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_CONSTRUCTOR)
265 inode_init_once(&ei->vfs_inode);
268 static int init_inodecache(void)
270 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
271 sizeof(struct socket_alloc),
273 (SLAB_HWCACHE_ALIGN |
274 SLAB_RECLAIM_ACCOUNT |
278 if (sock_inode_cachep == NULL)
283 static struct super_operations sockfs_ops = {
284 .alloc_inode = sock_alloc_inode,
285 .destroy_inode =sock_destroy_inode,
286 .statfs = simple_statfs,
289 static int sockfs_get_sb(struct file_system_type *fs_type,
290 int flags, const char *dev_name, void *data,
291 struct vfsmount *mnt)
293 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
297 static struct vfsmount *sock_mnt __read_mostly;
299 static struct file_system_type sock_fs_type = {
301 .get_sb = sockfs_get_sb,
302 .kill_sb = kill_anon_super,
305 static int sockfs_delete_dentry(struct dentry *dentry)
308 * At creation time, we pretended this dentry was hashed
309 * (by clearing DCACHE_UNHASHED bit in d_flags)
310 * At delete time, we restore the truth : not hashed.
311 * (so that dput() can proceed correctly)
313 dentry->d_flags |= DCACHE_UNHASHED;
318 * sockfs_dname() is called from d_path().
320 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
322 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
323 dentry->d_inode->i_ino);
326 static struct dentry_operations sockfs_dentry_operations = {
327 .d_delete = sockfs_delete_dentry,
328 .d_dname = sockfs_dname,
332 * Obtains the first available file descriptor and sets it up for use.
334 * These functions create file structures and maps them to fd space
335 * of the current process. On success it returns file descriptor
336 * and file struct implicitly stored in sock->file.
337 * Note that another thread may close file descriptor before we return
338 * from this function. We use the fact that now we do not refer
339 * to socket after mapping. If one day we will need it, this
340 * function will increment ref. count on file by 1.
342 * In any case returned fd MAY BE not valid!
343 * This race condition is unavoidable
344 * with shared fd spaces, we cannot solve it inside kernel,
345 * but we take care of internal coherence yet.
348 static int sock_alloc_fd(struct file **filep)
352 fd = get_unused_fd();
353 if (likely(fd >= 0)) {
354 struct file *file = get_empty_filp();
357 if (unlikely(!file)) {
366 static int sock_attach_fd(struct socket *sock, struct file *file)
368 struct qstr name = { .name = "" };
370 file->f_path.dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
371 if (unlikely(!file->f_path.dentry))
374 file->f_path.dentry->d_op = &sockfs_dentry_operations;
376 * We dont want to push this dentry into global dentry hash table.
377 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
378 * This permits a working /proc/$pid/fd/XXX on sockets
380 file->f_path.dentry->d_flags &= ~DCACHE_UNHASHED;
381 d_instantiate(file->f_path.dentry, SOCK_INODE(sock));
382 file->f_path.mnt = mntget(sock_mnt);
383 file->f_mapping = file->f_path.dentry->d_inode->i_mapping;
386 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
387 file->f_mode = FMODE_READ | FMODE_WRITE;
388 file->f_flags = O_RDWR;
390 file->private_data = sock;
395 int sock_map_fd(struct socket *sock)
397 struct file *newfile;
398 int fd = sock_alloc_fd(&newfile);
400 if (likely(fd >= 0)) {
401 int err = sock_attach_fd(sock, newfile);
403 if (unlikely(err < 0)) {
408 fd_install(fd, newfile);
413 static struct socket *sock_from_file(struct file *file, int *err)
415 if (file->f_op == &socket_file_ops)
416 return file->private_data; /* set in sock_map_fd */
423 * sockfd_lookup - Go from a file number to its socket slot
425 * @err: pointer to an error code return
427 * The file handle passed in is locked and the socket it is bound
428 * too is returned. If an error occurs the err pointer is overwritten
429 * with a negative errno code and NULL is returned. The function checks
430 * for both invalid handles and passing a handle which is not a socket.
432 * On a success the socket object pointer is returned.
435 struct socket *sockfd_lookup(int fd, int *err)
446 sock = sock_from_file(file, err);
452 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
458 file = fget_light(fd, fput_needed);
460 sock = sock_from_file(file, err);
463 fput_light(file, *fput_needed);
469 * sock_alloc - allocate a socket
471 * Allocate a new inode and socket object. The two are bound together
472 * and initialised. The socket is then returned. If we are out of inodes
476 static struct socket *sock_alloc(void)
481 inode = new_inode(sock_mnt->mnt_sb);
485 sock = SOCKET_I(inode);
487 inode->i_mode = S_IFSOCK | S_IRWXUGO;
488 inode->i_uid = current->fsuid;
489 inode->i_gid = current->fsgid;
491 get_cpu_var(sockets_in_use)++;
492 put_cpu_var(sockets_in_use);
497 * In theory you can't get an open on this inode, but /proc provides
498 * a back door. Remember to keep it shut otherwise you'll let the
499 * creepy crawlies in.
