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/socket.h>
63 #include <linux/file.h>
64 #include <linux/net.h>
65 #include <linux/interrupt.h>
66 #include <linux/rcupdate.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/mount.h>
81 #include <linux/security.h>
82 #include <linux/syscalls.h>
83 #include <linux/compat.h>
84 #include <linux/kmod.h>
85 #include <linux/audit.h>
86 #include <linux/wireless.h>
88 #include <asm/uaccess.h>
89 #include <asm/unistd.h>
91 #include <net/compat.h>
94 #include <linux/netfilter.h>
96 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
97 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
98 unsigned long nr_segs, loff_t pos);
99 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
100 unsigned long nr_segs, loff_t pos);
101 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
103 static int sock_close(struct inode *inode, struct file *file);
104 static unsigned int sock_poll(struct file *file,
105 struct poll_table_struct *wait);
106 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
108 static long compat_sock_ioctl(struct file *file,
109 unsigned int cmd, unsigned long arg);
111 static int sock_fasync(int fd, struct file *filp, int on);
112 static ssize_t sock_sendpage(struct file *file, struct page *page,
113 int offset, size_t size, loff_t *ppos, int more);
116 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
117 * in the operation structures but are done directly via the socketcall() multiplexor.
120 static const struct file_operations socket_file_ops = {
121 .owner = THIS_MODULE,
123 .aio_read = sock_aio_read,
124 .aio_write = sock_aio_write,
126 .unlocked_ioctl = sock_ioctl,
128 .compat_ioctl = compat_sock_ioctl,
131 .open = sock_no_open, /* special open code to disallow open via /proc */
132 .release = sock_close,
133 .fasync = sock_fasync,
134 .sendpage = sock_sendpage,
135 .splice_write = generic_splice_sendpage,
139 * The protocol list. Each protocol is registered in here.
142 static DEFINE_SPINLOCK(net_family_lock);
143 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
146 * Statistics counters of the socket lists
149 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
153 * Move socket addresses back and forth across the kernel/user
154 * divide and look after the messy bits.
157 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
158 16 for IP, 16 for IPX,
161 must be at least one bigger than
162 the AF_UNIX size (see net/unix/af_unix.c
167 * move_addr_to_kernel - copy a socket address into kernel space
168 * @uaddr: Address in user space
169 * @kaddr: Address in kernel space
170 * @ulen: Length in user space
172 * The address is copied into kernel space. If the provided address is
173 * too long an error code of -EINVAL is returned. If the copy gives
174 * invalid addresses -EFAULT is returned. On a success 0 is returned.
177 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
179 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
183 if (copy_from_user(kaddr, uaddr, ulen))
185 return audit_sockaddr(ulen, kaddr);
189 * move_addr_to_user - copy an address to user space
190 * @kaddr: kernel space address
191 * @klen: length of address in kernel
192 * @uaddr: user space address
193 * @ulen: pointer to user length field
195 * The value pointed to by ulen on entry is the buffer length available.
196 * This is overwritten with the buffer space used. -EINVAL is returned
197 * if an overlong buffer is specified or a negative buffer size. -EFAULT
198 * is returned if either the buffer or the length field are not
200 * After copying the data up to the limit the user specifies, the true
201 * length of the data is written over the length limit the user
202 * specified. Zero is returned for a success.
205 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
211 err = get_user(len, ulen);
216 if (len < 0 || len > MAX_SOCK_ADDR)
219 if (audit_sockaddr(klen, kaddr))
221 if (copy_to_user(uaddr, kaddr, len))
225 * "fromlen shall refer to the value before truncation.."
228 return __put_user(klen, ulen);
231 #define SOCKFS_MAGIC 0x534F434B
233 static struct kmem_cache *sock_inode_cachep __read_mostly;
235 static struct inode *sock_alloc_inode(struct super_block *sb)
237 struct socket_alloc *ei;
239 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
242 init_waitqueue_head(&ei->socket.wait);
244 ei->socket.fasync_list = NULL;
245 ei->socket.state = SS_UNCONNECTED;
246 ei->socket.flags = 0;
247 ei->socket.ops = NULL;
248 ei->socket.sk = NULL;
249 ei->socket.file = NULL;
251 return &ei->vfs_inode;
254 static void sock_destroy_inode(struct inode *inode)
256 kmem_cache_free(sock_inode_cachep,
257 container_of(inode, struct socket_alloc, vfs_inode));
260 static void init_once(void *foo, struct kmem_cache *cachep, unsigned long flags)
262 struct socket_alloc *ei = (struct socket_alloc *)foo;
264 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR))
265 == SLAB_CTOR_CONSTRUCTOR)
266 inode_init_once(&ei->vfs_inode);
269 static int init_inodecache(void)
271 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
272 sizeof(struct socket_alloc),
274 (SLAB_HWCACHE_ALIGN |
275 SLAB_RECLAIM_ACCOUNT |
279 if (sock_inode_cachep == NULL)
284 static struct super_operations sockfs_ops = {
285 .alloc_inode = sock_alloc_inode,
286 .destroy_inode =sock_destroy_inode,
287 .statfs = simple_statfs,
290 static int sockfs_get_sb(struct file_system_type *fs_type,
291 int flags, const char *dev_name, void *data,
292 struct vfsmount *mnt)
294 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
298 static struct vfsmount *sock_mnt __read_mostly;
300 static struct file_system_type sock_fs_type = {
302 .get_sb = sockfs_get_sb,
303 .kill_sb = kill_anon_super,
306 static int sockfs_delete_dentry(struct dentry *dentry)
309 * At creation time, we pretended this dentry was hashed
310 * (by clearing DCACHE_UNHASHED bit in d_flags)
311 * At delete time, we restore the truth : not hashed.
