567 vxdprintk(VXD_CBIT(net, 7),
568 "__sock_sendmsg: %p[%p,%p,%p;%d]:%d/%d",
570 (sock->sk)?sock->sk->sk_nx_info:0,
571 (sock->sk)?sock->sk->sk_vx_info:0,
572 (sock->sk)?sock->sk->sk_xid:0,
573 (unsigned int)size, len);
577 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
580 struct sock_iocb siocb;
583 init_sync_kiocb(&iocb, NULL);
584 iocb.private = &siocb;
585 ret = __sock_sendmsg(&iocb, sock, msg, size);
586 if (-EIOCBQUEUED == ret)
587 ret = wait_on_sync_kiocb(&iocb);
591 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
592 struct kvec *vec, size_t num, size_t size)
594 mm_segment_t oldfs = get_fs();
599 * the following is safe, since for compiler definitions of kvec and
600 * iovec are identical, yielding the same in-core layout and alignment
602 msg->msg_iov = (struct iovec *)vec,
603 msg->msg_iovlen = num;
604 result = sock_sendmsg(sock, msg, size);
609 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
610 struct msghdr *msg, size_t size, int flags)
613 struct sock_iocb *si = kiocb_to_siocb(iocb);
621 err = security_socket_recvmsg(sock, msg, size, flags);
625 len = sock->ops->recvmsg(iocb, sock, msg, size, flags);
626 if ((len >= 0) && sock->sk)
627 vx_sock_recv(sock->sk, len);
628 vxdprintk(VXD_CBIT(net, 7),
629 "__sock_recvmsg: %p[%p,%p,%p;%d]:%d/%d",
631 (sock->sk)?sock->sk->sk_nx_info:0,
632 (sock->sk)?sock->sk->sk_vx_info:0,
633 (sock->sk)?sock->sk->sk_xid:0,
634 (unsigned int)size, len);
638 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
639 size_t size, int flags)
642 struct sock_iocb siocb;
645 init_sync_kiocb(&iocb, NULL);
646 iocb.private = &siocb;
647 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
648 if (-EIOCBQUEUED == ret)
649 ret = wait_on_sync_kiocb(&iocb);
653 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
654 struct kvec *vec, size_t num,
655 size_t size, int flags)
657 mm_segment_t oldfs = get_fs();
662 * the following is safe, since for compiler definitions of kvec and
663 * iovec are identical, yielding the same in-core layout and alignment
665 msg->msg_iov = (struct iovec *)vec,
666 msg->msg_iovlen = num;
667 result = sock_recvmsg(sock, msg, size, flags);
672 static void sock_aio_dtor(struct kiocb *iocb)
674 kfree(iocb->private);
678 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
679 * area ubuf...ubuf+size-1 is writable before asking the protocol.
682 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
683 size_t size, loff_t pos)
685 struct sock_iocb *x, siocb;
691 if (size==0) /* Match SYS5 behaviour */
694 if (is_sync_kiocb(iocb))
697 x = kmalloc(sizeof(struct sock_iocb), GFP_KERNEL);
700 iocb->ki_dtor = sock_aio_dtor;
704 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
706 x->async_msg.msg_name = NULL;
707 x->async_msg.msg_namelen = 0;
708 x->async_msg.msg_iov = &x->async_iov;
709 x->async_msg.msg_iovlen = 1;
710 x->async_msg.msg_control = NULL;
711 x->async_msg.msg_controllen = 0;
712 x->async_iov.iov_base = ubuf;
713 x->async_iov.iov_len = size;
714 flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
716 return __sock_recvmsg(iocb, sock, &x->async_msg, size, flags);
721 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
722 * is readable by the user process.
