2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
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/mutex.h>
72 #include <linux/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/divert.h>
82 #include <linux/mount.h>
83 #include <linux/security.h>
84 #include <linux/syscalls.h>
85 #include <linux/compat.h>
86 #include <linux/kmod.h>
87 #include <linux/audit.h>
88 #include <linux/wireless.h>
90 #include <asm/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
96 #include <linux/netfilter.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);
111 static long compat_sock_ioctl(struct file *file,
112 unsigned int cmd, unsigned long arg);
114 static int sock_fasync(int fd, struct file *filp, int on);
115 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
116 unsigned long count, loff_t *ppos);
117 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
118 unsigned long count, loff_t *ppos);
119 static ssize_t sock_sendpage(struct file *file, struct page *page,
120 int offset, size_t size, loff_t *ppos, int more);
123 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
124 * in the operation structures but are done directly via the socketcall() multiplexor.
127 static struct file_operations socket_file_ops = {
128 .owner = THIS_MODULE,
130 .aio_read = sock_aio_read,
131 .aio_write = sock_aio_write,
133 .unlocked_ioctl = sock_ioctl,
135 .compat_ioctl = compat_sock_ioctl,
138 .open = sock_no_open, /* special open code to disallow open via /proc */
139 .release = sock_close,
140 .fasync = sock_fasync,
142 .writev = sock_writev,
143 .sendpage = sock_sendpage,
144 .splice_write = generic_splice_sendpage,
148 * The protocol list. Each protocol is registered in here.
151 static struct net_proto_family *net_families[NPROTO];
153 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
154 static atomic_t net_family_lockct = ATOMIC_INIT(0);
155 static DEFINE_SPINLOCK(net_family_lock);
157 /* The strategy is: modifications net_family vector are short, do not
158 sleep and veeery rare, but read access should be free of any exclusive
162 static void net_family_write_lock(void)
164 spin_lock(&net_family_lock);
165 while (atomic_read(&net_family_lockct) != 0) {
166 spin_unlock(&net_family_lock);
170 spin_lock(&net_family_lock);
174 static __inline__ void net_family_write_unlock(void)
176 spin_unlock(&net_family_lock);
179 static __inline__ void net_family_read_lock(void)
181 atomic_inc(&net_family_lockct);
182 spin_unlock_wait(&net_family_lock);
185 static __inline__ void net_family_read_unlock(void)
187 atomic_dec(&net_family_lockct);
191 #define net_family_write_lock() do { } while(0)
192 #define net_family_write_unlock() do { } while(0)
193 #define net_family_read_lock() do { } while(0)
194 #define net_family_read_unlock() do { } while(0)
199 * Statistics counters of the socket lists
202 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
205 * Support routines. Move socket addresses back and forth across the kernel/user
206 * divide and look after the messy bits.
209 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
210 16 for IP, 16 for IPX,
213 must be at least one bigger than
214 the AF_UNIX size (see net/unix/af_unix.c
219 * move_addr_to_kernel - copy a socket address into kernel space
220 * @uaddr: Address in user space
221 * @kaddr: Address in kernel space
222 * @ulen: Length in user space
224 * The address is copied into kernel space. If the provided address is
225 * too long an error code of -EINVAL is returned. If the copy gives
226 * invalid addresses -EFAULT is returned. On a success 0 is returned.
229 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
231 if(ulen<0||ulen>MAX_SOCK_ADDR)
235 if(copy_from_user(kaddr,uaddr,ulen))
237 return audit_sockaddr(ulen, kaddr);
241 * move_addr_to_user - copy an address to user space
242 * @kaddr: kernel space address
243 * @klen: length of address in kernel
244 * @uaddr: user space address
245 * @ulen: pointer to user length field
247 * The value pointed to by ulen on entry is the buffer length available.
248 * This is overwritten with the buffer space used. -EINVAL is returned
249 * if an overlong buffer is specified or a negative buffer size. -EFAULT
250 * is returned if either the buffer or the length field are not
252 * After copying the data up to the limit the user specifies, the true
253 * length of the data is written over the length limit the user
254 * specified. Zero is returned for a success.
257 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
262 if((err=get_user(len, ulen)))
266 if(len<0 || len> MAX_SOCK_ADDR)
270 if (audit_sockaddr(klen, kaddr))
272 if(copy_to_user(uaddr,kaddr,len))
276 * "fromlen shall refer to the value before truncation.."
