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/divert.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
89 #include <asm/uaccess.h>
90 #include <asm/unistd.h>
92 #include <net/compat.h>
95 #include <linux/netfilter.h>
97 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
98 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
99 unsigned long nr_segs, loff_t pos);
100 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
101 unsigned long nr_segs, loff_t pos);
102 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
104 static int sock_close(struct inode *inode, struct file *file);
105 static unsigned int sock_poll(struct file *file,
106 struct poll_table_struct *wait);
107 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
109 static long compat_sock_ioctl(struct file *file,
110 unsigned int cmd, unsigned long arg);
112 static int sock_fasync(int fd, struct file *filp, int on);
113 static ssize_t sock_sendpage(struct file *file, struct page *page,
114 int offset, size_t size, loff_t *ppos, int more);
117 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
118 * in the operation structures but are done directly via the socketcall() multiplexor.
121 static struct file_operations socket_file_ops = {
122 .owner = THIS_MODULE,
124 .aio_read = sock_aio_read,
125 .aio_write = sock_aio_write,
127 .unlocked_ioctl = sock_ioctl,
129 .compat_ioctl = compat_sock_ioctl,
132 .open = sock_no_open, /* special open code to disallow open via /proc */
133 .release = sock_close,
134 .fasync = sock_fasync,
135 .sendpage = sock_sendpage,
136 .splice_write = generic_splice_sendpage,
140 * The protocol list. Each protocol is registered in here.
143 static DEFINE_SPINLOCK(net_family_lock);
144 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
147 * Statistics counters of the socket lists
150 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
154 * Move socket addresses back and forth across the kernel/user
155 * divide and look after the messy bits.
158 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
159 16 for IP, 16 for IPX,
162 must be at least one bigger than
163 the AF_UNIX size (see net/unix/af_unix.c
168 * move_addr_to_kernel - copy a socket address into kernel space
169 * @uaddr: Address in user space
170 * @kaddr: Address in kernel space
171 * @ulen: Length in user space
173 * The address is copied into kernel space. If the provided address is
174 * too long an error code of -EINVAL is returned. If the copy gives
175 * invalid addresses -EFAULT is returned. On a success 0 is returned.
178 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
180 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
184 if (copy_from_user(kaddr, uaddr, ulen))
186 return audit_sockaddr(ulen, kaddr);
190 * move_addr_to_user - copy an address to user space
191 * @kaddr: kernel space address
192 * @klen: length of address in kernel
193 * @uaddr: user space address
194 * @ulen: pointer to user length field
196 * The value pointed to by ulen on entry is the buffer length available.
197 * This is overwritten with the buffer space used. -EINVAL is returned
198 * if an overlong buffer is specified or a negative buffer size. -EFAULT
199 * is returned if either the buffer or the length field are not
201 * After copying the data up to the limit the user specifies, the true
202 * length of the data is written over the length limit the user
203 * specified. Zero is returned for a success.
206 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
212 err = get_user(len, ulen);
217 if (len < 0 || len > MAX_SOCK_ADDR)
220 if (audit_sockaddr(klen, kaddr))
222 if (copy_to_user(uaddr, kaddr, len))
226 * "fromlen shall refer to the value before truncation.."
229 return __put_user(klen, ulen);
232 #define SOCKFS_MAGIC 0x534F434B
234 static kmem_cache_t *sock_inode_cachep __read_mostly;
236 static struct inode *sock_alloc_inode(struct super_block *sb)
238 struct socket_alloc *ei;
240 ei = kmem_cache_alloc(sock_inode_cachep, SLAB_KERNEL);
243 init_waitqueue_head(&ei->socket.wait);
245 ei->socket.fasync_list = NULL;
246 ei->socket.state = SS_UNCONNECTED;
247 ei->socket.flags = 0;
248 ei->socket.ops = NULL;
249 ei->socket.sk = NULL;
250 ei->socket.file = NULL;
252 return &ei->vfs_inode;
255 static void sock_destroy_inode(struct inode *inode)
257 kmem_cache_free(sock_inode_cachep,
258 container_of(inode, struct socket_alloc, vfs_inode));
261 static void init_once(void *foo, kmem_cache_t *cachep, unsigned long flags)
263 struct socket_alloc *ei = (struct socket_alloc *)foo;
265 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR))
266 == SLAB_CTOR_CONSTRUCTOR)
267 inode_init_once(&ei->vfs_inode);
270 static int init_inodecache(void)
272 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
273 sizeof(struct socket_alloc),
275 (SLAB_HWCACHE_ALIGN |
276 SLAB_RECLAIM_ACCOUNT |
280 if (sock_inode_cachep == NULL)
285 static struct super_operations sockfs_ops = {
286 .alloc_inode = sock_alloc_inode,
287 .destroy_inode =sock_destroy_inode,
288 .statfs = simple_statfs,
291 static int sockfs_get_sb(struct file_system_type *fs_type,
292 int flags, const char *dev_name, void *data,
293 struct vfsmount *mnt)
295 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
299 static struct vfsmount *sock_mnt __read_mostly;
301 static struct file_system_type sock_fs_type = {
303 .get_sb = sockfs_get_sb,
304 .kill_sb = kill_anon_super,
307 static int sockfs_delete_dentry(struct dentry *dentry)
311 static struct dentry_operations sockfs_dentry_operations = {
312 .d_delete = sockfs_delete_dentry,
316 * Obtains the first available file descriptor and sets it up for use.
