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>
87 #include <linux/nsproxy.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);
115 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
116 struct pipe_inode_info *pipe, size_t len,
120 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
121 * in the operation structures but are done directly via the socketcall() multiplexor.
124 static const struct file_operations socket_file_ops = {
125 .owner = THIS_MODULE,
127 .aio_read = sock_aio_read,
128 .aio_write = sock_aio_write,
130 .unlocked_ioctl = sock_ioctl,
132 .compat_ioctl = compat_sock_ioctl,
135 .open = sock_no_open, /* special open code to disallow open via /proc */
136 .release = sock_close,
137 .fasync = sock_fasync,
138 .sendpage = sock_sendpage,
139 .splice_write = generic_splice_sendpage,
140 .splice_read = sock_splice_read,
144 * The protocol list. Each protocol is registered in here.
147 static DEFINE_SPINLOCK(net_family_lock);
148 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
151 * Statistics counters of the socket lists
154 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
158 * Move socket addresses back and forth across the kernel/user
159 * divide and look after the messy bits.
162 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
163 16 for IP, 16 for IPX,
166 must be at least one bigger than
167 the AF_UNIX size (see net/unix/af_unix.c
172 * move_addr_to_kernel - copy a socket address into kernel space
173 * @uaddr: Address in user space
174 * @kaddr: Address in kernel space
175 * @ulen: Length in user space
177 * The address is copied into kernel space. If the provided address is
178 * too long an error code of -EINVAL is returned. If the copy gives
179 * invalid addresses -EFAULT is returned. On a success 0 is returned.
182 int move_addr_to_kernel(void __user *uaddr, int ulen, void *kaddr)
184 if (ulen < 0 || ulen > MAX_SOCK_ADDR)
188 if (copy_from_user(kaddr, uaddr, ulen))
190 return audit_sockaddr(ulen, kaddr);
194 * move_addr_to_user - copy an address to user space
195 * @kaddr: kernel space address
196 * @klen: length of address in kernel
197 * @uaddr: user space address
198 * @ulen: pointer to user length field
200 * The value pointed to by ulen on entry is the buffer length available.
201 * This is overwritten with the buffer space used. -EINVAL is returned
202 * if an overlong buffer is specified or a negative buffer size. -EFAULT
203 * is returned if either the buffer or the length field are not
205 * After copying the data up to the limit the user specifies, the true
206 * length of the data is written over the length limit the user
207 * specified. Zero is returned for a success.
210 int move_addr_to_user(void *kaddr, int klen, void __user *uaddr,
216 err = get_user(len, ulen);
221 if (len < 0 || len > MAX_SOCK_ADDR)
224 if (audit_sockaddr(klen, kaddr))
226 if (copy_to_user(uaddr, kaddr, len))
230 * "fromlen shall refer to the value before truncation.."
233 return __put_user(klen, ulen);
236 #define SOCKFS_MAGIC 0x534F434B
238 static struct kmem_cache *sock_inode_cachep __read_mostly;
240 static struct inode *sock_alloc_inode(struct super_block *sb)
242 struct socket_alloc *ei;
244 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
247 init_waitqueue_head(&ei->socket.wait);
249 ei->socket.fasync_list = NULL;
250 ei->socket.state = SS_UNCONNECTED;
251 ei->socket.flags = 0;
252 ei->socket.ops = NULL;
253 ei->socket.sk = NULL;
254 ei->socket.file = NULL;
256 return &ei->vfs_inode;
259 static void sock_destroy_inode(struct inode *inode)
261 kmem_cache_free(sock_inode_cachep,
262 container_of(inode, struct socket_alloc, vfs_inode));
265 static void init_once(struct kmem_cache *cachep, void *foo)
267 struct socket_alloc *ei = (struct socket_alloc *)foo;
269 inode_init_once(&ei->vfs_inode);
272 static int init_inodecache(void)
274 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
275 sizeof(struct socket_alloc),
277 (SLAB_HWCACHE_ALIGN |
278 SLAB_RECLAIM_ACCOUNT |
281 if (sock_inode_cachep == NULL)
286 static struct super_operations sockfs_ops = {
287 .alloc_inode = sock_alloc_inode,
288 .destroy_inode =sock_destroy_inode,
289 .statfs = simple_statfs,
292 static int sockfs_get_sb(struct file_system_type *fs_type,
293 int flags, const char *dev_name, void *data,
294 struct vfsmount *mnt)
296 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
300 static struct vfsmount *sock_mnt __read_mostly;
302 static struct file_system_type sock_fs_type = {
304 .get_sb = sockfs_get_sb,
305 .kill_sb = kill_anon_super,
308 static int sockfs_delete_dentry(struct dentry *dentry)
311 * At creation time, we pretended this dentry was hashed
312 * (by clearing DCACHE_UNHASHED bit in d_flags)
313 * At delete time, we restore the truth : not hashed.
314 * (so that dput() can proceed correctly)
316 dentry->d_flags |= DCACHE_UNHASHED;
321 * sockfs_dname() is called from d_path().
323 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
325 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
326 dentry->d_inode->i_ino);
329 static struct dentry_operations sockfs_dentry_operations = {
330 .d_delete = sockfs_delete_dentry,
331 .d_dname = sockfs_dname,
335 * Obtains the first available file descriptor and sets it up for use.
337 * These functions create file structures and maps them to fd space
338 * of the current process. On success it returns file descriptor
339 * and file struct implicitly stored in sock->file.