725 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
726 size_t size, loff_t pos)
728 struct sock_iocb *x, siocb;
733 if(size==0) /* Match SYS5 behaviour */
736 if (is_sync_kiocb(iocb))
739 x = kmalloc(sizeof(struct sock_iocb), GFP_KERNEL);
742 iocb->ki_dtor = sock_aio_dtor;
746 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
748 x->async_msg.msg_name = NULL;
749 x->async_msg.msg_namelen = 0;
750 x->async_msg.msg_iov = &x->async_iov;
751 x->async_msg.msg_iovlen = 1;
752 x->async_msg.msg_control = NULL;
753 x->async_msg.msg_controllen = 0;
754 x->async_msg.msg_flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
755 if (sock->type == SOCK_SEQPACKET)
756 x->async_msg.msg_flags |= MSG_EOR;
757 x->async_iov.iov_base = (void __user *)ubuf;
758 x->async_iov.iov_len = size;
760 return __sock_sendmsg(iocb, sock, &x->async_msg, size);
763 ssize_t sock_sendpage(struct file *file, struct page *page,
764 int offset, size_t size, loff_t *ppos, int more)
769 sock = SOCKET_I(file->f_dentry->d_inode);
771 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
775 return sock->ops->sendpage(sock, page, offset, size, flags);
778 int sock_readv_writev(int type, struct inode * inode, struct file * file,
779 const struct iovec * iov, long count, size_t size)
784 sock = SOCKET_I(inode);
788 msg.msg_control = NULL;
789 msg.msg_controllen = 0;
790 msg.msg_iov = (struct iovec *) iov;
791 msg.msg_iovlen = count;
792 msg.msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
794 /* read() does a VERIFY_WRITE */
795 if (type == VERIFY_WRITE)
796 return sock_recvmsg(sock, &msg, size, msg.msg_flags);
798 if (sock->type == SOCK_SEQPACKET)
799 msg.msg_flags |= MSG_EOR;
801 return sock_sendmsg(sock, &msg, size);
804 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
805 unsigned long count, loff_t *ppos)
809 for (i = 0 ; i < count ; i++)
810 tot_len += vector[i].iov_len;
811 return sock_readv_writev(VERIFY_WRITE, file->f_dentry->d_inode,
812 file, vector, count, tot_len);
815 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
816 unsigned long count, loff_t *ppos)
820 for (i = 0 ; i < count ; i++)
821 tot_len += vector[i].iov_len;
822 return sock_readv_writev(VERIFY_READ, file->f_dentry->d_inode,
823 file, vector, count, tot_len);
828 * Atomic setting of ioctl hooks to avoid race
829 * with module unload.
832 static DECLARE_MUTEX(br_ioctl_mutex);
833 static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
835 void brioctl_set(int (*hook)(unsigned int, void __user *))
837 down(&br_ioctl_mutex);
838 br_ioctl_hook = hook;
841 EXPORT_SYMBOL(brioctl_set);
843 static DECLARE_MUTEX(vlan_ioctl_mutex);
844 static int (*vlan_ioctl_hook)(void __user *arg);
846 void vlan_ioctl_set(int (*hook)(void __user *))
848 down(&vlan_ioctl_mutex);
849 vlan_ioctl_hook = hook;
850 up(&vlan_ioctl_mutex);
852 EXPORT_SYMBOL(vlan_ioctl_set);
854 static DECLARE_MUTEX(dlci_ioctl_mutex);
855 static int (*dlci_ioctl_hook)(unsigned int, void __user *);
857 void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
859 down(&dlci_ioctl_mutex);
860 dlci_ioctl_hook = hook;
861 up(&dlci_ioctl_mutex);
863 EXPORT_SYMBOL(dlci_ioctl_set);
866 * With an ioctl, arg may well be a user mode pointer, but we don't know
867 * what to do with it - that's up to the protocol still.
870 static int sock_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
874 void __user *argp = (void __user *)arg;
878 sock = SOCKET_I(inode);
879 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
880 err = dev_ioctl(cmd, argp);
883 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
884 err = dev_ioctl(cmd, argp);
886 #endif /* WIRELESS_EXT */
891 if (get_user(pid, (int __user *)argp))
893 err = f_setown(sock->file, pid, 1);
897 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
905 request_module("bridge");
907 down(&br_ioctl_mutex);
909 err = br_ioctl_hook(cmd, argp);
915 if (!vlan_ioctl_hook)
916 request_module("8021q");
918 down(&vlan_ioctl_mutex);
920 err = vlan_ioctl_hook(argp);
921 up(&vlan_ioctl_mutex);
925 /* Convert this to call through a hook */
926 err = divert_ioctl(cmd, argp);
931 if (!dlci_ioctl_hook)
932 request_module("dlci");
934 if (dlci_ioctl_hook) {
935 down(&dlci_ioctl_mutex);
936 err = dlci_ioctl_hook(cmd, argp);
937 up(&dlci_ioctl_mutex);
941 err = sock->ops->ioctl(sock, cmd, arg);
949 int sock_create_lite(int family, int type, int protocol, struct socket **res)
952 struct socket *sock = NULL;
954 err = security_socket_create(family, type, protocol, 1);
964 security_socket_post_create(sock, family, type, protocol, 1);
971 /* No kernel lock held - perfect */
972 static unsigned int sock_poll(struct file *file, poll_table * wait)
977 * We can't return errors to poll, so it's either yes or no.