318 * These functions create file structures and maps them to fd space
319 * of the current process. On success it returns file descriptor
320 * and file struct implicitly stored in sock->file.
321 * Note that another thread may close file descriptor before we return
322 * from this function. We use the fact that now we do not refer
323 * to socket after mapping. If one day we will need it, this
324 * function will increment ref. count on file by 1.
326 * In any case returned fd MAY BE not valid!
327 * This race condition is unavoidable
328 * with shared fd spaces, we cannot solve it inside kernel,
329 * but we take care of internal coherence yet.
332 static int sock_alloc_fd(struct file **filep)
336 fd = get_unused_fd();
337 if (likely(fd >= 0)) {
338 struct file *file = get_empty_filp();
341 if (unlikely(!file)) {
350 static int sock_attach_fd(struct socket *sock, struct file *file)
355 this.len = sprintf(name, "[%lu]", SOCK_INODE(sock)->i_ino);
357 this.hash = SOCK_INODE(sock)->i_ino;
359 file->f_dentry = d_alloc(sock_mnt->mnt_sb->s_root, &this);
360 if (unlikely(!file->f_dentry))
363 file->f_dentry->d_op = &sockfs_dentry_operations;
364 d_add(file->f_dentry, SOCK_INODE(sock));
365 file->f_vfsmnt = mntget(sock_mnt);
366 file->f_mapping = file->f_dentry->d_inode->i_mapping;
369 file->f_op = SOCK_INODE(sock)->i_fop = &socket_file_ops;
370 file->f_mode = FMODE_READ | FMODE_WRITE;
371 file->f_flags = O_RDWR;
373 file->private_data = sock;
378 int sock_map_fd(struct socket *sock)
380 struct file *newfile;
381 int fd = sock_alloc_fd(&newfile);
383 if (likely(fd >= 0)) {
384 int err = sock_attach_fd(sock, newfile);
386 if (unlikely(err < 0)) {
391 fd_install(fd, newfile);
396 static struct socket *sock_from_file(struct file *file, int *err)
401 if (file->f_op == &socket_file_ops)
402 return file->private_data; /* set in sock_map_fd */
404 inode = file->f_dentry->d_inode;
405 if (!S_ISSOCK(inode->i_mode)) {
410 sock = SOCKET_I(inode);
411 if (sock->file != file) {
412 printk(KERN_ERR "socki_lookup: socket file changed!\n");
419 * sockfd_lookup - Go from a file number to its socket slot
421 * @err: pointer to an error code return
423 * The file handle passed in is locked and the socket it is bound
424 * too is returned. If an error occurs the err pointer is overwritten
425 * with a negative errno code and NULL is returned. The function checks
426 * for both invalid handles and passing a handle which is not a socket.
428 * On a success the socket object pointer is returned.
431 struct socket *sockfd_lookup(int fd, int *err)
442 sock = sock_from_file(file, err);
448 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
454 file = fget_light(fd, fput_needed);
456 sock = sock_from_file(file, err);
459 fput_light(file, *fput_needed);
465 * sock_alloc - allocate a socket
467 * Allocate a new inode and socket object. The two are bound together
468 * and initialised. The socket is then returned. If we are out of inodes
472 static struct socket *sock_alloc(void)
477 inode = new_inode(sock_mnt->mnt_sb);
481 sock = SOCKET_I(inode);
483 inode->i_mode = S_IFSOCK | S_IRWXUGO;
484 inode->i_uid = current->fsuid;
485 inode->i_gid = current->fsgid;
487 get_cpu_var(sockets_in_use)++;
488 put_cpu_var(sockets_in_use);
493 * In theory you can't get an open on this inode, but /proc provides
494 * a back door. Remember to keep it shut otherwise you'll let the
495 * creepy crawlies in.
498 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
503 const struct file_operations bad_sock_fops = {
504 .owner = THIS_MODULE,
505 .open = sock_no_open,
509 * sock_release - close a socket
510 * @sock: socket to close
512 * The socket is released from the protocol stack if it has a release
513 * callback, and the inode is then released if the socket is bound to
514 * an inode not a file.