340 * Note that another thread may close file descriptor before we return
341 * from this function. We use the fact that now we do not refer
342 * to socket after mapping. If one day we will need it, this
343 * function will increment ref. count on file by 1.
345 * In any case returned fd MAY BE not valid!
346 * This race condition is unavoidable
347 * with shared fd spaces, we cannot solve it inside kernel,
348 * but we take care of internal coherence yet.
351 static int sock_alloc_fd(struct file **filep)
355 fd = get_unused_fd();
356 if (likely(fd >= 0)) {
357 struct file *file = get_empty_filp();
360 if (unlikely(!file)) {
369 static int sock_attach_fd(struct socket *sock, struct file *file)
371 struct dentry *dentry;
372 struct qstr name = { .name = "" };
374 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
375 if (unlikely(!dentry))
378 dentry->d_op = &sockfs_dentry_operations;
380 * We dont want to push this dentry into global dentry hash table.
381 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
382 * This permits a working /proc/$pid/fd/XXX on sockets
384 dentry->d_flags &= ~DCACHE_UNHASHED;
385 d_instantiate(dentry, SOCK_INODE(sock));
388 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
390 SOCK_INODE(sock)->i_fop = &socket_file_ops;
391 file->f_flags = O_RDWR;
393 file->private_data = sock;
398 int sock_map_fd(struct socket *sock)
400 struct file *newfile;
401 int fd = sock_alloc_fd(&newfile);
403 if (likely(fd >= 0)) {
404 int err = sock_attach_fd(sock, newfile);
406 if (unlikely(err < 0)) {
411 fd_install(fd, newfile);
416 static struct socket *sock_from_file(struct file *file, int *err)
418 if (file->f_op == &socket_file_ops)
419 return file->private_data; /* set in sock_map_fd */
426 * sockfd_lookup - Go from a file number to its socket slot
428 * @err: pointer to an error code return
430 * The file handle passed in is locked and the socket it is bound
431 * too is returned. If an error occurs the err pointer is overwritten
432 * with a negative errno code and NULL is returned. The function checks
433 * for both invalid handles and passing a handle which is not a socket.
435 * On a success the socket object pointer is returned.
438 struct socket *sockfd_lookup(int fd, int *err)
449 sock = sock_from_file(file, err);
455 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
461 file = fget_light(fd, fput_needed);
463 sock = sock_from_file(file, err);
466 fput_light(file, *fput_needed);
472 * sock_alloc - allocate a socket
474 * Allocate a new inode and socket object. The two are bound together
475 * and initialised. The socket is then returned. If we are out of inodes
479 static struct socket *sock_alloc(void)
484 inode = new_inode(sock_mnt->mnt_sb);
488 sock = SOCKET_I(inode);
490 inode->i_mode = S_IFSOCK | S_IRWXUGO;
491 inode->i_uid = current->fsuid;
492 inode->i_gid = current->fsgid;
494 get_cpu_var(sockets_in_use)++;
495 put_cpu_var(sockets_in_use);
500 * In theory you can't get an open on this inode, but /proc provides
501 * a back door. Remember to keep it shut otherwise you'll let the
502 * creepy crawlies in.
505 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
510 const struct file_operations bad_sock_fops = {
511 .owner = THIS_MODULE,
512 .open = sock_no_open,
516 * sock_release - close a socket
517 * @sock: socket to close
519 * The socket is released from the protocol stack if it has a release
520 * callback, and the inode is then released if the socket is bound to
521 * an inode not a file.
524 void sock_release(struct socket *sock)
527 struct module *owner = sock->ops->owner;
529 sock->ops->release(sock);
534 if (sock->fasync_list)
535 printk(KERN_ERR "sock_release: fasync list not empty!\n");
537 get_cpu_var(sockets_in_use)--;
538 put_cpu_var(sockets_in_use);
540 iput(SOCK_INODE(sock));
546 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
547 struct msghdr *msg, size_t size)
549 struct sock_iocb *si = kiocb_to_siocb(iocb);
557 err = security_socket_sendmsg(sock, msg, size);
561 return sock->ops->sendmsg(iocb, sock, msg, size);
564 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
567 struct sock_iocb siocb;
570 init_sync_kiocb(&iocb, NULL);
571 iocb.private = &siocb;
572 ret = __sock_sendmsg(&iocb, sock, msg, size);
573 if (-EIOCBQUEUED == ret)
574 ret = wait_on_sync_kiocb(&iocb);
578 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
579 struct kvec *vec, size_t num, size_t size)
581 mm_segment_t oldfs = get_fs();
586 * the following is safe, since for compiler definitions of kvec and
587 * iovec are identical, yielding the same in-core layout and alignment
589 msg->msg_iov = (struct iovec *)vec;
590 msg->msg_iovlen = num;
591 result = sock_sendmsg(sock, msg, size);
597 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
599 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
602 ktime_t kt = skb->tstamp;
604 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
606 /* Race occurred between timestamp enabling and packet
607 receiving. Fill in the current time for now. */
609 kt = ktime_get_real();
611 tv = ktime_to_timeval(kt);
612 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
615 /* Race occurred between timestamp enabling and packet
616 receiving. Fill in the current time for now. */
618 kt = ktime_get_real();
620 ts = ktime_to_timespec(kt);
621 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
625 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
627 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
628 struct msghdr *msg, size_t size, int flags)
631 struct sock_iocb *si = kiocb_to_siocb(iocb);
639 err = security_socket_recvmsg(sock, msg, size, flags);
643 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
646 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