279 return __put_user(klen, ulen);
282 #define SOCKFS_MAGIC 0x534F434B
284 static kmem_cache_t * sock_inode_cachep __read_mostly;
286 static struct inode *sock_alloc_inode(struct super_block *sb)
288 struct socket_alloc *ei;
289 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
292 init_waitqueue_head(&ei->socket.wait);
294 ei->socket.fasync_list = NULL;
295 ei->socket.state = SS_UNCONNECTED;
296 ei->socket.flags = 0;
297 ei->socket.ops = NULL;
298 ei->socket.sk = NULL;
299 ei->socket.file = NULL;
300 ei->socket.flags = 0;
302 return &ei->vfs_inode;
305 static void sock_destroy_inode(struct inode *inode)
307 kmem_cache_free(sock_inode_cachep,
308 container_of(inode, struct socket_alloc, vfs_inode));
311 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
313 struct socket_alloc *ei = (struct socket_alloc *) foo;
315 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
316 SLAB_CTOR_CONSTRUCTOR)
317 inode_init_once(&ei->vfs_inode);
320 static int init_inodecache(void)
322 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
323 sizeof(struct socket_alloc),
324 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
327 if (sock_inode_cachep == NULL)
332 static struct super_operations sockfs_ops = {
333 .alloc_inode = sock_alloc_inode,
334 .destroy_inode =sock_destroy_inode,
335 .statfs = simple_statfs,
338 static struct super_block *sockfs_get_sb(struct file_system_type *fs_type,
339 int flags, const char *dev_name, void *data)
341 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC);
344 static struct vfsmount *sock_mnt __read_mostly;
346 static struct file_system_type sock_fs_type = {
348 .get_sb = sockfs_get_sb,
349 .kill_sb = kill_anon_super,
351 static int sockfs_delete_dentry(struct dentry *dentry)
355 static struct dentry_operations sockfs_dentry_operations = {
356 .d_delete = sockfs_delete_dentry,
360 * Obtains the first available file descriptor and sets it up for use.
362 * These functions create file structures and maps them to fd space
363 * of the current process. On success it returns file descriptor
364 * and file struct implicitly stored in sock->file.
365 * Note that another thread may close file descriptor before we return
366 * from this function. We use the fact that now we do not refer
367 * to socket after mapping. If one day we will need it, this
368 * function will increment ref. count on file by 1.
370 * In any case returned fd MAY BE not valid!
371 * This race condition is unavoidable
372 * with shared fd spaces, we cannot solve it inside kernel,
373 * but we take care of internal coherence yet.
376 static int sock_alloc_fd(struct file **filep)
380 fd = get_unused_fd();
381 if (likely(fd >= 0)) {
382 struct file *file = get_empty_filp();
385 if (unlikely(!file)) {
394 static int sock_attach_fd(struct socket *sock, struct file *file)
399 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
401 this.hash = SOCK_INODE(sock)->i_ino;
403 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
404 if (unlikely(!file->f_dentry))
407 file->f_dentry->d_op = &sockfs_dentry_operations;
408 d_add(file->f_dentry, SOCK_INODE(sock));
409 file->f_vfsmnt = mntget(sock_mnt);
410 file->f_mapping = file->f_dentry->d_inode->i_mapping;
413 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
414 file->f_mode = FMODE_READ | FMODE_WRITE;
415 file->f_flags = O_RDWR;
417 file->private_data = sock;
422 int sock_map_fd(struct socket *sock)
424 struct file *newfile;
425 int fd = sock_alloc_fd(&newfile);
427 if (likely(fd >= 0)) {
428 int err = sock_attach_fd(sock, newfile);
430 if (unlikely(err < 0)) {
435 fd_install(fd, newfile);
440 static struct socket *sock_from_file(struct file *file, int *err)
445 if (file->f_op == &socket_file_ops)
446 return file->private_data; /* set in sock_map_fd */
448 inode = file->f_dentry->d_inode;
449 if (!S_ISSOCK(inode->i_mode)) {
454 sock = SOCKET_I(inode);
455 if (sock->file != file) {
456 printk(KERN_ERR "socki_lookup: socket file changed!\n");
463 * sockfd_lookup - Go from a file number to its socket slot
465 * @err: pointer to an error code return
467 * The file handle passed in is locked and the socket it is bound
468 * too is returned. If an error occurs the err pointer is overwritten
469 * with a negative errno code and NULL is returned. The function checks
470 * for both invalid handles and passing a handle which is not a socket.
472 * On a success the socket object pointer is returned.
475 struct socket *sockfd_lookup(int fd, int *err)
480 if (!(file = fget(fd))) {
484 sock = sock_from_file(file, err);
490 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
496 file = fget_light(fd, fput_needed);
498 sock = sock_from_file(file, err);
501 fput_light(file, *fput_needed);
507 * sock_alloc - allocate a socket
509 * Allocate a new inode and socket object. The two are bound together
510 * and initialised. The socket is then returned. If we are out of inodes
514 static struct socket *sock_alloc(void)
516 struct inode * inode;
517 struct socket * sock;
519 inode = new_inode(sock_mnt->mnt_sb);
523 sock = SOCKET_I(inode);
525 inode->i_mode = S_IFSOCK|S_IRWXUGO;
526 inode->i_uid = current->fsuid;
527 inode->i_gid = current->fsgid;
529 get_cpu_var(sockets_in_use)++;
530 put_cpu_var(sockets_in_use);
535 * In theory you can't get an open on this inode, but /proc provides
536 * a back door. Remember to keep it shut otherwise you'll let the
537 * creepy crawlies in.