517 void sock_release(struct socket *sock)
520 struct module *owner = sock->ops->owner;
522 sock->ops->release(sock);
527 if (sock->fasync_list)
528 printk(KERN_ERR "sock_release: fasync list not empty!\n");
530 get_cpu_var(sockets_in_use)--;
531 put_cpu_var(sockets_in_use);
533 iput(SOCK_INODE(sock));
539 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
540 struct msghdr *msg, size_t size)
542 struct sock_iocb *si = kiocb_to_siocb(iocb);
550 err = security_socket_sendmsg(sock, msg, size);
554 return sock->ops->sendmsg(iocb, sock, msg, size);
557 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
560 struct sock_iocb siocb;
563 init_sync_kiocb(&iocb, NULL);
564 iocb.private = &siocb;
565 ret = __sock_sendmsg(&iocb, sock, msg, size);
566 if (-EIOCBQUEUED == ret)
567 ret = wait_on_sync_kiocb(&iocb);
571 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
572 struct kvec *vec, size_t num, size_t size)
574 mm_segment_t oldfs = get_fs();
579 * the following is safe, since for compiler definitions of kvec and
580 * iovec are identical, yielding the same in-core layout and alignment
582 msg->msg_iov = (struct iovec *)vec;
583 msg->msg_iovlen = num;
584 result = sock_sendmsg(sock, msg, size);
589 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
590 struct msghdr *msg, size_t size, int flags)
593 struct sock_iocb *si = kiocb_to_siocb(iocb);
601 err = security_socket_recvmsg(sock, msg, size, flags);
605 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
608 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
609 size_t size, int flags)
612 struct sock_iocb siocb;
615 init_sync_kiocb(&iocb, NULL);
616 iocb.private = &siocb;
617 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
618 if (-EIOCBQUEUED == ret)
619 ret = wait_on_sync_kiocb(&iocb);
623 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
624 struct kvec *vec, size_t num, size_t size, int flags)
626 mm_segment_t oldfs = get_fs();
631 * the following is safe, since for compiler definitions of kvec and
632 * iovec are identical, yielding the same in-core layout and alignment
634 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
635 result = sock_recvmsg(sock, msg, size, flags);
640 static void sock_aio_dtor(struct kiocb *iocb)
642 kfree(iocb->private);
645 static ssize_t sock_sendpage(struct file *file, struct page *page,
646 int offset, size_t size, loff_t *ppos, int more)
651 sock = file->private_data;
653 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
657 return sock->ops->sendpage(sock, page, offset, size, flags);
660 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
661 struct sock_iocb *siocb)
663 if (!is_sync_kiocb(iocb)) {
664 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
667 iocb->ki_dtor = sock_aio_dtor;
671 iocb->private = siocb;
675 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
676 struct file *file, const struct iovec *iov,
677 unsigned long nr_segs)
679 struct socket *sock = file->private_data;
683 for (i = 0; i < nr_segs; i++)
684 size += iov[i].iov_len;
686 msg->msg_name = NULL;
687 msg->msg_namelen = 0;
688 msg->msg_control = NULL;
689 msg->msg_controllen = 0;
690 msg->msg_iov = (struct iovec *)iov;
691 msg->msg_iovlen = nr_segs;
692 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
694 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
697 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
698 unsigned long nr_segs, loff_t pos)
700 struct sock_iocb siocb, *x;
705 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
709 x = alloc_sock_iocb(iocb, &siocb);
712 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
715 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
716 struct file *file, const struct iovec *iov,
717 unsigned long nr_segs)
719 struct socket *sock = file->private_data;
723 for (i = 0; i < nr_segs; i++)
724 size += iov[i].iov_len;
726 msg->msg_name = NULL;
727 msg->msg_namelen = 0;
728 msg->msg_control = NULL;
729 msg->msg_controllen = 0;
730 msg->msg_iov = (struct iovec *)iov;
731 msg->msg_iovlen = nr_segs;
732 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
733 if (sock->type == SOCK_SEQPACKET)
734 msg->msg_flags |= MSG_EOR;
736 return __sock_sendmsg(iocb, sock, msg, size);
739 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
740 unsigned long nr_segs, loff_t pos)
742 struct sock_iocb siocb, *x;
747 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
750 x = alloc_sock_iocb(iocb, &siocb);
754 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
758 * Atomic setting of ioctl hooks to avoid race
759 * with module unload.
762 static DEFINE_MUTEX(br_ioctl_mutex);
763 static int (*br_ioctl_hook) (unsigned int cmd, void __user *arg) = NULL;
765 void brioctl_set(int (*hook) (unsigned int, void __user *))
767 mutex_lock(&br_ioctl_mutex);
768 br_ioctl_hook = hook;
769 mutex_unlock(&br_ioctl_mutex);
772 EXPORT_SYMBOL(brioctl_set);
774 static DEFINE_MUTEX(vlan_ioctl_mutex);
775 static int (*vlan_ioctl_hook) (void __user *arg);
777 void vlan_ioctl_set(int (*hook) (void __user *))
779 mutex_lock(&vlan_ioctl_mutex);
780 vlan_ioctl_hook = hook;
781 mutex_unlock(&vlan_ioctl_mutex);
784 EXPORT_SYMBOL(vlan_ioctl_set);
786 static DEFINE_MUTEX(dlci_ioctl_mutex);
787 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
789 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
791 mutex_lock(&dlci_ioctl_mutex);
792 dlci_ioctl_hook = hook;
793 mutex_unlock(&dlci_ioctl_mutex);
796 EXPORT_SYMBOL(dlci_ioctl_set);
799 * With an ioctl, arg may well be a user mode pointer, but we don't know
800 * what to do with it - that's up to the protocol still.
