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/wanrouter.h>
73 #include <linux/if_bridge.h>
74 #include <linux/if_frad.h>
75 #include <linux/if_vlan.h>
76 #include <linux/init.h>
77 #include <linux/poll.h>
78 #include <linux/cache.h>
79 #include <linux/module.h>
80 #include <linux/highmem.h>
81 #include <linux/mount.h>
82 #include <linux/security.h>
83 #include <linux/syscalls.h>
84 #include <linux/compat.h>
85 #include <linux/kmod.h>
86 #include <linux/audit.h>
87 #include <linux/wireless.h>
88 #include <linux/nsproxy.h>
90 #include <asm/uaccess.h>
91 #include <asm/unistd.h>
93 #include <net/compat.h>
97 #include <linux/netfilter.h>
99 static int sock_no_open(struct inode *irrelevant, struct file *dontcare);
100 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
101 unsigned long nr_segs, loff_t pos);
102 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
103 unsigned long nr_segs, loff_t pos);
104 static int sock_mmap(struct file *file, struct vm_area_struct *vma);
106 static int sock_close(struct inode *inode, struct file *file);
107 static unsigned int sock_poll(struct file *file,
108 struct poll_table_struct *wait);
109 static long sock_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
111 static long compat_sock_ioctl(struct file *file,
112 unsigned int cmd, unsigned long arg);
114 static int sock_fasync(int fd, struct file *filp, int on);
115 static ssize_t sock_sendpage(struct file *file, struct page *page,
116 int offset, size_t size, loff_t *ppos, int more);
117 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
118 struct pipe_inode_info *pipe, size_t len,
122 * Socket files have a set of 'special' operations as well as the generic file ones. These don't appear
123 * in the operation structures but are done directly via the socketcall() multiplexor.
126 static const struct file_operations socket_file_ops = {
127 .owner = THIS_MODULE,
129 .aio_read = sock_aio_read,
130 .aio_write = sock_aio_write,
132 .unlocked_ioctl = sock_ioctl,
134 .compat_ioctl = compat_sock_ioctl,
137 .open = sock_no_open, /* special open code to disallow open via /proc */
138 .release = sock_close,
139 .fasync = sock_fasync,
140 .sendpage = sock_sendpage,
141 .splice_write = generic_splice_sendpage,
142 .splice_read = sock_splice_read,
146 * The protocol list. Each protocol is registered in here.
149 static DEFINE_SPINLOCK(net_family_lock);
150 static const struct net_proto_family *net_families[NPROTO] __read_mostly;
153 * Statistics counters of the socket lists
156 static DEFINE_PER_CPU(int, sockets_in_use) = 0;
160 * Move socket addresses back and forth across the kernel/user
161 * divide and look after the messy bits.
164 #define MAX_SOCK_ADDR 128 /* 108 for Unix domain -
165 16 for IP, 16 for IPX,
168 must be at least one bigger than
169 the AF_UNIX size (see net/unix/af_unix.c
174 * move_addr_to_kernel - copy a socket address into kernel space
175 * @uaddr: Address in user space
176 * @kaddr: Address in kernel space
177 * @ulen: Length in user space
179 * The address is copied into kernel space. If the provided address is
180 * too long an error code of -EINVAL is returned. If the copy gives
181 * invalid addresses -EFAULT is returned. On a success 0 is returned.
184 int move_addr_to_kernel(void __user *uaddr, int ulen, struct sockaddr *kaddr)
186 if (ulen < 0 || ulen > sizeof(struct sockaddr_storage))
190 if (copy_from_user(kaddr, uaddr, ulen))
192 return audit_sockaddr(ulen, kaddr);
196 * move_addr_to_user - copy an address to user space
197 * @kaddr: kernel space address
198 * @klen: length of address in kernel
199 * @uaddr: user space address
200 * @ulen: pointer to user length field
202 * The value pointed to by ulen on entry is the buffer length available.
203 * This is overwritten with the buffer space used. -EINVAL is returned
204 * if an overlong buffer is specified or a negative buffer size. -EFAULT
205 * is returned if either the buffer or the length field are not
207 * After copying the data up to the limit the user specifies, the true
208 * length of the data is written over the length limit the user
209 * specified. Zero is returned for a success.
212 int move_addr_to_user(struct sockaddr *kaddr, int klen, void __user *uaddr,
218 err = get_user(len, ulen);
223 if (len < 0 || len > sizeof(struct sockaddr_storage))
226 if (audit_sockaddr(klen, kaddr))
228 if (copy_to_user(uaddr, kaddr, len))
232 * "fromlen shall refer to the value before truncation.."
235 return __put_user(klen, ulen);
238 #define SOCKFS_MAGIC 0x534F434B
240 static struct kmem_cache *sock_inode_cachep __read_mostly;
242 static struct inode *sock_alloc_inode(struct super_block *sb)
244 struct socket_alloc *ei;
246 ei = kmem_cache_alloc(sock_inode_cachep, GFP_KERNEL);
249 init_waitqueue_head(&ei->socket.wait);
251 ei->socket.fasync_list = NULL;
252 ei->socket.state = SS_UNCONNECTED;
253 ei->socket.flags = 0;
254 ei->socket.ops = NULL;
255 ei->socket.sk = NULL;
256 ei->socket.file = NULL;
258 return &ei->vfs_inode;
261 static void sock_destroy_inode(struct inode *inode)
263 kmem_cache_free(sock_inode_cachep,
264 container_of(inode, struct socket_alloc, vfs_inode));
267 static void init_once(void *foo)
269 struct socket_alloc *ei = (struct socket_alloc *)foo;
271 inode_init_once(&ei->vfs_inode);
274 static int init_inodecache(void)
276 sock_inode_cachep = kmem_cache_create("sock_inode_cache",
277 sizeof(struct socket_alloc),
279 (SLAB_HWCACHE_ALIGN |
280 SLAB_RECLAIM_ACCOUNT |
283 if (sock_inode_cachep == NULL)
288 static struct super_operations sockfs_ops = {
289 .alloc_inode = sock_alloc_inode,
290 .destroy_inode =sock_destroy_inode,
291 .statfs = simple_statfs,
294 static int sockfs_get_sb(struct file_system_type *fs_type,
295 int flags, const char *dev_name, void *data,
296 struct vfsmount *mnt)
298 return get_sb_pseudo(fs_type, "socket:", &sockfs_ops, SOCKFS_MAGIC,
302 static struct vfsmount *sock_mnt __read_mostly;
304 static struct file_system_type sock_fs_type = {
306 .get_sb = sockfs_get_sb,
307 .kill_sb = kill_anon_super,
310 static int sockfs_delete_dentry(struct dentry *dentry)
313 * At creation time, we pretended this dentry was hashed
314 * (by clearing DCACHE_UNHASHED bit in d_flags)
315 * At delete time, we restore the truth : not hashed.
316 * (so that dput() can proceed correctly)
318 dentry->d_flags |= DCACHE_UNHASHED;
323 * sockfs_dname() is called from d_path().
