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>
71 #include <linux/wanrouter.h>
72 #include <linux/if_bridge.h>
73 #include <linux/init.h>
74 #include <linux/poll.h>
75 #include <linux/cache.h>
76 #include <linux/module.h>
77 #include <linux/highmem.h>
78 #include <linux/divert.h>
79 #include <linux/mount.h>
80 #include <linux/security.h>
81 #include <linux/syscalls.h>
82 #include <linux/compat.h>
83 #include <linux/kmod.h>
85 #ifdef CONFIG_NET_RADIO
86 #include <linux/wireless.h> /* Note : will define WIRELESS_EXT */
87 #endif /* CONFIG_NET_RADIO */
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
95 #include <linux/netfilter.h>
96 #include <linux/vs_socket.h>
98 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
99 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
100 size_t size, loff_t pos);
101 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
102 size_t size, loff_t pos);
103 static int sock_mmap(struct file *file, struct vm_area_struct * vma);
105 static int sock_close(struct inode *inode, struct file *file);
106 static unsigned int sock_poll(struct file *file,
107 struct poll_table_struct *wait);
108 static long sock_ioctl(struct file *file,
109 unsigned int cmd, unsigned long arg);
110 static int sock_fasync(int fd, struct file *filp, int on);
111 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
112 unsigned long count, loff_t *ppos);
113 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
114 unsigned long count, loff_t *ppos);
115 static ssize_t sock_sendpage(struct file *file, struct page *page,
116 int offset, size_t size, loff_t *ppos, int more);
120 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
121 * in the operation structures but are done directly via the socketcall() multiplexor.
124 static struct file_operations socket_file_ops = {
125 .owner = THIS_MODULE,
127 .aio_read = sock_aio_read,
128 .aio_write = sock_aio_write,
130 .unlocked_ioctl = sock_ioctl,
132 .open = sock_no_open, /* special open code to disallow open via /proc */
133 .release = sock_close,
134 .fasync = sock_fasync,
136 .writev = sock_writev,
137 .sendpage = sock_sendpage
141 * The protocol list. Each protocol is registered in here.
144 static struct net_proto_family *net_families[NPROTO];
146 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
147 static atomic_t net_family_lockct = ATOMIC_INIT(0);
148 static DEFINE_SPINLOCK(net_family_lock);
150 /* The strategy is: modifications net_family vector are short, do not
151 sleep and veeery rare, but read access should be free of any exclusive
155 static void net_family_write_lock(void)
157 spin_lock(&net_family_lock);
158 while (atomic_read(&net_family_lockct) != 0) {
159 spin_unlock(&net_family_lock);
163 spin_lock(&net_family_lock);
167 static __inline__ void net_family_write_unlock(void)
169 spin_unlock(&net_family_lock);
172 static __inline__ void net_family_read_lock(void)
174 atomic_inc(&net_family_lockct);
175 spin_unlock_wait(&net_family_lock);
178 static __inline__ void net_family_read_unlock(void)
180 atomic_dec(&net_family_lockct);
184 #define net_family_write_lock() do { } while(0)
185 #define net_family_write_unlock() do { } while(0)
186 #define net_family_read_lock() do { } while(0)
187 #define net_family_read_unlock() do { } while(0)
192 * Statistics counters of the socket lists
195 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
198 * Support routines. Move socket addresses back and forth across the kernel/user
199 * divide and look after the messy bits.
202 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
203 16 for IP, 16 for IPX,
206 must be at least one bigger than
207 the AF_UNIX size (see net/unix/af_unix.c
212 * move_addr_to_kernel - copy a socket address into kernel space
213 * @uaddr: Address in user space
214 * @kaddr: Address in kernel space
215 * @ulen: Length in user space
217 * The address is copied into kernel space. If the provided address is
218 * too long an error code of -EINVAL is returned. If the copy gives
219 * invalid addresses -EFAULT is returned. On a success 0 is returned.
222 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
224 if(ulen<0||ulen>MAX_SOCK_ADDR)
228 if(copy_from_user(kaddr,uaddr,ulen))
234 * move_addr_to_user - copy an address to user space
235 * @kaddr: kernel space address
236 * @klen: length of address in kernel
237 * @uaddr: user space address
238 * @ulen: pointer to user length field
240 * The value pointed to by ulen on entry is the buffer length available.
241 * This is overwritten with the buffer space used. -EINVAL is returned
242 * if an overlong buffer is specified or a negative buffer size. -EFAULT
243 * is returned if either the buffer or the length field are not
245 * After copying the data up to the limit the user specifies, the true
246 * length of the data is written over the length limit the user
247 * specified. Zero is returned for a success.
250 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
255 if((err=get_user(len, ulen)))
259 if(len<0 || len> MAX_SOCK_ADDR)
263 if(copy_to_user(uaddr,kaddr,len))
267 * "fromlen shall refer to the value before truncation.."
