2 * NET An implementation of the SOCKET network access protocol.
4 * Version: @(#)socket.c 1.1.93 18/02/95
6 * Authors: Orest Zborowski, <obz@Kodak.COM>
8 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
11 * Anonymous : NOTSOCK/BADF cleanup. Error fix in
13 * Alan Cox : verify_area() fixes
14 * Alan Cox : Removed DDI
15 * Jonathan Kamens : SOCK_DGRAM reconnect bug
16 * Alan Cox : Moved a load of checks to the very
18 * Alan Cox : Move address structures to/from user
19 * mode above the protocol layers.
20 * Rob Janssen : Allow 0 length sends.
21 * Alan Cox : Asynchronous I/O support (cribbed from the
23 * Niibe Yutaka : Asynchronous I/O for writes (4.4BSD style)
24 * Jeff Uphoff : Made max number of sockets command-line
26 * Matti Aarnio : Made the number of sockets dynamic,
27 * to be allocated when needed, and mr.
28 * Uphoff's max is used as max to be
29 * allowed to allocate.
30 * Linus : Argh. removed all the socket allocation
31 * altogether: it's in the inode now.
32 * Alan Cox : Made sock_alloc()/sock_release() public
33 * for NetROM and future kernel nfsd type
35 * Alan Cox : sendmsg/recvmsg basics.
36 * Tom Dyas : Export net symbols.
37 * Marcin Dalecki : Fixed problems with CONFIG_NET="n".
38 * Alan Cox : Added thread locking to sys_* calls
39 * for sockets. May have errors at the
41 * Kevin Buhr : Fixed the dumb errors in the above.
42 * Andi Kleen : Some small cleanups, optimizations,
43 * and fixed a copy_from_user() bug.
44 * Tigran Aivazian : sys_send(args) calls sys_sendto(args, NULL, 0)
45 * Tigran Aivazian : Made listen(2) backlog sanity checks
46 * protocol-independent
49 * This program is free software; you can redistribute it and/or
50 * modify it under the terms of the GNU General Public License
51 * as published by the Free Software Foundation; either version
52 * 2 of the License, or (at your option) any later version.
55 * This module is effectively the top level interface to the BSD socket
58 * Based upon Swansea University Computer Society NET3.039
62 #include <linux/smp_lock.h>
63 #include <linux/socket.h>
64 #include <linux/file.h>
65 #include <linux/net.h>
66 #include <linux/interrupt.h>
67 #include <linux/netdevice.h>
68 #include <linux/proc_fs.h>
69 #include <linux/seq_file.h>
70 #include <linux/mutex.h>
71 #include <linux/wanrouter.h>
72 #include <linux/if_bridge.h>
73 #include <linux/if_frad.h>
74 #include <linux/if_vlan.h>
75 #include <linux/init.h>
76 #include <linux/poll.h>
77 #include <linux/cache.h>
78 #include <linux/module.h>
79 #include <linux/highmem.h>
80 #include <linux/divert.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
95 #include <linux/netfilter.h>
97 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
98 static ssize_t sock_aio_read(struct kiocb *iocb, char __user *buf,
99 size_t size, loff_t pos);
100 static ssize_t sock_aio_write(struct kiocb *iocb, const char __user *buf,
101 size_t size, loff_t pos);
102 static int sock_mmap(struct file *file, struct vm_area_struct * vma);
104 static int sock_close(struct inode *inode, struct file *file);
105 static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
107 static long sock_ioctl(struct file *file,
108 unsigned int cmd, unsigned long arg);
110 static long compat_sock_ioctl(struct file *file,
111 unsigned int cmd, unsigned long arg);
113 static int sock_fasync(int fd, struct file *filp, int on);
114 static ssize_t sock_readv(struct file *file, const struct iovec *vector,
115 unsigned long count, loff_t *ppos);
116 static ssize_t sock_writev(struct file *file, const struct iovec *vector,
117 unsigned long count, loff_t *ppos);
118 static ssize_t sock_sendpage(struct file *file, struct page *page,
119 int offset, size_t size, loff_t *ppos, int more);
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
126 static struct file_operations socket_file_ops = {
127 .owner = THIS_MODULE,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
132 .unlocked_ioctl = sock_ioctl,
134 .compat_ioctl = compat_sock_ioctl,
137 .open = sock_no_open, /* special open code to disallow open via /proc */
138 .release = sock_close,
139 .fasync = sock_fasync,
141 .writev = sock_writev,
142 .sendpage = sock_sendpage,
143 .splice_write = generic_splice_sendpage,
147 * The protocol list. Each protocol is registered in here.
