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/thread_info.h>
67 #include <linux/rcupdate.h>
68 #include <linux/netdevice.h>
69 #include <linux/proc_fs.h>
70 #include <linux/seq_file.h>
71 #include <linux/mutex.h>
72 #include <linux/thread_info.h>
73 #include <linux/wanrouter.h>
74 #include <linux/if_bridge.h>
75 #include <linux/if_frad.h>
76 #include <linux/if_vlan.h>
77 #include <linux/init.h>
78 #include <linux/poll.h>
79 #include <linux/cache.h>
80 #include <linux/module.h>
81 #include <linux/highmem.h>
82 #include <linux/mount.h>
83 #include <linux/security.h>
84 #include <linux/syscalls.h>
85 #include <linux/compat.h>
86 #include <linux/kmod.h>
87 #include <linux/audit.h>
88 #include <linux/wireless.h>
89 #include <linux/nsproxy.h>
91 #include <asm/uaccess.h>
92 #include <asm/unistd.h>
94 #include <net/compat.h>
98 #include <linux/netfilter.h>
100 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
101 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
102 unsigned long nr_segs, loff_t pos);
103 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
104 unsigned long nr_segs, loff_t pos);
105 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
107 static int sock_close(struct inode *inode, struct file *file);
108 static unsigned int sock_poll(struct file *file,
109 struct poll_table_struct *wait);
110 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
112 static long compat_sock_ioctl(struct file *file,
113 unsigned int cmd, unsigned long arg);
115 static int sock_fasync(int fd, struct file *filp, int on);
116 static ssize_t sock_sendpage(struct file *file, struct page *page,
117 int offset, size_t size, loff_t *ppos, int more);
118 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
119 struct pipe_inode_info *pipe, size_t len,
123 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
124 * in the operation structures but are done directly via the socketcall() multiplexor.
127 static const struct file_operations socket_file_ops = {
128 .owner = THIS_MODULE,
130 .aio_read = sock_aio_read,
131 .aio_write = sock_aio_write,
133 .unlocked_ioctl = sock_ioctl,
135 .compat_ioctl = compat_sock_ioctl,
138 .open = sock_no_open, /* special open code to disallow open via /proc */
139 .release = sock_close,
140 .fasync = sock_fasync,
141 .sendpage = sock_sendpage,
142 .splice_write = generic_splice_sendpage,
143 .splice_read = sock_splice_read,
147 * The protocol list. Each protocol is registered in here.
150 static DEFINE_SPINLOCK(net_family_lock);
151 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
154 * Statistics counters of the socket lists
157 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
161 * Move socket addresses back and forth across the kernel/user
162 * divide and look after the messy bits.
165 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
166 16 for IP, 16 for IPX,
169 must be at least one bigger than
170 the AF_UNIX size (see net/unix/af_unix.c
175 * move_addr_to_kernel - copy a socket address into kernel space
176 * @uaddr: Address in user space
177 * @kaddr: Address in kernel space
178 * @ulen: Length in user space
180 * The address is copied into kernel space. If the provided address is
181 * too long an error code of -EINVAL is returned. If the copy gives
182 * invalid addresses -EFAULT is returned. On a success 0 is returned.
185 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
187 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
191 if (copy_from_user(kaddr, uaddr, ulen))
193 return audit_sockaddr(ulen, kaddr);
197 * move_addr_to_user - copy an address to user space
198 * @kaddr: kernel space address
199 * @klen: length of address in kernel
200 * @uaddr: user space address
201 * @ulen: pointer to user length field
203 * The value pointed to by ulen on entry is the buffer length available.
204 * This is overwritten with the buffer space used. -EINVAL is returned
205 * if an overlong buffer is specified or a negative buffer size. -EFAULT
206 * is returned if either the buffer or the length field are not
208 * After copying the data up to the limit the user specifies, the true
209 * length of the data is written over the length limit the user
210 * specified. Zero is returned for a success.
213 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
219 err = get_user(len, ulen);
224 if (len < 0 || len > sizeof(struct sockaddr_storage))
227 if (audit_sockaddr(klen, kaddr))
229 if (copy_to_user(uaddr, kaddr, len))
233 * "fromlen shall refer to the value before truncation.."
236 return __put_user(klen, ulen);
239 #define SOCKFS_MAGIC 0x534F434B
241 static struct kmem_cache *sock_inode_cachep __read_mostly;
243 static struct inode *sock_alloc_inode(struct super_block *sb)
245 struct socket_alloc *ei;
247 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
250 init_waitqueue_head(&ei->socket.wait);
252 ei->socket.fasync_list = NULL;
253 ei->socket.state = SS_UNCONNECTED;
254 ei->socket.flags = 0;
255 ei->socket.ops = NULL;
256 ei->socket.sk = NULL;
257 ei->socket.file = NULL;
259 return &ei->vfs_inode;
262 static void sock_destroy_inode(struct inode *inode)
264 kmem_cache_free(sock_inode_cachep,
265 container_of(inode, struct socket_alloc, vfs_inode));
268 static void init_once(void *foo)
270 struct socket_alloc *ei = (struct socket_alloc *)foo;
272 inode_init_once(&ei->vfs_inode);
275 static int init_inodecache(void)
277 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
278 sizeof(struct socket_alloc),
280 (SLAB_HWCACHE_ALIGN |
281 SLAB_RECLAIM_ACCOUNT |
284 if (sock_inode_cachep == NULL)
289 static struct super_operations sockfs_ops = {
290 .alloc_inode = sock_alloc_inode,
291 .destroy_inode =sock_destroy_inode,
292 .statfs = simple_statfs,
295 static int sockfs_get_sb(struct file_system_type *fs_type,
296 int flags, const char *dev_name, void *data,
297 struct vfsmount *mnt)
299 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
303 static struct vfsmount *sock_mnt __read_mostly;
305 static struct file_system_type sock_fs_type = {
307 .get_sb = sockfs_get_sb,
308 .kill_sb = kill_anon_super,
311 static int sockfs_delete_dentry(struct dentry *dentry)
314 * At creation time, we pretended this dentry was hashed
315 * (by clearing DCACHE_UNHASHED bit in d_flags)
316 * At delete time, we restore the truth : not hashed.