502 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
507 const struct file_operations bad_sock_fops = {
508 .owner = THIS_MODULE,
509 .open = sock_no_open,
513 * sock_release - close a socket
514 * @sock: socket to close
516 * The socket is released from the protocol stack if it has a release
517 * callback, and the inode is then released if the socket is bound to
518 * an inode not a file.
521 void sock_release(struct socket *sock)
524 struct module *owner = sock->ops->owner;
526 sock->ops->release(sock);
531 if (sock->fasync_list)
532 printk(KERN_ERR "sock_release: fasync list not empty!\n");
534 get_cpu_var(sockets_in_use)--;
535 put_cpu_var(sockets_in_use);
537 iput(SOCK_INODE(sock));
543 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
544 struct msghdr *msg, size_t size)
546 struct sock_iocb *si = kiocb_to_siocb(iocb);
554 err = security_socket_sendmsg(sock, msg, size);
558 return sock->ops->sendmsg(iocb, sock, msg, size);
561 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
564 struct sock_iocb siocb;
567 init_sync_kiocb(&iocb, NULL);
568 iocb.private = &siocb;
569 ret = __sock_sendmsg(&iocb, sock, msg, size);
570 if (-EIOCBQUEUED == ret)
571 ret = wait_on_sync_kiocb(&iocb);
575 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
576 struct kvec *vec, size_t num, size_t size)
578 mm_segment_t oldfs = get_fs();
583 * the following is safe, since for compiler definitions of kvec and
584 * iovec are identical, yielding the same in-core layout and alignment
586 msg->msg_iov = (struct iovec *)vec;
587 msg->msg_iovlen = num;
588 result = sock_sendmsg(sock, msg, size);
594 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
596 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
599 ktime_t kt = skb->tstamp;
601 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
603 /* Race occurred between timestamp enabling and packet
604 receiving. Fill in the current time for now. */
606 kt = ktime_get_real();
608 tv = ktime_to_timeval(kt);
609 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
612 /* Race occurred between timestamp enabling and packet
613 receiving. Fill in the current time for now. */
615 kt = ktime_get_real();
617 ts = ktime_to_timespec(kt);
618 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
622 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
624 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
625 struct msghdr *msg, size_t size, int flags)
628 struct sock_iocb *si = kiocb_to_siocb(iocb);
636 err = security_socket_recvmsg(sock, msg, size, flags);
640 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
643 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
644 size_t size, int flags)
647 struct sock_iocb siocb;
650 init_sync_kiocb(&iocb, NULL);
651 iocb.private = &siocb;
652 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
653 if (-EIOCBQUEUED == ret)
654 ret = wait_on_sync_kiocb(&iocb);
658 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
659 struct kvec *vec, size_t num, size_t size, int flags)
661 mm_segment_t oldfs = get_fs();
666 * the following is safe, since for compiler definitions of kvec and
667 * iovec are identical, yielding the same in-core layout and alignment
669 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
670 result = sock_recvmsg(sock, msg, size, flags);
675 static void sock_aio_dtor(struct kiocb *iocb)
677 kfree(iocb->private);
680 static ssize_t sock_sendpage(struct file *file, struct page *page,
681 int offset, size_t size, loff_t *ppos, int more)
686 sock = file->private_data;
688 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
692 return sock->ops->sendpage(sock, page, offset, size, flags);
695 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
696 struct sock_iocb *siocb)
698 if (!is_sync_kiocb(iocb)) {
699 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
702 iocb->ki_dtor = sock_aio_dtor;
706 iocb->private = siocb;
710 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
711 struct file *file, const struct iovec *iov,
712 unsigned long nr_segs)
714 struct socket *sock = file->private_data;
718 for (i = 0; i < nr_segs; i++)
719 size += iov[i].iov_len;
721 msg->msg_name = NULL;
722 msg->msg_namelen = 0;
723 msg->msg_control = NULL;
724 msg->msg_controllen = 0;
725 msg->msg_iov = (struct iovec *)iov;
726 msg->msg_iovlen = nr_segs;
727 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
729 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
732 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
733 unsigned long nr_segs, loff_t pos)
735 struct sock_iocb siocb, *x;
740 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
744 x = alloc_sock_iocb(iocb, &siocb);
747 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
750 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
751 struct file *file, const struct iovec *iov,
752 unsigned long nr_segs)
754 struct socket *sock = file->private_data;
758 for (i = 0; i < nr_segs; i++)
759 size += iov[i].iov_len;
761 msg->msg_name = NULL;
762 msg->msg_namelen = 0;
763 msg->msg_control = NULL;
764 msg->msg_controllen = 0;
765 msg->msg_iov = (struct iovec *)iov;
766 msg->msg_iovlen = nr_segs;
767 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
768 if (sock->type == SOCK_SEQPACKET)
769 msg->msg_flags |= MSG_EOR;
771 return __sock_sendmsg(iocb, sock, msg, size);
774 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
775 unsigned long nr_segs, loff_t pos)
777 struct sock_iocb siocb, *x;
782 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
785 x = alloc_sock_iocb(iocb, &siocb);
789 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
793 * Atomic setting of ioctl hooks to avoid race
794 * with module unload.