540 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
545 const struct file_operations bad_sock_fops = {
546 .owner = THIS_MODULE,
547 .open = sock_no_open,
551 * sock_release - close a socket
552 * @sock: socket to close
554 * The socket is released from the protocol stack if it has a release
555 * callback, and the inode is then released if the socket is bound to
556 * an inode not a file.
559 void sock_release(struct socket *sock)
562 struct module *owner = sock->ops->owner;
564 sock->ops->release(sock);
569 if (sock->fasync_list)
570 printk(KERN_ERR "sock_release: fasync list not empty!\n");
572 get_cpu_var(sockets_in_use)--;
573 put_cpu_var(sockets_in_use);
575 iput(SOCK_INODE(sock));
581 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
582 struct msghdr *msg, size_t size)
584 struct sock_iocb *si = kiocb_to_siocb(iocb);
592 err = security_socket_sendmsg(sock, msg, size);
596 return sock->ops->sendmsg(iocb, sock, msg, size);
599 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
602 struct sock_iocb siocb;
605 init_sync_kiocb(&iocb, NULL);
606 iocb.private = &siocb;
607 ret = __sock_sendmsg(&iocb, sock, msg, size);
608 if (-EIOCBQUEUED == ret)
609 ret = wait_on_sync_kiocb(&iocb);
613 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
614 struct kvec *vec, size_t num, size_t size)
616 mm_segment_t oldfs = get_fs();
621 * the following is safe, since for compiler definitions of kvec and
622 * iovec are identical, yielding the same in-core layout and alignment
624 msg->msg_iov = (struct iovec *)vec,
625 msg->msg_iovlen = num;
626 result = sock_sendmsg(sock, msg, size);
631 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
632 struct msghdr *msg, size_t size, int flags)
635 struct sock_iocb *si = kiocb_to_siocb(iocb);
643 err = security_socket_recvmsg(sock, msg, size, flags);
647 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
650 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
651 size_t size, int flags)
654 struct sock_iocb siocb;
657 init_sync_kiocb(&iocb, NULL);
658 iocb.private = &siocb;
659 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
660 if (-EIOCBQUEUED == ret)
661 ret = wait_on_sync_kiocb(&iocb);
665 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
666 struct kvec *vec, size_t num,
667 size_t size, int flags)
669 mm_segment_t oldfs = get_fs();
674 * the following is safe, since for compiler definitions of kvec and
675 * iovec are identical, yielding the same in-core layout and alignment
677 msg->msg_iov = (struct iovec *)vec,
678 msg->msg_iovlen = num;
679 result = sock_recvmsg(sock, msg, size, flags);
684 static void sock_aio_dtor(struct kiocb *iocb)
686 kfree(iocb->private);
689 static ssize_t sock_sendpage(struct file *file, struct page *page,
690 int offset, size_t size, loff_t *ppos, int more)
695 sock = file->private_data;
697 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
701 return sock->ops->sendpage(sock, page, offset, size, flags);
704 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
705 char __user *ubuf, size_t size, struct sock_iocb *siocb)
707 if (!is_sync_kiocb(iocb)) {
708 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
711 iocb->ki_dtor = sock_aio_dtor;
715 siocb->async_iov.iov_base = ubuf;
716 siocb->async_iov.iov_len = size;
718 iocb->private = siocb;
722 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
723 struct file *file, struct iovec *iov, unsigned long nr_segs)
725 struct socket *sock = file->private_data;
729 for (i = 0 ; i < nr_segs ; i++)
730 size += iov[i].iov_len;
732 msg->msg_name = NULL;
733 msg->msg_namelen = 0;
734 msg->msg_control = NULL;
735 msg->msg_controllen = 0;
736 msg->msg_iov = (struct iovec *) iov;
737 msg->msg_iovlen = nr_segs;
738 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
740 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
743 static ssize_t sock_readv(struct file *file, const struct iovec *iov,
744 unsigned long nr_segs, loff_t *ppos)
747 struct sock_iocb siocb;
751 init_sync_kiocb(&iocb, NULL);
752 iocb.private = &siocb;
754 ret = do_sock_read(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
755 if (-EIOCBQUEUED == ret)
756 ret = wait_on_sync_kiocb(&iocb);
760 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *ubuf,
761 size_t count, loff_t pos)
763 struct sock_iocb siocb, *x;
767 if (count == 0) /* Match SYS5 behaviour */
770 x = alloc_sock_iocb(iocb, ubuf, count, &siocb);
773 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp,
777 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
778 struct file *file, struct iovec *iov, unsigned long nr_segs)
780 struct socket *sock = file->private_data;
784 for (i = 0 ; i < nr_segs ; i++)
785 size += iov[i].iov_len;
787 msg->msg_name = NULL;
788 msg->msg_namelen = 0;
789 msg->msg_control = NULL;
790 msg->msg_controllen = 0;
791 msg->msg_iov = (struct iovec *) iov;
792 msg->msg_iovlen = nr_segs;
793 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
794 if (sock->type == SOCK_SEQPACKET)
795 msg->msg_flags |= MSG_EOR;
797 return __sock_sendmsg(iocb, sock, msg, size);
800 static ssize_t sock_writev(struct file *file, const struct iovec *iov,
801 unsigned long nr_segs, loff_t *ppos)
805 struct sock_iocb siocb;
808 init_sync_kiocb(&iocb, NULL);
809 iocb.private = &siocb;
811 ret = do_sock_write(&msg, &iocb, file, (struct iovec *)iov, nr_segs);
812 if (-EIOCBQUEUED == ret)
813 ret = wait_on_sync_kiocb(&iocb);
817 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *ubuf,
818 size_t count, loff_t pos)
820 struct sock_iocb siocb, *x;
824 if (count == 0) /* Match SYS5 behaviour */
827 x = alloc_sock_iocb(iocb, (void __user *)ubuf, count, &siocb);
831 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp,
837 * Atomic setting of ioctl hooks to avoid race
838 * with module unload.