317 * (so that dput() can proceed correctly)
319 dentry->d_flags |= DCACHE_UNHASHED;
324 * sockfs_dname() is called from d_path().
326 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
328 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
329 dentry->d_inode->i_ino);
332 static struct dentry_operations sockfs_dentry_operations = {
333 .d_delete = sockfs_delete_dentry,
334 .d_dname = sockfs_dname,
338 * Obtains the first available file descriptor and sets it up for use.
340 * These functions create file structures and maps them to fd space
341 * of the current process. On success it returns file descriptor
342 * and file struct implicitly stored in sock->file.
343 * Note that another thread may close file descriptor before we return
344 * from this function. We use the fact that now we do not refer
345 * to socket after mapping. If one day we will need it, this
346 * function will increment ref. count on file by 1.
348 * In any case returned fd MAY BE not valid!
349 * This race condition is unavoidable
350 * with shared fd spaces, we cannot solve it inside kernel,
351 * but we take care of internal coherence yet.
354 static int sock_alloc_fd(struct file **filep, int flags)
358 fd = get_unused_fd_flags(flags);
359 if (likely(fd >= 0)) {
360 struct file *file = get_empty_filp();
363 if (unlikely(!file)) {
372 static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
374 struct dentry *dentry;
375 struct qstr name = { .name = "" };
377 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
378 if (unlikely(!dentry))
381 dentry->d_op = &sockfs_dentry_operations;
383 * We dont want to push this dentry into global dentry hash table.
384 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
385 * This permits a working /proc/$pid/fd/XXX on sockets
387 dentry->d_flags &= ~DCACHE_UNHASHED;
388 d_instantiate(dentry, SOCK_INODE(sock));
391 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
393 SOCK_INODE(sock)->i_fop = &socket_file_ops;
394 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
396 file->private_data = sock;
401 int sock_map_fd(struct socket *sock, int flags)
403 struct file *newfile;
404 int fd = sock_alloc_fd(&newfile, flags);
406 if (likely(fd >= 0)) {
407 int err = sock_attach_fd(sock, newfile, flags);
409 if (unlikely(err < 0)) {
414 fd_install(fd, newfile);
419 static struct socket *sock_from_file(struct file *file, int *err)
421 if (file->f_op == &socket_file_ops)
422 return file->private_data; /* set in sock_map_fd */
429 * sockfd_lookup - Go from a file number to its socket slot
431 * @err: pointer to an error code return
433 * The file handle passed in is locked and the socket it is bound
434 * too is returned. If an error occurs the err pointer is overwritten
435 * with a negative errno code and NULL is returned. The function checks
436 * for both invalid handles and passing a handle which is not a socket.
438 * On a success the socket object pointer is returned.
441 struct socket *sockfd_lookup(int fd, int *err)
452 sock = sock_from_file(file, err);
458 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
464 file = fget_light(fd, fput_needed);
466 sock = sock_from_file(file, err);
469 fput_light(file, *fput_needed);
475 * sock_alloc - allocate a socket
477 * Allocate a new inode and socket object. The two are bound together
478 * and initialised. The socket is then returned. If we are out of inodes
482 static struct socket *sock_alloc(void)
487 inode = new_inode(sock_mnt->mnt_sb);
491 sock = SOCKET_I(inode);
493 inode->i_mode = S_IFSOCK | S_IRWXUGO;
494 inode->i_uid = current->fsuid;
495 inode->i_gid = current->fsgid;
497 get_cpu_var(sockets_in_use)++;
498 put_cpu_var(sockets_in_use);
503 * In theory you can't get an open on this inode, but /proc provides
504 * a back door. Remember to keep it shut otherwise you'll let the
505 * creepy crawlies in.
508 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
513 const struct file_operations bad_sock_fops = {
514 .owner = THIS_MODULE,
515 .open = sock_no_open,
519 * sock_release - close a socket
520 * @sock: socket to close
522 * The socket is released from the protocol stack if it has a release
523 * callback, and the inode is then released if the socket is bound to
524 * an inode not a file.