150 static struct net_proto_family *net_families[NPROTO];
152 #if defined(CONFIG_SMP) || defined(CONFIG_PREEMPT)
153 static atomic_t net_family_lockct = ATOMIC_INIT(0);
154 static DEFINE_SPINLOCK(net_family_lock);
156 /* The strategy is: modifications net_family vector are short, do not
157 sleep and veeery rare, but read access should be free of any exclusive
161 static void net_family_write_lock(void)
163 spin_lock(&net_family_lock);
164 while (atomic_read(&net_family_lockct) != 0) {
165 spin_unlock(&net_family_lock);
169 spin_lock(&net_family_lock);
173 static __inline__ void net_family_write_unlock(void)
175 spin_unlock(&net_family_lock);
178 static __inline__ void net_family_read_lock(void)
180 atomic_inc(&net_family_lockct);
181 spin_unlock_wait(&net_family_lock);
184 static __inline__ void net_family_read_unlock(void)
186 atomic_dec(&net_family_lockct);
190 #define net_family_write_lock() do { } while(0)
191 #define net_family_write_unlock() do { } while(0)
192 #define net_family_read_lock() do { } while(0)
193 #define net_family_read_unlock() do { } while(0)
198 * Statistics counters of the socket lists
201 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
204 * Support routines. Move socket addresses back and forth across the kernel/user
205 * divide and look after the messy bits.
208 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
209 16 for IP, 16 for IPX,
212 must be at least one bigger than
213 the AF_UNIX size (see net/unix/af_unix.c
218 * move_addr_to_kernel - copy a socket address into kernel space
219 * @uaddr: Address in user space
220 * @kaddr: Address in kernel space
221 * @ulen: Length in user space
223 * The address is copied into kernel space. If the provided address is
224 * too long an error code of -EINVAL is returned. If the copy gives
225 * invalid addresses -EFAULT is returned. On a success 0 is returned.
228 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
230 if(ulen<0||ulen>MAX_SOCK_ADDR)
234 if(copy_from_user(kaddr,uaddr,ulen))
236 return audit_sockaddr(ulen, kaddr);
240 * move_addr_to_user - copy an address to user space
241 * @kaddr: kernel space address
242 * @klen: length of address in kernel
243 * @uaddr: user space address
244 * @ulen: pointer to user length field
246 * The value pointed to by ulen on entry is the buffer length available.
247 * This is overwritten with the buffer space used. -EINVAL is returned
248 * if an overlong buffer is specified or a negative buffer size. -EFAULT
249 * is returned if either the buffer or the length field are not
251 * After copying the data up to the limit the user specifies, the true
252 * length of the data is written over the length limit the user
253 * specified. Zero is returned for a success.
256 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr, int __user *ulen)
261 if((err=get_user(len, ulen)))
265 if(len<0 || len> MAX_SOCK_ADDR)
269 if (audit_sockaddr(klen, kaddr))
271 if(copy_to_user(uaddr,kaddr,len))
275 * "fromlen shall refer to the value before truncation.."
278 return __put_user(klen, ulen);
281 #define SOCKFS_MAGIC 0x534F434B
283 static kmem_cache_t * sock_inode_cachep __read_mostly;
285 static struct inode *sock_alloc_inode(struct super_block *sb)
287 struct socket_alloc *ei;
288 ei = (struct socket_alloc *)kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
291 init_waitqueue_head(&ei->socket.wait);
293 ei->socket.fasync_list = NULL;
294 ei->socket.state = SS_UNCONNECTED;
295 ei->socket.flags = 0;
296 ei->socket.ops = NULL;
297 ei->socket.sk = NULL;
298 ei->socket.file = NULL;
299 ei->socket.flags = 0;
301 return &ei->vfs_inode;
304 static void sock_destroy_inode(struct inode *inode)
306 kmem_cache_free(sock_inode_cachep,
307 container_of(inode, struct socket_alloc, vfs_inode));
310 static void init_once(void * foo, kmem_cache_t * cachep, unsigned long flags)
312 struct socket_alloc *ei = (struct socket_alloc *) foo;
314 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
315 SLAB_CTOR_CONSTRUCTOR)
316 inode_init_once(&ei->vfs_inode);
319 static int init_inodecache(void)
321 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
322 sizeof(struct socket_alloc),
323 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
326 if (sock_inode_cachep == NULL)
331 static struct super_operations sockfs_ops = {
332 .alloc_inode = sock_alloc_inode,
333 .destroy_inode =sock_destroy_inode,
334 .statfs = simple_statfs,
337 static int sockfs_get_sb(struct file_system_type *fs_type,
338 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
340 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);
977 err = security_socket_post_create(sock, family, type, protocol, 1);
990 /* No kernel lock held - perfect */
991 static unsigned int sock_poll(struct file *file, poll_table * wait)
996 * We can't return errors to poll, so it's either yes or no.