647 size_t size, int flags)
650 struct sock_iocb siocb;
653 init_sync_kiocb(&iocb, NULL);
654 iocb.private = &siocb;
655 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
656 if (-EIOCBQUEUED == ret)
657 ret = wait_on_sync_kiocb(&iocb);
661 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
662 struct kvec *vec, size_t num, size_t size, int flags)
664 mm_segment_t oldfs = get_fs();
669 * the following is safe, since for compiler definitions of kvec and
670 * iovec are identical, yielding the same in-core layout and alignment
672 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
673 result = sock_recvmsg(sock, msg, size, flags);
678 static void sock_aio_dtor(struct kiocb *iocb)
680 kfree(iocb->private);
683 static ssize_t sock_sendpage(struct file *file, struct page *page,
684 int offset, size_t size, loff_t *ppos, int more)
689 sock = file->private_data;
691 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
695 return sock->ops->sendpage(sock, page, offset, size, flags);
698 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
699 struct pipe_inode_info *pipe, size_t len,
702 struct socket *sock = file->private_data;
704 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
707 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
708 struct sock_iocb *siocb)
710 if (!is_sync_kiocb(iocb)) {
711 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
714 iocb->ki_dtor = sock_aio_dtor;
718 iocb->private = siocb;
722 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
723 struct file *file, const struct iovec *iov,
724 unsigned long nr_segs)
726 struct socket *sock = file->private_data;
730 for (i = 0; i < nr_segs; i++)
731 size += iov[i].iov_len;
733 msg->msg_name = NULL;
734 msg->msg_namelen = 0;
735 msg->msg_control = NULL;
736 msg->msg_controllen = 0;
737 msg->msg_iov = (struct iovec *)iov;
738 msg->msg_iovlen = nr_segs;
739 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
741 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
744 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
745 unsigned long nr_segs, loff_t pos)
747 struct sock_iocb siocb, *x;
752 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
756 x = alloc_sock_iocb(iocb, &siocb);
759 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
762 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
763 struct file *file, const struct iovec *iov,
764 unsigned long nr_segs)
766 struct socket *sock = file->private_data;
770 for (i = 0; i < nr_segs; i++)
771 size += iov[i].iov_len;
773 msg->msg_name = NULL;
774 msg->msg_namelen = 0;
775 msg->msg_control = NULL;
776 msg->msg_controllen = 0;
777 msg->msg_iov = (struct iovec *)iov;
778 msg->msg_iovlen = nr_segs;
779 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
780 if (sock->type == SOCK_SEQPACKET)
781 msg->msg_flags |= MSG_EOR;
783 return __sock_sendmsg(iocb, sock, msg, size);
786 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
787 unsigned long nr_segs, loff_t pos)
789 struct sock_iocb siocb, *x;
794 x = alloc_sock_iocb(iocb, &siocb);
798 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
802 * Atomic setting of ioctl hooks to avoid race
803 * with module unload.
806 static DEFINE_MUTEX(br_ioctl_mutex);
807 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
809 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
811 mutex_lock(&br_ioctl_mutex);
812 br_ioctl_hook = hook;
813 mutex_unlock(&br_ioctl_mutex);
816 EXPORT_SYMBOL(brioctl_set);
818 static DEFINE_MUTEX(vlan_ioctl_mutex);
819 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
821 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
823 mutex_lock(&vlan_ioctl_mutex);
824 vlan_ioctl_hook = hook;
825 mutex_unlock(&vlan_ioctl_mutex);
828 EXPORT_SYMBOL(vlan_ioctl_set);
830 static DEFINE_MUTEX(dlci_ioctl_mutex);
831 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
833 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
835 mutex_lock(&dlci_ioctl_mutex);
836 dlci_ioctl_hook = hook;
837 mutex_unlock(&dlci_ioctl_mutex);
840 EXPORT_SYMBOL(dlci_ioctl_set);
843 * With an ioctl, arg may well be a user mode pointer, but we don't know
844 * what to do with it - that's up to the protocol still.
847 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
851 void __user *argp = (void __user *)arg;
855 sock = file->private_data;
858 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
859 err = dev_ioctl(net, cmd, argp);
861 #ifdef CONFIG_WIRELESS_EXT
862 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
863 err = dev_ioctl(net, cmd, argp);
865 #endif /* CONFIG_WIRELESS_EXT */
870 if (get_user(pid, (int __user *)argp))
872 err = f_setown(sock->file, pid, 1);
876 err = put_user(f_getown(sock->file),
885 request_module("bridge");
887 mutex_lock(&br_ioctl_mutex);
889 err = br_ioctl_hook(net, cmd, argp);
890 mutex_unlock(&br_ioctl_mutex);
895 if (!vlan_ioctl_hook)
896 request_module("8021q");
898 mutex_lock(&vlan_ioctl_mutex);
900 err = vlan_ioctl_hook(net, argp);
901 mutex_unlock(&vlan_ioctl_mutex);
906 if (!dlci_ioctl_hook)
907 request_module("dlci");
909 if (dlci_ioctl_hook) {
910 mutex_lock(&dlci_ioctl_mutex);
911 err = dlci_ioctl_hook(cmd, argp);
912 mutex_unlock(&dlci_ioctl_mutex);
916 err = sock->ops->ioctl(sock, cmd, arg);
919 * If this ioctl is unknown try to hand it down
922 if (err == -ENOIOCTLCMD)
923 err = dev_ioctl(net, cmd, argp);
929 int sock_create_lite(int family, int type, int protocol, struct socket **res)
932 struct socket *sock = NULL;
934 err = security_socket_create(family, type, protocol, 1);
945 err = security_socket_post_create(sock, family, type, protocol, 1);
958 /* No kernel lock held - perfect */
959 static unsigned int sock_poll(struct file *file, poll_table *wait)
964 * We can't return errors to poll, so it's either yes or no.