312 * (so that dput() can proceed correctly)
314 dentry->d_flags |= DCACHE_UNHASHED;
317 static struct dentry_operations sockfs_dentry_operations = {
318 .d_delete = sockfs_delete_dentry,
322 * Obtains the first available file descriptor and sets it up for use.
324 * These functions create file structures and maps them to fd space
325 * of the current process. On success it returns file descriptor
326 * and file struct implicitly stored in sock->file.
327 * Note that another thread may close file descriptor before we return
328 * from this function. We use the fact that now we do not refer
329 * to socket after mapping. If one day we will need it, this
330 * function will increment ref. count on file by 1.
332 * In any case returned fd MAY BE not valid!
333 * This race condition is unavoidable
334 * with shared fd spaces, we cannot solve it inside kernel,
335 * but we take care of internal coherence yet.
338 static int sock_alloc_fd(struct file **filep)
342 fd = get_unused_fd();
343 if (likely(fd >= 0)) {
344 struct file *file = get_empty_filp();
347 if (unlikely(!file)) {
356 static int sock_attach_fd(struct socket *sock, struct file *file)
361 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
365 file->f_path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
366 if (unlikely(!file->f_path.dentry))
369 file->f_path.dentry->d_op = &sockfs_dentry_operations;
371 * We dont want to push this dentry into global dentry hash table.
372 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
373 * This permits a working /proc/$pid/fd/XXX on sockets
375 file->f_path.dentry->d_flags &= ~DCACHE_UNHASHED;
376 d_instantiate(file->f_path.dentry, SOCK_INODE(sock));
377 file->f_path.mnt = mntget(sock_mnt);
378 file->f_mapping = file->f_path.dentry->d_inode->i_mapping;
381 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
382 file->f_mode = FMODE_READ | FMODE_WRITE;
383 file->f_flags = O_RDWR;
385 file->private_data = sock;
390 int sock_map_fd(struct socket *sock)
392 struct file *newfile;
393 int fd = sock_alloc_fd(&newfile);
395 if (likely(fd >= 0)) {
396 int err = sock_attach_fd(sock, newfile);
398 if (unlikely(err < 0)) {
403 fd_install(fd, newfile);
408 static struct socket *sock_from_file(struct file *file, int *err)
410 if (file->f_op == &socket_file_ops)
411 return file->private_data; /* set in sock_map_fd */
418 * sockfd_lookup - Go from a file number to its socket slot
420 * @err: pointer to an error code return
422 * The file handle passed in is locked and the socket it is bound
423 * too is returned. If an error occurs the err pointer is overwritten
424 * with a negative errno code and NULL is returned. The function checks
425 * for both invalid handles and passing a handle which is not a socket.
427 * On a success the socket object pointer is returned.
430 struct socket *sockfd_lookup(int fd, int *err)
441 sock = sock_from_file(file, err);
447 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
453 file = fget_light(fd, fput_needed);
455 sock = sock_from_file(file, err);
458 fput_light(file, *fput_needed);
464 * sock_alloc - allocate a socket
466 * Allocate a new inode and socket object. The two are bound together
467 * and initialised. The socket is then returned. If we are out of inodes
471 static struct socket *sock_alloc(void)
476 inode = new_inode(sock_mnt->mnt_sb);
480 sock = SOCKET_I(inode);
482 inode->i_mode = S_IFSOCK | S_IRWXUGO;
483 inode->i_uid = current->fsuid;
484 inode->i_gid = current->fsgid;
486 get_cpu_var(sockets_in_use)++;
487 put_cpu_var(sockets_in_use);
492 * In theory you can't get an open on this inode, but /proc provides
493 * a back door. Remember to keep it shut otherwise you'll let the
494 * creepy crawlies in.
497 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
502 const struct file_operations bad_sock_fops = {
503 .owner = THIS_MODULE,
504 .open = sock_no_open,
508 * sock_release - close a socket
509 * @sock: socket to close
511 * The socket is released from the protocol stack if it has a release
512 * callback, and the inode is then released if the socket is bound to
513 * an inode not a file.