803 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
806 void __user *argp = (void __user *)arg;
809 sock = file->private_data;
810 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
811 err = dev_ioctl(cmd, argp);
813 #ifdef CONFIG_WIRELESS_EXT
814 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
815 err = dev_ioctl(cmd, argp);
817 #endif /* CONFIG_WIRELESS_EXT */
822 if (get_user(pid, (int __user *)argp))
824 err = f_setown(sock->file, pid, 1);
828 err = put_user(f_getown(sock->file),
837 request_module("bridge");
839 mutex_lock(&br_ioctl_mutex);
841 err = br_ioctl_hook(cmd, argp);
842 mutex_unlock(&br_ioctl_mutex);
847 if (!vlan_ioctl_hook)
848 request_module("8021q");
850 mutex_lock(&vlan_ioctl_mutex);
852 err = vlan_ioctl_hook(argp);
853 mutex_unlock(&vlan_ioctl_mutex);
857 /* Convert this to call through a hook */
858 err = divert_ioctl(cmd, argp);
863 if (!dlci_ioctl_hook)
864 request_module("dlci");
866 if (dlci_ioctl_hook) {
867 mutex_lock(&dlci_ioctl_mutex);
868 err = dlci_ioctl_hook(cmd, argp);
869 mutex_unlock(&dlci_ioctl_mutex);
873 err = sock->ops->ioctl(sock, cmd, arg);
876 * If this ioctl is unknown try to hand it down
879 if (err == -ENOIOCTLCMD)
880 err = dev_ioctl(cmd, argp);
886 int sock_create_lite(int family, int type, int protocol, struct socket **res)
889 struct socket *sock = NULL;
891 err = security_socket_create(family, type, protocol, 1);
902 err = security_socket_post_create(sock, family, type, protocol, 1);
915 /* No kernel lock held - perfect */
916 static unsigned int sock_poll(struct file *file, poll_table *wait)
921 * We can't return errors to poll, so it's either yes or no.
923 sock = file->private_data;
924 return sock->ops->poll(file, sock, wait);
927 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
929 struct socket *sock = file->private_data;
931 return sock->ops->mmap(file, sock, vma);
934 static int sock_close(struct inode *inode, struct file *filp)
937 * It was possible the inode is NULL we were
938 * closing an unfinished socket.
942 printk(KERN_DEBUG "sock_close: NULL inode\n");
945 sock_fasync(-1, filp, 0);
946 sock_release(SOCKET_I(inode));
951 * Update the socket async list
953 * Fasync_list locking strategy.
955 * 1. fasync_list is modified only under process context socket lock
956 * i.e. under semaphore.
957 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
958 * or under socket lock.
959 * 3. fasync_list can be used from softirq context, so that
960 * modification under socket lock have to be enhanced with
961 * write_lock_bh(&sk->sk_callback_lock).
965 static int sock_fasync(int fd, struct file *filp, int on)
967 struct fasync_struct *fa, *fna = NULL, **prev;
972 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
977 sock = filp->private_data;
987 prev = &(sock->fasync_list);
989 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
990 if (fa->fa_file == filp)
995 write_lock_bh(&sk->sk_callback_lock);
997 write_unlock_bh(&sk->sk_callback_lock);
1002 fna->fa_file = filp;
1004 fna->magic = FASYNC_MAGIC;
1005 fna->fa_next = sock->fasync_list;
1006 write_lock_bh(&sk->sk_callback_lock);
1007 sock->fasync_list = fna;
1008 write_unlock_bh(&sk->sk_callback_lock);
1011 write_lock_bh(&sk->sk_callback_lock);
1012 *prev = fa->fa_next;
1013 write_unlock_bh(&sk->sk_callback_lock);
1019 release_sock(sock->sk);
1023 /* This function may be called only under socket lock or callback_lock */
1025 int sock_wake_async(struct socket *sock, int how, int band)
1027 if (!sock || !sock->fasync_list)
1032 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1036 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1041 __kill_fasync(sock->fasync_list, SIGIO, band);
1044 __kill_fasync(sock->fasync_list, SIGURG, band);
1049 static int __sock_create(int family, int type, int protocol,
1050 struct socket **res, int kern)
1053 struct socket *sock;
1054 const struct net_proto_family *pf;
1057 * Check protocol is in range
1059 if (family < 0 || family >= NPROTO)
1060 return -EAFNOSUPPORT;
1061 if (type < 0 || type >= SOCK_MAX)
1066 This uglymoron is moved from INET layer to here to avoid
1067 deadlock in module load.
1069 if (family == PF_INET && type == SOCK_PACKET) {
1073 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1079 err = security_socket_create(family, type, protocol, kern);
1084 * Allocate the socket and allow the family to set things up. if
1085 * the protocol is 0, the family is instructed to select an appropriate
1088 sock = sock_alloc();
1090 if (net_ratelimit())
1091 printk(KERN_WARNING "socket: no more sockets\n");
1092 return -ENFILE; /* Not exactly a match, but its the
1093 closest posix thing */
1098 #if defined(CONFIG_KMOD)
1099 /* Attempt to load a protocol module if the find failed.
1101 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1102 * requested real, full-featured networking support upon configuration.
1103 * Otherwise module support will break!
1105 if (net_families[family] == NULL)
1106 request_module("net-pf-%d", family);
1110 pf = rcu_dereference(net_families[family]);
1111 err = -EAFNOSUPPORT;
1116 * We will call the ->create function, that possibly is in a loadable
1117 * module, so we have to bump that loadable module refcnt first.
1119 if (!try_module_get(pf->owner))
1122 /* Now protected by module ref count */
1125 err = pf->create(sock, protocol);
1127 goto out_module_put;
1130 * Now to bump the refcnt of the [loadable] module that owns this
1131 * socket at sock_release time we decrement its refcnt.