966 sock = file->private_data;
967 return sock->ops->poll(file, sock, wait);
970 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
972 struct socket *sock = file->private_data;
974 return sock->ops->mmap(file, sock, vma);
977 static int sock_close(struct inode *inode, struct file *filp)
980 * It was possible the inode is NULL we were
981 * closing an unfinished socket.
985 printk(KERN_DEBUG "sock_close: NULL inode\n");
988 sock_fasync(-1, filp, 0);
989 sock_release(SOCKET_I(inode));
994 * Update the socket async list
996 * Fasync_list locking strategy.
998 * 1. fasync_list is modified only under process context socket lock
999 * i.e. under semaphore.
1000 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1001 * or under socket lock.
1002 * 3. fasync_list can be used from softirq context, so that
1003 * modification under socket lock have to be enhanced with
1004 * write_lock_bh(&sk->sk_callback_lock).
1008 static int sock_fasync(int fd, struct file *filp, int on)
1010 struct fasync_struct *fa, *fna = NULL, **prev;
1011 struct socket *sock;
1015 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1020 sock = filp->private_data;
1030 prev = &(sock->fasync_list);
1032 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1033 if (fa->fa_file == filp)
1038 write_lock_bh(&sk->sk_callback_lock);
1040 write_unlock_bh(&sk->sk_callback_lock);
1045 fna->fa_file = filp;
1047 fna->magic = FASYNC_MAGIC;
1048 fna->fa_next = sock->fasync_list;
1049 write_lock_bh(&sk->sk_callback_lock);
1050 sock->fasync_list = fna;
1051 write_unlock_bh(&sk->sk_callback_lock);
1054 write_lock_bh(&sk->sk_callback_lock);
1055 *prev = fa->fa_next;
1056 write_unlock_bh(&sk->sk_callback_lock);
1062 release_sock(sock->sk);
1066 /* This function may be called only under socket lock or callback_lock */
1068 int sock_wake_async(struct socket *sock, int how, int band)
1070 if (!sock || !sock->fasync_list)
1073 case SOCK_WAKE_WAITD:
1074 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1077 case SOCK_WAKE_SPACE:
1078 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1083 __kill_fasync(sock->fasync_list, SIGIO, band);
1086 __kill_fasync(sock->fasync_list, SIGURG, band);
1091 static int __sock_create(struct net *net, int family, int type, int protocol,
1092 struct socket **res, int kern)
1095 struct socket *sock;
1096 const struct net_proto_family *pf;
1099 * Check protocol is in range
1101 if (family < 0 || family >= NPROTO)
1102 return -EAFNOSUPPORT;
1103 if (type < 0 || type >= SOCK_MAX)
1108 This uglymoron is moved from INET layer to here to avoid
1109 deadlock in module load.
1111 if (family == PF_INET && type == SOCK_PACKET) {
1115 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1121 err = security_socket_create(family, type, protocol, kern);
1126 * Allocate the socket and allow the family to set things up. if
1127 * the protocol is 0, the family is instructed to select an appropriate
1130 sock = sock_alloc();
1132 if (net_ratelimit())
1133 printk(KERN_WARNING "socket: no more sockets\n");
1134 return -ENFILE; /* Not exactly a match, but its the
1135 closest posix thing */
1140 #if defined(CONFIG_KMOD)
1141 /* Attempt to load a protocol module if the find failed.
1143 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1144 * requested real, full-featured networking support upon configuration.
1145 * Otherwise module support will break!
1147 if (net_families[family] == NULL)
1148 request_module("net-pf-%d", family);
1152 pf = rcu_dereference(net_families[family]);
1153 err = -EAFNOSUPPORT;
1158 * We will call the ->create function, that possibly is in a loadable
1159 * module, so we have to bump that loadable module refcnt first.
1161 if (!try_module_get(pf->owner))
1164 /* Now protected by module ref count */
1167 err = pf->create(net, sock, protocol);
1169 goto out_module_put;
1172 * Now to bump the refcnt of the [loadable] module that owns this
1173 * socket at sock_release time we decrement its refcnt.
1175 if (!try_module_get(sock->ops->owner))
1176 goto out_module_busy;
1179 * Now that we're done with the ->create function, the [loadable]
1180 * module can have its refcnt decremented
1182 module_put(pf->owner);
1183 err = security_socket_post_create(sock, family, type, protocol, kern);
1185 goto out_sock_release;
1191 err = -EAFNOSUPPORT;
1194 module_put(pf->owner);
1201 goto out_sock_release;
1204 int sock_create(int family, int type, int protocol, struct socket **res)
1206 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1209 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1211 return __sock_create(&init_net, family, type, protocol, res, 1);
1214 asmlinkage long sys_socket(int family, int type, int protocol)
1217 struct socket *sock;
1219 retval = sock_create(family, type, protocol, &sock);
1223 retval = sock_map_fd(sock);
1228 /* It may be already another descriptor 8) Not kernel problem. */
1237 * Create a pair of connected sockets.