270 return __put_user(klen, ulen);
273 #define SOCKFS_MAGIC 0x534F434B
275 static kmem_cache_t * sock_inode_cachep;
277 static struct inode *sock_alloc_inode(struct super_block *sb)
279 struct socket_alloc *ei;
280 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
283 init_waitqueue_head(&ei->socket.wait);
285 ei->socket.fasync_list = NULL;
286 ei->socket.state = SS_UNCONNECTED;
287 ei->socket.flags = 0;
288 ei->socket.ops = NULL;
289 ei->socket.sk = NULL;
290 ei->socket.file = NULL;
291 ei->socket.flags = 0;
293 return &ei->vfs_inode;
296 static void sock_destroy_inode(struct inode *inode)
298 kmem_cache_free(sock_inode_cachep,
299 container_of(inode, struct socket_alloc, vfs_inode));
302 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
304 struct socket_alloc *ei = (struct socket_alloc *) foo;
306 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
307 SLAB_CTOR_CONSTRUCTOR)
308 inode_init_once(&ei->vfs_inode);
311 static int init_inodecache(void)
313 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
314 sizeof(struct socket_alloc),
315 0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
317 if (sock_inode_cachep == NULL)
322 static struct super_operations sockfs_ops = {
323 .alloc_inode = sock_alloc_inode,
324 .destroy_inode =sock_destroy_inode,
325 .statfs = simple_statfs,
328 static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
329 int flags, const char *dev_name, void *data)
331 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
334 static struct vfsmount *sock_mnt;
336 static struct file_system_type sock_fs_type = {
338 .get_sb = sockfs_get_sb,
339 .kill_sb = kill_anon_super,
341 static int sockfs_delete_dentry(struct dentry *dentry)
345 static struct dentry_operations sockfs_dentry_operations = {
346 .d_delete = sockfs_delete_dentry,
350 * Obtains the first available file descriptor and sets it up for use.
352 * This function creates file structure and maps it to fd space
353 * of current process. On success it returns file descriptor
354 * and file struct implicitly stored in sock->file.
355 * Note that another thread may close file descriptor before we return
356 * from this function. We use the fact that now we do not refer
357 * to socket after mapping. If one day we will need it, this
358 * function will increment ref. count on file by 1.
360 * In any case returned fd MAY BE not valid!
361 * This race condition is unavoidable
362 * with shared fd spaces, we cannot solve it inside kernel,
363 * but we take care of internal coherence yet.
366 int sock_map_fd(struct socket *sock)
373 * Find a file descriptor suitable for return to the user.
376 fd = get_unused_fd();
378 struct file *file = get_empty_filp();
386 sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
388 this.len = strlen(name);
389 this.hash = SOCK_INODE(sock)->i_ino;
391 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
392 if (!file->f_dentry) {
398 file->f_dentry->d_op = &sockfs_dentry_operations;
399 d_add(file->f_dentry, SOCK_INODE(sock));
400 file->f_vfsmnt = mntget(sock_mnt);
401 file->f_mapping = file->f_dentry->d_inode->i_mapping;
404 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
405 file->f_mode = FMODE_READ | FMODE_WRITE;
406 file->f_flags = O_RDWR;
408 fd_install(fd, file);
416 * sockfd_lookup - Go from a file number to its socket slot
418 * @err: pointer to an error code return
420 * The file handle passed in is locked and the socket it is bound
421 * too is returned. If an error occurs the err pointer is overwritten
422 * with a negative errno code and NULL is returned. The function checks
423 * for both invalid handles and passing a handle which is not a socket.
425 * On a success the socket object pointer is returned.
428 struct socket *sockfd_lookup(int fd, int *err)
434 if (!(file = fget(fd)))
440 inode = file->f_dentry->d_inode;
441 if (!inode->i_sock || !(sock = SOCKET_I(inode)))
448 if (sock->file != file) {
449 printk(KERN_ERR "socki_lookup: socket file changed!\n");
456 * sock_alloc - allocate a socket
458 * Allocate a new inode and socket object. The two are bound together
459 * and initialised. The socket is then returned. If we are out of inodes
463 static struct socket *sock_alloc(void)
465 struct inode * inode;
466 struct socket * sock;
468 inode = new_inode(sock_mnt->mnt_sb);
472 sock = SOCKET_I(inode);
474 inode->i_mode = S_IFSOCK|S_IRWXUGO;
476 inode->i_uid = current->fsuid;
477 inode->i_gid = current->fsgid;
479 get_cpu_var(sockets_in_use)++;
480 put_cpu_var(sockets_in_use);
485 * In theory you can't get an open on this inode, but /proc provides
486 * a back door. Remember to keep it shut otherwise you'll let the
487 * creepy crawlies in.
490 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
495 struct file_operations bad_sock_fops = {
496 .owner = THIS_MODULE,
497 .open = sock_no_open,
501 * sock_release - close a socket
502 * @sock: socket to close
504 * The socket is released from the protocol stack if it has a release
505 * callback, and the inode is then released if the socket is bound to
506 * an inode not a file.