527 void sock_release(struct socket *sock)
530 struct module *owner = sock->ops->owner;
532 sock->ops->release(sock);
537 if (sock->fasync_list)
538 printk(KERN_ERR "sock_release: fasync list not empty!\n");
540 get_cpu_var(sockets_in_use)--;
541 put_cpu_var(sockets_in_use);
543 iput(SOCK_INODE(sock));
549 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
550 struct msghdr *msg, size_t size)
552 struct sock_iocb *si = kiocb_to_siocb(iocb);
560 err = security_socket_sendmsg(sock, msg, size);
564 return sock->ops->sendmsg(iocb, sock, msg, size);
567 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
570 struct sock_iocb siocb;
573 init_sync_kiocb(&iocb, NULL);
574 iocb.private = &siocb;
575 ret = __sock_sendmsg(&iocb, sock, msg, size);
576 if (-EIOCBQUEUED == ret)
577 ret = wait_on_sync_kiocb(&iocb);
581 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
582 struct kvec *vec, size_t num, size_t size)
584 mm_segment_t oldfs = get_fs();
589 * the following is safe, since for compiler definitions of kvec and
590 * iovec are identical, yielding the same in-core layout and alignment
592 msg->msg_iov = (struct iovec *)vec;
593 msg->msg_iovlen = num;
594 result = sock_sendmsg(sock, msg, size);
600 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
602 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
605 ktime_t kt = skb->tstamp;
607 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
609 /* Race occurred between timestamp enabling and packet
610 receiving. Fill in the current time for now. */
612 kt = ktime_get_real();
614 tv = ktime_to_timeval(kt);
615 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
618 /* Race occurred between timestamp enabling and packet
619 receiving. Fill in the current time for now. */
621 kt = ktime_get_real();
623 ts = ktime_to_timespec(kt);
624 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
628 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
630 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
631 struct msghdr *msg, size_t size, int flags)
634 struct sock_iocb *si = kiocb_to_siocb(iocb);
642 err = security_socket_recvmsg(sock, msg, size, flags);
646 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
649 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
650 size_t size, int flags)
653 struct sock_iocb siocb;
656 init_sync_kiocb(&iocb, NULL);
657 iocb.private = &siocb;
658 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
659 if (-EIOCBQUEUED == ret)
660 ret = wait_on_sync_kiocb(&iocb);
664 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
665 struct kvec *vec, size_t num, size_t size, int flags)
667 mm_segment_t oldfs = get_fs();
672 * the following is safe, since for compiler definitions of kvec and
673 * iovec are identical, yielding the same in-core layout and alignment
675 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
676 result = sock_recvmsg(sock, msg, size, flags);
681 static void sock_aio_dtor(struct kiocb *iocb)
683 kfree(iocb->private);
686 static ssize_t sock_sendpage(struct file *file, struct page *page,
687 int offset, size_t size, loff_t *ppos, int more)
692 sock = file->private_data;
694 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
698 return sock->ops->sendpage(sock, page, offset, size, flags);
701 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
702 struct pipe_inode_info *pipe, size_t len,
705 struct socket *sock = file->private_data;
707 if (unlikely(!sock->ops->splice_read))
710 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
713 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
714 struct sock_iocb *siocb)
716 if (!is_sync_kiocb(iocb)) {
717 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
720 iocb->ki_dtor = sock_aio_dtor;
724 iocb->private = siocb;
728 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
729 struct file *file, const struct iovec *iov,
730 unsigned long nr_segs)
732 struct socket *sock = file->private_data;
736 for (i = 0; i < nr_segs; i++)
737 size += iov[i].iov_len;
739 msg->msg_name = NULL;
740 msg->msg_namelen = 0;
741 msg->msg_control = NULL;
742 msg->msg_controllen = 0;
743 msg->msg_iov = (struct iovec *)iov;
744 msg->msg_iovlen = nr_segs;
745 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
747 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
750 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
751 unsigned long nr_segs, loff_t pos)
753 struct sock_iocb siocb, *x;
758 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
762 x = alloc_sock_iocb(iocb, &siocb);
765 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
768 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
769 struct file *file, const struct iovec *iov,
770 unsigned long nr_segs)
772 struct socket *sock = file->private_data;
776 for (i = 0; i < nr_segs; i++)
777 size += iov[i].iov_len;
779 msg->msg_name = NULL;
780 msg->msg_namelen = 0;
781 msg->msg_control = NULL;
782 msg->msg_controllen = 0;
783 msg->msg_iov = (struct iovec *)iov;
784 msg->msg_iovlen = nr_segs;
785 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
786 if (sock->type == SOCK_SEQPACKET)
787 msg->msg_flags |= MSG_EOR;
789 return __sock_sendmsg(iocb, sock, msg, size);
792 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
793 unsigned long nr_segs, loff_t pos)
795 struct sock_iocb siocb, *x;
800 x = alloc_sock_iocb(iocb, &siocb);
804 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
808 * Atomic setting of ioctl hooks to avoid race
809 * with module unload.
812 static DEFINE_MUTEX(br_ioctl_mutex);
813 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
815 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
817 mutex_lock(&br_ioctl_mutex);
818 br_ioctl_hook = hook;
819 mutex_unlock(&br_ioctl_mutex);
822 EXPORT_SYMBOL(brioctl_set);
824 static DEFINE_MUTEX(vlan_ioctl_mutex);
825 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
827 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
829 mutex_lock(&vlan_ioctl_mutex);
830 vlan_ioctl_hook = hook;
831 mutex_unlock(&vlan_ioctl_mutex);
834 EXPORT_SYMBOL(vlan_ioctl_set);
836 static DEFINE_MUTEX(dlci_ioctl_mutex);
837 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
839 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
841 mutex_lock(&dlci_ioctl_mutex);
842 dlci_ioctl_hook = hook;
843 mutex_unlock(&dlci_ioctl_mutex);
846 EXPORT_SYMBOL(dlci_ioctl_set);
849 * With an ioctl, arg may well be a user mode pointer, but we don't know
850 * what to do with it - that's up to the protocol still.
853 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
857 void __user *argp = (void __user *)arg;
861 sock = file->private_data;
864 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
865 err = dev_ioctl(net, cmd, argp);
867 #ifdef CONFIG_WIRELESS_EXT
868 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
869 err = dev_ioctl(net, cmd, argp);
871 #endif /* CONFIG_WIRELESS_EXT */
876 if (get_user(pid, (int __user *)argp))
878 err = f_setown(sock->file, pid, 1);
882 err = put_user(f_getown(sock->file),
891 request_module("bridge");
893 mutex_lock(&br_ioctl_mutex);
895 err = br_ioctl_hook(net, cmd, argp);
896 mutex_unlock(&br_ioctl_mutex);
901 if (!vlan_ioctl_hook)
902 request_module("8021q");
904 mutex_lock(&vlan_ioctl_mutex);
906 err = vlan_ioctl_hook(net, argp);
907 mutex_unlock(&vlan_ioctl_mutex);
912 if (!dlci_ioctl_hook)
913 request_module("dlci");
915 mutex_lock(&dlci_ioctl_mutex);
917 err = dlci_ioctl_hook(cmd, argp);
918 mutex_unlock(&dlci_ioctl_mutex);
921 err = sock->ops->ioctl(sock, cmd, arg);
924 * If this ioctl is unknown try to hand it down
927 if (err == -ENOIOCTLCMD)
928 err = dev_ioctl(net, cmd, argp);
934 int sock_create_lite(int family, int type, int protocol, struct socket **res)
937 struct socket *sock = NULL;
939 err = security_socket_create(family, type, protocol, 1);
950 err = security_socket_post_create(sock, family, type, protocol, 1);
963 /* No kernel lock held - perfect */
964 static unsigned int sock_poll(struct file *file, poll_table *wait)
969 * We can't return errors to poll, so it's either yes or no.