516 void sock_release(struct socket *sock)
519 struct module *owner = sock->ops->owner;
521 sock->ops->release(sock);
526 if (sock->fasync_list)
527 printk(KERN_ERR "sock_release: fasync list not empty!\n");
529 get_cpu_var(sockets_in_use)--;
530 put_cpu_var(sockets_in_use);
532 iput(SOCK_INODE(sock));
538 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
539 struct msghdr *msg, size_t size)
541 struct sock_iocb *si = kiocb_to_siocb(iocb);
549 err = security_socket_sendmsg(sock, msg, size);
553 return sock->ops->sendmsg(iocb, sock, msg, size);
556 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
559 struct sock_iocb siocb;
562 init_sync_kiocb(&iocb, NULL);
563 iocb.private = &siocb;
564 ret = __sock_sendmsg(&iocb, sock, msg, size);
565 if (-EIOCBQUEUED == ret)
566 ret = wait_on_sync_kiocb(&iocb);
570 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
571 struct kvec *vec, size_t num, size_t size)
573 mm_segment_t oldfs = get_fs();
578 * the following is safe, since for compiler definitions of kvec and
579 * iovec are identical, yielding the same in-core layout and alignment
581 msg->msg_iov = (struct iovec *)vec;
582 msg->msg_iovlen = num;
583 result = sock_sendmsg(sock, msg, size);
589 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
591 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
594 ktime_t kt = skb->tstamp;
596 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
598 /* Race occurred between timestamp enabling and packet
599 receiving. Fill in the current time for now. */
601 kt = ktime_get_real();
603 tv = ktime_to_timeval(kt);
604 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
607 /* Race occurred between timestamp enabling and packet
608 receiving. Fill in the current time for now. */
610 kt = ktime_get_real();
612 ts = ktime_to_timespec(kt);
613 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
617 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
618 struct msghdr *msg, size_t size, int flags)
621 struct sock_iocb *si = kiocb_to_siocb(iocb);
629 err = security_socket_recvmsg(sock, msg, size, flags);
633 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
636 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
637 size_t size, int flags)
640 struct sock_iocb siocb;
643 init_sync_kiocb(&iocb, NULL);
644 iocb.private = &siocb;
645 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
646 if (-EIOCBQUEUED == ret)
647 ret = wait_on_sync_kiocb(&iocb);
651 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
652 struct kvec *vec, size_t num, size_t size, int flags)
654 mm_segment_t oldfs = get_fs();
659 * the following is safe, since for compiler definitions of kvec and
660 * iovec are identical, yielding the same in-core layout and alignment
662 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
663 result = sock_recvmsg(sock, msg, size, flags);
668 static void sock_aio_dtor(struct kiocb *iocb)
670 kfree(iocb->private);
673 static ssize_t sock_sendpage(struct file *file, struct page *page,
674 int offset, size_t size, loff_t *ppos, int more)
679 sock = file->private_data;
681 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
685 return sock->ops->sendpage(sock, page, offset, size, flags);
688 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
689 struct sock_iocb *siocb)
691 if (!is_sync_kiocb(iocb)) {
692 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
695 iocb->ki_dtor = sock_aio_dtor;
699 iocb->private = siocb;
703 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
704 struct file *file, const struct iovec *iov,
705 unsigned long nr_segs)
707 struct socket *sock = file->private_data;
711 for (i = 0; i < nr_segs; i++)
712 size += iov[i].iov_len;
714 msg->msg_name = NULL;
715 msg->msg_namelen = 0;
716 msg->msg_control = NULL;
717 msg->msg_controllen = 0;
718 msg->msg_iov = (struct iovec *)iov;
719 msg->msg_iovlen = nr_segs;
720 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
722 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
725 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
726 unsigned long nr_segs, loff_t pos)
728 struct sock_iocb siocb, *x;
733 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
737 x = alloc_sock_iocb(iocb, &siocb);
740 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
743 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
744 struct file *file, const struct iovec *iov,
745 unsigned long nr_segs)
747 struct socket *sock = file->private_data;
751 for (i = 0; i < nr_segs; i++)
752 size += iov[i].iov_len;
754 msg->msg_name = NULL;
755 msg->msg_namelen = 0;
756 msg->msg_control = NULL;
757 msg->msg_controllen = 0;
758 msg->msg_iov = (struct iovec *)iov;
759 msg->msg_iovlen = nr_segs;
760 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
761 if (sock->type == SOCK_SEQPACKET)
762 msg->msg_flags |= MSG_EOR;
764 return __sock_sendmsg(iocb, sock, msg, size);
767 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
768 unsigned long nr_segs, loff_t pos)
770 struct sock_iocb siocb, *x;
775 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
778 x = alloc_sock_iocb(iocb, &siocb);
782 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
786 * Atomic setting of ioctl hooks to avoid race
787 * with module unload.
790 static DEFINE_MUTEX(br_ioctl_mutex);
791 static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL;
793 void brioctl_set(int (*hook) (unsigned int, void __user *))
795 mutex_lock(&br_ioctl_mutex);
796 br_ioctl_hook = hook;
797 mutex_unlock(&br_ioctl_mutex);
800 EXPORT_SYMBOL(brioctl_set);
802 static DEFINE_MUTEX(vlan_ioctl_mutex);
803 static int (*vlan_ioctl_hook) (void __user *arg);
805 void vlan_ioctl_set(int (*hook) (void __user *))
807 mutex_lock(&vlan_ioctl_mutex);
808 vlan_ioctl_hook = hook;
809 mutex_unlock(&vlan_ioctl_mutex);
812 EXPORT_SYMBOL(vlan_ioctl_set);
814 static DEFINE_MUTEX(dlci_ioctl_mutex);
815 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
817 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
819 mutex_lock(&dlci_ioctl_mutex);
820 dlci_ioctl_hook = hook;
821 mutex_unlock(&dlci_ioctl_mutex);
824 EXPORT_SYMBOL(dlci_ioctl_set);
827 * With an ioctl, arg may well be a user mode pointer, but we don't know
828 * what to do with it - that's up to the protocol still.