998 sock = file->private_data;
999 return sock->ops->poll(file, sock, wait);
1002 static int sock_mmap(struct file * file, struct vm_area_struct * vma)
1004 struct socket *sock = file->private_data;
1006 return sock->ops->mmap(file, sock, vma);
1009 static int sock_close(struct inode *inode, struct file *filp)
1012 * It was possible the inode is NULL we were
1013 * closing an unfinished socket.
1018 printk(KERN_DEBUG "sock_close: NULL inode\n");
1021 sock_fasync(-1, filp, 0);
1022 sock_release(SOCKET_I(inode));
1027 * Update the socket async list
1029 * Fasync_list locking strategy.
1031 * 1. fasync_list is modified only under process context socket lock
1032 * i.e. under semaphore.
1033 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1034 * or under socket lock.
1035 * 3. fasync_list can be used from softirq context, so that
1036 * modification under socket lock have to be enhanced with
1037 * write_lock_bh(&sk->sk_callback_lock).
1041 static int sock_fasync(int fd, struct file *filp, int on)
1043 struct fasync_struct *fa, *fna=NULL, **prev;
1044 struct socket *sock;
1049 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1054 sock = filp->private_data;
1056 if ((sk=sock->sk) == NULL) {
1063 prev=&(sock->fasync_list);
1065 for (fa=*prev; fa!=NULL; prev=&fa->fa_next,fa=*prev)
1066 if (fa->fa_file==filp)
1073 write_lock_bh(&sk->sk_callback_lock);
1075 write_unlock_bh(&sk->sk_callback_lock);
1082 fna->magic=FASYNC_MAGIC;
1083 fna->fa_next=sock->fasync_list;
1084 write_lock_bh(&sk->sk_callback_lock);
1085 sock->fasync_list=fna;
1086 write_unlock_bh(&sk->sk_callback_lock);
1092 write_lock_bh(&sk->sk_callback_lock);
1094 write_unlock_bh(&sk->sk_callback_lock);
1100 release_sock(sock->sk);
1104 /* This function may be called only under socket lock or callback_lock */
1106 int sock_wake_async(struct socket *sock, int how, int band)
1108 if (!sock || !sock->fasync_list)
1114 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1118 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1123 __kill_fasync(sock->fasync_list, SIGIO, band);
1126 __kill_fasync(sock->fasync_list, SIGURG, band);
1131 static int __sock_create(int family, int type, int protocol, struct socket **res, int kern)
1134 struct socket *sock;
1137 * Check protocol is in range
1139 if (family < 0 || family >= NPROTO)
1140 return -EAFNOSUPPORT;
1141 if (type < 0 || type >= SOCK_MAX)
1146 This uglymoron is moved from INET layer to here to avoid
1147 deadlock in module load.
1149 if (family == PF_INET && type == SOCK_PACKET) {
1153 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n", current->comm);
1158 err = security_socket_create(family, type, protocol, kern);
1162 #if defined(CONFIG_KMOD)
1163 /* Attempt to load a protocol module if the find failed.
1165 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1166 * requested real, full-featured networking support upon configuration.
1167 * Otherwise module support will break!
1169 if (net_families[family]==NULL)
1171 request_module("net-pf-%d",family);
1175 net_family_read_lock();
1176 if (net_families[family] == NULL) {
1177 err = -EAFNOSUPPORT;
1182 * Allocate the socket and allow the family to set things up. if
1183 * the protocol is 0, the family is instructed to select an appropriate
1187 if (!(sock = sock_alloc())) {
1188 if (net_ratelimit())
1189 printk(KERN_WARNING "socket: no more sockets\n");
1190 err = -ENFILE; /* Not exactly a match, but its the
1191 closest posix thing */
1198 * We will call the ->create function, that possibly is in a loadable
1199 * module, so we have to bump that loadable module refcnt first.