509 void sock_release(struct socket *sock)
512 struct module *owner = sock->ops->owner;
514 sock->ops->release(sock);
519 if (sock->fasync_list)
520 printk(KERN_ERR "sock_release: fasync list not empty!\n");
522 get_cpu_var(sockets_in_use)--;
523 put_cpu_var(sockets_in_use);
525 iput(SOCK_INODE(sock));
531 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
532 struct msghdr *msg, size_t size)
534 struct sock_iocb *si = kiocb_to_siocb(iocb);
542 err = security_socket_sendmsg(sock, msg, size);
546 len = sock->ops->sendmsg(iocb, sock, msg, size);
549 vx_sock_send(sock->sk, size);
551 vx_sock_fail(sock->sk, size);
553 vxdprintk(VXD_CBIT(net, 7),
554 "__sock_sendmsg: %p[%p,%p,%p;%d]:%d/%d",
556 (sock->sk)?sock->sk->sk_nx_info:0,
557 (sock->sk)?sock->sk->sk_vx_info:0,
558 (sock->sk)?sock->sk->sk_xid:0,
559 (unsigned int)size, len);
563 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
566 struct sock_iocb siocb;
569 init_sync_kiocb(&iocb, NULL);
570 iocb.private = &siocb;
571 ret = __sock_sendmsg(&iocb, sock, msg, size);
572 if (-EIOCBQUEUED == ret)
573 ret = wait_on_sync_kiocb(&iocb);
577 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
578 struct kvec *vec, size_t num, size_t size)
580 mm_segment_t oldfs = get_fs();
585 * the following is safe, since for compiler definitions of kvec and
586 * iovec are identical, yielding the same in-core layout and alignment
588 msg->msg_iov = (struct iovec *)vec,
589 msg->msg_iovlen = num;
590 result = sock_sendmsg(sock, msg, size);
595 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
596 struct msghdr *msg, size_t size, int flags)
599 struct sock_iocb *si = kiocb_to_siocb(iocb);
607 err = security_socket_recvmsg(sock, msg, size, flags);
611 len = sock->ops->recvmsg(iocb, sock, msg, size, flags);
612 if ((len >= 0) && sock->sk)
613 vx_sock_recv(sock->sk, len);
614 vxdprintk(VXD_CBIT(net, 7),
615 "__sock_recvmsg: %p[%p,%p,%p;%d]:%d/%d",
617 (sock->sk)?sock->sk->sk_nx_info:0,
618 (sock->sk)?sock->sk->sk_vx_info:0,
619 (sock->sk)?sock->sk->sk_xid:0,
620 (unsigned int)size, len);
624 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
625 size_t size, int flags)
628 struct sock_iocb siocb;
631 init_sync_kiocb(&iocb, NULL);
632 iocb.private = &siocb;
633 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
634 if (-EIOCBQUEUED == ret)
635 ret = wait_on_sync_kiocb(&iocb);
639 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
640 struct kvec *vec, size_t num,
641 size_t size, int flags)
643 mm_segment_t oldfs = get_fs();
648 * the following is safe, since for compiler definitions of kvec and
649 * iovec are identical, yielding the same in-core layout and alignment
651 msg->msg_iov = (struct iovec *)vec,
652 msg->msg_iovlen = num;
653 result = sock_recvmsg(sock, msg, size, flags);
658 static void sock_aio_dtor(struct kiocb *iocb)
660 kfree(iocb->private);
664 * Read data from a socket. ubuf is a user mode pointer. We make sure the user
665 * area ubuf...ubuf+size-1 is writable before asking the protocol.
668 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
669 size_t size, loff_t pos)
671 struct sock_iocb *x, siocb;
677 if (size==0) /* Match SYS5 behaviour */
680 if (is_sync_kiocb(iocb))
683 x = kmalloc(sizeof(struct sock_iocb), GFP_KERNEL);
686 iocb->ki_dtor = sock_aio_dtor;
690 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
692 x->async_msg.msg_name = NULL;
693 x->async_msg.msg_namelen = 0;
694 x->async_msg.msg_iov = &x->async_iov;
695 x->async_msg.msg_iovlen = 1;
696 x->async_msg.msg_control = NULL;
697 x->async_msg.msg_controllen = 0;
698 x->async_iov.iov_base = ubuf;
699 x->async_iov.iov_len = size;
700 flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
702 return __sock_recvmsg(iocb, sock, &x->async_msg, size, flags);
707 * Write data to a socket. We verify that the user area ubuf..ubuf+size-1
708 * is readable by the user process.