831 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
834 void __user *argp = (void __user *)arg;
837 sock = file->private_data;
838 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
839 err = dev_ioctl(cmd, argp);
841 #ifdef CONFIG_WIRELESS_EXT
842 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
843 err = dev_ioctl(cmd, argp);
845 #endif /* CONFIG_WIRELESS_EXT */
850 if (get_user(pid, (int __user *)argp))
852 err = f_setown(sock->file, pid, 1);
856 err = put_user(f_getown(sock->file),
865 request_module("bridge");
867 mutex_lock(&br_ioctl_mutex);
869 err = br_ioctl_hook(cmd, argp);
870 mutex_unlock(&br_ioctl_mutex);
875 if (!vlan_ioctl_hook)
876 request_module("8021q");
878 mutex_lock(&vlan_ioctl_mutex);
880 err = vlan_ioctl_hook(argp);
881 mutex_unlock(&vlan_ioctl_mutex);
886 if (!dlci_ioctl_hook)
887 request_module("dlci");
889 if (dlci_ioctl_hook) {
890 mutex_lock(&dlci_ioctl_mutex);
891 err = dlci_ioctl_hook(cmd, argp);
892 mutex_unlock(&dlci_ioctl_mutex);
896 err = sock->ops->ioctl(sock, cmd, arg);
899 * If this ioctl is unknown try to hand it down
902 if (err == -ENOIOCTLCMD)
903 err = dev_ioctl(cmd, argp);
909 int sock_create_lite(int family, int type, int protocol, struct socket **res)
912 struct socket *sock = NULL;
914 err = security_socket_create(family, type, protocol, 1);
925 err = security_socket_post_create(sock, family, type, protocol, 1);
938 /* No kernel lock held - perfect */
939 static unsigned int sock_poll(struct file *file, poll_table *wait)
944 * We can't return errors to poll, so it's either yes or no.
946 sock = file->private_data;
947 return sock->ops->poll(file, sock, wait);
950 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
952 struct socket *sock = file->private_data;
954 return sock->ops->mmap(file, sock, vma);
957 static int sock_close(struct inode *inode, struct file *filp)
960 * It was possible the inode is NULL we were
961 * closing an unfinished socket.
965 printk(KERN_DEBUG "sock_close: NULL inode\n");
968 sock_fasync(-1, filp, 0);
969 sock_release(SOCKET_I(inode));
974 * Update the socket async list
976 * Fasync_list locking strategy.
978 * 1. fasync_list is modified only under process context socket lock
979 * i.e. under semaphore.
980 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
981 * or under socket lock.
982 * 3. fasync_list can be used from softirq context, so that
983 * modification under socket lock have to be enhanced with
984 * write_lock_bh(&sk->sk_callback_lock).
988 static int sock_fasync(int fd, struct file *filp, int on)
990 struct fasync_struct *fa, *fna = NULL, **prev;
995 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1000 sock = filp->private_data;
1010 prev = &(sock->fasync_list);
1012 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1013 if (fa->fa_file == filp)
1018 write_lock_bh(&sk->sk_callback_lock);
1020 write_unlock_bh(&sk->sk_callback_lock);
1025 fna->fa_file = filp;
1027 fna->magic = FASYNC_MAGIC;
1028 fna->fa_next = sock->fasync_list;
1029 write_lock_bh(&sk->sk_callback_lock);
1030 sock->fasync_list = fna;
1031 write_unlock_bh(&sk->sk_callback_lock);
1034 write_lock_bh(&sk->sk_callback_lock);
1035 *prev = fa->fa_next;
1036 write_unlock_bh(&sk->sk_callback_lock);
1042 release_sock(sock->sk);
1046 /* This function may be called only under socket lock or callback_lock */
1048 int sock_wake_async(struct socket *sock, int how, int band)
1050 if (!sock || !sock->fasync_list)
1055 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1059 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1064 __kill_fasync(sock->fasync_list, SIGIO, band);
1067 __kill_fasync(sock->fasync_list, SIGURG, band);
1072 static int __sock_create(int family, int type, int protocol,
1073 struct socket **res, int kern)
1076 struct socket *sock;
1077 const struct net_proto_family *pf;
1080 * Check protocol is in range
1082 if (family < 0 || family >= NPROTO)
1083 return -EAFNOSUPPORT;
1084 if (type < 0 || type >= SOCK_MAX)
1089 This uglymoron is moved from INET layer to here to avoid
1090 deadlock in module load.
1092 if (family == PF_INET && type == SOCK_PACKET) {
1096 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1102 err = security_socket_create(family, type, protocol, kern);
1107 * Allocate the socket and allow the family to set things up. if
1108 * the protocol is 0, the family is instructed to select an appropriate
1111 sock = sock_alloc();
1113 if (net_ratelimit())
1114 printk(KERN_WARNING "socket: no more sockets\n");
1115 return -ENFILE; /* Not exactly a match, but its the
1116 closest posix thing */
1121 #if defined(CONFIG_KMOD)
1122 /* Attempt to load a protocol module if the find failed.
1124 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1125 * requested real, full-featured networking support upon configuration.
1126 * Otherwise module support will break!
1128 if (net_families[family] == NULL)
1129 request_module("net-pf-%d", family);
1133 pf = rcu_dereference(net_families[family]);
1134 err = -EAFNOSUPPORT;
1139 * We will call the ->create function, that possibly is in a loadable
1140 * module, so we have to bump that loadable module refcnt first.
1142 if (!try_module_get(pf->owner))
1145 /* Now protected by module ref count */
1148 err = pf->create(sock, protocol);
1150 goto out_module_put;
1153 * Now to bump the refcnt of the [loadable] module that owns this
1154 * socket at sock_release time we decrement its refcnt.