325 static char *sockfs_dname(struct dentry *dentry, char *buffer, int buflen)
327 return dynamic_dname(dentry, buffer, buflen, "socket:[%lu]",
328 dentry->d_inode->i_ino);
331 static struct dentry_operations sockfs_dentry_operations = {
332 .d_delete = sockfs_delete_dentry,
333 .d_dname = sockfs_dname,
337 * Obtains the first available file descriptor and sets it up for use.
339 * These functions create file structures and maps them to fd space
340 * of the current process. On success it returns file descriptor
341 * and file struct implicitly stored in sock->file.
342 * Note that another thread may close file descriptor before we return
343 * from this function. We use the fact that now we do not refer
344 * to socket after mapping. If one day we will need it, this
345 * function will increment ref. count on file by 1.
347 * In any case returned fd MAY BE not valid!
348 * This race condition is unavoidable
349 * with shared fd spaces, we cannot solve it inside kernel,
350 * but we take care of internal coherence yet.
353 static int sock_alloc_fd(struct file **filep, int flags)
357 fd = get_unused_fd_flags(flags);
358 if (likely(fd >= 0)) {
359 struct file *file = get_empty_filp();
362 if (unlikely(!file)) {
371 static int sock_attach_fd(struct socket *sock, struct file *file, int flags)
373 struct dentry *dentry;
374 struct qstr name = { .name = "" };
376 dentry = d_alloc(sock_mnt->mnt_sb->s_root, &name);
377 if (unlikely(!dentry))
380 dentry->d_op = &sockfs_dentry_operations;
382 * We dont want to push this dentry into global dentry hash table.
383 * We pretend dentry is already hashed, by unsetting DCACHE_UNHASHED
384 * This permits a working /proc/$pid/fd/XXX on sockets
386 dentry->d_flags &= ~DCACHE_UNHASHED;
387 d_instantiate(dentry, SOCK_INODE(sock));
390 init_file(file, sock_mnt, dentry, FMODE_READ | FMODE_WRITE,
392 SOCK_INODE(sock)->i_fop = &socket_file_ops;
393 file->f_flags = O_RDWR | (flags & O_NONBLOCK);
395 file->private_data = sock;
400 int sock_map_fd(struct socket *sock, int flags)
402 struct file *newfile;
403 int fd = sock_alloc_fd(&newfile, flags);
405 if (likely(fd >= 0)) {
406 int err = sock_attach_fd(sock, newfile, flags);
408 if (unlikely(err < 0)) {
413 fd_install(fd, newfile);
418 static struct socket *sock_from_file(struct file *file, int *err)
420 if (file->f_op == &socket_file_ops)
421 return file->private_data; /* set in sock_map_fd */
428 * sockfd_lookup - Go from a file number to its socket slot
430 * @err: pointer to an error code return
432 * The file handle passed in is locked and the socket it is bound
433 * too is returned. If an error occurs the err pointer is overwritten
434 * with a negative errno code and NULL is returned. The function checks
435 * for both invalid handles and passing a handle which is not a socket.
437 * On a success the socket object pointer is returned.
440 struct socket *sockfd_lookup(int fd, int *err)
451 sock = sock_from_file(file, err);
457 static struct socket *sockfd_lookup_light(int fd, int *err, int *fput_needed)
463 file = fget_light(fd, fput_needed);
465 sock = sock_from_file(file, err);
468 fput_light(file, *fput_needed);
474 * sock_alloc - allocate a socket
476 * Allocate a new inode and socket object. The two are bound together
477 * and initialised. The socket is then returned. If we are out of inodes
481 static struct socket *sock_alloc(void)
486 inode = new_inode(sock_mnt->mnt_sb);
490 sock = SOCKET_I(inode);
492 inode->i_mode = S_IFSOCK | S_IRWXUGO;
493 inode->i_uid = current_fsuid();
494 inode->i_gid = current_fsgid();
496 get_cpu_var(sockets_in_use)++;
497 put_cpu_var(sockets_in_use);
502 * In theory you can't get an open on this inode, but /proc provides
503 * a back door. Remember to keep it shut otherwise you'll let the
504 * creepy crawlies in.
507 static int sock_no_open(struct inode *irrelevant, struct file *dontcare)
512 const struct file_operations bad_sock_fops = {
513 .owner = THIS_MODULE,
514 .open = sock_no_open,
518 * sock_release - close a socket
519 * @sock: socket to close
521 * The socket is released from the protocol stack if it has a release
522 * callback, and the inode is then released if the socket is bound to
523 * an inode not a file.
526 void sock_release(struct socket *sock)
529 struct module *owner = sock->ops->owner;
531 sock->ops->release(sock);
536 if (sock->fasync_list)
537 printk(KERN_ERR "sock_release: fasync list not empty!\n");
539 get_cpu_var(sockets_in_use)--;
540 put_cpu_var(sockets_in_use);
542 iput(SOCK_INODE(sock));
548 static inline int __sock_sendmsg(struct kiocb *iocb, struct socket *sock,
549 struct msghdr *msg, size_t size)
551 struct sock_iocb *si = kiocb_to_siocb(iocb);
559 err = security_socket_sendmsg(sock, msg, size);
563 return sock->ops->sendmsg(iocb, sock, msg, size);
566 int sock_sendmsg(struct socket *sock, struct msghdr *msg, size_t size)
569 struct sock_iocb siocb;
572 init_sync_kiocb(&iocb, NULL);
573 iocb.private = &siocb;
574 ret = __sock_sendmsg(&iocb, sock, msg, size);
575 if (-EIOCBQUEUED == ret)
576 ret = wait_on_sync_kiocb(&iocb);
580 int kernel_sendmsg(struct socket *sock, struct msghdr *msg,
581 struct kvec *vec, size_t num, size_t size)
583 mm_segment_t oldfs = get_fs();
588 * the following is safe, since for compiler definitions of kvec and
589 * iovec are identical, yielding the same in-core layout and alignment
591 msg->msg_iov = (struct iovec *)vec;
592 msg->msg_iovlen = num;
593 result = sock_sendmsg(sock, msg, size);
599 * called from sock_recv_timestamp() if sock_flag(sk, SOCK_RCVTSTAMP)
601 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
604 ktime_t kt = skb->tstamp;
606 if (!sock_flag(sk, SOCK_RCVTSTAMPNS)) {
608 /* Race occurred between timestamp enabling and packet
609 receiving. Fill in the current time for now. */
611 kt = ktime_get_real();
613 tv = ktime_to_timeval(kt);
614 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMP, sizeof(tv), &tv);
617 /* Race occurred between timestamp enabling and packet
618 receiving. Fill in the current time for now. */
620 kt = ktime_get_real();
622 ts = ktime_to_timespec(kt);
623 put_cmsg(msg, SOL_SOCKET, SCM_TIMESTAMPNS, sizeof(ts), &ts);
627 EXPORT_SYMBOL_GPL(__sock_recv_timestamp);
629 static inline int __sock_recvmsg(struct kiocb *iocb, struct socket *sock,
630 struct msghdr *msg, size_t size, int flags)