1240 asmlinkage long sys_socketpair(int family, int type, int protocol,
1241 int __user *usockvec)
1243 struct socket *sock1, *sock2;
1245 struct file *newfile1, *newfile2;
1248 * Obtain the first socket and check if the underlying protocol
1249 * supports the socketpair call.
1252 err = sock_create(family, type, protocol, &sock1);
1256 err = sock_create(family, type, protocol, &sock2);
1260 err = sock1->ops->socketpair(sock1, sock2);
1262 goto out_release_both;
1264 fd1 = sock_alloc_fd(&newfile1);
1265 if (unlikely(fd1 < 0)) {
1267 goto out_release_both;
1270 fd2 = sock_alloc_fd(&newfile2);
1271 if (unlikely(fd2 < 0)) {
1275 goto out_release_both;
1278 err = sock_attach_fd(sock1, newfile1);
1279 if (unlikely(err < 0)) {
1283 err = sock_attach_fd(sock2, newfile2);
1284 if (unlikely(err < 0)) {
1289 err = audit_fd_pair(fd1, fd2);
1296 fd_install(fd1, newfile1);
1297 fd_install(fd2, newfile2);
1298 /* fd1 and fd2 may be already another descriptors.
1299 * Not kernel problem.
1302 err = put_user(fd1, &usockvec[0]);
1304 err = put_user(fd2, &usockvec[1]);
1313 sock_release(sock2);
1315 sock_release(sock1);
1321 sock_release(sock1);
1324 sock_release(sock2);
1332 * Bind a name to a socket. Nothing much to do here since it's
1333 * the protocol's responsibility to handle the local address.
1335 * We move the socket address to kernel space before we call
1336 * the protocol layer (having also checked the address is ok).
1339 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1341 struct socket *sock;
1342 char address[MAX_SOCK_ADDR];
1343 int err, fput_needed;
1345 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1347 err = move_addr_to_kernel(umyaddr, addrlen, address);
1349 err = security_socket_bind(sock,
1350 (struct sockaddr *)address,
1353 err = sock->ops->bind(sock,
1357 fput_light(sock->file, fput_needed);
1363 * Perform a listen. Basically, we allow the protocol to do anything
1364 * necessary for a listen, and if that works, we mark the socket as
1365 * ready for listening.
1368 asmlinkage long sys_listen(int fd, int backlog)
1370 struct socket *sock;
1371 int err, fput_needed;
1374 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1376 somaxconn = sock->sk->sk_net->sysctl_somaxconn;
1377 if ((unsigned)backlog > somaxconn)
1378 backlog = somaxconn;
1380 err = security_socket_listen(sock, backlog);
1382 err = sock->ops->listen(sock, backlog);
1384 fput_light(sock->file, fput_needed);
1390 * For accept, we attempt to create a new socket, set up the link
1391 * with the client, wake up the client, then return the new
1392 * connected fd. We collect the address of the connector in kernel
1393 * space and move it to user at the very end. This is unclean because
1394 * we open the socket then return an error.
1396 * 1003.1g adds the ability to recvmsg() to query connection pending
1397 * status to recvmsg. We need to add that support in a way thats
1398 * clean when we restucture accept also.
1401 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1402 int __user *upeer_addrlen)
1404 struct socket *sock, *newsock;
1405 struct file *newfile;
1406 int err, len, newfd, fput_needed;
1407 char address[MAX_SOCK_ADDR];
1409 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1414 if (!(newsock = sock_alloc()))
1417 newsock->type = sock->type;
1418 newsock->ops = sock->ops;
1421 * We don't need try_module_get here, as the listening socket (sock)
1422 * has the protocol module (sock->ops->owner) held.
1424 __module_get(newsock->ops->owner);
1426 newfd = sock_alloc_fd(&newfile);
1427 if (unlikely(newfd < 0)) {
1429 sock_release(newsock);
1433 err = sock_attach_fd(newsock, newfile);
1437 err = security_socket_accept(sock, newsock);
1441 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1445 if (upeer_sockaddr) {
1446 if (newsock->ops->getname(newsock, (struct sockaddr *)address,
1448 err = -ECONNABORTED;
1451 err = move_addr_to_user(address, len, upeer_sockaddr,
1457 /* File flags are not inherited via accept() unlike another OSes. */
1459 fd_install(newfd, newfile);
1462 security_socket_post_accept(sock, newsock);
1465 fput_light(sock->file, fput_needed);
1469 sock_release(newsock);
1471 put_unused_fd(newfd);
1475 put_unused_fd(newfd);
1480 * Attempt to connect to a socket with the server address. The address
1481 * is in user space so we verify it is OK and move it to kernel space.
1483 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1486 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1487 * other SEQPACKET protocols that take time to connect() as it doesn't
1488 * include the -EINPROGRESS status for such sockets.
1491 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1494 struct socket *sock;
1495 char address[MAX_SOCK_ADDR];
1496 int err, fput_needed;
1498 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1501 err = move_addr_to_kernel(uservaddr, addrlen, address);
1506 security_socket_connect(sock, (struct sockaddr *)address, addrlen);
1510 err = sock->ops->connect(sock, (struct sockaddr *)address, addrlen,
1511 sock->file->f_flags);
1513 fput_light(sock->file, fput_needed);
1519 * Get the local address ('name') of a socket object. Move the obtained
1520 * name to user space.