841 static DEFINE_MUTEX(br_ioctl_mutex);
842 static int (*br_ioctl_hook)(unsigned int cmd, void __user *arg) = NULL;
844 void brioctl_set(int (*hook)(unsigned int, void __user *))
846 mutex_lock(&br_ioctl_mutex);
847 br_ioctl_hook = hook;
848 mutex_unlock(&br_ioctl_mutex);
850 EXPORT_SYMBOL(brioctl_set);
852 static DEFINE_MUTEX(vlan_ioctl_mutex);
853 static int (*vlan_ioctl_hook)(void __user *arg);
855 void vlan_ioctl_set(int (*hook)(void __user *))
857 mutex_lock(&vlan_ioctl_mutex);
858 vlan_ioctl_hook = hook;
859 mutex_unlock(&vlan_ioctl_mutex);
861 EXPORT_SYMBOL(vlan_ioctl_set);
863 static DEFINE_MUTEX(dlci_ioctl_mutex);
864 static int (*dlci_ioctl_hook)(unsigned int, void __user *);
866 void dlci_ioctl_set(int (*hook)(unsigned int, void __user *))
868 mutex_lock(&dlci_ioctl_mutex);
869 dlci_ioctl_hook = hook;
870 mutex_unlock(&dlci_ioctl_mutex);
872 EXPORT_SYMBOL(dlci_ioctl_set);
875 * With an ioctl, arg may well be a user mode pointer, but we don't know
876 * what to do with it - that's up to the protocol still.
879 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
882 void __user *argp = (void __user *)arg;
885 sock = file->private_data;
886 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
887 err = dev_ioctl(cmd, argp);
889 #ifdef CONFIG_WIRELESS_EXT
890 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
891 err = dev_ioctl(cmd, argp);
893 #endif /* CONFIG_WIRELESS_EXT */
898 if (get_user(pid, (int __user *)argp))
900 err = f_setown(sock->file, pid, 1);
904 err = put_user(sock->file->f_owner.pid, (int __user *)argp);
912 request_module("bridge");
914 mutex_lock(&br_ioctl_mutex);
916 err = br_ioctl_hook(cmd, argp);
917 mutex_unlock(&br_ioctl_mutex);
922 if (!vlan_ioctl_hook)
923 request_module("8021q");
925 mutex_lock(&vlan_ioctl_mutex);
927 err = vlan_ioctl_hook(argp);
928 mutex_unlock(&vlan_ioctl_mutex);
932 /* Convert this to call through a hook */
933 err = divert_ioctl(cmd, argp);
938 if (!dlci_ioctl_hook)
939 request_module("dlci");
941 if (dlci_ioctl_hook) {
942 mutex_lock(&dlci_ioctl_mutex);
943 err = dlci_ioctl_hook(cmd, argp);
944 mutex_unlock(&dlci_ioctl_mutex);
948 err = sock->ops->ioctl(sock, cmd, arg);
951 * If this ioctl is unknown try to hand it down
954 if (err == -ENOIOCTLCMD)
955 err = dev_ioctl(cmd, argp);
961 int sock_create_lite(int family, int type, int protocol, struct socket **res)
964 struct socket *sock = NULL;
966 err = security_socket_create(family, type, protocol, 1);
976 security_socket_post_create(sock, family, type, protocol, 1);
983 /* No kernel lock held - perfect */
984 static unsigned int sock_poll(struct file *file, poll_table * wait)
989 * We can't return errors to poll, so it's either yes or no.
991 sock = file->private_data;
992 return sock->ops->poll(file, sock, wait);
995 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
997 struct socket *sock = file->private_data;
999 return sock->ops->mmap(file, sock, vma);
1002 static int sock_close(struct inode *inode, struct file *filp)
1005 * It was possible the inode is NULL we were
1006 * closing an unfinished socket.