797 static DEFINE_MUTEX(br_ioctl_mutex);
798 static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL;
800 void brioctl_set(int (*hook) (unsigned int, void __user *))
802 mutex_lock(&br_ioctl_mutex);
803 br_ioctl_hook = hook;
804 mutex_unlock(&br_ioctl_mutex);
807 EXPORT_SYMBOL(brioctl_set);
809 static DEFINE_MUTEX(vlan_ioctl_mutex);
810 static int (*vlan_ioctl_hook) (void __user *arg);
812 void vlan_ioctl_set(int (*hook) (void __user *))
814 mutex_lock(&vlan_ioctl_mutex);
815 vlan_ioctl_hook = hook;
816 mutex_unlock(&vlan_ioctl_mutex);
819 EXPORT_SYMBOL(vlan_ioctl_set);
821 static DEFINE_MUTEX(dlci_ioctl_mutex);
822 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
824 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
826 mutex_lock(&dlci_ioctl_mutex);
827 dlci_ioctl_hook = hook;
828 mutex_unlock(&dlci_ioctl_mutex);
831 EXPORT_SYMBOL(dlci_ioctl_set);
834 * With an ioctl, arg may well be a user mode pointer, but we don't know
835 * what to do with it - that's up to the protocol still.
838 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
841 void __user *argp = (void __user *)arg;
844 sock = file->private_data;
845 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
846 err = dev_ioctl(cmd, argp);
848 #ifdef CONFIG_WIRELESS_EXT
849 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
850 err = dev_ioctl(cmd, argp);
852 #endif /* CONFIG_WIRELESS_EXT */
857 if (get_user(pid, (int __user *)argp))
859 err = f_setown(sock->file, pid, 1);
863 err = put_user(f_getown(sock->file),
872 request_module("bridge");
874 mutex_lock(&br_ioctl_mutex);
876 err = br_ioctl_hook(cmd, argp);
877 mutex_unlock(&br_ioctl_mutex);
882 if (!vlan_ioctl_hook)
883 request_module("8021q");
885 mutex_lock(&vlan_ioctl_mutex);
887 err = vlan_ioctl_hook(argp);
888 mutex_unlock(&vlan_ioctl_mutex);
893 if (!dlci_ioctl_hook)
894 request_module("dlci");
896 if (dlci_ioctl_hook) {
897 mutex_lock(&dlci_ioctl_mutex);
898 err = dlci_ioctl_hook(cmd, argp);
899 mutex_unlock(&dlci_ioctl_mutex);
903 err = sock->ops->ioctl(sock, cmd, arg);
906 * If this ioctl is unknown try to hand it down
909 if (err == -ENOIOCTLCMD)
910 err = dev_ioctl(cmd, argp);
916 int sock_create_lite(int family, int type, int protocol, struct socket **res)
919 struct socket *sock = NULL;
921 err = security_socket_create(family, type, protocol, 1);
932 err = security_socket_post_create(sock, family, type, protocol, 1);
945 /* No kernel lock held - perfect */
946 static unsigned int sock_poll(struct file *file, poll_table *wait)
951 * We can't return errors to poll, so it's either yes or no.
953 sock = file->private_data;
954 return sock->ops->poll(file, sock, wait);
957 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
959 struct socket *sock = file->private_data;
961 return sock->ops->mmap(file, sock, vma);
964 static int sock_close(struct inode *inode, struct file *filp)
967 * It was possible the inode is NULL we were
968 * closing an unfinished socket.
972 printk(KERN_DEBUG "sock_close: NULL inode\n");
975 sock_fasync(-1, filp, 0);
976 sock_release(SOCKET_I(inode));
981 * Update the socket async list
983 * Fasync_list locking strategy.
985 * 1. fasync_list is modified only under process context socket lock
986 * i.e. under semaphore.
987 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
988 * or under socket lock.
989 * 3. fasync_list can be used from softirq context, so that
990 * modification under socket lock have to be enhanced with
991 * write_lock_bh(&sk->sk_callback_lock).