1156 if (!try_module_get(sock->ops->owner))
1157 goto out_module_busy;
1160 * Now that we're done with the ->create function, the [loadable]
1161 * module can have its refcnt decremented
1163 module_put(pf->owner);
1164 err = security_socket_post_create(sock, family, type, protocol, kern);
1172 err = -EAFNOSUPPORT;
1175 module_put(pf->owner);
1182 goto out_sock_release;
1185 int sock_create(int family, int type, int protocol, struct socket **res)
1187 return __sock_create(family, type, protocol, res, 0);
1190 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1192 return __sock_create(family, type, protocol, res, 1);
1195 asmlinkage long sys_socket(int family, int type, int protocol)
1198 struct socket *sock;
1200 retval = sock_create(family, type, protocol, &sock);
1204 retval = sock_map_fd(sock);
1209 /* It may be already another descriptor 8) Not kernel problem. */
1218 * Create a pair of connected sockets.
1221 asmlinkage long sys_socketpair(int family, int type, int protocol,
1222 int __user *usockvec)
1224 struct socket *sock1, *sock2;
1226 struct file *newfile1, *newfile2;
1229 * Obtain the first socket and check if the underlying protocol
1230 * supports the socketpair call.
1233 err = sock_create(family, type, protocol, &sock1);
1237 err = sock_create(family, type, protocol, &sock2);
1241 err = sock1->ops->socketpair(sock1, sock2);
1243 goto out_release_both;
1245 fd1 = sock_alloc_fd(&newfile1);
1246 if (unlikely(fd1 < 0))
1247 goto out_release_both;
1249 fd2 = sock_alloc_fd(&newfile2);
1250 if (unlikely(fd2 < 0)) {
1253 goto out_release_both;
1256 err = sock_attach_fd(sock1, newfile1);
1257 if (unlikely(err < 0)) {
1261 err = sock_attach_fd(sock2, newfile2);
1262 if (unlikely(err < 0)) {
1267 err = audit_fd_pair(fd1, fd2);
1274 fd_install(fd1, newfile1);
1275 fd_install(fd2, newfile2);
1276 /* fd1 and fd2 may be already another descriptors.
1277 * Not kernel problem.
1280 err = put_user(fd1, &usockvec[0]);
1282 err = put_user(fd2, &usockvec[1]);
1291 sock_release(sock2);
1293 sock_release(sock1);
1299 sock_release(sock1);
1302 sock_release(sock2);
1310 * Bind a name to a socket. Nothing much to do here since it's
1311 * the protocol's responsibility to handle the local address.
1313 * We move the socket address to kernel space before we call
1314 * the protocol layer (having also checked the address is ok).
1317 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1319 struct socket *sock;
1320 char address[MAX_SOCK_ADDR];
1321 int err, fput_needed;
1323 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1325 err = move_addr_to_kernel(umyaddr, addrlen, address);
1327 err = security_socket_bind(sock,
1328 (struct sockaddr *)address,
1331 err = sock->ops->bind(sock,
1335 fput_light(sock->file, fput_needed);
1341 * Perform a listen. Basically, we allow the protocol to do anything
1342 * necessary for a listen, and if that works, we mark the socket as
1343 * ready for listening.
1346 int sysctl_somaxconn __read_mostly = SOMAXCONN;
1348 asmlinkage long sys_listen(int fd, int backlog)
1350 struct socket *sock;
1351 int err, fput_needed;
1353 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1355 if ((unsigned)backlog > sysctl_somaxconn)
1356 backlog = sysctl_somaxconn;
1358 err = security_socket_listen(sock, backlog);
1360 err = sock->ops->listen(sock, backlog);
1362 fput_light(sock->file, fput_needed);
1368 * For accept, we attempt to create a new socket, set up the link
1369 * with the client, wake up the client, then return the new
1370 * connected fd. We collect the address of the connector in kernel
1371 * space and move it to user at the very end. This is unclean because
1372 * we open the socket then return an error.
1374 * 1003.1g adds the ability to recvmsg() to query connection pending
1375 * status to recvmsg. We need to add that support in a way thats
1376 * clean when we restucture accept also.
1379 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1380 int __user *upeer_addrlen)
1382 struct socket *sock, *newsock;
1383 struct file *newfile;
1384 int err, len, newfd, fput_needed;
1385 char address[MAX_SOCK_ADDR];
1387 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1392 if (!(newsock = sock_alloc()))
1395 newsock->type = sock->type;
1396 newsock->ops = sock->ops;
1399 * We don't need try_module_get here, as the listening socket (sock)
1400 * has the protocol module (sock->ops->owner) held.
1402 __module_get(newsock->ops->owner);
1404 newfd = sock_alloc_fd(&newfile);
1405 if (unlikely(newfd < 0)) {
1407 sock_release(newsock);
1411 err = sock_attach_fd(newsock, newfile);
1415 err = security_socket_accept(sock, newsock);
1419 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1423 if (upeer_sockaddr) {
1424 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1426 err = -ECONNABORTED;
1429 err = move_addr_to_user(address, len, upeer_sockaddr,
1435 /* File flags are not inherited via accept() unlike another OSes. */
1437 fd_install(newfd, newfile);
1440 security_socket_post_accept(sock, newsock);
1443 fput_light(sock->file, fput_needed);
1447 sock_release(newsock);
1449 put_unused_fd(newfd);
1453 put_unused_fd(newfd);
1458 * Attempt to connect to a socket with the server address. The address
1459 * is in user space so we verify it is OK and move it to kernel space.