711 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
712 size_t size, loff_t pos)
714 struct sock_iocb *x, siocb;
719 if(size==0) /* Match SYS5 behaviour */
722 if (is_sync_kiocb(iocb))
725 x = kmalloc(sizeof(struct sock_iocb), GFP_KERNEL);
728 iocb->ki_dtor = sock_aio_dtor;
732 sock = SOCKET_I(iocb->ki_filp->f_dentry->d_inode);
734 x->async_msg.msg_name = NULL;
735 x->async_msg.msg_namelen = 0;
736 x->async_msg.msg_iov = &x->async_iov;
737 x->async_msg.msg_iovlen = 1;
738 x->async_msg.msg_control = NULL;
739 x->async_msg.msg_controllen = 0;
740 x->async_msg.msg_flags = !(iocb->ki_filp->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
741 if (sock->type == SOCK_SEQPACKET)
742 x->async_msg.msg_flags |= MSG_EOR;
743 x->async_iov.iov_base = (void __user *)ubuf;
744 x->async_iov.iov_len = size;
746 return __sock_sendmsg(iocb, sock, &x->async_msg, size);
749 ssize_t sock_sendpage(struct file *file, struct page *page,
750 int offset, size_t size, loff_t *ppos, int more)
755 sock = SOCKET_I(file->f_dentry->d_inode);
757 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
761 return sock->ops->sendpage(sock, page, offset, size, flags);
764 static int sock_readv_writev(int type, struct inode * inode,
765 struct file * file, const struct iovec * iov,
766 long count, size_t size)
771 sock = SOCKET_I(inode);
775 msg.msg_control = NULL;
776 msg.msg_controllen = 0;
777 msg.msg_iov = (struct iovec *) iov;
778 msg.msg_iovlen = count;
779 msg.msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
781 /* read() does a VERIFY_WRITE */
782 if (type == VERIFY_WRITE)
783 return sock_recvmsg(sock, &msg, size, msg.msg_flags);
785 if (sock->type == SOCK_SEQPACKET)
786 msg.msg_flags |= MSG_EOR;
788 return sock_sendmsg(sock, &msg, size);
791 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
792 unsigned long count, loff_t *ppos)
796 for (i = 0 ; i < count ; i++)
797 tot_len += vector[i].iov_len;
798 return sock_readv_writev(VERIFY_WRITE, file->f_dentry->d_inode,
799 file, vector, count, tot_len);
802 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
803 unsigned long count, loff_t *ppos)
807 for (i = 0 ; i < count ; i++)
808 tot_len += vector[i].iov_len;
809 return sock_readv_writev(VERIFY_READ, file->f_dentry->d_inode,
810 file, vector, count, tot_len);
815 * Atomic setting of ioctl hooks to avoid race
816 * with module unload.
819 static DECLARE_MUTEX(br_ioctl_mutex);
820 static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
822 void brioctl_set(int (*hook)(unsigned int, void __user *))
824 down(&br_ioctl_mutex);
825 br_ioctl_hook = hook;
828 EXPORT_SYMBOL(brioctl_set);
830 static DECLARE_MUTEX(vlan_ioctl_mutex);
831 static int (*vlan_ioctl_hook)(void __user *arg);
833 void vlan_ioctl_set(int (*hook)(void __user *))
835 down(&vlan_ioctl_mutex);
836 vlan_ioctl_hook = hook;
837 up(&vlan_ioctl_mutex);
839 EXPORT_SYMBOL(vlan_ioctl_set);
841 static DECLARE_MUTEX(dlci_ioctl_mutex);
842 static int (*dlci_ioctl_hook)(unsigned int, void __user *);
844 void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
846 down(&dlci_ioctl_mutex);
847 dlci_ioctl_hook = hook;
848 up(&dlci_ioctl_mutex);
850 EXPORT_SYMBOL(dlci_ioctl_set);
853 * With an ioctl, arg may well be a user mode pointer, but we don't know
854 * what to do with it - that's up to the protocol still.
857 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
860 void __user *argp = (void __user *)arg;
863 sock = SOCKET_I(file->f_dentry->d_inode);
864 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
865 err = dev_ioctl(cmd, argp);
868 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
869 err = dev_ioctl(cmd, argp);
871 #endif /* WIRELESS_EXT */
876 if (get_user(pid, (int __user *)argp))
878 err = f_setown(sock->file, pid, 1);
882 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
890 request_module("bridge");
892 down(&br_ioctl_mutex);
894 err = br_ioctl_hook(cmd, argp);
900 if (!vlan_ioctl_hook)
901 request_module("8021q");
903 down(&vlan_ioctl_mutex);
905 err = vlan_ioctl_hook(argp);
906 up(&vlan_ioctl_mutex);
910 /* Convert this to call through a hook */
911 err = divert_ioctl(cmd, argp);
916 if (!dlci_ioctl_hook)
917 request_module("dlci");
919 if (dlci_ioctl_hook) {
920 down(&dlci_ioctl_mutex);
921 err = dlci_ioctl_hook(cmd, argp);
922 up(&dlci_ioctl_mutex);
926 err = sock->ops->ioctl(sock, cmd, arg);
932 int sock_create_lite(int family, int type, int protocol, struct socket **res)
935 struct socket *sock = NULL;
937 err = security_socket_create(family, type, protocol, 1);
947 security_socket_post_create(sock, family, type, protocol, 1);
954 /* No kernel lock held - perfect */
955 static unsigned int sock_poll(struct file *file, poll_table * wait)
960 * We can't return errors to poll, so it's either yes or no.