633 struct sock_iocb *si = kiocb_to_siocb(iocb);
641 err = security_socket_recvmsg(sock, msg, size, flags);
645 return sock->ops->recvmsg(iocb, sock, msg, size, flags);
648 int sock_recvmsg(struct socket *sock, struct msghdr *msg,
649 size_t size, int flags)
652 struct sock_iocb siocb;
655 init_sync_kiocb(&iocb, NULL);
656 iocb.private = &siocb;
657 ret = __sock_recvmsg(&iocb, sock, msg, size, flags);
658 if (-EIOCBQUEUED == ret)
659 ret = wait_on_sync_kiocb(&iocb);
663 int kernel_recvmsg(struct socket *sock, struct msghdr *msg,
664 struct kvec *vec, size_t num, size_t size, int flags)
666 mm_segment_t oldfs = get_fs();
671 * the following is safe, since for compiler definitions of kvec and
672 * iovec are identical, yielding the same in-core layout and alignment
674 msg->msg_iov = (struct iovec *)vec, msg->msg_iovlen = num;
675 result = sock_recvmsg(sock, msg, size, flags);
680 static void sock_aio_dtor(struct kiocb *iocb)
682 kfree(iocb->private);
685 static ssize_t sock_sendpage(struct file *file, struct page *page,
686 int offset, size_t size, loff_t *ppos, int more)
691 sock = file->private_data;
693 flags = !(file->f_flags & O_NONBLOCK) ? 0 : MSG_DONTWAIT;
697 return sock->ops->sendpage(sock, page, offset, size, flags);
700 static ssize_t sock_splice_read(struct file *file, loff_t *ppos,
701 struct pipe_inode_info *pipe, size_t len,
704 struct socket *sock = file->private_data;
706 if (unlikely(!sock->ops->splice_read))
709 return sock->ops->splice_read(sock, ppos, pipe, len, flags);
712 static struct sock_iocb *alloc_sock_iocb(struct kiocb *iocb,
713 struct sock_iocb *siocb)
715 if (!is_sync_kiocb(iocb)) {
716 siocb = kmalloc(sizeof(*siocb), GFP_KERNEL);
719 iocb->ki_dtor = sock_aio_dtor;
723 iocb->private = siocb;
727 static ssize_t do_sock_read(struct msghdr *msg, struct kiocb *iocb,
728 struct file *file, const struct iovec *iov,
729 unsigned long nr_segs)
731 struct socket *sock = file->private_data;
735 for (i = 0; i < nr_segs; i++)
736 size += iov[i].iov_len;
738 msg->msg_name = NULL;
739 msg->msg_namelen = 0;
740 msg->msg_control = NULL;
741 msg->msg_controllen = 0;
742 msg->msg_iov = (struct iovec *)iov;
743 msg->msg_iovlen = nr_segs;
744 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
746 return __sock_recvmsg(iocb, sock, msg, size, msg->msg_flags);
749 static ssize_t sock_aio_read(struct kiocb *iocb, const struct iovec *iov,
750 unsigned long nr_segs, loff_t pos)
752 struct sock_iocb siocb, *x;
757 if (iocb->ki_left == 0) /* Match SYS5 behaviour */
761 x = alloc_sock_iocb(iocb, &siocb);
764 return do_sock_read(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
767 static ssize_t do_sock_write(struct msghdr *msg, struct kiocb *iocb,
768 struct file *file, const struct iovec *iov,
769 unsigned long nr_segs)
771 struct socket *sock = file->private_data;
775 for (i = 0; i < nr_segs; i++)
776 size += iov[i].iov_len;
778 msg->msg_name = NULL;
779 msg->msg_namelen = 0;
780 msg->msg_control = NULL;
781 msg->msg_controllen = 0;
782 msg->msg_iov = (struct iovec *)iov;
783 msg->msg_iovlen = nr_segs;
784 msg->msg_flags = (file->f_flags & O_NONBLOCK) ? MSG_DONTWAIT : 0;
785 if (sock->type == SOCK_SEQPACKET)
786 msg->msg_flags |= MSG_EOR;
788 return __sock_sendmsg(iocb, sock, msg, size);
791 static ssize_t sock_aio_write(struct kiocb *iocb, const struct iovec *iov,
792 unsigned long nr_segs, loff_t pos)
794 struct sock_iocb siocb, *x;
799 x = alloc_sock_iocb(iocb, &siocb);
803 return do_sock_write(&x->async_msg, iocb, iocb->ki_filp, iov, nr_segs);
807 * Atomic setting of ioctl hooks to avoid race
808 * with module unload.
811 static DEFINE_MUTEX(br_ioctl_mutex);
812 static int (*br_ioctl_hook) (struct net *, unsigned int cmd, void __user *arg) = NULL;
814 void brioctl_set(int (*hook) (struct net *, unsigned int, void __user *))
816 mutex_lock(&br_ioctl_mutex);
817 br_ioctl_hook = hook;
818 mutex_unlock(&br_ioctl_mutex);
821 EXPORT_SYMBOL(brioctl_set);
823 static DEFINE_MUTEX(vlan_ioctl_mutex);
824 static int (*vlan_ioctl_hook) (struct net *, void __user *arg);
826 void vlan_ioctl_set(int (*hook) (struct net *, void __user *))
828 mutex_lock(&vlan_ioctl_mutex);
829 vlan_ioctl_hook = hook;
830 mutex_unlock(&vlan_ioctl_mutex);
833 EXPORT_SYMBOL(vlan_ioctl_set);
835 static DEFINE_MUTEX(dlci_ioctl_mutex);
836 static int (*dlci_ioctl_hook) (unsigned int, void __user *);
838 void dlci_ioctl_set(int (*hook) (unsigned int, void __user *))
840 mutex_lock(&dlci_ioctl_mutex);
841 dlci_ioctl_hook = hook;
842 mutex_unlock(&dlci_ioctl_mutex);
845 EXPORT_SYMBOL(dlci_ioctl_set);
848 * With an ioctl, arg may well be a user mode pointer, but we don't know
849 * what to do with it - that's up to the protocol still.
852 static long sock_ioctl(struct file *file, unsigned cmd, unsigned long arg)
856 void __user *argp = (void __user *)arg;
860 sock = file->private_data;
863 if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15)) {
864 err = dev_ioctl(net, cmd, argp);
866 #ifdef CONFIG_WIRELESS_EXT
867 if (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST) {
868 err = dev_ioctl(net, cmd, argp);
870 #endif /* CONFIG_WIRELESS_EXT */
875 if (get_user(pid, (int __user *)argp))
877 err = f_setown(sock->file, pid, 1);
881 err = put_user(f_getown(sock->file),
890 request_module("bridge");
892 mutex_lock(&br_ioctl_mutex);
894 err = br_ioctl_hook(net, cmd, argp);
895 mutex_unlock(&br_ioctl_mutex);
900 if (!vlan_ioctl_hook)
901 request_module("8021q");
903 mutex_lock(&vlan_ioctl_mutex);
905 err = vlan_ioctl_hook(net, argp);
906 mutex_unlock(&vlan_ioctl_mutex);
911 if (!dlci_ioctl_hook)
912 request_module("dlci");
914 mutex_lock(&dlci_ioctl_mutex);
916 err = dlci_ioctl_hook(cmd, argp);
917 mutex_unlock(&dlci_ioctl_mutex);
920 err = sock->ops->ioctl(sock, cmd, arg);
923 * If this ioctl is unknown try to hand it down
926 if (err == -ENOIOCTLCMD)
927 err = dev_ioctl(net, cmd, argp);
933 int sock_create_lite(int family, int type, int protocol, struct socket **res)
936 struct socket *sock = NULL;
938 err = security_socket_create(family, type, protocol, 1);
949 err = security_socket_post_create(sock, family, type, protocol, 1);
962 /* No kernel lock held - perfect */
963 static unsigned int sock_poll(struct file *file, poll_table *wait)
968 * We can't return errors to poll, so it's either yes or no.