1461 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1464 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1465 * other SEQPACKET protocols that take time to connect() as it doesn't
1466 * include the -EINPROGRESS status for such sockets.
1469 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1472 struct socket *sock;
1473 char address[MAX_SOCK_ADDR];
1474 int err, fput_needed;
1476 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1479 err = move_addr_to_kernel(uservaddr, addrlen, address);
1484 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1488 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1489 sock->file->f_flags);
1491 fput_light(sock->file, fput_needed);
1497 * Get the local address ('name') of a socket object. Move the obtained
1498 * name to user space.
1501 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1502 int __user *usockaddr_len)
1504 struct socket *sock;
1505 char address[MAX_SOCK_ADDR];
1506 int len, err, fput_needed;
1508 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1512 err = security_socket_getsockname(sock);
1516 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1519 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1522 fput_light(sock->file, fput_needed);
1528 * Get the remote address ('name') of a socket object. Move the obtained
1529 * name to user space.
1532 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1533 int __user *usockaddr_len)
1535 struct socket *sock;
1536 char address[MAX_SOCK_ADDR];
1537 int len, err, fput_needed;
1539 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1541 err = security_socket_getpeername(sock);
1543 fput_light(sock->file, fput_needed);
1548 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1551 err = move_addr_to_user(address, len, usockaddr,
1553 fput_light(sock->file, fput_needed);
1559 * Send a datagram to a given address. We move the address into kernel
1560 * space and check the user space data area is readable before invoking
1564 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1565 unsigned flags, struct sockaddr __user *addr,
1568 struct socket *sock;
1569 char address[MAX_SOCK_ADDR];
1574 struct file *sock_file;
1576 sock_file = fget_light(fd, &fput_needed);
1581 sock = sock_from_file(sock_file, &err);
1584 iov.iov_base = buff;
1586 msg.msg_name = NULL;
1589 msg.msg_control = NULL;
1590 msg.msg_controllen = 0;
1591 msg.msg_namelen = 0;
1593 err = move_addr_to_kernel(addr, addr_len, address);
1596 msg.msg_name = address;
1597 msg.msg_namelen = addr_len;
1599 if (sock->file->f_flags & O_NONBLOCK)
1600 flags |= MSG_DONTWAIT;
1601 msg.msg_flags = flags;
1602 err = sock_sendmsg(sock, &msg, len);
1605 fput_light(sock_file, fput_needed);
1611 * Send a datagram down a socket.
1614 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1616 return sys_sendto(fd, buff, len, flags, NULL, 0);
1620 * Receive a frame from the socket and optionally record the address of the
1621 * sender. We verify the buffers are writable and if needed move the
1622 * sender address from kernel to user space.
1625 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1626 unsigned flags, struct sockaddr __user *addr,
1627 int __user *addr_len)
1629 struct socket *sock;
1632 char address[MAX_SOCK_ADDR];
1634 struct file *sock_file;
1637 sock_file = fget_light(fd, &fput_needed);
1642 sock = sock_from_file(sock_file, &err);
1646 msg.msg_control = NULL;
1647 msg.msg_controllen = 0;
1651 iov.iov_base = ubuf;
1652 msg.msg_name = address;
1653 msg.msg_namelen = MAX_SOCK_ADDR;
1654 if (sock->file->f_flags & O_NONBLOCK)
1655 flags |= MSG_DONTWAIT;
1656 err = sock_recvmsg(sock, &msg, size, flags);
1658 if (err >= 0 && addr != NULL) {
1659 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1664 fput_light(sock_file, fput_needed);
1670 * Receive a datagram from a socket.
1673 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1676 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1680 * Set a socket option. Because we don't know the option lengths we have
1681 * to pass the user mode parameter for the protocols to sort out.
1684 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1685 char __user *optval, int optlen)
1687 int err, fput_needed;
1688 struct socket *sock;
1693 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1695 err = security_socket_setsockopt(sock, level, optname);
1699 if (level == SOL_SOCKET)
1701 sock_setsockopt(sock, level, optname, optval,
1705 sock->ops->setsockopt(sock, level, optname, optval,
1708 fput_light(sock->file, fput_needed);
1714 * Get a socket option. Because we don't know the option lengths we have
1715 * to pass a user mode parameter for the protocols to sort out.
1718 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1719 char __user *optval, int __user *optlen)
1721 int err, fput_needed;
1722 struct socket *sock;
1724 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1726 err = security_socket_getsockopt(sock, level, optname);
1730 if (level == SOL_SOCKET)
1732 sock_getsockopt(sock, level, optname, optval,
1736 sock->ops->getsockopt(sock, level, optname, optval,
1739 fput_light(sock->file, fput_needed);
1745 * Shutdown a socket.
1748 asmlinkage long sys_shutdown(int fd, int how)
1750 int err, fput_needed;
1751 struct socket *sock;
1753 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1755 err = security_socket_shutdown(sock, how);
1757 err = sock->ops->shutdown(sock, how);
1758 fput_light(sock->file, fput_needed);
1763 /* A couple of helpful macros for getting the address of the 32/64 bit
1764 * fields which are the same type (int / unsigned) on our platforms.