971 sock = file->private_data;
972 return sock->ops->poll(file, sock, wait);
975 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
977 struct socket *sock = file->private_data;
979 return sock->ops->mmap(file, sock, vma);
982 static int sock_close(struct inode *inode, struct file *filp)
985 * It was possible the inode is NULL we were
986 * closing an unfinished socket.
990 printk(KERN_DEBUG "sock_close: NULL inode\n");
993 sock_fasync(-1, filp, 0);
994 sock_release(SOCKET_I(inode));
999 * Update the socket async list
1001 * Fasync_list locking strategy.
1003 * 1. fasync_list is modified only under process context socket lock
1004 * i.e. under semaphore.
1005 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1006 * or under socket lock.
1007 * 3. fasync_list can be used from softirq context, so that
1008 * modification under socket lock have to be enhanced with
1009 * write_lock_bh(&sk->sk_callback_lock).
1013 static int sock_fasync(int fd, struct file *filp, int on)
1015 struct fasync_struct *fa, *fna = NULL, **prev;
1016 struct socket *sock;
1020 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1025 sock = filp->private_data;
1035 prev = &(sock->fasync_list);
1037 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1038 if (fa->fa_file == filp)
1043 write_lock_bh(&sk->sk_callback_lock);
1045 write_unlock_bh(&sk->sk_callback_lock);
1050 fna->fa_file = filp;
1052 fna->magic = FASYNC_MAGIC;
1053 fna->fa_next = sock->fasync_list;
1054 write_lock_bh(&sk->sk_callback_lock);
1055 sock->fasync_list = fna;
1056 write_unlock_bh(&sk->sk_callback_lock);
1059 write_lock_bh(&sk->sk_callback_lock);
1060 *prev = fa->fa_next;
1061 write_unlock_bh(&sk->sk_callback_lock);
1067 release_sock(sock->sk);
1071 /* This function may be called only under socket lock or callback_lock */
1073 int sock_wake_async(struct socket *sock, int how, int band)
1075 if (!sock || !sock->fasync_list)
1078 case SOCK_WAKE_WAITD:
1079 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1082 case SOCK_WAKE_SPACE:
1083 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1088 __kill_fasync(sock->fasync_list, SIGIO, band);
1091 __kill_fasync(sock->fasync_list, SIGURG, band);
1096 static int __sock_create(struct net *net, int family, int type, int protocol,
1097 struct socket **res, int kern)
1100 struct socket *sock;
1101 const struct net_proto_family *pf;
1104 * Check protocol is in range
1106 if (family < 0 || family >= NPROTO)
1107 return -EAFNOSUPPORT;
1108 if (type < 0 || type >= SOCK_MAX)
1113 This uglymoron is moved from INET layer to here to avoid
1114 deadlock in module load.
1116 if (family == PF_INET && type == SOCK_PACKET) {
1120 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1126 err = security_socket_create(family, type, protocol, kern);
1131 * Allocate the socket and allow the family to set things up. if
1132 * the protocol is 0, the family is instructed to select an appropriate
1135 sock = sock_alloc();
1137 if (net_ratelimit())
1138 printk(KERN_WARNING "socket: no more sockets\n");
1139 return -ENFILE; /* Not exactly a match, but its the
1140 closest posix thing */
1145 #if defined(CONFIG_KMOD)
1146 /* Attempt to load a protocol module if the find failed.
1148 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1149 * requested real, full-featured networking support upon configuration.
1150 * Otherwise module support will break!
1152 if (net_families[family] == NULL)
1153 request_module("net-pf-%d", family);
1157 pf = rcu_dereference(net_families[family]);
1158 err = -EAFNOSUPPORT;
1163 * We will call the ->create function, that possibly is in a loadable
1164 * module, so we have to bump that loadable module refcnt first.
1166 if (!try_module_get(pf->owner))
1169 /* Now protected by module ref count */
1172 err = pf->create(net, sock, protocol);
1174 goto out_module_put;
1177 * Now to bump the refcnt of the [loadable] module that owns this
1178 * socket at sock_release time we decrement its refcnt.
1180 if (!try_module_get(sock->ops->owner))
1181 goto out_module_busy;
1184 * Now that we're done with the ->create function, the [loadable]
1185 * module can have its refcnt decremented
1187 module_put(pf->owner);
1188 err = security_socket_post_create(sock, family, type, protocol, kern);
1190 goto out_sock_release;
1196 err = -EAFNOSUPPORT;
1199 module_put(pf->owner);
1206 goto out_sock_release;
1209 int sock_create(int family, int type, int protocol, struct socket **res)
1211 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1214 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1216 return __sock_create(&init_net, family, type, protocol, res, 1);
1219 asmlinkage long sys_socket(int family, int type, int protocol)
1222 struct socket *sock;
1225 /* Check the SOCK_* constants for consistency. */
1226 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1227 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1228 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1229 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1231 flags = type & ~SOCK_TYPE_MASK;
1232 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1234 type &= SOCK_TYPE_MASK;
1236 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1237 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1239 retval = sock_create(family, type, protocol, &sock);
1243 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1248 /* It may be already another descriptor 8) Not kernel problem. */
1257 * Create a pair of connected sockets.