962 sock = SOCKET_I(file->f_dentry->d_inode);
963 return sock->ops->poll(file, sock, wait);
966 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
968 struct socket *sock = SOCKET_I(file->f_dentry->d_inode);
970 return sock->ops->mmap(file, sock, vma);
973 int sock_close(struct inode *inode, struct file *filp)
976 * It was possible the inode is NULL we were
977 * closing an unfinished socket.
982 printk(KERN_DEBUG "sock_close: NULL inode\n");
985 sock_fasync(-1, filp, 0);
986 sock_release(SOCKET_I(inode));
991 * Update the socket async list
993 * Fasync_list locking strategy.
995 * 1. fasync_list is modified only under process context socket lock
996 * i.e. under semaphore.
997 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
998 * or under socket lock.
999 * 3. fasync_list can be used from softirq context, so that
1000 * modification under socket lock have to be enhanced with
1001 * write_lock_bh(&sk->sk_callback_lock).
1005 static int sock_fasync(int fd, struct file *filp, int on)
1007 struct fasync_struct *fa, *fna=NULL, **prev;
1008 struct socket *sock;
1013 fna=(struct fasync_struct *)kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1018 sock = SOCKET_I(filp->f_dentry->d_inode);
1020 if ((sk=sock->sk) == NULL) {
1028 prev=&(sock->fasync_list);
1030 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1031 if (fa->fa_file==filp)
1038 write_lock_bh(&sk->sk_callback_lock);
1040 write_unlock_bh(&sk->sk_callback_lock);
1047 fna->magic=FASYNC_MAGIC;
1048 fna->fa_next=sock->fasync_list;
1049 write_lock_bh(&sk->sk_callback_lock);
1050 sock->fasync_list=fna;
1051 write_unlock_bh(&sk->sk_callback_lock);
1057 write_lock_bh(&sk->sk_callback_lock);
1059 write_unlock_bh(&sk->sk_callback_lock);
1065 release_sock(sock->sk);
1069 /* This function may be called only under socket lock or callback_lock */
1071 int sock_wake_async(struct socket *sock, int how, int band)
1073 if (!sock || !sock->fasync_list)
1079 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1083 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1088 __kill_fasync(sock->fasync_list, SIGIO, band);
1091 __kill_fasync(sock->fasync_list, SIGURG, band);
1096 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1099 struct socket *sock;
1102 * Check protocol is in range
1104 if (family < 0 || family >= NPROTO)
1105 return -EAFNOSUPPORT;
1106 if (type < 0 || type >= SOCK_MAX)
1109 /* disable IPv6 inside vservers for now */
1110 if (family == PF_INET6 && !vx_check(0, VX_ADMIN))
1111 return -EAFNOSUPPORT;
1115 This uglymoron is moved from INET layer to here to avoid
1116 deadlock in module load.
1118 if (family == PF_INET && type == SOCK_PACKET) {
1122 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1127 err = security_socket_create(family, type, protocol, kern);
1131 #if defined(CONFIG_KMOD)
1132 /* Attempt to load a protocol module if the find failed.
1134 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1135 * requested real, full-featured networking support upon configuration.
1136 * Otherwise module support will break!
1138 if (net_families[family]==NULL)
1140 request_module("net-pf-%d",family);
1144 net_family_read_lock();
1145 if (net_families[family] == NULL) {
1146 err = -EAFNOSUPPORT;
1151 * Allocate the socket and allow the family to set things up. if
1152 * the protocol is 0, the family is instructed to select an appropriate
1156 if (!(sock = sock_alloc())) {
1157 printk(KERN_WARNING "socket: no more sockets\n");
1158 err = -ENFILE; /* Not exactly a match, but its the
1159 closest posix thing */
1166 * We will call the ->create function, that possibly is in a loadable
1167 * module, so we have to bump that loadable module refcnt first.
1169 err = -EAFNOSUPPORT;
1170 if (!try_module_get(net_families[family]->owner))
1173 if ((err = net_families[family]->create(sock, protocol)) < 0)
1174 goto out_module_put;
1176 * Now to bump the refcnt of the [loadable] module that owns this
1177 * socket at sock_release time we decrement its refcnt.
1179 if (!try_module_get(sock->ops->owner)) {
1181 goto out_module_put;
1184 * Now that we're done with the ->create function, the [loadable]
1185 * module can have its refcnt decremented
1187 module_put(net_families[family]->owner);
1189 security_socket_post_create(sock, family, type, protocol, kern);
1192 net_family_read_unlock();
1195 module_put(net_families[family]->owner);
1201 int sock_create(int family, int type, int protocol, struct socket **res)
1203 return __sock_create(family, type, protocol, res, 0);
1206 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1208 return __sock_create(family, type, protocol, res, 1);
1211 asmlinkage long sys_socket(int family, int type, int protocol)
1214 struct socket *sock;
1216 retval = sock_create(family, type, protocol, &sock);
1220 set_bit(SOCK_USER_SOCKET, &sock->flags);
1221 retval = sock_map_fd(sock);
1226 /* It may be already another descriptor 8) Not kernel problem. */
1235 * Create a pair of connected sockets.