970 sock = file->private_data;
971 return sock->ops->poll(file, sock, wait);
974 static int sock_mmap(struct file *file, struct vm_area_struct *vma)
976 struct socket *sock = file->private_data;
978 return sock->ops->mmap(file, sock, vma);
981 static int sock_close(struct inode *inode, struct file *filp)
984 * It was possible the inode is NULL we were
985 * closing an unfinished socket.
989 printk(KERN_DEBUG "sock_close: NULL inode\n");
992 sock_release(SOCKET_I(inode));
997 * Update the socket async list
999 * Fasync_list locking strategy.
1001 * 1. fasync_list is modified only under process context socket lock
1002 * i.e. under semaphore.
1003 * 2. fasync_list is used under read_lock(&sk->sk_callback_lock)
1004 * or under socket lock.
1005 * 3. fasync_list can be used from softirq context, so that
1006 * modification under socket lock have to be enhanced with
1007 * write_lock_bh(&sk->sk_callback_lock).
1011 static int sock_fasync(int fd, struct file *filp, int on)
1013 struct fasync_struct *fa, *fna = NULL, **prev;
1014 struct socket *sock;
1018 fna = kmalloc(sizeof(struct fasync_struct), GFP_KERNEL);
1023 sock = filp->private_data;
1033 spin_lock(&filp->f_lock);
1035 filp->f_flags |= FASYNC;
1037 filp->f_flags &= ~FASYNC;
1038 spin_unlock(&filp->f_lock);
1040 prev = &(sock->fasync_list);
1042 for (fa = *prev; fa != NULL; prev = &fa->fa_next, fa = *prev)
1043 if (fa->fa_file == filp)
1048 write_lock_bh(&sk->sk_callback_lock);
1050 write_unlock_bh(&sk->sk_callback_lock);
1055 fna->fa_file = filp;
1057 fna->magic = FASYNC_MAGIC;
1058 fna->fa_next = sock->fasync_list;
1059 write_lock_bh(&sk->sk_callback_lock);
1060 sock->fasync_list = fna;
1061 write_unlock_bh(&sk->sk_callback_lock);
1064 write_lock_bh(&sk->sk_callback_lock);
1065 *prev = fa->fa_next;
1066 write_unlock_bh(&sk->sk_callback_lock);
1072 release_sock(sock->sk);
1076 /* This function may be called only under socket lock or callback_lock */
1078 int sock_wake_async(struct socket *sock, int how, int band)
1080 if (!sock || !sock->fasync_list)
1083 case SOCK_WAKE_WAITD:
1084 if (test_bit(SOCK_ASYNC_WAITDATA, &sock->flags))
1087 case SOCK_WAKE_SPACE:
1088 if (!test_and_clear_bit(SOCK_ASYNC_NOSPACE, &sock->flags))
1093 __kill_fasync(sock->fasync_list, SIGIO, band);
1096 __kill_fasync(sock->fasync_list, SIGURG, band);
1101 static int __sock_create(struct net *net, int family, int type, int protocol,
1102 struct socket **res, int kern)
1105 struct socket *sock;
1106 const struct net_proto_family *pf;
1109 * Check protocol is in range
1111 if (family < 0 || family >= NPROTO)
1112 return -EAFNOSUPPORT;
1113 if (type < 0 || type >= SOCK_MAX)
1118 This uglymoron is moved from INET layer to here to avoid
1119 deadlock in module load.
1121 if (family == PF_INET && type == SOCK_PACKET) {
1125 printk(KERN_INFO "%s uses obsolete (PF_INET,SOCK_PACKET)\n",
1131 err = security_socket_create(family, type, protocol, kern);
1136 * Allocate the socket and allow the family to set things up. if
1137 * the protocol is 0, the family is instructed to select an appropriate
1140 sock = sock_alloc();
1142 if (net_ratelimit())
1143 printk(KERN_WARNING "socket: no more sockets\n");
1144 return -ENFILE; /* Not exactly a match, but its the
1145 closest posix thing */
1150 #ifdef CONFIG_MODULES
1151 /* Attempt to load a protocol module if the find failed.
1153 * 12/09/1996 Marcin: But! this makes REALLY only sense, if the user
1154 * requested real, full-featured networking support upon configuration.
1155 * Otherwise module support will break!
1157 if (net_families[family] == NULL)
1158 request_module("net-pf-%d", family);
1162 pf = rcu_dereference(net_families[family]);
1163 err = -EAFNOSUPPORT;
1168 * We will call the ->create function, that possibly is in a loadable
1169 * module, so we have to bump that loadable module refcnt first.
1171 if (!try_module_get(pf->owner))
1174 /* Now protected by module ref count */
1177 err = pf->create(net, sock, protocol);
1179 goto out_module_put;
1182 * Now to bump the refcnt of the [loadable] module that owns this
1183 * socket at sock_release time we decrement its refcnt.
1185 if (!try_module_get(sock->ops->owner))
1186 goto out_module_busy;
1189 * Now that we're done with the ->create function, the [loadable]
1190 * module can have its refcnt decremented
1192 module_put(pf->owner);
1193 err = security_socket_post_create(sock, family, type, protocol, kern);
1195 goto out_sock_release;
1201 err = -EAFNOSUPPORT;
1204 module_put(pf->owner);
1211 goto out_sock_release;
1214 int sock_create(int family, int type, int protocol, struct socket **res)
1216 return __sock_create(current->nsproxy->net_ns, family, type, protocol, res, 0);
1219 int sock_create_kern(int family, int type, int protocol, struct socket **res)
1221 return __sock_create(&init_net, family, type, protocol, res, 1);
1224 SYSCALL_DEFINE3(socket, int, family, int, type, int, protocol)
1227 struct socket *sock;
1230 /* Check the SOCK_* constants for consistency. */
1231 BUILD_BUG_ON(SOCK_CLOEXEC != O_CLOEXEC);
1232 BUILD_BUG_ON((SOCK_MAX | SOCK_TYPE_MASK) != SOCK_TYPE_MASK);
1233 BUILD_BUG_ON(SOCK_CLOEXEC & SOCK_TYPE_MASK);
1234 BUILD_BUG_ON(SOCK_NONBLOCK & SOCK_TYPE_MASK);
1236 flags = type & ~SOCK_TYPE_MASK;
1237 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1239 type &= SOCK_TYPE_MASK;
1241 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1242 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1244 retval = sock_create(family, type, protocol, &sock);
1248 retval = sock_map_fd(sock, flags & (O_CLOEXEC | O_NONBLOCK));
1253 /* It may be already another descriptor 8) Not kernel problem. */
1262 * Create a pair of connected sockets.