995 static int sock_fasync(int fd, struct file *filp, int on)
997 struct fasync_struct *fa, *fna = NULL, **prev;
1002 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1007 sock = filp->private_data;
1017 prev = &(sock->fasync_list);
1019 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1020 if (fa->fa_file == filp)
1025 write_lock_bh(&sk->sk_callback_lock);
1027 write_unlock_bh(&sk->sk_callback_lock);
1032 fna->fa_file = filp;
1034 fna->magic = FASYNC_MAGIC;
1035 fna->fa_next = sock->fasync_list;
1036 write_lock_bh(&sk->sk_callback_lock);
1037 sock->fasync_list = fna;
1038 write_unlock_bh(&sk->sk_callback_lock);
1041 write_lock_bh(&sk->sk_callback_lock);
1042 *prev = fa->fa_next;
1043 write_unlock_bh(&sk->sk_callback_lock);
1049 release_sock(sock->sk);
1053 /* This function may be called only under socket lock or callback_lock */
1055 int sock_wake_async(struct socket *sock, int how, int band)
1057 if (!sock || !sock->fasync_list)
1062 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1066 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1071 __kill_fasync(sock->fasync_list, SIGIO, band);
1074 __kill_fasync(sock->fasync_list, SIGURG, band);
1079 static int __sock_create(int family, int type, int protocol,
1080 struct socket **res, int kern)
1083 struct socket *sock;
1084 const struct net_proto_family *pf;
1087 * Check protocol is in range
1089 if (family < 0 || family >= NPROTO)
1090 return -EAFNOSUPPORT;
1091 if (type < 0 || type >= SOCK_MAX)
1096 This uglymoron is moved from INET layer to here to avoid
1097 deadlock in module load.
1099 if (family == PF_INET && type == SOCK_PACKET) {
1103 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1109 err = security_socket_create(family, type, protocol, kern);
1114 * Allocate the socket and allow the family to set things up. if
1115 * the protocol is 0, the family is instructed to select an appropriate
1118 sock = sock_alloc();
1120 if (net_ratelimit())
1121 printk(KERN_WARNING "socket: no more sockets\n");
1122 return -ENFILE; /* Not exactly a match, but its the
1123 closest posix thing */
1128 #if defined(CONFIG_KMOD)
1129 /* Attempt to load a protocol module if the find failed.
1131 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1132 * requested real, full-featured networking support upon configuration.
1133 * Otherwise module support will break!
1135 if (net_families[family] == NULL)
1136 request_module("net-pf-%d", family);
1140 pf = rcu_dereference(net_families[family]);
1141 err = -EAFNOSUPPORT;
1146 * We will call the ->create function, that possibly is in a loadable
1147 * module, so we have to bump that loadable module refcnt first.
1149 if (!try_module_get(pf->owner))
1152 /* Now protected by module ref count */
1155 err = pf->create(sock, protocol);
1157 goto out_module_put;
1160 * Now to bump the refcnt of the [loadable] module that owns this
1161 * socket at sock_release time we decrement its refcnt.
1163 if (!try_module_get(sock->ops->owner))
1164 goto out_module_busy;
1167 * Now that we're done with the ->create function, the [loadable]
1168 * module can have its refcnt decremented
1170 module_put(pf->owner);
1171 err = security_socket_post_create(sock, family, type, protocol, kern);
1179 err = -EAFNOSUPPORT;
1182 module_put(pf->owner);
1189 goto out_sock_release;
1192 int sock_create(int family, int type, int protocol, struct socket **res)
1194 return __sock_create(family, type, protocol, res, 0);
1197 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1199 return __sock_create(family, type, protocol, res, 1);
1202 asmlinkage long sys_socket(int family, int type, int protocol)
1205 struct socket *sock;
1207 retval = sock_create(family, type, protocol, &sock);
1211 retval = sock_map_fd(sock);
1216 /* It may be already another descriptor 8) Not kernel problem. */
1225 * Create a pair of connected sockets.
1228 asmlinkage long sys_socketpair(int family, int type, int protocol,
1229 int __user *usockvec)
1231 struct socket *sock1, *sock2;
1233 struct file *newfile1, *newfile2;
1236 * Obtain the first socket and check if the underlying protocol
1237 * supports the socketpair call.
1240 err = sock_create(family, type, protocol, &sock1);
1244 err = sock_create(family, type, protocol, &sock2);
1248 err = sock1->ops->socketpair(sock1, sock2);
1250 goto out_release_both;
1252 fd1 = sock_alloc_fd(&newfile1);
1253 if (unlikely(fd1 < 0))
1254 goto out_release_both;
1256 fd2 = sock_alloc_fd(&newfile2);
1257 if (unlikely(fd2 < 0)) {
1260 goto out_release_both;
1263 err = sock_attach_fd(sock1, newfile1);
1264 if (unlikely(err < 0)) {
1268 err = sock_attach_fd(sock2, newfile2);
1269 if (unlikely(err < 0)) {
1274 err = audit_fd_pair(fd1, fd2);
1281 fd_install(fd1, newfile1);
1282 fd_install(fd2, newfile2);
1283 /* fd1 and fd2 may be already another descriptors.