1238 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1240 struct socket *sock1, *sock2;
1244 * Obtain the first socket and check if the underlying protocol
1245 * supports the socketpair call.
1248 err = sock_create(family, type, protocol, &sock1);
1251 set_bit(SOCK_USER_SOCKET, &sock1->flags);
1253 err = sock_create(family, type, protocol, &sock2);
1256 set_bit(SOCK_USER_SOCKET, &sock2->flags);
1258 err = sock1->ops->socketpair(sock1, sock2);
1260 goto out_release_both;
1264 err = sock_map_fd(sock1);
1266 goto out_release_both;
1269 err = sock_map_fd(sock2);
1274 /* fd1 and fd2 may be already another descriptors.
1275 * Not kernel problem.
1278 err = put_user(fd1, &usockvec[0]);
1280 err = put_user(fd2, &usockvec[1]);
1289 sock_release(sock2);
1294 sock_release(sock2);
1296 sock_release(sock1);
1303 * Bind a name to a socket. Nothing much to do here since it's
1304 * the protocol's responsibility to handle the local address.
1306 * We move the socket address to kernel space before we call
1307 * the protocol layer (having also checked the address is ok).
1310 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1312 struct socket *sock;
1313 char address[MAX_SOCK_ADDR];
1316 if((sock = sockfd_lookup(fd,&err))!=NULL)
1318 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1319 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1324 err = sock->ops->bind(sock, (struct sockaddr *)address, addrlen);
1333 * Perform a listen. Basically, we allow the protocol to do anything
1334 * necessary for a listen, and if that works, we mark the socket as
1335 * ready for listening.
1338 int sysctl_somaxconn = SOMAXCONN;
1340 asmlinkage long sys_listen(int fd, int backlog)
1342 struct socket *sock;
1345 if ((sock = sockfd_lookup(fd, &err)) != NULL) {
1346 if ((unsigned) backlog > sysctl_somaxconn)
1347 backlog = sysctl_somaxconn;
1349 err = security_socket_listen(sock, backlog);
1355 err=sock->ops->listen(sock, backlog);
1363 * For accept, we attempt to create a new socket, set up the link
1364 * with the client, wake up the client, then return the new
1365 * connected fd. We collect the address of the connector in kernel
1366 * space and move it to user at the very end. This is unclean because
1367 * we open the socket then return an error.
1369 * 1003.1g adds the ability to recvmsg() to query connection pending
1370 * status to recvmsg. We need to add that support in a way thats
1371 * clean when we restucture accept also.
1374 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1376 struct socket *sock, *newsock;
1378 char address[MAX_SOCK_ADDR];
1380 sock = sockfd_lookup(fd, &err);
1385 if (!(newsock = sock_alloc()))
1388 newsock->type = sock->type;
1389 newsock->ops = sock->ops;
1391 err = security_socket_accept(sock, newsock);
1396 * We don't need try_module_get here, as the listening socket (sock)
1397 * has the protocol module (sock->ops->owner) held.
1399 __module_get(newsock->ops->owner);
1401 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1405 if (upeer_sockaddr) {
1406 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1407 err = -ECONNABORTED;
1410 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1415 /* File flags are not inherited via accept() unlike another OSes. */
1417 if ((err = sock_map_fd(newsock)) < 0)
1420 security_socket_post_accept(sock, newsock);
1427 sock_release(newsock);
1433 * Attempt to connect to a socket with the server address. The address
1434 * is in user space so we verify it is OK and move it to kernel space.
1436 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1439 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1440 * other SEQPACKET protocols that take time to connect() as it doesn't
1441 * include the -EINPROGRESS status for such sockets.
1444 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1446 struct socket *sock;
1447 char address[MAX_SOCK_ADDR];
1450 sock = sockfd_lookup(fd, &err);
1453 err = move_addr_to_kernel(uservaddr, addrlen, address);
1457 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1461 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1462 sock->file->f_flags);
1470 * Get the local address ('name') of a socket object. Move the obtained
1471 * name to user space.
1474 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1476 struct socket *sock;
1477 char address[MAX_SOCK_ADDR];
1480 sock = sockfd_lookup(fd, &err);
1484 err = security_socket_getsockname(sock);
1488 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1491 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1500 * Get the remote address ('name') of a socket object. Move the obtained
1501 * name to user space.
1504 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1506 struct socket *sock;
1507 char address[MAX_SOCK_ADDR];
1510 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1512 err = security_socket_getpeername(sock);
1518 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1520 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1527 * Send a datagram to a given address. We move the address into kernel
1528 * space and check the user space data area is readable before invoking
1532 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1533 struct sockaddr __user *addr, int addr_len)
1535 struct socket *sock;
1536 char address[MAX_SOCK_ADDR];
1541 sock = sockfd_lookup(fd, &err);
1549 msg.msg_control=NULL;
1550 msg.msg_controllen=0;
1554 err = move_addr_to_kernel(addr, addr_len, address);
1557 msg.msg_name=address;
1558 msg.msg_namelen=addr_len;
1560 if (sock->file->f_flags & O_NONBLOCK)
1561 flags |= MSG_DONTWAIT;
1562 msg.msg_flags = flags;
1563 err = sock_sendmsg(sock, &msg, len);
1572 * Send a datagram down a socket.