1260 asmlinkage long sys_socketpair(int family, int type, int protocol,
1261 int __user *usockvec)
1263 struct socket *sock1, *sock2;
1265 struct file *newfile1, *newfile2;
1268 flags = type & ~SOCK_TYPE_MASK;
1269 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1271 type &= SOCK_TYPE_MASK;
1273 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1274 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1277 * Obtain the first socket and check if the underlying protocol
1278 * supports the socketpair call.
1281 err = sock_create(family, type, protocol, &sock1);
1285 err = sock_create(family, type, protocol, &sock2);
1289 err = sock1->ops->socketpair(sock1, sock2);
1291 goto out_release_both;
1293 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
1294 if (unlikely(fd1 < 0)) {
1296 goto out_release_both;
1299 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
1300 if (unlikely(fd2 < 0)) {
1304 goto out_release_both;
1307 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
1308 if (unlikely(err < 0)) {
1312 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
1313 if (unlikely(err < 0)) {
1318 err = audit_fd_pair(fd1, fd2);
1325 fd_install(fd1, newfile1);
1326 fd_install(fd2, newfile2);
1327 /* fd1 and fd2 may be already another descriptors.
1328 * Not kernel problem.
1331 err = put_user(fd1, &usockvec[0]);
1333 err = put_user(fd2, &usockvec[1]);
1342 sock_release(sock2);
1344 sock_release(sock1);
1350 sock_release(sock1);
1353 sock_release(sock2);
1361 * Bind a name to a socket. Nothing much to do here since it's
1362 * the protocol's responsibility to handle the local address.
1364 * We move the socket address to kernel space before we call
1365 * the protocol layer (having also checked the address is ok).
1368 asmlinkage long sys_bind(int fd, struct sockaddr __user *umyaddr, int addrlen)
1370 struct socket *sock;
1371 struct sockaddr_storage address;
1372 int err, fput_needed;
1374 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1376 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1378 err = security_socket_bind(sock,
1379 (struct sockaddr *)&address,
1382 err = sock->ops->bind(sock,
1386 fput_light(sock->file, fput_needed);
1392 * Perform a listen. Basically, we allow the protocol to do anything
1393 * necessary for a listen, and if that works, we mark the socket as
1394 * ready for listening.
1397 asmlinkage long sys_listen(int fd, int backlog)
1399 struct socket *sock;
1400 int err, fput_needed;
1403 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1405 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1406 if ((unsigned)backlog > somaxconn)
1407 backlog = somaxconn;
1409 err = security_socket_listen(sock, backlog);
1411 err = sock->ops->listen(sock, backlog);
1413 fput_light(sock->file, fput_needed);
1419 * For accept, we attempt to create a new socket, set up the link
1420 * with the client, wake up the client, then return the new
1421 * connected fd. We collect the address of the connector in kernel
1422 * space and move it to user at the very end. This is unclean because
1423 * we open the socket then return an error.
1425 * 1003.1g adds the ability to recvmsg() to query connection pending
1426 * status to recvmsg. We need to add that support in a way thats
1427 * clean when we restucture accept also.
1430 long do_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1431 int __user *upeer_addrlen, int flags)
1433 struct socket *sock, *newsock;
1434 struct file *newfile;
1435 int err, len, newfd, fput_needed;
1436 struct sockaddr_storage address;
1438 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1441 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1442 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1444 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1449 if (!(newsock = sock_alloc()))
1452 newsock->type = sock->type;
1453 newsock->ops = sock->ops;
1456 * We don't need try_module_get here, as the listening socket (sock)
1457 * has the protocol module (sock->ops->owner) held.
1459 __module_get(newsock->ops->owner);
1461 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
1462 if (unlikely(newfd < 0)) {
1464 sock_release(newsock);
1468 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
1472 err = security_socket_accept(sock, newsock);
1476 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1480 if (upeer_sockaddr) {
1481 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1483 err = -ECONNABORTED;
1486 err = move_addr_to_user((struct sockaddr *)&address,
1487 len, upeer_sockaddr, upeer_addrlen);
1492 /* File flags are not inherited via accept() unlike another OSes. */
1494 fd_install(newfd, newfile);
1497 security_socket_post_accept(sock, newsock);
1500 fput_light(sock->file, fput_needed);
1504 sock_release(newsock);
1506 put_unused_fd(newfd);
1510 put_unused_fd(newfd);
1514 #ifdef HAVE_SET_RESTORE_SIGMASK
1515 asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1516 int __user *upeer_addrlen,
1517 const sigset_t __user *sigmask,
1518 size_t sigsetsize, int flags)
1520 sigset_t ksigmask, sigsaved;
1524 /* XXX: Don't preclude handling different sized sigset_t's. */
1525 if (sigsetsize != sizeof(sigset_t))
1527 if (copy_from_user(&ksigmask, sigmask, sizeof(ksigmask)))
1530 sigdelsetmask(&ksigmask, sigmask(SIGKILL)|sigmask(SIGSTOP));
1531 sigprocmask(SIG_SETMASK, &ksigmask, &sigsaved);
1534 ret = do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1536 if (ret < 0 && signal_pending(current)) {
1538 * Don't restore the signal mask yet. Let do_signal() deliver
1539 * the signal on the way back to userspace, before the signal
1543 memcpy(¤t->saved_sigmask, &sigsaved,
1545 set_restore_sigmask();
1548 sigprocmask(SIG_SETMASK, &sigsaved, NULL);
1553 asmlinkage long sys_paccept(int fd, struct sockaddr __user *upeer_sockaddr,
1554 int __user *upeer_addrlen,
1555 const sigset_t __user *sigmask,
1556 size_t sigsetsize, int flags)
1558 /* The platform does not support restoring the signal mask in the
1559 * return path. So we do not allow using paccept() with a signal
1564 return do_accept(fd, upeer_sockaddr, upeer_addrlen, flags);
1568 asmlinkage long sys_accept(int fd, struct sockaddr __user *upeer_sockaddr,
1569 int __user *upeer_addrlen)
1571 return do_accept(fd, upeer_sockaddr, upeer_addrlen, 0);
1575 * Attempt to connect to a socket with the server address. The address
1576 * is in user space so we verify it is OK and move it to kernel space.