1133 if (!try_module_get(sock->ops->owner))
1134 goto out_module_busy;
1137 * Now that we're done with the ->create function, the [loadable]
1138 * module can have its refcnt decremented
1140 module_put(pf->owner);
1141 err = security_socket_post_create(sock, family, type, protocol, kern);
1149 err = -EAFNOSUPPORT;
1152 module_put(pf->owner);
1159 goto out_sock_release;
1162 int sock_create(int family, int type, int protocol, struct socket **res)
1164 return __sock_create(family, type, protocol, res, 0);
1167 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1169 return __sock_create(family, type, protocol, res, 1);
1172 asmlinkage long sys_socket(int family, int type, int protocol)
1175 struct socket *sock;
1177 retval = sock_create(family, type, protocol, &sock);
1181 retval = sock_map_fd(sock);
1186 /* It may be already another descriptor 8) Not kernel problem. */
1195 * Create a pair of connected sockets.
1198 asmlinkage long sys_socketpair(int family, int type, int protocol,
1199 int __user *usockvec)
1201 struct socket *sock1, *sock2;
1205 * Obtain the first socket and check if the underlying protocol
1206 * supports the socketpair call.
1209 err = sock_create(family, type, protocol, &sock1);
1213 err = sock_create(family, type, protocol, &sock2);
1217 err = sock1->ops->socketpair(sock1, sock2);
1219 goto out_release_both;
1223 err = sock_map_fd(sock1);
1225 goto out_release_both;
1228 err = sock_map_fd(sock2);
1233 /* fd1 and fd2 may be already another descriptors.
1234 * Not kernel problem.
1237 err = put_user(fd1, &usockvec[0]);
1239 err = put_user(fd2, &usockvec[1]);
1248 sock_release(sock2);
1253 sock_release(sock2);
1255 sock_release(sock1);
1261 * Bind a name to a socket. Nothing much to do here since it's
1262 * the protocol's responsibility to handle the local address.
1264 * We move the socket address to kernel space before we call
1265 * the protocol layer (having also checked the address is ok).
1268 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1270 struct socket *sock;
1271 char address[MAX_SOCK_ADDR];
1272 int err, fput_needed;
1274 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1276 err = move_addr_to_kernel(umyaddr, addrlen, address);
1278 err = security_socket_bind(sock,
1279 (struct sockaddr *)address,
1282 err = sock->ops->bind(sock,
1286 fput_light(sock->file, fput_needed);
1292 * Perform a listen. Basically, we allow the protocol to do anything
1293 * necessary for a listen, and if that works, we mark the socket as
1294 * ready for listening.
1297 int sysctl_somaxconn __read_mostly = SOMAXCONN;
1299 asmlinkage long sys_listen(int fd, int backlog)
1301 struct socket *sock;
1302 int err, fput_needed;
1304 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1306 if ((unsigned)backlog > sysctl_somaxconn)
1307 backlog = sysctl_somaxconn;
1309 err = security_socket_listen(sock, backlog);
1311 err = sock->ops->listen(sock, backlog);
1313 fput_light(sock->file, fput_needed);
1319 * For accept, we attempt to create a new socket, set up the link
1320 * with the client, wake up the client, then return the new
1321 * connected fd. We collect the address of the connector in kernel
1322 * space and move it to user at the very end. This is unclean because
1323 * we open the socket then return an error.
1325 * 1003.1g adds the ability to recvmsg() to query connection pending
1326 * status to recvmsg. We need to add that support in a way thats
1327 * clean when we restucture accept also.
1330 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1331 int __user *upeer_addrlen)
1333 struct socket *sock, *newsock;
1334 struct file *newfile;
1335 int err, len, newfd, fput_needed;
1336 char address[MAX_SOCK_ADDR];
1338 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1343 if (!(newsock = sock_alloc()))
1346 newsock->type = sock->type;
1347 newsock->ops = sock->ops;
1350 * We don't need try_module_get here, as the listening socket (sock)
1351 * has the protocol module (sock->ops->owner) held.
1353 __module_get(newsock->ops->owner);
1355 newfd = sock_alloc_fd(&newfile);
1356 if (unlikely(newfd < 0)) {
1358 sock_release(newsock);
1362 err = sock_attach_fd(newsock, newfile);
1366 err = security_socket_accept(sock, newsock);
1370 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1374 if (upeer_sockaddr) {
1375 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1377 err = -ECONNABORTED;
1380 err = move_addr_to_user(address, len, upeer_sockaddr,
1386 /* File flags are not inherited via accept() unlike another OSes. */
1388 fd_install(newfd, newfile);
1391 security_socket_post_accept(sock, newsock);
1394 fput_light(sock->file, fput_needed);
1399 put_unused_fd(newfd);
1404 * Attempt to connect to a socket with the server address. The address
1405 * is in user space so we verify it is OK and move it to kernel space.
1407 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1410 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1411 * other SEQPACKET protocols that take time to connect() as it doesn't
1412 * include the -EINPROGRESS status for such sockets.
1415 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1418 struct socket *sock;
1419 char address[MAX_SOCK_ADDR];
1420 int err, fput_needed;
1422 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1425 err = move_addr_to_kernel(uservaddr, addrlen, address);
1430 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1434 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1435 sock->file->f_flags);
1437 fput_light(sock->file, fput_needed);
1443 * Get the local address ('name') of a socket object. Move the obtained
1444 * name to user space.