979 sock = SOCKET_I(file->f_dentry->d_inode);
980 return sock->ops->poll(file, sock, wait);
983 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
985 struct socket *sock = SOCKET_I(file->f_dentry->d_inode);
987 return sock->ops->mmap(file, sock, vma);
990 int sock_close(struct inode *inode, struct file *filp)
993 * It was possible the inode is NULL we were
994 * closing an unfinished socket.
999 printk(KERN_DEBUG "sock_close: NULL inode\n");
1002 sock_fasync(-1, filp, 0);
1003 sock_release(SOCKET_I(inode));
1008 * Update the socket async list
1010 * Fasync_list locking strategy.
1012 * 1. fasync_list is modified only under process context socket lock
1013 * i.e. under semaphore.
1014 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1015 * or under socket lock.
1016 * 3. fasync_list can be used from softirq context, so that
1017 * modification under socket lock have to be enhanced with
1018 * write_lock_bh(&sk->sk_callback_lock).
1022 static int sock_fasync(int fd, struct file *filp, int on)
1024 struct fasync_struct *fa, *fna=NULL, **prev;
1025 struct socket *sock;
1030 fna=(struct fasync_struct *)kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1035 sock = SOCKET_I(filp->f_dentry->d_inode);
1037 if ((sk=sock->sk) == NULL) {
1045 prev=&(sock->fasync_list);
1047 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1048 if (fa->fa_file==filp)
1055 write_lock_bh(&sk->sk_callback_lock);
1057 write_unlock_bh(&sk->sk_callback_lock);
1064 fna->magic=FASYNC_MAGIC;
1065 fna->fa_next=sock->fasync_list;
1066 write_lock_bh(&sk->sk_callback_lock);
1067 sock->fasync_list=fna;
1068 write_unlock_bh(&sk->sk_callback_lock);
1074 write_lock_bh(&sk->sk_callback_lock);
1076 write_unlock_bh(&sk->sk_callback_lock);
1084 release_sock(sock->sk);
1088 /* This function may be called only under socket lock or callback_lock */
1090 int sock_wake_async(struct socket *sock, int how, int band)
1092 if (!sock || !sock->fasync_list)
1098 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1102 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1107 __kill_fasync(sock->fasync_list, SIGIO, band);
1110 __kill_fasync(sock->fasync_list, SIGURG, band);
1115 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1119 struct socket *sock;
1122 * Check protocol is in range
1124 if (family < 0 || family >= NPROTO)
1125 return -EAFNOSUPPORT;
1126 if (type < 0 || type >= SOCK_MAX)
1129 /* disable IPv6 inside vservers for now */
1130 if (family == PF_INET6 && !vx_check(0, VX_ADMIN))
1131 return -EAFNOSUPPORT;
1135 This uglymoron is moved from INET layer to here to avoid
1136 deadlock in module load.
1138 if (family == PF_INET && type == SOCK_PACKET) {
1142 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1147 err = security_socket_create(family, type, protocol, kern);
1151 #if defined(CONFIG_KMOD)
1152 /* Attempt to load a protocol module if the find failed.
1154 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1155 * requested real, full-featured networking support upon configuration.
1156 * Otherwise module support will break!
1158 if (net_families[family]==NULL)
1160 request_module("net-pf-%d",family);
1164 net_family_read_lock();
1165 if (net_families[family] == NULL) {
1171 * Allocate the socket and allow the family to set things up. if
1172 * the protocol is 0, the family is instructed to select an appropriate
1176 if (!(sock = sock_alloc()))
1178 printk(KERN_WARNING "socket: no more sockets\n");
1179 i = -ENFILE; /* Not exactly a match, but its the
1180 closest posix thing */
1187 * We will call the ->create function, that possibly is in a loadable
1188 * module, so we have to bump that loadable module refcnt first.