1201 err = -EAFNOSUPPORT;
1202 if (!try_module_get(net_families[family]->owner))
1205 if ((err = net_families[family]->create(sock, protocol)) < 0) {
1207 goto out_module_put;
1211 * Now to bump the refcnt of the [loadable] module that owns this
1212 * socket at sock_release time we decrement its refcnt.
1214 if (!try_module_get(sock->ops->owner)) {
1216 goto out_module_put;
1219 * Now that we're done with the ->create function, the [loadable]
1220 * module can have its refcnt decremented
1222 module_put(net_families[family]->owner);
1224 err = security_socket_post_create(sock, family, type, protocol, kern);
1229 net_family_read_unlock();
1232 module_put(net_families[family]->owner);
1238 int sock_create(int family, int type, int protocol, struct socket **res)
1240 return __sock_create(family, type, protocol, res, 0);
1243 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1245 return __sock_create(family, type, protocol, res, 1);
1248 asmlinkage long sys_socket(int family, int type, int protocol)
1251 struct socket *sock;
1253 retval = sock_create(family, type, protocol, &sock);
1257 retval = sock_map_fd(sock);
1262 /* It may be already another descriptor 8) Not kernel problem. */
1271 * Create a pair of connected sockets.
1274 asmlinkage long sys_socketpair(int family, int type, int protocol, int __user *usockvec)
1276 struct socket *sock1, *sock2;
1280 * Obtain the first socket and check if the underlying protocol
1281 * supports the socketpair call.
1284 err = sock_create(family, type, protocol, &sock1);
1288 err = sock_create(family, type, protocol, &sock2);
1292 err = sock1->ops->socketpair(sock1, sock2);
1294 goto out_release_both;
1298 err = sock_map_fd(sock1);
1300 goto out_release_both;
1303 err = sock_map_fd(sock2);
1308 /* fd1 and fd2 may be already another descriptors.
1309 * Not kernel problem.
1312 err = put_user(fd1, &usockvec[0]);
1314 err = put_user(fd2, &usockvec[1]);
1323 sock_release(sock2);
1328 sock_release(sock2);
1330 sock_release(sock1);
1337 * Bind a name to a socket. Nothing much to do here since it's
1338 * the protocol's responsibility to handle the local address.
1340 * We move the socket address to kernel space before we call
1341 * the protocol layer (having also checked the address is ok).
1344 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1346 struct socket *sock;
1347 char address[MAX_SOCK_ADDR];
1348 int err, fput_needed;
1350 if((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1352 if((err=move_addr_to_kernel(umyaddr,addrlen,address))>=0) {
1353 err = security_socket_bind(sock, (struct sockaddr *)address, addrlen);
1355 err = sock->ops->bind(sock,
1356 (struct sockaddr *)address, addrlen);
1358 fput_light(sock->file, fput_needed);
1365 * Perform a listen. Basically, we allow the protocol to do anything
1366 * necessary for a listen, and if that works, we mark the socket as
1367 * ready for listening.
1370 int sysctl_somaxconn = SOMAXCONN;
1372 asmlinkage long sys_listen(int fd, int backlog)
1374 struct socket *sock;
1375 int err, fput_needed;
1377 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1378 if ((unsigned) backlog > sysctl_somaxconn)
1379 backlog = sysctl_somaxconn;
1381 err = security_socket_listen(sock, backlog);
1383 err = sock->ops->listen(sock, backlog);
1385 fput_light(sock->file, fput_needed);
1392 * For accept, we attempt to create a new socket, set up the link
1393 * with the client, wake up the client, then return the new
1394 * connected fd. We collect the address of the connector in kernel
1395 * space and move it to user at the very end. This is unclean because
1396 * we open the socket then return an error.
1398 * 1003.1g adds the ability to recvmsg() to query connection pending
1399 * status to recvmsg. We need to add that support in a way thats
1400 * clean when we restucture accept also.
1403 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr, int __user *upeer_addrlen)
1405 struct socket *sock, *newsock;
1406 struct file *newfile;
1407 int err, len, newfd, fput_needed;
1408 char address[MAX_SOCK_ADDR];
1410 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1415 if (!(newsock = sock_alloc()))
1418 newsock->type = sock->type;
1419 newsock->ops = sock->ops;
1422 * We don't need try_module_get here, as the listening socket (sock)
1423 * has the protocol module (sock->ops->owner) held.