1578 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1581 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1582 * other SEQPACKET protocols that take time to connect() as it doesn't
1583 * include the -EINPROGRESS status for such sockets.
1586 asmlinkage long sys_connect(int fd, struct sockaddr __user *uservaddr,
1589 struct socket *sock;
1590 struct sockaddr_storage address;
1591 int err, fput_needed;
1593 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1596 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1601 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1605 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1606 sock->file->f_flags);
1608 fput_light(sock->file, fput_needed);
1614 * Get the local address ('name') of a socket object. Move the obtained
1615 * name to user space.
1618 asmlinkage long sys_getsockname(int fd, struct sockaddr __user *usockaddr,
1619 int __user *usockaddr_len)
1621 struct socket *sock;
1622 struct sockaddr_storage address;
1623 int len, err, fput_needed;
1625 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1629 err = security_socket_getsockname(sock);
1633 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1636 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1639 fput_light(sock->file, fput_needed);
1645 * Get the remote address ('name') of a socket object. Move the obtained
1646 * name to user space.
1649 asmlinkage long sys_getpeername(int fd, struct sockaddr __user *usockaddr,
1650 int __user *usockaddr_len)
1652 struct socket *sock;
1653 struct sockaddr_storage address;
1654 int len, err, fput_needed;
1656 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1658 err = security_socket_getpeername(sock);
1660 fput_light(sock->file, fput_needed);
1665 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1668 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1670 fput_light(sock->file, fput_needed);
1676 * Send a datagram to a given address. We move the address into kernel
1677 * space and check the user space data area is readable before invoking
1681 asmlinkage long sys_sendto(int fd, void __user *buff, size_t len,
1682 unsigned flags, struct sockaddr __user *addr,
1685 struct socket *sock;
1686 struct sockaddr_storage address;
1692 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1696 iov.iov_base = buff;
1698 msg.msg_name = NULL;
1701 msg.msg_control = NULL;
1702 msg.msg_controllen = 0;
1703 msg.msg_namelen = 0;
1705 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1708 msg.msg_name = (struct sockaddr *)&address;
1709 msg.msg_namelen = addr_len;
1711 if (sock->file->f_flags & O_NONBLOCK)
1712 flags |= MSG_DONTWAIT;
1713 msg.msg_flags = flags;
1714 err = sock_sendmsg(sock, &msg, len);
1717 fput_light(sock->file, fput_needed);
1723 * Send a datagram down a socket.
1726 asmlinkage long sys_send(int fd, void __user *buff, size_t len, unsigned flags)
1728 return sys_sendto(fd, buff, len, flags, NULL, 0);
1732 * Receive a frame from the socket and optionally record the address of the
1733 * sender. We verify the buffers are writable and if needed move the
1734 * sender address from kernel to user space.
1737 asmlinkage long sys_recvfrom(int fd, void __user *ubuf, size_t size,
1738 unsigned flags, struct sockaddr __user *addr,
1739 int __user *addr_len)
1741 struct socket *sock;
1744 struct sockaddr_storage address;
1748 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1752 msg.msg_control = NULL;
1753 msg.msg_controllen = 0;
1757 iov.iov_base = ubuf;
1758 msg.msg_name = (struct sockaddr *)&address;
1759 msg.msg_namelen = sizeof(address);
1760 if (sock->file->f_flags & O_NONBLOCK)
1761 flags |= MSG_DONTWAIT;
1762 err = sock_recvmsg(sock, &msg, size, flags);
1764 if (err >= 0 && addr != NULL) {
1765 err2 = move_addr_to_user((struct sockaddr *)&address,
1766 msg.msg_namelen, addr, addr_len);
1771 fput_light(sock->file, fput_needed);
1777 * Receive a datagram from a socket.
1780 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1783 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1787 * Set a socket option. Because we don't know the option lengths we have
1788 * to pass the user mode parameter for the protocols to sort out.
1791 asmlinkage long sys_setsockopt(int fd, int level, int optname,
1792 char __user *optval, int optlen)
1794 int err, fput_needed;
1795 struct socket *sock;
1800 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1802 err = security_socket_setsockopt(sock, level, optname);
1806 if (level == SOL_SOCKET)
1808 sock_setsockopt(sock, level, optname, optval,
1812 sock->ops->setsockopt(sock, level, optname, optval,
1815 fput_light(sock->file, fput_needed);
1821 * Get a socket option. Because we don't know the option lengths we have
1822 * to pass a user mode parameter for the protocols to sort out.
1825 asmlinkage long sys_getsockopt(int fd, int level, int optname,
1826 char __user *optval, int __user *optlen)
1828 int err, fput_needed;
1829 struct socket *sock;
1831 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1833 err = security_socket_getsockopt(sock, level, optname);
1837 if (level == SOL_SOCKET)
1839 sock_getsockopt(sock, level, optname, optval,
1843 sock->ops->getsockopt(sock, level, optname, optval,
1846 fput_light(sock->file, fput_needed);
1852 * Shutdown a socket.
1855 asmlinkage long sys_shutdown(int fd, int how)
1857 int err, fput_needed;
1858 struct socket *sock;
1860 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1862 err = security_socket_shutdown(sock, how);
1864 err = sock->ops->shutdown(sock, how);
1865 fput_light(sock->file, fput_needed);
1870 /* A couple of helpful macros for getting the address of the 32/64 bit
1871 * fields which are the same type (int / unsigned) on our platforms.