1191 if (!try_module_get(net_families[family]->owner))
1194 if ((i = net_families[family]->create(sock, protocol)) < 0)
1195 goto out_module_put;
1197 * Now to bump the refcnt of the [loadable] module that owns this
1198 * socket at sock_release time we decrement its refcnt.
1200 if (!try_module_get(sock->ops->owner)) {
1202 goto out_module_put;
1205 * Now that we're done with the ->create function, the [loadable]
1206 * module can have its refcnt decremented
1208 module_put(net_families[family]->owner);
1210 security_socket_post_create(sock, family, type, protocol, kern);
1213 net_family_read_unlock();
1216 module_put(net_families[family]->owner);
1222 int sock_create(int family, int type, int protocol, struct socket **res)
1224 return __sock_create(family, type, protocol, res, 0);
1227 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1229 return __sock_create(family, type, protocol, res, 1);
1232 asmlinkage long sys_socket(int family, int type, int protocol)
1235 struct socket *sock;
1237 retval = sock_create(family, type, protocol, &sock);
1241 set_bit(SOCK_USER_SOCKET, &sock->flags);
1242 retval = sock_map_fd(sock);
1247 /* It may be already another descriptor 8) Not kernel problem. */
1256 * Create a pair of connected sockets.
1259 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1261 struct socket *sock1, *sock2;
1265 * Obtain the first socket and check if the underlying protocol
1266 * supports the socketpair call.
1269 err = sock_create(family, type, protocol, &sock1);
1272 set_bit(SOCK_USER_SOCKET, &sock1->flags);
1274 err = sock_create(family, type, protocol, &sock2);
1277 set_bit(SOCK_USER_SOCKET, &sock2->flags);
1279 err = sock1->ops->socketpair(sock1, sock2);
1281 goto out_release_both;
1285 err = sock_map_fd(sock1);
1287 goto out_release_both;
1290 err = sock_map_fd(sock2);
1295 /* fd1 and fd2 may be already another descriptors.
1296 * Not kernel problem.
1299 err = put_user(fd1, &usockvec[0]);
1301 err = put_user(fd2, &usockvec[1]);
1310 sock_release(sock2);
1315 sock_release(sock2);
1317 sock_release(sock1);
1324 * Bind a name to a socket. Nothing much to do here since it's
1325 * the protocol's responsibility to handle the local address.
1327 * We move the socket address to kernel space before we call
1328 * the protocol layer (having also checked the address is ok).
1331 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1333 struct socket *sock;
1334 char address[MAX_SOCK_ADDR];
1337 if((sock = sockfd_lookup(fd,&err))!=NULL)
1339 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1340 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1345 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1354 * Perform a listen. Basically, we allow the protocol to do anything
1355 * necessary for a listen, and if that works, we mark the socket as
1356 * ready for listening.
1359 int sysctl_somaxconn = SOMAXCONN;
1361 asmlinkage long sys_listen(int fd, int backlog)
1363 struct socket *sock;
1366 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1367 if ((unsigned) backlog > sysctl_somaxconn)
1368 backlog = sysctl_somaxconn;
1370 err = security_socket_listen(sock, backlog);
1376 err=sock->ops->listen(sock, backlog);
1384 * For accept, we attempt to create a new socket, set up the link
1385 * with the client, wake up the client, then return the new
1386 * connected fd. We collect the address of the connector in kernel
1387 * space and move it to user at the very end. This is unclean because
1388 * we open the socket then return an error.
1390 * 1003.1g adds the ability to recvmsg() to query connection pending
1391 * status to recvmsg. We need to add that support in a way thats
1392 * clean when we restucture accept also.
1395 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1397 struct socket *sock, *newsock;
1399 char address[MAX_SOCK_ADDR];
1401 sock = sockfd_lookup(fd, &err);
1406 if (!(newsock = sock_alloc()))
1409 newsock->type = sock->type;
1410 newsock->ops = sock->ops;
1412 err = security_socket_accept(sock, newsock);
1417 * We don't need try_module_get here, as the listening socket (sock)
1418 * has the protocol module (sock->ops->owner) held.
1420 __module_get(newsock->ops->owner);
1422 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1426 if (upeer_sockaddr) {
1427 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1428 err = -ECONNABORTED;
1431 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1436 /* File flags are not inherited via accept() unlike another OSes. */
1438 if ((err = sock_map_fd(newsock)) < 0)
1441 security_socket_post_accept(sock, newsock);
1448 sock_release(newsock);
1454 * Attempt to connect to a socket with the server address. The address
1455 * is in user space so we verify it is OK and move it to kernel space.