1425 __module_get(newsock->ops->owner);
1427 newfd = sock_alloc_fd(&newfile);
1428 if (unlikely(newfd < 0)) {
1430 sock_release(newsock);
1434 err = sock_attach_fd(newsock, newfile);
1438 err = security_socket_accept(sock, newsock);
1442 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1446 if (upeer_sockaddr) {
1447 if(newsock->ops->getname(newsock, (struct sockaddr *)address, &len, 2)<0) {
1448 err = -ECONNABORTED;
1451 err = move_addr_to_user(address, len, upeer_sockaddr, upeer_addrlen);
1456 /* File flags are not inherited via accept() unlike another OSes. */
1458 fd_install(newfd, newfile);
1461 security_socket_post_accept(sock, newsock);
1464 fput_light(sock->file, fput_needed);
1469 put_unused_fd(newfd);
1475 * Attempt to connect to a socket with the server address. The address
1476 * is in user space so we verify it is OK and move it to kernel space.
1478 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1481 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1482 * other SEQPACKET protocols that take time to connect() as it doesn't
1483 * include the -EINPROGRESS status for such sockets.
1486 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr, int addrlen)
1488 struct socket *sock;
1489 char address[MAX_SOCK_ADDR];
1490 int err, fput_needed;
1492 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1495 err = move_addr_to_kernel(uservaddr, addrlen, address);
1499 err = security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1503 err = sock->ops->connect(sock, (struct sockaddr *) address, addrlen,
1504 sock->file->f_flags);
1506 fput_light(sock->file, fput_needed);
1512 * Get the local address ('name') of a socket object. Move the obtained
1513 * name to user space.
1516 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1518 struct socket *sock;
1519 char address[MAX_SOCK_ADDR];
1520 int len, err, fput_needed;
1522 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1526 err = security_socket_getsockname(sock);
1530 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1533 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1536 fput_light(sock->file, fput_needed);
1542 * Get the remote address ('name') of a socket object. Move the obtained
1543 * name to user space.
1546 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr, int __user *usockaddr_len)
1548 struct socket *sock;
1549 char address[MAX_SOCK_ADDR];
1550 int len, err, fput_needed;
1552 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1553 err = security_socket_getpeername(sock);
1555 fput_light(sock->file, fput_needed);
1559 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 1);
1561 err=move_addr_to_user(address,len, usockaddr, usockaddr_len);
1562 fput_light(sock->file, fput_needed);
1568 * Send a datagram to a given address. We move the address into kernel
1569 * space and check the user space data area is readable before invoking
1573 asmlinkage long sys_sendto(int fd, void __user * buff, size_t len, unsigned flags,
1574 struct sockaddr __user *addr, int addr_len)
1576 struct socket *sock;
1577 char address[MAX_SOCK_ADDR];
1582 struct file *sock_file;
1584 sock_file = fget_light(fd, &fput_needed);
1588 sock = sock_from_file(sock_file, &err);
1596 msg.msg_control=NULL;
1597 msg.msg_controllen=0;
1600 err = move_addr_to_kernel(addr, addr_len, address);
1603 msg.msg_name=address;
1604 msg.msg_namelen=addr_len;
1606 if (sock->file->f_flags & O_NONBLOCK)
1607 flags |= MSG_DONTWAIT;
1608 msg.msg_flags = flags;
1609 err = sock_sendmsg(sock, &msg, len);
1612 fput_light(sock_file, fput_needed);
1617 * Send a datagram down a socket.
1620 asmlinkage long sys_send(int fd, void __user * buff, size_t len, unsigned flags)
1622 return sys_sendto(fd, buff, len, flags, NULL, 0);
1626 * Receive a frame from the socket and optionally record the address of the
1627 * sender. We verify the buffers are writable and if needed move the
1628 * sender address from kernel to user space.
1631 asmlinkage long sys_recvfrom(int fd, void __user * ubuf, size_t size, unsigned flags,
1632 struct sockaddr __user *addr, int __user *addr_len)
1634 struct socket *sock;
1637 char address[MAX_SOCK_ADDR];
1639 struct file *sock_file;
1642 sock_file = fget_light(fd, &fput_needed);
1646 sock = sock_from_file(sock_file, &err);
1650 msg.msg_control=NULL;
1651 msg.msg_controllen=0;
1656 msg.msg_name=address;
1657 msg.msg_namelen=MAX_SOCK_ADDR;
1658 if (sock->file->f_flags & O_NONBLOCK)
1659 flags |= MSG_DONTWAIT;
1660 err=sock_recvmsg(sock, &msg, size, flags);
1662 if(err >= 0 && addr != NULL)
1664 err2=move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1669 fput_light(sock_file, fput_needed);
1674 * Receive a datagram from a socket.