1447 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1448 int __user *usockaddr_len)
1450 struct socket *sock;
1451 char address[MAX_SOCK_ADDR];
1452 int len, err, fput_needed;
1454 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1458 err = security_socket_getsockname(sock);
1462 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1465 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1468 fput_light(sock->file, fput_needed);
1474 * Get the remote address ('name') of a socket object. Move the obtained
1475 * name to user space.
1478 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1479 int __user *usockaddr_len)
1481 struct socket *sock;
1482 char address[MAX_SOCK_ADDR];
1483 int len, err, fput_needed;
1485 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1487 err = security_socket_getpeername(sock);
1489 fput_light(sock->file, fput_needed);
1494 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1497 err = move_addr_to_user(address, len, usockaddr,
1499 fput_light(sock->file, fput_needed);
1505 * Send a datagram to a given address. We move the address into kernel
1506 * space and check the user space data area is readable before invoking
1510 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1511 unsigned flags, struct sockaddr __user *addr,
1514 struct socket *sock;
1515 char address[MAX_SOCK_ADDR];
1520 struct file *sock_file;
1522 sock_file = fget_light(fd, &fput_needed);
1526 sock = sock_from_file(sock_file, &err);
1529 iov.iov_base = buff;
1531 msg.msg_name = NULL;
1534 msg.msg_control = NULL;
1535 msg.msg_controllen = 0;
1536 msg.msg_namelen = 0;
1538 err = move_addr_to_kernel(addr, addr_len, address);
1541 msg.msg_name = address;
1542 msg.msg_namelen = addr_len;
1544 if (sock->file->f_flags & O_NONBLOCK)
1545 flags |= MSG_DONTWAIT;
1546 msg.msg_flags = flags;
1547 err = sock_sendmsg(sock, &msg, len);
1550 fput_light(sock_file, fput_needed);
1555 * Send a datagram down a socket.
1558 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1560 return sys_sendto(fd, buff, len, flags, NULL, 0);
1564 * Receive a frame from the socket and optionally record the address of the
1565 * sender. We verify the buffers are writable and if needed move the
1566 * sender address from kernel to user space.
1569 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1570 unsigned flags, struct sockaddr __user *addr,
1571 int __user *addr_len)
1573 struct socket *sock;
1576 char address[MAX_SOCK_ADDR];
1578 struct file *sock_file;
1581 sock_file = fget_light(fd, &fput_needed);
1585 sock = sock_from_file(sock_file, &err);
1589 msg.msg_control = NULL;
1590 msg.msg_controllen = 0;
1594 iov.iov_base = ubuf;
1595 msg.msg_name = address;
1596 msg.msg_namelen = MAX_SOCK_ADDR;
1597 if (sock->file->f_flags & O_NONBLOCK)
1598 flags |= MSG_DONTWAIT;
1599 err = sock_recvmsg(sock, &msg, size, flags);
1601 if (err >= 0 && addr != NULL) {
1602 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1607 fput_light(sock_file, fput_needed);
1612 * Receive a datagram from a socket.
1615 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1618 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1622 * Set a socket option. Because we don't know the option lengths we have
1623 * to pass the user mode parameter for the protocols to sort out.
1626 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1627 char __user *optval, int optlen)
1629 int err, fput_needed;
1630 struct socket *sock;
1635 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1637 err = security_socket_setsockopt(sock, level, optname);
1641 if (level == SOL_SOCKET)
1643 sock_setsockopt(sock, level, optname, optval,
1647 sock->ops->setsockopt(sock, level, optname, optval,
1650 fput_light(sock->file, fput_needed);
1656 * Get a socket option. Because we don't know the option lengths we have
1657 * to pass a user mode parameter for the protocols to sort out.
1660 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1661 char __user *optval, int __user *optlen)
1663 int err, fput_needed;
1664 struct socket *sock;
1666 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1668 err = security_socket_getsockopt(sock, level, optname);
1672 if (level == SOL_SOCKET)
1674 sock_getsockopt(sock, level, optname, optval,
1678 sock->ops->getsockopt(sock, level, optname, optval,
1681 fput_light(sock->file, fput_needed);
1687 * Shutdown a socket.
1690 asmlinkage long sys_shutdown(int fd, int how)
1692 int err, fput_needed;
1693 struct socket *sock;
1695 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1697 err = security_socket_shutdown(sock, how);
1699 err = sock->ops->shutdown(sock, how);
1700 fput_light(sock->file, fput_needed);
1705 /* A couple of helpful macros for getting the address of the 32/64 bit
1706 * fields which are the same type (int / unsigned) on our platforms.