1011 printk(KERN_DEBUG "sock_close: NULL inode\n");
1014 sock_fasync(-1, filp, 0);
1015 sock_release(SOCKET_I(inode));
1020 * Update the socket async list
1022 * Fasync_list locking strategy.
1024 * 1. fasync_list is modified only under process context socket lock
1025 * i.e. under semaphore.
1026 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1027 * or under socket lock.
1028 * 3. fasync_list can be used from softirq context, so that
1029 * modification under socket lock have to be enhanced with
1030 * write_lock_bh(&sk->sk_callback_lock).
1034 static int sock_fasync(int fd, struct file *filp, int on)
1036 struct fasync_struct *fa, *fna=NULL, **prev;
1037 struct socket *sock;
1042 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1047 sock = filp->private_data;
1049 if ((sk=sock->sk) == NULL) {
1056 prev=&(sock->fasync_list);
1058 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1059 if (fa->fa_file==filp)
1066 write_lock_bh(&sk->sk_callback_lock);
1068 write_unlock_bh(&sk->sk_callback_lock);
1075 fna->magic=FASYNC_MAGIC;
1076 fna->fa_next=sock->fasync_list;
1077 write_lock_bh(&sk->sk_callback_lock);
1078 sock->fasync_list=fna;
1079 write_unlock_bh(&sk->sk_callback_lock);
1085 write_lock_bh(&sk->sk_callback_lock);
1087 write_unlock_bh(&sk->sk_callback_lock);
1093 release_sock(sock->sk);
1097 /* This function may be called only under socket lock or callback_lock */
1099 int sock_wake_async(struct socket *sock, int how, int band)
1101 if (!sock || !sock->fasync_list)
1107 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1111 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1116 __kill_fasync(sock->fasync_list, SIGIO, band);
1119 __kill_fasync(sock->fasync_list, SIGURG, band);
1124 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1127 struct socket *sock;
1130 * Check protocol is in range
1132 if (family < 0 || family >= NPROTO)
1133 return -EAFNOSUPPORT;
1134 if (type < 0 || type >= SOCK_MAX)
1139 This uglymoron is moved from INET layer to here to avoid
1140 deadlock in module load.
1142 if (family == PF_INET && type == SOCK_PACKET) {
1146 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1151 err = security_socket_create(family, type, protocol, kern);
1155 #if defined(CONFIG_KMOD)
1156 /* Attempt to load a protocol module if the find failed.
1158 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1159 * requested real, full-featured networking support upon configuration.
1160 * Otherwise module support will break!
1162 if (net_families[family]==NULL)
1164 request_module("net-pf-%d",family);
1168 net_family_read_lock();
1169 if (net_families[family] == NULL) {
1170 err = -EAFNOSUPPORT;
1175 * Allocate the socket and allow the family to set things up. if
1176 * the protocol is 0, the family is instructed to select an appropriate
1180 if (!(sock = sock_alloc())) {
1181 printk(KERN_WARNING "socket: no more sockets\n");
1182 err = -ENFILE; /* Not exactly a match, but its the
1183 closest posix thing */
1190 * We will call the ->create function, that possibly is in a loadable
1191 * module, so we have to bump that loadable module refcnt first.
1193 err = -EAFNOSUPPORT;
1194 if (!try_module_get(net_families[family]->owner))
1197 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1199 goto out_module_put;
1203 * Now to bump the refcnt of the [loadable] module that owns this
1204 * socket at sock_release time we decrement its refcnt.
1206 if (!try_module_get(sock->ops->owner)) {
1208 goto out_module_put;
1211 * Now that we're done with the ->create function, the [loadable]
1212 * module can have its refcnt decremented
1214 module_put(net_families[family]->owner);
1216 security_socket_post_create(sock, family, type, protocol, kern);
1219 net_family_read_unlock();
1222 module_put(net_families[family]->owner);
1228 int sock_create(int family, int type, int protocol, struct socket **res)
1230 return __sock_create(family, type, protocol, res, 0);
1233 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1235 return __sock_create(family, type, protocol, res, 1);
1238 asmlinkage long sys_socket(int family, int type, int protocol)
1241 struct socket *sock;
1243 retval = sock_create(family, type, protocol, &sock);
1247 retval = sock_map_fd(sock);
1252 /* It may be already another descriptor 8) Not kernel problem. */
1261 * Create a pair of connected sockets.
1264 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1266 struct socket *sock1, *sock2;
1270 * Obtain the first socket and check if the underlying protocol
1271 * supports the socketpair call.
1274 err = sock_create(family, type, protocol, &sock1);
1278 err = sock_create(family, type, protocol, &sock2);
1282 err = sock1->ops->socketpair(sock1, sock2);
1284 goto out_release_both;
1288 err = sock_map_fd(sock1);
1290 goto out_release_both;
1293 err = sock_map_fd(sock2);
1298 /* fd1 and fd2 may be already another descriptors.
1299 * Not kernel problem.