1284 * Not kernel problem.
1287 err = put_user(fd1, &usockvec[0]);
1289 err = put_user(fd2, &usockvec[1]);
1298 sock_release(sock2);
1300 sock_release(sock1);
1306 sock_release(sock1);
1309 sock_release(sock2);
1317 * Bind a name to a socket. Nothing much to do here since it's
1318 * the protocol's responsibility to handle the local address.
1320 * We move the socket address to kernel space before we call
1321 * the protocol layer (having also checked the address is ok).
1324 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1326 struct socket *sock;
1327 char address[MAX_SOCK_ADDR];
1328 int err, fput_needed;
1330 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1332 err = move_addr_to_kernel(umyaddr, addrlen, address);
1334 err = security_socket_bind(sock,
1335 (struct sockaddr *)address,
1338 err = sock->ops->bind(sock,
1342 fput_light(sock->file, fput_needed);
1348 * Perform a listen. Basically, we allow the protocol to do anything
1349 * necessary for a listen, and if that works, we mark the socket as
1350 * ready for listening.
1353 int sysctl_somaxconn __read_mostly = SOMAXCONN;
1355 asmlinkage long sys_listen(int fd, int backlog)
1357 struct socket *sock;
1358 int err, fput_needed;
1360 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1362 if ((unsigned)backlog > sysctl_somaxconn)
1363 backlog = sysctl_somaxconn;
1365 err = security_socket_listen(sock, backlog);
1367 err = sock->ops->listen(sock, backlog);
1369 fput_light(sock->file, fput_needed);
1375 * For accept, we attempt to create a new socket, set up the link
1376 * with the client, wake up the client, then return the new
1377 * connected fd. We collect the address of the connector in kernel
1378 * space and move it to user at the very end. This is unclean because
1379 * we open the socket then return an error.
1381 * 1003.1g adds the ability to recvmsg() to query connection pending
1382 * status to recvmsg. We need to add that support in a way thats
1383 * clean when we restucture accept also.
1386 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1387 int __user *upeer_addrlen)
1389 struct socket *sock, *newsock;
1390 struct file *newfile;
1391 int err, len, newfd, fput_needed;
1392 char address[MAX_SOCK_ADDR];
1394 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1399 if (!(newsock = sock_alloc()))
1402 newsock->type = sock->type;
1403 newsock->ops = sock->ops;
1406 * We don't need try_module_get here, as the listening socket (sock)
1407 * has the protocol module (sock->ops->owner) held.
1409 __module_get(newsock->ops->owner);
1411 newfd = sock_alloc_fd(&newfile);
1412 if (unlikely(newfd < 0)) {
1414 sock_release(newsock);
1418 err = sock_attach_fd(newsock, newfile);
1422 err = security_socket_accept(sock, newsock);
1426 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1430 if (upeer_sockaddr) {
1431 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1433 err = -ECONNABORTED;
1436 err = move_addr_to_user(address, len, upeer_sockaddr,
1442 /* File flags are not inherited via accept() unlike another OSes. */
1444 fd_install(newfd, newfile);
1447 security_socket_post_accept(sock, newsock);
1450 fput_light(sock->file, fput_needed);
1454 sock_release(newsock);
1456 put_unused_fd(newfd);
1460 put_unused_fd(newfd);
1465 * Attempt to connect to a socket with the server address. The address
1466 * is in user space so we verify it is OK and move it to kernel space.
1468 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1471 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1472 * other SEQPACKET protocols that take time to connect() as it doesn't
1473 * include the -EINPROGRESS status for such sockets.
1476 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1479 struct socket *sock;
1480 char address[MAX_SOCK_ADDR];
1481 int err, fput_needed;
1483 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1486 err = move_addr_to_kernel(uservaddr, addrlen, address);
1491 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1495 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1496 sock->file->f_flags);
1498 fput_light(sock->file, fput_needed);
1504 * Get the local address ('name') of a socket object. Move the obtained
1505 * name to user space.
1508 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1509 int __user *usockaddr_len)
1511 struct socket *sock;
1512 char address[MAX_SOCK_ADDR];
1513 int len, err, fput_needed;
1515 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1519 err = security_socket_getsockname(sock);
1523 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1526 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1529 fput_light(sock->file, fput_needed);
1535 * Get the remote address ('name') of a socket object. Move the obtained
1536 * name to user space.