1766 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1767 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1768 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1771 * BSD sendmsg interface
1774 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1776 struct compat_msghdr __user *msg_compat =
1777 (struct compat_msghdr __user *)msg;
1778 struct socket *sock;
1779 char address[MAX_SOCK_ADDR];
1780 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1781 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1782 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1783 /* 20 is size of ipv6_pktinfo */
1784 unsigned char *ctl_buf = ctl;
1785 struct msghdr msg_sys;
1786 int err, ctl_len, iov_size, total_len;
1790 if (MSG_CMSG_COMPAT & flags) {
1791 if (get_compat_msghdr(&msg_sys, msg_compat))
1794 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1797 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1801 /* do not move before msg_sys is valid */
1803 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1806 /* Check whether to allocate the iovec area */
1808 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1809 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1810 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1815 /* This will also move the address data into kernel space */
1816 if (MSG_CMSG_COMPAT & flags) {
1817 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1819 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1826 if (msg_sys.msg_controllen > INT_MAX)
1828 ctl_len = msg_sys.msg_controllen;
1829 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1831 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1835 ctl_buf = msg_sys.msg_control;
1836 ctl_len = msg_sys.msg_controllen;
1837 } else if (ctl_len) {
1838 if (ctl_len > sizeof(ctl)) {
1839 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1840 if (ctl_buf == NULL)
1845 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1846 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1847 * checking falls down on this.
1849 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1852 msg_sys.msg_control = ctl_buf;
1854 msg_sys.msg_flags = flags;
1856 if (sock->file->f_flags & O_NONBLOCK)
1857 msg_sys.msg_flags |= MSG_DONTWAIT;
1858 err = sock_sendmsg(sock, &msg_sys, total_len);
1862 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1864 if (iov != iovstack)
1865 sock_kfree_s(sock->sk, iov, iov_size);
1867 fput_light(sock->file, fput_needed);
1873 * BSD recvmsg interface
1876 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1879 struct compat_msghdr __user *msg_compat =
1880 (struct compat_msghdr __user *)msg;
1881 struct socket *sock;
1882 struct iovec iovstack[UIO_FASTIOV];
1883 struct iovec *iov = iovstack;
1884 struct msghdr msg_sys;
1885 unsigned long cmsg_ptr;
1886 int err, iov_size, total_len, len;
1889 /* kernel mode address */
1890 char addr[MAX_SOCK_ADDR];
1892 /* user mode address pointers */
1893 struct sockaddr __user *uaddr;
1894 int __user *uaddr_len;
1896 if (MSG_CMSG_COMPAT & flags) {
1897 if (get_compat_msghdr(&msg_sys, msg_compat))
1900 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1903 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1908 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1911 /* Check whether to allocate the iovec area */
1913 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1914 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1915 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1921 * Save the user-mode address (verify_iovec will change the
1922 * kernel msghdr to use the kernel address space)
1925 uaddr = (void __user *)msg_sys.msg_name;
1926 uaddr_len = COMPAT_NAMELEN(msg);
1927 if (MSG_CMSG_COMPAT & flags) {
1928 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1930 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1935 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1936 msg_sys.msg_flags = 0;
1937 if (MSG_CMSG_COMPAT & flags)
1938 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1940 if (sock->file->f_flags & O_NONBLOCK)
1941 flags |= MSG_DONTWAIT;
1942 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1947 if (uaddr != NULL) {
1948 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1953 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1957 if (MSG_CMSG_COMPAT & flags)
1958 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1959 &msg_compat->msg_controllen);
1961 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1962 &msg->msg_controllen);
1968 if (iov != iovstack)
1969 sock_kfree_s(sock->sk, iov, iov_size);
1971 fput_light(sock->file, fput_needed);
1976 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1978 /* Argument list sizes for sys_socketcall */
1979 #define AL(x) ((x) * sizeof(unsigned long))
1980 static const unsigned char nargs[18]={
1981 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1982 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1983 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1989 * System call vectors.
1991 * Argument checking cleaned up. Saved 20% in size.
1992 * This function doesn't need to set the kernel lock because
1993 * it is set by the callees.
1996 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1999 unsigned long a0, a1;
2002 if (call < 1 || call > SYS_RECVMSG)
2005 /* copy_from_user should be SMP safe. */
2006 if (copy_from_user(a, args, nargs[call]))
2009 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2018 err = sys_socket(a0, a1, a[2]);
2021 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2024 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2027 err = sys_listen(a0, a1);
2031 sys_accept(a0, (struct sockaddr __user *)a1,
2032 (int __user *)a[2]);
2034 case SYS_GETSOCKNAME:
2036 sys_getsockname(a0, (struct sockaddr __user *)a1,
2037 (int __user *)a[2]);
2039 case SYS_GETPEERNAME:
2041 sys_getpeername(a0, (struct sockaddr __user *)a1,
2042 (int __user *)a[2]);
2044 case SYS_SOCKETPAIR:
2045 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2048 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2051 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2052 (struct sockaddr __user *)a[4], a[5]);
2055 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2058 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2059 (struct sockaddr __user *)a[4],
2060 (int __user *)a[5]);
2063 err = sys_shutdown(a0, a1);
2065 case SYS_SETSOCKOPT:
2066 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2068 case SYS_GETSOCKOPT:
2070 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2071 (int __user *)a[4]);
2074 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2077 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2086 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2089 * sock_register - add a socket protocol handler
2090 * @ops: description of protocol
2092 * This function is called by a protocol handler that wants to
2093 * advertise its address family, and have it linked into the
2094 * socket interface. The value ops->family coresponds to the
2095 * socket system call protocol family.