1677 asmlinkage long sys_recv(int fd, void __user * ubuf, size_t size, unsigned flags)
1679 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1683 * Set a socket option. Because we don't know the option lengths we have
1684 * to pass the user mode parameter for the protocols to sort out.
1687 asmlinkage long sys_setsockopt(int fd, int level, int optname, char __user *optval, int optlen)
1689 int err, fput_needed;
1690 struct socket *sock;
1695 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL)
1697 err = security_socket_setsockopt(sock,level,optname);
1701 if (level == SOL_SOCKET)
1702 err=sock_setsockopt(sock,level,optname,optval,optlen);
1704 err=sock->ops->setsockopt(sock, level, optname, optval, optlen);
1706 fput_light(sock->file, fput_needed);
1712 * Get a socket option. Because we don't know the option lengths we have
1713 * to pass a user mode parameter for the protocols to sort out.
1716 asmlinkage long sys_getsockopt(int fd, int level, int optname, char __user *optval, int __user *optlen)
1718 int err, fput_needed;
1719 struct socket *sock;
1721 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed)) != NULL) {
1722 err = security_socket_getsockopt(sock, level, optname);
1726 if (level == SOL_SOCKET)
1727 err=sock_getsockopt(sock,level,optname,optval,optlen);
1729 err=sock->ops->getsockopt(sock, level, optname, optval, optlen);
1731 fput_light(sock->file, fput_needed);
1738 * Shutdown a socket.
1741 asmlinkage long sys_shutdown(int fd, int how)
1743 int err, fput_needed;
1744 struct socket *sock;
1746 if ((sock = sockfd_lookup_light(fd, &err, &fput_needed))!=NULL)
1748 err = security_socket_shutdown(sock, how);
1750 err = sock->ops->shutdown(sock, how);
1751 fput_light(sock->file, fput_needed);
1756 /* A couple of helpful macros for getting the address of the 32/64 bit
1757 * fields which are the same type (int / unsigned) on our platforms.
1759 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1760 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1761 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1765 * BSD sendmsg interface
1768 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1770 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1771 struct socket *sock;
1772 char address[MAX_SOCK_ADDR];
1773 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1774 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1775 __attribute__ ((aligned (sizeof(__kernel_size_t))));
1776 /* 20 is size of ipv6_pktinfo */
1777 unsigned char *ctl_buf = ctl;
1778 struct msghdr msg_sys;
1779 int err, ctl_len, iov_size, total_len;
1783 if (MSG_CMSG_COMPAT & flags) {
1784 if (get_compat_msghdr(&msg_sys, msg_compat))
1786 } else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1789 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1793 /* do not move before msg_sys is valid */
1795 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1798 /* Check whether to allocate the iovec area*/
1800 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1801 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1802 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1807 /* This will also move the address data into kernel space */
1808 if (MSG_CMSG_COMPAT & flags) {
1809 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1811 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1818 if (msg_sys.msg_controllen > INT_MAX)
1820 ctl_len = msg_sys.msg_controllen;
1821 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1822 err = cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl, sizeof(ctl));
1825 ctl_buf = msg_sys.msg_control;
1826 ctl_len = msg_sys.msg_controllen;
1827 } else if (ctl_len) {
1828 if (ctl_len > sizeof(ctl))
1830 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1831 if (ctl_buf == NULL)
1836 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1837 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1838 * checking falls down on this.