1539 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1540 int __user *usockaddr_len)
1542 struct socket *sock;
1543 char address[MAX_SOCK_ADDR];
1544 int len, err, fput_needed;
1546 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1548 err = security_socket_getpeername(sock);
1550 fput_light(sock->file, fput_needed);
1555 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1558 err = move_addr_to_user(address, len, usockaddr,
1560 fput_light(sock->file, fput_needed);
1566 * Send a datagram to a given address. We move the address into kernel
1567 * space and check the user space data area is readable before invoking
1571 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1572 unsigned flags, struct sockaddr __user *addr,
1575 struct socket *sock;
1576 char address[MAX_SOCK_ADDR];
1581 struct file *sock_file;
1583 sock_file = fget_light(fd, &fput_needed);
1588 sock = sock_from_file(sock_file, &err);
1591 iov.iov_base = buff;
1593 msg.msg_name = NULL;
1596 msg.msg_control = NULL;
1597 msg.msg_controllen = 0;
1598 msg.msg_namelen = 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);
1618 * Send a datagram down a socket.
1621 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1623 return sys_sendto(fd, buff, len, flags, NULL, 0);
1627 * Receive a frame from the socket and optionally record the address of the
1628 * sender. We verify the buffers are writable and if needed move the
1629 * sender address from kernel to user space.
1632 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1633 unsigned flags, struct sockaddr __user *addr,
1634 int __user *addr_len)
1636 struct socket *sock;
1639 char address[MAX_SOCK_ADDR];
1641 struct file *sock_file;
1644 sock_file = fget_light(fd, &fput_needed);
1649 sock = sock_from_file(sock_file, &err);
1653 msg.msg_control = NULL;
1654 msg.msg_controllen = 0;
1658 iov.iov_base = ubuf;
1659 msg.msg_name = address;
1660 msg.msg_namelen = MAX_SOCK_ADDR;
1661 if (sock->file->f_flags & O_NONBLOCK)
1662 flags |= MSG_DONTWAIT;
1663 err = sock_recvmsg(sock, &msg, size, flags);
1665 if (err >= 0 && addr != NULL) {
1666 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1671 fput_light(sock_file, fput_needed);
1677 * Receive a datagram from a socket.
1680 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1683 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1687 * Set a socket option. Because we don't know the option lengths we have
1688 * to pass the user mode parameter for the protocols to sort out.
1691 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1692 char __user *optval, int optlen)
1694 int err, fput_needed;
1695 struct socket *sock;
1700 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1702 err = security_socket_setsockopt(sock, level, optname);
1706 if (level == SOL_SOCKET)
1708 sock_setsockopt(sock, level, optname, optval,
1712 sock->ops->setsockopt(sock, level, optname, optval,
1715 fput_light(sock->file, fput_needed);
1721 * Get a socket option. Because we don't know the option lengths we have
1722 * to pass a user mode parameter for the protocols to sort out.
1725 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1726 char __user *optval, int __user *optlen)
1728 int err, fput_needed;
1729 struct socket *sock;
1731 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1733 err = security_socket_getsockopt(sock, level, optname);
1737 if (level == SOL_SOCKET)
1739 sock_getsockopt(sock, level, optname, optval,
1743 sock->ops->getsockopt(sock, level, optname, optval,
1746 fput_light(sock->file, fput_needed);
1752 * Shutdown a socket.
1755 asmlinkage long sys_shutdown(int fd, int how)
1757 int err, fput_needed;
1758 struct socket *sock;
1760 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1762 err = security_socket_shutdown(sock, how);
1764 err = sock->ops->shutdown(sock, how);
1765 fput_light(sock->file, fput_needed);
1770 /* A couple of helpful macros for getting the address of the 32/64 bit
1771 * fields which are the same type (int / unsigned) on our platforms.
1773 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1774 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1775 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1778 * BSD sendmsg interface
1781 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1783 struct compat_msghdr __user *msg_compat =
1784 (struct compat_msghdr __user *)msg;
1785 struct socket *sock;
1786 char address[MAX_SOCK_ADDR];
1787 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1788 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1789 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1790 /* 20 is size of ipv6_pktinfo */
1791 unsigned char *ctl_buf = ctl;
1792 struct msghdr msg_sys;
1793 int err, ctl_len, iov_size, total_len;
1797 if (MSG_CMSG_COMPAT & flags) {
1798 if (get_compat_msghdr(&msg_sys, msg_compat))
1801 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1804 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1808 /* do not move before msg_sys is valid */
1810 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1813 /* Check whether to allocate the iovec area */
1815 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1816 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1817 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1822 /* This will also move the address data into kernel space */
1823 if (MSG_CMSG_COMPAT & flags) {
1824 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1826 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1833 if (msg_sys.msg_controllen > INT_MAX)
1835 ctl_len = msg_sys.msg_controllen;
1836 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1838 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1842 ctl_buf = msg_sys.msg_control;
1843 ctl_len = msg_sys.msg_controllen;
1844 } else if (ctl_len) {
1845 if (ctl_len > sizeof(ctl)) {
1846 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1847 if (ctl_buf == NULL)
1852 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1853 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1854 * checking falls down on this.