1302 err = put_user(fd1, &usockvec[0]);
1304 err = put_user(fd2, &usockvec[1]);
1313 sock_release(sock2);
1318 sock_release(sock2);
1320 sock_release(sock1);
1327 * Bind a name to a socket. Nothing much to do here since it's
1328 * the protocol's responsibility to handle the local address.
1330 * We move the socket address to kernel space before we call
1331 * the protocol layer (having also checked the address is ok).
1334 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1336 struct socket *sock;
1337 char address[MAX_SOCK_ADDR];
1338 int err, fput_needed;
1340 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1342 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1343 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1345 err = sock->ops->bind(sock,
1346 (struct sockaddr *)address, addrlen);
1348 fput_light(sock->file, fput_needed);
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;
1365 int err, fput_needed;
1367 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1368 if ((unsigned) backlog > sysctl_somaxconn)
1369 backlog = sysctl_somaxconn;
1371 err = security_socket_listen(sock, backlog);
1373 err = sock->ops->listen(sock, backlog);
1375 fput_light(sock->file, fput_needed);
1382 * For accept, we attempt to create a new socket, set up the link
1383 * with the client, wake up the client, then return the new
1384 * connected fd. We collect the address of the connector in kernel
1385 * space and move it to user at the very end. This is unclean because
1386 * we open the socket then return an error.
1388 * 1003.1g adds the ability to recvmsg() to query connection pending
1389 * status to recvmsg. We need to add that support in a way thats
1390 * clean when we restucture accept also.
1393 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1395 struct socket *sock, *newsock;
1396 struct file *newfile;
1397 int err, len, newfd, fput_needed;
1398 char address[MAX_SOCK_ADDR];
1400 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1405 if (!(newsock = sock_alloc()))
1408 newsock->type = sock->type;
1409 newsock->ops = sock->ops;
1412 * We don't need try_module_get here, as the listening socket (sock)
1413 * has the protocol module (sock->ops->owner) held.
1415 __module_get(newsock->ops->owner);
1417 newfd = sock_alloc_fd(&newfile);
1418 if (unlikely(newfd < 0)) {
1420 sock_release(newsock);
1424 err = sock_attach_fd(newsock, newfile);
1428 err = security_socket_accept(sock, newsock);
1432 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1436 if (upeer_sockaddr) {
1437 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1438 err = -ECONNABORTED;
1441 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1446 /* File flags are not inherited via accept() unlike another OSes. */
1448 fd_install(newfd, newfile);
1451 security_socket_post_accept(sock, newsock);
1454 fput_light(sock->file, fput_needed);
1459 put_unused_fd(newfd);
1465 * Attempt to connect to a socket with the server address. The address
1466 * is in user space so we verify it is OK and move it to kernel space.
1468 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1471 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1472 * other SEQPACKET protocols that take time to connect() as it doesn't
1473 * include the -EINPROGRESS status for such sockets.
1476 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1478 struct socket *sock;
1479 char address[MAX_SOCK_ADDR];
1480 int err, fput_needed;
1482 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1485 err = move_addr_to_kernel(uservaddr, addrlen, address);
1489 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1493 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1494 sock->file->f_flags);
1496 fput_light(sock->file, fput_needed);
1502 * Get the local address ('name') of a socket object. Move the obtained
1503 * name to user space.
1506 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1508 struct socket *sock;
1509 char address[MAX_SOCK_ADDR];
1510 int len, err, fput_needed;
1512 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1516 err = security_socket_getsockname(sock);
1520 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1523 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1526 fput_light(sock->file, fput_needed);
1532 * Get the remote address ('name') of a socket object. Move the obtained
1533 * name to user space.
1536 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1538 struct socket *sock;
1539 char address[MAX_SOCK_ADDR];
1540 int len, err, fput_needed;
1542 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1543 err = security_socket_getpeername(sock);
1545 fput_light(sock->file, fput_needed);
1549 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1551 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1552 fput_light(sock->file, fput_needed);
1558 * Send a datagram to a given address. We move the address into kernel
1559 * space and check the user space data area is readable before invoking
1563 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1564 struct sockaddr __user *addr, int addr_len)
1566 struct socket *sock;
1567 char address[MAX_SOCK_ADDR];
1572 struct file *sock_file;
1574 sock_file = fget_light(fd, &fput_needed);
1578 sock = sock_from_file(sock_file, &err);
1586 msg.msg_control=NULL;
1587 msg.msg_controllen=0;
1590 err = move_addr_to_kernel(addr, addr_len, address);
1593 msg.msg_name=address;
1594 msg.msg_namelen=addr_len;
1596 if (sock->file->f_flags & O_NONBLOCK)
1597 flags |= MSG_DONTWAIT;
1598 msg.msg_flags = flags;
1599 err = sock_sendmsg(sock, &msg, len);
1602 fput_light(sock_file, fput_needed);
1607 * Send a datagram down a socket.