1457 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1460 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1461 * other SEQPACKET protocols that take time to connect() as it doesn't
1462 * include the -EINPROGRESS status for such sockets.
1465 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1467 struct socket *sock;
1468 char address[MAX_SOCK_ADDR];
1471 sock = sockfd_lookup(fd, &err);
1474 err = move_addr_to_kernel(uservaddr, addrlen, address);
1478 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1482 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1483 sock->file->f_flags);
1491 * Get the local address ('name') of a socket object. Move the obtained
1492 * name to user space.
1495 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1497 struct socket *sock;
1498 char address[MAX_SOCK_ADDR];
1501 sock = sockfd_lookup(fd, &err);
1505 err = security_socket_getsockname(sock);
1509 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1512 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1521 * Get the remote address ('name') of a socket object. Move the obtained
1522 * name to user space.
1525 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1527 struct socket *sock;
1528 char address[MAX_SOCK_ADDR];
1531 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1533 err = security_socket_getpeername(sock);
1539 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1541 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1548 * Send a datagram to a given address. We move the address into kernel
1549 * space and check the user space data area is readable before invoking
1553 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1554 struct sockaddr __user *addr, int addr_len)
1556 struct socket *sock;
1557 char address[MAX_SOCK_ADDR];
1562 sock = sockfd_lookup(fd, &err);
1570 msg.msg_control=NULL;
1571 msg.msg_controllen=0;
1575 err = move_addr_to_kernel(addr, addr_len, address);
1578 msg.msg_name=address;
1579 msg.msg_namelen=addr_len;
1581 if (sock->file->f_flags & O_NONBLOCK)
1582 flags |= MSG_DONTWAIT;
1583 msg.msg_flags = flags;
1584 err = sock_sendmsg(sock, &msg, len);
1593 * Send a datagram down a socket.
1596 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1598 return sys_sendto(fd, buff, len, flags, NULL, 0);
1602 * Receive a frame from the socket and optionally record the address of the
1603 * sender. We verify the buffers are writable and if needed move the
1604 * sender address from kernel to user space.
1607 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1608 struct sockaddr __user *addr, int __user *addr_len)
1610 struct socket *sock;
1613 char address[MAX_SOCK_ADDR];
1616 sock = sockfd_lookup(fd, &err);
1620 msg.msg_control=NULL;
1621 msg.msg_controllen=0;
1626 msg.msg_name=address;
1627 msg.msg_namelen=MAX_SOCK_ADDR;
1628 if (sock->file->f_flags & O_NONBLOCK)
1629 flags |= MSG_DONTWAIT;
1630 err=sock_recvmsg(sock, &msg, size, flags);
1632 if(err >= 0 && addr != NULL)
1634 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1644 * Receive a datagram from a socket.
1647 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1649 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1653 * Set a socket option. Because we don't know the option lengths we have
1654 * to pass the user mode parameter for the protocols to sort out.
1657 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1660 struct socket *sock;
1665 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1667 err = security_socket_setsockopt(sock,level,optname);
1673 if (level == SOL_SOCKET)
1674 err=sock_setsockopt(sock,level,optname,optval,optlen);
1676 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1683 * Get a socket option. Because we don't know the option lengths we have
1684 * to pass a user mode parameter for the protocols to sort out.
1687 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1690 struct socket *sock;
1692 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1694 err = security_socket_getsockopt(sock, level,
1701 if (level == SOL_SOCKET)
1702 err=sock_getsockopt(sock,level,optname,optval,optlen);
1704 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1712 * Shutdown a socket.
1715 asmlinkage long sys_shutdown(int fd, int how)
1718 struct socket *sock;
1720 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1722 err = security_socket_shutdown(sock, how);
1728 err=sock->ops->shutdown(sock, how);
1734 /* A couple of helpful macros for getting the address of the 32/64 bit
1735 * fields which are the same type (int / unsigned) on our platforms.