1856 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1859 msg_sys.msg_control = ctl_buf;
1861 msg_sys.msg_flags = flags;
1863 if (sock->file->f_flags & O_NONBLOCK)
1864 msg_sys.msg_flags |= MSG_DONTWAIT;
1865 err = sock_sendmsg(sock, &msg_sys, total_len);
1869 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1871 if (iov != iovstack)
1872 sock_kfree_s(sock->sk, iov, iov_size);
1874 fput_light(sock->file, fput_needed);
1880 * BSD recvmsg interface
1883 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1886 struct compat_msghdr __user *msg_compat =
1887 (struct compat_msghdr __user *)msg;
1888 struct socket *sock;
1889 struct iovec iovstack[UIO_FASTIOV];
1890 struct iovec *iov = iovstack;
1891 struct msghdr msg_sys;
1892 unsigned long cmsg_ptr;
1893 int err, iov_size, total_len, len;
1896 /* kernel mode address */
1897 char addr[MAX_SOCK_ADDR];
1899 /* user mode address pointers */
1900 struct sockaddr __user *uaddr;
1901 int __user *uaddr_len;
1903 if (MSG_CMSG_COMPAT & flags) {
1904 if (get_compat_msghdr(&msg_sys, msg_compat))
1907 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1910 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1915 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1918 /* Check whether to allocate the iovec area */
1920 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1921 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1922 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1928 * Save the user-mode address (verify_iovec will change the
1929 * kernel msghdr to use the kernel address space)
1932 uaddr = (void __user *)msg_sys.msg_name;
1933 uaddr_len = COMPAT_NAMELEN(msg);
1934 if (MSG_CMSG_COMPAT & flags) {
1935 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1937 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1942 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1943 msg_sys.msg_flags = 0;
1944 if (MSG_CMSG_COMPAT & flags)
1945 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1947 if (sock->file->f_flags & O_NONBLOCK)
1948 flags |= MSG_DONTWAIT;
1949 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1954 if (uaddr != NULL) {
1955 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1960 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1964 if (MSG_CMSG_COMPAT & flags)
1965 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1966 &msg_compat->msg_controllen);
1968 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1969 &msg->msg_controllen);
1975 if (iov != iovstack)
1976 sock_kfree_s(sock->sk, iov, iov_size);
1978 fput_light(sock->file, fput_needed);
1983 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1985 /* Argument list sizes for sys_socketcall */
1986 #define AL(x) ((x) * sizeof(unsigned long))
1987 static const unsigned char nargs[18]={
1988 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1989 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1990 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1996 * System call vectors.
1998 * Argument checking cleaned up. Saved 20% in size.
1999 * This function doesn't need to set the kernel lock because
2000 * it is set by the callees.
2003 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2006 unsigned long a0, a1;
2009 if (call < 1 || call > SYS_RECVMSG)
2012 /* copy_from_user should be SMP safe. */
2013 if (copy_from_user(a, args, nargs[call]))
2016 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2025 err = sys_socket(a0, a1, a[2]);
2028 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2031 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2034 err = sys_listen(a0, a1);
2038 sys_accept(a0, (struct sockaddr __user *)a1,
2039 (int __user *)a[2]);
2041 case SYS_GETSOCKNAME:
2043 sys_getsockname(a0, (struct sockaddr __user *)a1,
2044 (int __user *)a[2]);
2046 case SYS_GETPEERNAME:
2048 sys_getpeername(a0, (struct sockaddr __user *)a1,
2049 (int __user *)a[2]);
2051 case SYS_SOCKETPAIR:
2052 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2055 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2058 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2059 (struct sockaddr __user *)a[4], a[5]);
2062 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2065 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2066 (struct sockaddr __user *)a[4],
2067 (int __user *)a[5]);
2070 err = sys_shutdown(a0, a1);
2072 case SYS_SETSOCKOPT:
2073 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2075 case SYS_GETSOCKOPT:
2077 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2078 (int __user *)a[4]);
2081 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2084 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2093 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2096 * sock_register - add a socket protocol handler
2097 * @ops: description of protocol
2099 * This function is called by a protocol handler that wants to
2100 * advertise its address family, and have it linked into the
2101 * socket interface. The value ops->family coresponds to the
2102 * socket system call protocol family.
2104 int sock_register(const struct net_proto_family *ops)
2108 if (ops->family >= NPROTO) {
2109 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2114 spin_lock(&net_family_lock);
2115 if (net_families[ops->family])
2118 net_families[ops->family] = ops;
2121 spin_unlock(&net_family_lock);
2123 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2128 * sock_unregister - remove a protocol handler
2129 * @family: protocol family to remove
2131 * This function is called by a protocol handler that wants to
2132 * remove its address family, and have it unlinked from the
2133 * new socket creation.