1840 if (copy_from_user(ctl_buf, (void __user *) msg_sys.msg_control, ctl_len))
1842 msg_sys.msg_control = ctl_buf;
1844 msg_sys.msg_flags = flags;
1846 if (sock->file->f_flags & O_NONBLOCK)
1847 msg_sys.msg_flags |= MSG_DONTWAIT;
1848 err = sock_sendmsg(sock, &msg_sys, total_len);
1852 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1854 if (iov != iovstack)
1855 sock_kfree_s(sock->sk, iov, iov_size);
1857 fput_light(sock->file, fput_needed);
1863 * BSD recvmsg interface
1866 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg, unsigned int flags)
1868 struct compat_msghdr __user *msg_compat = (struct compat_msghdr __user *)msg;
1869 struct socket *sock;
1870 struct iovec iovstack[UIO_FASTIOV];
1871 struct iovec *iov=iovstack;
1872 struct msghdr msg_sys;
1873 unsigned long cmsg_ptr;
1874 int err, iov_size, total_len, len;
1877 /* kernel mode address */
1878 char addr[MAX_SOCK_ADDR];
1880 /* user mode address pointers */
1881 struct sockaddr __user *uaddr;
1882 int __user *uaddr_len;
1884 if (MSG_CMSG_COMPAT & flags) {
1885 if (get_compat_msghdr(&msg_sys, msg_compat))
1888 if (copy_from_user(&msg_sys,msg,sizeof(struct msghdr)))
1891 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1896 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1899 /* Check whether to allocate the iovec area*/
1901 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1902 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1903 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1909 * Save the user-mode address (verify_iovec will change the
1910 * kernel msghdr to use the kernel address space)
1913 uaddr = (void __user *) msg_sys.msg_name;
1914 uaddr_len = COMPAT_NAMELEN(msg);
1915 if (MSG_CMSG_COMPAT & flags) {
1916 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1918 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1923 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1924 msg_sys.msg_flags = 0;
1925 if (MSG_CMSG_COMPAT & flags)
1926 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1928 if (sock->file->f_flags & O_NONBLOCK)
1929 flags |= MSG_DONTWAIT;
1930 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1935 if (uaddr != NULL) {
1936 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr, uaddr_len);
1940 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1944 if (MSG_CMSG_COMPAT & flags)
1945 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1946 &msg_compat->msg_controllen);
1948 err = __put_user((unsigned long)msg_sys.msg_control-cmsg_ptr,
1949 &msg->msg_controllen);
1955 if (iov != iovstack)
1956 sock_kfree_s(sock->sk, iov, iov_size);
1958 fput_light(sock->file, fput_needed);
1963 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1965 /* Argument list sizes for sys_socketcall */
1966 #define AL(x) ((x) * sizeof(unsigned long))
1967 static unsigned char nargs[18]={AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1968 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1969 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)};
1973 * System call vectors.
1975 * Argument checking cleaned up. Saved 20% in size.
1976 * This function doesn't need to set the kernel lock because
1977 * it is set by the callees.
1980 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1983 unsigned long a0,a1;
1986 if(call<1||call>SYS_RECVMSG)
1989 /* copy_from_user should be SMP safe. */
1990 if (copy_from_user(a, args, nargs[call]))
1993 err = audit_socketcall(nargs[call]/sizeof(unsigned long), a);
2003 err = sys_socket(a0,a1,a[2]);
2006 err = sys_bind(a0,(struct sockaddr __user *)a1, a[2]);
2009 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2012 err = sys_listen(a0,a1);
2015 err = sys_accept(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2017 case SYS_GETSOCKNAME:
2018 err = sys_getsockname(a0,(struct sockaddr __user *)a1, (int __user *)a[2]);
2020 case SYS_GETPEERNAME:
2021 err = sys_getpeername(a0, (struct sockaddr __user *)a1, (int __user *)a[2]);
2023 case SYS_SOCKETPAIR:
2024 err = sys_socketpair(a0,a1, a[2], (int __user *)a[3]);
2027 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2030 err = sys_sendto(a0,(void __user *)a1, a[2], a[3],
2031 (struct sockaddr __user *)a[4], a[5]);
2034 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2037 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2038 (struct sockaddr __user *)a[4], (int __user *)a[5]);
2041 err = sys_shutdown(a0,a1);
2043 case SYS_SETSOCKOPT:
2044 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2046 case SYS_GETSOCKOPT:
2047 err = sys_getsockopt(a0, a1, a[2], (char __user *)a[3], (int __user *)a[4]);
2050 err = sys_sendmsg(a0, (struct msghdr __user *) a1, a[2]);
2053 err = sys_recvmsg(a0, (struct msghdr __user *) a1, a[2]);
2062 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2065 * This function is called by a protocol handler that wants to
2066 * advertise its address family, and have it linked into the
2070 int sock_register(struct net_proto_family *ops)
2074 if (ops->family >= NPROTO) {
2075 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family, NPROTO);
2078 net_family_write_lock();
2080 if (net_families[ops->family] == NULL) {
2081 net_families[ops->family]=ops;
2084 net_family_write_unlock();
2085 printk(KERN_INFO "NET: Registered protocol family %d\n",
2091 * This function is called by a protocol handler that wants to
2092 * remove its address family, and have it unlinked from the
2096 int sock_unregister(int family)
2098 if (family < 0 || family >= NPROTO)
2101 net_family_write_lock();
2102 net_families[family]=NULL;
2103 net_family_write_unlock();
2104 printk(KERN_INFO "NET: Unregistered protocol family %d\n",
2109 static int __init sock_init(void)
2112 * Initialize sock SLAB cache.