1873 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1874 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1875 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1878 * BSD sendmsg interface
1881 asmlinkage long sys_sendmsg(int fd, struct msghdr __user *msg, unsigned flags)
1883 struct compat_msghdr __user *msg_compat =
1884 (struct compat_msghdr __user *)msg;
1885 struct socket *sock;
1886 struct sockaddr_storage address;
1887 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1888 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1889 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1890 /* 20 is size of ipv6_pktinfo */
1891 unsigned char *ctl_buf = ctl;
1892 struct msghdr msg_sys;
1893 int err, ctl_len, iov_size, total_len;
1897 if (MSG_CMSG_COMPAT & flags) {
1898 if (get_compat_msghdr(&msg_sys, msg_compat))
1901 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1904 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1908 /* do not move before msg_sys is valid */
1910 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1913 /* Check whether to allocate the iovec area */
1915 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1916 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1917 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1922 /* This will also move the address data into kernel space */
1923 if (MSG_CMSG_COMPAT & flags) {
1924 err = verify_compat_iovec(&msg_sys, iov,
1925 (struct sockaddr *)&address,
1928 err = verify_iovec(&msg_sys, iov,
1929 (struct sockaddr *)&address,
1937 if (msg_sys.msg_controllen > INT_MAX)
1939 ctl_len = msg_sys.msg_controllen;
1940 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1942 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1946 ctl_buf = msg_sys.msg_control;
1947 ctl_len = msg_sys.msg_controllen;
1948 } else if (ctl_len) {
1949 if (ctl_len > sizeof(ctl)) {
1950 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1951 if (ctl_buf == NULL)
1956 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1957 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1958 * checking falls down on this.
1960 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1963 msg_sys.msg_control = ctl_buf;
1965 msg_sys.msg_flags = flags;
1967 if (sock->file->f_flags & O_NONBLOCK)
1968 msg_sys.msg_flags |= MSG_DONTWAIT;
1969 err = sock_sendmsg(sock, &msg_sys, total_len);
1973 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1975 if (iov != iovstack)
1976 sock_kfree_s(sock->sk, iov, iov_size);
1978 fput_light(sock->file, fput_needed);
1984 * BSD recvmsg interface
1987 asmlinkage long sys_recvmsg(int fd, struct msghdr __user *msg,
1990 struct compat_msghdr __user *msg_compat =
1991 (struct compat_msghdr __user *)msg;
1992 struct socket *sock;
1993 struct iovec iovstack[UIO_FASTIOV];
1994 struct iovec *iov = iovstack;
1995 struct msghdr msg_sys;
1996 unsigned long cmsg_ptr;
1997 int err, iov_size, total_len, len;
2000 /* kernel mode address */
2001 struct sockaddr_storage addr;
2003 /* user mode address pointers */
2004 struct sockaddr __user *uaddr;
2005 int __user *uaddr_len;
2007 if (MSG_CMSG_COMPAT & flags) {
2008 if (get_compat_msghdr(&msg_sys, msg_compat))
2011 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
2014 sock = sockfd_lookup_light(fd, &err, &fput_needed);
2019 if (msg_sys.msg_iovlen > UIO_MAXIOV)
2022 /* Check whether to allocate the iovec area */
2024 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
2025 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
2026 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
2032 * Save the user-mode address (verify_iovec will change the
2033 * kernel msghdr to use the kernel address space)
2036 uaddr = (__force void __user *)msg_sys.msg_name;
2037 uaddr_len = COMPAT_NAMELEN(msg);
2038 if (MSG_CMSG_COMPAT & flags) {
2039 err = verify_compat_iovec(&msg_sys, iov,
2040 (struct sockaddr *)&addr,
2043 err = verify_iovec(&msg_sys, iov,
2044 (struct sockaddr *)&addr,
2050 cmsg_ptr = (unsigned long)msg_sys.msg_control;
2051 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
2053 if (sock->file->f_flags & O_NONBLOCK)
2054 flags |= MSG_DONTWAIT;
2055 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2060 if (uaddr != NULL) {
2061 err = move_addr_to_user((struct sockaddr *)&addr,
2062 msg_sys.msg_namelen, uaddr,
2067 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2071 if (MSG_CMSG_COMPAT & flags)
2072 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2073 &msg_compat->msg_controllen);
2075 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2076 &msg->msg_controllen);
2082 if (iov != iovstack)
2083 sock_kfree_s(sock->sk, iov, iov_size);
2085 fput_light(sock->file, fput_needed);
2090 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2092 /* Argument list sizes for sys_socketcall */
2093 #define AL(x) ((x) * sizeof(unsigned long))
2094 static const unsigned char nargs[19]={
2095 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2096 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2097 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2104 * System call vectors.
2106 * Argument checking cleaned up. Saved 20% in size.
2107 * This function doesn't need to set the kernel lock because
2108 * it is set by the callees.
2111 asmlinkage long sys_socketcall(int call, unsigned long __user *args)
2114 unsigned long a0, a1;
2117 if (call < 1 || call > SYS_PACCEPT)
2120 /* copy_from_user should be SMP safe. */
2121 if (copy_from_user(a, args, nargs[call]))
2124 err = audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2133 err = sys_socket(a0, a1, a[2]);
2136 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2139 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2142 err = sys_listen(a0, a1);
2146 do_accept(a0, (struct sockaddr __user *)a1,
2147 (int __user *)a[2], 0);
2149 case SYS_GETSOCKNAME:
2151 sys_getsockname(a0, (struct sockaddr __user *)a1,
2152 (int __user *)a[2]);
2154 case SYS_GETPEERNAME:
2156 sys_getpeername(a0, (struct sockaddr __user *)a1,
2157 (int __user *)a[2]);
2159 case SYS_SOCKETPAIR:
2160 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2163 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2166 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2167 (struct sockaddr __user *)a[4], a[5]);
2170 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2173 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2174 (struct sockaddr __user *)a[4],
2175 (int __user *)a[5]);
2178 err = sys_shutdown(a0, a1);
2180 case SYS_SETSOCKOPT:
2181 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2183 case SYS_GETSOCKOPT:
2185 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2186 (int __user *)a[4]);
2189 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2192 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2196 sys_paccept(a0, (struct sockaddr __user *)a1,
2198 (const sigset_t __user *) a[3],
2208 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2211 * sock_register - add a socket protocol handler
2212 * @ops: description of protocol
2214 * This function is called by a protocol handler that wants to
2215 * advertise its address family, and have it linked into the
2216 * socket interface. The value ops->family coresponds to the
2217 * socket system call protocol family.