2135 * If protocol handler is a module, then it can use module reference
2136 * counts to protect against new references. If protocol handler is not
2137 * a module then it needs to provide its own protection in
2138 * the ops->create routine.
2140 void sock_unregister(int family)
2142 BUG_ON(family < 0 || family >= NPROTO);
2144 spin_lock(&net_family_lock);
2145 net_families[family] = NULL;
2146 spin_unlock(&net_family_lock);
2150 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2153 static int __init sock_init(void)
2156 * Initialize sock SLAB cache.
2162 * Initialize skbuff SLAB cache
2167 * Initialize the protocols module.
2171 register_filesystem(&sock_fs_type);
2172 sock_mnt = kern_mount(&sock_fs_type);
2174 /* The real protocol initialization is performed in later initcalls.
2177 #ifdef CONFIG_NETFILTER
2184 core_initcall(sock_init); /* early initcall */
2186 #ifdef CONFIG_PROC_FS
2187 void socket_seq_show(struct seq_file *seq)
2192 for_each_possible_cpu(cpu)
2193 counter += per_cpu(sockets_in_use, cpu);
2195 /* It can be negative, by the way. 8) */
2199 seq_printf(seq, "sockets: used %d\n", counter);
2201 #endif /* CONFIG_PROC_FS */
2203 #ifdef CONFIG_COMPAT
2204 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2207 struct socket *sock = file->private_data;
2208 int ret = -ENOIOCTLCMD;
2210 if (sock->ops->compat_ioctl)
2211 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2217 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2219 return sock->ops->bind(sock, addr, addrlen);
2222 int kernel_listen(struct socket *sock, int backlog)
2224 return sock->ops->listen(sock, backlog);
2227 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2229 struct sock *sk = sock->sk;
2232 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2237 err = sock->ops->accept(sock, *newsock, flags);
2239 sock_release(*newsock);
2243 (*newsock)->ops = sock->ops;
2249 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2252 return sock->ops->connect(sock, addr, addrlen, flags);
2255 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2258 return sock->ops->getname(sock, addr, addrlen, 0);
2261 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2264 return sock->ops->getname(sock, addr, addrlen, 1);
2267 int kernel_getsockopt(struct socket *sock, int level, int optname,
2268 char *optval, int *optlen)
2270 mm_segment_t oldfs = get_fs();
2274 if (level == SOL_SOCKET)
2275 err = sock_getsockopt(sock, level, optname, optval, optlen);
2277 err = sock->ops->getsockopt(sock, level, optname, optval,
2283 int kernel_setsockopt(struct socket *sock, int level, int optname,
2284 char *optval, int optlen)
2286 mm_segment_t oldfs = get_fs();
2290 if (level == SOL_SOCKET)
2291 err = sock_setsockopt(sock, level, optname, optval, optlen);
2293 err = sock->ops->setsockopt(sock, level, optname, optval,
2299 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2300 size_t size, int flags)
2302 if (sock->ops->sendpage)
2303 return sock->ops->sendpage(sock, page, offset, size, flags);
2305 return sock_no_sendpage(sock, page, offset, size, flags);
2308 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2310 mm_segment_t oldfs = get_fs();
2314 err = sock->ops->ioctl(sock, cmd, arg);
2320 /* ABI emulation layers need these two */
2321 EXPORT_SYMBOL(move_addr_to_kernel);
2322 EXPORT_SYMBOL(move_addr_to_user);
2323 EXPORT_SYMBOL(sock_create);
2324 EXPORT_SYMBOL(sock_create_kern);
2325 EXPORT_SYMBOL(sock_create_lite);
2326 EXPORT_SYMBOL(sock_map_fd);
2327 EXPORT_SYMBOL(sock_recvmsg);
2328 EXPORT_SYMBOL(sock_register);
2329 EXPORT_SYMBOL(sock_release);
2330 EXPORT_SYMBOL(sock_sendmsg);
2331 EXPORT_SYMBOL(sock_unregister);
2332 EXPORT_SYMBOL(sock_wake_async);
2333 EXPORT_SYMBOL(sockfd_lookup);
2334 EXPORT_SYMBOL(kernel_sendmsg);
2335 EXPORT_SYMBOL(kernel_recvmsg);
2336 EXPORT_SYMBOL(kernel_bind);
2337 EXPORT_SYMBOL(kernel_listen);
2338 EXPORT_SYMBOL(kernel_accept);
2339 EXPORT_SYMBOL(kernel_connect);
2340 EXPORT_SYMBOL(kernel_getsockname);
2341 EXPORT_SYMBOL(kernel_getpeername);
2342 EXPORT_SYMBOL(kernel_getsockopt);
2343 EXPORT_SYMBOL(kernel_setsockopt);
2344 EXPORT_SYMBOL(kernel_sendpage);
2345 EXPORT_SYMBOL(kernel_sock_ioctl);