1708 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1709 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1710 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1713 * BSD sendmsg interface
1716 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1718 struct compat_msghdr __user *msg_compat =
1719 (struct compat_msghdr __user *)msg;
1720 struct socket *sock;
1721 char address[MAX_SOCK_ADDR];
1722 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1723 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1724 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1725 /* 20 is size of ipv6_pktinfo */
1726 unsigned char *ctl_buf = ctl;
1727 struct msghdr msg_sys;
1728 int err, ctl_len, iov_size, total_len;
1732 if (MSG_CMSG_COMPAT & flags) {
1733 if (get_compat_msghdr(&msg_sys, msg_compat))
1736 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1739 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1743 /* do not move before msg_sys is valid */
1745 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1748 /* Check whether to allocate the iovec area */
1750 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1751 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1752 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1757 /* This will also move the address data into kernel space */
1758 if (MSG_CMSG_COMPAT & flags) {
1759 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1761 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1768 if (msg_sys.msg_controllen > INT_MAX)
1770 ctl_len = msg_sys.msg_controllen;
1771 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1773 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1777 ctl_buf = msg_sys.msg_control;
1778 ctl_len = msg_sys.msg_controllen;
1779 } else if (ctl_len) {
1780 if (ctl_len > sizeof(ctl)) {
1781 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1782 if (ctl_buf == NULL)
1787 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1788 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1789 * checking falls down on this.
1791 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1794 msg_sys.msg_control = ctl_buf;
1796 msg_sys.msg_flags = flags;
1798 if (sock->file->f_flags & O_NONBLOCK)
1799 msg_sys.msg_flags |= MSG_DONTWAIT;
1800 err = sock_sendmsg(sock, &msg_sys, total_len);
1804 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1806 if (iov != iovstack)
1807 sock_kfree_s(sock->sk, iov, iov_size);
1809 fput_light(sock->file, fput_needed);
1815 * BSD recvmsg interface
1818 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1821 struct compat_msghdr __user *msg_compat =
1822 (struct compat_msghdr __user *)msg;
1823 struct socket *sock;
1824 struct iovec iovstack[UIO_FASTIOV];
1825 struct iovec *iov = iovstack;
1826 struct msghdr msg_sys;
1827 unsigned long cmsg_ptr;
1828 int err, iov_size, total_len, len;
1831 /* kernel mode address */
1832 char addr[MAX_SOCK_ADDR];
1834 /* user mode address pointers */
1835 struct sockaddr __user *uaddr;
1836 int __user *uaddr_len;
1838 if (MSG_CMSG_COMPAT & flags) {
1839 if (get_compat_msghdr(&msg_sys, msg_compat))
1842 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1845 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1850 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1853 /* Check whether to allocate the iovec area */
1855 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1856 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1857 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1863 * Save the user-mode address (verify_iovec will change the
1864 * kernel msghdr to use the kernel address space)
1867 uaddr = (void __user *)msg_sys.msg_name;
1868 uaddr_len = COMPAT_NAMELEN(msg);
1869 if (MSG_CMSG_COMPAT & flags) {
1870 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1872 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1877 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1878 msg_sys.msg_flags = 0;
1879 if (MSG_CMSG_COMPAT & flags)
1880 msg_sys.msg_flags = MSG_CMSG_COMPAT;
1882 if (sock->file->f_flags & O_NONBLOCK)
1883 flags |= MSG_DONTWAIT;
1884 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1889 if (uaddr != NULL) {
1890 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1895 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1899 if (MSG_CMSG_COMPAT & flags)
1900 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1901 &msg_compat->msg_controllen);
1903 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1904 &msg->msg_controllen);
1910 if (iov != iovstack)
1911 sock_kfree_s(sock->sk, iov, iov_size);
1913 fput_light(sock->file, fput_needed);
1918 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1920 /* Argument list sizes for sys_socketcall */
1921 #define AL(x) ((x) * sizeof(unsigned long))
1922 static const unsigned char nargs[18]={
1923 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1924 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1925 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1931 * System call vectors.
1933 * Argument checking cleaned up. Saved 20% in size.
1934 * This function doesn't need to set the kernel lock because
1935 * it is set by the callees.
1938 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
1941 unsigned long a0, a1;
1944 if (call < 1 || call > SYS_RECVMSG)
1947 /* copy_from_user should be SMP safe. */
1948 if (copy_from_user(a, args, nargs[call]))
1951 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
1960 err = sys_socket(a0, a1, a[2]);
1963 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
1966 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
1969 err = sys_listen(a0, a1);
1973 sys_accept(a0, (struct sockaddr __user *)a1,
1974 (int __user *)a[2]);
1976 case SYS_GETSOCKNAME:
1978 sys_getsockname(a0, (struct sockaddr __user *)a1,
1979 (int __user *)a[2]);
1981 case SYS_GETPEERNAME:
1983 sys_getpeername(a0, (struct sockaddr __user *)a1,
1984 (int __user *)a[2]);
1986 case SYS_SOCKETPAIR:
1987 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
1990 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
1993 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
1994 (struct sockaddr __user *)a[4], a[5]);
1997 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2000 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2001 (struct sockaddr __user *)a[4],
2002 (int __user *)a[5]);
2005 err = sys_shutdown(a0, a1);
2007 case SYS_SETSOCKOPT:
2008 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2010 case SYS_GETSOCKOPT:
2012 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2013 (int __user *)a[4]);
2016 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2019 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2028 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2031 * sock_register - add a socket protocol handler
2032 * @ops: description of protocol
2034 * This function is called by a protocol handler that wants to
2035 * advertise its address family, and have it linked into the
2036 * socket interface. The value ops->family coresponds to the
2037 * socket system call protocol family.