1575 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1577 return sys_sendto(fd, buff, len, flags, NULL, 0);
1581 * Receive a frame from the socket and optionally record the address of the
1582 * sender. We verify the buffers are writable and if needed move the
1583 * sender address from kernel to user space.
1586 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1587 struct sockaddr __user *addr, int __user *addr_len)
1589 struct socket *sock;
1592 char address[MAX_SOCK_ADDR];
1595 sock = sockfd_lookup(fd, &err);
1599 msg.msg_control=NULL;
1600 msg.msg_controllen=0;
1605 msg.msg_name=address;
1606 msg.msg_namelen=MAX_SOCK_ADDR;
1607 if (sock->file->f_flags & O_NONBLOCK)
1608 flags |= MSG_DONTWAIT;
1609 err=sock_recvmsg(sock, &msg, size, flags);
1611 if(err >= 0 && addr != NULL)
1613 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1623 * Receive a datagram from a socket.
1626 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1628 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1632 * Set a socket option. Because we don't know the option lengths we have
1633 * to pass the user mode parameter for the protocols to sort out.
1636 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1639 struct socket *sock;
1644 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1646 err = security_socket_setsockopt(sock,level,optname);
1652 if (level == SOL_SOCKET)
1653 err=sock_setsockopt(sock,level,optname,optval,optlen);
1655 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1662 * Get a socket option. Because we don't know the option lengths we have
1663 * to pass a user mode parameter for the protocols to sort out.
1666 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1669 struct socket *sock;
1671 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1673 err = security_socket_getsockopt(sock, level,
1680 if (level == SOL_SOCKET)
1681 err=sock_getsockopt(sock,level,optname,optval,optlen);
1683 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1691 * Shutdown a socket.
1694 asmlinkage long sys_shutdown(int fd, int how)
1697 struct socket *sock;
1699 if ((sock = sockfd_lookup(fd, &err))!=NULL)
1701 err = security_socket_shutdown(sock, how);
1707 err=sock->ops->shutdown(sock, how);
1713 /* A couple of helpful macros for getting the address of the 32/64 bit
1714 * fields which are the same type (int / unsigned) on our platforms.
1716 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1717 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1718 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1722 * BSD sendmsg interface
1725 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1727 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1728 struct socket *sock;
1729 char address[MAX_SOCK_ADDR];
1730 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1731 unsigned char ctl[sizeof(struct cmsghdr) + 20]; /* 20 is size of ipv6_pktinfo */
1732 unsigned char *ctl_buf = ctl;
1733 struct msghdr msg_sys;
1734 int err, ctl_len, iov_size, total_len;
1737 if (MSG_CMSG_COMPAT & flags) {
1738 if (get_compat_msghdr(&msg_sys, msg_compat))
1740 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1743 sock = sockfd_lookup(fd, &err);
1747 /* do not move before msg_sys is valid */
1749 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1752 /* Check whether to allocate the iovec area*/
1754 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1755 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1756 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1761 /* This will also move the address data into kernel space */
1762 if (MSG_CMSG_COMPAT & flags) {
1763 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1765 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1772 if (msg_sys.msg_controllen > INT_MAX)
1774 ctl_len = msg_sys.msg_controllen;
1775 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1776 err = cmsghdr_from_user_compat_to_kern(&msg_sys, ctl, sizeof(ctl));
1779 ctl_buf = msg_sys.msg_control;
1780 } else if (ctl_len) {
1781 if (ctl_len > sizeof(ctl))
1783 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1784 if (ctl_buf == NULL)
1789 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1790 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1791 * checking falls down on this.