2118 * Initialize skbuff SLAB cache
2123 * Initialize the protocols module.
2127 register_filesystem(&sock_fs_type);
2128 sock_mnt = kern_mount(&sock_fs_type);
2130 /* The real protocol initialization is performed in later initcalls.
2133 #ifdef CONFIG_NETFILTER
2140 core_initcall(sock_init); /* early initcall */
2142 #ifdef CONFIG_PROC_FS
2143 void socket_seq_show(struct seq_file *seq)
2148 for_each_possible_cpu(cpu)
2149 counter += per_cpu(sockets_in_use, cpu);
2151 /* It can be negative, by the way. 8) */
2155 seq_printf(seq, "sockets: used %d\n", counter);
2157 #endif /* CONFIG_PROC_FS */
2159 #ifdef CONFIG_COMPAT
2160 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2163 struct socket *sock = file->private_data;
2164 int ret = -ENOIOCTLCMD;
2166 if (sock->ops->compat_ioctl)
2167 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2173 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2175 return sock->ops->bind(sock, addr, addrlen);
2178 int kernel_listen(struct socket *sock, int backlog)
2180 return sock->ops->listen(sock, backlog);
2183 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2185 struct sock *sk = sock->sk;
2188 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2193 err = sock->ops->accept(sock, *newsock, flags);
2195 sock_release(*newsock);
2199 (*newsock)->ops = sock->ops;
2205 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2208 return sock->ops->connect(sock, addr, addrlen, flags);
2211 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2214 return sock->ops->getname(sock, addr, addrlen, 0);
2217 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2220 return sock->ops->getname(sock, addr, addrlen, 1);
2223 int kernel_getsockopt(struct socket *sock, int level, int optname,
2224 char *optval, int *optlen)
2226 mm_segment_t oldfs = get_fs();
2230 if (level == SOL_SOCKET)
2231 err = sock_getsockopt(sock, level, optname, optval, optlen);
2233 err = sock->ops->getsockopt(sock, level, optname, optval,
2239 int kernel_setsockopt(struct socket *sock, int level, int optname,
2240 char *optval, int optlen)
2242 mm_segment_t oldfs = get_fs();
2246 if (level == SOL_SOCKET)
2247 err = sock_setsockopt(sock, level, optname, optval, optlen);
2249 err = sock->ops->setsockopt(sock, level, optname, optval,
2255 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2256 size_t size, int flags)
2258 if (sock->ops->sendpage)
2259 return sock->ops->sendpage(sock, page, offset, size, flags);
2261 return sock_no_sendpage(sock, page, offset, size, flags);
2264 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2266 mm_segment_t oldfs = get_fs();
2270 err = sock->ops->ioctl(sock, cmd, arg);
2276 /* ABI emulation layers need these two */
2277 EXPORT_SYMBOL(move_addr_to_kernel);
2278 EXPORT_SYMBOL(move_addr_to_user);
2279 EXPORT_SYMBOL(sock_create);
2280 EXPORT_SYMBOL(sock_create_kern);
2281 EXPORT_SYMBOL(sock_create_lite);
2282 EXPORT_SYMBOL(sock_map_fd);
2283 EXPORT_SYMBOL(sock_recvmsg);
2284 EXPORT_SYMBOL(sock_register);
2285 EXPORT_SYMBOL(sock_release);
2286 EXPORT_SYMBOL(sock_sendmsg);
2287 EXPORT_SYMBOL(sock_unregister);
2288 EXPORT_SYMBOL(sock_wake_async);
2289 EXPORT_SYMBOL(sockfd_lookup);
2290 EXPORT_SYMBOL(kernel_sendmsg);
2291 EXPORT_SYMBOL(kernel_recvmsg);
2292 EXPORT_SYMBOL(kernel_bind);
2293 EXPORT_SYMBOL(kernel_listen);
2294 EXPORT_SYMBOL(kernel_accept);
2295 EXPORT_SYMBOL(kernel_connect);
2296 EXPORT_SYMBOL(kernel_getsockname);
2297 EXPORT_SYMBOL(kernel_getpeername);
2298 EXPORT_SYMBOL(kernel_getsockopt);
2299 EXPORT_SYMBOL(kernel_setsockopt);
2300 EXPORT_SYMBOL(kernel_sendpage);
2301 EXPORT_SYMBOL(kernel_sock_ioctl);