1737 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1738 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1739 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1743 * BSD sendmsg interface
1746 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1748 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1749 struct socket *sock;
1750 char address[MAX_SOCK_ADDR];
1751 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1752 unsigned char ctl[sizeof(struct cmsghdr) + 20]; /* 20 is size of ipv6_pktinfo */
1753 unsigned char *ctl_buf = ctl;
1754 struct msghdr msg_sys;
1755 int err, ctl_len, iov_size, total_len;
1758 if (MSG_CMSG_COMPAT & flags) {
1759 if (get_compat_msghdr(&msg_sys, msg_compat))
1761 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1764 sock = sockfd_lookup(fd, &err);
1768 /* do not move before msg_sys is valid */
1770 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1773 /* Check whether to allocate the iovec area*/
1775 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1776 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1777 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1782 /* This will also move the address data into kernel space */
1783 if (MSG_CMSG_COMPAT & flags) {
1784 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1786 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1793 if (msg_sys.msg_controllen > INT_MAX)
1795 ctl_len = msg_sys.msg_controllen;
1796 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1797 err = cmsghdr_from_user_compat_to_kern(&msg_sys, ctl, sizeof(ctl));
1800 ctl_buf = msg_sys.msg_control;
1801 } else if (ctl_len) {
1802 if (ctl_len > sizeof(ctl))
1804 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1805 if (ctl_buf == NULL)
1810 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1811 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1812 * checking falls down on this.
1814 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1816 msg_sys.msg_control = ctl_buf;
1818 msg_sys.msg_flags = flags;
1820 if (sock->file->f_flags & O_NONBLOCK)
1821 msg_sys.msg_flags |= MSG_DONTWAIT;
1822 err = sock_sendmsg(sock, &msg_sys, total_len);
1826 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1828 if (iov != iovstack)
1829 sock_kfree_s(sock->sk, iov, iov_size);
1837 * BSD recvmsg interface
1840 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1842 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1843 struct socket *sock;
1844 struct iovec iovstack[UIO_FASTIOV];
1845 struct iovec *iov=iovstack;
1846 struct msghdr msg_sys;
1847 unsigned long cmsg_ptr;
1848 int err, iov_size, total_len, len;
1850 /* kernel mode address */
1851 char addr[MAX_SOCK_ADDR];
1853 /* user mode address pointers */
1854 struct sockaddr __user *uaddr;
1855 int __user *uaddr_len;
1857 if (MSG_CMSG_COMPAT & flags) {
1858 if (get_compat_msghdr(&msg_sys, msg_compat))
1861 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1864 sock = sockfd_lookup(fd, &err);
1869 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1872 /* Check whether to allocate the iovec area*/
1874 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1875 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1876 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1882 * Save the user-mode address (verify_iovec will change the
1883 * kernel msghdr to use the kernel address space)
1886 uaddr = (void __user *) msg_sys.msg_name;
1887 uaddr_len = COMPAT_NAMELEN(msg);
1888 if (MSG_CMSG_COMPAT & flags) {
1889 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1891 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1896 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1897 msg_sys.msg_flags = 0;
1898 if (MSG_CMSG_COMPAT & flags)
1899 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1901 if (sock->file->f_flags & O_NONBLOCK)
1902 flags |= MSG_DONTWAIT;
1903 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1908 if (uaddr != NULL) {
1909 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1913 err = __put_user(msg_sys.msg_flags, COMPAT_FLAGS(msg));
1916 if (MSG_CMSG_COMPAT & flags)
1917 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1918 &msg_compat->msg_controllen);
1920 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1921 &msg->msg_controllen);
1927 if (iov != iovstack)
1928 sock_kfree_s(sock->sk, iov, iov_size);
1935 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1937 /* Argument list sizes for sys_socketcall */
1938 #define AL(x) ((x) * sizeof(unsigned long))
1939 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1940 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1941 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1945 * System call vectors.
1947 * Argument checking cleaned up. Saved 20% in size.
1948 * This function doesn't need to set the kernel lock because
1949 * it is set by the callees.