1610 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1612 return sys_sendto(fd, buff, len, flags, NULL, 0);
1616 * Receive a frame from the socket and optionally record the address of the
1617 * sender. We verify the buffers are writable and if needed move the
1618 * sender address from kernel to user space.
1621 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1622 struct sockaddr __user *addr, int __user *addr_len)
1624 struct socket *sock;
1627 char address[MAX_SOCK_ADDR];
1629 struct file *sock_file;
1632 sock_file = fget_light(fd, &fput_needed);
1636 sock = sock_from_file(sock_file, &err);
1640 msg.msg_control=NULL;
1641 msg.msg_controllen=0;
1646 msg.msg_name=address;
1647 msg.msg_namelen=MAX_SOCK_ADDR;
1648 if (sock->file->f_flags & O_NONBLOCK)
1649 flags |= MSG_DONTWAIT;
1650 err=sock_recvmsg(sock, &msg, size, flags);
1652 if(err >= 0 && addr != NULL)
1654 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1659 fput_light(sock_file, fput_needed);
1664 * Receive a datagram from a socket.
1667 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1669 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1673 * Set a socket option. Because we don't know the option lengths we have
1674 * to pass the user mode parameter for the protocols to sort out.
1677 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1679 int err, fput_needed;
1680 struct socket *sock;
1685 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
1687 err = security_socket_setsockopt(sock,level,optname);
1691 if (level == SOL_SOCKET)
1692 err=sock_setsockopt(sock,level,optname,optval,optlen);
1694 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1696 fput_light(sock->file, fput_needed);
1702 * Get a socket option. Because we don't know the option lengths we have
1703 * to pass a user mode parameter for the protocols to sort out.
1706 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1708 int err, fput_needed;
1709 struct socket *sock;
1711 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1712 err = security_socket_getsockopt(sock, level, optname);
1716 if (level == SOL_SOCKET)
1717 err=sock_getsockopt(sock,level,optname,optval,optlen);
1719 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1721 fput_light(sock->file, fput_needed);
1728 * Shutdown a socket.
1731 asmlinkage long sys_shutdown(int fd, int how)
1733 int err, fput_needed;
1734 struct socket *sock;
1736 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1738 err = security_socket_shutdown(sock, how);
1740 err = sock->ops->shutdown(sock, how);
1741 fput_light(sock->file, fput_needed);
1746 /* A couple of helpful macros for getting the address of the 32/64 bit
1747 * fields which are the same type (int / unsigned) on our platforms.
1749 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1750 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1751 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1755 * BSD sendmsg interface
1758 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1760 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1761 struct socket *sock;
1762 char address[MAX_SOCK_ADDR];
1763 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1764 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1765 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1766 /* 20 is size of ipv6_pktinfo */
1767 unsigned char *ctl_buf = ctl;
1768 struct msghdr msg_sys;
1769 int err, ctl_len, iov_size, total_len;
1773 if (MSG_CMSG_COMPAT & flags) {
1774 if (get_compat_msghdr(&msg_sys, msg_compat))
1776 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1779 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1783 /* do not move before msg_sys is valid */
1785 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1788 /* Check whether to allocate the iovec area*/
1790 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1791 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1792 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1797 /* This will also move the address data into kernel space */
1798 if (MSG_CMSG_COMPAT & flags) {
1799 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1801 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1808 if (msg_sys.msg_controllen > INT_MAX)
1810 ctl_len = msg_sys.msg_controllen;
1811 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1812 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
1815 ctl_buf = msg_sys.msg_control;
1816 ctl_len = msg_sys.msg_controllen;
1817 } else if (ctl_len) {
1818 if (ctl_len > sizeof(ctl))
1820 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1821 if (ctl_buf == NULL)
1826 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1827 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1828 * checking falls down on this.
1830 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1832 msg_sys.msg_control = ctl_buf;
1834 msg_sys.msg_flags = flags;
1836 if (sock->file->f_flags & O_NONBLOCK)
1837 msg_sys.msg_flags |= MSG_DONTWAIT;
1838 err = sock_sendmsg(sock, &msg_sys, total_len);
1842 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1844 if (iov != iovstack)
1845 sock_kfree_s(sock->sk, iov, iov_size);
1847 fput_light(sock->file, fput_needed);
1853 * BSD recvmsg interface
1856 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1858 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1859 struct socket *sock;
1860 struct iovec iovstack[UIO_FASTIOV];
1861 struct iovec *iov=iovstack;
1862 struct msghdr msg_sys;
1863 unsigned long cmsg_ptr;
1864 int err, iov_size, total_len, len;
1867 /* kernel mode address */
1868 char addr[MAX_SOCK_ADDR];
1870 /* user mode address pointers */
1871 struct sockaddr __user *uaddr;
1872 int __user *uaddr_len;
1874 if (MSG_CMSG_COMPAT & flags) {
1875 if (get_compat_msghdr(&msg_sys, msg_compat))
1878 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1881 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1886 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1889 /* Check whether to allocate the iovec area*/
1891 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1892 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1893 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1899 * Save the user-mode address (verify_iovec will change the
1900 * kernel msghdr to use the kernel address space)
1903 uaddr = (void __user *) msg_sys.msg_name;
1904 uaddr_len = COMPAT_NAMELEN(msg);
1905 if (MSG_CMSG_COMPAT & flags) {
1906 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1908 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1913 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1914 msg_sys.msg_flags = 0;
1915 if (MSG_CMSG_COMPAT & flags)
1916 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1918 if (sock->file->f_flags & O_NONBLOCK)
1919 flags |= MSG_DONTWAIT;
1920 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1925 if (uaddr != NULL) {
1926 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1930 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1934 if (MSG_CMSG_COMPAT & flags)
1935 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1936 &msg_compat->msg_controllen);
1938 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1939 &msg->msg_controllen);
1945 if (iov != iovstack)
1946 sock_kfree_s(sock->sk, iov, iov_size);
1948 fput_light(sock->file, fput_needed);
1953 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1955 /* Argument list sizes for sys_socketcall */
1956 #define AL(x) ((x) * sizeof(unsigned long))
1957 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1958 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1959 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1963 * System call vectors.