2097 int sock_register(const struct net_proto_family *ops)
2101 if (ops->family >= NPROTO) {
2102 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2107 spin_lock(&net_family_lock);
2108 if (net_families[ops->family])
2111 net_families[ops->family] = ops;
2114 spin_unlock(&net_family_lock);
2116 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2121 * sock_unregister - remove a protocol handler
2122 * @family: protocol family to remove
2124 * This function is called by a protocol handler that wants to
2125 * remove its address family, and have it unlinked from the
2126 * new socket creation.
2128 * If protocol handler is a module, then it can use module reference
2129 * counts to protect against new references. If protocol handler is not
2130 * a module then it needs to provide its own protection in
2131 * the ops->create routine.
2133 void sock_unregister(int family)
2135 BUG_ON(family < 0 || family >= NPROTO);
2137 spin_lock(&net_family_lock);
2138 net_families[family] = NULL;
2139 spin_unlock(&net_family_lock);
2143 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2146 static int __init sock_init(void)
2149 * Initialize sock SLAB cache.
2155 * Initialize skbuff SLAB cache
2160 * Initialize the protocols module.
2164 register_filesystem(&sock_fs_type);
2165 sock_mnt = kern_mount(&sock_fs_type);
2167 /* The real protocol initialization is performed in later initcalls.
2170 #ifdef CONFIG_NETFILTER
2177 core_initcall(sock_init); /* early initcall */
2179 #ifdef CONFIG_PROC_FS
2180 void socket_seq_show(struct seq_file *seq)
2185 for_each_possible_cpu(cpu)
2186 counter += per_cpu(sockets_in_use, cpu);
2188 /* It can be negative, by the way. 8) */
2192 seq_printf(seq, "sockets: used %d\n", counter);
2194 #endif /* CONFIG_PROC_FS */
2196 #ifdef CONFIG_COMPAT
2197 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2200 struct socket *sock = file->private_data;
2201 int ret = -ENOIOCTLCMD;
2203 if (sock->ops->compat_ioctl)
2204 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2210 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2212 return sock->ops->bind(sock, addr, addrlen);
2215 int kernel_listen(struct socket *sock, int backlog)
2217 return sock->ops->listen(sock, backlog);
2220 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2222 struct sock *sk = sock->sk;
2225 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2230 err = sock->ops->accept(sock, *newsock, flags);
2232 sock_release(*newsock);
2236 (*newsock)->ops = sock->ops;
2242 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2245 return sock->ops->connect(sock, addr, addrlen, flags);
2248 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2251 return sock->ops->getname(sock, addr, addrlen, 0);
2254 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2257 return sock->ops->getname(sock, addr, addrlen, 1);
2260 int kernel_getsockopt(struct socket *sock, int level, int optname,
2261 char *optval, int *optlen)
2263 mm_segment_t oldfs = get_fs();
2267 if (level == SOL_SOCKET)
2268 err = sock_getsockopt(sock, level, optname, optval, optlen);
2270 err = sock->ops->getsockopt(sock, level, optname, optval,
2276 int kernel_setsockopt(struct socket *sock, int level, int optname,
2277 char *optval, int optlen)
2279 mm_segment_t oldfs = get_fs();
2283 if (level == SOL_SOCKET)
2284 err = sock_setsockopt(sock, level, optname, optval, optlen);
2286 err = sock->ops->setsockopt(sock, level, optname, optval,
2292 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2293 size_t size, int flags)
2295 if (sock->ops->sendpage)
2296 return sock->ops->sendpage(sock, page, offset, size, flags);
2298 return sock_no_sendpage(sock, page, offset, size, flags);
2301 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2303 mm_segment_t oldfs = get_fs();
2307 err = sock->ops->ioctl(sock, cmd, arg);
2313 /* ABI emulation layers need these two */
2314 EXPORT_SYMBOL(move_addr_to_kernel);
2315 EXPORT_SYMBOL(move_addr_to_user);
2316 EXPORT_SYMBOL(sock_create);
2317 EXPORT_SYMBOL(sock_create_kern);
2318 EXPORT_SYMBOL(sock_create_lite);
2319 EXPORT_SYMBOL(sock_map_fd);
2320 EXPORT_SYMBOL(sock_recvmsg);
2321 EXPORT_SYMBOL(sock_register);
2322 EXPORT_SYMBOL(sock_release);
2323 EXPORT_SYMBOL(sock_sendmsg);
2324 EXPORT_SYMBOL(sock_unregister);
2325 EXPORT_SYMBOL(sock_wake_async);
2326 EXPORT_SYMBOL(sockfd_lookup);
2327 EXPORT_SYMBOL(kernel_sendmsg);
2328 EXPORT_SYMBOL(kernel_recvmsg);
2329 EXPORT_SYMBOL(kernel_bind);
2330 EXPORT_SYMBOL(kernel_listen);
2331 EXPORT_SYMBOL(kernel_accept);
2332 EXPORT_SYMBOL(kernel_connect);
2333 EXPORT_SYMBOL(kernel_getsockname);
2334 EXPORT_SYMBOL(kernel_getpeername);
2335 EXPORT_SYMBOL(kernel_getsockopt);
2336 EXPORT_SYMBOL(kernel_setsockopt);
2337 EXPORT_SYMBOL(kernel_sendpage);
2338 EXPORT_SYMBOL(kernel_sock_ioctl);