1523 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1524 int __user *usockaddr_len)
1526 struct socket *sock;
1527 char address[MAX_SOCK_ADDR];
1528 int len, err, fput_needed;
1530 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1534 err = security_socket_getsockname(sock);
1538 err = sock->ops->getname(sock, (struct sockaddr *)address, &len, 0);
1541 err = move_addr_to_user(address, len, usockaddr, usockaddr_len);
1544 fput_light(sock->file, fput_needed);
1550 * Get the remote address ('name') of a socket object. Move the obtained
1551 * name to user space.
1554 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1555 int __user *usockaddr_len)
1557 struct socket *sock;
1558 char address[MAX_SOCK_ADDR];
1559 int len, err, fput_needed;
1561 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1563 err = security_socket_getpeername(sock);
1565 fput_light(sock->file, fput_needed);
1570 sock->ops->getname(sock, (struct sockaddr *)address, &len,
1573 err = move_addr_to_user(address, len, usockaddr,
1575 fput_light(sock->file, fput_needed);
1581 * Send a datagram to a given address. We move the address into kernel
1582 * space and check the user space data area is readable before invoking
1586 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1587 unsigned flags, struct sockaddr __user *addr,
1590 struct socket *sock;
1591 char address[MAX_SOCK_ADDR];
1597 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1601 iov.iov_base = buff;
1603 msg.msg_name = NULL;
1606 msg.msg_control = NULL;
1607 msg.msg_controllen = 0;
1608 msg.msg_namelen = 0;
1610 err = move_addr_to_kernel(addr, addr_len, address);
1613 msg.msg_name = address;
1614 msg.msg_namelen = addr_len;
1616 if (sock->file->f_flags & O_NONBLOCK)
1617 flags |= MSG_DONTWAIT;
1618 msg.msg_flags = flags;
1619 err = sock_sendmsg(sock, &msg, len);
1622 fput_light(sock->file, fput_needed);
1628 * Send a datagram down a socket.
1631 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1633 return sys_sendto(fd, buff, len, flags, NULL, 0);
1637 * Receive a frame from the socket and optionally record the address of the
1638 * sender. We verify the buffers are writable and if needed move the
1639 * sender address from kernel to user space.
1642 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1643 unsigned flags, struct sockaddr __user *addr,
1644 int __user *addr_len)
1646 struct socket *sock;
1649 char address[MAX_SOCK_ADDR];
1653 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1657 msg.msg_control = NULL;
1658 msg.msg_controllen = 0;
1662 iov.iov_base = ubuf;
1663 msg.msg_name = address;
1664 msg.msg_namelen = MAX_SOCK_ADDR;
1665 if (sock->file->f_flags & O_NONBLOCK)
1666 flags |= MSG_DONTWAIT;
1667 err = sock_recvmsg(sock, &msg, size, flags);
1669 if (err >= 0 && addr != NULL) {
1670 err2 = move_addr_to_user(address, msg.msg_namelen, addr, addr_len);
1675 fput_light(sock->file, fput_needed);
1681 * Receive a datagram from a socket.
1684 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1687 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1691 * Set a socket option. Because we don't know the option lengths we have
1692 * to pass the user mode parameter for the protocols to sort out.
1695 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1696 char __user *optval, int optlen)
1698 int err, fput_needed;
1699 struct socket *sock;
1704 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1706 err = security_socket_setsockopt(sock, level, optname);
1710 if (level == SOL_SOCKET)
1712 sock_setsockopt(sock, level, optname, optval,
1716 sock->ops->setsockopt(sock, level, optname, optval,
1719 fput_light(sock->file, fput_needed);
1725 * Get a socket option. Because we don't know the option lengths we have
1726 * to pass a user mode parameter for the protocols to sort out.
1729 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1730 char __user *optval, int __user *optlen)
1732 int err, fput_needed;
1733 struct socket *sock;
1735 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1737 err = security_socket_getsockopt(sock, level, optname);
1741 if (level == SOL_SOCKET)
1743 sock_getsockopt(sock, level, optname, optval,
1747 sock->ops->getsockopt(sock, level, optname, optval,
1750 fput_light(sock->file, fput_needed);
1756 * Shutdown a socket.
1759 asmlinkage long sys_shutdown(int fd, int how)
1761 int err, fput_needed;
1762 struct socket *sock;
1764 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1766 err = security_socket_shutdown(sock, how);
1768 err = sock->ops->shutdown(sock, how);
1769 fput_light(sock->file, fput_needed);
1774 /* A couple of helpful macros for getting the address of the 32/64 bit
1775 * fields which are the same type (int / unsigned) on our platforms.