1793 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1795 msg_sys.msg_control = ctl_buf;
1797 msg_sys.msg_flags = flags;
1799 if (sock->file->f_flags & O_NONBLOCK)
1800 msg_sys.msg_flags |= MSG_DONTWAIT;
1801 err = sock_sendmsg(sock, &msg_sys, total_len);
1805 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1807 if (iov != iovstack)
1808 sock_kfree_s(sock->sk, iov, iov_size);
1816 * BSD recvmsg interface
1819 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1821 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1822 struct socket *sock;
1823 struct iovec iovstack[UIO_FASTIOV];
1824 struct iovec *iov=iovstack;
1825 struct msghdr msg_sys;
1826 unsigned long cmsg_ptr;
1827 int err, iov_size, total_len, len;
1829 /* kernel mode address */
1830 char addr[MAX_SOCK_ADDR];
1832 /* user mode address pointers */
1833 struct sockaddr __user *uaddr;
1834 int __user *uaddr_len;
1836 if (MSG_CMSG_COMPAT & flags) {
1837 if (get_compat_msghdr(&msg_sys, msg_compat))
1840 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1843 sock = sockfd_lookup(fd, &err);
1848 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1851 /* Check whether to allocate the iovec area*/
1853 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1854 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1855 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1861 * Save the user-mode address (verify_iovec will change the
1862 * kernel msghdr to use the kernel address space)
1865 uaddr = (void __user *) msg_sys.msg_name;
1866 uaddr_len = COMPAT_NAMELEN(msg);
1867 if (MSG_CMSG_COMPAT & flags) {
1868 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1870 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1875 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1876 msg_sys.msg_flags = 0;
1877 if (MSG_CMSG_COMPAT & flags)
1878 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1880 if (sock->file->f_flags & O_NONBLOCK)
1881 flags |= MSG_DONTWAIT;
1882 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1887 if (uaddr != NULL) {
1888 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1892 err = __put_user(msg_sys.msg_flags, COMPAT_FLAGS(msg));
1895 if (MSG_CMSG_COMPAT & flags)
1896 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1897 &msg_compat->msg_controllen);
1899 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1900 &msg->msg_controllen);
1906 if (iov != iovstack)
1907 sock_kfree_s(sock->sk, iov, iov_size);
1914 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1916 /* Argument list sizes for sys_socketcall */
1917 #define AL(x) ((x) * sizeof(unsigned long))
1918 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1919 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1920 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1924 * System call vectors.
1926 * Argument checking cleaned up. Saved 20% in size.
1927 * This function doesn't need to set the kernel lock because
1928 * it is set by the callees.
1931 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1934 unsigned long a0,a1;
1937 if(call<1||call>SYS_RECVMSG)
1940 /* copy_from_user should be SMP safe. */
1941 if (copy_from_user(a, args, nargs[call]))
1950 err = sys_socket(a0,a1,a[2]);
1953 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1956 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1959 err = sys_listen(a0,a1);
1962 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1964 case SYS_GETSOCKNAME:
1965 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
1967 case SYS_GETPEERNAME:
1968 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
1970 case SYS_SOCKETPAIR:
1971 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
1974 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1977 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
1978 (struct sockaddr __user *)a[4], a[5]);
1981 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
1984 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
1985 (struct sockaddr __user *)a[4], (int __user *)a[5]);
1988 err = sys_shutdown(a0,a1);
1990 case SYS_SETSOCKOPT:
1991 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
1993 case SYS_GETSOCKOPT:
1994 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
1997 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2000 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2009 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2012 * This function is called by a protocol handler that wants to
2013 * advertise its address family, and have it linked into the
2017 int sock_register(struct net_proto_family *ops)
2021 if (ops->family >= NPROTO) {
2022 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2025 net_family_write_lock();
2027 if (net_families[ops->family] == NULL) {
2028 net_families[ops->family]=ops;
2031 net_family_write_unlock();
2032 printk(KERN_INFO "NET: Registered protocol family %d\n",
2038 * This function is called by a protocol handler that wants to
2039 * remove its address family, and have it unlinked from the
2043 int sock_unregister(int family)
2045 if (family < 0 || family >= NPROTO)
2048 net_family_write_lock();
2049 net_families[family]=NULL;
2050 net_family_write_unlock();
2051 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2057 extern void sk_init(void);
2059 void __init sock_init(void)
2062 * Initialize sock SLAB cache.
2069 * Initialize skbuff SLAB cache
2075 * Initialize the protocols module.
2079 register_filesystem(&sock_fs_type);
2080 sock_mnt = kern_mount(&sock_fs_type);
2081 /* The real protocol initialization is performed when
2082 * do_initcalls is run.
2085 #ifdef CONFIG_NETFILTER
2090 #ifdef CONFIG_PROC_FS
2091 void socket_seq_show(struct seq_file *seq)
2096 for (cpu = 0; cpu < NR_CPUS; cpu++)
2097 counter += per_cpu(sockets_in_use, cpu);
2099 /* It can be negative, by the way. 8) */
2103 seq_printf(seq, "sockets: used %d\n", counter);
2105 #endif /* CONFIG_PROC_FS */
2107 /* ABI emulation layers need these two */
2108 EXPORT_SYMBOL(move_addr_to_kernel);
2109 EXPORT_SYMBOL(move_addr_to_user);
2110 EXPORT_SYMBOL(sock_create);
2111 EXPORT_SYMBOL(sock_create_kern);
2112 EXPORT_SYMBOL(sock_create_lite);
2113 EXPORT_SYMBOL(sock_map_fd);
2114 EXPORT_SYMBOL(sock_recvmsg);
2115 EXPORT_SYMBOL(sock_register);
2116 EXPORT_SYMBOL(sock_release);
2117 EXPORT_SYMBOL(sock_sendmsg);
2118 EXPORT_SYMBOL(sock_unregister);
2119 EXPORT_SYMBOL(sock_wake_async);
2120 EXPORT_SYMBOL(sockfd_lookup);
2121 EXPORT_SYMBOL(kernel_sendmsg);
2122 EXPORT_SYMBOL(kernel_recvmsg);