1265 SYSCALL_DEFINE4(socketpair, int, family, int, type, int, protocol,
1266 int __user *, usockvec)
1268 struct socket *sock1, *sock2;
1270 struct file *newfile1, *newfile2;
1273 flags = type & ~SOCK_TYPE_MASK;
1274 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1276 type &= SOCK_TYPE_MASK;
1278 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1279 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1282 * Obtain the first socket and check if the underlying protocol
1283 * supports the socketpair call.
1286 err = sock_create(family, type, protocol, &sock1);
1290 err = sock_create(family, type, protocol, &sock2);
1294 err = sock1->ops->socketpair(sock1, sock2);
1296 goto out_release_both;
1298 fd1 = sock_alloc_fd(&newfile1, flags & O_CLOEXEC);
1299 if (unlikely(fd1 < 0)) {
1301 goto out_release_both;
1304 fd2 = sock_alloc_fd(&newfile2, flags & O_CLOEXEC);
1305 if (unlikely(fd2 < 0)) {
1309 goto out_release_both;
1312 err = sock_attach_fd(sock1, newfile1, flags & O_NONBLOCK);
1313 if (unlikely(err < 0)) {
1317 err = sock_attach_fd(sock2, newfile2, flags & O_NONBLOCK);
1318 if (unlikely(err < 0)) {
1323 audit_fd_pair(fd1, fd2);
1324 fd_install(fd1, newfile1);
1325 fd_install(fd2, newfile2);
1326 /* fd1 and fd2 may be already another descriptors.
1327 * Not kernel problem.
1330 err = put_user(fd1, &usockvec[0]);
1332 err = put_user(fd2, &usockvec[1]);
1341 sock_release(sock2);
1343 sock_release(sock1);
1349 sock_release(sock1);
1352 sock_release(sock2);
1359 * Bind a name to a socket. Nothing much to do here since it's
1360 * the protocol's responsibility to handle the local address.
1362 * We move the socket address to kernel space before we call
1363 * the protocol layer (having also checked the address is ok).
1366 SYSCALL_DEFINE3(bind, int, fd, struct sockaddr __user *, umyaddr, int, addrlen)
1368 struct socket *sock;
1369 struct sockaddr_storage address;
1370 int err, fput_needed;
1372 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1374 err = move_addr_to_kernel(umyaddr, addrlen, (struct sockaddr *)&address);
1376 err = security_socket_bind(sock,
1377 (struct sockaddr *)&address,
1380 err = sock->ops->bind(sock,
1384 fput_light(sock->file, fput_needed);
1390 * Perform a listen. Basically, we allow the protocol to do anything
1391 * necessary for a listen, and if that works, we mark the socket as
1392 * ready for listening.
1395 SYSCALL_DEFINE2(listen, int, fd, int, backlog)
1397 struct socket *sock;
1398 int err, fput_needed;
1401 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1403 somaxconn = sock_net(sock->sk)->core.sysctl_somaxconn;
1404 if ((unsigned)backlog > somaxconn)
1405 backlog = somaxconn;
1407 err = security_socket_listen(sock, backlog);
1409 err = sock->ops->listen(sock, backlog);
1411 fput_light(sock->file, fput_needed);
1417 * For accept, we attempt to create a new socket, set up the link
1418 * with the client, wake up the client, then return the new
1419 * connected fd. We collect the address of the connector in kernel
1420 * space and move it to user at the very end. This is unclean because
1421 * we open the socket then return an error.
1423 * 1003.1g adds the ability to recvmsg() to query connection pending
1424 * status to recvmsg. We need to add that support in a way thats
1425 * clean when we restucture accept also.
1428 SYSCALL_DEFINE4(accept4, int, fd, struct sockaddr __user *, upeer_sockaddr,
1429 int __user *, upeer_addrlen, int, flags)
1431 struct socket *sock, *newsock;
1432 struct file *newfile;
1433 int err, len, newfd, fput_needed;
1434 struct sockaddr_storage address;
1436 if (flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
1439 if (SOCK_NONBLOCK != O_NONBLOCK && (flags & SOCK_NONBLOCK))
1440 flags = (flags & ~SOCK_NONBLOCK) | O_NONBLOCK;
1442 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1447 if (!(newsock = sock_alloc()))
1450 newsock->type = sock->type;
1451 newsock->ops = sock->ops;
1454 * We don't need try_module_get here, as the listening socket (sock)
1455 * has the protocol module (sock->ops->owner) held.
1457 __module_get(newsock->ops->owner);
1459 newfd = sock_alloc_fd(&newfile, flags & O_CLOEXEC);
1460 if (unlikely(newfd < 0)) {
1462 sock_release(newsock);
1466 err = sock_attach_fd(newsock, newfile, flags & O_NONBLOCK);
1470 err = security_socket_accept(sock, newsock);
1474 err = sock->ops->accept(sock, newsock, sock->file->f_flags);
1478 if (upeer_sockaddr) {
1479 if (newsock->ops->getname(newsock, (struct sockaddr *)&address,
1481 err = -ECONNABORTED;
1484 err = move_addr_to_user((struct sockaddr *)&address,
1485 len, upeer_sockaddr, upeer_addrlen);
1490 /* File flags are not inherited via accept() unlike another OSes. */
1492 fd_install(newfd, newfile);
1495 security_socket_post_accept(sock, newsock);
1498 fput_light(sock->file, fput_needed);
1502 sock_release(newsock);
1504 put_unused_fd(newfd);
1508 put_unused_fd(newfd);
1512 SYSCALL_DEFINE3(accept, int, fd, struct sockaddr __user *, upeer_sockaddr,
1513 int __user *, upeer_addrlen)
1515 return sys_accept4(fd, upeer_sockaddr, upeer_addrlen, 0);
1519 * Attempt to connect to a socket with the server address. The address
1520 * is in user space so we verify it is OK and move it to kernel space.
1522 * For 1003.1g we need to add clean support for a bind to AF_UNSPEC to
1525 * NOTE: 1003.1g draft 6.3 is broken with respect to AX.25/NetROM and
1526 * other SEQPACKET protocols that take time to connect() as it doesn't
1527 * include the -EINPROGRESS status for such sockets.
1530 SYSCALL_DEFINE3(connect, int, fd, struct sockaddr __user *, uservaddr,
1533 struct socket *sock;
1534 struct sockaddr_storage address;
1535 int err, fput_needed;
1537 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1540 err = move_addr_to_kernel(uservaddr, addrlen, (struct sockaddr *)&address);
1545 security_socket_connect(sock, (struct sockaddr *)&address, addrlen);
1549 err = sock->ops->connect(sock, (struct sockaddr *)&address, addrlen,
1550 sock->file->f_flags);
1552 fput_light(sock->file, fput_needed);
1558 * Get the local address ('name') of a socket object. Move the obtained
1559 * name to user space.