1777 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1778 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1779 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1782 * BSD sendmsg interface
1785 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1787 struct compat_msghdr __user *msg_compat =
1788 (struct compat_msghdr __user *)msg;
1789 struct socket *sock;
1790 char address[MAX_SOCK_ADDR];
1791 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1792 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1793 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1794 /* 20 is size of ipv6_pktinfo */
1795 unsigned char *ctl_buf = ctl;
1796 struct msghdr msg_sys;
1797 int err, ctl_len, iov_size, total_len;
1801 if (MSG_CMSG_COMPAT & flags) {
1802 if (get_compat_msghdr(&msg_sys, msg_compat))
1805 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1808 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1812 /* do not move before msg_sys is valid */
1814 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1817 /* Check whether to allocate the iovec area */
1819 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1820 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1821 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1826 /* This will also move the address data into kernel space */
1827 if (MSG_CMSG_COMPAT & flags) {
1828 err = verify_compat_iovec(&msg_sys, iov, address, VERIFY_READ);
1830 err = verify_iovec(&msg_sys, iov, address, VERIFY_READ);
1837 if (msg_sys.msg_controllen > INT_MAX)
1839 ctl_len = msg_sys.msg_controllen;
1840 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1842 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1846 ctl_buf = msg_sys.msg_control;
1847 ctl_len = msg_sys.msg_controllen;
1848 } else if (ctl_len) {
1849 if (ctl_len > sizeof(ctl)) {
1850 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1851 if (ctl_buf == NULL)
1856 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1857 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1858 * checking falls down on this.
1860 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1863 msg_sys.msg_control = ctl_buf;
1865 msg_sys.msg_flags = flags;
1867 if (sock->file->f_flags & O_NONBLOCK)
1868 msg_sys.msg_flags |= MSG_DONTWAIT;
1869 err = sock_sendmsg(sock, &msg_sys, total_len);
1873 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1875 if (iov != iovstack)
1876 sock_kfree_s(sock->sk, iov, iov_size);
1878 fput_light(sock->file, fput_needed);
1884 * BSD recvmsg interface
1887 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1890 struct compat_msghdr __user *msg_compat =
1891 (struct compat_msghdr __user *)msg;
1892 struct socket *sock;
1893 struct iovec iovstack[UIO_FASTIOV];
1894 struct iovec *iov = iovstack;
1895 struct msghdr msg_sys;
1896 unsigned long cmsg_ptr;
1897 int err, iov_size, total_len, len;
1900 /* kernel mode address */
1901 char addr[MAX_SOCK_ADDR];
1903 /* user mode address pointers */
1904 struct sockaddr __user *uaddr;
1905 int __user *uaddr_len;
1907 if (MSG_CMSG_COMPAT & flags) {
1908 if (get_compat_msghdr(&msg_sys, msg_compat))
1911 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1914 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1919 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1922 /* Check whether to allocate the iovec area */
1924 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1925 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1926 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1932 * Save the user-mode address (verify_iovec will change the
1933 * kernel msghdr to use the kernel address space)
1936 uaddr = (__force void __user *)msg_sys.msg_name;
1937 uaddr_len = COMPAT_NAMELEN(msg);
1938 if (MSG_CMSG_COMPAT & flags) {
1939 err = verify_compat_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1941 err = verify_iovec(&msg_sys, iov, addr, VERIFY_WRITE);
1946 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1947 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
1949 if (sock->file->f_flags & O_NONBLOCK)
1950 flags |= MSG_DONTWAIT;
1951 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
1956 if (uaddr != NULL) {
1957 err = move_addr_to_user(addr, msg_sys.msg_namelen, uaddr,
1962 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
1966 if (MSG_CMSG_COMPAT & flags)
1967 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1968 &msg_compat->msg_controllen);
1970 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
1971 &msg->msg_controllen);
1977 if (iov != iovstack)
1978 sock_kfree_s(sock->sk, iov, iov_size);
1980 fput_light(sock->file, fput_needed);
1985 #ifdef __ARCH_WANT_SYS_SOCKETCALL
1987 /* Argument list sizes for sys_socketcall */
1988 #define AL(x) ((x) * sizeof(unsigned long))
1989 static const unsigned char nargs[18]={
1990 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
1991 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
1992 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3)
1998 * System call vectors.
2000 * Argument checking cleaned up. Saved 20% in size.
2001 * This function doesn't need to set the kernel lock because
2002 * it is set by the callees.
2005 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2008 unsigned long a0, a1;
2011 if (call < 1 || call > SYS_RECVMSG)
2014 /* copy_from_user should be SMP safe. */
2015 if (copy_from_user(a, args, nargs[call]))
2018 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2027 err = sys_socket(a0, a1, a[2]);
2030 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2033 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2036 err = sys_listen(a0, a1);
2040 sys_accept(a0, (struct sockaddr __user *)a1,
2041 (int __user *)a[2]);
2043 case SYS_GETSOCKNAME:
2045 sys_getsockname(a0, (struct sockaddr __user *)a1,
2046 (int __user *)a[2]);
2048 case SYS_GETPEERNAME:
2050 sys_getpeername(a0, (struct sockaddr __user *)a1,
2051 (int __user *)a[2]);
2053 case SYS_SOCKETPAIR:
2054 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2057 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2060 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2061 (struct sockaddr __user *)a[4], a[5]);
2064 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2067 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2068 (struct sockaddr __user *)a[4],
2069 (int __user *)a[5]);
2072 err = sys_shutdown(a0, a1);
2074 case SYS_SETSOCKOPT:
2075 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2077 case SYS_GETSOCKOPT:
2079 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2080 (int __user *)a[4]);
2083 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2086 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2095 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2098 * sock_register - add a socket protocol handler
2099 * @ops: description of protocol
2101 * This function is called by a protocol handler that wants to
2102 * advertise its address family, and have it linked into the
2103 * socket interface. The value ops->family coresponds to the
2104 * socket system call protocol family.