1952 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1955 unsigned long a0,a1;
1958 if(call<1||call>SYS_RECVMSG)
1961 /* copy_from_user should be SMP safe. */
1962 if (copy_from_user(a, args, nargs[call]))
1971 err = sys_socket(a0,a1,a[2]);
1974 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1977 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1980 err = sys_listen(a0,a1);
1983 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1985 case SYS_GETSOCKNAME:
1986 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1988 case SYS_GETPEERNAME:
1989 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1991 case SYS_SOCKETPAIR:
1992 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1995 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1998 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
1999 (struct sockaddr __user *)a[4], a[5]);
2002 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2005 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2006 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2009 err = sys_shutdown(a0,a1);
2011 case SYS_SETSOCKOPT:
2012 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2014 case SYS_GETSOCKOPT:
2015 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2018 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2021 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2030 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2033 * This function is called by a protocol handler that wants to
2034 * advertise its address family, and have it linked into the
2038 int sock_register(struct net_proto_family *ops)
2042 if (ops->family >= NPROTO) {
2043 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2046 net_family_write_lock();
2048 if (net_families[ops->family] == NULL) {
2049 net_families[ops->family]=ops;
2052 net_family_write_unlock();
2053 printk(KERN_INFO "NET: Registered protocol family %d\n",
2059 * This function is called by a protocol handler that wants to
2060 * remove its address family, and have it unlinked from the
2064 int sock_unregister(int family)
2066 if (family < 0 || family >= NPROTO)
2069 net_family_write_lock();
2070 net_families[family]=NULL;
2071 net_family_write_unlock();
2072 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2078 extern void sk_init(void);
2080 void __init sock_init(void)
2085 * Initialize all address (protocol) families.
2088 for (i = 0; i < NPROTO; i++)
2089 net_families[i] = NULL;
2092 * Initialize sock SLAB cache.
2099 * Initialize skbuff SLAB cache
2105 * Initialize the protocols module.
2109 register_filesystem(&sock_fs_type);
2110 sock_mnt = kern_mount(&sock_fs_type);
2111 /* The real protocol initialization is performed when
2112 * do_initcalls is run.
2115 #ifdef CONFIG_NETFILTER
2123 struct module *tux_module = NULL;
2125 #ifdef CONFIG_TUX_MODULE
2127 asmlinkage long (*sys_tux_ptr) (unsigned int action, user_req_t *u_info) = NULL;
2128 spinlock_t tux_module_lock = SPIN_LOCK_UNLOCKED;
2130 asmlinkage long sys_tux (unsigned int action, user_req_t *u_info)
2134 if (current->tux_info)
2135 return sys_tux_ptr(action, u_info);
2138 spin_lock(&tux_module_lock);
2141 if (!try_module_get(tux_module))
2143 spin_unlock(&tux_module_lock);
2147 ret = sys_tux_ptr(action, u_info);
2149 spin_lock(&tux_module_lock);
2150 module_put(tux_module);
2152 spin_unlock(&tux_module_lock);
2157 EXPORT_SYMBOL_GPL(tux_module);
2158 EXPORT_SYMBOL_GPL(tux_module_lock);
2159 EXPORT_SYMBOL_GPL(sys_tux_ptr);
2161 EXPORT_SYMBOL_GPL(tux_Dprintk);
2162 EXPORT_SYMBOL_GPL(tux_TDprintk);
2165 #ifdef CONFIG_PROC_FS
2166 void socket_seq_show(struct seq_file *seq)
2171 for (cpu = 0; cpu < NR_CPUS; cpu++)
2172 counter += per_cpu(sockets_in_use, cpu);
2174 /* It can be negative, by the way. 8) */
2178 seq_printf(seq, "sockets: used %d\n", counter);
2180 #endif /* CONFIG_PROC_FS */
2182 /* ABI emulation layers need these two */
2183 EXPORT_SYMBOL(move_addr_to_kernel);
2184 EXPORT_SYMBOL(move_addr_to_user);
2185 EXPORT_SYMBOL(sock_create);
2186 EXPORT_SYMBOL(sock_create_kern);
2187 EXPORT_SYMBOL(sock_create_lite);
2188 EXPORT_SYMBOL(sock_map_fd);
2189 EXPORT_SYMBOL(sock_recvmsg);
2190 EXPORT_SYMBOL(sock_register);
2191 EXPORT_SYMBOL(sock_release);
2192 EXPORT_SYMBOL(sock_sendmsg);
2193 EXPORT_SYMBOL(sock_unregister);
2194 EXPORT_SYMBOL(sock_wake_async);
2195 EXPORT_SYMBOL(sockfd_lookup);
2196 EXPORT_SYMBOL(kernel_sendmsg);
2197 EXPORT_SYMBOL(kernel_recvmsg);