1562 SYSCALL_DEFINE3(getsockname, int, fd, struct sockaddr __user *, usockaddr,
1563 int __user *, usockaddr_len)
1565 struct socket *sock;
1566 struct sockaddr_storage address;
1567 int len, err, fput_needed;
1569 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1573 err = security_socket_getsockname(sock);
1577 err = sock->ops->getname(sock, (struct sockaddr *)&address, &len, 0);
1580 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr, usockaddr_len);
1583 fput_light(sock->file, fput_needed);
1589 * Get the remote address ('name') of a socket object. Move the obtained
1590 * name to user space.
1593 SYSCALL_DEFINE3(getpeername, int, fd, struct sockaddr __user *, usockaddr,
1594 int __user *, usockaddr_len)
1596 struct socket *sock;
1597 struct sockaddr_storage address;
1598 int len, err, fput_needed;
1600 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1602 err = security_socket_getpeername(sock);
1604 fput_light(sock->file, fput_needed);
1609 sock->ops->getname(sock, (struct sockaddr *)&address, &len,
1612 err = move_addr_to_user((struct sockaddr *)&address, len, usockaddr,
1614 fput_light(sock->file, fput_needed);
1620 * Send a datagram to a given address. We move the address into kernel
1621 * space and check the user space data area is readable before invoking
1625 SYSCALL_DEFINE6(sendto, int, fd, void __user *, buff, size_t, len,
1626 unsigned, flags, struct sockaddr __user *, addr,
1629 struct socket *sock;
1630 struct sockaddr_storage address;
1636 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1640 iov.iov_base = buff;
1642 msg.msg_name = NULL;
1645 msg.msg_control = NULL;
1646 msg.msg_controllen = 0;
1647 msg.msg_namelen = 0;
1649 err = move_addr_to_kernel(addr, addr_len, (struct sockaddr *)&address);
1652 msg.msg_name = (struct sockaddr *)&address;
1653 msg.msg_namelen = addr_len;
1655 if (sock->file->f_flags & O_NONBLOCK)
1656 flags |= MSG_DONTWAIT;
1657 msg.msg_flags = flags;
1658 err = sock_sendmsg(sock, &msg, len);
1661 fput_light(sock->file, fput_needed);
1667 * Send a datagram down a socket.
1670 SYSCALL_DEFINE4(send, int, fd, void __user *, buff, size_t, len,
1673 return sys_sendto(fd, buff, len, flags, NULL, 0);
1677 * Receive a frame from the socket and optionally record the address of the
1678 * sender. We verify the buffers are writable and if needed move the
1679 * sender address from kernel to user space.
1682 SYSCALL_DEFINE6(recvfrom, int, fd, void __user *, ubuf, size_t, size,
1683 unsigned, flags, struct sockaddr __user *, addr,
1684 int __user *, addr_len)
1686 struct socket *sock;
1689 struct sockaddr_storage address;
1693 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1697 msg.msg_control = NULL;
1698 msg.msg_controllen = 0;
1702 iov.iov_base = ubuf;
1703 msg.msg_name = (struct sockaddr *)&address;
1704 msg.msg_namelen = sizeof(address);
1705 if (sock->file->f_flags & O_NONBLOCK)
1706 flags |= MSG_DONTWAIT;
1707 err = sock_recvmsg(sock, &msg, size, flags);
1709 if (err >= 0 && addr != NULL) {
1710 err2 = move_addr_to_user((struct sockaddr *)&address,
1711 msg.msg_namelen, addr, addr_len);
1716 fput_light(sock->file, fput_needed);
1722 * Receive a datagram from a socket.
1725 asmlinkage long sys_recv(int fd, void __user *ubuf, size_t size,
1728 return sys_recvfrom(fd, ubuf, size, flags, NULL, NULL);
1732 * Set a socket option. Because we don't know the option lengths we have
1733 * to pass the user mode parameter for the protocols to sort out.
1736 SYSCALL_DEFINE5(setsockopt, int, fd, int, level, int, optname,
1737 char __user *, optval, int, optlen)
1739 int err, fput_needed;
1740 struct socket *sock;
1745 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1747 err = security_socket_setsockopt(sock, level, optname);
1751 if (level == SOL_SOCKET)
1753 sock_setsockopt(sock, level, optname, optval,
1757 sock->ops->setsockopt(sock, level, optname, optval,
1760 fput_light(sock->file, fput_needed);
1766 * Get a socket option. Because we don't know the option lengths we have
1767 * to pass a user mode parameter for the protocols to sort out.
1770 SYSCALL_DEFINE5(getsockopt, int, fd, int, level, int, optname,
1771 char __user *, optval, int __user *, optlen)
1773 int err, fput_needed;
1774 struct socket *sock;
1776 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1778 err = security_socket_getsockopt(sock, level, optname);
1782 if (level == SOL_SOCKET)
1784 sock_getsockopt(sock, level, optname, optval,
1788 sock->ops->getsockopt(sock, level, optname, optval,
1791 fput_light(sock->file, fput_needed);
1797 * Shutdown a socket.
1800 SYSCALL_DEFINE2(shutdown, int, fd, int, how)
1802 int err, fput_needed;
1803 struct socket *sock;
1805 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1807 err = security_socket_shutdown(sock, how);
1809 err = sock->ops->shutdown(sock, how);
1810 fput_light(sock->file, fput_needed);
1815 /* A couple of helpful macros for getting the address of the 32/64 bit
1816 * fields which are the same type (int / unsigned) on our platforms.
1818 #define COMPAT_MSG(msg, member) ((MSG_CMSG_COMPAT & flags) ? &msg##_compat->member : &msg->member)
1819 #define COMPAT_NAMELEN(msg) COMPAT_MSG(msg, msg_namelen)
1820 #define COMPAT_FLAGS(msg) COMPAT_MSG(msg, msg_flags)
1823 * BSD sendmsg interface
1826 SYSCALL_DEFINE3(sendmsg, int, fd, struct msghdr __user *, msg, unsigned, flags)
1828 struct compat_msghdr __user *msg_compat =
1829 (struct compat_msghdr __user *)msg;
1830 struct socket *sock;
1831 struct sockaddr_storage address;
1832 struct iovec iovstack[UIO_FASTIOV], *iov = iovstack;
1833 unsigned char ctl[sizeof(struct cmsghdr) + 20]
1834 __attribute__ ((aligned(sizeof(__kernel_size_t))));
1835 /* 20 is size of ipv6_pktinfo */
1836 unsigned char *ctl_buf = ctl;
1837 struct msghdr msg_sys;
1838 int err, ctl_len, iov_size, total_len;
1842 if (MSG_CMSG_COMPAT & flags) {
1843 if (get_compat_msghdr(&msg_sys, msg_compat))
1846 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1849 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1853 /* do not move before msg_sys is valid */
1855 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1858 /* Check whether to allocate the iovec area */
1860 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1861 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1862 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1867 /* This will also move the address data into kernel space */
1868 if (MSG_CMSG_COMPAT & flags) {
1869 err = verify_compat_iovec(&msg_sys, iov,
1870 (struct sockaddr *)&address,
1873 err = verify_iovec(&msg_sys, iov,
1874 (struct sockaddr *)&address,
1882 if (msg_sys.msg_controllen > INT_MAX)
1884 ctl_len = msg_sys.msg_controllen;
1885 if ((MSG_CMSG_COMPAT & flags) && ctl_len) {
1887 cmsghdr_from_user_compat_to_kern(&msg_sys, sock->sk, ctl,
1891 ctl_buf = msg_sys.msg_control;
1892 ctl_len = msg_sys.msg_controllen;
1893 } else if (ctl_len) {
1894 if (ctl_len > sizeof(ctl)) {
1895 ctl_buf = sock_kmalloc(sock->sk, ctl_len, GFP_KERNEL);
1896 if (ctl_buf == NULL)
1901 * Careful! Before this, msg_sys.msg_control contains a user pointer.