2219 int sock_register(const struct net_proto_family *ops)
2223 if (ops->family >= NPROTO) {
2224 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2229 spin_lock(&net_family_lock);
2230 if (net_families[ops->family])
2233 net_families[ops->family] = ops;
2236 spin_unlock(&net_family_lock);
2238 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2243 * sock_unregister - remove a protocol handler
2244 * @family: protocol family to remove
2246 * This function is called by a protocol handler that wants to
2247 * remove its address family, and have it unlinked from the
2248 * new socket creation.
2250 * If protocol handler is a module, then it can use module reference
2251 * counts to protect against new references. If protocol handler is not
2252 * a module then it needs to provide its own protection in
2253 * the ops->create routine.
2255 void sock_unregister(int family)
2257 BUG_ON(family < 0 || family >= NPROTO);
2259 spin_lock(&net_family_lock);
2260 net_families[family] = NULL;
2261 spin_unlock(&net_family_lock);
2265 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2268 static int __init sock_init(void)
2271 * Initialize sock SLAB cache.
2277 * Initialize skbuff SLAB cache
2282 * Initialize the protocols module.
2286 register_filesystem(&sock_fs_type);
2287 sock_mnt = kern_mount(&sock_fs_type);
2289 /* The real protocol initialization is performed in later initcalls.
2292 #ifdef CONFIG_NETFILTER
2299 core_initcall(sock_init); /* early initcall */
2301 #ifdef CONFIG_PROC_FS
2302 void socket_seq_show(struct seq_file *seq)
2307 for_each_possible_cpu(cpu)
2308 counter += per_cpu(sockets_in_use, cpu);
2310 /* It can be negative, by the way. 8) */
2314 seq_printf(seq, "sockets: used %d\n", counter);
2316 #endif /* CONFIG_PROC_FS */
2318 #ifdef CONFIG_COMPAT
2319 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2322 struct socket *sock = file->private_data;
2323 int ret = -ENOIOCTLCMD;
2330 if (sock->ops->compat_ioctl)
2331 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2333 if (ret == -ENOIOCTLCMD &&
2334 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2335 ret = compat_wext_handle_ioctl(net, cmd, arg);
2341 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2343 return sock->ops->bind(sock, addr, addrlen);
2346 int kernel_listen(struct socket *sock, int backlog)
2348 return sock->ops->listen(sock, backlog);
2351 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2353 struct sock *sk = sock->sk;
2356 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2361 err = sock->ops->accept(sock, *newsock, flags);
2363 sock_release(*newsock);
2368 (*newsock)->ops = sock->ops;
2374 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2377 return sock->ops->connect(sock, addr, addrlen, flags);
2380 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2383 return sock->ops->getname(sock, addr, addrlen, 0);
2386 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2389 return sock->ops->getname(sock, addr, addrlen, 1);
2392 int kernel_getsockopt(struct socket *sock, int level, int optname,
2393 char *optval, int *optlen)
2395 mm_segment_t oldfs = get_fs();
2399 if (level == SOL_SOCKET)
2400 err = sock_getsockopt(sock, level, optname, optval, optlen);
2402 err = sock->ops->getsockopt(sock, level, optname, optval,
2408 int kernel_setsockopt(struct socket *sock, int level, int optname,
2409 char *optval, int optlen)
2411 mm_segment_t oldfs = get_fs();
2415 if (level == SOL_SOCKET)
2416 err = sock_setsockopt(sock, level, optname, optval, optlen);
2418 err = sock->ops->setsockopt(sock, level, optname, optval,
2424 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2425 size_t size, int flags)
2427 if (sock->ops->sendpage)
2428 return sock->ops->sendpage(sock, page, offset, size, flags);
2430 return sock_no_sendpage(sock, page, offset, size, flags);
2433 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2435 mm_segment_t oldfs = get_fs();
2439 err = sock->ops->ioctl(sock, cmd, arg);
2445 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2447 return sock->ops->shutdown(sock, how);
2450 EXPORT_SYMBOL(sock_create);
2451 EXPORT_SYMBOL(sock_create_kern);
2452 EXPORT_SYMBOL(sock_create_lite);
2453 EXPORT_SYMBOL(sock_map_fd);
2454 EXPORT_SYMBOL(sock_recvmsg);
2455 EXPORT_SYMBOL(sock_register);
2456 EXPORT_SYMBOL(sock_release);
2457 EXPORT_SYMBOL(sock_sendmsg);
2458 EXPORT_SYMBOL(sock_unregister);
2459 EXPORT_SYMBOL(sock_wake_async);
2460 EXPORT_SYMBOL(sockfd_lookup);
2461 EXPORT_SYMBOL(kernel_sendmsg);
2462 EXPORT_SYMBOL(kernel_recvmsg);
2463 EXPORT_SYMBOL(kernel_bind);
2464 EXPORT_SYMBOL(kernel_listen);
2465 EXPORT_SYMBOL(kernel_accept);
2466 EXPORT_SYMBOL(kernel_connect);
2467 EXPORT_SYMBOL(kernel_getsockname);
2468 EXPORT_SYMBOL(kernel_getpeername);
2469 EXPORT_SYMBOL(kernel_getsockopt);
2470 EXPORT_SYMBOL(kernel_setsockopt);
2471 EXPORT_SYMBOL(kernel_sendpage);
2472 EXPORT_SYMBOL(kernel_sock_ioctl);
2473 EXPORT_SYMBOL(kernel_sock_shutdown);