2106 int sock_register(const struct net_proto_family *ops)
2110 if (ops->family >= NPROTO) {
2111 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2116 spin_lock(&net_family_lock);
2117 if (net_families[ops->family])
2120 net_families[ops->family] = ops;
2123 spin_unlock(&net_family_lock);
2125 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2130 * sock_unregister - remove a protocol handler
2131 * @family: protocol family to remove
2133 * This function is called by a protocol handler that wants to
2134 * remove its address family, and have it unlinked from the
2135 * new socket creation.
2137 * If protocol handler is a module, then it can use module reference
2138 * counts to protect against new references. If protocol handler is not
2139 * a module then it needs to provide its own protection in
2140 * the ops->create routine.
2142 void sock_unregister(int family)
2144 BUG_ON(family < 0 || family >= NPROTO);
2146 spin_lock(&net_family_lock);
2147 net_families[family] = NULL;
2148 spin_unlock(&net_family_lock);
2152 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2155 static int __init sock_init(void)
2158 * Initialize sock SLAB cache.
2164 * Initialize skbuff SLAB cache
2169 * Initialize the protocols module.
2173 register_filesystem(&sock_fs_type);
2174 sock_mnt = kern_mount(&sock_fs_type);
2176 /* The real protocol initialization is performed in later initcalls.
2179 #ifdef CONFIG_NETFILTER
2186 core_initcall(sock_init); /* early initcall */
2188 #ifdef CONFIG_PROC_FS
2189 void socket_seq_show(struct seq_file *seq)
2194 for_each_possible_cpu(cpu)
2195 counter += per_cpu(sockets_in_use, cpu);
2197 /* It can be negative, by the way. 8) */
2201 seq_printf(seq, "sockets: used %d\n", counter);
2203 #endif /* CONFIG_PROC_FS */
2205 #ifdef CONFIG_COMPAT
2206 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2209 struct socket *sock = file->private_data;
2210 int ret = -ENOIOCTLCMD;
2212 if (sock->ops->compat_ioctl)
2213 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2219 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2221 return sock->ops->bind(sock, addr, addrlen);
2224 int kernel_listen(struct socket *sock, int backlog)
2226 return sock->ops->listen(sock, backlog);
2229 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2231 struct sock *sk = sock->sk;
2234 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2239 err = sock->ops->accept(sock, *newsock, flags);
2241 sock_release(*newsock);
2246 (*newsock)->ops = sock->ops;
2252 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2255 return sock->ops->connect(sock, addr, addrlen, flags);
2258 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2261 return sock->ops->getname(sock, addr, addrlen, 0);
2264 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2267 return sock->ops->getname(sock, addr, addrlen, 1);
2270 int kernel_getsockopt(struct socket *sock, int level, int optname,
2271 char *optval, int *optlen)
2273 mm_segment_t oldfs = get_fs();
2277 if (level == SOL_SOCKET)
2278 err = sock_getsockopt(sock, level, optname, optval, optlen);
2280 err = sock->ops->getsockopt(sock, level, optname, optval,
2286 int kernel_setsockopt(struct socket *sock, int level, int optname,
2287 char *optval, int optlen)
2289 mm_segment_t oldfs = get_fs();
2293 if (level == SOL_SOCKET)
2294 err = sock_setsockopt(sock, level, optname, optval, optlen);
2296 err = sock->ops->setsockopt(sock, level, optname, optval,
2302 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2303 size_t size, int flags)
2305 if (sock->ops->sendpage)
2306 return sock->ops->sendpage(sock, page, offset, size, flags);
2308 return sock_no_sendpage(sock, page, offset, size, flags);
2311 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2313 mm_segment_t oldfs = get_fs();
2317 err = sock->ops->ioctl(sock, cmd, arg);
2323 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2325 return sock->ops->shutdown(sock, how);
2328 /* ABI emulation layers need these two */
2329 EXPORT_SYMBOL(move_addr_to_kernel);
2330 EXPORT_SYMBOL(move_addr_to_user);
2331 EXPORT_SYMBOL(sock_create);
2332 EXPORT_SYMBOL(sock_create_kern);
2333 EXPORT_SYMBOL(sock_create_lite);
2334 EXPORT_SYMBOL(sock_map_fd);
2335 EXPORT_SYMBOL(sock_recvmsg);
2336 EXPORT_SYMBOL(sock_register);
2337 EXPORT_SYMBOL(sock_release);
2338 EXPORT_SYMBOL(sock_sendmsg);
2339 EXPORT_SYMBOL(sock_unregister);
2340 EXPORT_SYMBOL(sock_wake_async);
2341 EXPORT_SYMBOL(sockfd_lookup);
2342 EXPORT_SYMBOL(kernel_sendmsg);
2343 EXPORT_SYMBOL(kernel_recvmsg);
2344 EXPORT_SYMBOL(kernel_bind);
2345 EXPORT_SYMBOL(kernel_listen);
2346 EXPORT_SYMBOL(kernel_accept);
2347 EXPORT_SYMBOL(kernel_connect);
2348 EXPORT_SYMBOL(kernel_getsockname);
2349 EXPORT_SYMBOL(kernel_getpeername);
2350 EXPORT_SYMBOL(kernel_getsockopt);
2351 EXPORT_SYMBOL(kernel_setsockopt);
2352 EXPORT_SYMBOL(kernel_sendpage);
2353 EXPORT_SYMBOL(kernel_sock_ioctl);
2354 EXPORT_SYMBOL(kernel_sock_shutdown);