1902 * Afterwards, it will be a kernel pointer. Thus the compiler-assisted
1903 * checking falls down on this.
1905 if (copy_from_user(ctl_buf, (void __user *)msg_sys.msg_control,
1908 msg_sys.msg_control = ctl_buf;
1910 msg_sys.msg_flags = flags;
1912 if (sock->file->f_flags & O_NONBLOCK)
1913 msg_sys.msg_flags |= MSG_DONTWAIT;
1914 err = sock_sendmsg(sock, &msg_sys, total_len);
1918 sock_kfree_s(sock->sk, ctl_buf, ctl_len);
1920 if (iov != iovstack)
1921 sock_kfree_s(sock->sk, iov, iov_size);
1923 fput_light(sock->file, fput_needed);
1929 * BSD recvmsg interface
1932 SYSCALL_DEFINE3(recvmsg, int, fd, struct msghdr __user *, msg,
1933 unsigned int, flags)
1935 struct compat_msghdr __user *msg_compat =
1936 (struct compat_msghdr __user *)msg;
1937 struct socket *sock;
1938 struct iovec iovstack[UIO_FASTIOV];
1939 struct iovec *iov = iovstack;
1940 struct msghdr msg_sys;
1941 unsigned long cmsg_ptr;
1942 int err, iov_size, total_len, len;
1945 /* kernel mode address */
1946 struct sockaddr_storage addr;
1948 /* user mode address pointers */
1949 struct sockaddr __user *uaddr;
1950 int __user *uaddr_len;
1952 if (MSG_CMSG_COMPAT & flags) {
1953 if (get_compat_msghdr(&msg_sys, msg_compat))
1956 else if (copy_from_user(&msg_sys, msg, sizeof(struct msghdr)))
1959 sock = sockfd_lookup_light(fd, &err, &fput_needed);
1964 if (msg_sys.msg_iovlen > UIO_MAXIOV)
1967 /* Check whether to allocate the iovec area */
1969 iov_size = msg_sys.msg_iovlen * sizeof(struct iovec);
1970 if (msg_sys.msg_iovlen > UIO_FASTIOV) {
1971 iov = sock_kmalloc(sock->sk, iov_size, GFP_KERNEL);
1977 * Save the user-mode address (verify_iovec will change the
1978 * kernel msghdr to use the kernel address space)
1981 uaddr = (__force void __user *)msg_sys.msg_name;
1982 uaddr_len = COMPAT_NAMELEN(msg);
1983 if (MSG_CMSG_COMPAT & flags) {
1984 err = verify_compat_iovec(&msg_sys, iov,
1985 (struct sockaddr *)&addr,
1988 err = verify_iovec(&msg_sys, iov,
1989 (struct sockaddr *)&addr,
1995 cmsg_ptr = (unsigned long)msg_sys.msg_control;
1996 msg_sys.msg_flags = flags & (MSG_CMSG_CLOEXEC|MSG_CMSG_COMPAT);
1998 if (sock->file->f_flags & O_NONBLOCK)
1999 flags |= MSG_DONTWAIT;
2000 err = sock_recvmsg(sock, &msg_sys, total_len, flags);
2005 if (uaddr != NULL) {
2006 err = move_addr_to_user((struct sockaddr *)&addr,
2007 msg_sys.msg_namelen, uaddr,
2012 err = __put_user((msg_sys.msg_flags & ~MSG_CMSG_COMPAT),
2016 if (MSG_CMSG_COMPAT & flags)
2017 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2018 &msg_compat->msg_controllen);
2020 err = __put_user((unsigned long)msg_sys.msg_control - cmsg_ptr,
2021 &msg->msg_controllen);
2027 if (iov != iovstack)
2028 sock_kfree_s(sock->sk, iov, iov_size);
2030 fput_light(sock->file, fput_needed);
2035 #ifdef __ARCH_WANT_SYS_SOCKETCALL
2037 /* Argument list sizes for sys_socketcall */
2038 #define AL(x) ((x) * sizeof(unsigned long))
2039 static const unsigned char nargs[19]={
2040 AL(0),AL(3),AL(3),AL(3),AL(2),AL(3),
2041 AL(3),AL(3),AL(4),AL(4),AL(4),AL(6),
2042 AL(6),AL(2),AL(5),AL(5),AL(3),AL(3),
2049 * System call vectors.
2051 * Argument checking cleaned up. Saved 20% in size.
2052 * This function doesn't need to set the kernel lock because
2053 * it is set by the callees.
2056 SYSCALL_DEFINE2(socketcall, int, call, unsigned long __user *, args)
2059 unsigned long a0, a1;
2062 if (call < 1 || call > SYS_ACCEPT4)
2065 /* copy_from_user should be SMP safe. */
2066 if (copy_from_user(a, args, nargs[call]))
2069 audit_socketcall(nargs[call] / sizeof(unsigned long), a);
2076 err = sys_socket(a0, a1, a[2]);
2079 err = sys_bind(a0, (struct sockaddr __user *)a1, a[2]);
2082 err = sys_connect(a0, (struct sockaddr __user *)a1, a[2]);
2085 err = sys_listen(a0, a1);
2088 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2089 (int __user *)a[2], 0);
2091 case SYS_GETSOCKNAME:
2093 sys_getsockname(a0, (struct sockaddr __user *)a1,
2094 (int __user *)a[2]);
2096 case SYS_GETPEERNAME:
2098 sys_getpeername(a0, (struct sockaddr __user *)a1,
2099 (int __user *)a[2]);
2101 case SYS_SOCKETPAIR:
2102 err = sys_socketpair(a0, a1, a[2], (int __user *)a[3]);
2105 err = sys_send(a0, (void __user *)a1, a[2], a[3]);
2108 err = sys_sendto(a0, (void __user *)a1, a[2], a[3],
2109 (struct sockaddr __user *)a[4], a[5]);
2112 err = sys_recv(a0, (void __user *)a1, a[2], a[3]);
2115 err = sys_recvfrom(a0, (void __user *)a1, a[2], a[3],
2116 (struct sockaddr __user *)a[4],
2117 (int __user *)a[5]);
2120 err = sys_shutdown(a0, a1);
2122 case SYS_SETSOCKOPT:
2123 err = sys_setsockopt(a0, a1, a[2], (char __user *)a[3], a[4]);
2125 case SYS_GETSOCKOPT:
2127 sys_getsockopt(a0, a1, a[2], (char __user *)a[3],
2128 (int __user *)a[4]);
2131 err = sys_sendmsg(a0, (struct msghdr __user *)a1, a[2]);
2134 err = sys_recvmsg(a0, (struct msghdr __user *)a1, a[2]);
2137 err = sys_accept4(a0, (struct sockaddr __user *)a1,
2138 (int __user *)a[2], a[3]);
2147 #endif /* __ARCH_WANT_SYS_SOCKETCALL */
2150 * sock_register - add a socket protocol handler
2151 * @ops: description of protocol
2153 * This function is called by a protocol handler that wants to
2154 * advertise its address family, and have it linked into the
2155 * socket interface. The value ops->family coresponds to the
2156 * socket system call protocol family.