1965 * Argument checking cleaned up. Saved 20% in size.
1966 * This function doesn't need to set the kernel lock because
1967 * it is set by the callees.
1970 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1973 unsigned long a0,a1;
1976 if(call<1||call>SYS_RECVMSG)
1979 /* copy_from_user should be SMP safe. */
1980 if (copy_from_user(a, args, nargs[call]))
1983 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
1993 err = sys_socket(a0,a1,a[2]);
1996 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
1999 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2002 err = sys_listen(a0,a1);
2005 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2007 case SYS_GETSOCKNAME:
2008 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2010 case SYS_GETPEERNAME:
2011 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2013 case SYS_SOCKETPAIR:
2014 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2017 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2020 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2021 (struct sockaddr __user *)a[4], a[5]);
2024 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2027 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2028 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2031 err = sys_shutdown(a0,a1);
2033 case SYS_SETSOCKOPT:
2034 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2036 case SYS_GETSOCKOPT:
2037 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2040 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2043 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2052 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2055 * This function is called by a protocol handler that wants to
2056 * advertise its address family, and have it linked into the
2060 int sock_register(struct net_proto_family *ops)
2064 if (ops->family >= NPROTO) {
2065 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2068 net_family_write_lock();
2070 if (net_families[ops->family] == NULL) {
2071 net_families[ops->family]=ops;
2074 net_family_write_unlock();
2075 printk(KERN_INFO "NET: Registered protocol family %d\n",
2081 * This function is called by a protocol handler that wants to
2082 * remove its address family, and have it unlinked from the
2086 int sock_unregister(int family)
2088 if (family < 0 || family >= NPROTO)
2091 net_family_write_lock();
2092 net_families[family]=NULL;
2093 net_family_write_unlock();
2094 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2099 static int __init sock_init(void)
2102 * Initialize sock SLAB cache.
2108 * Initialize skbuff SLAB cache
2113 * Initialize the protocols module.
2117 register_filesystem(&sock_fs_type);
2118 sock_mnt = kern_mount(&sock_fs_type);
2120 /* The real protocol initialization is performed in later initcalls.
2123 #ifdef CONFIG_NETFILTER
2130 core_initcall(sock_init); /* early initcall */
2132 #ifdef CONFIG_PROC_FS
2133 void socket_seq_show(struct seq_file *seq)
2138 for_each_possible_cpu(cpu)
2139 counter += per_cpu(sockets_in_use, cpu);
2141 /* It can be negative, by the way. 8) */
2145 seq_printf(seq, "sockets: used %d\n", counter);
2147 #endif /* CONFIG_PROC_FS */
2149 #ifdef CONFIG_COMPAT
2150 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2153 struct socket *sock = file->private_data;
2154 int ret = -ENOIOCTLCMD;
2156 if (sock->ops->compat_ioctl)
2157 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2163 /* ABI emulation layers need these two */
2164 EXPORT_SYMBOL(move_addr_to_kernel);
2165 EXPORT_SYMBOL(move_addr_to_user);
2166 EXPORT_SYMBOL(sock_create);
2167 EXPORT_SYMBOL(sock_create_kern);
2168 EXPORT_SYMBOL(sock_create_lite);
2169 EXPORT_SYMBOL(sock_map_fd);
2170 EXPORT_SYMBOL(sock_recvmsg);
2171 EXPORT_SYMBOL(sock_register);
2172 EXPORT_SYMBOL(sock_release);
2173 EXPORT_SYMBOL(sock_sendmsg);
2174 EXPORT_SYMBOL(sock_unregister);
2175 EXPORT_SYMBOL(sock_wake_async);
2176 EXPORT_SYMBOL(sockfd_lookup);
2177 EXPORT_SYMBOL(kernel_sendmsg);
2178 EXPORT_SYMBOL(kernel_recvmsg);