2039 int sock_register(const struct net_proto_family *ops)
2043 if (ops->family >= NPROTO) {
2044 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2049 spin_lock(&net_family_lock);
2050 if (net_families[ops->family])
2053 net_families[ops->family] = ops;
2056 spin_unlock(&net_family_lock);
2058 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2063 * sock_unregister - remove a protocol handler
2064 * @family: protocol family to remove
2066 * This function is called by a protocol handler that wants to
2067 * remove its address family, and have it unlinked from the
2068 * new socket creation.
2070 * If protocol handler is a module, then it can use module reference
2071 * counts to protect against new references. If protocol handler is not
2072 * a module then it needs to provide its own protection in
2073 * the ops->create routine.
2075 void sock_unregister(int family)
2077 BUG_ON(family < 0 || family >= NPROTO);
2079 spin_lock(&net_family_lock);
2080 net_families[family] = NULL;
2081 spin_unlock(&net_family_lock);
2085 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2088 static int __init sock_init(void)
2091 * Initialize sock SLAB cache.
2097 * Initialize skbuff SLAB cache
2102 * Initialize the protocols module.
2106 register_filesystem(&sock_fs_type);
2107 sock_mnt = kern_mount(&sock_fs_type);
2109 /* The real protocol initialization is performed in later initcalls.
2112 #ifdef CONFIG_NETFILTER
2119 core_initcall(sock_init); /* early initcall */
2121 #ifdef CONFIG_PROC_FS
2122 void socket_seq_show(struct seq_file *seq)
2127 for_each_possible_cpu(cpu)
2128 counter += per_cpu(sockets_in_use, cpu);
2130 /* It can be negative, by the way. 8) */
2134 seq_printf(seq, "sockets: used %d\n", counter);
2136 #endif /* CONFIG_PROC_FS */
2138 #ifdef CONFIG_COMPAT
2139 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2142 struct socket *sock = file->private_data;
2143 int ret = -ENOIOCTLCMD;
2145 if (sock->ops->compat_ioctl)
2146 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2152 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2154 return sock->ops->bind(sock, addr, addrlen);
2157 int kernel_listen(struct socket *sock, int backlog)
2159 return sock->ops->listen(sock, backlog);
2162 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2164 struct sock *sk = sock->sk;
2167 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2172 err = sock->ops->accept(sock, *newsock, flags);
2174 sock_release(*newsock);
2178 (*newsock)->ops = sock->ops;
2184 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2187 return sock->ops->connect(sock, addr, addrlen, flags);
2190 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2193 return sock->ops->getname(sock, addr, addrlen, 0);
2196 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2199 return sock->ops->getname(sock, addr, addrlen, 1);
2202 int kernel_getsockopt(struct socket *sock, int level, int optname,
2203 char *optval, int *optlen)
2205 mm_segment_t oldfs = get_fs();
2209 if (level == SOL_SOCKET)
2210 err = sock_getsockopt(sock, level, optname, optval, optlen);
2212 err = sock->ops->getsockopt(sock, level, optname, optval,
2218 int kernel_setsockopt(struct socket *sock, int level, int optname,
2219 char *optval, int optlen)
2221 mm_segment_t oldfs = get_fs();
2225 if (level == SOL_SOCKET)
2226 err = sock_setsockopt(sock, level, optname, optval, optlen);
2228 err = sock->ops->setsockopt(sock, level, optname, optval,
2234 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2235 size_t size, int flags)
2237 if (sock->ops->sendpage)
2238 return sock->ops->sendpage(sock, page, offset, size, flags);
2240 return sock_no_sendpage(sock, page, offset, size, flags);
2243 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2245 mm_segment_t oldfs = get_fs();
2249 err = sock->ops->ioctl(sock, cmd, arg);
2255 /* ABI emulation layers need these two */
2256 EXPORT_SYMBOL(move_addr_to_kernel);
2257 EXPORT_SYMBOL(move_addr_to_user);
2258 EXPORT_SYMBOL(sock_create);
2259 EXPORT_SYMBOL(sock_create_kern);
2260 EXPORT_SYMBOL(sock_create_lite);
2261 EXPORT_SYMBOL(sock_map_fd);
2262 EXPORT_SYMBOL(sock_recvmsg);
2263 EXPORT_SYMBOL(sock_register);
2264 EXPORT_SYMBOL(sock_release);
2265 EXPORT_SYMBOL(sock_sendmsg);
2266 EXPORT_SYMBOL(sock_unregister);
2267 EXPORT_SYMBOL(sock_wake_async);
2268 EXPORT_SYMBOL(sockfd_lookup);
2269 EXPORT_SYMBOL(kernel_sendmsg);
2270 EXPORT_SYMBOL(kernel_recvmsg);
2271 EXPORT_SYMBOL(kernel_bind);
2272 EXPORT_SYMBOL(kernel_listen);
2273 EXPORT_SYMBOL(kernel_accept);
2274 EXPORT_SYMBOL(kernel_connect);
2275 EXPORT_SYMBOL(kernel_getsockname);
2276 EXPORT_SYMBOL(kernel_getpeername);
2277 EXPORT_SYMBOL(kernel_getsockopt);
2278 EXPORT_SYMBOL(kernel_setsockopt);
2279 EXPORT_SYMBOL(kernel_sendpage);
2280 EXPORT_SYMBOL(kernel_sock_ioctl);