2158 int sock_register(const struct net_proto_family *ops)
2162 if (ops->family >= NPROTO) {
2163 printk(KERN_CRIT "protocol %d >= NPROTO(%d)\n", ops->family,
2168 spin_lock(&net_family_lock);
2169 if (net_families[ops->family])
2172 net_families[ops->family] = ops;
2175 spin_unlock(&net_family_lock);
2177 printk(KERN_INFO "NET: Registered protocol family %d\n", ops->family);
2182 * sock_unregister - remove a protocol handler
2183 * @family: protocol family to remove
2185 * This function is called by a protocol handler that wants to
2186 * remove its address family, and have it unlinked from the
2187 * new socket creation.
2189 * If protocol handler is a module, then it can use module reference
2190 * counts to protect against new references. If protocol handler is not
2191 * a module then it needs to provide its own protection in
2192 * the ops->create routine.
2194 void sock_unregister(int family)
2196 BUG_ON(family < 0 || family >= NPROTO);
2198 spin_lock(&net_family_lock);
2199 net_families[family] = NULL;
2200 spin_unlock(&net_family_lock);
2204 printk(KERN_INFO "NET: Unregistered protocol family %d\n", family);
2207 static int __init sock_init(void)
2210 * Initialize sock SLAB cache.
2216 * Initialize skbuff SLAB cache
2221 * Initialize the protocols module.
2225 register_filesystem(&sock_fs_type);
2226 sock_mnt = kern_mount(&sock_fs_type);
2228 /* The real protocol initialization is performed in later initcalls.
2231 #ifdef CONFIG_NETFILTER
2238 core_initcall(sock_init); /* early initcall */
2240 #ifdef CONFIG_PROC_FS
2241 void socket_seq_show(struct seq_file *seq)
2246 for_each_possible_cpu(cpu)
2247 counter += per_cpu(sockets_in_use, cpu);
2249 /* It can be negative, by the way. 8) */
2253 seq_printf(seq, "sockets: used %d\n", counter);
2255 #endif /* CONFIG_PROC_FS */
2257 #ifdef CONFIG_COMPAT
2258 static long compat_sock_ioctl(struct file *file, unsigned cmd,
2261 struct socket *sock = file->private_data;
2262 int ret = -ENOIOCTLCMD;
2269 if (sock->ops->compat_ioctl)
2270 ret = sock->ops->compat_ioctl(sock, cmd, arg);
2272 if (ret == -ENOIOCTLCMD &&
2273 (cmd >= SIOCIWFIRST && cmd <= SIOCIWLAST))
2274 ret = compat_wext_handle_ioctl(net, cmd, arg);
2280 int kernel_bind(struct socket *sock, struct sockaddr *addr, int addrlen)
2282 return sock->ops->bind(sock, addr, addrlen);
2285 int kernel_listen(struct socket *sock, int backlog)
2287 return sock->ops->listen(sock, backlog);
2290 int kernel_accept(struct socket *sock, struct socket **newsock, int flags)
2292 struct sock *sk = sock->sk;
2295 err = sock_create_lite(sk->sk_family, sk->sk_type, sk->sk_protocol,
2300 err = sock->ops->accept(sock, *newsock, flags);
2302 sock_release(*newsock);
2307 (*newsock)->ops = sock->ops;
2308 __module_get((*newsock)->ops->owner);
2314 int kernel_connect(struct socket *sock, struct sockaddr *addr, int addrlen,
2317 return sock->ops->connect(sock, addr, addrlen, flags);
2320 int kernel_getsockname(struct socket *sock, struct sockaddr *addr,
2323 return sock->ops->getname(sock, addr, addrlen, 0);
2326 int kernel_getpeername(struct socket *sock, struct sockaddr *addr,
2329 return sock->ops->getname(sock, addr, addrlen, 1);
2332 int kernel_getsockopt(struct socket *sock, int level, int optname,
2333 char *optval, int *optlen)
2335 mm_segment_t oldfs = get_fs();
2339 if (level == SOL_SOCKET)
2340 err = sock_getsockopt(sock, level, optname, optval, optlen);
2342 err = sock->ops->getsockopt(sock, level, optname, optval,
2348 int kernel_setsockopt(struct socket *sock, int level, int optname,
2349 char *optval, int optlen)
2351 mm_segment_t oldfs = get_fs();
2355 if (level == SOL_SOCKET)
2356 err = sock_setsockopt(sock, level, optname, optval, optlen);
2358 err = sock->ops->setsockopt(sock, level, optname, optval,
2364 int kernel_sendpage(struct socket *sock, struct page *page, int offset,
2365 size_t size, int flags)
2367 if (sock->ops->sendpage)
2368 return sock->ops->sendpage(sock, page, offset, size, flags);
2370 return sock_no_sendpage(sock, page, offset, size, flags);
2373 int kernel_sock_ioctl(struct socket *sock, int cmd, unsigned long arg)
2375 mm_segment_t oldfs = get_fs();
2379 err = sock->ops->ioctl(sock, cmd, arg);
2385 int kernel_sock_shutdown(struct socket *sock, enum sock_shutdown_cmd how)
2387 return sock->ops->shutdown(sock, how);
2390 EXPORT_SYMBOL(sock_create);
2391 EXPORT_SYMBOL(sock_create_kern);
2392 EXPORT_SYMBOL(sock_create_lite);
2393 EXPORT_SYMBOL(sock_map_fd);
2394 EXPORT_SYMBOL(sock_recvmsg);
2395 EXPORT_SYMBOL(sock_register);
2396 EXPORT_SYMBOL(sock_release);
2397 EXPORT_SYMBOL(sock_sendmsg);
2398 EXPORT_SYMBOL(sock_unregister);
2399 EXPORT_SYMBOL(sock_wake_async);
2400 EXPORT_SYMBOL(sockfd_lookup);
2401 EXPORT_SYMBOL(kernel_sendmsg);
2402 EXPORT_SYMBOL(kernel_recvmsg);
2403 EXPORT_SYMBOL(kernel_bind);
2404 EXPORT_SYMBOL(kernel_listen);
2405 EXPORT_SYMBOL(kernel_accept);
2406 EXPORT_SYMBOL(kernel_connect);
2407 EXPORT_SYMBOL(kernel_getsockname);
2408 EXPORT_SYMBOL(kernel_getpeername);
2409 EXPORT_SYMBOL(kernel_getsockopt);
2410 EXPORT_SYMBOL(kernel_setsockopt);
2411 EXPORT_SYMBOL(kernel_sendpage);
2412 EXPORT_SYMBOL(kernel_sock_ioctl);
2413 EXPORT_SYMBOL(kernel_sock_shutdown);